exploration des liens entre les propriétés optiques de
Transcription
exploration des liens entre les propriétés optiques de
Université du Québec Institut National de la Recherche Scientifique Centre Eau Terre Environnement EXPLORATION DES LIENS ENTRE LES PROPRIÉTÉS OPTIQUES DE LA MATIÈRE ORGANIQUE DISSOUTE ET LA COMPLEXATION DES MÉTAUX TRACES DANS LES EAUX DOUCES NATURELLES Par Kristin K. Mueller Thèse présentée pour l'obtention du grade de Philosophiae doctor (Ph.D.) en sciences de l'eau Jury d'évaluation Président du jury et examinateur externe William H. Hendershot Department of Renewable Resources Macdonald College, McGili University Examinateur externe D. Scott Smith Chemistry Department Wilfrid Laurier University Examinateur interne Bernard Vigneault Ressources naturelles Canada Directeur de recherche Peter G.c. Campbell INRS-ETE Codirecteur de recherche Claude Fortin INRS-ETE © Droits réservés de (Kristin K. Mueller), 2012 ii Résumé La matière organique dissoute (MOD) joue un rôle clé dans l'écotoxicologie des métaux puisqu'elle complexe les ions métalliques libres, et diminue ce faisant la concentration de l'ion libre. Une telle complexation atténue la biodisponibilité et la toxicité des métaux. Le modèle Windermere Humic Aqueous Model (WHAM) est un modèle d'équilibre chimique qui incorpore la complexation des métaux par la MOD. WHAM est partie intégrante du modèle du ligand biotique (BLM), lequel est couramment utilisé pour prédire la biodisponibilité de métaux environnementalement importants, tels que Cd, Cu, Ni et Zn, envers les organismes aquatiques dans les systèmes naturels. WH AM prend en compte la quantité mesurée de MOD, mais l'utilisateur doit définir la « qualité» de la MaD en imposant le pourcentage (%) de la MaD qui est réellement active dans la complexation des métaux. Ici, nous proposons une approche plus rigoureuse, laquelle consiste à mesurer le % de MaD active en utilisant des méthodes spectroscopiques bien établies. Nous avons mesuré les concentrations de Cd 2+, Cu 2+, Ni 2+ et Zn2+ dans huit lacs et trouvé des accords raisonnables entre les 2 2 mesures et les prédictions faites par WHAM pour Cd + et Zn +. Au contraire, des différences significatives ont été observées pour Cu 2+ et Ni 2+. Pour les métaux Cd, Ni et Zn, le % de MaD actif requis pour ajuster les prédictions du modèle aux mesures d'ion libre est bien expliqué par les variations inter-lacs dans la qualité spectroscopique de la MOD. La proportion de MOD active dans la complexation de Cd et Zn augmente avec l'augmentation de la fluorescence caractéristique de la MOD de type allochtone; une tendance inverse a été observée pour Ni, suggérant que les sites de complexation de Cd et Zn sur la MOD sont différents que pour Ni. L'utilisation de la qualité spectroscopique de la MaD pour estimer le % de MaD actif vis-à-vis de la complexation des métaux 2 a amélioré les capacités de WH AM à prédire les concentrations de Cu 2+ et particulièrement Ni + libres. Nous concluons que le raffinement du modèle WHAM pour qu'il incorpore les variations dans la qualité de la MOD doit aller au-delà du simple ajustement du pourcentage de la MOD qui est actif, par exemple en tenant compte de l'affinité des différents sites de complexation des métaux, notamment Cu et Ni. Nous discutons ensuite comment les variations dans la qualité de la MOD, telles que révélées par les différences observées dans les spectres de fluorescence, peuvent être utilisées pour améliorer les prédictions du modèle WHAM vis-à-vis des concentrations des métaux environnementale ment importants dans les systèmes aquatiques naturels. iii ~~~~-------- - Abstract Dissolved organic matter (DOM) plays a key role in metal ecotoxicology by complexing free metal ions, an action that decreases the free metal ion concentrations and thus attenuates metal bioavailability and toxicity. The Windermere Humic Aqueous Model (WHAM) is a chemical equilibrium model that incorporates DOM-metal complexation and is an integral part of the Biotic Ligand Model, which is widely used to predict the bioavailability of environmentally significant metals, such as Cd, Cu, Ni and Zn, to aquatic organisms in natural systems. WHAM incorporates the measured quantity of DOM, but users must define the "qua lity" of the DOM by choosing the % of DOM that is actively involved in metal complexation. The ability to estimate this % active DOM spectroscopically would be useful. We measured Cd 2+, Cu 2+, Ni 2+ and Zn2+ concentrations in eight 2 lakes and found reasonable agreement between measured and predicted concentrations of Cd + and Zn2+, but there were marked differences between measured and WHAM-modeled concentrations of Cu 2+ and Ni 2+. Relationships between the model-optimized % active DOM needed to fit modeled to measured free metal concentrations and the lake-to-Iake variation in the spectroscopie quality of the DOM were apparent for Cd, Ni and Zn. The proportion of DOM involved in Cd and Zn complexation was found to increase with an increase in the allochthonous DOM fluorescence signature; the opposite trend was observed for Ni, suggesting that the DOM binding sites involved in Cd and Zn binding differ from those involved in Ni complexation. Improved WHAM predictions of Cu 2+ and, especially, Ni 2+ concentrations were achieved when the DOM spectroscopie qua lity was used to estimate the % of DOM active in metal complexation. We conclude that the refinement of the WHAM model to incorporate variations in DOM quality must go beyond the simple adjustment of the "percent active DOM" and possibly include changes to DOM metal binding affinities, notably for Cu and Ni. We discuss the application of variations in DOM quality, as revealed by differences observed in the fluorescence spectra, to refine metal speciation predictions obtained with WHAM for environmentally significant trace metals in natural aquatic systems. iv Avant-propos La thèse comprend un sommaire récapitulatif (en français) ainsi qu'une première partie qui comporte une synthèse générale (en anglais) de l'ensemble des recherches, composée d'une introduction, des objectifs et hypothèses de recherche, d'une discussion des résultats majeurs et des perspectives de la recherche. La deuxième partie de la thèse consiste en trois publications scientifiques: la première est déjà publiée, la seconde est provisoirement acceptée pour publication (en révision), et la troisième est en préparation pour être soumise à une revue avec comité de lecture. Finalement, la troisième partie contient une annexe qui décrit une expérience complémentaire qui a été réalisée pour évaluer l'effet d'entreposage des échantillons sur la spéciation des métaux et la qualité de la matière organique dissoute. La contribution des auteurs aux différents articles s'établit comme suit: 1. Mueller, K.K., C. Fortin, P.G.c. Campbell. 2012. "Spatial variation in the optical properties of dissolved organic matter (DOM) in lakes on the Canadian Precambrian Shield and links to watershed characteristics", Aquatic Geochemistry, 18(1): 21-44. K.K. Mueller : Conception et réalisation du projet (échantillonnage, analyses, traitement et interprétation des données); rédaction initiale et finale de l'article. C. Fortin: Conception et réalisation du projet (contribution au traitement et l'interprétation de données); contribution à la rédaction finale de l'article. P.G.C. Campbell: Conception et réalisation du projet (contribution au traitement et l'interprétation de données); contribution à la rédaction finale de l'article. 2. Mueller, K.K., S. Lofts, C. Fortin, P.G.c. Campbell. 2012. "Trace metal speciation predictions in natural aquatic systems: incorporation of dissolved organic matter (DOM) spectroscopic quality", Environmental Chemistry, 9: 356-368. K.K. Mueller : Conception et réalisation du projet (échantillonnage, analyses, traitement et interprétation des données); rédaction initiale et finale de l'article. S. Lofts: Modifications du model WHAM (contribution au traitement et l'interprétation de données; optimisation du modèle); contribution à la rédaction finale de l'article. C. Fortin: Conception et réalisation du projet (contribution au traitement et l'interprétation de données); contribution à la rédaction finale de l'article. P.G.C. Campbell: Conception et réalisation du projet (contribution au traitement et l'interprétation de données); contribution à la rédaction finale de l'article. 3. Mueller, K.K., C. Fortin, P.G.c. Campbell. 2012. "Nickel and copper complexation by dissolved organic matter in lake waters", en preparation. K.K. Mueller : Conception et réalisation du projet (échantillonnage, analyses, t~aitement et interprétation des données); rédaction initiale et finale de l'article. C. Fortin: Conception et réalisation du projet (contribution au traitement et l'interprétation de données); contribution à la rédaction finale de l'article. P.G.C. Campbell: Conception et réalisation du projet (contribution au traitement et l'interprétation de données); contribution à la rédaction finale de l'article. v Forward The thesis consists of a summary (in French) followed bya more general synthesis (in English). The synthesis comprises the following sections: 1) an introduction; 2) the research hypothesis, objectives and originality of the thesis; 3) the methodology; 4) the results and discussion; and finally, 4) the main conclusions of the thesis. The second part of the thesis presents three scientific publications, the first of which is already published, the second has been accepted with revisions, whereas the third is in preparation for submission. The third part of the thesis is an appendix that describes the results of an experiment on the effects of sample storage on the trace metal speciation and the spectroscopic quality of dissolved organic matter in natural water samples. The contributions of each of the authors of the scientific articles included in the thesis are as follows: 1. Mueller, K.K., C. Fortin, P.G.c. Campbell. 2012. "Spatial variation in the optical properties of dissolved organic matter (DOM) in lakes on the Canadian Precambrian Shield and links to watershed characteristics", Aquatic Geochemistry, 18(1): 21-44. K.K. Mueller: Design and realization of the project (sampling, analyses, data analysis and interpretation); initial and final writing of the article. C. Fortin: Design and realization ofthe project (contribution to the analysis and interpretation of data); contribution to the final writing of the article. P.G.C. Campbell: Design and realization ofthe project (contribution to the analysis and interpretation of data); contribution to the final writing of the article. 2. Mueller, K.K., S. Lofts, C. Fortin, P.G.c. Campbell. 2012. "Trace metal speciation predictions in natural aquatic systems: incorporation of dissolved organic matter (DOM) spectroscopic quality", Environmental Chemistry, 9: 356-368. K.K. Mueller: Design and realization of the project (sampling, analyses, data analysis and interpretation); initial and final writing ofthe article. S. Lofts: Modifications to the model WHAM (contribution to the analysis and interpretation of data, model optimization); contribution to the final writing of the article. C. Fortin: Design and realization of the project (contribution to the analysis and interpretation of data); contribution to the final writing of the article. P.G.C. Campbell: Design and realization of the project (contribution to the analysis and interpretation of data); contribution to the final writing of the article. 3. Mueller, K.K., C. Fortin, P.G.c. Campbell. 2012. "Nickel and copper complexation by dissolved organic matter in lake waters", in preparation. K.K. Mueller: Design and realization of the project (sampling, analyses, data analysis and interpretation); initial and final writing ofthe article. C. Fortin: Design and realization of the project (contribution to the analysis and interpretation of data); contribution to the final writing of the article. P.G.C. Campbell: Design and realization ofthe project (contribution to the analysis and interpretation of data); contribution to the final writing of the article. vi Remerciements J'ai été privilégiée de pouvoir travailler, tout au long de mes études doctorales, avec un groupe de personnes exceptionnelles à l'Institut national de la Recherche scientifique. Je remercie d'abord mon superviseur Peter Campbell pour son mentorat, son support et sa patience. Tu as toujours été très généreux de ton temps et toujours disponible pour répondre à mes questions et réviser mes écrits rapidement, et cela peu importe où tu te trouvais sur la planète. Je remercie aussi mon co-superviseur, Claude Fortin, pour son support et son expertise analytique. Ton souci du détail et ta lecture attentive de mes travaux les ont grandement améliorés. Enfin, vous m'avez proposé des opportunités scientifiques qui m'ont permis d'élargir mes horizons et de développer un réseau de collègues qui me servira lors de ma carrière à venir. Parmi ces opportunités je suis particulièrement reconnaissante pour le séjour à Lancaster (Angleterre), l'atelier de travail à Grenade (Espagne) et la conférence à Kyoto (Japon) ou j'ai pu présenter mes résultats. Les autres professeurs du groupe de biogéochimie à l'INRS-ETE m'ont encouragée durant mes études. Je remercie Charles Gobeil, André Tessier, Landis Hare et Isabelle Laurion non seulement pour les discussions stimulantes, mais aussi pour les conseils sur la carrière et la vie de scientifique. Je remercie spécialement Scott Smith pour ses conseils sur la fluorescence de la MOD, son enthousiasme pour les sciences et sa pédagogie. Le mentorat que j'ai reçu du Professeur Ed Tipping et de Steve Lofts sur le modèle WHAM a été essentiel au succès de cette étude. Je vous remercie tous les deux de m'avoir généreusement accueilli au Centre for Ecology and Hydrology de Lancaster. Je remercie finalement mon comité examinateur pour leurs commentaires qui ont grandement amélioré la version finale de la thèse. J'exprime aussi ma gratitude aux professionnels de recherche et au personnel technique de l'INRSÉTÉ pour leur support et leur expertise tant au laboratoire que sur le terrain: Michelle Bordeleau, Sébastien Duval, Pauline Fournier, Philippe Girard, Pierre Marcoux, Julie Perreault, Stéfane Prémont, Lise Rancourt et René Rodrigue. Je remercie Marc-André Robin pour les données GIS et Isabelle Lavoie pour l'aide avec les analyses statistiques. Un support logistique essential a été fourni par John Gunn et la Freshwater Ecology Unit de Laurentian University Cooperative, de même que par Louis Jourdain du Ministère de Ressources naturelles et de la Faune du Québec. Ce fut un grand plaisir de côtoyer le groupe d'étudiants et de chercheurs de l'INRS-ETE, en particulier Julie Breton, Dany Dumont, Dominique Lapointe, Michel Lavoie, Isabelle Lavoie, Séverine· Le Faucheur et Isabelle Prou Ix. J'aimerais exprimer ma reconnaissance à mes amis et ma famille pour leur support et leur amour. En particulier Kimberley Tasker, qui a été une amie si généreuse durant ces années; mes parents, Lyle et Wendi, qui m'ont enseigné la ténacité et l'indépendance; ma sœur Johanna, à qui j'ai toujours pu me confier malgré la distance qui nous sépare; mes beaux-parents Gilles et Sylvie, qui m'ont généreusement accueillie durant mes nombreux séjours à Québec durant la fin de mes études. Enfin, je remercie Raoul-Marie Couture pour ces encouragements, sa patience et son support. J'ai apprécié tes conseils et tout le temps que tu aies passé à revoir mes écrits. vii Acknowledgements 1was very fortunate, throughout my doctoral studies, to have been able to work with a great group of people at the Institut national de la Recherche scientifique. 1would first like to thank my supervisor Peter Campbell for his mentorship, guidance, support and most of ail, his patience. Vou were always very generous with your time and 1appreciate your quick response to my questions and written work, even from the other side of the planet! 1would also like to my co-supervisor, Claude Fortin, for his guidance, support and analytical expertise. Vour attention to detail in reviewing written work was remarkable and greatly improved the quality of my work. Finally, the scientific opportunities Vou both provided me with, from an internship in Lancaster, to a workshop in Granada and a conference in Kyoto, have helped me to broaden my horizons and develop a network of colleagues that will serve me weil in my future career. Other professors in the biogeochemistry research group at INRS-ETE were also encouraging during my studies. Thank Vou to Charles Gobeil, André Tessier, Landis Hare and Isabelle Laurion for not only stimulating scientific discussions, but also for candid discussions on career and life balance. 1would especially like to thank Scott Smith for insight and advice on DOM fluorescence, your enthusiasm for science and your pedagogy. The training 1received from Professor Ed Tipping and Steve Lofts on the model WHAM were essential for the completion of this work. Thank vou for generously hosting me at the Centre for Ecology and Hydrology in Lancaster and for your continuing encouragement. 1also thank my examining committee who took the time to evaluate my thesis and whose comments have greatly improved this thesis. My gratitude goes to the following technical staff at INRS-ETE for their support and expertise in the lab and in the field: Michelle Bordeleau, Sébastien Duval, Pauline Fournier, Philippe Girard, Pierre Marcoux, Julie Perreault, Stéfane Prémont, Lise Rancourt and René Rodrigue. 1also thank MarcAndré Robin for providing GIS data and Isabelle Lavoie for help with statistical analyses. Essential logistical assistance in the field was provided by John Gunn and the Laurentian University Cooperative Freshwater Ecology Unit, as weil as Louis Jourdain from the Ministère de Ressources naturelles et de la Faune du Québec. 1was also fortune enough to work with a great group of students and researchers at INRS-ETE. would especially like to acknowledge Julie Breton, Dany Dumont, Dominique Lapointe, Michel Lavoie, Isabelle Lavoie, Séverine Le Faucheur and Isabelle Proulx who made daily life at INRS-ETE a pleasure. 1am forever grateful to my friends and family for their continuai love and support. In particular Kimberley Tasker, who has been such a generous friend over the years; my parents, Lyle and Wendi, who taught me to be tenacious and independent; my sister, Johanna, who has always been only a phone cali away; and my in-Iaws, Gilles and Sylvie, who graciously housed me many times during my many trips to Québec City near the end of my studies. Finally, 1am grateful for the endless encouragement, patience and support of Raoul-Marie Couture. Thank vou for your advice, comments and ail the time Vou spent editing my written work. viii Table des matières Résumé .........................................................................................................................................................iii Abstract ................................... :.................................................................................................................... iv Avant-propos .................................................................................................................................................v Forward ........................................................................................................................................................vi Remerciements ............................................................................................................................................ vii Acknowledgements .................................................................................................................................... viii Liste des figures .......................................................................................................................................... xiii Liste des tableaux ........................................................................................................................................ xv Sommaire récapitulatif ................................................................................................................................ 1. 1 Introduction ............................................................................................................................................. 3 1.1 Métaux et lacs d'intérêt ............................................................................................................... 3 1.2 Spéciation des métaux en eau douce ........................................................................................... 4 1.3 Matière organique dissoute (MOD) ............................................................................................. 5 1.4 Modélisation de la spéciation à l'équilibre .................................................................................. 8 2 Hypothèse de recherche, objectifs et originalité .................................................................................. 10 2.1 Hypothèse de recherche ............................................................................................................ 10 2.2 Objectifs de recherche ............................................................................................................... 10 2.3 Originalité ....................................... :........................................................................................... 11 3 Méthodologie ........................................................................................................................................ 11 3.1 Échantillonnage .......................................................................................................................... 12 3.2 Analyses ....................................................................................................................................... 12 4 Résultats et Discussion .......................................................................................................................... 13 4.1 Bassins versants .......................................................................................................................... 14 4.2 Chimie des eaux lacustres .......................................................................................................... 14 4.3 Matière organique dissoute (MOD) ........................................................................................... 15 4.4 Spéciation des métaux traces ..................................................................................................... 17 4.5 Spéciation des métaux traces mesurée et prédite ..................................................................... 17 ix 4.6 Paramètres de complexation mesurés ....................................................................................... 19 5 Conclusions ............................................................................................................................................ 22 Première partie: Synthèse ......................................................................................................................... 27 1 Introduction ........................................................................................................................................... 29 1.1 Metals of interest ........................................................................................................................ 29 1.2 Sites of interest ........................................................................................................................... 29 1.3 Freshwater trace metal speciation ............................................................................................. 31 1.3.1 Measuring freshwater trace metal speciation ............................................................. 33 1.4 Dissolved organic matter (DOM) ................................................................................................ 34 1.5 Chemical equilibrium speciation modelling ................................................................................ 39 1.5.1 WHAM field validations ............................................................................................... 41 2 Research hypothesis, objectives and origina lity .................................................................................... 44 2.1 Research hypothesis ................................................................................................................... 44 2.2 Research objectives ..................................................................................................................... 45 2.3 Originality .................................................................................................................................... 46 3 Methodology ......................................................................................................................................... 46 3.1 Sampling ...................................................................................................................................... 46 3.2 Sample analysis ........................................................................................................................... 49 3.2.1 Lake DOM quality ......................................................................................................... 49 3.2.2 Cd, Cu, Ni and Zn metal speciation .............................................................................. 50 3.2.3 DOM titrations with Cu and Ni. .................................................................................... 50 3.2.4 Effect of sample storage on metal speciation and DOM quality ................................. 51 4 Results and Discussion ........................................................................................................................... 52 4.1 Lake watersheds ......................................................................................................................... 52 4.2 Lake water chemistry .................................................................................................................. 54 4.3 Dissolved organic matter (DOM) ................................................................................................ 55 4.4 Trace metal speciation ................................................................................................................ 60 4.5 Measured vs. modeled trace metal speciation ............................ ,.............................................• 61 4.6 Measured metal binding parameters ......................................................................................... 66 x 5 Conclusions ............................................................................................................................................ 70 5.1 Improving trace metal speciation predictions ............................................................................ 70 5.2 Research perspectives ................................................................................................................ 74 Deuxième partie: Articles ......................................................................................................................... 77 Spatial variation in the optical properties of dissolved organic matter (DOM) in lakes on the Canadian Precambrian Shield and links to watershed characteristics .................................................. 79 Trace metal speciation predictions in natural aquatic systems: incorporation of dissolved organic matter (DOM) spectroscopic quality ................................................................................................... 117 Nickel and copper complexation by natural dissolved organic matter in lake waters ............................ 161 Troisième partie: Annexes ...................................................................................................................... 181 Appendix A: Effect of sample storage on metal speciation and DOM quality / Annexe A : Effet de l'entreposage des échantillons sur la spéciation de métaux et la qualité de la MOD ........................ 183 Appendix B: Lake DOM spectroscopic quality results / Annexe B : Résultats de l'analyse spectroscopique de la MOD lacustre .................................................................................................. 191 Appendix C: Effects of aluminum (AI) and iron (Fe) on WHAM VI predictions of Cd, Cu, Ni and Zn speciation / Annexe C : Effets de l'aluminium (AI) et du fer (FE) sur les prédictions de la spéciation du Cd, du Cu, du Ni et du Zn ................................................................................................................ 201 Quatrième partie: Bibliographie ............................................................................................................ 205 xi ----------- ---- - -------------- xii Liste des figures Figure 1. Schematic representation of the interconnectedness between water chemistry and bioavailability / Représentation schématique de la relation entre la chimie des eaux naturelles et la biodisponibilité des métaux ...................................................... 32 Figure 2. Representation of a proposed molecular structure for humic substances (from Schulten and Schnitzer 1997) / Structure moléculaire probable de la substance humique (tiré de Schulten and Schnitzer, 1997) .............................................................. 37 Figure 3. Example ofthe sampling apparatus / Montage utilisé pour l'échantillonnage .............. .48 Figure 4. Map of the two study areas showing the relative position of the two regions in eastern Canada (panel a), the geographicallocation of the Rouyn-Noranda lakes (panel b) and the Sudbury lakes (panel cl. Reproduced fram Mueller et al. (2012b) / Carte de l'Est du Canada montrant les positions relatives de deux régions étudiées (A), ainsi que l'emplacement de lacs de la région de RouynNoranda (B) et de la région de Sudbury (C). Reproduit de Mueller et al. (2012b) ........... 53 Figure s. Comparison of the concentrations of the free metal ion [M 2+] measured using lET (Cd, Ni and Zn) and ISE (Cu) and those calculated with WHAM VI assuming 65% aFA (large dark circles) and a range of 33% (smaillight circles) to 130% aFA (smaillight squares). The dashed line represents the l:lline / Comparaison 2 entre les concentrations d'ion métallique libre [M +] mesurées par lET (Cd, Ni et Zn) ou ISE (CU) et les concentrations prédites par WHAM VI en supposant des valeurs de % aFA de 65 % (cercles pleins) ou entre 33 et 130 % (cercles ouverts). La ligne pointillée représente l'équivalence 1 : 1. ............................................................ 64 Figure 6. Relationship between the WHAM-calculated percent fulvic acid active in complexation (%aFA optimised) for Cd, Cu, Ni and Zn and the relative contribution of the PARAFAC fluorescence components 1 (panels a-dl and 3 (panels e-h) to the overall fluorescence spectrum of lake DOM. Points of the same shape and fill represent replicate samples. Solid lines represent regression equations of statistical significance at the 0.05 alpha level. Reproduced from Mueller et al. (2012c) / Pourcentage optimal d'acide fulvique actif dans la complexation (%aFA optimisé) calculé par WHAM pour Cd, Cu, Ni et Zn en fonction de la contribution, relative à la fluorescence totale de la MOD, des composantes 1 (panneaux a-dl et 3 (panneaux e-h) de fluorescence PARAFAC. xiii Les points de même forme et de même couleur représentent des échantillons en répliqua, et les lignes continues représentent les régressions significatives au niveau statistique de alpha égal à 0.05. Reproduit de Mueller et al. (2012c) .................. 65 Figure 7. z Comparison ofthe mean (± SD, n=3) concentrations offree metal ion [M +] calculated with WH AM VI (with the percentage active fulvic acid (%aFA) estimated from Figure 6) and those measured (± SD, n=3) using the ion exchange technique (lET) (Cd, Ni and Zn) or ion selective electrode {ISE) (Cu) methods: (a) cadmium; (b) copper; (c) nickel and (d) zinc. The dashed line represents the l:lline / Concentrations moyennes (± ET, n=3) du métal libre z [M +] calculé par WHAM (en employant le pourcentage d'acide fulvique actif (%aFA) estimé d'après la Figure 6) comparées à celles mesurées (± ET, n=3) avec la technique d'échange d'ions (IET)(Cd, Ni, Zn) ou une électrode sélective (ISE) (Cu) : (a) cadmium; (b) cuivre; (c) nickel et (d) zinc. La ligne pointillée représente l'équivalence 1 : 1. ............................................................................................................ 67 Figure 8. Normalized Ni (A and C) and Cu (B and D) complexation curves (M-DOM/DOC) for Lake Geneva (A and B) and Lake Opasatica (C and D). Measured values are shown in colour; values calculated with WHAM VI are shown in gray triangles. Modified from Mueller et al. (2012a) / Courbe de complexation normalisée (MMOD/MOD) pour Ni (panneaux A et D) et Cu (panneaux B et D) dans les lacs Geneva (carrées; panneaux A et B) et Opasatica (cercles, panneaux C et D). Les valeurs mesurées sont montrées en couleur alors que celles prédites par WHAM VI sont représentées par des triangles gris. Modifié d'après Mueller et al. (2012a) ............................................................................................................................. 68 xiv Liste des tableaux Table 1. Lake names, sampling regions, locations and years sampled. Modified from Mueller et al. (2012b) / Noms des lacs, ainsi que les régions, sites et dates d'échantillonnage. Modifié d'après Mueller et al. (2012b) .............................................. 54 Table 2. Measured pH and mean (± SD, N=3) dissolved organic matter (DOM) concentrations, total dissolved concentrations of major cations and trace metals, and % free metal values for lakes sampled from Rouyn-Noranda (QC) and Sudbury (ON) in 2008/ Valeurs de pH, de concentrations moyennes (± SD, N=3) de matière organique dissoute (MOD), de concentrations totales de cations majeurs dissous et de métaux traces ainsi que les pourcentages (%) de métal libre pour les lacs échantillonnés dans les régions de Rouyn-Noranda (Qué.) et de Sudbury (Ont.) en 2008 .................................................................................................................................. 56 Table 3. Measured mean (± SD, N=3) dissolved organic matter (DOM) concentration, specifie UV-absorbance (SUVA 24S ), fluorescence index (FI), and relative proportions of the Cl and C3 PARAFAC components for the eighteen lakes sampled in 2007 and the eight lakes sampled in 2008/ Concentrations moyennes (± SD, N=3) de matière organique dissoute (MOD), absorbance UV spécifique (SUVA 24S ), indice de fluorescence (FI) et les proportions relatives des composantes de fluorescence Cl et C3 pour les 18 lacs échantillonnés en 2007 et les 8 lacs échantillonnés en 2008 .................................................................................................................................. 59 xv -------------------------- xvi Sommaire récapitulatif 2 1 1.1 Introduction Métaux et lacs d'intérêt Plusieurs substances présentes dans l'environnent naturel sont considérées comme potentiellement problématiques envers la santé de l'écosystème et la santé humaine. Le cuivre (Cu), le cadmium (Cd), le nickel (Ni) et le zinc (Zn), émis par les raffineries et fonderies de cuivre et zinc vers l'atmosphère et l'environnement terrestre et aquatique, ont été identifiés comme « prioritaires» par la Loi canadienne sur la protection de l'environnement (CEPA, 2001)1. La toxicité potentielle et le risque écotoxicologique de ces métaux doivent être déterminés. Les régions du Canada qui sont les plus affectées par la présence des raffineries et fonderies de cuivre et zinc sont les régions de Rouyn-Noranda, au Québec, et de Sudbury, en Ontario. Ces régions sont toutes deux localisées sur le Bouclier canadien précambrien (Carignan and Nriagu, 1985). Dans la région de Rouyn-Noranda, la roche mère de type granitique est le plus souvent recouverte d'une épaisse couche de dépôts superficiels de l'ère glacière, bien que parfois elle soit exposée (Veillette et aL, 2005). En revanche, dans la région de Sudbury, la roche mère est parfois recouverte d'une mince couche de dépôt «1 m), mais est le plus souvent exposée (Barnett and Bajc, 2002). Ces différences dans l'exposition de la roche-mère contribuent aux différences dans la topographie et, probablement, à celles dans la végétation, les sols et l'hydrologie des bassins versants des deux régions (Sobek et aL, 2007). Les deux régions d'intérêt possèdent un long historique d'activité minière et de raffinage de métaux, datant du début du XIXe siècle (Carignan and Nriagu, 1985; Couillard et aL, 2004; Gallon et aL, 2006; Nriagu et aL, 1982). Les concentrations de métaux trouvées dans les aires entourant les raffineries sont tributaires de la direction des vents dominants et décroissent au fur et à mesure que l'on s'éloigne de la raffinerie. Certains lacs dans la région de RouynNoranda sont aussi directement affectés par des rejets miniers et par le drainage minier acide (Goulet and Couillard, 2009). 1 Toutes les références se trouvent à la fin de la Quatrième partie de la thèse (page 205). 3 ---------------------- 1.2 Spéciation des métaux en eau douce La toxicité des métaux envers les organismes dans l'environnement aquatique dépend de leur biodisponibilité, laquelle dépend des formes spécifiques que revêtent les métaux (Allen and Hansen, 1996; Batley et aL, 2004; Janssen et aL, 2003). En particulier, la concentration de l'ion métallique libre est généralement reconnue comme représentative de la biodisponibilité des métaux envers les organismes aquatiques (Campbell, 1995). En eau douce, les ligands inorganiques simples tels que OH-, HC0 3- et C0 32- sont souvent présents à des concentrations 2 2 permettant la formation de complexes avec les ions métalliques tels que Cu 2+, Cd +, Ni + et Zn2+ (Stumm and Morgan, 1996). Des monomères organiques comme le citrate, la glycine ou les acides polycarboxyliques ont aussi une grande affinité pour les ions métalliques, mais était donné leur grande labilité vis-à-vis de la dégradation par l'activité microbienne, ils ne soht normalement pas présents en concentration suffisante, dans les eaux douces, pour influencer la spéciation des métaux (Morel and Hering, 1993). La complexation des métaux par la matière organique dissoute (MOD) est plus difficile à quantifier que celle par les ligands inorganiques mentionnés précédemment. En effet, l'hétérogénéité ainsi que les propriétés polycoordinantes et polyélectrolytiques de la MOD rendent impossible la détermination de sites de complexation précis. Des constantes conditionnelles sont par conséquent déterminées expérimentalement en titrant de la MOD isolée chimiquement avec des métaux dans des conditions très spécifiques (ex. : pH, force ionique, concentrations d'ions majeurs). Ces constantes conditionnelles ne peuvent toutefois être utilisées dans des conditions autres que celles dans lesquelles elles ont été déterminées. L'effet de ces différentes variables sur la complexation des métaux par la MOD a été exploré en détail par Tipping (2002). D'abord, la complexation des métaux dépend de la concentration du z métal libre, [M +], laquelle dépend de la concentration totale de métal. La complexation dépend aussi de l'affinité du métal étudié pour les sites de complexation de la MOD, ainsi que du pH de la solution. Lorsque le pH diminue, la compétition entre le cation métallique et les protons pour les sites de complexation augmente (ex. : Benedetti et al. (1995)). Les autres cations en solution peuvent aussi compétitionner avec le cation métallique, en fonction de leur 4 affinité relative et de leur concentration en solution (ex. : Cao et al. (2006)). Enfin, l'effet de la force ionique sur la complexation des métaux par la MaD est expliqué par le fait que les autres ions atténuent les attractions électrostatiques entre la MaD et le cation métallique (Tipping, 2002). Pour les métaux cationiques, la mesure d'une ou plusieurs espèces chimiques dans les eaux douces naturelles n'est pas une opération de routine. Selon la composition chimique de l'échantillon, des changements réversibles dans la coordination ou dans l'état d'oxydation des éléments d'intérêt peuvent survenir. De plus, la mesure elle-même peut perturber l'équilibre initialement présent. En dépit de ces difficultés, des méthodes analytiques pour déterminer la forme spécifique des métaux dans les eaux naturelles se sont grandement améliorées au cours des dernières décennies. Ces techniques comprennent par exemple les électrodes sélectives (ISE), la voltampérométrie anodique (ASV) et cathodique (CSV) et la diffusion sur gel en couches minces (DGT) (Batley et aL, 2004). 1.3 Matière organique dissoute (MOO) La nature ubiquiste et persistante de la MaD dans l'environnement terrestre et aquatique est responsable de son rôle clé dans plusieurs processus biogéochimiques. Le cycle du carbone (Schiff et al., 1990), la pénétration de la lumière dans les eaux de surface (Scully and Lean, 1994), les réactions photochimiques (Molot and Dillon, 1997), l'acidification des eaux de surface (Schindler and Curtis, 1997), les cycles des nutriments (Dillon and Molot, 1997) et la spéciation de métaux (Buffle, 1988; Perdue and Ritchie, 2003; Wood et al., 2011) sont autant d'exemples de processus dans lesquels la MaD joue un rôle. La MaD se définit par un mélange complexe et désordonné de molécules polyélectrolytiques provenant des plantes, microbes, et produits animaux à divers stades de décomposition (Ai ken et aL, 1985; Thurman, 1985; Wetzel, 2001). C'est la présence d'une variété de groupements fonctionnels qui permet à la MaD d'exercer un grand nombre de fonctions écologiques. La variabilité dans les sources de la MaD, son transport et les mécanismes de sa dégradation 5 contribuent aussi à sa nature hétérogène. Dans la plupart des lacs, la MaD provient principalement de sources allochtones extérieures au système aquatique, c'est-à-dire depuis les sols et les plantes terrestres. La MaD allochtone consiste en un mélange de matière organique provenant du sol, à la fois la matière organique humifiée au cours du temps et la matière organique fraichement issue des plantes et du sol (Thurman, 1985). Dans un lac, les sources de matière organique autochtone sont le résultat de la sécrétion, de la décomposition et la de lyse des macrophytes, algues et cyanobactéries présents dans les eaux littorales et profondes (Thurman, 1985; Wetzel, 2001). La composition de la MaD dans un lac reflète donc la contribution relative de la matière organique provenant de sources autochtones et allochtones, et la transformation de la MaD qui survient le long de son transport dans le bassin versant vers le lac, et dans le lac lui-même (affectés par exemple par la sorption de la MaD sur les particules de sol (Schumacher et aL, 2006) et par sa photo- et biodégradation (Këhler et aL, 2002)). Par conséquent, les caractéristiques du bassin versant comme sa géomorphologie, son régime hydrologique et l'existence de lacs en amont, la couverture végétative, et l'utilisation du territoire influencent la nature de la MaD (Jaffé et aL, 2008; Mattsson et aL, 2005). La caractérisation de la MaD requiert généralement la séparation de cette dernière en isolat en utilisant la méthode développée par Thurman et Malcolm (1981). D'abord la MaD est séparée en substances humiques et en substances non humiques. Les substances humiques sont d'origine biologique et sont composées de molécules à haut poids moléculaire, réfractaires à la biodégradation. Ces substances sont réputées représenter jusqu'à 70 ou 80 % de la MaD dans les eaux naturelles (Thurman, 1985). Les substances humiques peuvent être ensuite séparées en acides humiques et en acides fulviques, ou humine, en tirant avantage de leurs différentes solubilités en fonction du pH (Thurman and Malcolm, 1981). En effet, les acides humiques sont solubles seulement à pH plus grand que 2, alors que les acides fulviques sont solubles à tous les pH. Les acides fulviques sont plus solubles parce qu'ils contiennent une grande proportion de groupes fonctionnels carboxyliques et hydroxyles. Leurs masses moléculaires sont aussi plus petites que celle des acides humiques, qui sont plus lourds et contiennent une plus grande proportion de groupements ayant des pKa similaires à ceux des groupements phénoliques 6 (Thurman, 1985). Malgré les efforts, une description précise de la structure de la matière organique dissoute n'est toujours pas disponible (Abbt-Braun et aL, 2004; Sutton and Sposito, 2005). La qualité de la MaD est un paramètre qu'il est utile d'estimer afin de mieux comprendre les rôles de la MaD dans le milieu aquatique, incluant son rôle dans la complexation et le transport des métaux. Le terme « qualité» est utilisé ici pour classifier la MaD quant à sa composition et à ces sources. Les propriétés optiques de la MaD, en particulier son absorbance (Gondar et aL, 2008; Weishaar et aL, 2003) et sa fluorescence (Baker, 2002; Coble, 1996; McKnight et aL, 2001), sont souvent utilisées pour distinguer entre différents types de MaD provenant d'environnements contrastés. Par exemple, AI-Reasi et al. (2011) ont récemment publié un article de synthèse portant sur l'utilisation des propriétés spectroscopiques de la MaD pour distinguer entre différents types de matière organique et pour prédire leurs capacités à protéger les organismes aquatiques contre la toxicité des métaux. Les mesures d'absorbance et de fluorescence sont particulièrement utiles puisqu'elles sont relativement simples, rapides et non destructives. En général, la MaD est caractérisée par la fluorescence de type humique ou de type protéique d'origine allochtone ou d'origine autochtone. Les mesures de fluorescence sont maintenant effectuées à l'aide de matrices d'excitation-émission (MEE), puisque cette approche permet d'obtenir une grande quantité d'information et de cartographier la fluorescence de la MaD en trois dimensions. De plus, l'analyse statistique multivariée, comme l'analyse en composantes principales (ACP), a été utilisée avec succès pour extraire de l'information des spectres MEE complexes (Andersen and Bro, 2003; Stedmon et aL, 2003). Lorsque l'ACP est utilisée conjointement avec les MEE, des composantes de fluorescence individuelles peuvent être identifiées ainsi que leurs contributions relatives à la fluorescence totale. Cette approche permet ainsi de caractériser la MaD le long des gradients environnementaux (AI-Reasi et aL, 2011; Baken et aL, 2011; Jaffé et aL, 2008; Miller and McKnight, 2010). 7 Étant donné que la MOD forme des complexes avec les métaux traces dans l'environnement aquatique, et puisque les caractéristiques de la MOD vis-à-vis de la complexation de métaux dépendent du type et de la source de cette dernière (Ba ken et aL, 2011; Cao et aL, 2006; Cheng et aL, 2005; Xue and Sigg, 1999), les propriétés optiques, en tant que mesures de la qualité de la MOD, sont susceptibles d'être un outil puissant pour prédire l'intensité des interactions entre la MOD et les métaux cationiques. En particulier, la qualité de la MOD, telle que déterminée en utilisant les MEE de fluorescence et l'ACP, pourrait permettre de raffiner les modèles de spéciation à l'équilibre conçus pour prédire la biodisponibilité des métaux dans les eaux naturelles (Dudal and Gérard, 2004). 1.4 Modélisation de la spéciation à l'équilibre Les modèles de spéciation à l'équilibre sont souvent utilisés pour prédire la forme spécifique des métaux dans les eaux naturelles en utilisant une approche thermodynamique avec des équations d'équilibres chimiques à stœchiométries précises. Toutefois, la nature hétérogène de la MOD rend difficile la définition de constantes thermodynamiques spécifiques qui pourraient être associées à des complexes. Néanmoins, en utilisant des hypothèses raisonnables, basées sur les connaissances actuelles sur les groupements fonctionnels de la MOD, sa propension à complexer les cations dans les eaux naturelles a pu être modélisée avec un certain succès (ex. : Kinniburgh et al. (1996), Tipping (1998)). Le logiciel de spéciation à l'équilibre Windermere Humic Aqueous Model (WHAM), combiné au modèle Humic lon-Binding Model VI (Tipping, 1998), prend en compte l'hétérogénéité naturelle de la matière organique naturelle (MON), la compétition des ions, la stœchiométrie des réactions, et les interactions électrostatiques. Le logiciel WHAM-VI incorpore des paramètres de complexation issus de l'analyse de centaines de jeux de données existant pour la complexation du proton et des métaux par les acides fulviques et humiques, obtenus pour une large gamme de conditions expérimentales. WHAM-VI a été abondamment testé en laboratoire pour étudier les interactions entre la MOD et les ions métalliques dissous (Tipping, 2002). La robustesse du modèle et de ces paramètres 8 généraux a été testée fréquemment en comparant ces prédictions avec les résultats d'expériences en laboratoire. Toutefois, il ya peu d'exemples d'études où WHAM VI a été testé pour les eaux naturelles (ex. : Bryan et al. 2002; Cheng et al. 2005; Guthrie et al. 2005; Fortin et al. 2010; Doig and Liber 2007). " est important de spécifier que le modèle WHAM-VI est basé sur l'hypothèse que l'affinité de la MOD pour les cations dans les eaux naturelles peut être représentée par l'affinité respective des acides humiques et fulviques, lesquels peuvent être isolés des eaux naturelles et étudiés en laboratoire. "est de plus supposé que le rapport « matière organique dissoute: carbone organique dissous» (MOD : COD) est de 2 : 1. En effet, puisqu'aucune technique permettant de mesurer directement la concentration de substances humiques dans les eaux naturelles n'a été développée, il est plus simple de mesurer le COD et d'extrapoler la quantité de MOD. Enfin, puisqu'une telle estimation de la quantité de MOD inclut à la fois les substances humiques et non humiques, l'utilisateur du modèle doit aussi estimer la proportion de la MOD qui correspond aux acides fulviques et humiques menant à la complexation des métaux. Cette , proportion est souvent considérée comme un paramètre ajustable lors de l'ajustement du modèle aux données mesurées. "en ressort que les choix, faits par l'utilisateur, quant à la proportion de la MOD jouant un rôle dans la complexation des métaux (fraction dite « active»), affectent considérablement (par 4 ordres de grandeur pour le Cu) les prédictions faites par la suite (Kalis et al. 2006). La sélection de la fraction active de la MOD est un défi de taille, et les études dans lesquelles ce paramètre est réellement mesuré, plutôt que simplement estimé par ajustement des données, sont rares. Par conséquent, des études qui mesurent à la fois la spéciation des métaux dans les eaux naturelles et la qualité de la MOD doivent être menées à bien pour combler cette lacune (Lofts and Tipping, 2011). De telles études permettront de mieux comprendre la qualité de la MOD, et en principe permettront d'éliminer le besoin de faire des suppositions quant à la proportion « active» de la MOD lors de l'utilisation des modèles. Elles permettront aussi de diminuer les écarts entre la spéciation des métaux prédite et celle mesurée. Toutefois, il faut 9 -------- -- admettre que même si on réussit à éliminer les artefacts provoqués par l'utilisation de la MOD isolée en laboratoire, il restera néanmoins des fluctuations temporelles et spatiales à l'échelle du bassin versant qui induiront aussi une variabilité vis-à-vis de la complexation des métaux. Ainsi, les modèles d'équilibres chimiques bénéficieraient de recherches dans les domaines suivants: 1) l'obtention de données de haute qualité sur la complexation des métaux par la matière organique sous une large gamme de conditions environnementales, 2) la mesure de la spéciation des métaux dans les eaux naturelles, 3) la mesure de la qualité de la NOM naturelle en ce qui a trait à son affinité pour les métaux et 4) la validation des modèles de complexation avec des données de terrain (Tipping, 2002). 2 2.1 Hypothèse de recherche, objectifs et originalité Hypothèse de recherche D'après des études récentes (Fortin et aL, 2010; Guthrie et aL, 2005; Unsworth et aL, 2006), les concentrations d'ions métalliques libres ([M z+]) mesurées pour le Cu, le Cd, le Ni et le Zn dans les eaux naturelles divergent souvent de celles modélisées en utilisant le logiciel WHAM-VI. Sur la base de ces résultats, nous avons fait l'hypothèse que l'échec du modèle à prédire la spéciation est causé par des variations dans la qualité de la MOD dans les différents plans d'eau étudiés, et que par conséquent l'incorporation de la qualité de la MOD dans le modèle de spéciation à l'équilibre permettrait de mieux prédire la spéciation des métaux dans les eaux naturelles. 2.2 Objectifs de recherche L'objectif de ce projet de recherche est d'améliorer les modèles de spéciation à l'équilibre en tenant en compte la variabilité spatiale de la qualité de la MOD. Un tel modèle pourrait ainsi mieux prédire la spéciation, et, par conséquent la biodisponibilité du Cu, du Cd, du Ni et du Zn dans l'environnement aquatique naturel. Pour atteindre cet objectif, il est important: 1- D'identifier une série de lacs représentant un large gradient de concentrations de métaux, de MOD et de conditions physicochimiques. 10 2- De déterminer les qualités spectroscopiques de la MOD et de relier ces variations entre les lacs aux caractéristiques de ces derniers et de leurs bassins versants. 3- De déterminer expérimentalement les concentrations de métal libre ([M z+]) dans des lacs choisis et d'explorer les liens entre la concentration de métal libre et différents facteurs environnementaux (concentration de métal total, pH, quantité et qualité de la MOD). 4- De s'assurer que les conditions expérimentales s'approchent des conditions naturelles, par exemple en titrant des échantillons de MOD provenant de lacs contrastés (quantité et qualité de la MOD, spéciation des métaux) avec les métaux dont les concentrations en ions libres prédites et mesurées divergent le plus et de comparer les constantes de complexation conditionnelles observées avec les propriétés spectrales de la MOD. 5- D'améliorer les prédictions du modèle WHAM-VI vis-à-vis de la spéciation des métaux dans les lacs choisis en incorporant la qualité de la MOD dans le modèle. Cette dernière étape est effectuée en estimant la proportion de MOD active dans la complexation des métaux (% d'acides fulviques actifs). 2.3 Originalité À notre connaissance, cette étude est la première à utiliser l'analyse APC de matrices de fluorescence comme une mesure de la qualité de la MOD, et à incorporer cette information dans les modèles de spéciation chimique à l'équilibre pour les eaux douces naturelles. Il s'agit de plus d'une des rares études à comparer les concentrations d'ions métalliques libres prédites avec celles mesurées expérimentalement, et ce dans variété d'échantillons provenant de 18 lacs dans deux régions contrastées, dans lesquelles la qualité de la MOD et les concentrations de Cu2+, Cd 2+, Ni 2+ et Zn2+ ont été mesurées. le projet contribue donc à une meilleure connaissance de la biodisponibilité de métaux d'importance environnementale. 3 Méthodologie les méthodes et le matériel employés pour l'échantillonnage et l'analyse des échantillons d'eau étudiés sont décrits succinctement ci-dessous. Une description détaillée est donnée dans les 11 articles faisant partie de cette thèse (Mueller et aL, 2012a; Mueller et aL, 2012b; Mueller et aL, 2012c) et dans la Deuxième partie de la thèse. 3.1 ÉChantillonnage Dix-huit lacs répartis dans les régions de Rouyn-Noranda et de Sudbury ont été échantillonnés à l'été 2007 avec des dialyseurs (cellules à diffusion; voir la figure 3 de la Première partie, p.48). Huit d'entre eux ont été échantillonnés à nouveau en 2008. La qualité de la MOD a été caractérisée dans chaque lac (Mueller et aL, 2012b), puis la spéciation des métaux Cd, Cu, Ni et Zn, mesurée et prédite, a été comparée pour les échantillons prélevés en 2008 (Mueller et aL, 2012c). Une paire de lacs témoins avec de faibles concentrations de métaux totaux (Lac Opasatica dans la région de Rouyn-Noranda et Lac Geneva dans la région de Sudbury) ont été échantillonnés en 2009 pour déterminer les constantes conditionnelles de complexation des métaux Ni et Cu par la MOD à l'aide d'un titrage. Enfin, les lacs Vaudray, dans la région de Rouyn-Noranda, et Geneva, dans la région de Sudbury, ont été échantillonnés en 2009 pour vérifier l'effet du stockage de la MOD sur sa qualité spectroscopique et la spéciation des métaux. Une description détaillée des lacs est donnée dans le tableau 1 (p.54). 3.2 Analyses Le matériel de laboratoire et d'échantillonnage en plastique a été lavé à l'acide HN0 3 10% (v/v)~ pour au moins 24 h, rincé avec l'eau ultrapure (>18 Mohms·cm), et séché sous une hotte à flux laminaire de classe 100. La verrerie a été lavée à l'acide HCI 2 N, puis rincée à l'eau ultrapure. Les flacons de polystyrène pour l'analyse des anions ont été rincés à l'eau ultrapure uniquement. Des sous-échantillons ont été prélevés pour l'analyse des cations majeurs, des anions, des concentrations totales de métaux, du carbone organique total (COT) ainsi que pour les mesures d'absorbance et de fluorescence. La mesure d'absorbance a été utilisée pour calculer l'absorbance spécifique à 254 nm (SUVA 254 ), tandis que les mesures de fluorescencè ont été utilisées pour calculer l'indice de fluorescence (FI, le ratio entre l'intensité de l'émission de 12 fluorescence à 470 nm sur celle de 520 nm, pour une excitation à 370 nm) et, en utilisant l'analyse de composantes principales (ACP ou PARAFAC pour paraI/el factor analysis), afin d'identifier les fluorophores individuels qui contribuent à la fluorescence totale dans chaque lac (Gondar et aL, 2008; Koster and Schmuckler, 1967; McKnight et aL, 2001). les caractéristiques des bassins versants ont aussi été identifiées en utilisant des données hydrographiques et topographique géospatiales intégrées dans le logiciel ArcGIS 10 (Environmental Systems Research Institute, Inc., Redlands, CA). les concentrations des ions Cd 2+, Ni 2+ et Zn2+ dans les lacs ont été mesurées par la technique 2 d'échange d'ions (lET), et la concentration de Cu + par électrode sélective {ISE). Enfin, les concentrations d'ions métalliques libres dans chaque lac ont été calculées avec le logiciel WHAM version 6.1. les échantillons des lacs Opasatica et Geneva prélevés en 2009 ont été filtrés sur des membranes de polycarbonate (membrane finale 0,2 ~m) et titrés avec le Cu et le Ni. Pour les titrages avec le Cu, des quantités connues d'une solution mère de Cu (0,2 mM, Cu(N0 3h) ont été ajoutées à l'échantillon, réparti en huit portions de 200 ml. le pH des solutions a été ajusté à celui de l'eau naturelle du lac en utilisant de petits volumes de solutions concentrées de HN0 3 et de NaOH; les solutions étaient ensuite équilibrées pendant 12 h sous la hotte à flux laminaire et à l'abri de lumière. la concentration de Cu 2+ libre a été ensuite mesurée par ISE. Une approche similaire a été utilisée pour les titrages au Ni, avec l'échantillon réparti en sept portions de 250 ml et des ajouts d'une solution de Ni(N0 3 h (1 mM), en laissant le pH s'équilibrer pendant 12 h avant d'effectuer la mesure par la méthode lET. les mesures de Cu 2+ 2 et Ni + ont été prises en triple sur un total de 27 portions pour le Cu et 24 pour le Ni. Après les 2 mesures de Cu + et Ni 2+, les sous-échantillons ont été analysés pour le Cu et le Ni total. 4 Résultats et Discussion Cette section contient une brève discussion des résultats principaux de la thèse. Ces derniers sont présentés en détail dans les articles de la Deuxième partie. 13 4.1 Bassins versants La recherche a été effectuée dans des lacs de la région de Rouyn-Noranda, dans le Nord-est québécois, ainsi que dans la région de Sudbury, dans le Nord ontarien. Les lacs ont été échantillonnés durant l'été au cours des trois années d'études (voir la figure 4 et le tableau 1 de la Première partie, pp. 53-54). Les caractéristiques des bassins versants des lacs diffèrent d'une région à l'autre, et d'un lac à l'autre. En effet, les lacs de la région de Rouyn-Noranda ont un temps de résidence plus court que les lacs de la région de Sudbury, et par conséquent ces derniers sont moins influencés par les apports du bassin versant. Il résulte de ces différences que la matière organique dissoute (MOD) provient davantage des lacs eux-mêmes dans la région de Sudbury et davantage des bassins versants dans la région de Rouyn-Noranda. 4.2 Chimie des eaux lacustres Les différentes caractéristiques des bassins versants, comme la géologie, l'hydrologie et le couvert végétal, jouent un rôle important dans la variabilité régionale de la chimie des eaux de surface. Les paramètres tels que le pH et la concentration des ions majeurs et des métaux sont influencés, de même que la qualité et la quantité de la MOD (Sobek et al., 2007; Wetzel, 2001). Néanmoins, le pH et les concentrations de cations majeurs n'ont varié que relativement peu entre les deux régions d'échantillonnages (voir le tableau 2 de la Première partie, p.56). Les lacs des deux régions ont un pH près de la neutralité et de faibles concentrations de Ca et Mg, ce qui est typique des eaux douces du Bouclier canadren. Les concentrations totales en Cd, Cu, Ni et Zn dissous varient largement entre les régions et entre les lacs (voir le tableau 2 de la Première partie, p.56), ce qui témoigne du choix judicieux des sites d'étude. Les plus grandes concentrations de Cd et de Zn se trouvent dans la région de Rouyn-Noranda, de même que les plus grands écarts de concentrations entre lacs. Dans la région de Sudbury, le Ni et le Cd montrent les plus grands gradients de concentrations, tandis que les concentrations de Cu et de Zn sont moins variables. La variabilité inter-lacs dans les. concentrations totales de métaux dans une région donnée est principalement due aux 14 gradients dans le dépôt atmosphérique des contaminants, ces dépôts étant influencés par les vents dominants qui transportent les émissions des fonderies et raffineries de métaux (Borgmann et aL, 1998; Dixit et aL, 2007). Certains lacs sont également influencés par des sources locales de métaux, par exemple le drainage minier acide provenant de résidus miniers. 4.3 Matière organique dissoute (MOD) les concentrations de MOD dans les lacs d'une région donnée sont influencées par la végétation terrestre, les sols et l'hydrologie, lesquels sont en retour influencés par le climat et la topographie. Néanmoins, à l'échelle locale, les propriétés du bassin versant et du lac luimême influencent les concentrations et la nature de la MOD (Sobek et aL, 2007). Pour les dixhuit lacs échantillonnés, la variabilité dans les concentrations de MOD est bien expliquée par les caractéristiques des bassins versants (voir le tableau 3 de la Première partie, p.59). En général, les lacs de Rouyn-Noranda affichent une plus grande concentration moyenne de MOD (7.2 ± 2.4 mg c·r l ) ainsi qu'une plus large gamme de concentrations (3.8 à 13.4 mg C'l- l ) que les lacs de la région de Sudbury région (moyenne de 3.4 ± 1.1 mg c-r\ concentrations de 0.8 à 6.0 mg C·l- l ). Nous formulons l'hypothèse que les lacs de la région de Rouyn-Noranda reçoivent de grandes quantités de MOD de leurs bassins versants comparativement plus grands, alors que les lacs de la région de Sudbury reçoivent moins de MOD, cette dernière étant alors produite principalement dans le lac même (MOD autochtone). Outre les différences dans la quantité de MOD, sa qualité, telle que mesurée par spectroscopie, varie aussi grandement entre les lacs (voir le tableau 3 de la Première partie, p.59). les lacs de l la région de Rouyn-Noranda possèdent un indice SUVA 254 (8.3 ± 1.5 l·m- ·mg Cl) plus élevé et l un indice FI (1.32 ± 0.04) plus bas que ceux de la région de Sudbury (SUVA 254=5.3 ± 1.3 l·m- ·mg Cl et FI=1.36 ± 0.08), témoignant d'apports accrus de MOD allochtone dans les lacs de la région de Rouyn-Noranda. le rapport élevé entre l'aire des bassins versants et celle des lacs (BIL), la grande proportion de couvert végétatif et l'épaisse couche de sols dans les bassins versants de cette région font en sorte que la MOD y provienne davantage du bassin versant. Des études précédentes ont par ailleurs trouvé une relation similaire entre les sources de MOD et la 15 fluorescence et l'absorption de cette dernière (McKnight et aL, 2001; Weishaar et aL, 2003). Par exemple, Jaffé et al. (2008) ont trouvé des valeurs de SUVA faibles (0.5 à 5 l·m-l·mg Cl) et des valeurs de FI élevées (1.4 to 1.75) pour des eaux de surface où il y avait une haute proportion de matériel d'origine microbienne, associé à des sources autochtones. Dans d'autres eaux de surface, ils ont noté des valeurs de SUVA élevées (1 à 7.5 l·m-l·mg Cl) et des valeurs de FI faibles (1.1 à 1.5), indicatives de matière organique allochtone, provenant de sources terrestres. les matrices tridimensionnelles d'excitation-émission de fluorescence (MEE) contiennent des données reflétant l'intégralité du spectre de fluorescence de la MOD. Un outil d'analyse statistique multivariée, plus spécifiquement l'analyse de composantes principales (ACP), a été utilisé pour identifier les quatre fluorophores distincts contribuant à la fluorescence totale de la MOD dans les lacs, lesquels ont été attribués à différents groupements fonctionnels de la MOD en s'appuyant sur des données acceptées dans la littérature (Coble, 1996; Cory and McKnight, 2005; Koster and Schmuckler, 1967). les composantes 1 et 2 sont attribuées aux fluorophores typiques de la quinone réduite de la matière humique présente dans une grande variété d'environnements aquatiques. la composante 3 est attribuée à un fluorophore typique de la quinone oxydée et réfractaire de la matière humique d'origine allochtone, tandis que la composante 4 est attribuée à un fluorophore typique des groupements tryptophanes des protéines d'origine autochtone. Selon l'analyse de la fluorescence de leur MOD, les lacs de la région de Rouyn-Noranda possèdent une plus grande proportion de MOD d'origine allochtone que ceux de la région de Sudbury, c'est-à-dire que ces derniers ont une plus grande proportion de MOD d'origine autochtone. Puisque la MOD est un ligand important qui influe sur la spéciation des métaux dans les eaux naturelles, nous formulons l'hypothèse que la spéciation des métaux dissous dans ces systèmes est influencée par les différences dans les types de MOD. 16 4.4 Spéciation des métaux traces Le terme « spéciation » réfère ici à la distribution des métaux entre leurs différentes formes spécifiques. Parmi ces formes possibles, l'ion libre revêt un intérêt particulier pour les écotoxicologues. La complexation de l'ion libre dans les eaux naturelles dépend des concentrations de métaux et de ligand disponibles, qu'ils soient organiques ou inorganiques, de même que de l'influence d'autres cations comme le proton et les cations majeurs qui compétitionnent pour les sites de complexation des ligands (Stumm and Morgan, 1996). Pour faciliter la comparaison entre les lacs, la proportion relative d'ions libres, plutôt que la concentration en M Z+ seule, a été calculée en divisant cette dernière par la concentration totale du métal dissous. Ainsi, dans la région de Rouyn-Noranda, le lac Opasatica possède le plus bas z pourcentage de métal libre (%M +) pour tous les métaux à l'exception de Cu, alors que le lac Dasserat affiche le pourcentage le plus élevé. Dans la région de Sudbury, le lac Bethel possède le plus petit %MZ+ pour tous les métaux, alors que le lac Raft montre les plus grands %M z + pour z tous les métaux, à l'exception de Cu. En général, le %M + augmente lorsque la concentration de métal total augmente et que le pH décroit, témoignant respectivement de l'importante du rapport métal/ligand et de la compétition entre le proton et Mz+ pour des ligands. 4.5 Spéciation des métaux traces mesurée et prédite Les modèles qui calculent la spéciation à l'équilibre des métaux, tels que le modèle WHAM, sont très utiles pour prédire la concentration d'ions métalliques libres dans les systèmes naturels qui contiennent de la MOD. Afin d'utiliser le modèle WHAM pour prédire la spéciation des métaux dans les eaux naturelles, l'utilisateur doit estimer la proportion d'acides humiques et fulviques présents dans l'eau. Cette procédure ne prend cependant pas en compte les variations dans la qualité de la MOD provenant de différentes sources. Dans cette thèse, nous avons d'abord prédit les concentrations d'ions métalliques libres pour Cd 2 +, Cu 2 +, Ni 2 + et Zn 2 + en utilisant la proportion généralement suggérée de MOD active dans la complexation des métaux, c'est-à-dire 65% (Bryan et al., 2002). Pour tous les lacs étudiés, WHAM VI prédit des concentrations de Cd 2 + similaires à celles mesurées (voir la figure 5 de la Première partie, p.64). Les prédictions du modèle sont bonnes pour le Zn aussi, bien qu'à de faibles concentrations de 17 Zn total et libre, le modèle WHAM surestime ces dernières. Au contraire, dans le cas de Cu, WHAM sous-estime l'ion libre dans la moitié des lacs, plus précisément pour ceux dans lesquels les concentrations totales de Cu sont les plus faibles. De plus, WHAM prédit des concentrations de Ni libre systématiquement plus élevées que celles mesurées, et ce pour tous les lacs dans lesquels des quantités mesurables de Ni libre ont été trouvées. Afin de prendre en compte la variabilité naturelle de la Mao entre les différents lacs, ainsi que les effets de cette variation sur la spéciation des métaux traces, nous avons testé la sensibilité du modèle WHAM vis-à-vis de la proportion de Mao active définie par l'utilisateur (%aFA). Un intervalle arbitraire a été défini en diminuant de moitié (33 %) et en doublant (130 %) la proportion de 65 % utilisée pour calculer les concentrations d'ions métalliques libres (voir la figure 5 de la Première partie, p.64). Selon cette deuxième série de simulations, le modèle WHAM prédit une proportion de métal libre similaire à celle mesurée pour la plupart des lacs pour le Cd et le Zn. Toutefois, malgré la plus large gamme de %aFA employée, l'accord entre les prédictions et les mesures fut satisfaisant seulement pour la moitié des lacs pour Cu et dans aucun cas pour Ni. Nous avons ensuite calculé, pour chaque lac et pour chaque métal, la proportion d'acide fulvique optimale (%aFAopt) qui permettait aux prédictions de WHAM et aux mesures de concentrations d'ions métalliques libres de correspondre parfaitement. Bien que l'augmentation ou la diminution de %aFA ait pour effet de modifier le nombre de sites impliqués dans la complexation des métaux, cela n'affecte pas l'affinité des métaux pour les sites. Pour le Cu, les valeurs de %aFA optimisés varient entre 7 et 90 % des concentrations de Mao totales. Pour le Cd, le Zn et le Ni, les valeurs de %aFAopt sont toutes très élevées, souvent au-dessus de la limite raisonnable de 100 % : entre 61 et 250 % pour le Cd, entre 65 et 410 % pour le Zn et entre 440 et 1900 % pour le Ni. Ces résultats suggèrent fortement que le seul ajustement de la capacité de complexation des acides fulviques (c'est-à-dire le nombre de sites) ne suffit pas, et qu'il faudra par conséquent ajuster aussi l'affinité de complexation actuellement utilisée par WHAM pour le Cd, le Zn et particulièrement le Ni. 18 Enfin, nous avons comparé la proportion optimisée de la MOD active dans la complexation des métaux aux variations inter-lacs de la qualité spectroscopique de la MOD. Une relation entre les valeurs de %aFAopt calculées et la proportion relative de la composante de fluorescence humique (Cl) est observée pour le Cu et le Ni; de même, une relation se dessine entre les valeurs de %aFAopt et la composante de fluorescence allochtone (C3) pour le Cd et le Zn (voir la figure 6 de la Première partie, p.65). De plus, la direction des relations entre les rapports Cl/CT ou C3/CT et les %aFAopt pour le Cd, le Cu et le Zn diffère nettement de la direction de ces relations pour le Ni. Ces tendances distinctes suggèrent que les sites de complexation de la MOD pour le Cd, le Cu et le Zn sont différents de ceux pour le Ni. Les relations entre les valeurs de %FAopt calculées et la proportion relative des composantes de fluorescence Cl et C3 ont été utilisées pour calculer à nouveau la concentration des ions métalliques libres pour le Cd, le Cu, le Ni et le Zn dans les lacs d'étude en utilisant WHAM VI. Cette approche a permis d'améliorer les performances du modèle WHAM VI, en particulier quant aux prédictions des concentrations de Ni libre (une amélioration jusqu'à un facteur de 6 pour le Ni; voir la figure 7 de la Première partie, p.67). Ces résultats appuient notre hypothèse selon laquelle la qualité spectroscopique de la MOD pourra être utilisée pour prédire la proportion de MOD active dans la complexation des métaux. 4.6 Paramètres de complexation mesurés Les raisons suivantes peuvent expliquer les différences entre les concentrations d'ions métalliques libres mesurées et celles prédites par le modèle WHAM : 1) la sous-estimation de la concentration des sites de complexation de métaux par unité de carbone organique dissous, et 2) la sous-estimation de l'affinité de ces sites pour les métaux d'intérêt. La première explication a été explorée ci-haut en calculant les valeurs de %FAopt pour chaque métal et pour chaque lac. Toutefois pour le Cd, le Cu, le Zn et le Ni, les valeurs de %FAopt étaient supérieures à la valeur maximale raisonnable de 100 %. Cela nous amène à supposer que c'est la sousestimation de l'affinité des sites complexation qu'il nous fait maintenant explorer. En effet, les constantes de complexation prises en compte par WHAM sont obtenues depuis des titrages en 19 laboratoire en utilisant des isolats d'acides humiques et fulviques. De plus, de telles expériences sont souvent réalisées en utilisant des rapports métaux: MOD beaucoup plus élevés que ceux normalement observés dans les eaux naturelles. Nous avons mesuré la complexation de Ni et Cu par la MOD par l'entremise de titrages pour deux lacs où les concentrations naturelles de métaux totaux sont faibles et pour lesquels la qualité spectroscopique de la MOD est différente. Ensuite, nous avons comparé les courbes de titrage obtenues à celles prédites par WHAM VI en utilisant les paramètres suggérés par le fabricant du logiciel, en fixant à 65 % la proportion de la MOD active dans la complexation des métaux, et en supposant que toute cette fraction active pourrait être représentée par les acides fulviques (voir la figure 8 de la Première partie, p.68). Les résultats montrent que, d'une part, la MOD présente dans le lac Geneva complexe plus de Ni par unité de carbone que celle du lac Opasatica, et que, d'autre part, elle complexe moins de Cu par unité de carbone. Nous nous sommes ensuite appuyés sur ces résultats pour calculer des constantes conditionnelles pour la complexation de Ni et de Cu par la MOD des lacs Opasatica et Geneva. Considérons d'abord la complexation du Ni. Son affinité pour les sites faibles de la MOD du lac Opasatica (Log Kl,Ni=6.98) est significativement plus élevée (P=0.009) que pour la MOD du lac Geneva (Log Kl ,Ni=6.30), malgré le fait que les sites faibles dans le lac Opasatica (L l ,Ni=34.2 nmole·mg Cl) soient présents à des concentrations quatre fois plus basses (P=0.01) que dans le lac Geneva (Ll ,Ni=137 nmole·mg Cl). L'affinité du Ni pour les sites forts dans la MOD n'est pas significativement différente (Log K2,Ni=11.6 et 11.07 pour le lac Opasatica et Geneva, respectivement), encore une fois malgré le fait qu'il y ait environ six fois moins (P=0.025) de sites forts par unité de carbone dans le lac Opasatica (L 2,Ni=0.70 nmole·mg Cl) que dans le lac Geneva (L 2,Ni=3.98 nmole·mg Cl). Malgré que l'affinité de complexation de Cu pour les sites faibles de la MOD ne soit pas significativement différente entre les deux lacs (Log Kl ,cu=7.59 pour le lac Opasatica et Log Kl ,cu=7.63 pour le lac Geneva), le nombre de sites faibles par unité de carbone est environ trois 20 fois plus élevé (P<O.OOl) dans le lac Opasatica (Ll ,cu=244 nmole·mg Cl) que dans le lac Geneva (Ll ,cu=75.6 nmole·mg Cl). Les sites forts pour la complexation du Cu, quant à eux, ont une affinité (Log K2,cu=9.46) significativement plus forte (P=O.014) et sont deux fois plus nombreux (P=O.003) dans la MOD du lac Opasatica que dans la MOD du lac Geneva (Log K2,cu=8.90 et L2,cu=27.0 nmole·mg Cl). Ces résultats suggèrent fortement que la MOD diffère entre les lacs Opasatica et Geneva en ce qui a trait à son affinité pour le Ni et le Cu. Les mesures d'absorbance et de fluorescence suggèrent que la MOD du lac Opasatica possède les caractéristiques d'une MOD provenant de sources allochtones, alors que celle du lac Geneva affiche une signature d'origine autochtone. Il est par conséquent raisonnable de conjecturer que les différentes affinités de complexation pour Cu et Ni trouvées pour la MOD de ces deux lacs sont reliées aux différentes proportions de MOD allochtones et autochtones. La MOD de type aromatique provenant de sources allochtones possède une affinité de complexation significativement plus grande pour Cu - aux sites forts, K2 - que la MOD de type protéique provenant de sources autochtones. Au contraire, une tendance opposée est observée pour Ni. En effet, l'affinité de Ni pour les sites faibles Ki est significativement moindre pour le lac possédant la MOD plus aromatique de type humique provenant de source allochtone. La capacité de complexation des deux types de sites pour le Ni était significativement plus faible pour le lac possédant la MOD aromatique de type humique provenant de source allochtone. De nouveau, ces résultats suggèrent que le Ni et le Cu se lient à des types de sites de complexation différents. À notre connaissance, ces résultats sont les premiers à indiquer la présence de sites spécifiques pour certains métaux dans la structure complexe de la MOD. Ces résultats indiquent l'importance de spécifier la qualité de la MOD afin de mieux prédire la complexation des métaux par la MOD dans les systèmes naturels. 21 5 Conclusions L'objectif principal de cette thèse était de démontrer que les calculs de la spéciation des métaux dans les eaux naturelles puissent être améliorés en incorporant des indices de la qualité de la MOD dans les modèles de spéciation à l'équilibre existants: nous avons formulé l'hypothèse que la qualité spectroscopique de la MOD peut être utilisée à cette fin. Fort des résultats d'études précédentes montrant les différences entre les concentrations d'ions libres mesurées et celles prédites par le code WHAM dans les lacs (Fortin et al., 2010; Unsworth et al., 2006), nous avons caractérisé la spéciation des métaux dans une série des lacs situés dans deux régions différentes du Bouclier canadien. Malgré un accord raisonnable entre les concentrations de métaux libres prédites par WHAM quant aux concentrations libres de Cd 2+ et Zn2+ pour la majorité des lacs, des différences notables ont été observées pour le Cu 2+ et le Ni 2+. En particulier, des différences marquées ont été observées pour le Cu 2+ lors que les concentrations de Cu 2+ libre étaient sous 0.4 nM (c'est-àdire pour des concentrations totales de Cu sous 40 nM). Or, de telles concentrations sont fréquentes dans les eaux naturelles. Dans le cas de Ni, WHAM surestime systématiquement les concentrations de Ni 2+ dans chaque lac. Les raisons possibles pour les différences entre la spéciation mesurée et celle prédite par WHAM ont récemment été résumées par Lofts et Tipping (2011). Ces derniers ont noté que les facteurs suivants peuvent influencer les prédictions de WHAM : 1) la précision des intrants du modèle ainsi que celle de ces paramètres, 2) l'incertitude sur les mesures de concentrations d'ions métalliques libres et 3) les différences entre les propriétés complexantes des substances humiques présentes dans les eaux naturelles et celles des isolats humiques et fulviques utilisés pour calibrer le modèle. C'est précisément cette dernière explication qui a été explorée durant cette thèse, en caractérisant la MOD et la spéciation des métaux dans les lacs d'étude. 22 Nous nous sommes intéressés à une méthode de caractérisation de la MOD qui, en principe, pourrait être utilisée par des non-spécialistes. Nous avons donc choisi des techniques spectroscopiques simples, lesquelles sont disponibles dans la plupart des laboratoires de chimie de l'eau, par opposition aux techniques fastidieuses utilisées pour isoler les substances humiques. Premièrement, nous avons mis en évidence des différences dans la composition de la MOD dans un ensemble de petits lacs du Bouclier canadien en utilisant différents paramètres optiques tels que l'absorbance spécifique UV (SUVA 254 L les indices de fluorescence (FI) et les matrices d'excitation-émission de fluorescence (MEE) interprétées par l'analyse de composantes principales (ACP). Nous avons ensuite exploré comment les propriétés optiques de la MOD peuvent être utilisées pour améliorer la prédiction, par les modèles d'équilibres chimiques, de la spéciation des métaux dans les eaux naturelles. Nous avons fait l'hypothèse que la mesure spectroscopique de la qualité de la MOD peut être utilisée pour estimer la proportion de MOD active dans la complexation des métaux. Comme mentionné précédemment, l'utilisateur de WHAM doit d'abord estimer la proportion du carbone organique dissous présent dans l'échantillon naturel considéré (mesuré en mg c·r 1 ) qui est composé de substances humiques, ainsi que la répartition de ces substances humiques entre acides fulviques et acides humiques. La pratique courante consiste à présumer un rapport MOD : COD de 2 (Thurman, 1985) et à supposer que 65 % de cette matière organique dissoute possède des propriétés de complexation des cations similaires à celles des isolats d'acides fulviques (Bryan et al., 2002). En utilisant ces principes suggérés par les créateurs de WHAM, nous avons trouvé de grandes différences entre les concentrations de Cu 2+ et de Ni 2+ mesurées et celles prédites par le modèle. Nous avons travaillé à améliorer les prédictions du modèle WHAM en calculant le pourcentage optimal d'acide fulvique impliqué dans la complexation des métaux (%FAopt) qui permet de reproduire précisément les mesures d'ion libre pour chaque métal, et dans chaque lac. Les relations entre la valeur de %FAopt et les propriétés spectroscopiques de la MOD ont été explorées. Des relations significatives ont été trouvées entre la composante de fluorescence de 23 type humique (Cl) et le %FAopt pour le Cu et le Ni, et entre la composante d'origine allochtone (C3) et le %FAopt pour le Cd et le Zn (voir la figure 6 de la Première partie, p.65). Ainsi, nous avons amélioré les capacités de WHAM à prédire la spéciation de Cu 2 + et Ni 2 + libres dans les lacs en utilisant les mesures de fluorescence pour estimer la proportion de MOD active dans la complexation des métaux. La plus grande amélioration a été observée quant aux prédictions de Ni 2+ libre (jusqu'à un facteur de 6), suggérant que la signature de fluorescence de la MOD peut être utilisée comme indice de son activité de complexation dans les modèles d'équilibres chimiques, tels que WHAM. En fait, il serait utile de systématiquement déterminer: i) la matrice d'excitation-émission pour une eau lacustre donnée, ii) de calculer la proportion relative de la composante de fluorescence Cl/CT ou C3/CT, et iii) d'utiliser cette proportion pour estimer %FA dans le lac en se basent sur une relation telle que celle illustrée dans la figure 6 de la Première partie (p.65). Bien que cette approche soit prometteuse, elle devrait idéalement être vérifiée sur une plus large gamme d'échantillons de MOD dans différents types d'eau naturelle. Afin de permettre au modèle WHAM d'estimer encore plus précisément la spéciation de Cu et de Ni, des ajustements non pas seulement à sa capacité de complexation mais aussi à son affinité pour ces métaux seront éventuellement nécessaires. En effet, nous avons trouvé que les différences dans la capacité de complexation de la MOD, c'est-à-dire dans le nombre de sites, pour Cu et Ni pouvaient être expliquées par les qualités spectroscopiques de la MOD. Dans un petit nombre de lacs (N=2) nous avons aussi trouvé que les variations entre lacs et entre métaux de l'affinité de complexation de la MOD pour Ni et Cu pouvant également être expliquée par la qualité spectroscopique de la MOD. Puisque le nombre de lacs utilisé pour cette dernière observation est petit, la relation tirée est conjecturale et devrait être étudiée dans une plus grande variété d'eaux de surface. Néanmoins, ces résultats confortent notre hypothèse à l'effet que la qualité de la MOD, telle qu'exprimée par la signature fluorescente de 24 cette dernière, soit un paramètre clé qui permettra d'estimer le rôle de la MOD dans la spéciation des métaux traces dans les eaux naturelles. l'amélioration des modèles d'équilibres chimiques permet de mieux prédire la spéciation des métaux traces dans les écosystèmes aquatiques. la justesse de ces précisions est essentielle pour pouvoir estimer la biodisponibilitédes métaux et, par conséquent, leur toxicité envers les organismes aquatiques. le modèle du ligand biotique (BlM) est couramment utilisé par les scientifiques et les législateurs pour prédire la toxicité des métaux envers les organismes aquatiques. Ce modèle utilise WHAM pour calculer la concentration d'ion libre aqueuse à proximité du ligand biotique considéré. Bien que le BlM fonctionne raisonnablement bien lorsqu'il considère des concentrations de métaux élevées et leur effets toxiques aigues, il doit être amélioré pour mieux prédire la toxicité des métaux à faible concentrations de ces derniers, c'est-à-dire lorsqu'ils sont susceptibles d'engendrer des effets toxique chroniques (Cé3mpbell et al., 2006). Augmenter la justesse des prédictions du modèle WHAM lorsqu'il considère de faibles concentrations de métaux permettrait directement de mieux prédire la toxicité chronique des métaux. la portée de cette recherche est donc basée sur le fait qu'une amélioration des prédictions de la spéciation de métaux traces dans l'environnement aquatique est nécessaire afin de mieux estimer et gérer le risque écologique posé par les métaux traces (luoma and Rainbow, 2008). 25 26 Première partie: Synthèse 27 28 1 1.1 Introduction Metals of interest Many substances present in our natural environment may be considered as prablematic with respect to human or ecosystem health. The Canadian Environmental Protection Act defines a toxic substance as one that "entering or may enter the environment in ~ quantity or a concentration or under conditions that: a) have or may have an immediate or long-term harmful effect on the environment or its biological diversity, or b) constitute or may constitute a danger to the environ ment on which life depends, or c) constitute or may constitute a danger in Canada to human life or health". Enviranment Canada and Health Canada published a Priority Substances List Assessment Report in 2001 (CEPA, 2001) within which Cu, Cd, Ni and Zn released fram copper and zinc smelters and refineries to the atmospheric, terrestrial and aquatic environments were identified as priority substances, as defined by CEPA, the potential toxicity of which should be determined. What was not clear fram this assessment was the chemical speciation of the metals released and their fate in the eventual receiving water, therefore limiting the regulators' ability to estimate their bioavailability and toxicity (Hornbrook et al., 1986). 1.2 Sites of interest Areas in Canada that are greatly affected by the presence of copper smelting and refining include the Rouyn-Noranda region of Québec and the Sudbury region of Ontario. RouynNoranda, Québec, is home to the Xstrata Copper-Horne smelter, which has been in operation since 1927, and lakes in the area have been contaminated by atmospheric de position of smelter emissions (Couillard et al., 2004; Gallon et al., 2006). The Xstrata Copper-Horne smelter is the highest producing copper smelter in Canada (190000 tonnes in 2001) and releases large amounts of Cu, Zn and Cd into the atmosphere (42, 18 and 2.5 tonnes in 2001) (Bonham-Carter, 2005). Due to the changes in smelter operations and legislated emission controls, the atmospheric emissions of Cd and Zn from the Horne smelter have decreased significantly (75% and 90%, respectively) in recent history (1989-2001), although the emission of Cu has varied considerably (Bonham-Carter, 2005). Furthermore, the concentrations of metals found in the 29 surrounding area depend on the direction of the prevailing winds and decrease with increasing distance from the smelter (Bonham-Carter et aL, 2006; Telmer et aL, 2006). There have been many studies of the total dissolved concentrations of metals in the lakes in the surrounding area of Rouyn-Noranda (e.g., Borgmann et al. (2004), Croteau et al. (2002), Guthrie et al. (2005)). In recent years, the concentration of total dissolved Cu in area lakes has been shown to vary from 33 nM (Giguère et aL, 2006) to 608 nM (Fortin et aL, 2010). The concentration Cd varies from 0.06 nM to 6.8 nM (Fortin et aL, 2010). The widest concentration range is that of Zn from 7 nM (Bonneris et aL, 2005) to 3700 nM (Fortin et aL, 2010). Conversely, Ni varied the least in the region with concentrations ranging from 8 to 12 nM (Giguère et aL, 2006). It should also be noted that certain lakes in the Rouyn-Noranda area are affected by direct run-off from past or present mining activities in addition to atmospheric de position (Goulet and Couillard, 2009). Sudbury, Ontario, also has a long history of mining, smelting and refinery activities dating back to 1929. There are currently two nickeljcopper smelters (Vale-Copper Cliff and Xstrata NickelSudbury) and one copper refinery (Vale-Copper Cliff) operating in the area (CEPA 2001). The Vale-Copper Cliff smelter was reported to release 132, 9.5, 87, and 2.4 tonnesfyear of Cu, Zn, Ni and Cd, respectively to the atmosphere in 1995 (CEPA 2001). The Xstrata Nickel-Sudbury smelter was reported to release 9.0, 1.9, 10.2, and 4.5 tonnesfyear of Cu, Zn, Ni and Cd, respectively to the atmosphere in 1995 (CEPA 2001). Similar to the Rouyn-Noranda lakes, the Sudbury lakes are greatly affected by the mining and smelting activities in the area (Carignan and Nriagu, 1985; Nriagu et aL, 1982). Moreover, the metal concentrations in lakes surrounding the smelters decrease with increasing distance from the smelters (Keller et aL, 1999; Pyle et aL, 2005; Shuhaimi-Othman et aL, 2006). Similar to the Rouyn-Noranda region mentioned above, a substantial decrease in lake water metal concentrations, especially Cu and Ni, over the past 30 years has been observed following reductions in smelter emissions (Keller et aL, 1999; Tropea et aL, 2010). Nriagu et al. (1998) found that the concentrations of total dissolved metal in lakes in the region varied from 27-4600 nM for Cu, 0.03-11 nM for Cd, 871100 nM for Ni and 3-1300 nM for Zn. 30 Both the Rouyn-Noranda and Sudbury regions are found on the Canadian Precambrian Shield characterized by bedrock made up of quartzite, gabbro and felsic gneisses (Carignan and Nriagu, 1985). However, the superficial deposits covering this underlying geology vary among regions. In the Rouyn-Noranda region, located in the Abitibi sub-province and containing the Abitibi Greenstone Belt, the bedrock is covered with a thick layer of glacial deposits, although clusters of bedrock do break through to the surface occasionally (Veillette et aL, 2005). In contrast, the Sudbury region, located on three structural provinces (Grenville, Southern and Superior) of the Shield (Rousell et aL, 2002), is covered bya thin «1 m) and discontinuous layer of glacial deposits (Barnett and Bajc, 2002). These differences in surficial geology contribute to differences in the sensitivity of the two regions to acid precipitation (Jeffries et al. 2003), and they also affect the movement of organic matter from the terrestrial to the aquatic environ ment (see section 1.4). The differences in surficial geology also contribute to differences in the topography and thus, presumably, to the differences in vegetation, soils and hydrology of the watersheds within these two regions (Sobek et aL, 2007). 1.3 Freshwater trace metal speciation The toxicity of the metals to organisms in the aquatic environment depends on their bioavailability and in turn, the bioavailability of metals in aquatic systems depends on their chemical speciation (Allen and Hansen, 1996; Batley et aL, 2004; Janssen et aL, 2003). In particular, it is the free metal ion that best predicts the bioavailability of a metal to aquatic organisms (Campbell, 1995). The free metal ion may form complexes with many ligands present in natural freshwater system. For a given complexation reaction between a metal (M) and a ligand (L) at equilibrium, M+L~ML (1) K - [ML] - [MIL] (2) the equilibrium constant K, is a measure of the stability of the formed complex (charges are omitted for brevity). In freshwater systems, simple inorganic ligands, such as OH-, HC0 3-, CO/- are present at significant 31 2 2 concentrations and can form complexes with metals ions such as Cu 2+, Cd +, Ni + and Zn2+ (Stumm and Morgan, 1996). Simple monomeric organic ligands such as citrate, glycine or polycarboxylates have high affinities for many metal ions, but due to their microbiallability, these ligands are not normally present in natural freshwater systems at high enough concentrations to influence the speciation of most metal ions (Morel and Hering, 1993). The speciatiol1 of metal ions in simple aquatic systems (i.e. inorganic and simple organic ligands) can be successfully modeled using computer speciation codes (Turner, 1995). The speciation of metals in natural waters depends on their affinity for the available ligands, and on the concentrations of these ligands. Equilibrium calculations for the four metals of interest in this thesis (Cd, Cu, Ni and Zn) suggest that the most abundant inorganic metal species in natural freshwater are: CuC0 30 and Cu(OHho; Cd 2+ and CdC0 30; Ni 2+ and NiC030; Zn2+ and ZnC0 3 0 (Stumm and Morgan, 1996). The carbonate ligand is very important in natural freshwater due to its presence at relatively high concentrations and its affinity for many diva lent metals. However, the availability of the carbonate ligand will de pend on the pH of the system (Stumm and Morgan, 1996). Alkalinity and pH, therefore, play significant roles in the speciation of many metals in natural aquatic systems. ....---- , ...... , '\ \ -..._--- .."l ' -------------" M( OH-, HC0 CÛ32-) @ ,,-----------------------, 3-, Figure 1. / Schematic representation of the interconnectedness between water chemistry and bioavailability / Représentation schématique de la relation entre la chimie des eaux naturelles et la biodisponibilité des métaux. 32 The complexation of dissolved metals with natural dissolved organic matter (DOM) is less easily quantified than is their binding to the inorganic ligands mentioned above. The heterogeneous, multi-dentate and poly-electrolyte properties of DOM make it impossible to determine specifie metal binding constants for DOM (Tipping, 2002). Binding characteristics of metal ions to DOM have, however, been estimated in the past by titrating chemically isolated organic matter fractions with metals under specifie conditions of pH, ionic strength and concentration of competing ions (see, for example, Cabaniss and Shuman (1988), Ephraim et al. (1989), Gondar et al. (2006), Higgo et al. (1993), Saar and Weber (1980) ). However, these conditional binding constants define the complexation under very specifie conditions and theoretically cannot be applied to other scenarios. Tipping (2002) summarised the effect of many variables on the binding of metal ions to DOM. First, the binding of metal ions to DOM depends on the concentration of the free metal ion, [M z+] (which depends on the total metal concentration) and the affinity of the DOM binding sites for the metal in question. The binding of metal cations to DOM will also de pend on the pH of the solution. As the pH increases, the competition between the metal cations and protons for binding sites decreases (e.g. Benedetti et al. (1995)). Depending on their relative concentrations and binding strengths, other cations in the solution can also compete with metal cations for DOM binding sites (e.g. Cao et al. (2006)). Finally, the effect of ionic strength on the binding of cations to DOM is explained by the shielding of electrostatic attractions by other ions in solution (Tipping, 2002). Chemical equilibrium models that take into account the binding of metal cations to DOM have been developed and will be discussed in Section 1.5. 1.3.1 Measuring freshwater trace metal speciation According to the International Union of Pure and Applied Chemistry (lUPAC), a chemical species is defined as "the specifie form of an element defined as to isotopie composition, electronic or oxidation state, and/or complex or molecular structure" and therefore, the distribution of such chemical species in a system is referred to as "speciation" (Templeton et al., 2000). The measurement of one or more chemical species in a natural freshwater sample is, however, not routine. Depending on the chemistry of the sample, reversible changes in the coordinate 33 bonding or oxidation state of the element of interest may occur. Furthermore, the very act of measuring the chemical speciation in a sample may perturb the equilibrium state initially present. Despite these difficulties, analytical methods to measure the speciation in natural waters have largely improved in recent decades. The review by Batley et al. (2004) gives a detailed description of the recent improvements in the detection limits, sensitivities and selectivities of methods, such as those using ion-selective electrodes (ISE), anodic stripping voltammetry (ASV), cathodic stripping voltammetry (CSV), diffusive gradients in thin films (DGT) and permeation liquid membranes (PLM). 1.4 Dissolved organic matter (DOM) The ubiquitous and persistent nature of organic matter in both the terrestrial and aquatic environments is responsible for its key role in many biogeochemical processes. In aquatic systems, an arbitrary separation of dissolved organic matter (DOM) from particulate organic matter (POM) is achieved by filtering natural water samples through 0.45 ~m filters(Wetzel, 2001). Important aquatic biogeochemical processes affected by the presence of DOM include, carbon cycling (Schiff et aL, 1990), light penetration into surface waters (Scully and Lean, 1994), photochemical processes (Molot and Dillon, 1997), the acidification of surface waters (Schindler and Curtis, 1997), nutrient cycling (Dillon and Molot, 1997) and metal speciation and transport (Buffle, 1988; Perdue and Ritchie, 2003; Wood et aL, 2011). The foremost characteristic of DOM leading to its many important roles in the environment is its complex heterogeneity. In order for DOM to be involved in such a vast range of ecologically important biogeochemical processes, a heterogeneous nature is necessary (MacCarthy and Rice, 1991). To this end, DOM is defined as a complex mixture of disordered, poly-electrolytic molecules, generated from plant, microbial, and animal products at various stages of decomposition (Aiken et aL, 1985; Thurman, 1985; Wetzel, 2001). The presence of a wide variety of functional groups that, together, express average characteristics, permits the ecological functions which rely on the presence of DOM. The importance of the nature of organic matter is weil described by MacCarthy and Rice (1991): tiAn ecologically based rationale for the chemically heterogeneous nature of humus: in the case of humus the mixture 34 is the message". In order to persist in aquatic systems, DOM must also be resistant to rapid degradation (MacCarthy and Rice, 1991; Wetzel, 2001). The complex heterogeneous nature of DOM is in part due to disorderly decomposition processes as opposed to the ordered synthesis of specifie biological compounds (MacCarthy and Rice, 1991; Thurman, 1985). Variability in sources, transport and degradation pathways also contribute to the heterogeneous nature of DOM. In most lakes, DOM originates predominantly from allochthonous sources outside the aquatic system, Le. from soil or terrestrial plant sources (Thurman, 1985; Wetzel, 2001). Allochthonous DOM may consist of a mixture of old soil organic matter that has been degraded and humified over a long period of time and relatively young organic matter that has recently decomposed from plant or soil organic matter (Thurman, 1985). Autochthonous sources of DOM from within the lake have been described as resulting from the active secretion, decomposition and Iysis of macrophytes, algae and cyanobacteria in littoral and open water (Thurman, 1985; Wetzel, 2001). The composition of DOM in a given lake will reflect the relative contributions of its allochthonous and autochthonous sources, as weil as the processing of the DOM occurring along its path from the watershed to the lake and within the lake itself (such as adsorption to soil particles (Schumacher et aL, 2006) and photo- and bio-degradation (Kahler et aL, 2002)). Watershed characteristics su ch as geomorphology, hydrologie regime, land cover, upstream waterbodies, and land use are likely to influence the composition or nature of DOM (Jaffé et aL, 2008; Mattsson et aL, 2005). Seasonal variability will also play a role in the relative importance of each DOM source. For example, stream and runoff flow as weil as growth and decay cycles of terrestrial and aquatic vegetation vary greatly with seasonal temperature changes (Wetzel, 2001). Attempts to characterize DOM often involve its separation into isolated fractions using a method developed by Thurman and Malcolm (1981). First, DOM is separated into humic and non-humic substances. Non-humic substances include, among other low molecular weight organic substances, carbohydrates, proteins, peptides and amino acids, which are relatively 35 easily utilised or degraded by microorganisms and, therefore, are present at relatively low concentrations in aquatic systems (Ai ken and Wershaw, 1985; Thurman, 1985). Humic substances, on the other hand, are coloured, high molecular weight organic substances of biogenic origin that are resistant to biological degradation. Humic substances are thought to make up 70 - 80% of the DOM in natural waters (Thurman, 1985). Humic substances can be further separated into humic acids, fulvic acids or humin based on their acid-base solubility (Thurman and Malcolm, 1981). Humic acids are soluble only at a pH greater than 2, whereas fulvic acids are soluble at any pH. Humin is insoluble in aqueous solutions at any pH. There has been some debate over the wide pH range used in the isolation of fulvic and humic acids from DOM and there is still no consensus as to whether or not isolation procedures lead to any major chemical decomposition of the DOM (Abbt-Braun et al., 2004). Fulvie acids are thought to be more water soluble because of their higher proportion of carboxylic and hydroxyl functional groups and their smaller molecular size, relative to humic acids, which are generally larger and have a higher phenolic content (Thurman, 1985). The greater aqueous solubility of fulvic acids leads to their presence in greater proportions than humic acid in natural waters. Despite many efforts, a precise description of the structure of DOM has not been achieved (Abbt-Braun et al., 2004; Sutton and Sposito, 2005). The model of DOM structure outlined by Schulten and Schnitzer (1997) illustrates its heterogeneous nature (Figure 2). These authors point out the usefulness of such models for the investigation of the structure and function of aqueous humic material under environmental conditions. Sutton and Sposito (2005), however, describe a supramolecular view of the humic substances. They describe humic substances as mixture of diverse, low molecular mass components held together by hydrogen bonds and hydrophobie interactions, ultimately forming micellar structures. A similar description of humic substances was first advanced by Khan and Schnitzer (1971). The importance of describing the character of humic substances by molecular interactions is emphasised over molecular components. Sutton and Sposito, therefore, raise the need to re-evaluate the biogeochemical formation and pathways of humic substances in the natural environment. 36 Figure 2. Representation of a proposed molecular structure for humic substances (fram Schulten and Schnitzer 1997) 2 / Structure moléculaire probable de la substance humique (tiré de Schulten and Schnitzer, 1997). ln order to better understand the role of DOM in the aquatic environment, including its role in trace metal speciation and transport, the quality of DOM is often estimated. The term quality, here, is used to describe the type of DOM based on its composition and source. Abbt-Braun et al. (2004) present a useful review of quantitative and qualitative methods used to characterize DOM. Spectroscopie methods of characterisation include UV-visible, Infrared (IR), nuclear magnetic resonance (NMR) as weil as absorbance and fluorescence spectroscopy. Optical properties of DOM have been used to distinguish among different types of DOM from contrasting environments, including absorbance (Gondar et aL, 2008; Weishaar et aL, 2003) and fluorescence (Baker, 2002; Coble, 1996; McKnight et aL, 2001). For example, a recent Note that this structure does not account for the trace amounts of sulfur that normally are found in isolated humic substances. 37 review by AI-Rea si et al. (2011) discusses the use of various spectroscopic properties of natural organic matter (NOM) to distinguish between different sources of NOM with different protective effects against metal toxicity towards aquatic organisms. Absorbance and fluorescence measurements are useful tools in that they are relatively simple, rapid, sensitive and non-destructive. In general, DOM fluorescence has been characterized as either humic-like or protein-like and has been attributed to either allochthonous or autochthonous sources. More recently, Cory and McKnight (2005) have used fluorescence spectroscopy to identify oxidised and reduced quinone-like fluorophores in DOM from a variety of aquatic environments. Fluorescence spectroscopy may therefore be used to further characterize DOM based on its redox state. Fluorescence excitation-emission matrices (EEMs) have become a popular method of DOM fluorescence measurement in recent years due to the great amount of information obtained from a three-dimensional picture of DOM fluorescence. EEMs have been used to discriminate among different types of DOM from a wide range of environments such as river water (Baker, 2002), marine and estuarine environments (Coble, 1996) and Antarctic lakes and ponds (McKnight et aL, 2001). Although the three-dimensional scans offer more information than traditional single wavelength scans, these methods usually involve the visual inspection of the spectra and identification of zone of fluorescence representative of different types of DOM. More recently, multivariate data analyses, such as parallel factor analysis (PARAFAC) methods, have been applied to EEM spectra to facilitate the extraction of useful data from the complex spectra (Andersen and Bro, 2003; Stedmon et aL, 2003). When PARAFAC analysis is used in conjunction with fluorescence EEMs, individual fluorescence components may be identified along with their relative contribution to the overall fluorescence, allowing for a more refined characterisation of DOM along a variety of environmental gradients (AI-Reasi et aL, 2011; Baken et aL, 2011; Jaffé et aL, 2008; Miller and McKnight, 2010). As DOM is known to form complexes with trace metals in aquatic environments, and since DOM metal-complexing characteristics depend on the character and source of DOM (Baken et 38 aL, 2011; Cao et aL, 2006; Cheng et aL, 2005; Xue and Sigg, 1999), optical properties, as a measure of DOM quality, may be an important tool for predicting the role of DOM in metal speciation and transport. However, the percent of DOM that is actually fluorescent has been described to be as low as 1% (Green and Blough, 1994; Senesi, 1990). Although we are unable to directly determine the proportion of DOM that is represented by the fluorescence spectrum, this approach has been successfully used to estimate metal-DOM complexation in titrations of isolated organic matter samples (Saar and Weber, 1980) and in natural untreated water samples (Yamashita and Jaffé, 2008). The incorporation of DOM quality, determined from fluorescence EEMs and PARAFAC analysis, would be a valuable addition to existing chemical equilibrium models used for the prediction of metal bioavailability in natural freshwater (Dudal and Gérard, 2004). 1.5 Chemical equilibrium speciation modelling Chemical equilibrium modelling is often employed in natural aqueous systems to predict the chemical speciation of various aqueous solutes based on established thermodynamic equilibrium reactions with defined stoichiometries. Simple chemical equilibrium modelling systems, su ch as MINEQL+, calculate the speciation, precipitation-dissolution and adsorption of aqueous inorganic chemical species using an extensive thermodynamic data base included in the model (Schecher and McAvoy, 1992). The heterogeneous nature of DOM, however, makès it difficult to define specific thermodynamic data associated with its complexes. Nevertheless, with the use of sorne reasonable assumptions based on the current knowledge of the character of DOM, the complexation of DOM with cations in natural aqueous systems has been modeled with success (e.g. Kinniburgh et al. (1996), Tipping (1998)). The calculation of complexation reactions in natural aquatic systems is possible using the Windermere Humic Aqueous Model (WHAM) speciation code combined with the HumiclonBinding Model VI 3 (Tipping, 1998). The model takes into account the heterogeneity of natural 3 WHAM has very recently been updated to include the Humic lon-binding model VII (Tipping et al. 2011). The present thesis work was, however, based on model VI which was available throughout the majority of the research project. 39 organic matter (NOM), ion competition, reaction stoichiometry variability and electrostatic interactions, but it is not yet weil tested in natural aquatic systems (Tipping, 2002). Tipping (1998) describes the Model VI as a "discrete site-electrostatic model of the interactions of protons and metals with fulvic and humic acids". The dissociation of protons on the surface of humic and fulvic acids, assumed to be rigid spheres of uniform size, is represented byeight groups having different acidities characterized by intrinsic equilibrium constants (Tipping, 1998). (3) The eight groups are further divided into four strongly acidic type A groups representing carboxylic acids and four weakly acidic type B groups representing phenolic acids. The pK; values associated with these groups are calculated as follows: for i = 1-4 pK.l =pKA . for 1 = 4-8 pK; + (2i -5) A"KA 6 ~ = pKB + (2i -13) 6 !1pKB (4) (5) where pKA and pKB are average pK values of each type of group and I1pKA and I1pKB signify the spread of the pK values around means. Furthermore, it is written into the model that the type A groups are twice as abundant as the type B groups. With respect to metal binding to type A and B groups, metal ions will compete with each other and with protons. Monodentate binding is described by the following general equation and equilibrium constants: (6) Type A groups; K(i) = logK MA + (2i -5) MK1 (7) Type B groups; K(i) = logKMB + (2i -13) MK1 (8) 6 6 where ALK1 is a constant estimated from the fitting of experimental data. Bidentate and tridentate binding is also incorporated into the model. log K(i,j) = log K(i)+ log K(j)+ x· t.LK 2 (9) log K(i, j,k) = log K(i)+ log K(j)+ log K(k)+ y. t.LK 2 (10) 40 4 Tridentate binding sites and adjustable parameters f).LK1 and f).LK2 were added to take into account the heterogeneity of binding sites, resulting in 80 different sites in ail. In general, WHAM assumes fulvic acids to have a higher density of binding sites, as weil as stronger strong acid sites and weaker weak acid sites than humic acids (Tipping, 2002). The non-specific binding caused by the electrostatic attraction between the generally negative humic substances and the cations in their immediate environ ment must also be taken into account. The accumulation of counterions in the electrical double layer formed at the surface of humic substances is approximated bya simple Donnan sub-model. The non-specific binding of counterions in the double layer adjacent to the binding surface is most pronounced at low ionic strengths, such as is the case for most freshwaters on the Canadian Precambrian Shield. A set of generic default parameters for the WHAM/Model VI has been determined by Tipping (1998). Hundreds of existing data sets for proton and metal binding to humic and fulvic acids, obtained over a wide range of experimental conditions, were analysed to generate average binding parameters. Generic default parameters offer a good starting point when applying the models to experimental data. Once the models are applied to experimental data, and based on the overall goodness of fit, several model parameters can be adjusted. In addition, in situations where experimental data are limited, the number of parameters needing to be optimised may be reduced; in the case of WHAM/Model VI, the number of adjustable parameters can be reduced to a single parameter, log KMA' where K MA is the intrinsic equilibrium constant for the binding of metal M to the type "A" binding sites (Tipping, 1998). 1.5.1 WHAM field validations WHAM/Model VI has been extensively applied and tested in laboratory experiments ta investigate the interaction between DOM and dissolved metal ions (Tipping, 2002). The robustness of this model and its default parameters have been tested by comparing simulated 4 During the development of WHAM Model VII, Tipping et al. (2011) found a coding error in Model VI, which prompted the removal of LK1 parameters from Model VII. Ail calculations using the Model VI are still correct, but do not refer to the model described by Tipping (1998). 41 results to those obtained from laboratory experiments. The reviews by Tipping (2002) and Dudal and Gérard (2004) describe at length the laboratory application and optimisation of the most recent versions of the WHAM/Model VI. Applications of WHAM to predict the speciation of trace metals in natural waters, as opposed to laboratory solutions, should be made while keeping in mind that it is a semi-empirical model based on some major assumptions, such as the size and shape of the natural organic matter as weil as density, distribution and heterogeneity of the metal-binding sites (Tipping, 2002). Another major assumption of WHAM worth mentioning is that the cation-bindingproperties of organic matter in natural waters can be represented by those of humic and fulvic acids that have been isolated from natural waters and studied in the laboratory. Note too that no technique has been developed that directly measures the concentration of humic substances in natural waters. To circumvent this difficulty, it is normally assumed that the ratio of DOM to DOC is about 2:1, the concentration of DOC being readily available analytically. However, since this estimate of DOM includes both humic and non-humic components, the model user must also estimate the proportion of DOM that corresponds to fulvic and humic acids and is active in metal complexation. This proportion is often used as an optimisation parameter to empirically fit model predictions to experimental data. For example, Bryan et al. (2002) optimised the proportion of DOM as "active" in metal complexation to be used in WHAM VI as 65% by comparing predicted and measured Cu speciation in 15 natural water samples. Cheng et al. (2005) also found that a similar 64% of DOM was "active" in Zn complexation, but only after the binding constant for Zn to fulvic acid was optimised to fit predicted to experimental speciation data. Similarly, Van Laer et al. (2006) found that the best agreement between predicted and experimental Ni speciation data for a series of natural surface water samples was found when the fraction of FA "active" in complexing Ni was optimised to 65%, along with the KMA constant for Ni, and the binding parameter determining the range of binding strengths of multidentate sites, ~LK2. It is clear that the model user's choice of the proportion of DOM that is active in metal complexation greatly affects the predicted metal speciation in natural waters (Kalis et aL, 2006). However, choosing the "correct" active fraction of DOM is a major challenge and studies 42 where this parameter is actually measured in the natural water samples, as opposed to being estimated àfter-the-fact by optimisation, are lacking. Studies that simultaneously measure the speciation of metals in natural waters and the character of DOM are needed (lofts and Tipping, 2011). One of the few studies to test the ability of chemical equilibrium speciation models (such as WHAM) to predict measured metal speciation in natural fresh waters is described in the paper by Unsworth et al. (2006). The free metal concentrations were measured by hollow fiber permeable liquid membrane and Donnan membrane techniques and 50% of the DOC was assumed to be "active" in metal complexation and to consist of 90% fulvic acids and 10% humic acids. The authors of this study found that the model predictions of free-ion activities often diverged from the observed values. The authors found that although there was reasonable agreement between the predicted and measured free metal ion concentrations for Cd 2 +, there was a poor agreement between the predicted and measured free ion concentrations of Cu 2 + and Pb 2+. The authors suggested that the poor prediction of free Cu 2+ and Pb 2+ concentrations might be due to the use of default model parameters determined in laboratory experiments with isolated humic substances. Kalis et al. (2007) found modeled free Ni 2 + to be consistently higher than that measured using the Donnan membrane technique (DMT) and cited the binding of Ni bya ligand unaccounted for by the model as an expia nation. A better understanding of the quality of DOM would in principle alleviate the need to make assumptions about the activity of DOM and minimise the discrepancies between model predicted and measured free metal concentrations. The ideal application of chemical speciation models to the binding of metal ions to organic matter in a complex, natural aquatic system draws upon an appropriate set of known parameters that describe the binding characteristics of individual metal ions under conditions where the pH, ionic strength, metal-ion concentration and speciation, and presence of competing ions are known (Milne et aL, 2003). Unfortunately, this is rarely the case. One of the key assumptions outlined by lofts and Tipping (2011) is that the cation-binding data 43 generated from laboratory experiments on samples of isolated humic substances used to parameterise chemical equilibrium models, such as WHAM, represent the cation-binding properties of DOM in natural waters. First, there is a large variability in the data sets used to generate default model parameters (Tipping, 2002). The variability may result from the use of various experimental conditions and analytical methods, and also from different sample origins. Furthermore, there is no consensus on whether or not isolation procedures used to isolate humic and fulvic acids from DOM affect the proportion of reactive sites and binding site affinities. In addition to the variability in DOM that may result from laboratory procedures, there is natural spatial and temporal variability in DOM with respect to its cation binding character. Chemical equilibrium models wou Id, therefore, benefit from future research in the following areas: 1) the generation of high quality data sets of metal binding to organic matter under a wide range of environmentally significant conditions; 2) the simultaneous prediction and measurement of metal speciation in natural samples; 3) a measure of the natural quality of NOM with respect to metal binding activity; and 4) the validation of modeled metal binding characteristics with reliable field data (Tipping, 2002). 2 2.1 Research hypothesis, objectives and originality Research hypothesis Based on recently published results (Fortin et aL, 2010; Guthrie et aL, 2005; Unsworth et aL, 2006), it was anticipated that the measured [M z+] values for Cu, Cd, Ni and Zn would not agree weil with those predicted by the "off the shelf" or default versions of WHAM VI. It was thus hypothesised that this failure of the models to predict the free metal ion concentrations accurately is primarily due to variations in the quality of the dissolved organic matter (DOM) in the different water bodies, and that the incorporation of DOM quality into the chemical equilibrium models would result in predicted values in closer agreement with the measured values in natural freshwater samples. 44 2.2 Research objectives The overall objective for this research was to enhance existing chemical equilibrium models by taking into account spatial variability in the quality of DOM. The resulting model should better predict the speciation, and therefore, bioavailability of Cu, Cd, Ni and Zn in the natural aquatic environment. In order to attain this proposed objective, five specifie sub-objectives were identified. Sub-objective 1 Identify a series of lakes representing a wide gradient of dissolved metal concentrations, DOM concentrations and physico-chemical conditions. Sub-objective 2 Determine the quality of DOM using UV-visible absorbance, 3D-EEM fluorescence and PARAFAC analysis, and link variations among lakes to watershed and lake characteristics. Sub-objective 3 Experimentally determine the concentrations of the free metal ion ([M z+]) for Cd, Cu, Ni and Zn in the chosen lakes and explore the link between the concentration of the free metal ion and its proportion (Le., its contribution to the total dissolved metal concentration) and various environmental factors (total metal concentration, pH, quantity and quality of DOM). Sub-objective 4 Based on the results of sub-objectives 2 and 3, identify a subset of lakes that exhibit marked differences in DOM (quantity and quality) and in metal speciation, and titrate lake water samples from these lakes with Cu and Ni under environmentally realistic conditions (Le., natural DOM lake samples and low metal concentrations). Generate conditional binding constants for each sample, and investigate potentiallinks between the observed binding constants and the spectral properties of the lake water DOM. 45 Sub-objective 5 Improve WHAM VI metal speciation predictions in the chosen lakes through the incorporation of DOM qua lity by adjusting the estimated proportion of DOM active in metal complexation (% active FA). 2.3 Originality Several factors contribute to the originality and significance of this doctoral thesis. First, to the best of our knowledge, this is the first study to apply fluorescence PARAFAC analysis as a measure of DOM quality to chemical equilibrium modelling for the prediction of metal speciation in natural freshwater systems. Second, the study measured DOM quality and the free Cu 2+, Cd 2+, Ni 2+ and Zn2+ concentrations in severallakes (N=18) in two different regions of Canada with wide gradients in dissolved organic carbon and metal concentrations and water chemistry. This is therefore one of the rare studies that compare predicted free metal concentrations with experimentally measured concentrations in a wide variety of natural samples. Finally, the project was designed within the research objectives of the Metals in the Human Environment Research Network (MITHE-RN). The project yields a better understanding of the bioavailability of environmentally significant metals. 3 Methodology The following section briefly outlines the mater:ials and methods used for sampling and analysing the lake water samples collected throughout the study. For a more detailed description, the reader is referred to Mueller et al. (2012a; 2012b; 2012c) found in Part Il of the thesis. 3.1 Sampling As described in the Introduction section (Section 1.2), the areas of Rouyn-Noranda, Québec and Sudbury, Ontario have been impacted by past and present copper/nickel mining, smelting and refining activities. A detailed description of the study areas can be found in Mueller et al. 46 (2012b). lakes where chosen to represent geochemical gradients in trace metal concentrations, pH and DOM quantity and quality. Eighteen lakes from both of the Rouyn-Noranda and Sudbury regions were sampled in the summer of 2007 and eight of these lakes were re-sampled in the summer of 2008. The 2007 and 2008 samples were used to characterize the DOM in the chosen lakes (Mueller et al., 2012b), while the eight samples taken in 2008 were subsequently used to compare measured and predicted trace metal speciation of Cd, Cu, Ni and Zn (Mueller et al., 2012c). A further two reference lakes with low total metal concentrations (lake Opasatica from the Rouyn-Noranda region and Lake Geneva from the Sudbury region) were sampled again in 2009 to determine conditional binding constants for Ni and Cu to DOM by titration (Mueller et al., 2012a). Lake Vaudray from the Rouyn-Noranda region and Lake Geneva from the Sudbury region were also sampled in 2009 to measure the effects of long-term sample storage on DOM spectroscopie quality and trace metal speciation. See Table 1 for a description of the lakes sampled. Lake water was collected from the littoral zones of the chosen lakes using equilibrium diffusion samplers in the summers of 2007 and 2008 (details can be found in Mueller et al. (2012b)). This sampling technique is based on the equilibrium of dissolved species in the surface waters and the ultrapure water initially held within the sampler; the diffusion occurs across a membrane of known porosity (0.2 Ilm). The sample containers consisted of 250 ml polypropylene (Nalgene) jars with custom-made plastic lids fitted with a 0.2 Ilm polysulfone filter membrane (HT Tuffryn, Pail). Before being transported into the field, the assembled jars were filled with ultrapure water (>18 Mohms) and bagged to avoid contamination. Two sample jars were horizontally fixed to plastic rods and anchored to the lake bottom (Figure 3). Three rods (Le., six sample jars) were installed in each lake at a depth of 1 m from the lake surface. The samplers were left to equilibrate in the lakes for 13 to 19 days before being removed. 47 Figure 3. Example of the sampling apparatus / Montage utilisé pour l'échantillonnage. The membrane of each sample jar was rinsed with ultrapure water and then gently pierced with a clean plastic pipette tip before replacing the membrane cover with a clean polypropylene screw lid. The samplers were bagged and stored in the dark at 4°C until transported back to the lab. In the field, lake water temperature, conductivity and pH were measured. ln the summer of 2009, large volume grab samples were collected from the surface of the lakes for the titration and sample storage experiments. For the water needed for the titration experiments, two 4-L amber glass bottles (I-CHEM Brand) were used for each of the two lakes sampled. For the sample storage experiments, one 10-L polypropylene carboy (Nalgene) was used for each of the two lakes sampled. Ali grab samples were taken from approximately the 48 same location as the equilibrium diffusion samplers and transported back the laboratory where they were stored in the dark and at 4°C. 3.2 Sample analysis Ali plastic laboratory and sampling equipment was soaked in a 10% (v/v) HN0 3 solution for at least 24 h and then rinsed a minimum of six times with ultrapure (>18 Mohms·cm) water and dried under a Class 100 laminar flow hood. Ali glassware was soaked in a 2 N HCL solution before being rinsed with ultrapure water. The polysulfone membrane for the diffusion samplers was soaked in ultrapure water, changed daily, for 1 week. The polycarbonate filters for vacuum filtration were soaked in 10% HN0 3 for at least 24 h and then ultrapure water for at least 24 h. Polystyrene vials for anion analyses were rinsed three times with ultrapure water only. 3.2.1 Lake DOM quality Details of the sample preparation and analyses of the samples taken in 2007 and 2008 to characterize the lake DOM can be found in Mueller et al. (2012b). Briefly, subsamples were taken for analyses of major cation concentrations, total organic carbon (TOC) as weil as absorbance and fluorescence measurements. Absorbance measurements were used to calculate the specifie absorbance at 254 nm (SUVA 254 ), while the fluorescence measurements were used to calculate the fluorescence index (FI; the ratio between the fluorescence emission intensity at 470 nm over that at 520 nm, for an excitation at 370 nm) and, using parallel factor analysis (PARAFAC), to identify the individual fluorescent components contributing to the overall DOM fluorescence for each lake (Cory et al., 2010; Gondar et al., 2008; Koster and Schmuckler, 1967). Lake watershed characteristics were also identified using hydrographie and topographie geospatial data integrated using ArcGIS 10 (Environmental Systems Research Institute, Inc., Redlands, CA). 49 3.2.2 Cd, Cu, Ni and Zn metal speciation Details of the sample preparation and analyses of the samples taken in 2008 to compare measured and predicted trace metal speciation ean be found in Mueller et al. (2012c). Briefly, subsamples were taken for determination of major cation, anion and total metal concentrations. Total organic carbon (TOC), total inorganic carbon (TOC), as weil as absorbance (SUVA 254 ) and fluorescence (FI and PARAFAC components) measurements were also made. The concentrations of Cd 2+, Ni 2+ and Zn 2+ in the lake samples were measured using an ion exchange technique (lET), whereas the concentration of Cu 2+ was measured using a cupric ion selective electrode (ISE). The concentrations of free metal ions for each lake were also calculated using the Windermere Humic Aqueous Model (WHAM model 6.1). 3.2.3 DOM titrations with Cu and Ni The grab samples taken from Lake Opasatica (8 l) and Lake Geneva (8 l) in 2009 were transported back to the laboratory and vacuum filtered under a clean laminar flow hood using polycarbonate filter holders (Advantec MFS, Inc.) and a series of 47-mm polycarbonate filters (S, 2, 0.4 and 0.2 !lm; Millipore and AMD Manufaeturing Ine.). Before filtering the lake water samples, 1 lof ultrapure water and a small volume of sample were filtered through the filtration apparatus and discarded. Following filtration, the samples were stored in the dark and at 4°C until analysis. Samples from both lakes Opasatica and Geneva were individually titrated with Cu and Ni. For the Cu titrations, known amounts of a stock solution of 0.2 mM copper nitrate (CU(N03h, PlasmaCal) were added to eight 200-ml aliquots of the sample, which had been amended with KN0 3 (99.995%, Fluka; a known amount of a 1.23 M stock solution was added to achieve a final concentration of 10 mM KN0 3 for ail titration aliquots). The pH of the titration aliquots was adjusted to the naturallake pH using small volumes of either HN0 3 or NaOH, and left to equilibrate overnight under a laminar flow hood and in the dark before measuring the free Cu 2 + concentration using a cupric ISE. An initial sample aliquot with no added Cu was also analysed. For details on the calibration of the electrode and measurement of the free Cu 2 + ion in natural 50 lake water samples, the reader is referred to Mueller et al. (2012c). Triplicate Cu titrations were carried out for a total of 27 Cu titration aliquots analysed. After the measurement of the free Cu 2 + concentration, subsamples of each titration aliquot were taken for measurement of the total dissolved Cu concentration. For the Ni titrations, known amounts of a stock solution of 1 mM nickel nitrate (Ni(N0 3h, PlasmaCal) were added to seven aliquots of 250 ml of sample, which had been amended with Ca(N0 3 h (99.0%, Sigma-Aldrich; a known amount of a 0.025 M stock solution was added to achieve a final concentration of 0.46 mM Ca(N0 3 h for ail titration aliquots). Again, the pH of the titration aliquots was adjusted to the naturallake pH and left to equilibrate overnight before measuring the free Ni 2 + concentration using the lET. An initial sample aliquot with no added Ni was also analysed. For details on the calibration and measurement of the free Ni 2+ ion in naturallake water samples, the reader is referred to Mueller et al. (2012c). Triplicate Ni titrations were carried out for a total of 24 Ni titration aliquots analysed. In the same manner as the Cu titrations, subsamples of each titration aliquot were taken for measurement of the total dissolved Ni concentration. 3.2.4 Effect of sam pie storage on metal speciation and DOM quality Following the same protocol as described above (Section 3.2.3), the 10-l samples collected from Lake Vaudray and Lake Geneva in July 2009 were filtered upon return to the laboratory. Subsamples were taken for analyses of pH, major ion concentrations, DOM optical properties (absorbance and fluorescence) as weil as total and free Cd, Cu, Ni and Zn concentrations immediately after filtration and at regular intervals (approximately 3 months) over an 8-month period. Between subsampling events, the large volume samples were stored in the dark at 4°C, as was the case for the lake water samples collected with the in situ diffusion samplers. See Appendix A for further details. 51 4 Results and Discussion The following section includes a discussion of the major results found during the completion of this thesis. More detail can be found in the papers found in Part Il of the thesis. The overall objective of this thesis was to enhance existing chemical equilibrium models by incorporating the quality of DOM, resulting in improved predictions of Cd, Cu, Ni and Zn speciation, and therefore, bioavailability in natural aquatic environments. To this end, the effects of watershed characteristics and in-Iake processes on lake water chemistry are first presented, with an emphasis on the DOM present in the various lakes. For a subset of these lakes, chosen to represent differences in DOM and in water chemistry (including total dissolved metal concentrations), we then examine the free-metal ion concentrations for four metals (Cd, Cu, Ni and Zn) and consider how the trace metal speciation varies among the metals and for a given metal among the various lakes. The measured trace metal speciation is then compared to that predicted by the commonly used chemical equilibrium speciation model, the Windermere Humic Aqueous Model (WHAM) and possible reasons for differences between the measured and WHAM modeled trace metal speciation are discussed. Finally, the conditional binding parameters for the complexation of Cu and Ni by DOM are calculated and compared between two lakes with low ambient concentrations of metals and spectroscopically different DOM. 4.1 Lake watersheds During the thesis research, lakes from the Rouyn-Noranda region of north-western Québec and from the Sudbury region of north-central Ontario were sampled during the summer over the course of three years (Figure 4 and Table 1). The watershed characteristics of these lakes were found to vary between sampling regions. In general, the lakes from the more northerly Rouyn-Noranda region had a higher mean watershed-to-Iake surface area (WjL) ratio (25), % vegetation (83%; dominated by coniferous forest cover) and % wetland cover (1%), fewer rock outcrops and a thicker soillayer than the lakes from the more southerly Sudbury region (mean W/L of 9; 41% vegetation; no wetland cover and 28% rock outcrops) (see Table 2 in Mueller et al. (2012b)). 52 1 ' 4! ... :l ... ......... b a. Rouyn No nda 'lt:... l t 1I _ w.1 a Hudson by , , .~ .; v....ro. L .. " Quêbec • M ,• , !l- ..... c -,.'". Y-- , - - - - Onlario ~ ~ y \ 1 1---_ L l Sudbury * r i 1l1-:wrvv Figure 4. Map of the two study areas showing the relative position of the two regions in eastern Canada (panel a), the geographicallocation ofthe Rouyn-Noranda lakes (panel b) and the Sudbury lakes (panel cl. Reproduced from Mueller et al. (2012b) / Carte de l'Est du Canada montrant les positions relatives de deux régions étudiées (A), ainsi que l'emplacement de lacs de la région de Rouyn-Noranda (8) et de la région de Sudbury (C). Reproduit de Mueller et al. (2012b). It follows that the lakes from the Rouyn-Noranda region tend to have shorter water residence times than the lakes from the Sudbury area, and accordingly are strongly influenced by inputs from their catchments. With respect to natural organic matter, which can originate either within the lake or in the catchment, the Rouyn-Noranda lakes receive a larger proportion of DOM from their catchments than do the Sudbury area lakes. 53 Table 1. Lake names, sampling regions, locations and years sampled. Modified from Mueller et al. (2012b) / Noms des lacs, ainsi que les régions, sites et dates d'échantillonnage. Modifié d'après Mueller et al. (2012b). Years Sam pied Lake Region Location 2007 2008 Adeline Rouyn-Noranda 48°12'12" N 79°10'17" W y Bousquet Rouyn-Noranda 48°12'56" N 78°37'06" W y D'Alembert Rouyn-Noranda 48°23'01" N 79°00'35" W y Dasserat Rouyn-Noranda 48°15'06" N 79°24'24"W y y Dufault Rouyn-Noranda 48°18'30" N 78°59'25" W y y Dufay Rouyn-Noranda 48°01'42" N 79°27'33" W y Joannès Rouyn-Noranda 48°10'53" N 78°40'02" W y Opasatica Rouyn-Noranda 48°04'32" N 79°17'49" W y y Vaudray Rouyn-Noranda 48°05'45" N 78°40'50" W y y Bethel Sudbury 46°46'17" N 80°57'43" W y y Crowley Sudbury 46°23'06" N 80°58'55" W y Geneva Sudbury 46°45'52" N 81°32'46" W y Hannah Sudbury 46°26'36" N 81°02'17" W y Middle Sudbury 46°26'22" N 81°01'26" W y Nelson Sudbury 46°43'36" N 81°05'39" W y Raft Sudbury 46°24'35" N 80°56'44" W y Ramsey Sudbury 46°28'57" N 80°57'00" W y Whitson Sudbury 46°35'03" N 80°58'20" W y 4.2 y 2009 y y y y Lake water chemistry Differences in watershed characteristics, such as geology, hydrology and vegetation cover, between regions can play an important role in the regional variability of surface water chemistry parameters such as the pH and the concentrations of major ions and metals, as weil as the quantity and quality of DOM (Sobek et al., 2007; Wetzel, 2001). However, the water chemistry, Le. pH and major cation concentrations, varied only slightly between the two 54 pHs,with low sampling regions(seeTable2). Lakes from both regionshavecircumneutral concentrations of Caand Mg,typicalof softwaterlakeson the Canadian Shield. Thetotaldissolved concentrations of Cd,Cu,Ni andZn did varybetweenregionsandamong the eightlakessampledin 2008(seeTable2),aswasthe aimof the experimental design.The Rouyn-Noranda regionhadthe highestconcentrations of CdandZn andshowedthe largest gradients concentration amongthe lakesfor thesemetals,whereasCuand Ni concentrations variedless.LakesDufaultand Dasserat hadthe highestconcentrations of totaldissolved metals,whileLakeOpasatica hadthe lowest.In the Sudburyregion,Ni and Cdshowedthe gradients largestconcentration amongthe lakes,whereasCuandZn variedless.LakesBethel andGenevahadthe lowestconcentrations of metals,whileLakesRaftandWhitsonhadthe highest.Thelake-to-lake variability in measured total metalconcentrations withina given regionis dueto variations in atmospheric deposition down-wind, with respectto the distance presentin individual from miningandsmeltingactivities, andto localpointsourcedischarges (Borgmann catchments et al.,1998;Dixitet al.,2007;GouletandCouillard, 2009). 4.3 Dissolvedorganicmotter (DOM) Dissolved organicmatterplaysan importantrolein controlling the speciation and bioavailability of manytracemetalsin naturalaquaticsystems, by complexing the metalsand loweringthe concentration of the free metalion (Buffle,1984).lt is reasonable to hypothesise, however, giventhe recognized diversity of the originsof DOM,that its abilityto complextracemetals mayvaryfrom laketo lake,dueto differences in precursor materialwith differentmetalgroups.Forexample, specific functional the Cubindingby DOM,aswellasthe uptakeof Cuby fishgills,wasfoundto differamongsamples with differentsources of DOM(Luideret al.,2004). 55 Table2. MeasuredpH and mean(+ SD,N=3)dissolved organicmatter(DOM)concentrations, total dissolvedconcentrations of majorcations andtrace metals,ando/o (QC)and Sudbury(ON)in 2OO8 free metalvaluesfor lakessampledfrom Rouyn-Noranda /Valeursde pH, de concentrations moyennes(t SD,N=3)de matidreorganiquedissoute(MOD),de concentrations totalesde cationsmajeurs (%)de m6tallibrepour leslacs6chantillonn6s dissouset de m6tauxtracesainsique lespourcentages danslesr6gionsde RouynNoranda(Qu6.1et de Sudbury(Ont.)en2008. Dasserat pH 6.92 7.59 (mg DOM C'L-1) 4.85r 0 03 5 . 2x 0 . 1 Ca (uM) 2 0 7r . 1 4 0 4t 3 Ms (pM) 93.7i 0.3 118*1 Cd (nM) 0 . 8i 0 . 2 2 . 0r . 0 . 2 Cu (nM) 112t3 246t3 Ni(nM) 8 r.2 5 . 1i 0 . 3 Zn (nM) 464 t 51 342r.60 cd2.1nM; r 0.5 0.2 1 . 1 4i 0 . 0 6 cu2.1nM; 0.37r 0.04 0.44r 0.09 < 0.62 Ni2-1nM; 0.6' zn2.1nM; 279 x32 188r23 o/o Cd2* o/o Cu2* %oNi2* o/oZnz* Opasatica Vaudray 7.69 6.81 6 . 3r 0 . 1 8 . 1 0i 0 . 0 8 218r.2 83rl 1 1 4r . 1 3 7 . 1r 0 . 6 0.06r 0.01 0.608r 0.005 43t1 3 5 . 1* 0 . 2 1 2 . 5t 0 . 4 12t1 15x4 5 2x s 0.021r 0.008 0.26t0.02 0.08r 0.01 0.20r 0 08 < 0.61 <0.12 2 . 3i 0 . 9 31i18 63r36 58i5 3 7x 1 7 42t 4 0.32t 0.04 0.17r 0.04 0.21t 0.04 0.47t 0.2 NA NA NA 8' 60f10 5 5r . 1 2 15t7 60r36 8.08 4.89i 0.03 4 6 4t 3 274t3 0 . 0 1 8i 0 . 0 0 9 24.9t 0.1 226 r.2 12t3 6.9 2 . 9* . 0 . 1 68r3 2 6t 1 0.021r 0.005 1 0 . 5r 0 . 6 1 8r . 2 2 1t 8 0.05r 0.01 7.3f09 2 . 3t 0 . 4 NA 0.48r 0.09 4 . 0t 0 . 7 1 0 . 7i 0 . 9 70 4 . 6r 0 . 9 22ts 51 t20 o.o1' 0.20r 0.05 3.2 tO.4 20r6 Raft 6.4 2.03r 0.03 8 1t 2 46r1 0.56r 0.02 8 6 . 0i 0 . 9 775 i.4'l 62t22 0 402 t 0.004 3 . 2r 0 . 5 545r9 5 1r 1 6 71t2 3.7r 0.6 70t4 82i39 Whitson 6.9 3.77x0.05 1 4 3* 1 78.4x0.4 0.51i 0.04 1 6 5t 3 1136i 54 52t2 0.31r 0.05 1 . 0r 0 . 1 489t72 32r.4 61r10 0.57i 0.08 43t7 62r8 An asterisk* indicatesonly one sampleabovelimit of quantification (LOQ;calculated asthreetimesthe standarddeviationof triplicateanalyses of each lake).NA indicatesno samplesabovethe LOQ.Modifiedfrom Muelleret al.(2012c). (LOQ;calculdecomme6tanttrois fois l'6carttype sur Lesymbole* d6noteun r6pliquadont lesconcentrations sontau-delide la limitede quantification une analyseen triple danschaquelac)et le symboleNAd6notequ'aucunechantillonn'estau-delide la LOQ.Modifi6d'aprEsMuelleret al. (2012c). DOM Quontity Theconcentration of DOM,hereafterexpressed in termsof DOC(mgC.L-1), in lakesin a given regionis influenced by terrestrialvegetation, soilsand hydrology, whichare in turn regulated by climateandtopography.Howeverat a morelocalscale,catchment andlakeproperties influence the DOMconcentration in individual lakes(Sobeket al.,2007).Indeedfor the playedan importantrolein eighteenlakessampledin 2007,the lakewatershed characteristics the variability in DOMconcentration amonglakes(seeTable3). In general, the lakesof the Rouyn-Noranda regionhadhighermeanconcentrations of DOM(7.2 t 2.4 mgc.t-1)and 1) spanneda widerconcentration range(3.8to 13.4mg C.L thanthe lakesof the Sudburyregion (meanof 3.4t 1.1mg C.L-l;rangefrom 0.8to 6.0mg C.t-1).LakesBousquet andJoannds, in the Rouyn-Noranda region,hadthe highestconcentrations of DOM(13.0and 10.0mg C.L-1, (1.4 respectively), whereasLakeNelson, of the Sudburyregion,hadthe lowestconcentration mg C.L1). As described in section4.L,we suspected that the lakesfrom Rouyn-Noranda region receivehighquantities whereaslakesfrom the Sudbury of DOMfrom their largecatchments, regionreceiveloweramountsof DOMfrom theircatchments, and relatively higheramounts from withinthe watercolumn.Thewiderangeof DOMconcentrations in the lakessampled fulfilledoneof our objectives a seriesof lakesrepresenting of identifying a widegradientof DOMconcentrations. DOM Quality In additionto differences in the quantityof organicmatter,the qualityof the lakeDOM(as measured spectroscopically) alsodifferedbetweenregionsandamonglakes(seeTable3). The (8.31 L.5L.m1'mgC-1) Rouyn-Noranda lakesdisplayed higherSUVAzsq andslightlylowerFl values(1,.32 !0.04) thanthe Sudburylakes(5.3t 1.3L.m-l.mg L.36t 0.08,respectively), C-1and indicative of greaterinputsof allochthonous DOMto the Rouyn-Noranda lakesthanthe Sudburylakes.Thehighwatershed-to-lake area(WL) ratiosandthe largeproportionof surfacevegetation in the catchments lakesalsopointto a higher of the Rouyn-Noranda proportionof DOMoriginating from the catchment thanin the caseof the Sudburylakes. Amonglakes,LakesBousquet, VaudrayandJoannds hadthe highestSUVA254 valuesandthe 57 lowestFlvalues, whileLakesBetheland Ramsey valuesandthe highest hadthe lowestSUVAzsq Flvalues.Lakes Joannds andVaudraybothflow into LakeBousquet andthe priorin-lake processing of DOMin theselakespresumably natureof the DOM to the recalcitrant contributes presentin LakeBousquet.In otherwords,the DOMoriginating from the catchments of Lakes Joannds andVaudray,andfor that generated to withintheirwatercolumnsis exposed pathways degradation on itswayto LakeBousquet.A lowerproportionof easilydegradable DOManda higherproportionof recalcitrant to be presentin Lake DOMarethereforeexpected Bousquet.BothLakeRamsey and LakeBethel,a smalleutrophiclakethat flowsinto Lake Ramsey, arelikelyproducing DOM,thusleadingto their a highproportionof autochthonous low specific absorptivities and highFlvalues.Previous studieshavefoundsimilarrelationships (McKnight betweenthe sourceof DOMandsimpleabsorption measurements andfluorescence et al.,2001;Weishaar et al.,2003).Forexample, Jaff6et al.(2008)foundlow SUVAvalues(0.5 t to 5 L.m-l.mg C ) and highFl (1.4to 1.75)values for surfacewaterswith a highmicrobial foundin autochthonous contribution samples.In othercases, theyfoundhighSUVAvalues(1 to 7.5L.m-l.mg C-1) andlow Flvalues(1.1to L.5),indicative materialfound of terrestrial-derived in allochthonous samples. Balcarczyk et al.(2009)alsofounda rangeof stream,springand thermokarst samples to exhibitFlvaluesto rangebetween1.3and 1.4,representative of terrestrially-derived allochthonous DOM. Unlikethe simpleabsorbance andfluorescence measurements that useonlya smallportion (two-dimensional) partof the absorbance or fluorescence spectrum, three-dimensional excitation-emission matrix(EEM)fluorescence incorporates datafrom the entire spectroscopy DOMfluorescence spectrum.Theuseof a multivariate statistical analysis tool,i.e.parallel factor(PARAFAC) analysis, on EEMfluorescence four distinctfluorescent spectraidentified components contributing to the overalllakeDOMfluorescence. Components l and2 were presentin a widevarietyof aquatic ascribed to reducedquinone,humic-like fluorophores quinone,humic-like environments, component3 to refractory, fluorophores of oxidized proteinfluorescence allochthonous origin,andcomponent 4 to tryptophan-like of 58 Table3. Measuredmean(+ SD,N=3)dissolved (SUVAzas), organicmatter(DoM)concentration, fluorescence index(Fl), specific UV-absorbance and relativeproportions of the C1andC3PARAFAC for the eighteenlakessampledin 2OO7 components andthe eightlakessampledin 2OO8 moyennes(t SD,N=3)de matidreorganiquedissoute(MOD),absorbance indicede UV sp6cifique(SUVA2or), / Concentrations (Fl)et lesproportionsrelativesdescomposantes fluorescence de fluorescence C1et C3 pour les18 lacs6chantillonn6s en 2007et les8 lacs6chantillonn6s en 2008. Lake Region DOC suvA 254 SAC340 (mg t-'') (L m-l mg c'1) (t m'l mg c-1) FI cLlcf czlcr c3lcr c4lcr Adeline R o u y n - N o r a n d a 4 . 6 0! O . 7 9 6 . 9 6! r . 2 7 r . 4 6! 0 . 2 5 1 . 3 3r 0 . 0 1 0 . 2 7! 0 . o 2 0.33r 0.02 0.26t 0.02 0.15r 0.01 Bousquet R o u y n - N o r a n d a 1 2 . 5 8r 0 . 6 9 1 1 . 0 3r 0 . 6 2 3 . 5 2I 0 . 2 0 1 . 2 9r 0 . 0 1 0.15r 0.02 0.07r 0.01 0.76r 0.05 0.0210.00 D'Alembert Rouyn-Noranda 8.00t 0.80 8.67t 0.90 2 . 2 0! 0 . 2 2 1 . 3 0r 0 . 0 1 0 . 2 1 1 0 . 0 30.24r 0.03 0.48r 0.07 0.08r 0.01 Dasserat Rouyn-Noranda 5 . 1 4r 0 . 5 8 8.00r 0.40 1 . 9 8i 0 . 1 8 1 . 3 5I 0 . 0 3 0.25r 0.19 0.34r 0.25 0 . 3 0 1 0 . 3 2 0.10t 0.07 Dufault Rouyn-Noranda 5 . 1 0r 0 . 3 6 6.86i 0.70 L . 4 8! 0 . 2 6 1 . 3 2r 0 . 0 3 o . 2 7! O . L 2 0 . 3 3 1 0 . 1 4o . 2 7! O . 7 0 0 . 1 4 1 0 . 0 5 Dufay Rouyn-Noranda 8 . 2 51 0 . 3 7 8.36r 0.44 2 . 1 1I 0 . 1 3 1 . 3 0 1 0 . 0 2 o.22! O.03 0.24r 0.03 0.45! 0.05 0.09r 0.01 Joannds Rouyn-Noranda 1 0 . 2 6! 0 . 2 6 9 . 9 8 10 . 2 5 2 . 9 0 1 0 . 0 6 1 . 2 8I 0 . 0 1 0.18i 0.04 0.16r 0.03 0.61t 0.16 0.05r 0.01 Opasatica Rouyn-Noranda 6.00r 0.40 7 .47! 0.34 1 . 6 5i 0 . 1 4 1 . 3 5r 0 . 0 2 0.26r 0.05 0.32r 0.05 0.29r 0.05 o.tz!o.o2 Vaudray Rouyn-Noranda 8 . 0 7 1 0 . 0 9 9 . 8 31 0 . 4 3 2 . 8 8! 0 . 2 4 1 . 2 61 0 . 0 1 0.27r 0.08 0 . 2 4! O . O 7 0.43t 0.10 0.07r 0.02 Bethel Sudbury 5 . 2 2! 0 . 4 6 4.64t0.47 0 . 8 4r 0 . 2 5 1 . 4 91 0 . 0 5 0.2410.08 0.37t 0.13 0 . 1 5 1 0 . 0 5 0.23i 0.09 Crowley Sudbury 2 . 5 9I 0 . 2 8 4.82! 0.44 1 . 0 2! 0 . 1 2 1 . 3 31 0 . 0 8 0.29!0.o7 0 . 3 6 1 0 . 0 80.16r 0.05 0.19r 0.04 Geneva Sudbury 3 . 2 6r 0 . 3 9 5 . 2 31 0 . 2 8 1.0910.25 1 . 3 31 0 . 0 6 0.29r 0.09 0.37r 0.10 0 . 1 81 0 . 0 7 0.17r 0.06 Hannah Sudbury 3.61!O.r2 5.05r 0.26 1.03r 0.08 1.31t 0.04 0.28i 0.04 0.34r 0.04 0.18r 0.03 0.2010.03 Middle Sudbury 3.32r 0.40 5 . L 7! 0 . 5 7 1.01r 0.08 1.35i 0.07 0 . 2 5 1 0 . 0 30.34r 0.04 0.20r 0.03 0.20t 0.03 Nelson Sudbury 1 . 3 6t 0 . 4 6 4 . 5 3! 2 . O 2 0 . 8 31 0 . 4 2 1.32I 0.03 0.28r 0.08 0.35t 0.09 0.14r 0.03 0.2310.07 Raft Sudbury 2 . 2 6! O . 3 r 5 . 9 9! 2 . 3 2 1 . 3 5r 0 . 7 1 I . 2 8 ! O . 0 7 0.29r 0.07 0.34t 0.08 0.15t 0.16 0.22r 0.06 Ramsey Sudbury 3 . 5 5r 0 . 5 3 4 . 3 21 0 . 6 0 0 . 7 41 0 . 0 9 1 . 4 1 1 0 . 0 2 0.24t 0.05 0.33r 0.05 0.17r 0.04 0.25r 0.05 Whitson Sudbury 3 . 8 2! 0 . L 2 6 . 8 7r 0 . 5 6 t . 7 3! 0 . 3 4 7 . 3 4! O . 0 4 0.30r 0.06 0.36r 0.08 0.19r 0.04 0.15r 0.03 values(Coble,1996; literature autochthonous origin,basedon the comparison with established Coryand McKnight,2005;Kosterand Schmuckler, L9671. DOMfluorescence from the lakesof the Rouyn-Noranda regionhada higherrelativeproportion of allochthonous DOMfluorescence thanthe Sudburylakes,whereasthe Sudburylakeshada higherproportionof autochthonous DOMfluorescence. OnceagainLakesBetheland Ramsey origin, werefoundto havehighproportions DOMof autochthonous of proteinfluorescent fluorescent whereasLakes Vaudray, Joannds hadhighproportions of humic-like and Bousquet DOMof allochthonous the speciation origin.Giventhat DOMis an importantligandregulating metalsin of tracemetalsin naturalwaters,we hypothesised of dissolved that the speciation freshwatersystemswouldbe sensitiveto the differenttypesof DOMpresent. 4.4 Tracemetalspeciation Forthe purposeof the thesis,tracemetalspeciation of trace is definedasthe distribution is the free metal metalsamongsttheirchemical forms;of particular interestto ecotoxicologists ion. Thecomplexation of the freemetalion in naturalwaterswill dependon the concentration of the metalandavailable ligands(bothorganicandinorganic), of other aswellasthe influence cations(suchas protonsandmajorcations)competing for complexation siteson suchligands (Stummand Morgan,1996).Of the lakesstudied,thosethat hadthe highesttotal dissolved (seeTable2). The metalconcentrations alsohadthe highestfreemetalion concentrations relativeproportionof the freemetalion in individual in orderto lakeswasalsoconsidered accountfor differences in the totalmetalconcentrations amonglakes.In the Rouyn-Noranda region,LakeOpasatica hadthe lowestTofreemetalforall metals(exceptCu)and LakeDasserat hadthe highest.In the Sudburyregion,LakeBethelhadthe lowesto/ofree metalfor all metals and LakeRaft,the highestpercentmetals(exceptCu).Thepercentfreemetalincreased as a pH,reflecting functionof increasing totalmetalconcentration the importance anddecreasing of the metal-to-ligand ratioandprotoncompetition respectively. on metalspeciation, 50 on the DOMwill varyasa functionof the metal-toThestrengthof the bindingsitesavailable bindingsites,the available ligandratio. At low M/L ratios,i.e.earlyin the titrationof available highaffinityfor the metal.At highM/L ratiosthe strongbindingsites siteswill havea relatively siteswill tendto haveloweraffinityfor the available will havebeenoccupied andthe remaining metalthanthosethat wereoccupied earlyin the titration.lt followsthat the effectivebinding andaccordingly the %freemetalwill constantwill becomeweakerasthe M/L ratioincreases, that asthe pH of naturalwaters be higherundertheseconditions.lt is alsowelldocumented with metalionsfor bindingsites increases andcompetes the protonconcentration decreases, L991). ligands, the percentfreemetalion (NelsonandCampbell, on available thusincreasing higherin the the percentfreemetalwasgenerally to resultsfoundin the literature, Compared lakes,mostlikelydueto lowerpH andalkalinity QuebecandOntariolakesthan in the European lakessampled(Kaliset al.,2006; ratiosin the Canadian values,aswellas highermetal:NOM Sigget al.,2006;Van Laeret al.,2006). 4.5 Meosuredvs.modeledtrocemetol speciotion the models,suchasWHAM,areextremelyusefulin predicting Chemical equilibrium speciation containing DOM.WHAMassumes that free-metal concentration of cationsin naturalsystems DOMis bestrepresented by fulvicandhumicacidsthat bindcationsat typeA sites to phenolicgroups acidsgroupsandtype B sitesthat correspond corresponding to carboxylic (Tipping, 2OO2). Whenapplying the WHAMmodelto a naturalwatersample,the usermust, on the however,estimatethe fulvicand humicacidcontentof the water. Basedon the analysis (lakes, ponds,streamsand rivers),Bryanet al. samples averageCubindingof 15freshwater (2002)suggested that an averageof 65%of the DOMis activelyinvolvedin metalcomplexation therefore,doesnot take by fulvicacidsonly. Thisapproach, andthat it couldbe represented of variousorigins. intoaccountvariations in DOMqualityamongsamples of Cd2*,Cu2*, In the presentwork,we startedby predictingthe free metalion concentrations Forallthe lakes of DOMactivein metalcomplexation. Ni2*andZn2*usingthe suggest ed 650/o that weresimilarto thosemeasured freeCd2*concentrations WHAMVl predicted sampled, 61 (seeFigure5). Modelpredictions werealsosimilarto measured valuesfor 7n,althoughat low totalandfreeZn concentrations WHAMslightlyover-predicted of freeZn2*. the concentration Forthe Cu,however,WHAMunder-predicted in halfof the lakes itsfreeion concentration sampled, i.e.,thosewith the lowesttotalCuconcentrations. Worsestill,WHAMsystematically over-predicted the concentration of free Ni2*in all lakesfor whichthe free Ni2*concentration waswithinthe analytically measurable window. In an attemptto accountfor the variability in the natureof the DOMamonglakesanditseffect on theirtracemetalspeciation, we testedthe sensitivity of WHAMto the user-defined proportionof DOMactive(%aFAlin metalcomplexation; an arbitraryrangeof halving(33%l anddoublingf.39%l the 65%valuewasusedto calculate the freemetalion concentrations (seeFigure5). WHAMpredicted a similaro/ofree metaltothat measured in mostlakesfor Cd andZn,but despitethe widerangeof %aFAemployed, for wasstillonlysatisfactory agreement halfof the lakesfor Cuandfor noneof lakesfor Ni. We thencalculated the optimalpercentFA (%aFAopt) activein metalcomplexation and that wasrequiredto exactlymatchthe predicted measured freemetalionconcentrations. The%aFAopt valuesfor eachmetalandeachlake werecalculated by repeatedly runningWHAMVl andadjusting untilthe the FAconcentration predicted (seeMuelleret al.(2072cl freemetalion concentrations matchedthosemeasured for moredetailson howthe %aFAopt valueswerecalculated). Increasing the or decreasing %aFAchanges the numberof bindingsitesinvolvedin metalcomplexation, but the intrinsic metalbindingaffinities at thesebindingsitesremainthe same.lt shouldbe notedthat increasing the %aFAto beyond100%simplyincreases the numberof fulvicacidbindingsites involvedin metalcomplexation to a valuehigherthanthe defaultcapacity assumed by WHAM. ForCu,the %aFAoptvaluesvariedbetween8 and90%of the total DOMconcentration.For Cdz* Ni2*,however,the %aFAopt valueswereall quitehigh,often , Zn2'andparticularly exceeding the notionallimit of LOO%: 6Ito 25O% for Cd2*;65 to 4IO%forZnz*;440to I1OO% (i.e.,the number for Ni. Theseresultssuggest that in additionto adjusting the bindingcapacity of bindingsites)of the fulvicacid,the defaultbindingaffinities Ni will for Cd,Zn andespecially needto be adjustedin WHAM. 62 In a finalstep,we compared the optimisedproportionof DOMactivein metalcomplexation to qualityof the DOM. Relationships the lake-to-lake variationin the spectroscopic betweenthe calculated %aFAopt valuesandthe relativeproportionof the ubiquitous humic-like DOM fluorescence PARAFAC component(C1)werefoundfor Cuand Ni;similarly, relationships between%aFAopt andthe allochthonous fluorescence PARAFAC component(C3)werefound for CdandZn (seeFigure6). Moreover,the trendsin the relationships between%aFAopt and the Cl./Crand C3/Crratiosfor Ni were foundto be oppositeto thosefor the samerelationships for Cd,CuandZn. Thesedistinctly for the firsttime,that the DOM differenttrendssuggest, bindingsitesactivein Cd,CuandZn complexation aredifferentfrom thosebindingNi. Admittedlythe relationships shownin Figure6 will haveto be exploredovera widerrangeof DOMsamples, but theydo supportour originalideathat the proportionof DOMactivein metal complexation couldbe estimatedon the basisof itsfluorescence signature andthen introduced into chemical equilibrium modelssuchasWHAM. Usingthe relationships foundbetweenthe calculated %aFAopt valuesandthe relativeproportionof the allochthonous fluorescence component(C3),a C3-optimised%aFAwas usedto calculate, anew,the free metalCd2*, Ni2* andZn2* concentrations in the sampledlakesusingWHAMVl. An improvement in the WHAM (Figure7). Vl prediction of the freemetalion concentration wasobserved for Ni especially Theseresultssupportour hypothesis qualityof DOM that,especially for Ni,the spectroscopic maypossibly be usedasa proxyfor the proportionof DOMactivein metalcomplexation. yieldedan improvement Althoughthisapproach in the WHAMVl prediction of the freemetal ion concentrations, it is apparentthat thereis stillroomfor improvement, especially for the lakeswith low ambientmetalconcentrations. 63 o 65% aFA 33% aFA 130%aFA a I 1e-8 I f 3 1)3'" 1e-9 + = + (\| ,/- N tt ./.. o E t e -ro o ttq) 1e-10 t' ./ o o 1e-11 o 'o1 E E o t./ ?' oO' f- '/'.'r /.t , 3 1e-12 1e-11 1e-11 1e-10 1e-9 1e-6 ,/ lr 1e-12 1e-11 1e-10 1e-9 1e.8 1) 2 + a ft-7 q|. t z .r/ o € o Q ,r/ t 1e-7 ,r/ ?/ '-1' N g N E E [ 1e-8 E o It ;O 1e-8 1e-9 1e-9 1e-8 1e-7 MeasuredM2+(M) Figure5. 1e-6 1e-9 1e-9 1e-8 1e-7 MeasuredM2+(M) Comparison of the concentrations of the free metalion [M2*]measuredusingIET(Cd,Ni andZn)andISE(Cu)andthosecalculated with WHAMVl assuming 65%aFA(largedark circles)and a rangeof 33%(smalllightcircles)to L30%aFA(smalllightsquares). The dashedlinerepresents the 1:1line/ Comparaison d'ionm6tallique entrelesconcentrations par IET(Cd,Ni et Zn)ou ISE(CU)et lesconcentrations pr6ditespar libre[M2.]mesur6es WHAMVl en supposantdesvaleursde % aFAde 65 % (cerclespleins)ou entre33 et I3Oo/o (cercles ouverts).Lalignepointill6erepr6sentel'6quivalence L : 1. 64 R'?=068 ! o I'-- 250 o E 150 < r o 100 "- 50 \ 024 o26 028 030 032 010 034 100 o20 Y I v v o80 030 040 tr tr A R60 E o40 A L o Szo .t "€ o24 05 o2a 030 032 010 034 o20 030 040 R'z=0 34 o 2 1500 o N E a 1000 o 024 026 028 030 o32 034 005 010 015 t(o a N € E o soo a a N E oo I -o i o F - nv 100 'A 024 026 028 C1/C1 Figure5. 030 032 034 010 020 030 040 C3/CT percentfulvicacidactivein complexation Relationship betweenthe WHAM-calculated (%aFAoptimised)for Cd,Cu,Ni and Zn andthe relativecontributionof the PARAFAC fluorescence L (panels a-d)and3 (panelse-h)to the overallfluorescence components spectrumof lakeDOM,Pointsof the sameshapeand fill representreplicatesamples.Solid significance linesrepresentregression equationsof statistical at the 0.05alphalevel. Reproduced from Muelleret al. (2OL2cl optimald'acidefulviqueactifdansla / Pourcentage (o/oaFA compfexation optimis6)calcul6par WHAMpourCd,Cu,Ni et Zn en fonctionde la 1 (panneaux contribution, relativei la fluorescence totalede la MOD,descomposantes apoints (panneaux Les fluorescence PARAFAC. de m6me forme et de m6me d) et 3 e-h)de en r6pliqua,et leslignescontinuesrepr6sentent les couleurrepr6sentent des6chantillons r6gressions au niveaustatistiquede alpha6gald 0.05.Reproduitde Muelleret significatives al. (2012c). 65 4.6 porameters Measuredmetalbinding Possible reasons for discrepancies betweenmeasured andWHAMmodeledfree metalion includet) the underestimation sitesper concentrations of metal-binding of the concentration unitof dissolved carbon,or 2) the underestimation sites of the strengthof thesemetal-binding for the metalsof interest.Theformerreasonwasaddressed of the abovewith the calculation valuesfor eachmetalandeachlake.ForCd,Znand Ni,however, values %aFAopt the %aFAopt exceededthe notionallimit of LOOo/o. A possiblereasonfor the underestimation of the strength of the bindingsitesmightbe that the genericbindingconstants usedin WHAMweregenerated from laboratory titrationsof isolatedhumicandfulvicacidsandoftenat highermetal:DOC ratiosthanthosenormallyencountered in naturalwaters.Underthesetitrationconditions, onlythe highlyabundantsiteswith low metal-binding affinitywouldbe probed. Theneedto generatebindingaffinities for the bindingof metalsto DOMat andcapacities environmentally relevantmetalconcentrations occurring DOMhasalsobeen to naturally expressed by Tippingand Lofts(2011)in theirreviewof the performance of WHAMwhen appliedto naturalwaters.We thereforemeasured the bindingof Ni andCuto DOMthrough titrationsfor two lakeswith spectroscopically metal DOMand low ambientdissolved differ.ent concentrations, andthencompared the resulting titrationcurvesto thosepredictedby WHAM (seeFigure8). Vl usingdefaultparameters and650/o of fulvicacidsactivein metalcomplexation In general,at ambientmetalconcentrations LakeGenevawasfoundto bindmoreNi per unit DOCthan LakeGenevaDOM. In contrast,LakeOpasatica DOMwasfoundto bindmoreCuper unit DOCthan LakeGenevaDOM.Thetitrationcurveswerethen usedto calculate conditional bindingparameters for Ni andCubindingto DOMin LakesOpasatica andGeneva. 66 1e-8 (a) ^ DU' 1e-9 ,/q + E 3 &= 1e-9 (J N () E o $ te-to E o o E o Iz. te-to E E o o .N N E E te-tt E ^o te-1t o o aYt o 1e-11 e 1e-6 1e-10 (d) rc/ ,r2 /wh z p 1e-7 -g o E o = + N c N E 1 0e-7 o o) T' o = ff 1e-8 N E E o .N E 1.0e-8 o CL o 1e-11 1.0e-6 (c) + F 1e-12 1e-12 1e-9 1e-10 N o (b) o 1e-9 (t o 1e-9 1e-8 1e-7 MeasuredM2.(M) Figure7. 1.0e-9 1 0e-9 1 0e-8 '1.0e-7 MeasuredM2.(M) with of free metalion [Mz*]calculated Comparison of the mean(t SD,n=3)concentrations estimatedfrom Figure6)and those astivefulvicacid(o/oaFAl WHAMVl (withthe percentage technique(lET)(Cd,Ni andZn)or ion selective measured(t SD,n=3)usingthe ion exchange electrode(lSE)(Cu)methods:(a)cadmium;(b)copper;(c)nickeland (d)zinc.Thedashed moyennes(t ET,n=3)du m6tallibre[Mz.] linerepresents the L:Lline/ Concentrations par WHAM(enemployantle pourcentage d'acidefulviqueactif(%aFA) estim6 calcu16 (t ET, n=3) la d'6change avec technique i cellesmesur6es d'aprdsla Figure6) compar6es (lSE)(Cu): (a)cadmium;(b)cuivre;(c) s6lective Ni,Zn)ou une6lectrode d'ions(lET)(Cd, repr6sente l'6quivalence 1 : L. nickelet(d)zinc.Lalignepointill6e 67 I v -4 -3.5 Lake Geneva WHAM modeled I lv o-5 o w .fil -4.0 o o o tfo J ow I I z-o o ct o o) o w -7 -11 . a -4 -10 -9 -7 o o o olA o aA I I aoo z-o o) o o g -8 -9 -7 -6 D -4.0 A 6- A a 4.5 A et) o -5.0 -7 EF -3.5 G ..ttf o-5 -5.0 -6.0 -10 -6 LakeOpasatica WHAMmodeled o gE -5.5 -8 Ett.il ry -4.5 #B 6 4 6 A -8 -5.5 -11 -10 -7 log Ni2+ Figure8. -6 -11 -10 -9 -8 -7 -6 log Cu2+ Normalized Ni (A andC)andCu(BandD) complexation curves(M-DOM/DOC)for LakeGeneva(A and B) and LakeOpasatica (Cand D).Measuredvaluesare shown in colour;valuescalculated with WHAMVl are shownin graytriangles.Modified (M-MOD/MOD) from Muelleret al. (z}L2al/ Courbede complexation normalis6e pour Ni (panneaux A et D) et Cu(panneaux B et D) dansleslacsGeneva(carr6es; panneaux (cercles, panneauxC et D).Lesvaleursmesur6es A et B) et Opasatica sontmontr6esen couleuralorsquecellespr6ditesparWHAMVl sont par destrianglesgris.Modifi6d'aprdsMuelleret al. (2012a). repr6sent6es 68 Considering Ni bindingfirst,the affinityof the weaksites(LogKr,rui=6.98) on DOMfrom Lake is significantly higher(P=0.009) Opasatica thanfor LakeGeneva(LogKr,rui=6.30), but theseweak (Lr,r.ri=34.2 sitesare presentat a concentration nmole'mgC-1) nearlyfour timeslower(P=0.01) nmole.mgC 1).the Ni bindingaffinities for the strongsitesin than in LakeGeneva(Lr,r.ri=137 for LakesOpasatica both lakesare not significantly different(LogKz,r,ri=11.6 and L'1..7 and Geneva, respectively), althoughthereare nearlysixtimesless(P=0.025) of thesebindingsites per unit DOCin LakeOpasatica (12,x;=9.70 nmole.mgC-t),comparedto the DOMfrom Lake Geneva(Lz,rrri=3.98 nmole'mgC-1). Althoughthe strengthof the bindingof Cuto weaksiteson the DOMwasnot significantly for LakeOpasatica and LogKr,cu=7.63 differentfor both lakes(LogKr,c,=7.59 for LakeGeneva), the numberof Lrsitesper unit DOCwasapproximately threetimeshigher(P<0.001) in Lake (Lt,cu=244 Opasatica nmole.mgC 1;thanin LakeGeneva(11,6u=75.6 nmole'mgC-1).Therewere (P=0.014) significantly stronger(LogKz,cu=9.46) andapproximately twiceas many(P=0.003) nmole'mgC 1)on the DOMfrom LakeOpasatica strongbindingCusitesper unit DOC(12,6u=2t4.5 comparedto LakeGeneva(LogK2,6u=8.90 nmole'mgC-1).Theseresultsstrongly and L2,gy=27.O suggest that the DOMfrom LakesOpasatica andGenevaaredifferentin termsof their Ni and Cucomplexation affinityandcapacity. Basedon the absorbance measurements described above(section 4.3),Lake andfluorescence Opasatica DOMwasfoundto be largelycharacteristic of allochthonous sources, whereasLake GenevacontainsDOMfrom largelyautochthonous sources.lt is thustemptingto speculate that the differences in Cuand Ni bindingexhibitedby the DOMin the two lakesare relatedto the differentweightingof allochthonous andautochthonous originsin the two lakes.Themore mayhavea significantly aromatichumic-like DOMfrom allochthonous sources higher for Cuthanthe moreproteincomplexation affinity(at the strongbindingsites,K2)andcapacity likeDOMfrom autochthonous a contrasting trendwasfoundfor Ni. The sources.Interestingly, Ni bindingaffinityat the weakbindingsites(Kr)wassignificantly lowerfor the lakewith the 69 morearomatichumic-like at both DOMfrom allochthonous sources.Thebindingcapacities typesof bindingsiteswassignificantly lowerfor the lakewith the morearomatichumic-like DOMfrom allochthonous sources, againsuggesting that Ni andCuare bindingto different for typesof bindingsiteson DOM. Bakenet al. (2011)foundhigherCu,Ni,Zn andCdaffinities morearomaticDOM(basedon SUVAmeasurements), althoughthe affinityvalueswerebased on total complexedmetal,not at two differentbindingsites.Thesuggestion from the titration studythat Ni andCuarebindingto differentDOMsitesis analogous to the earliersuggestion, basedon the relationships between% aFAoptandthe CI/CrandC3/Crratios,that Ni binding sitesare distinctfrom thoseoccupiedby CdandZn. To our knowledge, theseareamongthe firstresultsto demonstrate siteswithinthe DOM the presence metal-binding of independent mosaic.Theseresultsfurtherindicatethe importance the qualityof DOMwhen of specifying predicting the complexation of tracemetalsby DOMin naturalsystems. 5 5.1 Conclusions lmprovingtrocemetol speciotionpredictions Themainobjective of the thesiswasto demonstrate of tracemetal that the calculation in naturalwaterscouldbe improvedby incorporating speciation a measure of the qualitvof quality DOMintoexisting chemical equilibrium models;we hypothesised that the spectroscopic of DOMmightbe usedto thisend. Eachof the followingsub-objectives wasaddressed during the courseof the thesis:1) a seriesof lakesrepresenting metal a widegradientof dissolved concentrations, DOMconcentrations 2)we and physico-chemical conditions wasidentified; qualityof the lakewaterDOM,whichwaslinkedto successfully determined the spectroscopic lakeand watershedcharacteristics; Ni2*and Zn2*were 3) the concentrations of free Cdz*,Cu2*, measured in the seriesof lakesandcomparedto variousenvironmental factors(e.g.,pH,total quantityandqualityof DOM);4) linksbetweenthe DOMconditional metalconcentration, bindingconstants for CuandNi andthe spectralproperties DOMwere of the lakewater exploredfor two lakesthat exhibiteddifferentDOMcharacteristics andfor whichmetal speciation alsodiffered;and5) WHAMpredictions were of the free metalionconcentrations improved(esp.for Ni)by estimating the proportionof DOMactivein metalcomplexation using 70 qualityof the DOM.Themainconclusions the spectroscopic of the thesisare described in more detailin the followingparagraphs. Basedon previous studiesthat showeddiscrepancies betweenmeasured andWHAM-predicted free metalconcentrations in lakes(Fortinet al.,2010;Unsworthet al.,2006),we setout to characterize the free metalspeciationin a seriesof lakesfrom two differentgeographical regionson the Canadian Shield.Althoughgoodagreement wasfoundbetweenmeasured and WHAMpredictedfree metalconcentrations for Cd2*and Znz*for mostlakes,departuresfrom thisagreement wereobserved for Cu2*and Ni2*.Markeddivergences werefoundfor Cu2* whenthe freeCu2*concentration wasbelow0.4 nM (i.e.,belowtotaldissolved Cu concentrations of about40 nM);similarlow freeCu2*concentrations areoftenfoundin natural waters.In the caseof Ni,WHAMconsistently the free Ni2*concentration over-predicted in all lakes. Thepresence of non-humic origin,sometimes foundin natural organicligandsof anthropogenic waters(butnot accounted for in chemical modelssuchasWHAM),hasbeen equilibrium suggested asa reasonfor discrepancies betweenmeasured andmodeledfree metal (Bakenet al.,20L1).Thisexplanation concentrations was,however,discounted for the present studyas remotelakes,littleinfluenced by anthropogenic sources, werepurposely chosen. Otherpossible reasons for discrepancies betweenmeasured andWHAM-modeled speciation valueswererecentlysummarised by LoftsandTipping(2011).Theseauthorsnotedthat the followingfactorscouldinfluence the accuracy of WHAMpredictions: 1) the precision of model inputvariables in measured andmodelparameters, 2)the uncertainty free metal properties concentrations, and3) differences betweenthe metal-complexing of the humic presentin naturalwatersandthe isolatedhumicandfulvicacidsusedto calibrate substances the model.lt wasthislatterexplanation that wasexploredin thisthesis,by characterising the DOM,in additionto the metalspeciation, in the studylakes. 7I We wereinterested in a DOMcharacterisation be applied methodthat could,in principle, naturalDOMusingsimple broadlyby non-specialists. We,therefore,choseto characterize laboratories, asopposedto the spectroscopic techniques available in typicalwaterchemistry that moreinvolvedhumicsubstances isolation Moreover,it hasbeensuggested techniques. thesehumicsubstance isolation techniques mayalterthe structureand metal-complexing properties of naturalorganicmatter(Abbt-Braun et al.,2004;De Haan,1992).First,we showed in the composition Shield that differences of DOMin a smallsetof lakeson the Precambrian (SUVAzsa), (specific weredetectable usingvariousopticalparameters UVabsorbance fluorescence indices(Fl)andexcitation-emission measurements matrix(EEM)fluorescence (PARAFAC)). interpreted with parallelfactoranalysis We thenexploredhowtheseoptical parameters predictions mightbe usedto improvemetalspeciation in naturalwatersusing chemical equilibrium models. We hypothesised that the spectroscopic measurements of DOMqualitycouldbe usedto estimatethe proportionof DOMactivein metalcomplexation. As described earlier(section 4.5),WHAMusersmustfirstestimatewhat proportionof the dissolved organiccarbonpresent in a naturalwatersample(measured andthen is constituted of humicsubstances, as mg C.L-l) estimatewhat proportionof this humicmaterialis presentas humicvs.fulvicacids.Current practiceisto assumethat humicsubstances 1985) areapproximately 50%carbon(Thurman, properties andthat "65Yo of the DOCis dueto organicmatterthat possesses the cation-binding (asmentioned of averageisolated fulvicacids"(Bryanet al.,2002).Usingtheseassumptions earlierin thissection)we foundmarkeddivergences andWHAMmodeled betweenmeasured concentrations of free Cu2*and Ni2*. Thepossibility of improving the optimal the agreement wasthenexploredby calculating percentof FAactivelyinvolvedin metalcomplexation (%aFAopt) requiredto exactlypredictthe measured freemetalionconcentration for eachmetalandfor eachlakesampled. properties Relationships betweenthis%aFAopt for a givenlakeand its DOMspectroscopic wereinvestigated. Themostsignificant relationships werefoundbetweenthe ubiquitous 72 humic-like DOMfluorescence component(C1)andthe %aFAopt for Cuand Ni,andbetweenthe humic-like fluorescence componentof allochthonous origin(C3)andthe %aFAopt for CdandZn (Figure6). Giventhe measured fluorescence of thesecomponents for a givenlake,its %aFAopt wasthen estimatedandusedto predictthe speciation of Cd,Cu,Ni andZn. Whenfluorescence measurements wereusedto estimatethe proportionof DOMactively involvedin metalcomplexation, the WHAMpredictions of free Ni2*concentration were (upto a factorof 6),comparedto the resultsobtainedwith improvedfor allthe lakesmeasured versionof WHAMandthe arbitrary65%aFAvalue.TheWHAMpredictions the off-the-shelf of the freeCu2*concentration improvedby a factorof 17f or LakeBethelandby a factorof 2 f or LakesOpasatica however,no improvementwasobserved.The andVaudray.ForCd2*andZn2*, fluorescence signatureof DOMmay,therefore,potentiallybe usedas a proxyfor its metal bindingactivityin chemical equilibrium modelssuchasWHAM. Ultimately, it wouldbe of practical useto determinethe EEMfingerprint of a givenlakewater,calculate the relative proportionof eitherthe ubiquitous fluorescence component(Cl/Cr)or the allochthonous fluorescence component(C3/Cr), andusethisvalueto estimatethe %aFAfor the givenlake waterfrom a relationship similarto thoseshownin Figure6. Althoughthisapproach appears to be promising, it wouldfirsthaveto be exploredovera widerrangeof DOMsamples and shownto holdtrue for othernaturalwaters. It wasalsoapparentthat in orderto furtherimprovethe WHAMprediction of the speciation of Cuandespecially Ni,adjustments to the metalbindingcapacity of the DOMandthe affinity constants Indeed,we foundthat differences in the usedby WHAMwouldbe necessary. i.e.the numberof bindingsites,for Cuand Ni experimentally measured DOMbindingcapacity, in the spectroscopic character couldbe explained by differences of DOM. In a smallnumberof lakes(N=2)wefoundbetween-lake and between-metal differences in the bindingaffinities of in the spectroscopic Ni andCuthat couldalsobe relatedto differences character of the DOM. Dueto the smallnumberof samples, theserelationships arespeculative onlyandshouldbe furtherinvestigated theseresultssupport for a widevarietyof surfacewaters.Nonetheless, 73 our hypothesis that the qualityof DOMis importantin determining its rolein the speciation of tracemetalin naturalwatersandthat the fluorescence signature of DOMmaybe usedasa proxyfor its metalbindingactivityin chemical modelssuchasWHAM. However, equilibrium we havecalculated the conditional bindingproperties of DOMfor Cuand Ni for onlytwo lakes. A muchlargersetof watersamples wouldbe requiredto comparethe WHAMVl Cuand Ni bindingparameters studieshave to thoseexperimentally measured here.Someprevious compared the conditional in the sampledlakewatersto bindingparameters for DOMmeasured the average intrinsicequilibrium with datasetson the constants usedby WHAMVl (calibrated bindingof metalsto isolatedhumicsubstances) for each WHAMmodelparameters by adjusting lakeuntila bestfit betweenmeasured was andpredicted freemetalion concentrations (Bryanet al.,2002;Chenget al.,2005;VanLaeret al.,2006).Following achieved thismethod, properties relationships betweenthe fittedWHAMKyavaluesandmeasured of spectroscopic DOMin naturalwaterscouldbe compared. 5.2 Researchperspectives Themainobjectiveof thisthesiswasto enhance modelsby taking equilibrium existingchemical intoaccountthe spatialvariability of the qualityof DOM. Alongtheselines,we successfully demonstrated the importance of takingintoaccountthe qualityof DOMwhenusingchemical equilibrium modelsto predictthe speciation of tracemetalsin naturalwaters.We also "signature", demonstrated that the opticalproperties may itsfluorescence of DOM,including (i.e.,the numberof bindingsitesper unitof be usedto estimateits metalbindingcapacity organiccarbon).Forexample, by measuring of differentPARAFAC the relativeproportions fluorescent components in a givensample,the % of MODactivein the complexation of Cd,Cu, Ni andZn in the samplemaybe estimated inputparameterin a prioriandusedasa measured chemical speciation models,suchasWHAM. lmprovements in the WHAMpredictions of the concentration of the free metalrangedfrom up to a factorof 6 for Ni and up to factorof 17 for Cu. Theconditional bindingconstants for the metaI-DOM complexation of certaintracemetals may,however,needto be updatedfor conditions wherelow metalconcentrations arefound (i.e.,whereIM]/DOCratiosare low). 74 Enhancing modelsleadsto improvedpredictions chemical equilibrium of tracemetalspeciation in aquaticsystems.Theaccurateprediction of the speciation of tracemetalsin aquaticsystems is essential for predicting theirbioavailability and,thus,theirtoxicityto aquaticorganisms. The BioticLigandModel(BLM)is widelyusedamongscientific, regulated andregulatory (Paquin communities to predictthe toxicityof tracemetalstowardsaquaticorganisms et al., 2OO2l.TheBLMusesWHAMto calculate in the aqueous the free metalion concentration phasenearthe bioticligandconsidered. Althoughthe BLMworksreasonably wellfor high, improvements acutelytoxicmetalconcentrations, areneededto betterpredictmetaltoxicity at low,chronictoxicitylevels(Campbell et al.,2006).Our resultsshowimprovements in WHAM (upto a factorof 5 for Ni)that candirectly speciation calculations at low metalconcentrations improveour abilityto predictchronicmetaltoxicity.Thewiderimplication of our research is that improvedprediction in naturalaquatic of tracemetalspeciation systems is neededfor toxicitytesting,environmental impactassessment, environmental monitoring, andecological (Luomaand Rainbow, riskassessment 2008). and management Althoughthe resultsandconclusions of the thesisaresignificant and novel,moreresearch questions perspective, wereraised.Firstfrom an analytical are currentanalytical techniques lmprovements ableto measure the free metalion accurately? are neededin the analytical (Tipping, measurement of manydifferentfreemetalionsat low environmental concentrations 2002),, including the measurement of free FeandAl concentrations in naturalwaters,so asto improvethe waychemical equilibrium modelsaccountfor FeandAl competition for complexation signature sites.Second, canthe DOMfluorescence be usedto estimatethe metal bindingactivityof othertypesof DOM(e.g.,non-lakeDOM)andfor metalsotherthanCd,Cu, Zn and Ni? In orderto address thisquestion, the comparison betweenmeasured and modeled free metalionconcentrations shouldbe conducted for moretypesof and DOMfluorescence DOM(e.g.,moreautochthonous andfor moremetals.Third,do the metalend-members) humicsubstances usedby WHAMsufficiently represent metaI-DOM bindingparameters bindingin naturalwaters?Shouldfutureresearch effortsmoveawayfrom usingisolatedhumic 75 naturalDOMin chemicalspeciation modelsandworkon building substances to represent intodetermining modelsusingDOMas it existsin naturalwatersamples?Again,moreresearch the bindingparameters of varioustypesof DOMwith manymetals,for widerconcentration rangesof totalandfreemetaland DOM,is needed.Finally, what isthe bestwayto incorporate DOMqualityintochemical speciation modelssuchasWHAM?Fornow,thiscanbe done "signature", but externally by estimating the metalbindingactivityfrom the DOMfluorescence a newsample-specific inputparameter for DOMqualitywouldbe mostuseful. 76 partie: Articles DeuxiEme 78 Article1 Spatialvariationin the opticalpropertiesof dissolvedorganicmatter (DOM)in lakeson the CanadianPrecambrian Shieldand linksto watershedcharacteristics KristinK.Mueller,ClaudeFortinand PeterG.C.Campbell Canada INRSEauTerreet Environnement, Universit6 du Qu6bec,490 de la Couronne, Qu6bec(Qu6bec), G1K9A9 18(1):2t-44. Publi6en 2012dansAquaticGeochemistry -y DOf: 10.L007/ stO498-OLL-9t47 79 Abstract In the presentstudywe exploredthe useof variousopticalparameters in to detectdifferences the composition of the dissolved organicmatter(DOM)in a setof lakesthat areall locatedon predictions the Canadian Precambrian were Shield,but withinwhichCuand Ni speciation previously shownto divergefrom measured valuesin somelakesbut not in others.Water (N = 18 lakes)and2008(N = 8 werecollected samples with in situdiffusionsamplers in 2OO7 lakes).Significant differences in DOMqualitywereidentifiedbetweenthe sampling regions (Rouyn-Noranda, organic Ontario)and amonglakes,basedon dissolved Qu6becandSudbury, ([DOC]), (SUVAz5a), carbonconcentrations fluorescence indices(Fl)and specificUVabsorbance excitation-emission matrix(EEM)fluorescence measurements. Parallelfactor analysis (PARAFAC) quinone of the EEMspectrarevealed four components, two of which(C3,oxidized proteinfluorescence fluorophore of allochthonous origin,andC4,tryptophan-like of autochthonous origin)showedthe greatestinter-regionalvariation. Theinter-lake differences in DOMqualitywereconsistent from with the regional watershed characteristics asdetermined satelliteimagery(e.g.,watershed-to-lake of surface surfacearearatiosand relativepercentages water,rockoutcrops,vegetativecover,and urbandevelopment).Sourceapportionmentplots, builtuponPARAFAC components ratioscalculated for our lakes,wereusedto discriminate amongDOMsources andto comparethemto sources identifiedin the literature.Theseresults haveimplications for otherareasof research, variations in the lake-to-lake suchasquantifying influence of organicmatteron the speciation of traceelementsin naturalaquatic environments. 80 1 Introduction with in the aquaticenvironment, organicmatter(DOM)is ubiquitous Naturaldissolved rangingfrom 0.5 mg C L-lin the openoceanto morethan30 mg C L-1in bog concentrations 1985).Althoughthe quantityof DOMis importantwith respectto its waters(Thurman, processes, for exampleasan its rolein biogeochemical transportandflux withinan ecosystem, (Aeschbacher et al.,2O7Ol, et al.,1990),asa mediatorof redoxreactions organicacid(Shuman (Campbellet will depend ligand(Tipping, 2OO2l asa metal-binding al.,L9971and asa surfactant With respectto the latterrole,attemptshavebeen on its nature(i.e.,sourceandcomposition). fractions of DOM,e.g.fulvicand humicacids,in madeto includevariousisolatedchemical models(DudalandG6rard equilibrium chemical ,2OO4l.Thesemodels(e.g.,the Windermere Adsorption Competitive et al.,1998)or the Non-ldeal HumicAqueousModel,WHAM(Tipping laboratory titrationdata,but whenusedto Model,NICA(Milneet al.,2003))canreproduce (Unsworth et predictmetalspeciation theyare muchlesssuccessful in naturalwatersamples in applying thesemodelsto naturalwaters al.,2006).Oneof the majorproblemsencountered of "active"fulvicor humicacidpresentin the natural is the needto estimatethe concentration organiccarbon isthe concentration of dissolved information water. Oftenthe onlyavailable (tDOCl) asto the natureof the havehadto makeassumptions andin suchcasesresearchers of the fulvicor humicacidsandthe other andthe relativecontributions DOCin theirsamples dissolved organicconstituents. between Theimpetusfor the presentstudywasthe recentobservationof markeddifferences Precambrian Shield(Fortin in lakeson the Canadian measured andpredictedmetalspeciation the WHAMand et al.,2010;Muelleret al.,2010).On the basisof the ambientwaterchemistry, ion concentrations of a modelswereusedto predictthe free-metal NICAchemicalequilibrium , u ,N i a n dZ n )i n l a k e si n t h e R o u y n - N o r a n(dQau 6 b e c ) a n d numbeo r f t r a c em e t a l s( C d C valueswasgenerally and predicted betweenmeasured Sudbury(Ontario)areas.Agreement werewellbelowthe ion concentrations goodfor CdandZn,but for Cuthe predicted free-metal the measured valuesconsistently exceeded valueswhereasfor Ni2*the predicted measured betweenmeasured free-ionconcentration.In the lattertwo cases,the degreeof divergence 81 valuesvariedfrom laketo lake.We speculated in and predicted that lake-to-lake differences the composition to the differences of the DOMmightbe oneof the majorfactorscontributing betweenobserved andpredictedmetalspeciation. polyelectrolytic Dissolved organicmatterhasbeendefinedasa complexmixtureof disordered, generated molecules, from plant,microbial, andanimalproductsat variousstagesof (Wetzel decomposition TheDOMpoolin lakesmaycontainsomesimplelabile,low ,2OO1-). molecular weightorganiccompounds, high but normallyis dominatedby recalcitrant, molecular weightfulvicand humicmaterial(Thurman, L985).With respectto the originof (production DOM,bothallochthonous from the sources environment of DOMin the terrestrial (production degradation of soiland plantorganicmatter),andautochthonous within sources the watercolumnfrom the excretion, algaeand decomposition andlysisof macrophytes, cyanobacteria in the littoraland pelagic zones)contributeto the poolof DOMin naturalaquatic (Thurman, environments 1985;Wetzel,2001).Thecomposition of DOMin a givenlakewill reflectthe relativecontribution of eachof thesesources, but it will alsodependon the processes to whichthe organicmolecules havebeensubjected on theirpathfrom the watershed to the lakeandin the lakeitself(e.g.,differential sorptionon inorganic soilhorizons (Schumacher et al.,2006);photochemical in the tributary degradation and biodegradation streamsfeedingthe lakeand in the lakewatercolumn(Kcihler et al.,2002;Yoshioka et al., 2OO7D.lt followsthat watershedcharacteristics, hydrologicregime, suchasgeomorphology, landcover,the presence of upstreamlakesand landuseare likelyto influence the quality(i.e., the natureor composition) of DOM(Jaff6et al.,2008;Mattssonet al.,2005). Thespectroscopic character of DOMhasprovedusefulin assessing itssourceandquality. Opticalproperties of DOMthat havebeenusedto distinguish betweenthe typesof DOM (Gondaret al.,2008;Weishaar (Coble,1-996; includeabsorbance et al.,2003)andfluorescence KosterandSchmuckler, l-967;McKnight et al.,2001).Absorbance andfluorescence spectroscopy areparticularly usefultoolsin the analysis of DOMqualityin that theyare relatively simple,rapid,sensitive andnon-destructive. Moststudies, however,have \ 82 with DOMfrom verycontrasting discriminated amonga largenumberof samples andto the for examplecomparing DOMfrom wetlandsto rivers,to estuaries, environments, openocean(Jaff6et al.,2008). to detect Thepurposeof the presentstudywasto explorethe useof variousopticalparameters closeto one in the qualityof the DOMin a smallsetof lakesthat aregeographically differences predictions Precambrian Shield,andwithinwhichCuand Ni speciation anotheron the Canadian in DOM values.We reporton the variability havebeenshownto divergefrom measured measurements, in lakes andfluorescence opticalquality,asdeterminedby simpleabsorbance Precambrian Shield.Parallel from the Rouyn-Noranda andSudburyregionson the Canadian (PARAFAC) that contributeto the overall factoranalysis is alsousedto identifyfluorophores to the spectra.Clusteranalysis DOMfluorescence andto estimatetheirrelativecontributions are usedto identifygroupsof lakeswith similarDOM and principal components analysis properties, catchments areexplored.A source and linksto the landusewithinthe individual trendsandto identifythe lakes approach is then usedto illustratelake-to-lake apportionment alongthe spatialgradientof the lakessampled. that constitute"end-members" Methodology 1.1 Studyorea The lakewater qualitydatain this studywerecollectedfrom two differentregionsof the in north-western and Precambrian Canadian Shield:one centredon Rouyn-Noranda Qu6bec, Ontario(Figure1). In termsof the underlying the otherlocatednearSudburyin north-central gabbroandfelsic geology, Shield,madeup of quartzite, both regionsarefoundon the Canadian (Carignan gneisses is foundin the Abitibisuband Nriagu,1985).Thecityof Rouyn-Noranda provinceof the Canadian Belt. Forthe mostpart,the the AbitibiGreenstone Shieldcontaining glaciofluvial andglaciolacustrian deposits, bedrockis coveredwith a thicklayerof glacial, et al.,2005). breakthroughthe surfacedeposits(Veillette althoughbedrockdoesoccasionally is dominatedby Shield,the locallandscape Whilethe cityof Sudburyis alsoon the Canadian 83 glacialdeposits(Barnett the presenceof exposedbedrockcoveredby thin and discontinuous and Bajc,2@2; Carignan and Nriagu,1985). :\ L- -"_)-:,'^ a ln'r i* : t -f =lY Figure1. Mapof thetwo studyareasshowing position in eastern the relative of thetwo regions (panela),thegeographical Canada lakes(panelb)andthe location ofthe Rouyn-Noranda Sudbury lakes(panelc). Bothstudyareasare richin ore depositsand havebeengreatlyimpactedby metalminingand smeltingactivities,particularlyby atmospheric depositionof acidand metals(Borgmannet al., 1998;Dixitet al.,2007).ThepH of the lakestendsto decrease with a decreasein the downwind distancefrom the metalsmelters,particularlyin the Sudburyarea(Yanand Miller,1984). Acidminedrainage from pointsources, suchasabandoned minesor mineralized may outcrops, alsodecrease the pH of lakewater locally.In eachregion,lakeswere chosento representa gradientin water quality(pH,[DOC]andtracemetalconcentrations). 84 2.2 Lakewoter sompling Lakewatersamples weretakenfrom ninelakesin eachof the studyareasduringJulyand Augustof 2OO7 siteswherechosenin the littoralzonesof small, and2008(Table1). Sampling well mixedlakesandthe sampling to be representative siteswereconsidered of the wholelake of waterquality(i.e.,pH andconcentrations of DOCand epilimnion.Thespatialvariability majorcations)is knownto be minimalwithinmanyof the sampledlakes(Fortinet al.,2010). in Fortinet al. (Fortin Lakewaterwassampledusingequilibrium diffusionsampling asdescribed jars(Nalgene) plasticlids et al.,2010).Briefly,250mL polypropylene toppedwith custom-made fittedwith a 0.2prmfiltermembrane(HTTuffryn,Pall)werefilledwith ultrapurewater(>18 to the lakebottom.Thesamplers Mohmscm)andfixedto plasticrodsthat wereanchored for 14 to 19 days. wereinstalled at a depthof 1 m from the lakesurfaceandleftto equilibrate period,the samplers werebroughtto the surface, Following thisequilibration the membrane piercedwith a cleanplasticpipettetip and surfacerinsedwith ultrapurewater,the membranes with cleanpolypropylene the customjar lidsreplaced screwlids.Thesamplejarswerebagged and storedin the darkat 4oCin fieldcoolersfor a maximumof 12 daysuntiltheywere were transported backto the laboratory.LakewaterpH,conductivity andtemperature measured in the fieldat the time of samplecollection. 85 Table1. Location,average(t SD)measuredpH,dissolvedorganiccarbon(DOC)concentration, dissolvedconcentrations of Caand Mg and yearsthe lakesweresampled(N=3-6). Lake (code) Region Location pH DOC (mg'L-') Ca (rrM) Mg (uM) Years Sampled 2007 Adeline(AD) Rouyn-Noranda 48"12',12" N 7 9 " 1 0 ' 1 7W" 7.53 4.6f 0.8 1 7 2 . 6t 0 . 8 86.7r 0.5 Y Bousquet(BO) Rouyn-Noranda 48"12'56" N 78'37'06"W 6.35 1 3 . 0r 0 . 7 1 2 2 . 2* . 0 . 4 54.4r.0.1 Y D'Alembert(DA) Rouyn-Noranda 48'23'01"N 79'00'35"W 7.48 8 . 0* 0 . 8 1 8 0 . 8r 0 . 5 6 2 . 5r 0 . 1 Y Dasserat(DS) Rouyn-Noranda 48"15'06" N 79"24',24"W 6.64 5.2t 0.7 208.9t.2.1 96.4r 3.0 Y Dufault(DU) Rouyn-Noranda 48'18'30"N 78"59'25" W 7.41 5 . 1i 0 . 4 413.2*.10.9 1 2 2 . 7i . 4 . 9 Y Dufay(DF) Rouyn-Noranda 48'01'42"N 79"27',33" W 6.60 8 . 3r 0 . 4 7 5 . 9r 0 . 5 55.6r 0.2 Joannes(JO) Rouyn-Noranda 48"10',53" N 78"40'02"W 6.82 1 0 . 0i 0 . 3 1 9 0 . 0i 1. 4 77.6r 0.3 Opasatica(OP) Rouyn-Noranda 48"04'32"N 79"17',49" W 7.47 6 . 0r 0 . 4 215.4r.3 O 1 1 4 . 2r . 0 . 9 Vaudray(VA) Rouyn-Noranda 48'05'45"N 78'40'50"W 6.59 81r0.1 8.93 5 . 2r 0 . 5 466.1r 3.3 282.6t9.4 8 3 . 6i 0 . 8 Bethel(BE) Sudbury 46'46'17"N Crowley(CR) Sudbury 46'23'06',N 80'58'55'W 6.46 2 . 6r 0 . 3 Geneva(GE) Sudbury 46"45'52',N 81'32',46" W 6.89 3 . 4r . 0 . 4 6 8 . 6r 1 . 9 Hannah(HA) Sudbury 46'26'36"N W 81"02'17" Middle(Ml) Sudbury 46'2d22" N Nelson(NE) Sudbury Raft (RA) 80"57'43"W 57.4t0.2 3 7 . 9i 1 . 1 Y Y Y Y Y 34.2i.0.1 Y 2 7 . 3r . 1 . 2 Y 7.40 3 . 6i 0 . 1 2 7 4 . 9t 1 . 0 1 5 6 . 9i 0 . 3 Y 81'01'26"W 7 .46 3 . 3r 0 . 4 246.9r 0.3 140.5t0.4 Y 46"43'36"N 81'05'39"W 5.86 1 . 4r 0 . 5 5 8 . 9r 0 . 1 28.3r01 Y Sudbury 46'24'35"N 80'56'44"W 6.91 23r0.3 8 1 . 6r 1 . 1 4 7 . 7t 1 . 6 Y Ramsey(RM) Sudbury 46'28'57"N 80"57'00"w 7.83 3.6i 0.5 395.0r 0.9 198.7t O.4 Y Whitson(WH) Sudbury 46'35'03"N 80'58'20"W 6.24 3 . 8r 0 . 1 145.1t 1.9 Y 8 0 . 5r 2 . 3 2008 Y Y 2.3 Sompleanalysis All plasticlaboratory andsampling equipmentwassoakedinLO%(w/v)nitricacidfor at least24 h, rinseda minimumof sixtimeswith ultrapurewateranddriedundera ClassL00laminarflow hood. Similarly, allglasswarewassoakedin 2 N HCIsolutionbeforebeingrinsedwith ultrapure water. jarshadarrivedin the laboratory, Oncethe polypropylene theywereopenedundera clean ClassL00laminarflow hoodandthe followingsubsamples weretaken.A 1.0-mL samplewas transferred to a polypropylene vialandacidified to 2%(v/v)with concentrated ultra-trace nitric acid(BDHAristar,VWRInternational) for analysis of majorcations(Ca,Mg)by atomicemission (VistaAXCCDSimultaneous spectroscopy ICP-AES, Varian). Next,20-mLsubsamples weretransferred to amberborosilicate vials(l-ChemBrand)for total (TOC-Vcpn, (Cary300BioUV-Visible organiccarbonanalysis aswellasabsorption Shimadzu) (CaryEclipse Spectrophotometer, Varian)andfluorescence Fluorescence Spectrophotometer, Varian)measurements. All opticalparameters weremeasured at the naturalpH of the sample andwithoutdilution.Absorption spectrawererecordedusing1-cmpolymethacrylate cuvettes. Thespectrawere measuredfrom 24Oto 700 nm at intervalsof L nm with ultrapurewateras a reference.Excitation-emission matrix(EEM)fluorescence spectraweremeasured using1-cm quartzcuvettesat excitation wavelengths from 2L0to 400 nm at increments of 5 nm andat emission wavelengths from 300to 580nm at increments of 2 nm. Theprotocolfor data collection andtreatmentrecommended in Stedmonand Bro(Stedmon and Bro,2008)was (usingcorrection followed,including filesprovidedby the instrument spectralcorrection (McKnight manufacturer), innerfiltercorrection et al.,200I;Mobedet al.,1996),calibration to Ramanunits(Lawaetz 2009)andfinallythe subtraction andStedmon, of ultrapurewaterblank and removalof spectralscatter. (SUVAzsa, Absorbance measurements wereusedto calculate the specificUVabsorbance L ffi'1 mg C t) by dividingthe absorbance at 254nm in inversemeters(m-t)by the concentration of 87 DOCmeasured in milligrams of carbonper litre(mgC L-')(Weishaar et al.,2003).The fluorescence indexwascalculated emission intensityat asthe ratiobetweenthe fluorescence 470 nm overthat at 520nm for an excitation et al., at 370nm (Coryet al.,2010;McKnight (PARAFAC), 2001).Finally, the multivariable statistical was tool,parallelfactoranalysis employedusingthe procedure in a outlinedby Stedmonand Bro(2008)as implemented MATLABenvironment,usingthe "N-waytoolboxfor MATLAB"(Andersson and Bro,2000). Excitation-emission wavelengths from 240 spectrafor 87 individual lakesamples with excitation to 400nm andemission wavelengths from 350to 550nm wereincludedin the PARAFAC model.Themeasured valuesat excitation wavelengths from 210to 235nm andemission wavelengths from 300to 348nm werenot includedin the modelso asto reducethe influence of datawith a low signal-to-noise with ratioandto constrain the modelto the wavelengths appreciable fluorescence. Thewiderangeof DOCconcentrations resulted in the lakesamples (in Ramanunits,RU).To avoidthe in a correspondingly widerangeof fluorescence intensities identification of the DOMfluorescence intensityextremesasoutliersby PARAFAC andto compareallof the samples on the samescale,the normalized fluorescence intensityforagiven samplewascalculated by dividingthe measured fluorescence intensityat eachpairof excitation andemission wavelengths intensityof the entire by the average fluorescence spectrumfor eachlake.Following the identification numberof PARAFAC of the appropriate fluorescence components, the maximumfluorescence fluorescent intensities of the identified components for eachlakewerethen restoredby multiplication by the average spectrum fl uorescence intensity. 2.4 Lokewatershedcharacteristics Lakewatershed delimitation wasachieved data usingNationalHydroNetwork(NHN)geospatial portalGeobase from the geospatial andthe NationalTopographic DataBase(NTDB)generated (NRCan). geospatial by NaturalResources Canada Usinghydrographic andtopographic data, the waterdividelinewasidentified wascalculated. andthe surfaceareaof eachwatershed Watershedlandcover(%surfacewater,% rockoutcrops,% vegetationand%wetland)and landuse(%agriculture ando/ourbandevelopment) for each werealsocalculated aspercentages 88 Thelandcover landcoverdataprovidedby Geobase. usinggeospatial of the lakewatersheds Landsatortho-images.Dataintegration dataconsistedof vectorizedrasterdatafrom classified Research Systems weredoneusingArcGlS10 (Environmental andsurfaceareacalculation Institute,Inc.,Redlands, CA). 2.5 Stotisticolanalysis level at the alphaequals0.05confidence werecalculated Thefollowingstatistical analyses lL). Parametric correlation analyses, t-tests 1L.0(Systat SoftwareInc.,Chicago, usingSigmaPlot procedures usingthe Holm-Sidak with pairwisemultiplecomparison andone-wayANOVA's where of variances amongdatasets methodwereusedwhenthe normalityand homogeneity Rank testswereemployed(Spearman non-parametric otherwise, the appropriate confirmed; Mann-Whitney RankSumtest in placeof correlation, in placeof the Pearson Ordercorrelation one-wayANOVAon Rankswith pairwisemultiplecomparison the t-tesqKruskal-Wallis procedures usingDunn'smethod,insteadof an ANOVA).Percentlandcoverand usagedata,as werearcsinesquare components, intensities of the PARAFAC wellasthe relativefluorescence tests. root transformedbeforeperformingthesestatistical (pH,the of waterchemistry basedon the similarity A hierarchical clusteranalysis (%surfacewater,% rockoutcrops,% of Caand Mg),watershedcharacteristics concentrations ([DOC], Flandthe SUVA25a, and DOMproperties vegetation cover,and% urbandevelopment) Ct, C2,C3andC4)wasconductedusingthe components relativeproportions of the PARAFAC priorto analysis. distance.All datawerestandardized Wardlinkagemethodwith Euclidean lL).A principal SoftwareInc.,Chicago, wereperformedusingSystat12 (Systat Clusteranalyses (PCA) represent the distribution of the to graphically wasalsoconducted componentanalysis gradients, inputdata.The usingthe above-mentioned studiedlakesalongthe environmental International, Wageningen, NL; version4.5(PlantResearch PCAwasperformedusingCANOCO in the (2002)).Theresultsof thesemultivariate analyses are presented ter BraackandSmilauer S u p p l e m e n t aM r ya t e r i a(lS M ) . 89 3 Results Thenaturalvariability in dissolved organicmatter(DOM)wasassessed for the sampledlakes andthe resultsare presented in the followingfour sections: 1) the generalwaterqualityof the lakessampled, 2) the lakewatershed characteristics, by 3) the quantityof DOMas measured the concentration of DOC,4) the qualityof DOMasevaluated on the basisof its absorbance andfluorescence, and5) the statistical analysis of theselatterdatato identifyDOM components. 3.1 Generallake water quality In general, the waterqualityof the lakessampled from the two regionsis quitesimilar(Table 1). Lakes from the Rouyn-Noranda regionarecircumneutral, with a meanpH of 6.79anda narrowrangebetweenpH 6.35and7.53.ThemeanpH for the Sudburylakes(6.53)is similarto that of the Rouyn-Noranda lakes,but theyspana widerpH range(5.86to 8.93).Thelakesin bothsampling regionshavesimilarly low concentrations of calcium(Ca:57 to 41.3pM) and (MC:27to 199pM),contributing magnesium to generally softwater. LakesRamsey and Bethel, from the Sudburyregion,hadthe highestpH andthe highestCaand Mg concentrations of all the lakes. 3.2 Lokewatershedchoracteristics Thewatershed characteristics of the lakessampledweresignificantly differentbetween sampling regions(Table2). Thewatershed-to-lake surfacearearatios(W/L)aregenerally higherin the Rouyn-Noranda region(mean= 25)thanin the Sudburyregion(mean= 9). Moreover,lakewatersheds in the Rouyn-Noranda regionhavea significantly highermean percentvegetationcover(83%landwetlandcover(L%),whereaspercentsurfacewater (L3%l androckoutcrops(O%larelowerthanin the Sudburyregion(4L%,O%,23%and28%, respectively). Forestcoverdominated the vegetative landcoverfor the watersheds near (69-99o/olandSudbury(96- IOO%), Rouyn-Noranda from shrubsbeingmuch the contribution lessimportant.Themorenortherlywatersheds nearRouyn-Noranda alsocontained a higher 90 proportionof coniferous with the moresoutherly Sudburyregion forests,compared watersheds. 3.3 DOMquontity differedsignificantly betweenthe two study Thequantityof DOM,i.e.the DOCconcentration, (Figure2a). ThemeanDOCconcentration lakeswashigher of Rouyn-Noranda areas(P<0.001) (7.2mg L-1)andspanneda widerrange(3.8to 13.4mg L-1) thanin the lakesfrom the Sudbury 1; region(meanof 3.4mg L rangefrom0.8to 6.0mg L-1).Withinthe entiredataset,Lakes regionhadthe highestIDOC], whereasLake Bousquet andJoannds from the Rouyn-Noranda (Figure 2a). Significant Nelson, from the Sudburyregion,hadthe lowestconcentration (P<0.001) in [DOC]werefoundamongmostlakes.In anygivenyear,the variability differences (N=3)DOCvaluesin a givenlakewaslow (i.e.,coefficient of variation of replicate , CV< L7yol, whichhada CVof 33%.Forthoselakesthat weresampled exceptin LakeNelson(Sudbury), variability in the [DOC]waslessthanI4%. duringmorethanoneyear,the inter-annual 3.4 DOM quolity (SUVAzsq), the specificUVabsorbance To probethe qualityof the DOM,we calculated valuesvariedsignificantly between measured at a wavelength of 254 nm. TheSUVAzs4 (Figure lakes vdlu€for the Rouyn-Noranda 2b). ThemeanSUVA25a samplingregions(P<0.00L) 1 but washigher(8.3L m mg C-1) thanfor the lakesfrom the Sudburyregion(5.3L m-1mg C-1), thanthat for the Sudburylakes(3.2to 10.3 L m-tmg C-1) spanneda narrowerrange(5.8to 11-.5 (P<O.OOl)were foundbetween differences L m-l mg C-1).Amongall lakesamples, significant (P<0.05); hadthe highestSUVAzsa LakeBousquet LakesBetheland Ramsey and LakeBousquet fromthe Sudburyregionandwhichhada moderate value,whereasLakesBetheland Ramsey, of replicate values.In anygivenyear,the variability [DOC],hadsomeof the lowestSUVAzsa (N=3)SUVA25a similarto that of the [DOC](i.e.,CV< t8%1, v?lu€sin a givenlakewasgenerally Forthoselakes respectively). exceptin two Sudburylakes(Nelsonand RafUCVs45%and57%o, in the SUVAzs4 values variability that weresampledduringmorethanoneyear,the inter-annual was lessthan L0%,exceptfor LakeRaftfor whichthe CVwas39%. 91 Tabfe 2. Lake (code) Adeline(AD) Watershed-to-lake surfacearearatio (W/L),percentwatershedlandcover(%surfacewater,% rockoutcrops,% vegetationand% wetland)and landuse(%agriculture ando/o urbandevelopment)for the lakessampled. Region Rouyn-Noranda w/L 7 Land Goverage Land Usage % surface o/o foCk water 16 outcrops 0 82 development 0 % vegetation % wetland % agricurture 7ourban Bousquet(BO) Rouyn-Noranda 129 6 0 90 1 (DA) Rouyn-Noranda D'Alembert 22 o 0 81 1 Dasserat(DS) Rouyn-Noranda 19 16 0 83 1 Dufault(DU) Rouyn-Noranda 7 21 0 69 7 Dufay(DF) Rouyn-Noranda 11 13 0 86 0 Joannds(JO) Rouyn-Noranda 10 11 0 87 0 Opasatica(OP) Rouyn-Noranda 11 13 0 78 1 Vaudray(VA) Rouyn-Noranda 8 15 0 81 I Bethel(BE) Sudbury 5 25 18 29 29 Crowley(CR) Sudbury 23 19 28 52 0 Geneva(GE) Sudbury 5 25 1 74 0 Hannah(HA) Sudbury 6 18 63 11 7 Middle(Ml) Sudbury 11 17 70 6 6 Nelson(NE) Sudbury 4 34 0 66 0 Raft(RA) Sudbury 15 23 25 52 0 Ramsey(RM) Sudbury 6 25 26 24 25 Whitson(WH) Sudbury 7 19 21 50 8 16 t Sudbury Rouyn-Noranda 14 r ^12 I I iro CJ g8 o6 o o4 Itt ol X 5 8308 2 8f 6O of o 0 14 b f()t z E oo P 10 c '-8 u E J v6 otI E t ^nn.) Sloo o' oo s .E IEI83^gE ro Nt U u)z 0 1.6 a a 1.5 a tr 1.4 1.3 1.2 as 8g'3:8 o a o c "o a $'a:u\2"$e.s'"$.9-*""" *q$b"*&\ffi Figure2. (panela), specificUV Replicate measures of dissolved organiccarbon(DOC)concentration (SUVA254) (panel (Fl) (panel index b) andfluorescence c) for samplestakenfrom absorbance lakesin 2007(opencircles;N=18)and 2008(closedcircles;N=8).Lakesto the left of the region,whereasthoseto the rightarefrom the dashedlinearefrom the Rouyn-Noranda Sudburyregion. 93 Thequalityof DOMwasalsoevaluated matrix(EEM)fluorescence by excitation-emission Thefluorescence spectroscopy. index(Fl)iatio (inherently wasused normalized to the tDOCI) (Fl>1.4) to distinguish betweenorganicmatterfrom autochthonous sources andallochthonous (Coryet al.,2OIO; sources(Fl<1.4) McKnight between et al.,200L).TheFldifferedsignificantly (Figure2c);the meanFlfor the Rouyn-Noranda samplingregions(P<0.05) lakes(1.32)was lowerthanthat for the Sudburylakes(1.36),andspanneda narrowerrange(L.26to 1.38)than did the Flfor the Sudburylakes(1.20to 1.54).Amongall lakesamples, significant differences (P<0.001) werefoundbetweenLakesBethelandVaudrayonly(P<0.05). Thewithin-year (N=3)Flvaluesin a givenlakewasminimal(CV<7%1,aswastemporal variability of replicate (CV< 6%). variability Furtherevidence for differences betweenthe two sampling areaswasfoundwhenthe bivariate relationships betweenthe DOMquantity([DOC]), andthe qualityparameters SUVAzsa and Fl (Figure3). A positiverelationship wereexamined is apparentbetweenSUVAzsa and [DOC]for all lakes(Figure 3a). No relationship betweenthe Fland [DOC]is apparentfor the RouynNorandalakes,whereasa positiverelationship betweenFland [DOC]is apparentfor the Sudburylakes(Figure 3b). Notethat sincebothSUVAz5a with respectto dnd Flare normalized the [DOC],it wouldbe statistically inappropriate coefficient or to calculate a correlation (r = regression betweenSUVAzsa or Fland[DOC].A weak,but significant negative correlation 0.38,P = 0.001)wasfoundbetweenthe two DOMqualityparameters, (Figure FlandSUVAzsq 4). Therelationship is anchoredby LakesBetheland Ramsey valuesand at highFland low SUVAzs4 by LakesBousquet, Joannds andVaudraywith low Fland highSUVA254 values. In summary, theserelatively ([DOC], routinemeasurements 5UVA254 and Fl)allowedusto identifyLakesBousquet, Joannds andVaudrayfrom the Rouyn-Noranda region,and Lakes Betheland Ramsey from the Sudburyregion,as lakesthat consistently emergefrom the general trendsandarethereforeof particular interest.Furtherevidence for inter-lake differences in DOMqualitywasobtainedusingPARAFAC to characterize the DOMtakenfromthe lakes sampled. 94 12 1.55 a 1.50 510 o o) E I 1.45 8 1.40 E J tr 1.35 s6 (\ 1.30 't, *F 1.25 Jo J4 a Bo 1.20 12 46810 14 DOC(mgC L-1) Figure3. 468101214 DoC (mgc L-1) (panela) and Fl (panelb) with DOCconcentration Relationships betweenSUVA254 for (squares) samplestakenfrom lakesin the Rouyn-Noranda and Sudbury(circles) areas.Error (N=3-7)coveringall sampling barsindicatethe standarddeviationof replicatemeasures years. 1.55 1.50 145 140 LL 1.35 1.30 1.25 1.20 12 SUVA254(L m-1 mg 6-1; Figure4. Relationship for samplestakenfrom lakesin the Rouyn-Noranda betweenFl and SUVA25a (squares)and Errorbarsindicate Sudbury(circles)areas. the standard deviation of replicate (N=3-7) years. measures covering all sampling 95 (PARAFAC) Fouruniquefluorescent components wereidentifiedusingparallelfactoranalysis year(N=87)(Figure5). Component 1 (C1) on replicate EEMspectrafor eachlakeandsampling peakat 250nm anda secondary peakat 340nm. Thiscomponentalso hasa primaryexcitation peakfrom 480to 500nm. Component hasa broademission 2 (C2lhasa similarfluorescence peakat 310nm. peakat just below240nm anda secondary signature with a primaryexcitation Itsemission is,however,blueshiftedto a shorterwavelength 3 of 410to 430nm. Component (C3)hasan excitation peakbelow240nm anda verybroademission peakfrom 420to 450 nm. peakat peakat lessthan240nm anda secondary 4 (C4lhasa primaryexcitation Component peakoccursat 355nm.Someresidualfluorescence 285nm. ltsemission in the upperleft fluorescence cornerof the spectrumwasunaccounted for by the PARAFAC model.Thisresidual model is mostlikelya resultof unresolved secondary Rayleigh scatter.Overall, the PARAFAC explained 998%of the variancein the EEMspectra. In severalof our lakesthe absorbance valuesexceeded 0.3at 254 nm (3 of 18 lakesin 2007and (Milleret L of 8 lakesin 2008),andthusthe innerfiltercorrection mayhavebeenincomplete results, we al.,20L0).To checkfor the possible influence of theselakeson the PARAFAC removedthemfrom the datasetandreranthe PARAFAC the samefour components analysis; wereidentifiedin the smallerdatasetandthe % explained remainedabove99%.We variance alsoperformeda split-half validation exercise andthe splitdatagaveverysimilarresults(see sM). Forcomparison was amonglakesamples, the maximumfluorescence of eachcomponent(F,n.,) normalized by dividingby the totalfluorescence the lakes of allfour components.In general, proportionof C3(40t from the Rouyn-Noranda regionhavea significantly higher(P<0.001) L6%)thando the Sudburylakes(17t 5%)(Figure 6a),whereasthe Sudburylakeshavea proportionof C (2014%)thanthe Rouyn-Noranda higher(P<0.00L) significantly lakes(10t 4%)(Figure 6b);in particular, LakesBethel,Nelson,Raftand Ramsey hadthe highest from thislatterfluorescent contribution component. 96 E g o, c o E tu!i; G 550- 3 .9 U' .g TU (nm) Excitation Wavelength Figure5. Parallelfactor components1 (panela), 2 (panelb), 3 (panelc) and 4 analysis(PARAFAC) (paneld) identifiedfrom the excitation-emission matrix(EEM)fluorescence spectraof DOM fluorescence from all lakesoverall samplingyears.Contourlinesrepresentincreasing intensityin arbitraryunits. 97 1.0 o E + 0.8 co. o E 0.6 E x (U E 0.4 t.|c o c g. o.2 E o o 0.0 1.0 "**i""*.j.1*O"lasee{ato,}r"'}\\outulrunnnlou+t'\uuduJ c o t A 0.8 co. I E 0.6 E x (5 IL E o.+ E o c g 0.2 E ()o 0.0 Figure6. (F..,) of PARAFAC Proportionof maximumfluorescence components1 (diagonallines),2 (cross-hatch), 3 (horizontallines)and4 (verticallines)relativeto the total component (panela) and Sudbury(panelb) regions fluorescence for lakesin the Rouyn-Noranda sampledin2OO7. 98 Table3. Component number (this study) peakin parentheses) Literaturebaseddescriptionand positionof the fluorescence maxima(secondary and descriptionof the four fluorescentcomponentsidentifiedby the PARAFAC model. Excitation Emission max (nm) max (nml Component Gomponent Component assignmentby assignmentby assignmentby Stedmonand Cory and Goble(1996) Markager McKnight (2005) (200s) Description Ubiquitous origin. c1 250(340) 480to 500 c SQ1 Humic-like fluorescence. Reducedquinonefluorophore associated with higherplantmatter. origin. Ubiquitous c2 <240(310) 410 to 430 M so2 Humic-likefluorescence. Reducedquinonefluorophoreassociated with microbialorganicmatter. Allochthonous origin. c3 <240 420 to 450 A Q2 Humic-like fluorescence. Oxidizedquinonefluorophore. Refractory. c4 <250(285) 355 to 355 Tryptophan-like Autochthonous origin. proteinfluorescence. Tryptophan-like 4 Discussion Significant differences in the quantityandqualityof DOMwerehighlighted in the preceding section, bothamonglakesandbetweenregions.Theobservation is consistent with the of suchdifferences (2007),who concluded analysis of Sobekandco-workers from theirstudyof a verylargelake (7574lakes database from sixcontinents)that regulate terrestrialvegetation, climateandtopography soilsand hydrology, andthat thesemastervariables in turn determinethe rangeof possible DOM concentrations in lakesin a givenregion.The[DOC]of eachindividual lakewithinthe regionwill be affectedby the locallakeandcatchment characteristics, suchasthe proportionof wetlandsin the catchment, the presence of upstreamlakes,andwaterretentiontime. Usingthisconstruct, in the followingsections we will firstrelatedifferences in [DOC]to the characteristics of the lake's watershed,suchasthe watershed-to-lake surfacearearatio(hereafternotedasW/L; seeTable2) (SUVAzsa andwatershed landcover.We will thenapplythe routinelymeasured opticalproperties and Fl),alongwith the relativeproportionof fluorescence asidentifiedusingPARAFAC, components, to characterize the DOMin the sampledlakes,identifyits possible its upstream originsanddiscuss andin-lakeprocessing. Althoughthereis a knowneffectof pH on the absorbance over andfluorescence of humicsubstances a widerange(generally 2 to 10)of pH values(Mobedet al.,1996),littleeffectis reportedfor lake naturalorganicmattersamples withinthe pH rangereportedfor our lakes(Senesi, 1990;Weishaar et al.,2003).In fact,Weishaar et al. (2003)suggest that withinthe pH rangeof 2 to 8.6thereis no need to adjustthe pHto a constantvaluewhencomparing UVabsorbance valuesamongsamples.Baker (2OO2l alsofound no variabilityin DOMfluorescence as a resultof changesin pH for a seriesof circumneutral riversites.Therefore, allopticalproperties weremeasured at the naturalpH of the (SUVA25a, sample.We foundno correlations betweenpH andanyof the opticalmeasurements Fl, PARAFAC components; datanot shown),suggesting that pH effectson the opticalmeasurements are not significant for theselakessamples andat their naturalpH valuesreported. L00 4.1 Lakewatershedcharacteristics(regional differences) Thedatain thisstudywerecollected fromtwo geographically differentregionson the Canadian Precambrian of theseregionsplayan importantrolein determining Shield; the geologyand hydrology variability suchas pH,hardness, the inter-lake of surfacewaterqualityparameters and DOMquantity andquality.Indeed,the observed difference in meanlake[DOC]betweenthe two sampling regions (Figure2a: [DOC]p is consistentwith the differentwatershedcharacteristics in the r,r> [DOC]sudourv) regionhavehigherW/L two regions.Thewatersheds of the lakessampledfrom the Rouyn-Noranda ratios,higherpercentvegetation coverandlowerpercentrockoutcrops, compared to the watersheds of the lakessampledfromthe Sudburyregion(Table2). lt hasalsobeenreportedthat the Rouyn-Noranda regionhasgreatersoilcoverthanthe Sudburyregion(Veillette et al.,2005), whichhasverylittlesoilcoverandwherebedrockoutcroppings are muchmorefrequent(Barnett and Bajc,2002l..Also,because of acidification anderosionof the soilin the Sudburyareadueto past (Barnettand miningand smeltingactivities, lossof soilA- and B-horizons therehasbeena significant Bajc,2OO2). we conclude that the Rouyn-Noranda lakesreceivea larger Basedon thisinformation, proportionof DOMfrom theircatchments in the Sudburyregion,whereas thando theircounterparts producea higherproportionof DOMwithinthe watercolumn. the Sudburylakespresumably 4.2 DOM quolity 4.2.1 SpecificUVAbsorbonce(SUVA25a) lndex(Fl) ond Fluorescence Opticalproperties haveoftenbeenusedin the pastto characterize DOMand identifyits source(s) (Coble,1996;KosterandSchmuckler, 1967;McKnight et al.,2001).Althoughthe absorption spectra of naturalDOMare relatively featureless andsimilaramongnaturalDOMenvironments, the (SUVA254) calculation of the specificUVabsorption canbe usedasa tool to estimateandcomparethe qualityof DOMfrom differentsources (Weishaar et al.,2003).We successfully discriminated values(Figure2b)andon the trendsin betweenthe two sampling regionsbasedon the SUVA254 valuesfor the lakesin the Rouynwith increasing SUVAzsq [DOC](Figure3). ThehigherSUVAzs+ giventheir higherW/L ratiosand Norandaregion,comparedto the Sudburyregion,wereexpected (i.e.,forestcoverage) in the region.Thesefactorswouldbe the greateramountof vegetation DOM,knownto havehigheraromaticity expected and to favourgreaterinputsof allochthonous 101 absorptivity thanautochthonous DOM(Chinet al.,1994;McKnight et al., et al.,2001;Weishaar 2003).Temperateconiferousforestsof the type foundin the Rouyn-Noranda regionare recognized of leachable values(Sobek sources organiccarbonwith relatively higharomaticity and highSUVA254 et al.,2007).In addition,giventhe greatersoildepthin the Rouyn-Noranda region,somepartitioning of the leachable organicmatterlikelyoccursin the lowermineralsoilhorizons, with consequent removalof the morehydrophobic components.In theirstudyof DOMmovementin a largeAlaskan permafrost, catchment underlain with discontinuous Balcarczyck et al. (2009)suggested that such selective adsorption of hydrophobic components wouldleadto a furtherenrichment of hydrophilic moietiesin solution, with higharomaticity values.Notehoweverthat in theirstudy andhighSUVA254 of the originsof dissolved organicmatterin streamsof the YukonRiverbasin,O'Donnell et al.(2010) suggested the oppositetrend. Duringwinterlow flowsin theirsystem,SUVA25a decreased andthey attributedthischangeto selective removalof aromaticcompounds, suchas lignin-derived polyphenols, in the deepersoillayers. In additionto thisdirecteffecton the relativeproportions DOM andautochthonous of allochthonous enteringa lake,differences in WL ratioswouldalsobe expected to affectDOMindirectly.For example, the higherthe WL ratio,the higherthe relativeinputof allochthonous DOM,but alsothe lowerthe residence time of the lakeand,therefore, of in-lakeprocessing of the lowerthe likelihood DOM(Sobeket al.,2OO7l,. LakeBousquetwasfoundto have,by far, the highestW/L ratio(129)of all the lakessampled, andthe SUVAzs+ valuefor the DOMin this lakewasthe highestvaluethat we recorded.lt shouldalsobe notedthat Lakes and Joannds andVaudraybothflow into LakeBousquet therefore, contributeto its DOMload. Microbialprocessing of the DOMin the upstreamlakes,with consequent removalofthe morelabilefractions, wouldbe expected to leadto an increased proportionof recalcitrant DOMin LakeBousquet.On the otherhand,lakesNelsonand Bethelhad the lowestW/L ratios(4 and5, respectively). Despitethesesimilarlow W/L ratiosfor the two lakes, the SUVAzsa valuefor LakeBethelwaslowerthanthat for LakeNelson.LakeNelsonis a clear, oligotrophic lakewith a low [DOC], whereasLakeBethelis a coloured, eutrophiclakewith a moderate (Table isfoundin a watershed that is highlydeveloped [DOC].LakeBethel,aswellas LakeRamsey, 2). Furthermore, LakeBethel,a smalllakethat flowsinto LakeRamsey, wasusedasa sewagelagoon LOz until1986(Pearson et al.,2002),whichcontributed to itseutrophication. BothLakesBetheland Ramsey are,therefore,mostlikelyproducing a higherproportionof autochthonous DOMthan Lake Nelson, thusleadingto theirlowerspecific absorptivities. SimilarspecificUVabsorbance valueshavebeenreportedfor otherlakesystems.Forexample,Miller et al. (20L0)foundlow SUVA254 valuesfor alpinelakessampledin Colorado duringthe summer rangingbetween0.5and2.5L m-1mg-t,indicative of autochthonous DOMproducedby in-lake processes. biogeochemical higherSUVAzsc Theyalsoreportsignificantly valuesfor an adjacent t), subalpine streamduringthe summermonths(2.5to 4.0 L m-l mg attributedto the inputsof allochthonous DOM. We werealsoableto discriminate regionsbasedon Flvalues amonglakesfromthe two sampling (Figure2c). Fluorescence spectraof naturalDOMare morecomplexthanabsorption vary spectra, amongenvironments indicative andrevealstructural characteristics of DOMsourcematerial(Fellman et al.,2010).TheFlvaluesmeasured for the lakesvariedfrom 1.3to L.5,enoughto indicate significant differences in the sourceof DOM(Balcarczyk et al.,2009).Thelakesfrom the RouynNorandaregionhavea low meanFl,indicative DOM,whereasthe lakesfromthe of allochthonous Sudburyregionhavea slightlyhighermeanFl,indicative of a greatercontribution by autochthonous DOM. Ourcalculated Flvaluesaresimilarto thosereportedfor otherfreshwater environments in centralandsouthernOntario(Keltonet al.,2007). Thedifferences betweenthe two samplingregionsarealsoevidentwhenwe comparethe quantityof DOM([DOC]) to the DOMqualityparameters SUVAzsa and Fl (Figure 3). Forthe Sudburylakes, SUVAzsq and Fl increased with a highersteady-state of the Sudbury [DOC]in the lake.Thewatersheds areahavelow WL ratios,low percentvegetation cover,highpercentrockoutcropsanda thin soil layer,andthereforeanyincrease in the DOCconcentration in the lakesis likelyattributable to an increased contribution of autochthonous DOMin theselakes.Forthe Rouyn-Noranda lakes,the increases SUVAzsq with an increase in [DOC], whereasthe Fl appears to decrease. A shiftto higher DOM)at higher[DOC]suggests SUVA25a valuesand lowerFlvalues(indicative of allochthonous an 103 is increase in the proportionof allochthonous DOMderivedfromthe catchment.Thisreasoning consistentwith the highW/L ratios,higherpercentvegetationcover,lowerpercentrockoutcrops areaascomparedto the andthickersoillayerin the catchments of the lakesin the Rouyn-Noranda Sudburyarea. (Fl) parameters Otherstudieshaveshownthat the relationship andfluorescence betweenSUVAzs+ canbe usedto identifyautochthonous end members(Jaff6et al.,2008;Millerand andallochthonous whichallowedusto McKnight, 2010).We thereforecompared thesetwo DOMqualityparameters, isolateend membersin termsof allochthonous sources of DOMwithinour andautochthonous dataset(Figure 4). Again,LakesBetheland Ramsey, foundin the upperleft handcornerof Figure4, DOM.ln contrast, are isolated.TheirhighFland low SUVA25a of autochthonous valuesare indicative Lakes Vaudray, Joannds and Bousquet arefoundin the lowerrighthandcornerwith low Fland high valuesindicative SUVAzsq of allochthonous DOM.Theseresultsaresupportedby the resultsof the clusterand PCAanalyses whichfoundLakesBousquet, andVaudray(alongwith Lakes Joannds (see D'Alembert and Dufay)tobe groupedtogetherbasedon theirDOMopticalproperties Material,SM). Supplementary 4.2.2 PARAFAC Althoughsimpleabsorption andfluorescence DOMqualityparameters havefrequentlybeenusedin the pastto characterize naturalDOM(Gondar et al.,2008;Greenand Blough,1994),theyare not alwaysthe besttoolsto employ.Forexample, aresubjectto absorbance measurements interferences from particles and inorganic et al.,2003), species suchas ironand nitrate(Weishaar (Greenand theseinterferences beingmoreimportantat low [DOC]thanin high[DOC]situations Blough,1994;Millerand McKnight, 20L0).Furthermore, whetherSUVAzsa or Fl is usedto describe the qualityof DOM,both parameters arecalculated usingonlya smallportionof the available informationin eitherthe absorbance therehas or fluorescence spectrum.In recentyears,however, (PARAFAC) beena significant increase in the useof parallelfactoranalysis asa statistical tool to interpretexcitation emission matrix(EEM)fluorescence fluorophores or spectraand identifyspecific groupsof fluorophores (Stedmon (Jaff6et al.,2008;Miller et al.,2003)tocharacterize DOMsources 104 described above,thisapproach 2010;Yamashita et al.,201.0).Unlikethe techniques and McKnight, spectrum. incorporates datafrom the entireDOMfluorescence are similarto thosefound identifiedin our lakesusingPARAFAC TheDOMfluorescence components (Coble,1996;Coryand McKnight, 2005;Koster environments in the literaturefor similarfreshwater we ascribeCLandC2to reducedquinone, L967).Basedon theseauthors'findings, andSchmuckler, C3to refractory, humic-like fluorophores that are presentin a widevarietyof aquaticenvironments, protein quinone,humic-like originandC4to tryptophan-like fluorophores of allochthonous oxidized fluorescence of autochthonous origin(Table3). fluorescence component(C3)in the Rouyn-Noranda Therelativeproportionof the allochthonous whereasthe higherthan in the Sudburylakes(17! 5 o/ol, lakes(40t L6%;Figure6a)wassignificantly lower(10t 4 compared from the proteinfluorescence component(C ) wassignificantly contribution with the anticipated trendsareconsistent to 20 t 4 %)(P< 0.001in bothcases).Theseinter-regional lowerinputsof DOMfromthe catchments of the Sudburylakes,giventheir lowerWL ratios,lower percentvegetation cover,higherpercentrockoutcropsandthinnersoilcover.Whenallthe lakes (P<0.001) in statistically significant differences of variance, werepooledandsubjected to an analysis and Bethel,and betweenLakes the relativeproportions of C3werefoundbetweenLakesBousquet in the relativeproportions of C4werefound differences and Bethel.Similarly, significant Joannds Nelsonand Bethel.LakeBethelhasthe highest betweenLakeBousquet and LakesRamsey, proportionof the protein-like with its component, C4,consistent fluorescent autochthonous increase in primaryproduction.The earlier,andthe consequent eutrophicnature,asdescribed as lakeswith supportthe groupingof LakesBetheland Ramsey resultsof the clusterandPCAanalyses from LakesBousquet and origin,andtheirseparation highproportions of DOMof autochthonous DOM(seeSM). of allochthonous characterized by highcontributions Joannds, 105 5 0.8 E o o. o o- I o.o (U E x G' E rr- 0.4 c o c o o- E o ^^ o u.z Vaudray Figure7. Joannes Bousquet (F.",)of PARAFAC Proportionof maximumfluorescence components1 (diagonallines),2 (crosshatch),3 (horizontallines)and4 (verticallines)relativeto the total componentfluorescence, and (opencircles)for three hydrologically DOCconcentration connectedlakesin the Rouyn-Noranda areasampledin 2007.Errorbarsindicatethe standarddeviationof three replicateDOCmeasures. Vaudray, Lakes Joannds and Bousquet areall locatedin closeproximityandareconnected hydrologically; theyarecharacterized component by relatively highproportions of the allochthonous C3andlow proportions of ubiquitous componentC2andthe proteincomponentC4. Figure7 presents for the trendsin the relativeproportions aswellas [DOC], components, of the fluorescence thesethreelakes.Asthe [DOC]increases the relative from LakeVaudrayto LakeBousquet, proportionof the allochthonous VaudrayandJoannds both componentC3increases. Lakes independently flow into LakeBousquet.Therefore, must in [DOC]in LakeBousquet the increase represent inputsfrom its immediatewatershed.Notethat thisreasoning is alsosupportedby the factthat LakeBousquet is hasthe highestWL ratio(L29,or 83 if onlyits immediate watershed (10)andthen LakeVaudray(8). followedby LakeJoannds considered), plotsto furtherdiscriminate We usedsourceapportionment betweenallochthonous and autochthonous sources of DOMfor our studylakes.Thefourfluorescence components are represented with threedifferentcomponentratiosandvisuallypresented in two plots:C3/C1vs. L05 C2/CL(Figure8a)andC4/CLvs.C2/Ct(Figure8b). The ratioC2/CLwaschosenfor the dependent 2005),i.e.,CLandC2exhibit axison the basisof itssimilarity to the Fl (Coryand McKnight, for an excitation respectively, wavelength of emission at 470and520 nm wavelengths, appreciable Flvaluesfor the lakes 370nm (seeMethods).Indeed,plottingCZ/CLratiosagainstthe calculated (r=0.65, yieldeda strongpositivecorrelation P<0.001), supporting the analogybetweenC2/CLand Fl. betweenthe allochthonous Thefluorescence and componentratioplotsclearlydiscriminate quinone autochthonous sources of DOMto the lakesstudied.ln Figure8a,we comparethe oxidized (C1andC2).Asthe fluorophores of allochthonous origin(C3)to the reducedquinonehumic-like relativeproportionof C2to CLincreases, the relativeproportionof C3decreases.LakesBousquet, in the upperleft handcornerof the plotwherethereis Joannds andVaudrayanchorthe relationship whereasLakesBetheland Ramsey fluorophores, anchorthe a higherproportionof the allochthonous of the allochthonous fluorophore.ln Figure8b, lowerrighthandcornerwherethereis lessinfluence proteinfluorescence (C4)to the reduced we comparethe relativeproportionof the tryptophan-like quinonehumic-like (CLandC2):asthe relativeproportionof C2to CLincreases, fluorophores the relativeproportionof C4increases. Again,lakesBousquet, Joannds andVaudrayanchorthe relationship in the lowerleft handcornerof the plotwherethereis a lowerproportionof the fluorophores, whereaslakesBetheland Ramsey anchorthe upperrighthandcorner autochthonous fluorophore. wherethereis higherproportionof the autochthonous 107 0.2 o.4 0.6 0.8 1 . 0 1 . 2 1 . 4 1 . 6 c2tc1 1.0 () 0.8 !t o 0.6 0.4 Jo-olL 0.2 VA-08 0.0 0 .2 0 .4 0 . 6 0 . 8 1.0 1.2 1.4 1.6 C2IC1 Figure8. (panel (panela) andautochthonous Variation in fluorescence component ratiosfor allochthonous b)DOMfrom lakesin the Rouyn-Noranda andSudburyregions sampledin2007(-7)and2008(-8). Reference valuesarefrom Stedmonet al. 2003(X=forest stream;upwardtriangle=agricultural stream;downwardtriangle=lake; square=marine). star=wetland; circle=urban drainage; 108 of DOM,we haveaddedthe ratiosof known the DOMin our lakeswith sources In orderto associate valuesfor signature from Stedmonet al. (2003)toFigure8, as reference calculated components quinoneof the oxidized differentDOMsources.With respectto C3(Figure8a),whichrepresents andVaudrayarefoundin the sameregionas DOM Joannds origin,LakesBousquet, allochthonous arefoundnearthe regionfor DOMfrom urban from foreststreams.LakesBetheland Ramsey similarto thosefor DOM drainage.Thelakesfoundbetweentheseend membersexhibitsignatures tryptophanlands.ForC4(Figure 8b),representing from lakesandfrom streamsdrainingagricultural arefoundnearthe origin,LakesBetheland Ramsey of autochthonous likeproteinfluorescence VaudrayandJoannds areonceagainfoundnear DOM. LakesBousquet, for urbandrainage signature for foreststreamDOM.Thelakeslyingbetweentheseend membersarefoundnear the signature streams. for DOMfrom lakes,wetlandsandagricultural the signatures with the resultsof the Therelationships andend membersidentifiedin theseplotsareconsistent in Figure3. To our knowledge, thisis the betweenour SUVAzs+ and Flvaluespresented relationship derivedDOMhasbeen andautochthonous betweenallochthonous firsttime that the discrimination plotsbuiltuponPARAFAC components. usingsourceapportionment demonstrated 5 Conclusions (SUVAzs+; fluorescence In thisresearch we setout to explorethe useof variousopticalparameters to detect coupledto PARAFAC analysis) spectroscopy matrixfluorescence index;excitation-emission Precambrian Shield of the DOMin a smallsetof lakeson the Canadian differences in the composition predictions values.On hadbeenshownto divergefrom measured withinwhichCuand Ni speciation revealed lakesto Sudburylakes,allthreetechniques Rouyn-Noranda a regionalscale,i.e.,comparing areas. in the composition of the lakewaterDOMbetweenthe two sampling distinctdifferences quinonefluorophore of allochthonous origin")andC4 C3("oxidized PARAFAC components proteinfluorescence origin")werethosethat showedthe ("tryptophan-like of autochthonous characteristics with the regionalwatershed greatestinter-regional a resultthat is consistent variation, the optical to the lake.Bycombining andthe presumedroutingof the DOMfrom the watershed and usingclusteranalysis parameters, landcoverinformation, waterqualitydataand percentage 109 principal components analysis, we wereableto identifygroupsof lakeswith similarDOM composition andrelatethesegroupings to the localwatershed characteristics of the lakes.The relativepercentages of vegetative(forest)cover,rockoutcropsand surfacewaterwereshownto be the keywatershedcharacteristics. Currentmetalspeciation modelsthat includedissolved organicmatter,suchasthe Windermere HumicAqueousModel(WHAM),incorporate quantityof DOM,but usersmustdefine the measured in DOMquality(i.e.,chemical the variability character andsource)by choosing the % of DOMthat is (Bryanet al.,2002).Thepresentresultsopenthe doorto the activelyinvolvedin metalcomplexation useof opticalmeasures of DOMqualityto refineuserestimates of thiskeyparameter. Acknowledgements Theauthorsacknowledge providedby M.G.Bordeleau, the technical assistance S.DuvalandJ. Perreault in the laboratory and P.Girardand P.Marcouxin the field. We alsothankM.-A.Robinfor providing GISdata,aswellasS.SmithandC.Stedmonfor helpfuldiscussions on PARAFAC modelling. l. Lavoieprovidedinvaluable assistance with the multivariate was statistical analyses. Sampling greatlyfacilitatedby the personnelof the Laurentian Freshwater EcologyUnit, UniversityCooperative led by JohnGunn,in the Sudburyareaand by L.Jourdainof the Ministdredesressources noturelleset de lofaunedu Quibecin the Rouyn-Noranda providedby R.-M.Coutureon earlier area.Comments versions of the MSweregreatlyappreciated. Financial supportwasprovidedby the NaturalSciences and Engineering Research Councilof Canada Research andby the Metalsin the HumanEnvironment Network(www.mithe-rn.ors). P.G.C. Campbell andC.Fortinaresupportedby the Canada Research ChairProgram.Theconstructive commentsof two anonymous referees aregratefully acknowledged. 110 Material Supplementary 6 PARAFACmodelvalidation 6.1 matrix(EEMlfluorescence modelbasedon the excitation-emission Thefour-component PARAFAC followingthe stepsoutlinedin Stedmonand Bro lakesamples wasvalidated spectraof 87 individual (2008).Outliertests,comparing in the excitation andemission data, the sumof squaredresiduals the resultsof the wereperformed.FigureS1presents and randominitialisation split-half analysis procedure PARAFAC model.Thevalidation of the chosenfour-component splithalfvalidation Sla,bandS1c,d)andcomparing the excitation and involvedsplittingthe dataintotwo halves(Figures models(compareFigure51-panelsa Thetwo individual loadings for the four components. emission modelwas,therefore,foundto and b with panelsc andd) are nearlyidentical.Thefour-component most(>99%lof the variancein the data. adequately describe the dataandexplained alt E) to o fr o l! 450 Em (nm1 Ex (nm) 0.6 v.z o & € ors X n, E) d o E o.r o o * 0.2 E oos 5so FigureS1. -\---- t o IJJ 400 450 Em (nm) 500 550 0r , 250 300 Ex (nm) 350 400 PARAFAC model.Panels a andc showthe emission Splithalfvalidation of the four-component for the two loadingsand panelsb and d showthe excitationloadingsof the four components 1, the dashed line is for component 2, the independent models.Thesolidlineisfor component dottedlineis for component3 and the dash-dotlineis for component4. 111 6.2 Multivariotestatisticalanolyses 6.2.1 ClusterAnalysis Clusteranalysis wasemployedto identifygroupsof lakeswith similarwaterquality,watershed characteristics andDOMopticalproperties.First,by considering waterchemistry data(pHandCa and Mg concentrations) and watershedlandcoverdata(%surfacewater,Torockoutcrops,% vegetation and%urbandevelopment S2a). )only,threemaingroupsof lakeswereidentified(Figure All of the lakesfrom the Rouyn-Noranda regionweregroupedtogether(Group3),whilethe lakes from the Sudburyregionwereseparated intoonegroupcontaining LakesWhitson,Raft,Crowley, Genevaand Nelson(Group2),andanothergroupcontaining LakesMiddle,Hannah, Ramsey and Bethel(Group1-).A secondclusteranalysis wasconducted usingthe DOMdataonly(DOC (Figure concentration, SUVA25a, Flandthe relativeproportions of the PARAFAC components S2b)). groupedthe Rouyn-Noranda Thissecondclusteranalysis D'Alembert, Dufay LakesBousquet, Joannds, andVaudraytogether(Group1'). Theremaining lakesweregroupedtogetheralong Rouyn-Noranda with the SudburyLakeWhitson(Group2'). TheSudburyLakesRaft,Geneva, Crowley, Nelsonwere re-grouped togetherwith LakesHannah, Middle,Ramsey andBethel(Group3'). In the firstanalysis, (Group1, FigureS2a) the SudburyLakesHannah, Middle,Betheland Ramsey were clearlydistinct,reflectingthe higho/ourbandevelopmentin their watersheds, their highpH valuesandtheirhighdissolved concentrations of Caand Mg. However, if the clusteranalysis was limitedto the DOMopticalproperties, theselakesweremixedintothe generalSudburylakegroup (Group3', FigureS2b).Theanalysis usingonlythe DOMdataidentified Lakes the Rouyn-Noranda Bousquet, Joannds, Vaudray, D'Alembert and Dufay(GroupL', FigureS2b)asa distinctgroup.These lakeshavehighDOCconcentrations, highSUVA254 valuesandhighproportions of the allochthonous fluorescence provides component C3. Theclustering of lakesbasedon DOMopticalproperties additional information on the natureof lakeDOM,whichmightotherwisehaveremainedhidden. 1.t2 BE-07 BE-08 RM.O7 HA-07 Ml-07 NE-07 GE-07 GE.08 cR-07 RA-08 RA.O7 WH-08 WH-07 BO-07 VA-07 VA.O8 JO-07 DF.O7 DS-o7 DS-08 AD-07 DA-07 DU.07 DU.08 oP-08 oP-07 BO-07 JO-07 DA-07 DF-07 VA.O7 VA-08 DS-07 oP-08 oP-07 DU-07 AD.O7 WH-07 DU-08 DS-08 WH-08 RA.O7 RA-08 GE.O8 GE-07 cR-07 HA-07 Mt-07 NE RM- Lakesare identifiedby a two-lettercode(see resultingfrom the clusteranalyses. Figure52. Dendrograms of the analysisperformedusingwaterchemistry(pH, TableL) and a two-digityearcode.Results andwatersheddata(%surfacewater,rockoutcropsvegetation,and Caand Mg concentrations) urbandevelopment) are shownin panela). Resultsof the analysisusingDOMproperties(DOC components C1,C2,C3 Fl andthe relativeproportionsof the PARAFAC concentration, SUVA25a, the levelat whichthe clusterswere and C4)are presentedin panelb).Thedashedlinerepresents established andthe numbersrepresentthe groupsidentified. Analysis 6.2.2 PrincipolComponents (PCA)wasemployedto graphically the lakesalong distribute analysis Principal component in site gradients (Figure 8L%of the variance S3).Thefirsttwo PCAaxesexplained environmental ThePCAordinationshowsthat the gradientof maximumvariance(Axis1) is associated distribution. the relative rangingfrom highvaluesfor DOCconcentrations, variables with a seriesof environmental proportions in quadrant| (Q-l)tohighvaluesfor the relative of C3,SUVA25a and%vegetation proportions of CL,C2andC4,aswellasfor % surfacewaterand%rockoutcropsin quadrantll (Q-ll). (Caand Mg variables Thesecondgradient(Axis2) spannedfrom highvaluesof waterchemistry and Fl in quadrantlV (Q-lV)to highvaluesof andpH)aswellas%ourbandevelopment concentrations in quadrantll (Q-ll). % vegetation 113 Thelakeswerefirstdiscriminated DOMgradientspanning alongAxis1, definedasan allochthonous diagonally from Q-l (withhighvaluesfor DOCconcentration, from C3andSUVAzs+) the contribution to Q-lfl(with highcontributions from C4,C2,C7,losurface waterand%rockoutcrops).Mostof the Rouyn-Noranda lakeswerefoundin Q-1,reflecting the importance of allochthonous sources of DOM for theselakes.ln contrast,nearlyallof the Sudburylakeswerefoundin Q-lll,reflecting the lesser importance of allochthonous DOMin theselakes.On a secondlevel,the lakeswerediscriminated gradientspanning alongAxis2, definedasan autochthonous from Q-lV,with highvaluesfor Fl,% urbandevelopment, Mg andCaconcentration, andpH to Q-llwith high% vegetation.LakesBethel and Ramsey werefoundat the top of thisgradientQ,-lV, indicating the importance of the production graphically autochthonous of DOMin theselakes.Thisordination integrated the site distribution we achieved (Figure with the two independent clusteranalyses S2a,b).General discrimination gradientandsecondarily basedprimarilyon the allochthonous on the autochthonous gradientis illustrated by the schematic in the lowerrighthandcornerof Figure53. tt4 o-/\/ C\f BE-070 .2 x nnn DU47 \a Ml-07r M-07 o oDF-07 Torock .afif.=-.--;-t .DS-08 ve-;s--- Crl /o / Q-/,J Figure53. Axis1 %veg CR-07o GE{g / C1 vA-07 RA-08. ruf,oz *"rf'@ Q-1/ watershed Principal componentanalysisusingwaterchemistry(pH,Caand Mg concentrations), and DOM information(%surfacewater,rockoutcrops,vegetationand urbandevelopment) properties(DOCconcentration, components SUVA,Fl andthe relativeproportionof the PARAFAC axesare labeledAxis1 and Axis2, respectively. CL,C2,C3and C4).Theprimaryandsecondary variablescontributingto the gradientsandthe dark Thegreyarrowsrepresentthe environmental circlesrepresentthe individuallakesamples(identifiedby lakecode- yearsampled).Each quadrantof the figureis enumeratedQ-1,Q-ll,Q-lllor Q-lV.Theschematicin the lowerrighthand gradientsidentified. the majorenvironmental cornerof Q-ll illustrates 115 116 Article2 Tracemetalspeciationpredictionsin naturalaquaticsystems:incorporationof quality dissolvedorganicmatter (DOM)spectroscopic KristinK.Mueller',StephenLofts',ClaudeFortin'andPeterG.C.Campbell' ' INRSEauTerreet Environnement, 490de la Couronne, Universit6 du Qu6bec, Qu6bec(Qu6bec), Canada G1K9A9 2 Centrefor Ecology Lancaster Environment and Hydrology, Centre,LibraryAvenue,Bailrigg, Lancaster, LA14AP,UnitedKingdom Publi6en 2Ot2 dansEnvironmentalChemistry9: 356-368 DOI:1-0.1071/8N11156 Abstract To calculate metalspeciation in naturalwaters,modellers mustchoosethe proportionof dissolved organicmatter(DOM)thatis activelyinvolvedin metalcomplexation, definedhereasthe % active fulvicacid(FA);to be ableto estimatethisproportionspectroscopically wouldbe veryuseful.In the presentstudy,we determinedthe free Cd2*,Cu2*,Ni2*andZn2* concentrations in eightCanadian Shieldlakesandcompared thesemeasured concentrations to thosepredictedby the Windermere proportions HumicAqueousModel(WHAMVl). Forsevenof the eightlakes,the measured of Cd2* andZnz*fell withinthe rangeof valuespredictedby WHAM;the measuredproportionof Cu2*fell withinthis rangefor onlyhalfof the lakessampled, whereasfor Ni,WHAMsystematically the proportionof Ni2*.With the aim of ascribing overestimated betweenmeasured the differences and modeledmetalspeciation to variationsin DOMquality,the % activeFAneededto fit modeledto measuredfree metalconcentrations variationin the spectroscopic wascomparedto the lake-to-lake qualityof the DOM,as determined measurements. Relationships by absorbance andfluorescence between% activeFAand DOMqualitywere apparentfor Cd,Cu,Ni andZn,suggesting the possibility of estimating the % activeFAspectroscopically to refinemodel andthen usingthisinformation predictions.Therelationships for Ni differedmarkedlyfrom thoseobservedfor the other metals, suggesting that the DOMbindingsitesactivein Cd,Cuand Zn complexation are differentfrom those involvedin Ni complexation. To our knowledge, hasbeen thisisthe firsttime that sucha distinction resolved in naturalwatersamples. 118 RunningHead Predicting metalspeciationin lakewaters EnvironmentalContext To assess contaminants suchas metals,one needsto be ableto the riskposedby environmental identifythe keychemical stumbling species that prevailin naturalwaters.Oneof the recognized dissolved organicmatter(DOM).Here blocksis the needto quantifythe influence of heterogeneous we explorethe possibility of DOMto determineits quality,to alleviate of usingthe opticalsignature properties,andto improvethe predictionof aboutits metal-binding the needto makeassumptions in naturalwaters. tracemetalspecies distributions r.19 1 Introduction as a keyto Knowledge of the speciationof tracemetalsin naturalwatersis widelyrecognized impacts. In natural boththeirgeochemical mobilityandtheirecotoxicological understanding including the common freshwater systems, tracemetalsmayinteractwith a varietyof ligands, of natural inorganic anions(e.g.,OH-,HCOg-, organicmolecules CO32-, Cl-,F-,SOaz-), simplemonomeric suchas origin(e.g.,metabolites or anthropogenic suchascitrateor glycine,or polycarboxylates (fulvicandhumicacids)(Batleyet al.,2004). nitrilotriacetic acid),anddissolved humicsubstances perspective, Froma geochemical or ecotoxicological the keyspeciesis the free metalion (M'*),which ligands andwith is usuallypresumed or particulate to be in equilibrium with thesevariousdissolved from which it is the masterspecies the epithelial or cellsurfaces of the residentaquaticorganisms; the concentrations of all otherspecies of interestcanin principlebe calculated. in natural A numberof approaches havebeenusedto determinefree metalion concentrations (lET)hasbeenemployedto determineCd2*,Ni2*and waters.Forexample, equilibrium ion exchange (Doigand Liber,2007;Fortinet al.,2010;Wormsand Zn2*concentrations in environmental samples potentiometry Wilkinson, 2008).Variouselectrochemicaltechniques havealsobeenused,including (Rachou with ion-selective electrodes etal'.,2007; XueandSunda,1997)andcathodicstripping (XueandSigg,1993;XueandSunda,L9971, voltammetry technique as hasthe Donnanmembrane (Sigget al.,2006;Wenget al.,2011). However,giventhe complexityof the analyticalmatricesin whichthe free metalion is found,andthe oftenvery low ambientmetalconcentrations, the determination of [M'*] in naturalwatersis not a routinemeasurement. A complementary approach to suchanalytical wouldbe the useof chemical determinations equilibrium modelsto calculate Forsimplesystems, containing the freemetalion concentration. inorganic is reasonably cationsandanionstogetherwith simplemonomericligands, suchan approach straightforward. However, for waterscontaining naturaldissolved organicmatter(DOM),the presence chemical equilibrium calculations mustdealwith the challenging of fulvicandhumicacids. Theseubiquitous but poorlydefinedmaterials ligands and poly-electrolytes behavelikemulti-dentate in solution.DudalandG6rard(2004)havereviewedvariousapproaches to that havebeendeveloped L20 two that are equilibrium models,including accountfor naturalorganicmatterin aqueouschemical HumicAqueousModel(WHAM)andthe Non-ldeal currentlywidelyused:the Windermere (NICA)-Donnanmodel. In additionto fulvicandhumicacids,DOMalso Adsorption Competitive (aminoacids,proteins, of whichtend carbohydrates), the concentrations includes biogenicmolecules and uptakeby heterotrophic at verylow levelsdueto theirdegradation to be maintained of the DOMpoolnormallywill not affecttrace microorganisms. Asa result,theseothercomponents elementspeciation(Moreland Hering,1993). In watersaffectedby wastewatereffluents,the DOM origin. agentsof anthropogenic mayalsoincludestrongchelating measuredin naturalaquaticsystemswith the Recentcomparisons of free metalion concentrations predictedfor thesesystems modelshave with the NICAandWHAMspeciation concentrations majordifferences betweenmeasuredand predictedvaluesfor somemetals,notably demonstrated Cu,Ni and Pb(Fortinet al.,2010;Guthrieet al.,2005;Unsworthet al.,2006).LoftsandTipping (2011)havediscussed possible including errorsin measuring the free reasons for thesedifferences, that are usedas inputdatafor the model,anddifferences the variables metalion,errorsin measuring and humicacids betweenthe DOMpresentin the naturalwatersandthe typesof (isolated)fulvic models.In the chemical equilibrium or parameterize beenusedto calibrate that havetraditionally the presentpaperwe haveexploredthis latterfactor. equilibrium modelto a naturalwateris to involvedin applyinga chemical Oneof the challenges modelsthat include estimatethe fulvicandhumicacidcontentof the water. Currentspeciation quantityof DOM,but the measured dissolved organicmatter(DOM),suchasWHAM,incorporate the percentof the DOMthat is activelyinvolvedin usersmustdefinethe DOMqualityby choosing what proportions of this"active"DOMshouldbe designated metalcomplexation, and by deciding would be very active"DOMspectroscopically fulvicand humicacid. To be ableto estimatethis"%o a measureof the of the DOM(SUVA25a), ultraviolet absorbance the specific useful.Forexample, aromaticity of the organicmatter,hasbeenusedto estimatethe proportionof DOMthat is present in the complexation of Cuin reconstituted as humicor fulvicacidin solutionandthat participates of Cu et al.,2001)and in the complexation naturalwatersamples(Luideret al.,2004;Richards tzL (Ameryet al.,2008)andCd(Cornuet al.,2009;Cornuet al.,2011)in soilextracts.A bettero priori of the qualityof DOMin a givensamplewouldalleviate the needto makeassumptions understanding free betweenpredicted andmeasured aboutthe activityof DOMand mightreducethe discrepancies metalconcentrations. (i) to determinethe free Cd2*,Cu2*,Ni2*and Zn2* Thepurposeof this paperis thusthr:ee-fold: gradient;(ii)to comparethese in a setof lakeslocatedalonga metalcontamination concentrations predictedby WHAMVl; and(iii)to measured with the free metalconcentrations concentrations of estimatingthe explorethe useof the opticalsignatureof the DOMin eachlakeasa mea"ns proportionof the naturalDOMthat is activein metalcomplexation. 2 Methodology 2.1 Studyareo Althougha detaileddescription a brief of the studysitescanbe foundin Muelleret al.(2012b); descriptionis givenhere. Lakewaterwascollectedfrom lakesin two differentregionson the Canadian Shield:nearRouyn-Noranda in north-western Qu6becandnearSudburyin north-central Ontario.Bothstudyareashavebeengreatlyimpactedby metalminingandsmeltingactivities, particularly by atmospheric deposition et al.,1998;Dixitet al.,2007). of acidand metals(Borgmann ThepH of the lakestendsto decrease tend to increasewith a andtotal metalconcentrations particularly decrease in the down-winddistance in the Sudburyarea(Yan from the metalsmelters, andMiller,1984).Acidminedrainage from pointsources, minesor mineralized suchasabandoned mayalsodecrease outcrops, the pH of lakewaterlocally.In eachregion,lakeswerechosento represent a gradientin waterquality(pH,dissolved tracemetal organiccarbon(DOC))and concentrations. 2.2 Lakewatersampling Lakewatersamples weretakenfromfour lakesin eachof the studyareasduringJulyandAugustof 2008.Sampling siteswerechosenin the littoralzonesof small,well mixedlakesandthe sampling siteswereconsidered to be representative of of the wholelakeepilimnion.Thespatialvariability r22 (i.e.,pH andconcentrations waterchemistry of DOCandmajorcations)is knownto be minimal withinmanyof the sampledlakes(Fortinet al.,2010). in Fortinet al. Lakewaterwascollectedpassively usingequilibrium diffusionsampling asdescribed jars(Nalgene, (20L0)and Muelleret al.(Muelleret al.,20L2b).Briefly,250-mLpolypropylene Nalge plasticlidsfittedwith a NuncInternational Corporation, Rochester, NY)toppedwith custom-made PallCanadaLtd,VilleSt-Laurent, were 0.2-pmfiltermembrane(HTTuffrynMembrane, QC,Canada) MA) ultrapurewater(>18Mohms-cm) filledwith Milli-Q(Millipore, EMDMillipore,Billerica, andfixed wereinstalled to plasticrodsthat wereanchored to the lakebottom.Thesamplers at a depthof 1 m period,the from the lakesurfaceandleftto equilibrate for 13to 14 days.Following thisequilibration samplers werebroughtto the surface, the membrane surfacerinsedwith ultrapurewater,the piercedwith a cleanplasticpipettetip andthe customjar lidsreplaced with clean membranes polypropylene screwlids. Thesamplejarswere baggedand storedin the darkat 4oCin field coolers for a maximumof 12 daysuntilthey were transportedbackto the laboratory.Notethat lakewater pH,conductivity weremeasured in the fieldat the time of samplecollection. andtemperature 2.3 Sompleanolysis d in tO%(v/v)nitricacidfor at least24 h and All plasticlaboratory andsampling equipmentwassoake then rinseda minimumof sixtimeswith ultrapurewateranddriedundera Class100laminarflow vialsusedfor storingthe subsamples hood. Theonly exceptionwasfor the polystyrene for anion wassoakedin a analyses; thesevialswererinsedthreetimeswith ultrapurewateronly. Allglassware 2 N HCIsolutionbeforebeingrinsedwith ultrapurewater. jarshadarrivedin the centrallaboratory, theywereopenedin a cleanClass Oncethe polypropylene werecollected asfollows.A 1-0-mL subsample was L00laminarflow hood,andsubsamples ultra-tracenitricacid vialand acidifiedto2o/o(v/v)with concentrated transferredto a polypropylene (BDHAristarUltra,VWRlnternational, Mississauga, ON,Canada)for analysis of majorcations(Ca, (lCP-OES, VarianVistaAXCCD,AgilentTechnologies Canada Mg)by atomicemission spectroscopy Inc.,Mississauga, coupledplasmamass ON)andtotal metals(Cd,Cu,Ni andZn)by inductively t23 (lCP-MS, A secondL0ElementX Series, spectrometry ThermoScientific, Mississauga, ON,Canada). (F-, mLsubsample wastransferred of anionconcentrations to a polystyrene vialfor the determination (DionexICS-2000, SOa2and POa3-) ThermoScientific). Cl-,NO3-, by ion chromatography Detailsconcerning organicmatter the quantitativeand qualitativecharacterization of dissolved (DOM)canbe foundin Muelleret al. (Muelleret al.,2OL2b).Briefly,20-mLsubsamples were transferredto amberborosilicate vials(l-ChemBrand)for total organiccarbonanalysis(TOC-Vcpn, (1.mL)weretransferred Shimadzu Scientific Instruments, to 3-mLglass Columbia, MD). Subsamples (Kendall vacutainers Monoject,Mansfield, carbonanalysis by gas MA)for totalinorganic (VarianUV-Vis (Varian3800GC,AgilentTechnologies chromatography lnc.).Theabsorption Canada (VarianCaryEclipse Cary300Spectrophotometer, AgilentTechnologies CanadaInc.)and fluorescence Fluorescence Spectrophotometer, AgilentTechnologies CanadaInc.)spectraof lakewater DOMwere (SUVA, ThespecificUVabsorbance alsomeasured. by dividingthe L m-1mg C-1)was alsocalculated measuredabsorbance at254 nm by the concentration et al.,2003). Excitationof DOC(Weishaar emission matrix(EEM)fluorescence spectraweremeasured using1-cmquartzcuvettesat excitation wavelengths from 210to 400 nm at incrementsof 5 nm and at emissionwavelengths from 300to 580 nm at increments of 2 nm. Thefluorescence indexwascalculated asthe ratiobetweenthe fluorescence emission intensityat 47Onm overthat at 520nm for an excitation at 370nm (Coryet al.,2OLO; McKnight et al.,2001).Finally, tool,parallelfactoranalysis the multivariable statistical (PARAFAC), wasemployedto deconvolute the EEMspectrausingthe procedure outlinedby Stedmon and Bro(2008)as implementedin a MATLABenvironment,usingthe "N-waytoolboxfor MATLAB" (Andersson and Bro,2000). 2.4 Metol speciationmeosurements All sub-samples for metalspeciation filter;AMD measurements werefiltered(0.2-pmpolycarbonate Manufacturing Inc.,Mississauga, was ON)undera laminarflow hoodbeforeanalysis. A precipitate just priorto the metal observed in a few samples, whichprompteda filtrationstepfor all samples speciation analyses. WHAMmetalspeciation simulations wereperformedusingthe element concentrations measured on thesefilteredsamples.All calibration werepreparedusing solutions r24 ultrapurewater. Stocksolutions of Cd,Cu,Ni andZn wereprepared from ICP-MS standardsolutions (4%HNO3, PlasmaCAL, SPCScience, BaieD'Urf6,QC,Canada). To takeinto accountthe variable naturalconditions of eachlakesampled, e.g.pH,ionicstrengthandtotal metalconcentrations, a (LOQ)wascalculated lake-specific limitof quantification for the measurement of the free metalion for eachlake,basedon the standarddeviationof triplicatemeasurements. Whenthe free metal valuefor a lakefell belowits LOQ,it wasdiscardedandthe free metalion was not reported. 2.4.7 lon exchangetechnique(lET) Theconcentrations of Cd2*,Ni2*and Zn2*were measuredfollowingthe methodsdescribedby Fortin (L998)and updatedby Fortinet al. (2010).Briefly, andCampbell the ion-exchange technique involves the equilibration of the free metalionsin an unknownsamplewith a cationexchange resin (Dowexmesh,Sigma-Aldrich CanadaLtd.,Oakville, ON,Canada)that hasbeencalibrated with standardsolutions according to the followingreaction, M2n +(3 -;r) . R.CatZ RrM + (3 - x).Cat'* (1) whereM 2*is the free metalion, R is the resin,R,Cat is the concentration of resinbindingsitesfilled by a mono-or divalentcation,.r is the chargeon the cationandR"M is the concentration of resin bindingsitesfilledby the metalion of interest.Theconditional equilibrium constant, Kfu, for the abovereactionis calculated usingthe followingequation, u, _IR,M ]. [*,'. KL=ffi ]'-' (2) Underswamping electrolyte conditions, to remainconstantand lcot'.1 anO[n.Cot] areassumed equation2 is rearranged to givethe resindistributioncoefficientat fixedionicstrengthand pH, Ao.,.,, (t-.e'). Lo,i.pn : Kio .fcot'-ft-' =[!rrJ (3) [R,cor]'-' lu'.l Theresinwascalibrated with standardsolutions with a pH rangeof 5.3to 8.9andwith known concentrations of free Cd2*(0.24to 42 nM), Ni2.(13to 202nM) and zn2*14.9to 148 nM) andthe 125 resindistributioncoefficient(tr)wascalculated for eachmetalbeforeapplyingthe methodto the (seeSupplementary naturalunknownsamples Material,FigureS1). Experimentally, the resinwaspre-equilibrated with a matrixsolutionthat hadan ionicstrength (Ca(NOa)z asthe swamping solutionandthe natural electrolyte) and pH similarto the calibrating (99.0%, samples; a sufficient volumefrom a 25 mM Ca(NOg)z Sigma-Aldrich CanadaLtd.)stock solutionwasalsoaddedto eachnaturalsampleso asto obtaina constantCaconcentration of 0.46 mM in all naturalsamples.By levelling we ensuredthat the the ionicstrengthof the naturalsamples, and time neededto reachequilibrium betweenthe resinandthe samplewassimilarfor all samples that a singlemetal-specific distributioncoefficientcouldbe used.Aftera steadystatehad been reachedbetweenthe resinandthe sample(190mL),the resinwasrinsedwith ultra-pure waterand elutedwith a volume(V,6 mL)of IO%(v/v)ultra-tracenitricacid(BDHAristarUltra,VWR lnternational). Theamountof free metaloriginally boundto the resinat steadystate( R M ) was calculated usingthe followingequation, ln t11 lMr,*,"f'V m, LRzM l7!_____!JJ!!!r!_ (4) where lM u,.*] is the concentration of metalmeasuredin the eluateand m,is the massof the resin used(approximately 7 mg;weighedprecisely). 3 and4, the free Finally, by combining equations metalconcentration in the samplewascalculated usingequation5. lu*l=Wu'*"|l' (s) Ao,i,pH'm, 2.4.2 Cupricion selectiveelectrode(Cu ISE) Because the lETtendsto overestimate freeCu2*whenthe lakepH isgreaterthan6.5(Fortinet al., 2OLO), we determinedCu2*in the lakesamplesusinga cupricion selectiveelectrode(CulSE,Orion, ThermoScientific) followingthe methodsoutlinedby Rachou et al.(2007).TheCuISEwascalibrated dailywith a solutioncontaining 0.1 mM coppernitrate(Cu(NOr)2, SPCScience), 1 mM PlasmaCAL, iminodiacetic acid(lDA,Sigma-Aldrich hydrogenphthalate Canada Ltd.),2.5mM potassium (KHCaHaO+,99.95%, Sigma-Aldrich Fluka, Canada Ltd.),10 mM potassium nitrate(KNO3,99.995%, 126 (NaOH,98%, CanadaLtd.)and6 mM sodiumhydroxide Sigma-Aldrich Sigma-Aldrich CanadaLtd.). Thecalibration solutionwassplitinto 5 to 11 subsamples with pH valuesvaryingfrom2 to 11-, adjustedby the additionsmallknownamountsof eitherHNOgor NaOH,andallowedto equilibrate overnight.Thefree Cu2*concentration calculatedrangedfrom the methoddetectionlimit (three timesSDof sixreplicate blankmeasurements) of 2.5x10-11to 6.3x10-s M. Thecalibration subsamples with pH valuesabove6 were bubbledwith Nzgas,to avoida decreasein pH with the dissolutionof carbondioxide,COz(aq). atmospheric ThepH andtemperature of samples weremonitoredusingan Orion(Orion,ThermoScientific) electrode and probe,respectively. DuringCuISEmeasurements, constantstirringandtemperature conditions werealsomaintained. Thecalibrated response of the wasdetermined CuISEelectrode fromthe plot of the measured ISEpotentialagainstthe calculated pCufor eachcalibration subsample. Forthe naturalwatersamples, the concentration of Cu2*was then calculated usingthe CuISEcalibration equationandthe measured electrodepotential.The naturalsamples wereamendedwith KN03(99.995Yo,Fluka, Sigma-Aldrich Canada Ltd.;a specific volumeof a 1.0M stocksolutionwasaddedto achievea finalconcentration of 10 mM KNOa) to ensurea constantionicstrengthamongthe naturalsamples. 2.5 Metal speciationcalculations Thefree Cd2*,Cuz*,Ni2*andZn2*concentrations measuredfor eachlakewerecomparedto values usingthe Windermere calculated HumicAqueousModel(WHAM,modelversion6.1)(Tipping, 1-998). Several weremadein the application assumptions of the modelto the lakewatersamples. First,we that the DOC:DOM assumed ratiowas2 (Buffle,1988),that 65%of the DOMwasactivein the complexation of metals(Bryanet al.,2002),andthat thisactivefractionwascomposed of fulvicacids only. We alsoassumed that both Fe(lll)andAl(lll)activities arecontrolledby the solubility of their hydroxides, as calculatedusingthe empiricalequationsgivenby Loftset al. (2008)andTipping(2005), respectively. Othermeasured WHAMinputparameters includedpH andthe totalconcentrations of Na,Mg, K,Ca,Cr,Co,Ni, Cu,7n,Cd,Cl,NO3,SOa,CO3(measured astotal inorganiccarbon)and F. The measured surfacetemperature of the lakessampled{usedin the WHAMsimulations) variedbetween 20 and 23 degrees Celsius. Separate WHAMsimulations wererun for lakesamples analysed usingthe IETfree ion measurement method(forthe analysis of freeCdz*,Ni2*andZn2*)andfor thoseanalysed L27 corresponded with the ISEmethod(forthe analysis of freeCu");the inputdatafor thesesimulations for thusaccounting exactlyto the composition of the samples usedfor the freemetalion analyses, (additions the differences in theirioniccomposition and KN03,respectively). of Ca(NOe)z 2.6 Statisticalanalysis All statisticalanalyses were calculated at the alphaequals0.05confidencelevelusingSigmoPlot11.0 (SystatSoftwareInc.,SanJose,CA). Parametric were t-testsand linearand non-linearregression usedwhenthe normalityand homogeneity amongdatasetswere confirmed;otherwise, of variances the appropriate RankSumtest,in placeof the non-parametric testswereemployed(Mann-Whitney percent t-test. Notethat for the regression equationsgeneratedbetweenthe WHAMcalculated (%aFAootimized) fulvicacidactivein metalcomplexation and the relativecontributionof PARAFAC fluorescence componentsl and 3 to the overallfluorescence spectrumof lakeDOM (seesection 3.4),the assumptions wereconfirmed.Theonlyexception of normalityandhomogeneity of variance wasfor Ni,wherethe variance amongthe datasetswasnot constant(P=0.01). 3 3.1 Resultsand Discussion Waterquality Thespeciationof dissolved of DOM, metalsin freshwatersystemsis sensitive to the concentration whichtendsto complexfree metalions,but mayalsobe affectedby otherwater qualityvariables, notablythe concentrations of Ca,Mg,Al, Feand the H*-ion,thesebeingpotentialcompetitorsfor the cationbindingsiteson the DOM(Tipping et al.,2002).In addition,for a waterof givencomposition, the total dissolved concentration of the metalmayaffectits speciation(i.e.,a limitednumberof high affinitybindingsites).In the followingtwo subsections how the waterchemistry varied we describe amongthe studiedlakes. 3.2 Backgroundwater chemistry pH)of the lakeswasdescribed Thewaterquality(i.e.[majorcations], in Muelleret al. (2012b)for 2007and2008.Onlythe 2008dataareconsidered measurements here,sincethe metalspeciation (pH,Ca,Mg and DOCconcentrations) datefrom that year.Widerangesin generalwaterchemistry L28 werefoundamongthe eightlakessampled(Table1),reflecting both regionalgeological differences and localdifferences at the watershed scale(Muelleret al.,2OL2bl.Thelakesin bothsampling regionshavelow concentrations of calcium(Ca:68 to 464prM)and magnesiumlMg:27 to 274 gMl, contributingto generallysoftwater. LakeBethel,from the Sudburyregion,hadthe highestpH and the highestCaand Mg concentrations of all the lakes;for the IETmeasurements, all otherlakewater (seesection2.4.tl'. werethusadjustedupwardsto theseCaconcentrations samples Lakesfrom the Rouyn-Noranda regionare circumneutral, with a meanpH of 7.25anda narrowrange of valuesbetweenpH 6.81and7.69.ThemeanpHfor the Sudburylakes(7.07)is slightlylowerthan that of the Rouyn-Noranda lakesand pH valuesspana widerrange(6.40to 8.08).Historically, the Rouyn-Noranda lakeswere lessaffectedby anthropogenic acidification than werethosein the Sudburyarea(Dupont,t9921,despitemassiveSOzemissions from the Hornesmelter;the RouynNorandalakesare locatedin the Abitibiclaybelt and are effectivelybetterbufferedagainstacid pH of the lakesin the Sudburyregionsis evidence deposition.Althoughthe average circumneutral of the recoveryof theselakesfrom historicalacidification in recentyears(Kelleret al., 2OO7l,pH values in the lakesof thisregionspanneda widerrangethan in the Rouyn-Noranda lakes. (1-4pg chl-aL-1; Basedon their historical chlorophyll-a concentrations Campbell, unpublished data), the four studiedlakesin the Rouyn-Noranda regionareoligotrophic, asare LakesRaftandWhitsonin the GreaterSudburyarea(2-3pg chl-aL-1; h t t p :" v w w . g r e a t e r s u d b u r v . c a / c m s ' i n d e x . c f m ? al apkpe=wd ai vt e r q u a l i t v & l a n g = e n & c u r r l D = 6 1 9 . accessed 7 August2OL2).We were unableto find comparabledatafor LakeGeneva,whichlies outsidethe GreaterSudburyarea,but giventhe lackof development in itswatershed, it is alsolikely oligotrophic.Theonlyexceptionis LakeBethel,whichis muchmore productive(L2-24pg chl-aL-1; http://www.greatersudburv.ca/cms/index.cfm?app=div lakewaterqualitv&lang=en&currlD=61-9, 7 August2OL2)and canbe classified accessed aseutrophic. I29 3.3 Dissolvedorgoniccarbon ThemeanDOCconcentration of Rouyn-Noranda lakeswashigher(6.1mg C t-1)andspanneda wider range(4.9to 8.1mg C t-1)thanin the lakesfrom the Sudburyregion(meanof 3.4mg L 1;rangefrom regionthan 2.Oto4.9 mg t{). the higherDOCconcentration in the lakesin the Rouyn-Noranda arearatiosfor the Sudburylakesreflectsthe greatersoilcoverandthe higherwatershed-to-lake Rouyn-Noranda lakes.As described in Muelleret al. (z)Lzbl,thequantityandthe qualityof DOMin andaswellas DOM the lakessampledvaryasa functionof regionalwatershed characteristics processing withinthe watershed andwithinthe lakeitself. DOMopticalquality(SUVA254 between and Fluorescence Index)wasfoundto differsignificantly regionsandamonglakes(seeTable51 in the Supplementary sampling Materialformoredetails)and (Muelleretal.,2OI2b).Theauthorsalsousedparallelfactor is described elsewhere analysis (PARAFAC) to extractfluorescence matrix(EEM)fluorescence componentsfrom excitation-emission including spectrameasured for 19 lakesin the Rouyn-Noranda the eightlakes andSudburyregions, retainedfor the presentstudy. Of particularinterestwerethe humic-likefluorescence components of ubiquitous origin(C1)andallochthonous to a reduced CLcorresponds origin(C3).Component quinonefluorophore to an oxidized associated with higherplantmatter,whereasC3corresponds quinonefluorophore(Muelleret al.,2OL2b).We usedthe proportipnof eachof theseDOM fluorescence i.e., components(relativeto the sumof all four PARAFAC fluorescence components; as proxies CI/C1or C3/Cr)as measures of the qualityof DOMsampledin our lakes,andin particular for the fractionof DOMactivein the complexation 3.3.2). of traceelements(section 3.4 Lake-to-lakevariationsin metal concentrationsand metol speciotion 3.4.1 Totoldissolvedmetal ion concentrations (Cd,Cu,Ni,Zn)variedmarkedlyamongthe eightlakes Thetotaldissolved metalconcentrations gradientratios(i.e.,the sampled(Table1). Forlakesin the Rouyn-Noranda region,the concentration + minimummetalconcentration, ratioof the maximummetalconcentration IM]r."/[M]rin) decreased in the sequence Cd(33)- Zn (31)> Cu(7)> Ni (2.5).In the Sudburyregion,the ratiosdecreased in the orderNi (63)> Cd(3L)> Cu(16)> Zn (5). Lakesin the Rouyn[M]r.,./[M]n';n 130 Norandaregionhadsignificantly highermeantotal concentrations of Cd (P=0.004) and Zn (P=0.030) thanthe lakesin the Sudburyarea,but lowerconcentrations of Ni (P<O.OOl). No significant differencein total dissolvedCuwasfound betweenthe samplingregions. Thelake-to-lake variability in the measured totaldissolved metalconcentrations withina givenregion is largelydueto variations in atmospheric loadingfrom the localminingandsmeltingactivities (Borgmann et al.,1998;Couillard et al.,2004).Despite the markedreductions in smelteremissions that havebeenachievedoverthe past30 years,lakesin closeproximityto and down-windfrom the metalsmelters(e.g.,LakesDufaultand Dasserat in the Rouyn-Noranda region,LakesRaftand Whitsonin the Sudburyarea)stillhaveconsiderably highertotaldissolved metalconcentrations than do thoselocatedupwindor far downwind. 3.4.2 Freemetal ion concentrotions Thefree Cdz*,Cu2*, Ni2*and Zn2*concentrations alsovariedamonglakes(TableL). The ratioof for the Rouyn-Noranda lakesdecreased asfollows:Zn (I2Il > Cd (54)> Cu(5.5). [M2.],nr"/[M2*]n,;n FreeNi2*concentrations wereonlymeasurable in one replicate samplefrom one lakefor the Rouyngradientratiowascalculated.The ratioof Norandaregionand,therefore,no concentration for the Sudburylakesdecreased in the sequence Ni (136)> Cu(64)> Cd(40)> Zn [M2*],.*/[M2*],n;n (22).Asanticipated, lakesthat hadthe highestconcentrations metalalsoexhibitedthe of dissolved highestfree metalion concentrations. 131 Table1. MeasuredpH and mean(t SD,n=3)dissolved organiccarbon(DOC), total dissolved concentrations of majorcationsand trace metals,aswell ' (QC)andSudbury(ON)in 2008. as measured free metalconcentrations and%treemetalvalues for lakessampledfrom Rouyn-Noranda (LOQcalculated indicatesonlyone sampleabovelimit of quantification asthreetimesthe standarddeviationof triplicateanalyses of each lake).NA indicatesno samplesabovethe LOQ. Dasserat(DS) pH DOC(mg.L-l) Ca(pM) MS (rrM) Cd(nM) Cu(nM) tt E = o o c) = Ni(nM) Zn (nM) cd'?.(nM) cu2*1nM; Ni2*(nM) zn2*1nM; o/o Cdz* o/oCu2' 7o Niz* o/oZn2* 6.92 4.85r 0.03 2 0 7! 1 93.7r 0.3 0.8i 0.2 112t3 8x2 464t 51 0.5r 0.2 0.37r 0.04 0.6' 279!32 63r36 0.32r 0.04 8 60r10 Rouyn-Norandaregion lakes Dufault(DU) Opasatica(OP) 7.59 5 . 2t 0 . 1 4 0 4! 3 118!1 7.69 6 . 3r 0 . 1 2 1 8t 2 Vaudray(VA) Bethel(be) 6.81 8 . 1 0r 0 . 0 8 8.08 4.89i 0.03 464r.3 83r1 Sudbury region lakes Geneva(ge) Raft (ra) 6.4 2 . 0 3r 0 . 0 3 6.9 3 . 7 7r 0 . 0 5 68r3 81t2 0.56r 0.02 86.0r 0.9 143x1 78.4t0.4 0 . 5 1r 0 . 0 4 165r3 7 7 5! 4 1 1136r54 2 7 4t 3 26L1 0 . 0 1 8 1 0 . 0 0 9 0.021r 0.005 2 4 . 9t O . 1 10.5r 0.6 226t2 1 8! 2 114t1 0 . 0 6r 0 . 0 1 3 5 . 1r 0 . 2 3 7 . 1r 0 . 6 0.608r 0.005 43t1 12.5t 0.4 15!4 1 2r . 1 52t5 1 2t 3 1 . 1 4r 0 . 0 6 0.021* 0.008 NA 0.44r 0.09 0 . 0 8i 0 . 0 1 0.26r 0.02 0.20i 0.08 21xB 0.01' NA 1 8 8+ 2 3 NA NA 0 . 0 5r 0 . 0 1 7 . 3r 0 . 9 0.48r 0.09 4.0t0.7 2 " 3r 0 . 9 3 7r . 1 7 31r18 2 . 3t 0 . 4 1 0 . 7* 0 . 9 51r16 42t4 0.47t0.2 NA 0.20i 0.05 70 4 . 6r 0 . 9 7 1t 2 3 . 7r 0 . 6 2 . 0! O . 2 2 4 6x 3 5 . 1r 0 . 3 342 t60 58r5 0 . 1 7t 0 . 0 4 O.21t0.04 Whitson(wh) 6.9 2 . 9r 0 . 1 46t1 62t22 0.402r.0.004 3.2 i 0.5 545r9 5 2t 2 0 . 3 1f 0 . 0 5 1 . 0r 0 . 1 489r.72 32r.4 61*10 NA NA NA 3 . 2t 0 . 4 22t5 70t4 0.57r 0.08 43t7 55!12 1 5! 7 60r36 20r6 51t20 82r39 62!8 In the presentcontext,wherewe arecomparing measured andpredictedmetalspeciation in differentlakes,it is importantto considerthe relativeproportionof the free metalion in eachlake,as in totaldissolved metalconcentrations this measure takesinto accountthe differences amonglakes. thanthe otherthreemetals- the mean Asanticipated, copperwasmuchmorestronglycomplexed percentfree metalfor copperwasonly L.3%o,whereas the meanpercentages for Cd2*(54%),Ni2* (29%)andZnz*(50%)weremuchhigher(Table1). Forall four metals,and in boththe Rouyn-Noranda pointingto asthe pH increased, andSudburyregions, the percentfreemetaltendedto decrease competitionby protonsasan importantfactoraffectingmetalspeciation.A general complexation for Cd,Cuand Ni asthe DOCconcentration declinein the percentfree metalwasalsoobserved presumably ratio(seeFigs.53 and 54 in the increased, reflectingan increasein the ligand-to-metal Materialfor thesepH and DOCrelationships). Supplementary 3.4.3 Comporisonof the observedmetol speciationwith literature volues In comparingmetalspeciationin differentfreshwaterenvironments, the reportedfree metalion are not particularly concentrations useful,sincetheywill varyaccording to the metalloadings affectingthe water body. Comparisons of free metalion proportionsare a more meaningfulway of evaluating how metalspeciation variesfrom one lakeor riverto another.In Table2, we have compiledfreemetalion percentages that havebeenreportedin the literaturefor Cd,Cu,Ni andZn (the recentreviewpaperby LoftsandTipping(201L)was usedasa startingpointto identifypapers wherefree metalion concentrations for naturalwatershad beenreported).Table2 cannotclaimto be comprehensive, asthe representation of surfacewatersfrom Switzerland and Canadais high,but the compilation in metal abnormally doesoffersomeinsightinto metal-to-metal differences speciation. is the mostextensive Copperis the metalfor whichthe Table2 database andit is alsoconsistently the moststronglycomplexedmetal,followedby Ni > Cd> Zn. Fora givenmetal,however,there are distinctregional differences; the percentfree metalis virtuallyalwayslowerin the alkalineSwiss Precambrian lakesthan in moreacidicand moredilutelakeson the Canadian Shield.On a relative wherethe percentage is particularly scale,thisdifference notablefor Cu2*, canbe lessthan0.00L%in 133 manyof the Swisslakesbut consistently above1%in the Canadian surfacewaters. Factors (carbonate and contributing to thistrendincludethe aforementioned is pH andalkalinity differences in the hydroxide complexation), but alsothe metalloadings. metalconcentrations Totaldissolved lakesare muchhigherthan in the Swisssurfacewaters(Table2), presumably reflecting Canadian Shield.Highertotal inputsfrom the miningandsmeltingactivities that are prevalent on the Canadian to leadto lower metalconcentrations, highermetal:DOM andthe resulting ratios,wouldbe expected degreesof complexation. 3.5 Lake-to-lokevoriotionsin predictedmetal speciotionond compariion with measuredvolues 3.5.1 lnitialWHAMsimulations Freemetalion concentrotions.TheWHAMVl chemicalequilibriummodelhasbeendescribedin (ModelV), Model Tippinget al.,2011).In comparisons detailby Tipping(2OO2; with its predecessor Vl doesa betterjob of simulating metalbindingto fulvicandhumicacids,particularly at low metal concentrations, for low [M]/DOCratios,andfor metalsthat tendto bindto the strongmetal-binding groups)that wereexpressly sites(putativeN andS functional introduced into ModelVl. In brief, fulvicand humicacidsareassumed to be rigidspheres of uniformsize(radiiof 0.8and 1-.7nm, groupspositioned respectively)with two classes of ion-binding typeA randomlyon theirsurface: sitescorrespond to monoproticcarboxylic acidgroups,whereastype B sitesare modeledas weaker (e.g.phenolic) products(MOUtn-tl*, acidgroups.Metalionsandtheirfirsthydrolysis competewith eachother,andwith protons,for the typeA andtype B groups.Bi-dentate sitesare metal-binding generatedby combiningpairsof proton-binding sites(A-Aor A-Bcombinations).Onlyproton-binding sitesthat are sufficiently closeto eachother are ableto form bi-dentatesites(s 0.45nm apart; estimated statistically). sites. Similarrulesare usedto generatetri-dentate metal-binding L34 Table2. Freemetalion percentages in variousnaturalwatersamples. Total dissolved Metal Sampledescription cd Ni Cu 7n Percentfree metal ion Mean SD Reference 54 13 (thisstudy) Canadian Shield 7 tMl (nMl 0.06-2.0 C a n a d i aSnh i e l d 13 0.07-9.3 72 11 Fortinet al.(20L0) Canadian Shield 4 0.9-8.0 20 16 Guthrieet al.(2005) Swisslake,UKriver 4 0.07-0.20 3.7 3.5 Sigget al.(2006) Swisslake,UKriver 5 0.06-0.87 4.9 3.5 Kaliset al.(2006) Canadian Shield 5 8-1140 29 28 (thisstudy) Canadian Shield 6 490-1760 lL 6 Gopalapillai et al. (2008) Swisslake,UKriver 4 6-20 6.6 Swisslake,UKriver 7 6-s9 3.0 Belgiansurface 6 44-75 Canadian Shield 8 Canadian Shield 1,.4 '1,.4 Sigget al.(2005) 22 16 VanLaeret al.(2006) Lr-246 1.3 1.8 (thisstudy) 2 42-L50 1-534 Swisslake,UKriver 4 L9-30 0.25 o.r2 Sigget al.(2006) Swisslake,UKriver 7 L3-76 o.2\ 0.22 Kaliset al.(2006) Swisslakes(neutral) 2 5 4-L7 <0.0010 Swisslakes(acidic) 5 3-71. Swissrivers 20 1_0-68 <0.0010 Swisslake 2 65-71. 0.3 O.2 XueandSunda(1997) C a n a d i aSnh i e l d 8 L2-464 50 23 (thisstudy) Canadian Shield 10 10-2010 86 L5 Fortinet al.(2010) Canadian Shield 4 46-1300 25 4 Guthrieet al.(2005) Swisslake,UKriver 4 28-282 10 6 Kaliset al.(2006) Swisslakes 6 Lt-25 8 L XueandSieg(1-994) Kaliset al.(2006) Guthrieet al.(2005) Xueand Siee(L996) 4 XueandSiee(1996) XueandSiee(1996) oThese " Concentration range(N=2)ratherthan the meanvalue. very low valueswere obtainedby competitiveligand exchange and cathodicstrippingvoltammetry.LoftsandTipping(2011)haverecentlysuggested that thistechniquemay underestimate free Cu2*concentrations. 135 Whenusingthe WHAMmodelto predictthe free metalconcentrations in the studiedlakes,we initiallyassumed in metalcomplexation that the components of the DOCpoolthat participate reactions couldbe represented entirelyby fulvicacid(100%FA,O%HA),andthat this FAportion corresponds to 65%of the DOC;thesechoiceswere basedon the recommendation of Bryanet wasobserved al.(Bryan et al.,2002)forcoppercomplexation in naturalwaters.Goodagreement range betweenpredictedand measuredvaluesof [Cd2.](Figure1a),overa 1-00-fold concentration (0.01(Figure1d),overa similar100-fold to 1.1 nM). Agreementfor l1n2*lwasalsoreasonable (2 to 280 nM) that were considerably higher concentration rangebut for free Zn2*concentrations thanfor Cd2*.In the leastcontaminated the predicted lakes(Geneva, Betheland notablyOpasatica), freeZnz*concentrations that WHAMVl were somewhathigherthan the measuredvalues,suggesting mayunderestimate Zn complexation at low dissolved Zn:DOC molarratios. (> 0.4nM) our ISEmeasurements At highfreeCu2*concentrations andthe freeCu2*valuespredicted by WHAMVl were in verygoodagreement(Figure1b),but belowthisthreshold,in LakesVaudray, Geneva, Opasatica and Bethel,our measured valueswere4 to 22 timeshigherthanthe predicted values.In theirstudyof Cuspeciation Cu in threenaturalwaterswith similarlow totaldissolved (19-30nM),Unsworthet al. (2006)alsoreportedmeasured concentrations valuesof Cu2*that exceeded WHAMVl predictions; techniques in theirstudy,freeCu2*asestimatedby two independent (Donnanmembrane; permeable Hollow-fibre liquidmembrane) wasL to 3 ordersof magnitude higherthanthe WHAMVl predictions. 136 1e-8 A B A _,/ ' t"ll' r uxy to" + N E 9 o/ *n/ 1e-9 ,r/r, + N ,/ 1e-10 lo-rn VA /./ ' a ' o) o q) ,/ o / 1e-11 |ege 1e-11 top / l--U i{'oe ,/ 1 e - 10 1e-'t1 1e-12 1e-12 1e-9 cv r-o be 1e-11 1e-9 1e-10 1e-8 1e-6 D DS ./'" DUj n 2_ te-z 'le-7 {_ z + be l,/ :o o o) E 1e-8 o ||,/ *^q(^ N ,/ c N o -9 o / E ge/ o 2 l,/ Ds/ T 1e-8 +#+ /vA n"tf/ /^ +{ / iA/ OPr 7b" 1e-9 1e-8 1e-7 Measured M" (M) Figure1. 1e-9 1.0e-9 1.0e-8 1.Oe-7 1.0e-6 Measured M'. (M) of the free metalion [Mz*]calculated with WHAM of mean(t SD,n=3)concentrations Comparison (withthe 65%aFAassumption) andthosemeasured usingIET(Cd,Ni andZn)and ISE(Cu).The dashedlinerepresents the 1:1line. Thegenericconstantsusedin WHAMVl havebeengeneratedfrom laboratorytitrationsof isolated ratiosin thesetitrationsareveryoftenhigherthanthose humicandfulvicacids.Copper/DOC with the andthusone mighthaveexpected calculations encountered in naturalwatersamples, genericconstants we observethe oppositeresult However, to underestimate coppercomplexation. (overestimation of Cucomplexatioh, aswasthe casefor Unsworthet al. (2006)).Clearlymore L37 to determinethe conditional experiments equilibrium for the complexation of Cuto DOM constants shouldbe carriedout underconditions approaching thosefoundin naturalwaters(e.g.,low Cu/DOC ratios;naturalratherthanisolatedorganicmatter). In our experiments, explanation but not in thoseof Unsworthet al. (2006)thereis a secondpossible of an analytical naturefor the discrepancy betweenmeasured and modeledfreeCu2*concentrations. In the four lakeswith low total Cuconcentrations, may be below the total dissolved Cuconcentration that whichis necessary to properlymeasure the freeCu2*concentration usingan ion-selective electrode(-fO-t M) in a partiallybufferedsystem(Buffle,1988;Mota and CorreiaDosSantos,1995; Rachouet al.,2007). In suchsystems,if the bufferingcapacityis insufficient, from Cucontamination duringthe potentiometric the electrode increase the apparent[Cu2.] measurement mayartificially value.However, if thiswerethe case,allfour lakesmightbe expected to exhibitsimilarCu2* concentrations, whichis not the case.Notetoo that in all our lakes,the freeCu2*concentration is bufferedby the DOMnaturallypresentin the samples. Finally, WHAMVl consistently over-predicted in allthe lakeswherethe the free Ni2*concentration ambientconcentration wasabovethe methoddetectionlimit (FigureLc),i.e.the oppositeresultfrom that with Cu,suggesting that the defaultequilibrium for the complexation of Ni to natural constants DOMaretoo low for the environmental conditions foundin our lakes.VanLaeret al. (2006)cameto a similarconclusion in theirstudyof the complexation surface of Ni by the DOMpresentin sixBelgian (40to 80 nM),the measuredfree-ionfraction,as waters.At their background Ni concentrations measured usingthe Donnanmembrane variedfrom 4 to 45%;WHAMVl overestimated technique, the free-ionproportionmorethan 2-fold,evenif it wasassumedthat allthe DOMwaspresentas fulvicacidandactivelyparticipating in Ni complexation. 138 100 A t 80 Tre I; 660 o E IL s40 T + L I t 1,1 TI 1 . r '" l: 7'' t': It lF T 7 *'' !,,: !F ' * a * i!: - 't' 7 i:. 20 120 z60 q) E IL s40 D .80 N o ,fl uo s u$*o%suj)"."$*"i.."]',"odu"o u$*o{.$..'$*"i".{"*0"*uuo Figure2. Comparison betweenmeasuredpercentfree Cd2*(A),Cu2'(B),Ni2*(C)and Zn2-1D1and WHAM predictedpercentfree Mz*for eightlakessampledfrom the regionsof Rouyn-Noranda, QCand (LOQ)are not shown. Sudbury,ON.Measuredpointsbelowthe methodlimit of quantification inputvalues. for a rangeof percentactivefulvicacid(%aFA) WHAMpredictions were calculated Lakeson the x-axisare arrangedfrom highpH on the left to low pH on the right.Solidcirclesare measuredo/ofree Mz*valueswith error barsrepresentstandarddeviationof triplicatemeasures. Greyverticalbarsrepresentthe rangeof WHAMpredicted%free Mz*valuesusing33%to t3O% aFA.Solidhorizontallinesare WHAMpredicted%free Mz*valuesusing65%aFA.Dashed horizontallinesrepresentthe LOQ. 139 Freemetolion percentages. A comparison wasalsomadebetweenmeasuredand modeled specific "percentfree metal"valuesto take into accountdifferences in the total metalconcentrations (LOQ)wascalculated percentfreemetallimitof quantification 2). A lake-specific to eachlake(Figure for eachlakebasedon threetimesthe standarddeviations of boththe of triplicatemeasurements freemetalandtotal metalconcentrations measured for eachlake.Anypercentfree metalvalue calculated for a lakethat yieldeda valuebelowits LOQwasdiscarded. ln additionto usingthe defaultassumption that 65%of the DOMis presentasfulvicacid(FA)andis activein metalcomplexation, we alsocalculateda rangeof modelpercentfree metalpredictionsby the %aFA halving(33o/o) arbitrarily anddoubline$3O%) thisvalue.lt shouldbe notedthat increasing to a valuehigherthan 100%simplyincreases the numberof fulvicacidbindingsitesinvolvedto a (pmolbindingsitesper unitcarbon)of the defaultfulvic valuethat exceeds the complexation capacity (Figure percentfree acid.In the presentation of the resultsof thesesimulations 2),the calculated metalis depictedas an envelope to the default of valuesfor eachlake:the middlevaluecorresponds valueobtainedwith the assumption to of 65%aFA,whereasthe upperandlowervaluescorrespond the percentfree metalcalculated asthe proportionsof DOMactive with 33%andI3O%,respectively, in metalcomplexation. Thebracketof upperand lower%freemetalvaluesincludedthe measured percentfree metalvaluesfor both Cdand Zn in most lakes(7 of 8). Thelakesfor whichthe measured valuefell outsidethe predictedenvelopewerethosewith the lowesttotal andfree metal concentrations. Evenwith the assumption WHAMVl of only33%aFA,i.e.,minimumcomplexation, stillpredicted a lowerpercentage valuefor the lakeswith the lowest freeCu2*thanthe measured (4 of 8). Similarly, totalCuconcentrations i.e.,maximum of L30%aFA, evenwith the assumption complexation, the measured valuesof percentfree Ni2*allfell belowthe minimumpredicted value. 3.5.2 Optimizationof WHAM Vl predictions Thechemical equilibrium simulations discussed to this pointwererunwith WHAMVl in its default mode,usingthe assumptions that wereoutlinedearlier.Amongthe adjustable WHAMinput parameters proportionof DOMactivein metalcomplexation are(i)the user-defined and(ii)the relativeproportions of humicandfulvicacids(FAand HA)makingup the DOMthat is activein metal 140 complexation.The latter parametervariesasa functionof the type of organicmatter,e.g.,whether it is derivedfrom soilwater,peatwater,lakewater,etc.(Bryanet al.,2002).The proportionof DOM that is activein metalcomplexation is usuallyunknownandmustbe estimatedby the userof WHAM. At this point,we retainedthe assumption that the DOMactivelyinvolvedin metalcomplexation asfulvicacid,but we removedanyconstrainton the "percentactiveFA"and couldbe represented for eachlakeandeachmetal,individually. Theoptimalpercentof FA ran multiplesimulations (%aFAool) requiredto exactlypredictthe measuredfree metal activelyinvolvedin metalcomplexation adjusting ion wascalculated by repeatedly runningWHAM/ModelVl, the FAconcentration untilthe predictedfree metalion concentration equalledthe observedvalue. A computercodewas usedto run WHAMandoptimisethe valuesof %aFA.The%aFAoot automatically wascalculated in this mannerfor eachmetalandfor eachlakeas o/aF\,= (mgL-1) (mgt 1). isthe optimisedFAconcentration where[FA]opt and [DOCIis the DOCconcentration fn a givenlakeandfor a particular metal,an optimalo/oaFAfor eachof the triplicatesamples was allthe measured calculated by findingthe FAconcentration that bestdescribed freeion (Table3). Thiswasdoneby adjusting for eachsamplesimultaneously concentrations the modelFA as before,and minimising concentration, the errorterm : ) ([MJ,0,,, [M].,",,)' freeion concentrations where[M]our, andcalculated for metali. i and [M]carc, i ar€the observed Byincreasing or decreasing numberof FAbindingsitesinvolvedin metalcomplexation the %oaFA,the is eitherincreased or decreased, whereasthe intrinsicmetalbindingaffinities at thesebindingsites ln otherwords,the ligandconcentration is adjusted.Ascanbe seenin Table3, the remainconstant. reflecting interoptimal%aFAvaluesobtainedin thismannervariedfrom laketo lake,presumably valuesall fell in the 8to 90%range, lakedifferences in the qualityof the DOM. ForCu2*the %aFAoo, 147 the notional Ni2*,the valueswereall muchhigher,oftenexceeding but for Cd2*,2n2* andparticularly limit of too%:6t to 250%for cd2*;65 to 4Lo%f or zn2t;44oto Lgoo%for Ni. 3.5.3 Possiblereasonsfor the divergencebetweenWHAMcalculationsand measuredmetal speciation Althoughincreasing the %aFA,andthusaugmenting the numberof bindingsiteson the FA(as thisvalueup to three in the preceding section), is an acceptable increasing described adjustment, that in Thisresultsuggests higherthanthe defaultvalueis clearlyunrealistic. ordersof magnitude to adjust additionto adjustingthe bindingcapacitvof the fulvicacid,it wilt tiketyalsobe necessary Ni. Theremaywell be natural metalbindingaffinities(Kvd (Tipping,2002) for Cd,Zn and especially ligandspresentin the lakewaterthat havea higheraffinityfor Ni thandoesfulvicacid. Further to determinethe bindingconstants experiments of thesemetalswith naturalDOM,at realistic metalconcentrations, environmentally areclearly needed. couldalso Discrepancies betweenWHAMpredictions free-metal ion concentrations and measured Forexample, ariseif ligands otherthanfulvicand humicacidswereinvolvedin metalcomplexation. surfacewatersand Bakenet al. (2011)detectedamino-polycarboxylate anionsin someurbanBelgian for the greaterthan expected suggested that theseligandsof anthropogenic originwere responsible (asseenherefor Ni). However, seems metalcomplexation in thesesamples suchan explanation of urban unlikelyin the presentcase,giventhe remotelocationof our lakesandthe absence wastewaterinputs. r42 Table3. (QC)and Sudbury lakessampledfrom Rouyn-Noranda Mean(t SD,n=3)calculated %oaFAorrfor ( O N ) i n2 0 0 8 . YoaFAop. cd Cu Ni Zn (DS) Dasserat 93141 80r12 140Oa 95r33 Dufault(DU) 61r 3 5618 NAO 65r21 (OP) Opasatica 7 5! 5 1 6 13 NA 3L0t 33 Vaudray(VA) 92!1.4 32!7 NA 9s190 Bethel(be) NA 9!2 1900r 130 1_90 t 36 Geneva(ge) t40' 8r1 890r 170 290!120 Raft(ra) 2 5 0! 2 3 86r8 440!74 4LO!4 Whitson(wh) 140t 31 90 t 8 580t 75 180t 51 [ake (code] (LOQ calculated asthreetimesthe standarddeviationof " indicatesonlyone sampleabovelimit of quantification triplicateanalyses of eachlake). b NA indicates valuesnot available. 3.5 Relotionshipsbetween %aFAor,and DOM spectroscopicproperties we exploredpossiblerelationships in %aFAspe between Giventhe markedlake-to-lake differences of the DOMin the originallakewatersamples.lnitiallywe thisvalueandthe opticalproperties values,giventhe promising resultsobtainedby soilscientists lookedfor relationships with SUVAzsq in soilsolutions(Ameryet al.,2008;Cornuet al.,2011).However, workingon CuandCdspeciation (see plotsof YoaFAool indexfailedto yieldusefulrelationships or the fluorescence againstSUVAzs+ Figs.55 and56 in the Supplementary Material).Ourinterestthenturnedto the fluorescence spectrum, to determineif it couldbe usedto estimatethe proportionof DOMthat is activein metal fluorescence of eachof the four PARAFAC complexation.We plottedthe relativefluorescence to freemetalconcentration againstthe %aFAneededto forcethe WHAM-calculated components (%aFAoor) (seeFigs.57 and58 in the Supplementary free metalionconcentration equalthemeasured L43 for the four metalsemerged relationships with %aFAopt Material).Themoststatistically significant for CUC1andC3lCr(Figure3). In seeking suchrelationships, we are usingthe fluorescence matrixas a "proxy"for the compositionof the DOM. Although excitation-emission probethe functionalgroupsor the exactsitesinvolvedin spectrofluorimetry doesnot necessarily in the metalcomplexation by the DOM,it is nevertheless a usefulandeasywayto trackchanges qualityof DOMin naturalaquaticsamples (Fellman et al.,2010;Muelleret al.,2012b). (y=380e(-6'4x)) (P<O.OOOL) relationship ForCd,the strongest(R2=0.79) and moststatistically significant significant emergedbetweenyoFAoptand the relativeproportionof C3(Figure3e). A statistically (P<O.OO1) positivelinearrelationship (y=1100x-260; R2=O.621was found betweenCuand CL/CI (y=-20 000x+6800) (Figure3b). ForNi a strong(R2=0.96) linearrelationship and significant(P<0.001) (P=0.01) but statistically significant wasobservedwith Ct/Cr (Figure3c). Finally,a weak(R2=0.33), (y=560e(-s'3x)) 3h). relationship decreasing exponential wasfoundbetweenZn andC3/q (Figure 144 I E () 250 I € zoo o N 200 150 150 100 100 E o o < L 6 sso 010 030 r ar(b t V V V o80 (D 020 030 D tr t\ ^^ N v w c = o40 AA I o S20 ot "€ 030 032 034 2000 z "P rsoo () N E Eo 1000 o LL o s s00 024 030 034 o soo 015 020 025 a o N E 6 200 010 t 400 tr N € o 005 oo I I s F - AV 100 'A o24 0 26 028 0 C1/C1 Figure3. G3/C1 Relationship betweenthe WHAMcalculatedpercentfulvicacidactivein complexation (%aFAopl;.;,"6) for Cd, Cu,Ni andZn andthe relativecontributionof PARAFAC fluorescence (panels (panels components 1 a d) and3 e h) to the overallfluorescence spectrumof lake DOM.Pointsof the sameshapeandfill representreplicatesamples.Solidlinesrepresent regression equationsof statistical significance at the 0.05alphalevel. 145 ln termsof the ubiquitous humic-like DOMfluorescence component(C1),the positivelinear relationships between%oaFAorl and C1/Crfor Cdand Cu (Figures 3a and 3b) suggestthat with an in the relativecontribution increase the proportionof of thiscomponentin the lakewater samples, the DOMthat is involvedin CdandCucomplexation In contrast, a negativelinear increases. relationship wasfoundfor Ni (Figure in the bindingcapacity of the DOMfor 3c),indicating a decrease Ni. With respectto the allochthonous-like DOMcomponent(C3),the negative exponential relationships between%aFAoo, andC3/Crfor CdandZn (Figures 3e and3h)suggest that with an increase in the allochthonous signature of the DOMin the lakewater samples, the proportionof the DOMthat is involvedin CdandZn complexation In contrast,a weakbut positive decreases. relationship with the C3/q ratiowasfoundfor Ni - asthe allochthonous signature of the DOMin the lakesamples decreases, the bindingactivityof the DOMfor Ni increases. Thissingularbehaviour of plotsfor the Ni,as demonstrated by the contrastbetweenFigures 3c and39 andthe comparable othermetals,suggests that the DOMbindingsitesactivein Cd,CuandZn complexation aredifferent from thoseinvolvedin Ni complexation. To our knowledge, thisis the firsttime that sucha hasbeenresolved distinction in naturalwatersamples. Admittedly the relationships shownin Figure3 will haveto be exploredovera widerrangeof DOM samples, but theydo supportour originalideathat the proportionof DOMactivein metal complexation couldbe estimated into on the basisof itsfluorescence andthen introduced signature chemical equilibrium modelssuchasWHAM.To testthisidea,we incorporated eachlake'soptimized %aFAvalue,ascalculated fromthe relationship between%aFA optimized andC1/Cror C3/Gfor each metalandlake(i.e.,the regressions shownin Figure3), intoWHAMandre-ranthe chemical equilibrium simulations to predictthe free M2*concentrations; thiswasdonefor eachmetal,since the optimized %aFAvalueis metal-specific. Figure4 depictsthe comparison betweenthese "improved"WHAMcalculations to of free M2*andthe measured free M2*concentrations. Compared the resultspresented in Figure1, the greatestimprovement usingthe CL/Crratiowasfoundfor nickel,for whichthe modeledfree Ni2*decreased from the to withinat mosta factorof 1.3difference measured free Ni2*concentrations. ForCd2*andZn2*,metalsfor whichthe originalWHAM predictions werealreadygood,no improvement wasnoted(indeedthe agreement wasslightlyworse t46 thanwith the arbitraryvalueof 65%activefulvicacid).Althoughthe WHAMcalculation of freeNi2*in our lakeswasimprovedby estimating the proportionof fulvicacidsactivein Ni complexation from qualityof DOM,the estimated%aFAwas the spectroscopic stillunrealistically high,as mentioned earlier.Nevertheless, the ideaof usingthe spectroscopic qualityof DOMas a proxyfor the proportionof DoM that is activein metalcomplexation remainspromising. 4 Conclusions Themaingoalof thisstudywasto explorethe possibility of incorporating a measureof the qualityof DOMinto chemical speciation models,suchasWHAM,to improvetheirpredictions of tracemetal speciation in naturalaquaticsystems.ForCd,Cu,Ni andZn,we haveshownthat suchan approach is feasible,usingsimplefluorescence measurements on lakewaterDOMto estimatethe proportionof DOMthat is activelyinvolvedin metalcomplexation (aswasdonein generating Figure4). However, our resultshavealsoindicated that in additionto adjustments to the metal-binding capacity of the DOM,it mayalsobe necessary to adjustthe affinityconstantsthat are usedby WHAMto calculate the speciationof somemetals,especially Ni;the possibility of usingthe spectroscopic propertiesof DOMto predictits metal-binding affinityshouldbe explored.Thepossible involvement of ligands otherthanfulvicand humicacidsin metalcomplexation shouldalsobe considered, particularly for samples with low totaltracemetalconcentrations. lmprovedtracemetalspeciation predictions, especially at low,environmentally significant metalconcentrations, will be importantnot onlyfor geochemical studies,but alsofor the prediction of chronictoxicityin naturalaquaticsystems (e.g.,as partof modelssuchasthe BioticLigandModelor BLM(Slaveykova andWilkinson, 2005)). t47 1e-8 (a) ^ 1e-9 (b) '/" e 3 /o, ./. + *= 1e-9 () & E O ! E -o $ te-to o ! E o) o.Pz o .N .N {-1 E 1e-11 Oo / o Dsry ,/ c) o .9 te-to z. E ar) qu,irt*n /t E E- te-tt /r o / -l ,//'FA-rvA bd$ | ,rl /t oi lQ se OP (, ln" l 1e-12 1e.10 1e-9 1e-6 e N f c N a 1.0e-7 L z u 1e-7 o q) E o E o S 1e-8 .N E ! o .F o o o - o c) 'F '1 .0e-8 c o 1e-9 co () ' 1.0e-9 1e-9 1e8 1e-7 MeasuredMt. (M) Figure4. 1e-6 1.0e-8 1.0e-7 Measured M'. (M) with WHAM of the free metalion [Mz*]calculated of mean(t SD,n=3)concentrations Comparison IET(Cd,Ni and (withthe %aFAestimatedfrom Figure3) andthosemeasured(t SD,n=3)usingthe the 1:1|ine' Thedashedlinerepresents Zn)or |SE(Cu)methods. L48 Acknowledgements providedby M.G.Bordeleau, Theauthorsacknowledge assistance the technical S.Duval,P.Fournier andJ. Perreault in the laboratory in the field. Sampling and P.GirardandA. Cr6mazy wasgreatly facilitatedby the personnelof the LaurentianUniversityCooperative Freshwater EcologyUnit,led by J.Gunn,in the Sudburyareaand by L.Jourdainof the Ministiredes Ressources naturelleset de lo Faunedu Qudbecin the Rouyn-Noranda datafor the Sudburyarealakes area. Accessto limnological was providedby StephenMonetof the Cityof GreaterSudbury.Financial supportwas providedby Research Councilof Canada and by the Metalsin the Human the NaturalSciences and Engineering Environment Research P.G.C. Campbell andC.Fortinaresupportedby Network(www.mithe-rn.ore). the CanadaResearch ChairProgram.S.Loftsacknowledges the supportof the UKNatural Research Environment Council. 149 5 5.1 SupplementaryMaterial (IET)calibrationresults lon ExchangeTechnique Thecalibrationof the IETfor Cd2*,Ni2*andZn2*wasperformedwith standardsolutionsin a pH range of 5.29to 8.86andwith knownconcentrations of freecd2*(0.24to 405 nM),Ni2*(12.8to L750nM) (I)for Cd,Ni in the distribution coefficient andZnz*(4.86to L310nM). FigureSl depictsthe changes andZn asa functionof the free metalconcentration. stable.Thelt weregenerally Acrossthis pH andconcentration range,the distribution coefficients at the highestpH valueof 8.86(closedsquares in Figure5L),however,aremuch valuescalculated l, values.At highpHvalues,metalhydroxocomplexes tendto interact higherthanthe remaining andincreasing the h valuesfor all with the resin(Fortin et al.,2010),creatinga positiveinterference pH range(5.29to 7.88)the l, valuescanbe considered to be threemetals.Forthe remaining range(N=56, constant.ForNi,the hvalueswerefoundto be stableacrossthe wholeconcentration (0.24-O.26 ForCdand Zn,the lowestconcentration nM for Cdand4.86-5.42nM for Zn) CV<L8%1. introduced a considerable amountof variability in the I values.At the lowestfreemetal (strong)ligandmayhavecomplexed the metal, traceamountsof an unidentified concentrations, yieldingan artificiallylow valuefor 1,.Theh valuesfor Cdand Zn were,however,foundto be stable and 14.6to acrossthe free metalconcentration of rangeof 4.08to 405 nM Cd2*(N=46,Cv<23%) of the free-metalion 1310nM Zn2*1N=46,CV<94%) and thesevalueswere usedfor the calculation concentrations. 150 14 tz 'o) 1 0 T 8 E O o 4 oo z s 0 oP 6gu9 \gu$ \Fu$ ?gos abrs tgu$ o5u$ rgu$ tcd'z.l(M) 14 12 10 o) 2 8 b e e{oae 4 2 o 0 sS ugu$ os.$ sgu$ nlux nb"x n9"x l.\ux tNi'z.l(M) 25 20 'g) E N 15 10 u g %"e@ 5 0 g9 zguo lgu$ ogu$ ugu$ ngux ny/ n.o"x 1zn2.11tvt; Figure51. (l\)for Cd,Ni andZn asa functionof free metalconcentration. IETdistributioncoefficients Open circlesrepresenttr calculated at pH 5.29to 7.88,whereasclosedsquareswere calculated at pH 8.86, 151 5.2 lon ExchangeTechnique(lET)calibrationresults(Rachouet al., 2007) freeCu2* basedon measured Althoughthe calibration curve(Figure 52)wascalculated as low as 4.5x 10-1s M, a methoddetectionlimit of 2.6x LO-llM Cuz*wascalculated concentrations methodblankmeasurements. of sixreplicate basedon the threetimesthe standarddeviation pcu 15 13 11 a 10 -40 -90 g -140 lu -190 -240 y=-29x+112 R2= 0.99 -290 -340 Figure52. ISEpotentialresponseasa functionof Cuactivity(pCu).N=66. L52 Relationships betweenpercentfree metal ond pH and DOCconcentrations 5.3 Forallfour metals(Cd,Cu,Ni andZn),andin boththe Rouyn-Noranda andSudburyregions, the percentfreemetalasdetermined by the IETand ISEmethodstendedto decrease asthe lakepH pointingto complexation increased, competitionby protonsasan importantfactoraffectingmetal speciation. T + N o o gl Ir N T- 60 E l! s o3 o' I IL x- 2 tr..i 20 80 120 T + qL 100 *80 I z o,^ E*' u- s C N 90u IL s i 40 TT a 6.0 6.4 6.8 1-L 7.2 pH Figure53. /.b 6.4 I.b 8.0 pH Measured%free metal(tSD,n=3)asa functionof lakepH. 153 percentfree metalwasalsoobserved for Cd,Cuand Ni asthe DOC A generaldeclinein the measured presumably metalbindingsiteson the in available increased, reflecting an increase concentration ratio. DOMwith an increasein the ligand-to-metal I I.T + N E o R 60 g o o? o- {I T l- s 1 N 1 tL TI LL t{ >s -2 t^ i 20 rG.I 80 120 100 *80 sz N o,^ g*t o^^ EbU u- tL s s I 40 TTT I I I t I't-t I' l. -T1T1 a a 54JbI DOC(m9 C L-1) IT {l 3456 DOC (mg C L-1) Figure 54. Measured n=3)asa functionof lakeDOCconcentration. %freemetal(+SD, 5.4 Relationships betweenthe optimized%oFAoos and DOMopticalproperties below Themeasured opticalqualityof the DOMin the lakessampledfor thisstudyis presented (Table51)andis described (Muelleret al.,201-2b). in detailelsewhere L54 Table51. Lake Measured(+SD,n=3)DOMopticalquality(SUVA2'4, Fluorescence Index(Fl)and the relative proportionof PARAFAC fluorescence components1 to 4)for lakessampledfrom Rouyn-Noranda (QC)andSudbury(ON)in 2008. SUVAzsa Fl ctlcr czlcr c3lcr c4lcl (L m-t mg c-1) Dasserat Dufault 7 . 8t 0 . 3 7.3! O.4 Opasatica 7.5!0.4 Vaudray L 0 .0r 0 .3 Bethel Geneva Raft Whitson 5 .0r 0 .3 5.3r 0.2 5.8r 0.8 7 .31 0 .5 0.384r 0.34r 0.0L 0.325r 0 . 2 8 1 0 . 0 2 0.254! 4.246! 0.380r 0.30' 0.39' 0.32L! 0.384r 0.3L0r 0.385r 1 . 3 3 1 0 . 0 2 0.287! 1.301r 0.28r 0.01 1.3611 0.27Lt L .2 7IO.OL 1 .5 1 r0 .0 4 1.30' 1.30t 0.09 1 -.3r60 .0 3 0.2L6r 0.25r 0.03 0.29r 0.0L 0.40r 0.02 0.L67t o.Lla 0.08r 0.0L 0.161! 0.114r 0.133t 0.116r 0.068r 0.2061 O.r7^ 0.218t 0.1.44! " indicatesonlyone replicatesamplemeasured. We exploredpossible relationships betweenthe optimalpercentof FAactivelyinvolvedin metal (%aFAool) complexation of the DOMin the lakewatersamples, in orderto andthe opticalproperties estimatethe proportionof DOMthat is activein metalcomplexation.The moststatistically for the four metalsemergedfor CL/G and C3/Cr(seeFigure3 in significantrelationships with %aFA6p1 the manuscript). bivariate relationships Thefollowingfigures(S5to 58)showthe lackof significant between%oaFAopl and the other opticalproperties. 155 120 t I t' lC O V \9 2oo o OAn 150 Vn Y A AA V tr^ L LL o s % q) .N E60 .N E lY r AA o s ooal AA 50 20 a^ ^ao ovv 10 "t800 z x 1600 ! 11 c N 400 ! soo a raoo .N E o o >< - a .N .E 12oo t a 1000 8oo ogvv A LL 600 +v 400 200 V Y " 100 5^ A An '4. 200 4 56789456789't0 suvA2s4 Figure55. 11 suvA2s4 (%aFAoo,) Relationship betweenthe WHAMcalculatedpercentfulvicacidactivein complexation (SUVA254) for Cd,Cu,Ni andZn andthe specificUVabsorbance of lakeDOM.Pointsof the same shapeandfill representreplicatesamples. 156 120 300 100 250 o () o 1t o .N 200 980 v E 150 V o. o II o s Vv tr 100 A^ o o ,s B AA 20 a aa 0 U 1;15 tr A s 4$ Y I E c) .N 'at b u LL (! 't 50 lV o V I 1.20 1.30 1.25 1.35 1.40 1.1 1.45 1.2 1.3 1.4 1.5 1.6 2000 z I 800 o 1600 E N o E 1l 400 o 1400 .N E 1200 oo 1000 $ soo t! LL C' s o a o .N .E o o 800 v 600 _tr 100 Y t 400 ^' I o o '|1 200 I A A 200 1.2 1.3 1.4 FI Figure56. t.c 1.6 1.1 1.2 1? 1.4 t.c FI percentfulvicacidactivein complexation (%aFAoo,) Relationship betweenthe WHAMcalculated index(Fl)of lakeDOM.Pointsof the sameshapeand for Cd,Cu,Ni andZn andthe fluorescence fill representreplicatesamples. r57 120 I I io o b I r rv L e 200 Oan o .N E 150 o o V V o VV tr IL (5 A 'l (! s tr E q) .N EOU .E AA s tr O aa N vv 0 0.24 0.28 0.32 0.36 0.40 0 0.24 0.44 0.28 0.32 0.36 0.40 500 z E 1800 c N o 1600 8 r+oo V o a a $ eoo .N .N .E 12oo E o o 200 CL o looo k s- 1l 400 8oo €v LL G' - 600 E 100 vl 400 200 0.37 0.38 I 0.38 0.39 C2ten 0.39 0.40 0.40 0 0.24 0.28 0.32 0.36 0.40 0.44 c2tcr (%aFA"o1) percentfulvicacidactivein complexation Figure57. Relationship betweenthe WHAMcalculated for Cd,Cu,Ni andZn andthe relativecontributionof PARAFAC fluorescence component2 to the overallfluorescence spectrumof lakewaterDOM.Pointsof the sameshapeandfill represent replicatesamples. 158 120 I 250 o I B 2oo E €80 o o .N .N o .E 150 tE s -o .F 0.04 A u A A, LL (U s A ft- 50 0.08 0.12 v v o oo ! tr ^. t IL Y I a a 0.16 0.20 0.24 0.04 0.08 0.12 0.16 0.20 0.24 2200 2000 1800 c N o z .E 1600 E i4oo .N .E 12oo o TL (s s' a a .N E o voo 200 Ll- n o OUU Y Y V f,LY 100 A lr 400 I 200 0.10 0.12 0.14 0.16 0.18 0.20 0.22 0.24 c4tcr Figure58. a $ soo 1000 600 lg 400 0.04 0.08 0.12 0.16 0.20 0.24 c4lcr (%aFAool) Relationship betweenthe WHAMcalculatedpercentfulvicacidactivein complexation for Cd,Cu,Ni andZn and the relativecontributionof PARAFAC fluorescence component4 to the overallfluorescence spectrumof lakewaterDOM.Pointsof the sameshapeandfill represent replicatesamples. L59 Article3 Nickeland coppercomplexation by naturaldissolved organicmatterin lake KristinK.Mueller,ClaudeFortinand PeterG.C.Campbell INRSEauTerreet Environnement, du Qu6bec, Universit6 490 de la Couronne, Qu6bec (Qu6bec), GlK 9A9 Canada Abstract TheWindermere HumicAqueousModel(WHAM),usedto estimatethe bindingof cationsby usingthe resultsof is calibrated dissolved organicmatter(DOM)in naturalaquaticsystems, titrationexperiments on isolatedandpurifiedhumicandfulvicacids,at DOMandtotalmetal concentrations that areoftenunrealistic for naturalaquaticsystems.Usingwell-characterized conditions, that the naturalwatersamples, we demonstrate, for realistic environmental for the complexation conditional bindingparameters of Ni andCuwith naturalDOMvaryasa qualityof the DOM. Watersfrom two lakeswith contrasting functionof the spectroscopic bindingparameters typesof DOMweretitratedwith CuandNi andthe calculated conditional were comparedbetweenlakesand betweenmetals,andthe titrationcurvesfor eachlakewere comparedto thosepredictedby the off-the-shelf versionof WHAM6.1-.Bindingaffinitiesand capacities of DOMfor Ni and Cuwerefoundto be significantly differentnot only between the amountof DOMlakes,but alsobetweenmetals.WHAMwasfoundto underestimate the titrationresults boundNi,whilethe oppositetrendwasfoundfor DOM-bound Cu. Overall, sourcesmayhavea suggest that the morearomatichumic-like DOMfrom allochthonous DOM highercomplexation significantly affinityandcapacity for Cuthanthe moreprotein-like that Ni from autochthonous whereasthe oppositetrendwasfoundfor Ni,indicating sources, andCuare bindingto differenttypesof bindingsiteswithinthe DOMmatrix. L62 1 lntroduction organicmatter(DOM),an important Thecomplexation of tracemetalsby naturaldissolved chemicalreactivity, organiccomplexing agentin mostnaturalwaters,affectstheir speciation, HumicAqueous models,suchasthe Windermere and bioavailability. Chemical equilibrium of metalsby DOMin natural Model(WHAM),areoftenusedto predictthe complexation equilibrium reactions. The thermodynamic aquaticsystems basedon established heterogeneous natureof DOM,however,makesit difficultto assignspecificthermodynamic assumptions basedon with the useof somereasonable datato its complexes. Nevertheless, of cationswith DOMhas currentknowledgeof the characteristics of DOM,the complexation (Tipping,2002;Tippinget al.,ZO11-1. beenmodeledwith somesuccess gainedfrom laboratory studieson TheWHAMmodelis parameterised basedon information of metalsto organic isolatedhumicsubstances, but it is often usedto predictthe complexation havebeenfoundbetweenthe matterin naturalwatersamples.Unfortunately, discrepancies for Cuand Ni,andthat measured WHAM-predicted in speciation of somemetals,especially (Fortinet al.,2010;Muelleret al.,2OL2c; Unsworthet al.,2006;Van naturalfreshwater samples andmodeledfreemetal Laeret al.,2006).Thediscrepancies betweenmeasured in metal-binding concentrations in naturalfreshwater systemsmaybe dueto differences usedto character betweenthe DOMin naturalwatersandthe isolatedhumicsubstances parameterise the model(LoftsandTipping,zOtL). with Cuand Ni andcompared ln the presentstudy,we havetitratedtwo lakewatersamples the resultingtitrationcurveswith thosepredictedby WHAM. Thetwo lakeswere chosen and because their because earlierwork hadshownthat their DOMdifferedspectroscopically, ratios). ambientmetallevelswerelow (thusallowingusto explorea wide rangeof metal:DOM propertiesof two contrasting Theobjectiveof the work wasto comparethe metal-binding for the complexation of Cuand Ni with the typesof DOM. Conditional bindingparameters naturalDOMhavebeencalculated, comparedbetweenlakesand betweenmetals,and related to the opticalqualityof the lakeDOM. 163 2 2.1 Materialsand Methods Sitedescriptionand sompling TheCanadian areasof Rouyn-Noranda, Ontariohavebeenimpactedby Qu6becandSudbury, pastand presentcopper/nickel mining,smeltingand refiningactivities.A detaileddescription of thesestudyareascanbe foundin Muelleret al. (2012b).Two lakeswith low total metal (48"04'32"N; were selectedfor the titrationexperiments: concentrations LakeOpasatica 79"L7'49"W) from the Rouyn-Noranda regionand LakeGeneva(79"17'49"W;8L"32'46"W) from the Sudburyregion.Basedon the findingsreportedby Muelleret al. (2012b;2012c),Lake pH (7.47), hasa moderateconcentration Opasatica of DOC(6.0t 0.4 mg.L-1), a near-neutral very low total dissolved concentrations of Ni and low concentrations of Cd,Cuand Zn (see pH Table1). LakeGenevahasa low concentration a near-neutral of DOC(3.4t 0.4mg.L-1), (6.89),andthe totaldissolved concentrations of Cd,Cu,Ni andZn areallverylow (seeTable1). qualityof DOMbetweenthe lakeswere alsoreported(Mueller in the spectroscopic Differences et al.,2012b). Grabsamplesweretakenfrom the littoralzonesof LakesOpasatica and Genevausingtwo 4-L polypropylene (l-CHEM amberglassjugswith Teflon@-lined, Brand)for eachlakein closures JuneandJulyof 2009,respectively. Thesamples werethentransported backto the laboratory within24 h andvacuumfilteredundera cleanlaminarflow hoodusingpolycarbonate filter holders(Advantec MFS,Inc.)anda seriesof 47-mmpolycarbonate filters{.5,2,O.4 and0.2 pm; MilliporeandAMD Manufacturing Inc.).Beforefilteringthe lakewatersamples, 1 L of ultrapurewateranda smallvolumeof samplewerefilteredthroughthe filtrationapparatus Following anddiscarded. filtration,the samples werestoredin the darkandat 4"Cuntil (storage analysis time < 72 d). 2.2 Analyses (v/v) All plasticware usedfor sampling andsampleanalyses wassoakedin a solutionof 1:O% nitricacidfor at least24 h, rinsedseveraltimes with ultrapurewater(>18Mohms.cm) and driedundera laminarflow hood. Polystyrene for anion vialsusedfor storingthe subsamples 1.64 analyses were,however,rinsedthreetimeswith ultrapurewateronly.All glassware wassoaked in a solutionof 2 N HCIand rinsedwith ultrapurewater. Following samplefiltration,subsamples weretakenundera laminarflow hoodto measure the ambientwaterchemistry for LakesOpasatica andGeneva.Subsamples (10-mL) were transferredto polypropylene vialsand acidifiedto 2o/o (v/v) nitricacid(ultra-tracenitricacid, BDHAristar)foranalysis of majorcations(Ca,Mg, Na,K)by atomicemission spectroscopy (VistaAXCCDSimultaneous lcP-AEs, Varian)andtotaltracemetals(Cd,Cu,Ni,Zn)by inductively coupledplasmamassspectrometry (lCP-MS, ThermoElementX Series).Similar 10-mLsubsamples were alsotransferredto polystyrene vials(VWRlnternational)formajor anions(cl, N03,Soa)analysisby ion chromatography (lcS-2000lon chromatography system, Dionex), and20-mLsubsamples weretransferred to amberglassvials(l-cHEMBrand)for total organiccarbonanalysis (TOC-VCpH, Shimadzu). Finally, L-mLsubsamples weretransferred to 3-mLglassvacutainers (KendalMonojet)for total inorganiccarbonanalysisby gas chromatography (3800Varian). Fordetailson the calibration and measurement of the free Ni2*concentrations usingan ion exchange technique(lET)andthe freecu2*concentrations usinga cupricion selective electrode (lSE), the readeris referredto a previouspaperby Muelleret al. (2012c).Briefly,when using the lET,the free Niz*ion in a lakewatersamplewasequilibratedin a flow-through systemwith a cationexchange resin(Dowex50w-X8,50-100mesh)thathadbeencalibrated usingstandard solutionswith a pH rangeof 5.3to 7.9and with a knownconcentration of free Ni2*ranging from L3 nM to 1.8pM (seesupplementary materialin Muelleret al. (2012c)). The concentration of Cain the calibration andnaturallakesamples wasadjustedusinga stock solutionof 25 mM Ca(Nog)z (99%,Sigma-Aldrich) to a finalconcentration of 0.46mM in order to attaina constantionicstrengthamongsamples andto ensurethat the time neededto reach equilibrium betweenthe resinandthe samplewassimilarfor all samples.Following the elution of the resinwith an acidicsolutionanddetermination of the amountof Ni that had been retainedby the resin,the concentration of freeNi2*in the originalunknownsamplewasthen 165 Forthe (r) for Ni (Fortinandcampbell'1998)' coefficient usingthe resindistribution calculated potential in the lakewater samples'the electrode Cu2* of concentration the of analysis rangingfrom the method overa free cu2*concentration calibrated was lsE, cu the of response of 0'03 deviationof sixrepricatebrankmeasurements) detectionrimit(threetimesthe standard material of Rachouet al' (2007) seethe supplementary method the pM following 63 to nM wereamendedwith KNOg andlakesamples Thecalibration for Muelleret al. (2012c)). wasaddedto achievea final amountof a L.23M stocksolut'ton known a lgg.gg9%,Fluka); and a simirarelectrode thusensurea constantionicstrength and KNOg mM r.0 of concentration analysed' amongall samples response 2.3 Titrationexperimentolset-up and Ni' Forthe cu were intrividuatytitratedwith cu Geneva and opasatica Lakes from sampres Plasmacal) of 0'2 mM coppernitrate(cu(NOr)z' titrations,knownamountsof a stocksolution which BDFalcon)' container' the sample(polypropylene wereaddedto eight200-mLaliquotsof to the rraturallake pH of the titrationaliquotswasadjusted The KNog. with amended been had wereleftto equilibrate of eitherHNOaor NaOH,andthe solutions pH usingsmallvolumes the freeCu2* in the darkbeforemeasuring overnightundera laminarflow hoodand 24 time of approximately An equilibration usinga cupricion selectiveelectrode' concentration hourswasconsideredaconservativeestimateofthetimeneededfortheaddedmetaltoreach by Tipping(2002)'ThepH of presented values literature on based DoM, the with equilibrium to the and re-adjusted the free Cu2*concentration the aliquotswascheckedbeforemeasuring analysed' aliquotwith no addedcu wasalso sample initial An pH necessary. if lake natural After of 27 cutitrationariquotsanarysed' cu titrationswere carriedout for a totar Tripricate of eachtitrationaliquotwere subsamples of the free Cu2*concentration, the measurement takenforthemeasurementoftotaldissolvedcu(5mLina13-mLpolypropylenevial'sarstedt) Brand)' (10mLin a 40-mLamberglassvial,l-cHEM and Doc concentrations (Ni(NO3)2' a stocksolutionof 1 mM nickelnitrate Forthe Ni titrations,knownamountsof container'BD sample(polypropylene plasmacal) wereaddedto seven220-mLaliquotsof 166 Falcon), whichhadbeenamendedwith Ca(NO3)2. Again,the pH of the titrationaliquotswas adjustedto the naturallakepH andthe solutions wereleftto equilibrate overnightbefore measuring the free Ni2*concentration usingthe IET(thepH wasre-adjusted to the naturallake just priorto IETanalysis). pH if necessary An initialsamplealiquotwith no addedNi wasalso analysed. Triplicate Ni titrationswerecarriedout for a totalof 24 Ni titrationaliquotsanalysed. In the samemannerasfor the Cutitrations, subsamples of eachtitrationaliquotweretakenfor the measurement of total dissolved Ni (5 mLin a 13-mLpolypropylene vial,Sarstedt) and DOC (10mLin a 40-mLamberglassvial,l-CHEM concentrations Brand). 2.4 Calculations Triplicatetitrationscurvesof measuredfree Ni2*and Cu2*asa functionof addedtotal metal weregenerated for eachlakeandfor eachmetal.Bindingparameters werecalculated from eachreplicatetitrationcurveand then the averages of thesetriplicatebindingparameters were calculated alongwith theirassociated variability. Theequilibrium formationof a 1:1complexes betweenmetal(tvlz*) anda two-siteligand(Li),with an associated conditionalstability (K;)wasassumed. constants M2++Li=MLi xi=ffi (1) The bindingparameters, Lt, K7,Lzand K2,w€r€ calculated from titrationcurvesusingthe followingtwo-siteLangmuir-type modelequation: IMu=ffi*m el where[ML]isthe molarconcentration (version11.0),nonof boundmetal. UsingSigmaPlot linearregressions wereperformedon triplicatetitrationsfor eachlakeandeachmetalfor a total of 12 titrationcurves. Forcomparison with the measured concentrations of boundmetal,[ML],in the sampledlakes, usingthe off-the-shelf versionof WHAM6.1and measured valuesfor [ML]wasalsocalculated pH,DOCandthetotalconcentrationsof Na,Mg, K,Ca,Ni, Cu,Zn,Cd,Cl,NO:,SOq,COg (measured astotal inorganic carbon)and F. We assumed that the DOC:DOM ratiowas2 t67 of metals(Bryanet al', (Buffle,1988),that 65%of the DoM wasactivein the complexation that that this activefractionwascomposedof fulvicacidsonly. we alsoassumed 2OO2l,and ascalculated of their hydroxides, by the solubility arecontrolled activities both Fe(lll)andAl(il1) givenby Loftset al. (2008)andTipping(2005),respectively' equations usingthe empirical (version1L.0)at the alphaequals0'05 in sigmaPlot wereall conducted analyses statistical level. confidence Resultsand Discussion 3 3.7 Lakewater and DOMchorocteristics andSudburyhavebeenimpactedby pastandpresentmetal Theregionsof Rouyn-Noranda givingriseto lakeswith a wide rangeof metalconcentrations miningand smeltingactivities, to theirlocationup-or down-windof the smelters.LakesOpasatica and pH valuesaccording andwerechosenfor theirnaturallylow total andGenevaare locatedupwindfrom the smelters (Table1). Thespectroscopic and differingDOMconcentrations dissolvedmetalconcentrations lake qualityof the DOMwasalsofoundto differbetweenthe two lakes(Muelleret al.,2012b); whereasthe sources, of allochthonous wasfoundto haveDOMlargelycharacteristic Opasatica a markedcontributionfrom autochthonous of LakeGenevaDOMsuggested characteristics metalconcentrations, with low ambienttotaldissolved sources.Bytitratingnaturalsamples by metalsandthe strongbindingsites,of low fewerDOMbindingsitesarealreadyoccupied pHvalues,LakeGenevais canbe probed.Althoughbdthlakeshavecircumneutral abundance, of major LakeGenevaalsohasa lowerconcentration slightlymoreacidicthanLakeOpasatica. metal(Cd'Cu'Ni cations(Ca,Mg, NaandK),majoranions(Cl,NOa,SOqand F),totaldissolved (Table1)' andZn)andDOCthan LakeOpasatica 3.2 Titrotion curves and with naturalligandsin LakesOpasatica Thetitrationcurvesfor Ni andCucomplexation ligandsin the of Ni andCuby inorganic Genevaare plottedin Figure1. Thecomplexation presence of DOMfor the lakessampledwasestimatedusingWHAM' ForLakeOpasatica' Ni and t.7-3.9%of Cuwasboundin carbonate' totaldissolved ely4.8-5.Z%of approximat 168 sulfateandchloridespecies.Althougha similarL.t-4.3%of totaldissolved Cuwasboundby theseinorganicligandsfor LakeGeneva,44-47%of Ni wascalculated to be presentas inorganic with theseligands.Therefore complexes the concentration of boundmetalto DOM,[ML],was to be equalto the total dissolved metalconcentration minusthe concentration assumed of free metalandthat boundto inorganic complexes. Fora giventotal dissolvedmetalconcentration, LakeGenevahada higherfree metal (Figure1). Because LakeGenevahasa lowerDOC concentration than LakeOpasatica presumably concentration, lessDOMis available to bindNi andCu. ln addition,at the (slightly) lower pH in LakeGeneva,the protonwould be expectedto competesomewhatmore for cationbindingsiteson the DOM.As Ni is knownto be moreweaklyboundto successfully DOMthan Cu(Buffle,1988),the effectof protoncompetitionwould be expectedto be greater for Ni than for Cu. Thiswas in fact observedfor both lakeswherethe concentration of free Ni wasgenerallyhigherthan free Cufor a giventotal metalconcentration.The presenceof strong Cu-binding siteson the DOMis suggested by Figure18,wherewith the first few additionsof total Curesultin littleincrease in the freeCu2*concentration. Thisbehaviour is characteristic of Cufillingup the strongDOMbindingsitesfirst,followedby the weakersites(Sundaand Hanson,19791. Alternatively, maybe a resultof the observedasymptoteat low total Cuconcentrations reaching the detectionlimitof the CuISEmethod.ln four of the eightlakessampledin 2008by Muelleret al. (20L2b), i.e.in thosewith the lowesttotalambientdissolved Cuconcentrations (Opasatica, Vaudray, freeCu2*concentrations werehigher BethelandGeneva), the measured that the measuredconcentrations were than thosepredictedby WHAM,possiblysuggesting affectedby dissolution of Cufrom the lSE.However,whenthe free Cu2*activityin solutionis M canbe achieved well bufferedwith organicligands,an ISEdetectionlimit as low as l-O-14 (Rachou et al.,2007).In the presentstudy,aswellasin the Muelleret al.(2012c)study,the in the calibration werebufferedwith iminodiacetic acid(lDA)and free Cu2*activities solutions with DOMin the naturallakewatersamples. 159 Table1. concentrations of majorcations,anionsandtrace metalsfor Lakes Totaldissolved conditions Opasatica andGenevaunderIETandISEexperimental LakeOpasatica LakeGeneva IET rsE IET ISE pH 7.69 7.77 7.02 6.97 Doc (melL) 6.92 7.47 3.36 3.54 N a( m M ) 0.13 0.13 0.027 0.027 Mg(mM) 0.11 0.11 0.025 0.025 K (mM) 0.023 1.0 0.0070 10 Ca(mM) 0.46 O.zt 0.46 0.055 Ni(nM) 16 16 22 23 Cu(nM) 39 38 tr 13 Zn (nM) 33 20 31 L2 cd (nM) 0.062 0.0s0 o.oL2 0.0s3 cl(pM) 8s 89 7.0 7.0 NOg(mM) 0.93 10 0.93 10 SO+(mM) 0.058 0.058 0.044 0.044 CO3(mM) 0.s2 0.52 0.05 0.05 2.6 2.6 2.8 2.8 F (pM) to the with WHAMVl andcompared Thenickelandcoppertitrationcurvesweresimulated i.e.,for low Ni-DOMratios,WHAM 2). Earlyin the Nititrations, observed values(Figure 2A and 2C),but for total in both lakes(Figures overestimated the free Ni2*concentration and greaterthanabout10-6M the agreement betweenobserved dissolved Ni concentrations predictedvaluesimproved.In the caseof Cu,the two lakewatersbehaveddifferently.ForLake throughoutthe titration(Figure the free Cu2*concentrations Geneva,WHAMunderestimated andthe ISE betweenWHAMpredictions the agreement 2B),whereasfor LakeOpasatica M Cu);for higher goodearlyin the titration(upto about10-6'7 wasreasonably measurements the free Cu2* Cuconcentrations, however,WHAMagainunderestimated total dissolved (Figure2D). concentrations 170 -5.5 -6.0 . I A Lake Opasatica Lake Geneva l a' l . r -6.5 I l.' + qt -7.0 lo z -7.5 I q, -9 a -8.0 I T -8.5 -9.0 t -9.5 -8.0 -7.5 -7.0 -6.0 -6.5 -5.5 log Ni1e1 I + I N J O (') o tr a a -7.5 -7.0 -6.5 -6.0 -5.5 logCu1s1 Figure1. (circles) Comparison of titrationcurvesfor Ni (A)and Cu (B),for LakeOpasatica and Lake Geneva(squares). L7t -5.5 r -6.0 v s!z E E -6.5 + B LakeGeneva WHAM modeled -7.0 I TT I t N f vl I o) -7.5 -9 rr' -7 + o-8 I o) o -9 -8.0 I -8.5 -10 -9.0 -8.0 -11 -7.5 -7.0 -6.5 -6.0 v I 4a oee a + o o) o) o 9 -s.o oA -9.0 -10.5 -9.5 -7.0 -6.5 -6.0 log Ni1s1 Figure2. -5.5 L I -9.5 -10.0 -11.0 -8.0 A 6 -9.0 -8.5 D t a -8.5 N 5 -7.5 0 -8.0 + qt z -7.5 3.3 -7.5 c LakeOpasatica WHAM modeled -7'o -5.5 log Cu1e1 -5.5 -6.5 -6.0 -6.5 -7.0 -7.5 log Ni1q1 . -6.0 a v v -8.0 -5.5 rl v 9v v ae3 -7.5 -7.0 -6.5 -6.0 -5.5 log Cu1e1 (AandB)andLake for LakeGeneva Nickel(AandC)andcopper(BandD)titrationcurves withWHAM (CandD). Measured calculated Opasatica values areshownin colour;values gray Vl areshownin triangles. Metolbindingporometers are Thecomplexation curvesfor Ni andCuwith naturalligandsin LakeGenevaandOpasatica presented canbe foundin Table in Figure3 andtheirassociated bindingparameters calculated model(Tipping,2OO2l. were derivedusinga two-siteLangmuir-type 2. The bindingparameters It shouldbe notedthat a successful fit usingtwo or more discretesitesdoesnot constitute 772 for the true existence of suchsites.As pointedout by Tipping,it simplymeansthat evidence ([DOM],pH,pM),the effectsof siteheterogeneity canbe withina givenrangeof conditions degree. accountedfor to someacceptable the concentrations of boundNi (M) in the titrations, Forthe rangeof free Ni concentrations of DOC(e/L)areslightlyhigherin LakeGenevathan in Lake normalized to the concentration (Figure34). Thecalculated reflectthe resultsof the titration bindingparameters Opasatica = 6.98)on DOMfrom LakeOpasatica is curves;the Ni bindingaffinityof the weaksites(LogK1,p; the higher(P= 0.009)thanfor LakeGeneva(LogKr,u= 6.30).In contrast, significantly = 34.2nmole.mgC t) is nearlyfourtimeslower(P= 0.01) of theseweaksites(Lr.rui concentration = 137nmole.mgC-1) (Table2). TheNi bindingaffinities for the strong thanin LakeGeneva(Lr,ni = L1.6and 11.7for LakesOpasatica different(LogK2,x; sitesin both lakesare not significantly respectively), andGeneva, althoughthereare nearlysixtimesless(P= 0.025)of thesebinding = 0.70nmole'mgC-1), (Lz,r..r to the DOMfrom compared sitesper unit DOCin LakeOpasatica betweenthe LakeGeneva(Lz,r,ri= 3.98nmole.mgC-1).Theseresultspointto the difference discussed with respectto their Ni-binding characteristics, DOMin LakesOpasatica andGeneva, furtherin section3.4. L73 LakeOpasatica LakeOpasaticamodeled -4.0 o o o -4.5 fz_ -uo o) 9 -s.s -6.0 -6.5 - t1 1 I -10 -9 -8 log Ni2+ -3.5 B -4,0 o 6 :I -4.5 f O 6o - s o a. C' a o -5.5 -11 -10 -9 -8 -7 -6 log Cu2+ Figure3. Comparison of complexation (circles) curvesfor Ni (A)andCu (B),for LakeOpasatica and LakeGeneva(squares). Themodelledcurveswere calculated with the two-siteLangmuir bindingmodel(seetext for details). 174 sitedensities(L;,pmole/gC) Conditional stabilityconstants(logK) and carbonnormalized and Genevamodeledusinga 2-site by DOMfor LakesOpasatica of Ni and Cucomplexation (P)valuesfor the non-linearregressions ranged significance Langmuirmodel.Thestatistical from 0.001to0.024. Table2. Lake Lake Opasatica Geneva 6.98r 0.12 6.30r 0.23 Lr,rui(nmole.mg C-t) 34.2!3.2 L37!24 Log Kz,r.1 1 1 . 6t 0 . 6 1 1 . 7t 0 . 5 0.7010.40 3 . 9 8r 0 . 1 1 7 . 5 9! 0 . L 2 7 . 6 3! 0 . 4 6 2 4 4! L O 75.6!19.4 9.46t 0.13 8.90r 0.04 4 4 . 5! 7 . 4 2 7 . O! 8 . 1 IET Log Kr,rui L2,x;(nmole.mg C-1) rsE Log K1,6, ( nmole.mgC-1) 11,6, Log K2,6, 12,6,(nmole.mgCt) than in Lake A higherconcentration of Cuwasboundpergramof DOCin LakeOpasatica (Figure3B). Furthermore, the for LakeGeneva, at allfreeCu2*concentrations Geneva, in the with an increase of boundCuappearsto tendtowardsan asymptote concentration indicating bindingsitesaturation.Althoughthe strengthof the of freeCu2*, concentration differentfor both lakes(Log bindingof Cuto weaksiteson the DOMwasnot significantly = 7.63for LakeGeneva), the numberof Lrsitesper and LogK1,6, Kr,c,= 7.59for LakeOpasatica (h,cu=244 threetimeshigher(P< 0.001)in LakeOpasatica unit DOCwasapproximately t). in LakeGeneva(Lr,c,= 75.6nmole.mgC Thereweresignificantly nmole.mgC-1)than = 9.46)andtwiceas many(P= 0.003)strongbindingCusitesper (P= 0.014)stronger(LogKz,cu = 44.5nmole.mgC-1) comparedto Lake on the DOMfrom LakeOpasatica unit DOC(Lz,cu = 27.0nmole.mgC-1).Clearly = 8.90dhd 12,6u theseresultsfor Ni andCu Geneva(LogK2,6u L75 demonstrate complexation that the DOMin LakeOpasatica differsfrom that in LakeGeneva, with respectto theiraffinityfor thesemetalsandtheircomplexation capacity.These differences are exploredin section3.4. We comparedour calculated free concentrations of boundmetalasa functionof the measured metalconcentration to thosepredictedby WHAM.Throughout the entireNi2*concentration range,the measured amountof boundNi in bothlakesis muchhigherthanthat predictedby WHAM(Figures 44 and4C).A contrasting for Cu. Theconcentration of trendwasobserved boundCupredictedby WHAMis verysimilarto that measured for LakesOpasatica andGeneva, overnearlythe entirefreeCu2*concentration 48 and4D). Onlyat freeCu2* range(Figures concentrations above1O-7 M do the WHAMpredictions divergefrom the measuredvalues,in LakeGeneva, wherethe amountof boundCupredictedby WHAMis higherthanthat measured. Notethat sincethe bindingof Cuto DOMis muchmoreimportantthanis the bindingof Ni,the plotsof boundmetal(IM-DOMI) versus[M2.]for Cuappearto be verycloseto the WHAM predictions. Theearlierplotsof IlVt2.] versus[M]oissorueo, are more sensitiveand moreusefulfor predicted comparing and measured values(e.g.,compareFigures 34 and5A). In addition,from perspective, an ecotoxicological is muchmoreimportantthan the freemetalion concentration isthe concentration of boundmetal.Fromthispointof view,the typeof DOMinvolvedin metalcomplexation, i.e.,the contrastbetweenLakesGenevaandOpasatica, is importantfor b o t hC ua n dN i . 776 r v -3.5 LakeGeneva WHAM modeled ..|f o-5 o o lv o o tJo w -4.0 slrf Q -4.5 -J J 2-o o) o 9 -s.o o) o -7 g -5.5 -6.0 -10 -8 -11 -10 -9 -8 -7 -8 C LakeOpasatica WHAM modeled a.tt' C)-5 o ooF o J a I z-o -7 -6 log Cu2+ tog Ni2+ o a Y1t. ow utrl EF I I #B aoo -3.5 o o o ) j D 6 -4.0 ^A a -4.5 A !L o) o o, o -7 6 4 A -5.0 A -8 -11 -10 -9 -8 tog Ni2+ Figure 4. -7 -5.5 41 -10 -9 -8 -7 log Cu2+ (N|-DOM/DOC)for Lake curves Ni(AandC)andCu(BandD)complexation Normalized (CandD). Measured (AandB)andLakeOpasatica values areshownin colour; Geneva values withWHAMVl areshownin graytriangles, calculated areconditional andapplyonlyto the Ni andCubindingparameters Althoughthe calculated it is usefulto comparethemto underwhichtheyweredetermined, experimental conditions therearefew studiesthat reportthe foundin the literature.Unfortunately otherparameters measured of metalsto naturalorganicmatter,asopposedto isolatedhumic bindingproperties (e.9.(Anteloet al.,1.998; Sundaand Hanson in freshwater systems substances, , 1979;Xueand L77 Sunda,1997)).In naturallakewatersandat low,environmentally Ni significant concentrations, wasfoundto bindto strongligandswith logK valuesfrom L2to 14 (Xueet al.,2001).Xueand Sigg(1999)reportedsimilarly strongbindingof Cuto naturalligandsin lakewater(logK values from 10 to 15). lt has,however,beensuggested from competitive that thesevaluescalculated (CLE-AdSV) ligandexchange-adsorptive stripping voltammetry measurements under-estimated the freemetalconcentrations andthereforeover-estimated the complexation of metalswith naturalligands(VanLeeuwen andTown,2005).Nevertheless, it is apparentthat overa wide rangeof Ni concentrations in naturalwatersWHAMoverestimates the free Ni2*concentration (Figures 2A and2C;seealsoMuelleret al. (2OL2c) andVanLaeret al. (2006)).Theopposite trendis observed for Cu,i.e.,an underestimate but the absolute of the freeCu2*concentration, differences betweenmeasured andpredicted valuesaresmallerthanfor Ni2*,and in addition (Figures the differences aremoreimportantin LakeGenevathanin LakeOpasatica 28 and2D), suggesting that the bindingof Cuto naturalligandsmaydependof the typeof DOMpresent. 3.4 Relotionship betweenmetal bindingand DOM charocteristics prevailing Forthe ambientconditions in the two lakes,LakeOpasatica DOMwasfoundto bind moreNi andCuthan LakeGenevaDOM(Figure1). Whenthe concentration of metalbinding siteswasnormalized to the totalorganiccarbonconcentration, LakeOpasatica DOMhada higherapparentnumberof metalbindingsitesfor Cubut a lowernumberof sitesfor Ni (Table 2). Theconditional stabilityconstants alsodifferedbetweenthe two lakes.Basedon the qualityof the DOMsampledfrom the samelakesin 2OO7and2008,Muelleret al. spectroscopic (2012b)foundLakeOpasatica to haveDOMlargelycharacteristic of allochthonous sources, whereasthe characteristics of LakeGenevaDOMsuggested a markedcontribution from autochthonous sources.Thiswasevidenced UVabsorbance at 254nm by a higherspecific (SUVA254 values), a higherrelativeproportionin LakeOpasatica DOMof the PARAFAC fluorescence component attributedto reduced-quinone, fluorophores humic-like of allochthonous origin,anda lowerproportionof the component attributedto tryptophan-like proteinfluorescence of autochthonous origin.Theauthorssupported theirfindingwith watershed characteristic datathat showedthat LakeOpasatica liesin a regionwith a thicker L78 soillayerand hasa muchlargerwatershed-to-lake surfacearearatiothan LakeGeneva, contributing to higherinputsof DOMfrom the lake'swatershed asopposedto beingproduced withinthe lakeitself.Theseresultssuggest that the morearomatichumic-like DOMfrom allochthonous sourcesmayhavea highercomplexation capacity for Cuthandoesthe proteinlikeDOMfrom autochthonous whereasfor Nithe oppositetrendis apparent.A sources, positiverelationship betweenthe Ni andCubindingaffinityandthe opticalquality(SUVA254) wasalsofoundfor naturalsurfacewatersby Bakenet al. (2011).Similarpositiverelationships for Cuhavebeenreportedby Al-Reasi et al. (20L2).Thisis,however,the firsttime that the parallelfactoranalysis (PARAFAC) of the fluorescence character of DOMhasbeencomparedto its metalbindingaffinityandcapacity. 3.5 Environmentolimplications Currently, chemical equilibrium models,suchasWHAM,areusedto estimatethe bindingof cationsby DOMin naturalaquaticsystems(Tipping,2OO2l.Thesemodelsare,however, calibrated usingthe resultsof titrationexperiments on isolatedandpurifiedhumicandfulvic acidat highDOCandtotalmetalconcentrations that areoftenenvironmentally unrealistic for naturalaquaticsystems.Herewe demonstrate, for low,environmentally realistic concentrations, for the complexation that the conditional bindingparameters of Ni andCuwith naturalDOMvaryasa functionof the qualityof the DOM,as revealed by previous fluorescence measurements. Thisresulthasimportantconsequences for thoseinterested in estimating the speciation of thesetracemetalsin naturalaquaticsystems.Forexample, the BioticLigandModel(BLM), usedto predictthe toxicityof tracemetalsto aquaticorganisms, incorporates WHAMto metalspeciation calculate in the exposure medium(Paquin et al.,2002).lmprovedbinding parameters for Ni andCuwouldimprovethe prediction of the bioavailability andtoxicityof thesemetalsat low,chroniclevelexposures. 179 Acknowledgments providedby M.G.Bordeleau, S.Duval,P. Theauthorsacknowledge the technical assistance in the laboratory.Financial supportwasprovidedby the Natural FournierandJ. Perreault andby the Metalsin the Human Research Council of Canada Sciences andEngineering andC.Fortinare P.G.C. Campbell Research Network(www.mithe-rn.org). Environment Research ChairProgram. supportedby the Canada 180 Troisidmepartie: Annexes AppendixA: Effectof samplestorageon metalspeciationand DOM quatity/ AnnexeA : Effet de l'entreposage des6chantillonssur la sp6ciationde m6tauxet la qualit6de la MOD 1 Introduction Oncea naturalaqueoussampleistakenfrom the environment, possible changes in tracemetal speciation mayoccurdueto abioticchanges (e.g.pH,adsorption andchemicalreactions) or (e.g.microbial bioticchanges activity)occurring in the sampleduringstorage.Althoughvery few studieson the stabilityof samples duringstoragecanbe foundin the literature, filtration through0.2-pmfilterscoupledwith storageat 4"Cand in the darkisgenerally accepted asa meansto minimizethesechanges (Batleyet al.,2004)s.Herewe testedthe effectof sample storageon tracemetalspeciation and DOMspectroscopic quality.Forthisexperiment we identified two lakes(LakeGenevaand LakeVaudray) that exhibitmarkeddifferences in DOM (quantityandquality)and in totalandfreemetalconcentrations. 2 Methods Largevolume(10L,polypropylene, Nalgene) surfacewatersamples weretakenfrom lakes GenevaandVaudrayin July2009,returnedto the labandsequentially filteredthrougha series of acid-cleaned polycarbonate filters(5 Um,2 pm,0.4prmand0.2 pm; AMD Manufacturing Inc.).Subsamples weretakenfor analyses of pH,majorion concentrations, DOMoptical properties(absorbance and fluorescence) aswell astotal andfree Cd2*,Cuz*,Ni2*and Zn2* concentrations immediately afterfiltrationandat regularintervals (approximately 3 months) overan 8-monthperiod.Betweensampleanalyses, samples werestoredin the darkat 4'C,as wasthe casefor the lakewatersamples collected with the in situdiffusionsamplers. Fora moredetaileddescription of the sampleanalyses, the readeris referredto Muelleret al. (Muelleret al.,20L2a; Muelleret al.,2O!2b;Muelleret al.,2O].2c) foundin partll of the thesis. Briefly,subsamples weretakenfor determination of total metalconcentrations. The concentrations of cd2*,Ni2*andznz*in the lakesamples werealsomeasured usingan ion 'All references canbe found on page205of the thesis. 183 usinga cupricion of Cuz*wasmeasured whereasthe concentration technique(lET), exchange aswellas organiccarbon(Doc)concentration' dissolved (lsE).Finally, electrode selective werealso measurements CompOnentS) (Fland PARAFAC fluorescence (suvA2s4)and absorbance made. levelusing confidence at the alpha=0'05 werecalculated analyses Thefollowingstatistical sigmaPlotlt.O(systatsoftwarelnc.,chicago,lL).Linearregressionsandone-wayANOVA',s methodwereusedwhen usingthe Holm-sidak procedures with pairwisemultiplecomparison thenormaIityandhomogeneityofvariancesamongdatasetswereconfirmed;otherwise,the one-wayANOVAon Ranks testswereemployed(Kruskal-Wallis non-parametric appropriate usingDunn'smethod'insteadof a parametric procedures with pairwisemultiplecomparison ANoVA).A||proportionaIdatawerearcsinesquareroottransformedbeforeperformingthese statisticaltests'Dateswereconvertedto Juliandays' 3 Resultsand discussion and relative in thetotal metalconcentrations (P<0.05) differences significant statistically as a functionof time werefound proportionthe free metalion to the total metalconcentration a|thoughsma||, significant, A].andA2). Statistica||y for somemeta|sin both|akes(Figures over8 the total metalconcentration in the relativeproportionof the free metalion to increases for Zn in LakeVaudrayandCuin Lake monthswerefoundfor all metalsin bothlakesexcept for Lake andfreeCdconcentrations Geneva(TableA1). lt shouldbe notedthat the total Genevaatthetwofina|timepointswerebe|owthedetection|imitsofthemeasurements. or any DOMspectroscopic (D0.05) in DOCconcentrations differences significant No statisticaily (P<0'05)in DOC differences parameterswerefoundfor LakeGeneva.However,significant LakeVaudray(Figure andC3/Crasa functionof time werenotedfor SUVA25a concentration, trendsin the significant no statistically in theseparameters, A3). Despitethe variability of from LakesVaudrayandGenevaasa function qualityof the DOMsampled spectroscopic time were observed(TableA1)' 184 Althoughsomevariability in the totaldissolved metalconcentrations, the free metalion qualitywasobserved concentrations and DOMspectroscopic overthe 8-monthsamplestorage time,the temporaltrends(whenobserved) weremodest.TheNiz*concentration appeared to changemoreovertime thandid the otherfree metalion concentrations (Figure A2),which againmaypointto the bindingof Ni to differentsitesor differentligands withinthe DOM mosaic.We speculate that theseNi-binding sitesmaybe morelabileor susceptible to change overtime thanthe ligandsinvolvedin complexing the othermetals.Despite the possible liberation of Ni2*ionsas ligands aredegraded overtime,our measured free Ni2*concentrations were,nevertheless, muchlowerthan thosevaluespredictedby WHAM. We thereforehave confidence that therewasno significant bias,dueto sampledegradation, in the tracemetal speciation and DOMqualityresultsreportedin the thesis. 185 5.0e-8 a o 8 0e-10 LakeVaudray LakeGeneva a I 4.0e-8 qp b? T ? 6.0e-10 3.0e-8 F C) 20e-8 2.0e-10 1 0e-8 0.0 00 D o bb OG 7.0e-8 3.0e-8 TI 6 0e-8 2.5e-8 I 5 0e-8 2.0e-8 1.5e-8 z F = qI F I C N 1.0e-8 I 4.0e-8 b 3 0e-8 o 2.0e-8 5.0e-9 T t_ I I 1.0e-8 00 0.0 Nov Jan Month May Nov Jan May Month in filteredsamplesof lakesVaudray(filled in total metal(Mr)concentrations FigureA1. Changes Pointswith similarletters(a-d) circles)and Geneva(opencircles)overtime (2009-2010). points. An asteriskindicatesonly (P>0.05) the between difference indicateno significant totalesde les dans concentrations one replicateabovethe detectionlimit/ Changement pleins) et Geneva(cercles filtr6sdeslacsVaudray(cercles m6taux(Mr) dansles6chantillons Leslettresidentiques(a-d)indiquentqu'il n'y a ouverts)en fonctiondu temps(2009-2010). (P>0.05) pasde diff6rencesignificative entrelespoints.Lesymbole* d6noteun r6pliqua sont au-delAde la limited6tection. dont lesconcentrations 186 1.4 . o 1.2 Lakevaudray LakeGeneva ab 1.0 F F P IJ nn () vva - + b N 0.6 I 002 a 00 000 0.5 1.4 b rI 1.2 b 1.0 c N z + z I o) L N na 0.6 T t 0.4 Ia 0.2 00 0.0 sep Nov Jan Month May sep Nov Jan May Month FigureA2. Changes in the ratioof the free metalCd2*, cu2*,Ni2*andZn2*to their total metal(M1) concentrations in filteredsamplesof lakesVaudray(filledcircles)and Geneva(opencircles), Pointswith similarletters(a-c)indicateno significant overtime (2009-2010). difference (P>0.05) betweenthe points.An asteriskindicatesonlyone replicateabovethe detection limit / Evolutiondu rapportentrelesconcentrations de m6tauxlibresCd2*,Cu2*,Ni2*et Zn2* filtr6sdeslacsVaudray(cercles et lesconcentrations de m6taltotale(M1)dans6chantillons pleins)et Geneva(cerclesouverts)au coursdu temps(2009-2010). Leslettresidentiques (a-c)indiquentqu'il n'y a pasde diff6rencesignificative (P>0.05) entrelespoints.Le symbole* d6noteun r6pliquadont lesconcentrations sontau-delide la limited6tection. t87 12 14 o o LakeVaudray LakeGeneva ito o ab I J 11 1 q () I g) E9 a 5 I g8 I ,tn E^ -J () o ob d d lA xI a I I 1.45 T b a 4 I 1.35 tr 130 C) (r) () t 1 1.25 d o rI 0.1 120 rI 1.15 sep Nov Jan Month May Nov Jan May Month Fl and properties, SUVAzsc, andthe DOMspectroscopic in the DOCconcentration FigureA3. Changes C3/Cr,in filteredsamplesof lakesVaudray(filledcircles)and Geneva(opencircles)over difference(P>0.05) Pointswith similarletters(a-d)indicateno significant time (2009-2010). propri6t6s et des COD de points betweenthe / Evolutiondesconcentrations filtr6sdeslacs Fl et C3/Cr)dansdes6chantillons de la MOD(SUVA2'4, spectroscopiques (2009-2010)' Les ouverts)au coursdu temps Vaudray(cerclespleins)et Geneva(cercles (P>0.05) entreles lettresidentiques(a-d)indiquentqu'il n'y a pasde diff6rencesignificative points. 188 TableA1. coefficientR'; F statistic,P value)for changesin metal Regression statistics(regression speciationand DOMqualityin filteredsamplesof lakesVaudrayand Genevaoverthe 8monthsamplestoragetime.P valuelessthan0.05indicates a slopenot equalto zero,NA indicatesnot enoughdatato performthe test / R6gression statistique(coefficient de de la sp6ciationdes r6gression R2,statistiqueF et valeursde P) pour leschangements filtr6sdeslacsVaudrayet Geneva m6tauxet de la qualit6de la MODdansdes6chantillons plus petite que0.05indiquentune pente pendant8 moisd'entreposage. Lesvaleursde P non6galed z6ro,tandisque le symboleNAindiquequ'iln'y avaitpasassezde donn6es poureffectuerle test statistique. disponibles 189 qualityresults/ AnnexeB : R6sultatsde l'analyse AppendixB: LakeDOMspectroscopic spectroscopique de la MODlacustre resultsfor the eighteenlakes Thefollowingsectionis a compilation of the spectroscopic ON. Figure81 consists of the sampledfrom the regionsof Rouyn-Noranda, QCandSudbury, excitation-emission matrix(EEM)fluorescence spectrafor the lakessampledin 2007(presented on the rightsideof on the left sideof the page)andfor the lakessampledin 2008(presented resultscalculated for the lakessampled. the page).Table81 consists of all of the spectroscopic L91 LakeAdeline2007 4m 380 360 e 340 i P 320 g o I ooo € g 0 280 u 2@ 240 220 320 340 360 380 400 420 440 460 480 500 520 500 520 500 520 LakeBousouet 2007 o o I E c 6 o U 320 340 360 380 400 120 440 4ffi 480 LakeD'Alembert2007 E c D o o G 3 a ul 320 340 360 380 400 420 440 460 480 Emission wavelength (nm) L92 400 400 380 380 360 e 3oo e 340 ; ; o o lrn e o I $ aoo I goo o ii $ zeo 280 IJJ U 260 260 240 240 220 220 320 340 360 380 400 420 440 460 480 500 320 520 400 400 380 380 360 360 e 340 340 360 380 400 420 410 460 480 500 520 340 360 380 400 420 440 460 480 500 520 e 340 ; p ozo E 320 o o .9 I soo I soo :: zeo E o o 240 U uj 260 260 240 240 220 320 340 360 380 400 420 440 460 480 500 520 320 LakeDufay2007 400 380 C a4o P 320 o o I soo o E zso ul 260 240 220 320 340 360 380 400 420 440 460 480 500 520 193 LakeJoannes2007 400 380 360 e 340 ; E sro g o E .oo o E r* a x IJJ 260 240 20 320 U0 360 380 400 420 440 460 480 500 520 Lake Opasatica2007 LakeOpasatica2008 400 400 380 380 360 e 340 € ; E uo o ; p szo o o o I soo I aoo E ,to ,to E o IJJ UJ c o 340 c 260 260 240 240 220 220 320 340 360 380 400 420 440 460 480 500 320 520 340 360 400 380 420 440 460 480 500 520 500 520 LakeVaudray2008 LakeVaudray2007 400 400 380 380 360 s4o E g C a4o ; p azo p szo o o .g o I soo I soo E o o zeo E o zoo E o uJ U 260 260 240 240 220 320 340 360 380 4@ 420 440 460 480 Emission (nm) wa\€length 500 520 320 340 360 380 400 420 440 460 Emissionwavelength(nm) 480 400 400 380 380 360 360 340 e g a4o C g -g szo p szo .c o o o I soo $ soo p zeo 0 zeo E o ul uJ c o o 260 2& 240 24 220 220 320 340 360 380 400 420 440 460 480 500 320 520 340 360 380 LakeCrowley2007 400 420 440 460 480 500 520 Emissionwavelength(nm) 400 380 € 340 ; P 320 o o I aoo o zeo E o uJ 240 220 320 340 360 380 400 420 440 460 480 500 520 LakeGeneva2007 400 400 380 380 350 360 e 340 € g ; p szo 340 E E .ro .c o o o I soo $ soo o o E zso zso E o x ul U 260 260 240 240 220 220 320 340 360 380 400 420 440 460 480 500 520 320 340 360 380 400 420 440 460 480 500 520 400 380 360 340 € ; P 320 .c o 5 goo o zso E I UJ 260 240 220 320 340 360 380 400 420 440 460 480 500 520 LakeMiddle2007 E E o C o o g E o =oo lu 320 340 360 380 400 420 440 460 480 500 520 400 380 360 e 340 ; .ro E o o E*o E o zeo E o uJ 260 240 220 320 340 360 380 400 420 440 460 480 500 520 (nm) Emission wavelength \ 196 400 400 380 380 360 360 e 340 e 340 ; ; P g2o o o E', 320 .g o I soo I soo c o zeo E o € zeo lu tu 260 260 240 240 20 220 320 340 360 380 400 420 440 4@ 480 500 520 500 520 ?,20 uo 360 380 400 120 440 460 480 s00 520 LakeRamsey2007 o o G ' 0 o U 320 340 360 380 400 420 440 460 480 LakeWhitson2007 400 380 e 340 E o c o o o P ,oo o o ; $ soo G a U $ zao c o ul 260 240 220 320 340 360 380 400 420 440 460 Emission wa\relength (nm) 480 500 520 320 340 360 380 400 420 140 460 480 500 520 spectraof lakeDOMfrom matrix(EEM)fluorescence Figure81. Average(N=3)3Dexcitation-emission (left)andthe eightlakessampledin 2008(right).Each the eighteenlakessampledin2OOT to Ramanunits spectrumwascorrectedfor instrumentand innerfilter effects,normalised was Raman scatter where the regions Jagged regions represent and blanksubtracted. en d'excitation-6mission3D de la removed./ Moyenne(N=3)desspectresde fluorescence et leshuit lacs en 2007(i gauche) MODprovenant desdix-huitlacs6chantillonn6s en 2008(d droite).Chaquespectrea 6t6 corrig6pourdeseffets 6chantillonn6s d'instruments et deseffetsli6si l'att6nuationde l'6missionpar la MODelle-mAme(< inner d'un filter effects>),normalis6par rapportauxunit6sde Raman,et corrig6parsoustraction a 6t6 enlev6e. repr6sentent desr6gionsoir la dispersion blanc.Lesr6gionsembrouill6es 198 Table81. (SUVAzas), specificabsorbance Measuredmean(+ SD,N=3)dissolvedorganicmatter(DOM)concentration, specificUV-absorbance fluorescence index(Fl),and relativeproportionof the PARAFAC components(C1,C2,C3and C4)for eachof the coefficient(SACa4o), moyennes(t SD,N=3)de matidreorganique eighteenlakessampledin2OOT andthe eightlakessampledin 2008/ Concentrations (Fl) l'indicede fluorescence le coefficientd'absorbance sp6cifique(SACa4o), dissoute(MOD),l'absorbance UV sp6cifique(SUVA245), (CL,C2,C3et C4)pour chacundes 18 lacs6chantillonn6s en 2007et les8 de fluorescence et la proportionrelativedescomposantes lacs6chantillonn6s en 2008. Adeline Bousquet D'Alembert Dasserat Dufault Dufay Joannds Opasatica Vaudray Bethel Crowley Geneva Hannah Middle Nelson Raft Ramsey Whitson DOC SUVA254 sAc340 (mg Ltl (L m-l mgc-l) (L m-1mgc'l) 4 . 6 0! 0 . 7 9 L2.58t 0.69 6 . 9 6! 7 . 2 7 11.03t 0.62 8.00t 5.14r 5.10r 8.25r 8.67r 8.00r 6.85r 8.36r 1.46r 0.25 3 . 5 2r 0 . 2 0 2.20!0.22 1.98t 0.18 1.48r 0.25 2 . 1 1r 0 . 1 3 2.90t 0.06 1.55t 0.14 2 . 8 8! O . 2 4 0.84r 0.25 L.O2!0.r2 1.09r0.25 1.03t 0.08 1 . 0 1 r0 . 0 8 0.83t 0.42 1 . 3 5r 0 . 7 1 0.74t 0.09 7 . 7 3! 0 . 3 4 0.80 0.58 0.35 0.37 70.26!0.26 6.00r 0.40 8.07r 0.09 5.22!0.46 2.59t 0.28 3.2510.39 3 . 6 1 r0 . L 2 3.32r 0.40 1.36r 0.46 2 . 2 6! 0 . 3 r 3.55r 0.53 3 . 8 2t 0 . 7 2 0.90 0.40 0.70 0.44 9.98! 0.25 7.4r!0.34 9.83t 0.43 4.64!O.47 4.82t0.44 5 . 2 3r 0 . 2 8 5.05r 0.25 5.17tO.57 4 . 5 3! 2 . 0 2 5 . 9 9! 2 . 3 2 4.32r 0.60 6.87r 0.55 1.33r 0.01 1.29r 0.0L 1.30r 0.01 L . 3 51 0 . 0 3 1.32r 0.03 L.30r 0.02 1.28r 0.01 1.35r 0.02 1 . 2 5r 0 . 0 1 1.49t 0.06 1.33! 0.08 1.33r 0.05 1.31r 0.04 1.36r 0.07 1.32r 0.03 1.28r 0.07 7.4I!0.02 1.34r 0.04 ctlc'r c2lcr c3lcr c4lcr 0 . 2 7! 0 . 0 2 0.15r 0.02 0 . 2 1 r0 . 0 3 0.26r 0.19 0 . 2 7! O . L 2 0.22r 0.03 0.18r 0.04 0.25r 0.05 0.27r 0.08 0.24r 0.08 0.29r 0.07 0.29r 0.09 0.28t 0.04 0.25r 0.03 0.28t 0.08 0.29t 0.07 0.24r 0.05 0.30r 0.06 0.33r 0.02 0.07t 0.01 0.24r 0.03 0.34r 0.25 0.33r 0.14 0.24r 0.03 0.15t 0.03 0.32r 0.05 o . 2 4! 0 . 0 7 0.37! 0.13 0.36r 0.08 0.37r 0.10 0.34r 0.04 0.34t 0.04 0.35r 0.09 0.34r 0.08 0.33r 0.06 0.36r 0.08 0.26!0.02 0.75r 0.05 0.48r 0.07 0.30r 0.32 o . 2 7! 0 . 1 0 0.45r 0.05 0 . 5 1 r0 . 1 6 0.29r 0.05 0.43r 0.10 0.15t 0.06 0.15i 0.05 0.18r 0.07 0 . 1 8 10 . 0 3 0.20r 0.03 0.14r 0.03 0 . 1 5r 0 . 1 6 0.17t 0.04 0.19r 0.04 0.15t 0.01 0.02r 0.00 0.08t 0.01 0.10r 0.07 0.14r 0.05 0.09r 0.01 0.05r 0.01 o.t2!0.o2 0.07r 0.02 0.23r 0.09 0.19r 0.04 0.L7r 0.05 0.20r 0.03 0.20r 0.03 0.23r 0.07 o.22!0.06 0.25r 0.05 0.15r 0.03 200 AppendixC: Effectsof aluminum(Al)and iron (Fe)on WHAM Vl predictionsof Cd,Cu,Ni and Zn speciation/ AnnexeC : Effetsde l'aluminium(Al)et du fer (FE)surles pr6dictionsde la sp6ciationdu Cd,du Cu,du Ni et du Zn for bindingsites with traceelements suchasAl3*and Fe3*, Thecompetition of trivalentcations, how we on DOMis accountedfor in WHAMVl (Tipping2OO2l.Thefollowingsectiondescribes of Cd,Cu,Ni andZn choseto dealwith the effectsof Al and Feon WHAMVl predictions in the sampledlakes. speciation in the lakessampled of Al and Feweremeasured concentrations Althoughthe total dissolved Al and Fe (seeTableC1,it is unknownwhatfractionof the measured totaldissolved andavailable to is colloidal Al and Feandwhatfractionis trulydissolved concentrations by Ed competewith tracemetalsfor bindingsiteson the DOM. First,followingsuggestions using the activityof Al3*and Fe3*in the samples TippingandSteveLofts,we calculated of Al3* presented in Tipping(2005)and Loftset al. (2008)wherethe concentrations equations (s), (s)andFe(OH)3 of AI(OH)s asa functionof pH andthe solubility and Fe3*arecalculated ThesevaluesweretheninputteddirectlyintoWHAM. In the secondapproach, respectively. form,andthat all Al and Fewasin the trulydissolved we assumed that all of the total measured the ironwaspresentas Fe(lll).In thiscase,WHAMVl allowsthe userto inputthe total the Al3*and Fe3*concentrations andthencalculates Al(lll)and Fe(lll)concentrations, dissolved usedby WHAMareslightlydifferentfromthe empirical internally.Thesolubility equations that the in Tipping(2005)and Loftset al. (2008).FigureC1,demonstrates relations suggested by aresimilarto thosecalculated WHAMcalculated Cd,Cu,Ni andZnfree metalconcentrations equations. WHAMwhenusingthe externalpH-solubility 20L TableC1. Lake Comparison of measuredtotal dissolved Al3*and Fe3*activities Al and Feconcentrations, calculated usingWHAMVl assuming Al andFeistrulydissolved, all measured andAl3*and Fe3*activitiescalculatedusingthe externalpH-solubility relationsfor Al and Fehydroxides / Comparaison de concentrations totalesdissoutes mesur6espour l'Al et le Fe,lesactivit6s d'Al3*et de Fe3*calcul6es que tout l'Alet Femesur6 en utilisantWHAMVl et en consid6rant est dissous, et lesactivit6sd'Al3*et de Fe3*calcul6es d partirdesrelationspH-solubilitd externespour leshydroxydes d'Al et de Fe. W.HA,M v1 wHAIvlv1 Measured calculated'Measured calculated^ Bethel Dasserat Dufault Geneva Opasatica Raft Vaudray Whitson (M) "i:liii,'J" Ar(r)a rFelr (M) 'i:l,f,Tt Fe(u)a tArtr 7.6E-08 7.3E-1.8 4.9E-O7 4 . 1 E - L 3 8.1E-08 6 . 0 E - 1 6 1.6E-08 1 . 1 E - L 5 3.7E-O7 1 . 2 E - 1 5 NA NA 3.4E-06 1 . 9 E - 1 1 1.5E-07 7.5E-14 L.2E-L6 6.7E-L2 1.8E-14 7.9E-r2 6.5E-15 3.3E-10 1.6E-11 7.9E-I2 NA 2.6E-O7 L.2E-07 3.9E-08 1,.5E-07 1.6E-08 s.0E-07 2.2E-O7 NA 4.4E-L7 2.6E-L9 2.tE-Lg 9.4E-20 4.6E-18 3.5E-16 2.7E-r7 5.7E-20 7.4E-77 1.5E-18 6.3E-r7 7.OE-rg L.7E.Ls t.2E-L6 9.6E-L7 *indicatesAl activitycalculated usingthe equationoutlinedin Tippinget al. (2005). ^indicatedFeactivitycalculated usingthe equationoutlinedin LoftsandTipping(2008). 202 4e-9 1.2e-9 T 1.0e-9 I E J I *, I 3 6.0e- E & () 2e-9 f () 4.0e- T 2.0e- o tIl I n 3.5e-7 0.0 1e-6 3.0e-7 8e-7 2.5e-7 6e-7 s- z 2.0e-7 + N 4e-7 c N 1.5e-7 1.0e-7 2e-7 5.0e-8 0 0.0 rI t-L "$*'%";1))"*$"""),."t..od€"u$+-%.q))'"$*'")".'1'"".'d"'. FigureC1. Comparison betweenmeasuredconcentrations of free Cd2*,Cu2*,Ni2*and Zn2*(circles) and predictedby WHAMVl (i) assuming free M2*concentrations all measuredAl and Fe istruly dissolved(squares), or (ii) usingthe Al3*and Fe3'activities calculated from empiricalpH(triangles). solubilityrelationsfor Al and Fehydroxides Results for eightlakessampledfrom the regionsof Rouyn-Noranda, entre les QCand Sudbury,ON./ Comparaison (cercles) concentrations de Cd2*,Cu2*,Ni2*etZn2*libresmesur6es et lesconcentrations des que tout l'Al et Femesur6est m6tauxlibresM2*pr6ditespar WHAMVl (i) en consid6rant dissous(carr6s), ou (ii)en utilisantlesactivit6sd'Al3*et de Fe3*calcul6es i partirde relationspH-solubilit6 empiriquespour leshydroxydes d'Al et de Fe(triangles). R6sultats pour leshuit lacs6chantillonn6s danslesr6gionsde Rouyn-Noranda, QCet Sudbury,ON. 203 Quatridmepartie: Bibliographie characterization of aquatichumicsubstances F.H.,2004.Structural Abbt-Braun, G.,Lankes, U.,Frimmel, - the needfor a multiplemethodapproach.AquaticSciences 66, 151-170. approach to assess the R.P.,20L0.Novelelectrochemical M., Sander, M.,Schwarzenbach, Aeschbacher, Environ.Sci.Technol.44,87-93. redoxpropertiesof humicsubstances. An introduction in P.,1.985. to humicsubstances R.1.,MacCarthy, Aiken,G.R.,McKnight, D.M.,Wershaw, R.1.,MacCarthy, P.(Eds.), D.M.,Wershaw, soil,sedimentandwater,in:Aiken,G.R.,McKnight, lsolation,and Characterization. HumicSubstances in Soil,Sediment,and Water.Geochemistry, JohnWiley& Sons,Inc.,NewYork,NY,pp. 1-9. isolation in soil,sedimentandwater:geochemistry, Humicsubstances Aiken,G.R.,Wershaw,R.1.,1.985. P.(Eds.). John D.M.,Wershaw,R.1.,MacCarthy, in:Aiken,G.R.,McKnight, andcharacterization, p.477. Wiley,NewYork, the ameliorative effectof naturaldissolved Al-Reasi, H.A.,Smith,D.S.,Wood,C.M.,2012.Evaluating the BLM. organicmatter(DOM)qualityon coppertoxicityto Daphniamagna:improving 21, 524-537. Ecotoxicology propertiesof natural and spectroscopic Al-Reasi, H.A.,Wood,C.M.,Smith,D.S.,2011.Physicochemical for ameliorative effectson metal organicmatter(NOM)fromvarioussourcesand implications toxicityto aquaticbiota.Aquat.Toxicol.103,179-190. to waterqualitycriteria. of tracemetalspeciation Allen,H.E.,Hansen, D.J.,L996.Theimportance Wat.Environ.Res. 58, 42-54. E.,2008,The UVAmery,F.,Degryse, F.,Cheyns,K.,DeTroyer,1.,Mertens,J.,Merckx,R.,Smolders, of dissolved organicmatterpredictsthe fivefoldvariationin its affinityfor mobilizing absorbance Cuin an agricultural soilhorizon.Eur.J.SoilSci.59, 1087-1095. excitation-emission aspectsof Parafacmodelingof fluorescence Andersen, C.M.,Bro,R.,2003.Practical data.J.Chemom.17,2OO-275. Chem.lntell.Lab.Syst. 52, !-4. Andersson, C.A.,Bro,R.,2000.TheN-waytoolboxfor MATLAB. changeson the complexing of Cu(ll)by Antelo,J.M.,Arce,F.,Penedo,F.J.,L998.Effectof seasonal dissolved organicmatterin riverwater.WaterRes.32, 2714-2720. propertiesof F.,Verheyen,1.,Merckx,R.,Smolders Baken,S.,Degryse, , E.,20t1.Metalcomplexation freshwaterdissolved organicmatterare explainedby its aromaticityand by anthropogenic figands.Environ.Sci.Technol.45,2584-2590. of riverdissolvedorganicmatter.Hydrol.Proc.15, discrimination Baker,A.,2002.Spectrophotometric 3203-32L3. organicmatterbioavailability and K.,Jones,J., JafI6,R.,Maie,N.,2009.Streamdissolved Balcarczyk, permafrost.Biogeochemistry 94, 255underlainwith discontinuous compositionin watersheds 270. K.J.(Eds.), Thephysical in: Rousell, D.H.,Jansons, Geology, Barnett,P.J.,Bajc,A.F.,2002. Quaternary Surveypp.67-85. environmentof the city of GreaterSudbury.OntarioGeological and bioavailability of tracemetalsin water: Speciation Batfey,G.E.,Apte,S.C., Stauber, J.L.,2004. Progress since1982.Aust,J. Chem.57, 903-91.9. W.H.,Koopal,1.K.,1995.Metalion binding D.G.,VanRiemsdijk, Benedetti, M.F.,Milne,C.J.,Kinniburgh, model.Environ. Sci. adsorption application of the non-idealcompetitive to humicsubstances: Technol.29,446-457. and at Rouyn-Noranda, aroundsmelters Quebec, Bonham-Carter, G.F.,2005.Metalsin the environment MITE Point Project. of the GSC Sources resultsand conclusions New Brunswick: Belledune, Surveyof Canada, Ottawa,Ontario,Canada,p. 265. Geological 1.M.,2006.Comparison of metaldistributions P.J.,Kliza,D.A.,Kettles, Bonham-Carter, G.F.,Henderson, peat, in western Canada. Geochem. a Cu smelter lakes humus around in snow, and Quebec, Explor.Environ.Anal.5, 275-228. 207 partitioning Bonneris, E.,Perceval, O.,Masson, 2005.Sub-cellular of Cd,Cu S.,Hare,1.,Campbell, P.G.C., gradientandlinksto andZn in tissuesof indigenous livingalonga metalexposure unionidbivalves metal-induced effects.Environ.Pollut.135,195-208. Borgmann, U.,Nowierski,M., Grapentine, 1.C.,Dixon,D.G.,2004.Assessing the causeof impactson benthicorganisms nearRouyn-Noranda, Quebec.Environ.Pollut.129,39-48. Borgmann, U.,Reynoldson, T.8.,Rosa,F.,Norwood,W.P.,1998.Finalreporton the effectsof atmospheric ReportNo. depositionof metalsfrom the Sudburysmelterson aquaticecosystems. pp. 1-61. 01-023.Environment Canada, Burlington, ON,Canada, Bryan,S.E., Tipping, E.,Hamilton-Taylor, copperbinding J.,2002.Comparison of measured and modelled by naturalorganicmatterin freshwaters. 133,37C:Toxicol.Pharmacol. Comp.Biochem.Physiol., 49. Buffle,J.,1.984. Naturalorganicmatterand metal-organic interactions in aquaticsystems,in: Sigel,A., (Eds.), Sigef,H. Met. lonsBiol.Syst.MarcelDekkerInc.,NewYork,NY,pp. t65-22I. Buffle,J.,1988.Complexation Reactions in AquaticSystems: An Analytical Approach.EllisHorwoodLtd., Chichester, UK. Cabaniss, S.E., Shuman, M.S.,1988.Copperbindingby dissolved Riverfulvic organicmatter:l. Suwannee acidequilibria. Geochim. Cosmochim. Acta52,185-193. P.G.C., Campbell, 1995.Interactions betweentracemetalsand organisms: a critiqueof the free-ion activitymodel,in: Tessier, A.,Turner,D. (Eds.),MetalSpeciation in Aquatic and Bioavailability Systems. J. Wiley& Sons,Chichester, UK,pp. 45-102. Campbell, P.G.C., Chapman, P.M.,Hale,8.A.,2006.Riskassessment lssues of metalsin the environment. Environ.Sci.Tech nol. 27, L02-I31. Campbef f, P.G.C., Twiss,M.R.,Wilkinson, K.J.,L997.Accumulation of naturalorganicmatteron the surfaces of livingcells:implications for the interactionof toxicsoluteswith aquaticbiota.Can.J. Fish.Aquat.Sci.54, 2543-2554. Cao,J.,Xue,H.,Sigg,1.,2006.Effectsof pH and Cacompetitionon complexation of cadmiumby fulvic acidsandby naturalorganicligands from a riveranda lake.Aquat.Geochem. \2,375-387. Carignan, R.,Nriagu, J.O.,1985.Tracemetaldeposition andmobilityin the sediments of two lakesnear Sudbury,Ontario.Geochim.Cosmochim. Acta49, t753-t764. CEPA, 2001.Canadian Environmental ProtectionAct 1999.PrioritySubstances ListAssessment Report: Releases from primaryand secondary Releases from coppersmeltersandcopperrefineries, primaryand secondary zincsmeltersand zincrefineries. Environment Canada/ HealthCanada, Ottawa,ON,p. 164. T.,DeSchamphelaere, Cheng, K.,Lofts,S.,Janssen, C.,Allen,H.E,,2005.Measurement andcomputation of zincbindingto naturaldissolved organicmatterin Europeansurfacewaters.Anal.Chim.Acta 542,230-239. Chin,Y.P.,Aiken,G.R.,O'Loughlin, E.,1.994. Molecularweight,polydispersity, and spectroscopic properties of aquatichumicsubstances. Environ. Sci.Technol. 28, 1853-1858. Coble,P.G.,1996.Characterization of marineandterrestrialDOMin seawaterusingexcitation-emission matrixspectroscopy. MarineChem.51,325-346. Cornu,J.Y.,Parat,C.,Schneider, A.,Authier,1.,Dauthieu, M., Sappin-Didier, V.,Denaix, 1.,2009. geochemical Cadmiumspeciationassessed in soil by voltammetry,ion exchange and calculation solutionscollectedaftersoilrewetting.Chemosphere 75, 502-508. Cornu,J.Y.,Schneider, A.,Jezequel, K.,Denaix, L.,20LL.Modelling the complexation of Cdin soilsolution at differenttemperaturesusingthe UV-absorbance of dissolvedorganicmatter.Geoderma162, 65-70. Cory,R,M.,McKnight,D.M.,2005.Fluorescence revealsubiquitouspresence spectroscopy of oxidized and reducedquinonesin dissolved organicmatter.Environ.Sci.Technol.39,8742-8749. 208 Cory,R.M.,Miller,M.P.,McKnight,D.M.,Guerard,J.J.,Miller,P.1.,2010.Effectof instrument-specific responseon the analysis of fulvicacidfluorescence spectra.Limnol.Oceanogr. Methods8,67-78. Couillard, Y.,Courcelles, M,, Cattaneo, A., Wunsam,S.,2004.A test of the integrityof metalrecordsin sedimentcoresbasedon the documentedhistoryof metalcontamination in LacDufault(Qu6bec, Canada). J. Paleolimn ol. 32, L49-762. Croteau,M.N.,Hare,1.,Tessier, A.,2002.Increases in food web cadmiumfollowingreductionsin atmospheric inputsto somelakes.Environ.Sci.Technol.36, 3079-3082. De Haan,H., L992.lmpacts of environmental changeson the biogeochemistry of aquatichumic substances.Hydrobiologia229,59-7t. Diffon, P.J.,Molot, L.A.,L997.Effectof landscape form on exportof dissolved organiccarbon,iron,and phosphorus from forestedstreamcatchments. Wat.Resour.Res. 33, 2591-2600. Dixit,A.S.,Alpay,S.,Dixit,S.S.,Smol,J.P.,2007. Paleolimnological reconstructions of Rouyn-Noranda lakeswithin the zoneof influenceof the HorneSmelter,Quebec,Canada. J. Paleolimnol. 38, 209226. Doig,1.E.,Liber,K.,2007.Nickelspeciationin the presenceof differentsourcesandfractionsof dissolved organicmatter.Ecotoxicol. Environ.Saf.55, L69-!77. Dudaf,Y.,G6rard,F.,2004. Accounting for naturalorganicmatterin aqueouschemical equilibrium models:a reviewof the theoriesand applications. Earth-Sci. Rev.55, 799-276. Dupont,J.,1992.Quebeclakesurvey:ll. Originand extentof acidification. WaterAir SoilPollut.61.,t25t37. properties Ephraim, J.H.,Marinsky, J.A.,Cramer, S.J.,L989.Complex-forming of naturalorganicacids. Fufvicacidcomplexes with cobalt,zincand europium.Talanta36,437-443. Fellman,J.B.,Hood,E.,Spencer,R.G.M.,2010.Fluorescence spectroscopy opensnew windowsinto dissolvedorganicmatterdynamicsin freshwaterecosystems: A review.Limnol.Oceanogr. 55, 2452-2462. Fortin,C.,Campbell,P.G.C., 1-998. (Cd2*, An ion-exchange techniquefor free-metalion measurements Zn2*): applications to complexaqueousmedia,Int.J. Environ. Anal.Chem.72, t73-794. Fortin,C.,Couillard, Y.,Vigneault,8., Campbell, P.G.C., 2010.Determination of free Cd,CuandZn concentrations in lakewatersby in situdiffusionfollowedby columnequilibrationion-exchange. Aquat.Geochem.16,75t-172. Gallon,C.,Tessier, A.,Gobeil,C.,Carignan, R.,2006.Historicalperspective of industrialleademissions to the atmosphere from a Canadian smelter.Environ. Sci.Technol.40,74I-747. partitioningof cadmium,copper, Gigudre,A.,Campbell,P.G.C., Hare,1.,Couture,P.,2006.Sub-cellular yellowperch(Perco nickefand zincin indigenous sampled alonga polymetallic flovescens) gradient.Aquat.Toxicol.77, 178-189. Gondar,D.,L6pez,R.,Fiol,S.,Antelo,J.M.,Arce,F.,2006.Cadmium, lead,andcopperbindingto humic acidandfulvicacidextractedfrom an ombrotrophicpeatbog.Geoderma135,196-203. Gondar,D.,Thacker,S.A.,Tipping,E.,Baker,A., 2008.Functionalvariability of dissolvedorganicmatter from the surfacewaterof a productivelake.WaterRes.42, 81-90. Gopalapillai, Y.,Fasfous, 1.1., Murimboh,J.D.,Yapici, T.,Chakraborty, P.,Chakrabarti, C.1.,2008. Determination usingthe ion exchange of free nickelion concentrations technique:application to aqueousminingand municipal effluents. Aquat.Geochem. 14,99-1L6. Goulet,R.R.,Couillard, Y.,2009.Weight-of-evidence assessment of impactsfrom an abandonedmine siteto the Dasserat lakewatershed, in: Miranda,F.R.,Bernard, Canada, L.M.(Eds.), Lake Quebec, PollutionResearch lnc.,Hauppauge, Progress. NovaScience Publishers, NY,USA,pp. 355-369. propertiesof chromophoric Green,S.A.,Blough,N.V.,1994.Opticalabsorptionandfluorescence dissolvedorganicmatterin naturalwaters.Limnol.Oceanogr. 39, L903-1916. 209 C.1., J.W.,Hassan, N.M.,Salam,M.S.A., Fasfous, 1.1., Murimboh,C.A.,Murimboh,J.,Chakrabarti, Guthrie, Gr6goire,D.C.,2005.Complexation of Ni, Cu,Zn,and Cd by DOCin somemetal-impacted determinationof free-metalfreshwaterlakes:a comparison usingelectrochemical of approaches model,WHAMV andVl. Anal.Chim.Acta528, ion andlabilecomplexes anda computerspeciation 205-218. and Ca2*by of Co2*,Ni2*,UO22t Higgo,J.J.W.,Kinniburgh, D.,Smith,8., Tipping,E.,1993.Complexation humicsubstances in groundwaters. Radiochimica Acta61,91.-103. sediments, of aquaticandterrestrial Hornbrook, E.H.W., Kettles, 1.M.,Shilts,W.W.,1986.Geochemistry Precambrian Shieldof southeastern Ontario.WaterAir SoilPollut.3I,97L-98t. J.1.,2008.Spatialand D.M.,Maie,N.,Cory,R.M.,McDowell, W.H.,Campbell, Jaff6,R.,McKnight, in DOMcompositionin ecosystems: Theimportanceof long-termmonitoring temporalvariations ofopticalproperties. J.Geophys.Res.113,G04032. riskassessment Janssen, C.R.,Heijerick, D.G.,DeSchamphelaere, Allen,H.E.,2003.Environmental K.A.C., of metals:Toolsfor incorporating Environ.Int. 28,793-800. bioavailability. Measuringfree W.H.,2006. Kalis,E.J.J., Weng,1.P.,Dousma, F.,Temminghoff, VanRiemsdijk, E.J.M., Technique. metalion concentrations in situin naturalwatersusingthe DonnanMembrane Environ. Sci.Technol. 40,955-961. free metalion Kalis,E.J.J., Weng,1.P.,Temminghoff, E.J.M., VanRiemsdijk, W.H.,2007.Measuring Anal.Chem. technique. concentrations in multicomponent solutions usingthe Donnanmembrane 79, 1555-1563. from the Sudburysmelterson Keller,W., Heneberry, J.,Gunn,J.,L999.Effectsof emissionreductions StressRecovery 6, 189-198. the recoveryof acid-and metal-damaged lakes.J.Aquat.Ecosyst. from Sudbury, Kefler,W., Yan,N.,Gunn,J.,Heneberry, . Recovery of acidifiedlakes:Lessons J.,2OO7 Ontario,Canada. WaterAir SoilPollut.:Focus7, 3L7-322. propertiesof dissolved Kefton,N.,Molot, 1.A.,Dillon,P.J.,2007. organicmatter Spectrofluorometric from Centraland SouthernOntariostreamsandthe influenceof iron and irradiation.WaterRes. 4I,638-646. Khan,S.U.,Schnitzer, M., t97L. Furtherinvestigations of the chemistryof fulvicacid,a soil humic fraction.Can.J. Chem.49,2302-2309. W.H., Kinniburgh, D.G.,Milne,C.J.,Benedetti, J.,Koopal, 1.K.,VanRiemsd'tjk, M.F.,Pinheiro, J.P.,Filius, 1996.Metalionbindingby humicacid:application model.Environ. Sci. of the NICA-Donnan Technol.30, 1587-1698. Kdhfer, S.,Buffam,1.,Jonsson, of stream A., Bishop,K.,2002.Photochemical and microbialprocessing and soilwaterdissolved organicmatterin a borealforestedcatchmentin northernSweden. Aquat.Sci.54, 269-281,. Koster,G.,Schmuckler, G.,1967.Separation of noblemetalsfrom basemetalsby meansof a new chelationresin.Anal.Chim.Acta,179-t84. Lawaetz, A.J.,Stedmon,C.A.,2009.Fluorescence intensitycalibrationusingthe Ramanscatterpeakof water.Appl.Spectrosc. 63,936-940. Lofts,S.,Tipping,E.,207'J.. Assessing of free metalion WHAM/ModelVll againstfield measurements concentrations: modelperformance and inputs.Environ, andthe roleof uncertaintyin parameters C h e m . 85, 0 1 - 5 L 6 . Lofts,S.,Tipping, E.,Hamilton-Taylor, Aquat. of Fe(lll)in freshwaters, J.,2008.Thechemical speciation 14,337-358. Geochem. Luider,C.D.,Crusius, J.,Playle, R.C., Curtis,P.J.,2004.Influence of naturalorganicmattersourceon gills, copperspeciation asdemonstrated by Cu bindingto fish by ion selectiveelectrode,and by DGTgefsampler.Environ.Sci.Technol.38,2865-2872. 2L0 Luoma,S.N.,Rainbow, P.S., 2008.MetalContamination in AquaticEnvironments: Science and Lateral Management. Cambridge University Press, Cambridge, UK. MacCarthy, P.,Rice,J.A.,1991.An ecological rationalefor the heterogeneity of humicsubstances: a holosticperspective on humus., in:Schneider, S.H.,Boston,P.J.(Eds.), Scientists on Gaia.MIT pp. 339-345. Press,Cambridge, Massachusetts, Mattsson,T., Kortelainen, P.,Rdike,A., 2005.Exportof DOMfrom borealcatchments: lmpactsof land usecoverand climate.Biogeochemistry 76, 373-394. McKnight,D.,M., Boyer,E.W.,Westerhoff,P.K.,Doran,P.T.,Kulbe,T.,Andersen,D.T.,2001. Spectrofluorometric characterization of dissolved organicmatterfor identification of precursor organicmaterialand aromaticity.Limnol.Oceanogr. 46, 38-48. Miller,M.P.,McKnight, D.M.,2010.Comparison of seasonalchanges in fluorescent dissolved organic matteramongaquaticlakeandstreamsitesin the GreenLakes Valley.J.Geophys.Res. 115. Miller,M.P.,Simone,B.E.,McKnight, D.M.,Cory,R.M.,Williams, M.W.,Boyer,E.W.,20i.0. Newlighton a darksubject:comment.Aquat.Sci.72, 269-275. Milne,C.J.,Kinniburgh, D.G.,VanRiemsdijk, W.H.,Tipping, E.,2003.GenericNICA-Donnan model parameters for metal-ionbindingby humicsubstances. Environ.Sci.Technol.37,958-97t. Mobed,J.J.,Hemmingsen, S.1.,Autry,J.1.,McGown,1.8.,1996.Fluorescence characterization of IHSS humicsubstances: Totalluminescence spectrawith absorbance correction. Environ. Sci.Technol. 30,3061-3065. Mofot,1.A., Dillon,P.J.,1997. Photolytic regulation of dissolved organic-carbon in northernlakes.Global Biogeochem. Cycles1L,357-355. Morel,F.M.M.,Hering, J.G.,1993.Complexation, ln: Principles andApplications of AquaticChemistry. J. Wiley& SonsLtd.,NewYork,NY,USA,pp. 358-395. Mota,A.M.,CorreiaDosSantos, M.M.,L995.Tracemetalspeciation of labilechemicalspecies: Electrochemical methods,in:Tessier, A.,Turner,D. (Eds.), MetalSpeciation andBioavailability in AquaticSystems. J. Wiley& Sons,Chichester, UK,pp. 205-257. Mueller,K.K.,Fortin,C.,Campbell, P.G.C., 2010.Exploring linksbetweenfluorophores and metalbinding by naturalorganicmatter,International DOMSpectroscopy Workshopin, Granada, Spain. Mueffer, K.K.,Fortin,C.,Campbell,P,G.C., 2Ot2a.Nickeland coppercomplexation by naturalorganic matterin lakewaters.Draftmanuscript. Muefler,K.K.,Fortin,C.,Campbell,P.G.C., 2OtZb.SpatialVariationin the OpticalProperties of Dissolved OrganicMatter(DOM)in Lakeson the Canadian Precambrian Shieldand Linksto Watershed Characteristics. AquaticGeochem istry18,2t-44. Mueffer,K.K.,Lofts,S.,Fortin,C.,Campbell,P.G.C., 2Ot2c.lmprovingtracemetalspeciationpredictions in naturalaquaticsystems:incorporation of dissolved quality. organicmatter(DOM)spectroscopic Environ. Chem.9, 356-368. Nelson,W.O.,Campbell,P.G.C., 1991.Theeffectsof acidification on the geochemistry of Al, Cd,Pband Hgin freshwaterenvironments: a literaturereview.Environ.Pollut.7L,97-130. Nriagu, J.O.,Wong,H.K.T., Coker,R.D.,1982.Deposition andchemistry of pollutantmetalsin lakes aroundthe smelters at Sudbury, Ontario.Environ. Sci.Technol. 16,551-550. Nriagu,J.O.,Wong,H.K.T., Lawson,G.,Daniel,P.,1998.Saturation of ecosystems with toxic metalsin Sudburybasin,Ontario,Canada.Sci.TotalEnviron.223,99-LL7 . O'Donnell, J.A.,Aiken,G.R.,Kane,E.S., Jones, J.8.,2010.Sourcewatercontrolson the character and originof dissolved organicmatterin streamsof the YukonRiverbasin,Alaska. J.Geophys. Res.(G Biogeosci.) 115,G03025. Paquin,P.R.,Gorsuch, J.W.,Apte,S.,Batley,G.E.,Bowles, K.C., Campbell, P.G.C., Delos,C.G.,Di Toro, D.M.,Dwyer,R.1.,Galvez, F,,Gensemer, R.W.,Goss,G.G.,Hogstrand, C.,Janssen, C.R.,McGeer, J.C.,Naddy,R.8.,Playle, R.C., Santore, R.C., Schneider, U.,Stubblefield, W.A.,Wood,C.M.,Wu, 27L Physiol. C: Pharmacol. Comp.Biochem. Thebioticligandmodel:ahistorical overview. K.8.,2002. Toxicol.133,3-35. lakes,in: Rousell, W., 2002.Thepast,presentandfutureof Sudbury's D.A.B., Gunn,J.M.,Keller, Pearson, Ontario Thephysical of the cityof GreaterSudbury. D.H.,Jansons, K.J.(Eds.), environment Geological Survey,pp, L95-215. 5, Treatiseon Geochemistry Perdue,E.M.,Ritchie,J.D.,2003.Dissolved organicmatterin freshwaters. 273-3L8. alonga Pyle,G.G.,Rajotte,J.W.,Couture,P.,2005.Effectsof industrialmetalson wild fish populations metaI contaminationgradient.Ecotoxicol. Environ.Saf.61.,287-3t2. electrodefor free coppermeasurements Rachou, J.,Gagnon,C.,Sauv6,S.,2007.Useof an ion-selective Environ. Chem.4,90-97. in low salinityand low ionicstrengthmatrices. Richards, J.G.,Curtis,P.J.,Burnison,8.K.,Playle,R.C.,2001..Effectsof naturalorganicmattersourceon mykissland on metalbindingto their gills. reducingmetaltoxicityto rainbowtrout (Oncorhynchus Environ.Toxicol.Chem.20, 1159-1165. D.H., Rousell,D.H.,Meyer,W., Prevec,S.A.,2002.Bedrockgeologyand mineraldeposits,in: Rousell, Thephysicalenvironmentof the city of GreaterSudbury.OntarioGeological K.J.(Eds.), Jansons, Survey,pp. 21-55. and ion potentiometryfor of spectrofluorometry Saar,R.A.,Weber,J.H.,1980.Comparison ions.Anal.Chem.,2095-2100. betweenfulvicacidand heavy-metal determination of complexes W.D.,McAvov,D.C.,1992.MINEQL+: a softwareenvironmentfor chemicalequilibrium Schecher, modeling.Comput.Environ.Urban Syst.L6, 65-76, organiccarboncyclingin forested Schiff,S.1.,Aravena,R.,Trumbore,S.E.,Dillon,P.J.,1990.Dissolved . r.Res.26, 2949-2957 watersheds : a carbon isotopeapproach.Wat.Resou subjectedto climatic D.W.,Curtis,P.J.,t997. Theroleof DOCin protectingfreshwaters Schindler, 36, 1-8. warmingand acidification from UVexposure.Biogeochemistry for soilorganicmatterandsoils.Soil H.R.,Schnitzer, M., 1997.Chemical modelstructures Schulten, Science152,115-130. R.,2006.Chemical M., Christl,1.,Vogt,R.,Barmettler,K.,Jacobsen, C.,Kretzschmar, Schumacher, sampledin spring compositionof aquaticdissolved organicmatterin five borealforestcatchments Biogeochemistry andfall seasons. 80, 263-275. Scully,N.M.,Lean,D.R.S., 1994.Theattenuationof ultravioletradiationin temperatelakes. Arch.Hydrobiol.Beih. 43, L35-t44. Senesi,N., 1990.Molecularand quantitativeaspectsof the chemistryof fulvicacidand its interactions Anal. approach. spectroscopy with metalionsandorganicchemicals. Partll. Thefluorescence chim.Acta232,77-106. metalsconcentrations M., Pascoe, D.,Borgmann, Shuhaimi-Othman, U.,Norwood,W.P.,2005.Reduced of water,sedimentand Hyalellaaztecafrom lakesin the vicinityof the Sudburymetalsmelters, Environ. Ontario,Canada. Monit,Assess. t17,27-44. A., Kramer, J.R.,Mannio,J.Y., M.S.,Calmano, W., De Haan,H.,Fredrickson, H.1.,Henriksen, Shuman, Morel,F.M.M.,Niemeyer, J.,Ohman,1.O.,Perdue,E.M.,Weis,M., 1990.Groupreport:Howare propertiesof "DOC"measuredand modeled,and how do they affectaquatic acid-base J.Wiley& Acidsin AquaticEcosystems. ecosystems?, in: Perdue, E.M.,Gjessing, E.T.(Eds.), Organic SonsLtd.,NewYork,pp. 141-149. D., J.,Gunkel,P.,Kalis,E.,Kistler, R.,Davison, W.,Galceran, Sigg,1.,Black,F.,Buffle,J.,Cao,J.,Cleven, E.,TercierW.,Temminghoff, Martin,M., Noel,S.,Nur,Y.,Odzak,N.,Puy,J.,VanRiemsdijk, Waeber,M.1.,Toepperwien, E.,Warnken,K.W.,Weng,1.P,,Xue,H.B., S.,Town,R.M.,Unsworth, in natural Zhang,H.,2006.Comparison for dynamictracemetalspeciation of analytical techniques freshwaters. Environ.Sci.Technol.40,L934-t947. 212 of metalsand metalcomplexes: the bioavailability V.1.,Wilkinson,K.J.,2005.Predicting Slaveykova, Chem.2,9-24. Criticalreview of the BioticLigandModel.Environ. P.,Cole,J.J.,2007.Patternsand regulationof dissolved Sobek,S.,Tranvik,1.J.,Prairie,Y.T.,Kortelainen, 52, L2O8-L2L9. of 7,500widelydistributedlakes.Limnol.Oceanogr. organiccarbon:An analysis with parallelfactor dissolvedorganicmatterfluorescence Stedmon,C.A.,Bro,R.,2008.Characterizing Methods6, 572-579. anafysis: a tutorial.Limnol.Oceanogr. organicmatterin aquaticenvironments Stedmon,C.A.,Markager,S.,Bro,R.,2003.Tracingdissolved Marine Chemistry82,239-254. fluorescence spectroscopy. usinga new approachto Equilibria and Ratesin NaturalWaters.John Chemical Stumm,W., Morgan,J.J.,1995,AquaticChemistry. Wileyand SonsInc.,NewYork,NY,USA. of copperin riverwater,in:Jenne,E.A.(Ed.), P.J.,1979.Chemical speciation Sunda,W.G.,Hanson, ACSSymposium SeriesNo.93,pp.147-180. Systems, Chemical Modelingin Aqueous the newview.Environ.Sci. Sutton,R.,Sposito,G.,2005.Molecularstructurein soil humicsubstances: Technol.39, 9009-9015. D.,Rancourt, D.G.,2006.Theroleof smelter M.S.,Kliza-Petelle, Telmer,K.H.,Daneshfar,B., Sanborn, in sediment chemistry of 99 lakesaroundthe Horne emissions andelementremobilizationthe smelter,Quebec.Geochem.Explor.Environ.Anal.6, t87-202. H.P.,Lobinski, R., l.G.,Muntau,H.,vanLeeuwen, R.,Danielsson, Templeton, D.M.,Ariese,F.,Cornelis, andfractionationof elements. 2000.Guidelines for termsrelatedto chemicalspeciation PureAppl.Chem.72,I453-L470. approaches. and methodological Definitions, structuralaspects, for Windowsuser'sguide: manualandCanoDraw ter Braack, C.J.F., Smilau er,P.,2002.Reference communityordination,4.5 ed. Micropower,lthaca,NY. Softwarefor canonical Geochemistry of NaturalWaters.MartinusNijoff/ Dr.W. JunkPublishers Thurman,E.M.,L985.Organic (KluwerAcademicPublishers Netherlands. Dordrecht, Group), Environ. Sci. isolation of aquatichumicsubstances. Thurman,E.M.,Malcolm,R.1.,1981.Preparative Technol.75,463-466. of protons of the interactions ModelVl: an improveddescription Tipping,E.,1998.Humicion-binding 4,3-48. Aquat.Geochem. andmetalionswith humicsubstances. UniversityPress,Cambridge, UK. Cambridge Tipping,E.,2002.CationBindingby HumicSubstances. organicmatterin soil Tipping,E.,2005.ModellingAl competitionfor heavymetalbindingby dissolved andsurfacewatersof acidand neutralpH.Geodermat27,293-304. Tipping,E.,Lofts,S.,Lawlor,A.J.,1998.Modelingthe chemicalspeciationof tracemetalsin the surface waters.Sci.TotalEnviron.2IO,63-77. of ModelVll:a revisedparameterisation Tipping,E.,Lofts,S.,Sonke, J.E.,2011.Humiclon-Binding Chem.8,225-235. Environ. cation-binding by humicsubstances. J.,2002.Al(lll)and Fe(lll)bindingby humic Tipping,E.,Rey-Castro, C.,Bryan,S.E.,Hamilton-Taylor, Acta for tracemetalspeciation. Geochim.Cosmochim. in freshwaters, and implications substances 66,32rL-3224. revisited:evaluating Sudbury sediments W.,Smol,J.P.,2OLO. Tropea,A.E.,Paterson, A.M.,Keller, WaterAir SoilPollut.210,377-333. environment. limnofogical recoveryin a multiple-stressor Metal modeling, in:Tessier, A.,Turner,D. (Eds.), in tracemetalspeciation Turner,D.R.,1995.Problems UK,pp. 149-203. J.Wiley& Sons,Chichester, in AquaticSystems. Speciation and Bioavailability W., Black,F.,Buffle,J.,Cao,J.,Cleven,R.,Galceran, Unsworth,E.R.,Warnken,K.W.,7hang,H.,Davison, H,P.,Martin,M., Noel,S.,Nur,Y.,Odzak, O.,Puy, D.,vanLeeuwen, J.,Gunkel,P.,Kalis,E.,Kistler, Town, E.,Tercier-Waeber, M.1.,Toepperwein, S., W., Sigg,1.,Temminghoff, J.,Van Riemsdijk, of metalspeciationin freshwaters comparedto R.M.,Weng,1.,Xue,H.,2006.Modelpredictions Environ.Sci.Technol.40, L942-7949. measurements by in situtechniques. 213 K.A.C., 2006.Speciation of nickel Van Laer,1.,Smolders, E.,Degryse, F.,Janssen, C.,DeSchamphelaere, in surface watersmeasured Anal.Chim.Acta578,195with the Donnanmembrane technique. 202. of metalcomplexes by VanLeeuwen, H.P.,Town,R.M.,2005.Kineticlimitations in measuring stabilities Environ.Sci.Technol. competitiveligandexchange-adsorptive strippingvoltammetry(CLE-AdSV). 39,72L7-7225. Veillette,J.J.,Paradis, S.J.,Buckle,J.,2005.Bedrockand surficialgeologyof the generalareaaround Rouyn-Noranda, G.F.(Ed.),Metalsin the Environment QuebecandOntario,in: Bonham-Carter, Results AroundSmeltersat Rouyn-Noranda, New Brunswick: and Belledune, and Quebec, p. Conclusions of the GSCMITEPointSourcesProject.Geological Surveyof Canada, 15. of Weishaar, J.1.,Aiken,G.R.,Bergamaschi, B.A.,Fram,M.S.,Fujii,R.,Mopper,K.,2003.Evaluation specificultravioletabsorbance asan indicatorof the chemicalcompositionand reactivityof dissofved organiccarbon.Environ.Sci.Technol.37,4702-4708. Weng,1.,AlonsoVega,F.,Van Riemsdijk, W.H.,20L1.Strategies in the applicationof the Donnan membrane technique. Environ. Chem.8,466. - Lakeand RiverEcosystems. Wetzel,R.G.,2001.Limnology AcademicPress,SanDiego,CA,USA. Wood,C.M.,Al-Reasi, H.A.,Smith,D.S.,201L.Thetwo facesof DOC.Aquat.Toxicol.105,3-8. ion exchange Worms,l.A.M.,Wilkinson, K.J.,2008.Determination of Ni2*usingan equilibrium lmportantchemical Anal.Chim. technique: factorsandapplicability samples. to environmental Acta516,95-102. Xue,H.,Sigg,1.,L999.Comparison of the complexation of Cuand Cd by humicor fulvicacidsand by ligandsobservedin lakewaters.Aquat.Geochem. 5, 313-335. kineticsin freshwater Xue,H.B.,Jansen, S.,Prasch, A.,Sigg,1.,2001.Nickelspeciation and complexation by ligandexchange and DPCSV. Environ.Sci.Technol.35, 539-546. in Xue,H.B.,Oestreich, A., Kistler,D.,Sigg,1.,L996.Freecupricion concentrations and Cucomplexation selectedSwisslakesand rivers.Aquat.Sci.58, 69-87. Xue,H.B.,Sigg,1.,1993.Freecupricion concentration in a eutrophiclake.Limnol. andCu(ll)speciation Oceanogr.38, L2OO-L213. Xue,H.B.,Sigg,1.,1994.Zincspeciationin lakewatersand its determinationby ligandexchange with EDTAanddifferentialpulseanodicstrippingvoltammetry.Anal.Chim.Acta284,505-515. Xue,H.B.,Sunda, W.G.,1997.Comparison in lakewaterdetermined by ligand measurements of [Cu2*] exchange and cathodicstrippingvoltammetryand by ion-selective electrode.Environ.Sci. Technol.31, 1902-1909. Yamashita, Y.,Jaff6,R.,2008.Characterizing the interactions betweentracemetalsand dissolved parallel Environ. organicmatterusingexcitation-emission factoranalysis. Sci.Technol. matrixand 42,7374-7379. Yamashita, Y.,Scinto,1.,Maie,N.,Jaff6,R.,2010.Dissolved acrossa organicmattercharacteristics wetland'slandscape: subtropical Applicationof opticalpropertiesin the assessment of environmental dynamics.Ecosystems 13, 1005-1019. Yan,N.D.,Miller,G.E.,1984.Effectsof depositionof acidsand metalson chemistryand biologyof lakes nearSudbury, JohnWiley& Ontario,in: Nriagu, impactsof smelters, J.O.(Ed.),Environmental Sons,pp. 243-282. Yoshioka, T.,Mostofa,K.M.G., Konohira, K.,Takahashi, M., Ueda,S., E.,Tanoue,E.,Hayakawa, Katsuyama, M., Khodzher, 1., T., Bashenkhaeva, 1.,Sorokovikova, 1.,Dorbunova, N.,Korovvakova, 2007.Distribution and characteristics organic of molecularsizefractionsof freshwaterdissolved matterin watershed environments: its implication Limnology 8,29-44. to degradation. 2t4