BONUS FERRYSCOPE year 1 report
Transcription
BONUS FERRYSCOPE year 1 report
FerryScope Summary of Annual Report 1 July 2014 – June 2015 Version 1.0 30.06.2015 Martin Boettcher, Brockmann Consult Jenni Attila, Mikko Kervinen, Seppo Kaitala, SYKE Tiit Kutser, EMI Project full title: Bridging the divide between satellite and shipborne sensing for Baltic Sea water quality assessment Project coordinator: Dr. Martin Boettcher, Brockmann Consult GmbH phone: ++49 4152 889 315 email: [email protected] Project applicants: Brockmann Consult GmbH (BC), Germany Finnish Environment Institute (SYKE), Finland Estonian Marine Institute (EMI), Estonia Grant identifier: BONUS call2012inno-38 FerryScope Annual Report 1 Brockmann Consult GmbH Table of Contents 1 Project FerryScope ..................................................................................................................... - 1 - 2 Work performed in 1st year of FerryScope ............................................................................... - 1 2.1 In situ Rrs data provision framework (WP1) ....................................................................... - 1 - 2.2 Innovative approaches in EO data interpretation (WP2) .................................................... - 2 - 2.3 Data assimilation engine (WP3) ........................................................................................ - 10 - 2.4 Service Deployment (WP4) ............................................................................................... - 11 - 3 Main results, achievements, and impact ................................................................................. - 11 - 4 References ............................................................................................................................... - 12 - v1.0, 30.06.2015 FerryScope Annual Report 1 Brockmann Consult GmbH ii v1.0, 30.06.2015 Brockmann Consult GmbH FerryScope Annual Report 1 1 Project FerryScope FerryScope aims at improving water quality assessment of the Baltic Sea by the combination of satellite data, time series of shipborne Rrs measurements, other in-situ data, and improved algorithms and models. The FerryScope project has started in July 2014. FerryScope is a BONUS project. The FerryScope objectives are • To provide quality assured, freely accessible NRT Rrs from ships of opportunity (SOOP) • • To improve the accuracy of EO products by data assimilation with shipborne Rrs To investigate the sources of uncertainty in EO products, particularly near the coast, and to improve the underlying algorithms using large volumes of aggregated optical observations • To familiarize users (managers, researchers) with the new information system and to streamline adoption of the system in national monitoring agencies • To develop an open framework for automated and continuous in-situ and satellite measurement ingestion, processing and provision • To develop the commercial service model on top of the open source technical model 2 Work performed in 1st year of FerryScope Main focus of the first year was on the in-situ data framework and the development of the spectral library for data interpretation. 2.1 In situ Rrs data provision framework (WP1) During WP1 an online OGC-compliant Web Feature Service for the Rflex in-situ data time series was developed. FerryScope in-situ data server serves a combination of remote-sensing reflectance (Rrs) and time-matched flow-through ferrybox data. The service can be accessed online at http://ferryscope.ymparisto.fi/Rflex/index.xhtml The interface is fully documented inFerryScope deliverable D1.2 Rflex WFS API Reference also accessible at http://ferryscope.org/wp-content/uploads/2014/10/FerryScope-D1.2-RflexAPIQuickStart-v1.0.4.pdf The set of measurements served is extended daily shortly after one of the two ferrys reaches a harbour. The Web Feature Service is a machine-to-machine interface. The provided web pages with a graphical user interface only provide limited capabilities compared to the WFS. A detailed description is given in [D1.2]. v1.0, 30.06.2015 -1- FerryScope Annual Report 1 Brockmann Consult GmbH Figure 2-1: Spatial distribution and temporal coverage of the FerryScope Rrs in-situ data time series (continuously extended) An OGC Web Feature Service client has been implemented to retrieve Rrs data from the FerryScope insitu data server. This client can be downloaded by users from the FerryScope web site (www.ferryscope.org). It is implemented in Python. The client is installed in the Calvalus processing environment at Brockmann Consult to provide all the Rrs data to the FerryScope Data Assimilation Engine where also the Earth observation satellite data is processed. After the FerryScope in-situ data service had been extended also to serve FerryBox data matched with Rrs measurements the client has been extended to retrieve this combined in-situ data records, too. 2.2 Innovative approaches in EO data interpretation (WP2) Remote sensing data can be interpreted in two main ways. The “classical” approach is developing bandratio-type or more sophisticated algorithm that describe the shape and magnitude of reflectance spectrum with one number and then studying the regression between these numbers and water quality parameters (like chlorophyll, turbisity, etc.). These statistical algorithms are computationally simple and easy to apply. However, the relationships between the band-ratio-type algorithms and water parameters are varying from site to site (i.e. there is need in tuning the algorithms to local conditions) or the algorithms may not describe the water reflectance spectra at all (e.g. blue to green band ratios do not work in optically complex waters). More innovative approaches use the full spectrum measured by a remote sensing sensor and retrieve either IOPs (absorption and backscattering coefficients) or concentrations of optically active substances (CDOM, Chl, TSM) simultaneously. In principle, both measured in situ reflectance spectral libraries and modelled spectral libraries can be used. For example SIOCS (The Sensor-Independent Ocean Colour Processor) retrieves concentrations of optically active substances from reflectance data based on a LUT derived from a spectral library. Using only Rflex and ferrybox data collected during the FerryScope project would be biased towards the open parts of the Baltic Sea and towards the summer -2- v1.0, 30.06.2019 Brockmann Consult GmbH FerryScope Annual Report 1 season. Therefore, we used also a modelling approach that allowed us to create a spectral library that covers concentrations of optically active substances that occur in coastal waters and represents both spring and summer conditions. We use both “classical” approach and the LUT approach in the FerryScope project in order to find optimal methods for interpreting remote sensing data. 2.2.1 WP 2.1 Developing of data quality filters (identification of glint, processing errors, and other artefacts) Rflex spectra collected in the frame of the FerryScope project (together with the water quality parameters collected simultaneously by ferybox systems) can be used as an in situ spectral library for processing satellite data, as a reference data needed for evaluating the performance of atmospheric correction, or as a remote sensing database on its own right that can be used to retrieve water quality parameters from it in case the ferrybox data is missing (i.e. if the ship of opportunity has only Rflex system on board). In all these cases high quality reflectance data is needed. The Rflex system is designed to collect data under angles that should provide the best quality and avoid artefacts like glint or ship shadows. Nevertheless, the Rflex data has to be checked carefully and all suspicious data has to be removed. In this task we investigated several methods how to improve filtering of the Rflex data in order to provide output reflectances we are confident in. The automated Rflex filtering utilizes methods described in Simis and Olsson (2013). Additional rules below are used for removing bad Rrs spectra out of the measured Rflex dataset. Rrs spectra are filtered based on the spectral information only. R400<RPEAK> RNIR(800) = no anomalies White errors removed only with threshold Rflex(λ400 )< 0.001 Rflex wavelengths used for filtering: λNIR =(λRrs >= 795 nm & λRrs < 810 nm ) λPEAK = (λRrs >= 570 nm & λRrs < 590 nm ) λ400 =(λRrs >= 395 nm & λRrs <= 405 nm) λ500 = (λRrs >= 495 nm & λRrs <= 505 nm ) λRED = (λRrs >= 648 nm & λRrs <= 672 nm ) λ762 .2= (λRrs == 762.2nm ) λ765.5 = (λRrs ==765.5nm ) λ450 = (λRrs >= 445nm & λRrs <= 455nm) λ650 = (λRrs >= 645nm & λRrs <= 655nm) Filtering rules: • Rrs (λPEAK) > Rrs (λ650) AND • • Rrs (λ450) > Rrs (λ400) AND Rrs (λNIR) < 0.001 AND • • Rrs(λRED)/ Rrs (λNIR) > TH AND Rrs (λ400)< Rrs (λ500) AND • • Rrs(λ765.5 )/Rrs(λ762 .2)< 1.015) TH = 1 -1.5 (varies according to season) Figure 2-2 shows the resulting spectras after using the filtering rules above. Note that the Figures show the wavelength range ( RRS >= 400 nm & RRS <= 900 nm). v1.0, 30.06.2015 -3- FerryScope Annual Report 1 Brockmann Consult GmbH Figure 2-2: Examples of filtered Rflex spectra on different periods of annual measurement period. a) Spring cases a) 20.4.2014 and b) 13.4.2014, c) summer minimum 6.6.2014 and d) cyanobacteria bloom period 21.7.2014. Median is taken at based on wavelength 2.2.2 Parameterising HydroLight radiative transfer model for simulation of Baltic Sea Rrs spectra As was mentioned above, collecting the spectral library that contains examples from spring and summer bloom conditions and from varios coastal waters is time consuming and expensive even if we have such powerful data collection system like the Rflex-ferrybox pair. For example, extreme blooms occur with such frequency that capturing them within a two-year project is unrealistic. Therefore, it is reasonable to create a modelled spectral library. For example, using the HydroLight radiative transfer model. In the FerryScope project the parameterization of the Hydrolight model was done based on extensive field campaigns on the open Baltic Sea and on coastal waters of Estonia and Sweden. The procedure is described in detail in D2.1. The outcomes of this work are Look-upTables that can be futher utilized together with SIOCS (The Sensor-Independent Ocean Colour Processor) processor, see 2.2.5 below. Specific inherent optical properties of Estonian and Swedish coastal waters proved to be extremely variable and often different from the SIOPs of open parts of the Baltic Sea. Therefore, the D2.1 contains a modelled spectral library for the open parts of the sea and for two seasons with distinctly different SIOPs as well as in situ spectral libraries for near coastal waters. Analysis on the near coastal SIOPs are still work in progress and it is not determined yet in which parts of the Baltic Sea we can use the open sea spectral library and which part of the near coastal waters require different libraries for processing the remote sensing data. -4- v1.0, 30.06.2019 Brockmann Consult GmbH FerryScope Annual Report 1 2.2.3 Creating Baltic Sea Rrs spectral library for interpretation of satellite data The open Baltic Sea spectral library has been created using the procedure documented in D2.1 and given as a D2.2 LUT file. This spectral library is complemented with Excel files of the actual measumements made on ship cruises on the open Baltic sea and on coastal waters of Estonia and Sweden. The LUT delivered is compatible to SIOCS model and their joint use will be tested and utilized during the second year of the FerryScope project. The work on developing further the coastal model continues during the FerryScope project. b) a) Figure 2-3: Example of modelled spectras with a) varying pigment absorption (from 0.0061 to 2.27 with 14 steps), nadir viewing angle and sun angle 30o and b) Example of modelled spectras with varying b_part (from 0.01 to 10.013), nadir viewing angle and sun angle 30o. 2.2.4 Improving empirical algorithms for retrieving water characteristics from satellites Evaluating the performance of different band-ratio-type remote sensing algorithms requires statistically large amount of test data. We have reflectance data from 42 sampling stations in Estonian and Swedish waters that is accompanied by the concentrations of optically active substances as well as SIOPs and IOPs. Minimum and maximum concentrations of optically active substances in these sites are given in Table 2-1. Table 2-1: Concentrations of optically active substances observed in Estonian and Swedish coastal waters and used in testing the empirical algorithms Chl, mg/m3 TSM, mg/l aCDOM(400), m-1 Min 0.79 1.2 0.18 Max 9.03 8.8 3.92 The in situ database is too small to draw any major conclusions about suitability of the empirical algorithms. Therefore, we used the modelled spectral library created in the WP2 and described in the D2.2 in testing the empirical algorithms. The modelled LUT contains 655 200 simulated Rrs spectra in for spring and 561 600 for summer. The creation of spectral library and its input parameters have been documented in [D2.1] (Attila et al., 2015). The first version of spectral library is included in [D2.2]. Concentrations used in creating the spring and summer LUTs are shown in Table 2-2 and Table 2-3. Table 2-2: Concentration steps used in the spring simulations v1.0, 30.06.2015 -5- FerryScope Annual Report 1 Brockmann Consult GmbH Variable N Chl (µg l-1) 0.10 1.0 2.0 4.0 6.0 8.0 10.0 14.0 20.0 26.0 32.0 42.0 TSM(mg l-1) 0.05 0.4 0.8 1.3 1.8 2.3 4.0 5.7 7.4 8.9 18.0 50.0 aCDOM412(m-1) 0.05 0.2 0.3 0.5 0.7 0.9 1.2 1.5 3.0 20.0 84.0 250.0 14 12 10 Table 2-3: Concentration steps used in the summer simulations Variable N Chl (µg l-1) 0.10 0.8 1.7 2.5 3.3 4.2 6.2 8.2 10.2 12.8 26.0 120.0 12 TSM(mg l-1) 0.05 0.2 0.3 0.8 1.3 1.8 3.4 5.0 6.6 8.1 16.0 50.0 12 aCDOM412(m- 0.05 0.2 0.3 0.5 0.7 0.9 1.2 1.5 3.0 20.0 10 1 ) The band ratio algorithms used in this study were mainly chosen from those developed for optically complex coastal and inland waters rather than open ocean waters. Authors of the algorithms and equations are provided in Table 2-4. -6- v1.0, 30.06.2019 Brockmann Consult GmbH FerryScope Annual Report 1 Table 2-4: Summary table of the empirical remote sensing algorithms used in this study Reference Equation Regression type General form Code Chlorophyll Zimba, 2006 (1/R650-1/R710)*R740/0,0003+17,33 linear (1/R650-1/R710)*R740 CHL1 Moses ,2009a (R665e-1-R708e-1)*R753 linear (1/R665-1/R708)*R753 CHL2 Gitelson, 2009 (R670e-1-R710e-1)*R750 linear (1/R670-1/R710)*R750 CHL3 Mayo, 1995 0,164((R450_520-R620_690)/R530_610)^-0,98 power (R485-R660)/R570 CHL4 Hunter, 2008b log10(L710/L670) logaritmic log10(R710/R670) CHL5 Han, 2005 -9,5126+12,8315*(log R450_515/log R630_690) linear logR482.5/logR660 CHL6 Schalles, 1998 max(670-850)-670-850 line value at the location of maximum linear max(670-850)-670-850 line value at the location of maximum CHL7 Brezonik, 2005 -1,7237*L450_515/L630_690+9,6487 linear R482.5/R660 CHL8 Östlund, 2001 1409*(R525_605/(R450_515+R525_605+R630_690)-421,4 linear R565/(R482,5+R565+R660) CHL9 Wang, 2006 381,932-259,602*R630_690/R525_605 linear R660/R565 CHL10 Dierberg, 1994 151,6*L689_698/L673_685-114,6 linear R693.5/R679 CHL11 Duan, 2007 93,67*R700/R670-90,4 linear R700/R670 CHL12 Menken, 2006 5,91*(R700/R670)^4,96 power R700/R670 CHL12 Dierberg, 1994 174,5*R690_710/R673_687-156,6 linear R700/R670 CHL12 Kutser, 1999 89,8*L701/L673-64,1 linear R701/R673 CHL13 Kallio, 2001 L699_705/L670_677 linear R702/R674 CHL13 Koponen, 2007 166*L705/L663-106 linear R705/R663 CHL14 Ammenberg, 2002 85.01*R(705)/R(664)-51 linear R705/R664 CHL14 Kallio, 2003 108,5*(L700_710/L680_665)-68,7 linear R705/R673 CHL15 Kallio, 2003 112,1*(L700_710/L680_665)-77,1 linear R705/R673 CHL15 Kallio, 2001 L699_714/L670_685 linear R706.5/R677.5 CHL16 Kallio, 2001 L699_714/L661_667 linear R707.5/R664 CHL17 Moses, 2009a R708/R665 linear R708/R665 CHL17 Kallio, 2001 L705_714/L670_677 linear R709.5/R673.5 CHL18 Jiao, 2006 0,0282*(R719/R665)^3,0769 power R719/R665 CHL19 Härmä, 2001 (L705-L754)/(L665-L754) linear R730/R710 CHL20 Härmä, 2001 (L705-L775)/(L665-L775) linear R735/R720 CHL21 Moses, 2009b R748/R667 linear R748/R667 CHL22 Yacobi, 1995 -32,35+43,08*Rmax/R670 linear Rmax(670-850)/R670 CHL23 Schalles, 1998 sum 670-850 - sum 670-850 linear - % linear sum 670-850 - sum 670-850 linear - % CHL24 Total Suspended Matter Dekker, 2002 0,7517*e^65,736((R500_590+R610_680)/2) exponential (R545+R645)/2 TSM1 Dekker, 2002 0,7581*e^61,683((R525_605+R630_690)/2) exponential (R565+R660)/2 TSM2 Kutser, 1999 6,2*(Lmax-L750)/(L476-L750)-8,59 linear (Rmax-R750)/(R476-R750) TSM3 Kutser, 2014/2015?? 812-(770-840)base linear 812-(770-840)base TSM4 Wang, 2001 e^(5,6394+1,5493*ln((R630_690+R750_900)/(R450_515+R525_605)) logaritmic ln((R660+R825)/(R482,5+R565)) TSM5 Neukermans, 2009 38,02*R635/(0,162-R635)+2,32 linear R635/(0.162-R635) TSM6 Miller, 2004 -1,91+1140,25*M1 linear R645 TSM7 Doxaran, 2006 29,022 e^(0,0335 R750_900/R525_605) exponential R660/R565 TSM8 Wang, 2006 46,638-27,062*R630_690/R525_605 linear R660/R565 TSM8 Kallio, 2001 R699_705 linear R702 TSM9 Kallio, 2001 R699_705 - R747_755 linear R702-R751 TSM10 Koponen, 2007 1,47*L705+0,13 linear R705 TSM11 Ammenberg, 2002 174.8*R(705)-0.12 linear R705 TSM11 Thiemann, 2000 -52,9+73,6*R705/R678 linear R705/R678 TSM12 Härmä, 2001 L705-L754 linear R705-R754 TSM13 Kallio, 2001 R705_714 linear R709,5 TSM14 Doxaran, 2006 27,424 e^(0,0279 R790_890/R500_590) exponential R825/R565 TSM15 Doxaran, 2003 29,022*e^0,0335(R750_900/R525_605) exponential R825/R565 TSM15 Onderka, 2008 4,17*L(TM4)-43,22 linear R840/R545 TSM16 Doxaran, 2003 27,424*e^0,0279(R790_890/R500_590) exponential R840/R545 TSM16 Doxaran, 2003 18,895*e^0,0322(R790_890/R500_590) exponential R840/R545 TSM16 Doxaran, 2005 3.2846*(R850/R550)*100-7,3959 linear R850/R550 TSM17 Doxaran, 2002a 162,03x^3-394,45x^2+339,88x+1,027 polynomial R850/R550 TSM17 Doxaran, 2003 26,083*e^0,0365(R855/R555) exponential R855/R555 TSM18 Colored dissolved organic matter Brezonik, 2005 23,65-0,3528*L450_515-0,657(L450_515/L750_900) linear L482,5-0.657(L482,5/L825) CDOM1 Koponen, 2007 4,41*L663/L490-0,52 linear L663/L490 CDOM2 Doxaran, 2005 10,253*(R400/R600*100)^-0,9149 power R400/R600 CDOM3 Kallio, 2008 23,33*e^(-0,970 TM2/TM3) exponential R560/R660 CDOM4 Kallio, 2008 a*e^(b*TM2/TM3) exponential R560/R660 CDOM4 Kutser, 2005 5,13*R525_605/R630_690-2,67 power R565/R660 CDOM5 Ammenberg, 2002 5.894*R(664)/R(550)-1.53 linear R664/R550 CDOM6 Menken, 2006 R670/R571 linear R670/R571 CDOM7 v1.0, 30.06.2015 -7- FerryScope Annual Report 1 Brockmann Consult GmbH Analysis of the results indicate that there was no algorithm that perfprmed well in the case of in situ data. It does not mean necessarily that the algorithms are unsuitable in the Baltic Sea conditions. The reason may be the the number of samples is small (42) and the range of concentrations isnot representative for the whole Baltic Sea. Although the data from the sampling stations was carefully evaluated and stations with suboptimal data quality were excluded, there still may be measuring errors both in the case of reflectance measurements and further laboratory analysis. The modelled spectral library covered wider range of concentrations than has be published for the Baltic Sea in different studies we were abe to find. Also the number of samples used in the study was over 1.2 million. There were several band ratio type chlorophyll algorithms that had very high correlation (R2>0.8 and up to 0.97) with the chlorophyll concentration used in the model. Most of these algorithms used spectral bands in the reflectance peak near 700-710 nm and/or from the chlorophyll-a absorption feature near 660-680 nm. There were also several algorithms that produces good correlation with TSM. These algorithms used spectral bands either in the 700-750 nm range or hight of the reflectance peak near 810 nm. Three CDOM algorithms had correlation that was higher than 0.8. All of these algorithm used different red to green band ratios. This was not surprising as the reflectance in blue bands (where the CDOM influence on water colour is the highest) is negligible in such high CDOM waters like the Baltic Sea. 2.2.5 Developing analytical method for retrieving optically active substances concentrations from satellite data The Sensor Independent Ocean Colour processor SIOCS (under development by BC) is an Earth observation data processor for water quality that will provide spectral inversion algorithms for use with several optical satellite sensors. It retrieves water constituents from optical satellite data using sensorspecific lookup tables (LUTs) that relate spectra to water constituents. Such LUTs can be generated using radiative transfer models for the respective water type. The LUT’s described in D2.1 and delivered as D2.2 were tested with current version of SIOCS. Initial test results were made using MERIS/OLCI band configurations. The configuration of Baltic LUT and SIOCS model is under progress. Using Rflex dataset from April 2015, the first experiments using SIOCS processor and Baltic LUT have been generated. The results indicate that, using Rflex data, modified to MERIS band configurations, the Baltic LUT and current version of SIOCS gives pigment absorption estimates that can be used to derive realistic chlorophyll-a concentrations. The first test set contained Rflex reflectances at the end of spring chl-a bloom period. The simulations with Baltic LUT and SIOCS led to concentrations between 3.2 - 5.45 µg/l. The algorithm testing and further specification will continue in the WP3. The following figure shows the LUT browser tool for LUT inspection. -8- v1.0, 30.06.2019 Brockmann Consult GmbH FerryScope Annual Report 1 Figure 2-4: Part of the Baltic Sea LUT that has been derived from the model-generated Baltic Sea Spectral Library. The spectra ara shown graphically and in tabular form in the LUT browser for LUT inspection. During the second year of FerryScope project, the joint use of SIOCS and Baltic Sea LUT will be experimented and tested thoroughly in varying optical conditions typical for the Baltic Sea. It is foreseen, that both Baltic LUT and SIOCS will undergo updates before their full cababilities can be utilized. However, the combination of Baltic LUT and SIOCS is a first attempt to generate a Baltic Sea specific spectral inversion algorithm package that can be updated and modified for varying satellite and field instruments. 2.2.4 Adjustment of vicarious calibration using Rflex data – example with MODIS instrument data Prior to the launch of Sentinel3a OLCI (Ocean Land Colour Instrument) instrument, the available satellite instrument for testing the usability of Rflex data with Calvalus processing (match-up) system is MODIS (Moderate Resolution Imaging Spectroradiometer) by NASA. Unfortunately, MODIS is at the end of its lifetime and is degradating in time. The atmospheric correction by MODIS instrument on the Baltic Sea shows very poorly corrected reflectance spectra - thus resulting in nonrealistic chl-a concentrations [Darecki and Stramski 2004]. However, MODIS data allowed us to experiment Rflex data capabilities using an instrument with non-optimal radionmetric calibration. A method for adjusting the MODIS gain settings utilizing Rflex reflectance data as a reference was tested. First, initial tests were made using 4 sample dates. The method and more specific Baltic gain parameters can be futher optimized using Calvalus to generate automated Rflex and MODIS match-ups. Examples of filtered Rflex spectra and different settings for gain values for MODIS data for different seasons during the annual measurement period are given in Figure 2-5. Spring cases a) 20.4.2014 and b) 13.4.2014, c) summer minimum 6.6.2014 and d) cyanobacteria bloom period 21.7.2014. The atmospherically corrected MODIS data with original gain provided by SEADAS is given as green line, other colors represent adjusted gains. The first versions of the Baltic Sea adjusted gains are given in Table 2-5. Change is different for each MODIS band, varying between 0.2% to 3.2% from the original gain (original given on the first column). No changes were made for bands with central wavelength longer than 748 nm. Table 2-5: MODIS default gain (by NASA, available on Seadas) and Baltic adjusted versions v1.0, 30.06.2015 -9- FerryScope Annual Report 1 Brockmann Consult GmbH Gains/wavelenths 412 443 488 531 551 667 678 748 MODIS default 0.9731 0.9910 1.0132 0.9935 1.0002 0.9994 1.0012 1.0280 Baltic + SD 0.9858 1.0118 1.0472 1.0277 1.0358 0.9784 0.9902 1.0280 Baltic Sea gain 0.9809 1.0009 1.0416 1.0253 1.0322 0.9694 0.9812 1.0280 Baltic - SD 0.9760 0.9900 1.0359 1.0229 1.0286 0.9604 0.9722 1.0280 - 10 - v1.0, 30.06.2019 Brockmann Consult GmbH FerryScope Annual Report 1 Figure 2-5: Examples of filtered Rflex spectra and different settings for gain values for MODIS data. The examples represent different seasons during the annual measurement period. Spring cases a) 20.4.2014 and b) 13.4.2014, c) summer minimum 6.6.2014 and d) cyanobackteria bloom period 21.7.2014. The atmospherically corrected MODIS data with original gain provided by SEADAS is given as green line, other colors represent adjusted gains. 2.3 Data assimilation engine (WP3) In WP 3 FerryScope develops the software environment for data harvesting, processing, assimilation, and aggregation. This software environment is based on the Calvalus processing system platform at Brockmann Consult. As counterpart to the in-situ data service at SYKE the retrieval client for Rrs in-situ data and for FerryBox in-situ data has been integrated into the processing system. All in-situ data acquired by Finnmaid and Transpaper is available as input for FerryScope processing and validation on Calvalus. Newly acquired data can be retrieved. In preparation of the FerryScope processing chain the systematic harvesting and retrieval of Earth observation satellite data from MODIS and VIIRS has been configured and is operational. All MODIS L1A data since the beginning of 2015 and all VIIRS Level 1 data from beginning of March has been retrieved systematically and is available for FerryScope on Calvalus. This is a prerequisite for a later NRT water quality service of FerryScope. The time series of data is also required in FerryScope for algorithm development, validation and improvement. v1.0, 30.06.2015 11 - FerryScope Annual Report 1 Brockmann Consult GmbH The FerryScope processing chains on Calvalus automatically generate matchups between in-situ and satellite data. This will be combined with the different processors and processing steps to be integrated for FerryScope in the next period. Processing is currently being tested with the standard NASA processor SeaDAS with different parameterisations and the comparison of results with actual FerryScope Rflex data. 2.4 Service Deployment (WP4) This WP formally has not yet started except for preparatory work. A list of users and groups interested in FerryScope data and results is continuously extended. A request has been submitted to the European Maritime Day 2016 coordinators in order to have a FerryScope user workshop at that event. 3 Main results, achievements, and impact In agreement with the plan the WP1 (In-situ data framework) is finished, WP 2 (Earth observation data interpretation methods) and WP3 (Data assimilation engine) are ongoing. WP 4 (Service deployment) is planned to be started in January 2016. Figure 3-1: Deliverable D1.2 – detailed description how to access the in-situ Rrs data service of FerryScope The deliverables D1.1 and D1.2 have been delivered (and accepted). D1.1 is the in-situ data service itself that is accessible online (http://ferryscope.ymparisto.fi/Rflex/index.xhtml and http://ferryscope.org/) and continuously serves data that is acquired daily. D1.2 is the report that describes structure, content and access to the data. - 12 v1.0, 30.06.2019 Brockmann Consult GmbH FerryScope Annual Report 1 - Figure 3-2: Deliverable D2.1 – Description of the Spectral Library for the Baltic Sea Also deliverables D2.1 and D2.2 have been delivered (and accepted). D2.1 is the description of the Hydrolight model for the Baltic Sea. The document also contains the format description of the Baltic Sea Spectral Library. The Spectral Library itself is a database that is identified in D2.2. Scientific publications on the Spectral Library are in preparation. After their publication the Spectral Library will be made accessible online on the FerryScope web site. 4 References The following documents are referenced in this document. ID Title Issue Date [DoW] BONUS project FerryScope Document of Work, Brockmann Consult GmbH, Geesthacht 1.0 23.04.2014 [D1.1] Rflex WFS Service Web Layer, BONUS project FerryScope Deliverable D1.1, SYKE, Helsinki 1.0 13.04.2015 [D1.2] Rflex WFS API Reference, BONUS Deliverable D1.2, SYKE, Helsinki FerryScope 1.0 13.04.2015 [D2.1] Hydrolight Baltic, BONUS project FerryScope Deliverable D2.1, 1.2 SYKE, Helsinki and EMI, Estonia 30.06.2015 [D2.2] Baltic Sea Rrs Spectral Library, BONUS project FerryScope 1.0 Deliverable D2.2, SYKE, Helsinki and EMI, Estonia 30.06.2015 [D2.3] Algoithm Theoretical Basis Document forthcoming v1.0, 30.06.2015 project 13 - FerryScope Annual Report 1 [Darecki and Stramski 2004] Brockmann Consult GmbH Darecki M. and D. Stramski, 2004. An evaluation of MODIS and SeaWiFS bio-optical algorithms in the Baltic Sea. Remote Sensing of Environment, 89(3), 326-350. 2004 [Simis and Simis, S. G. H., and J. Olsson. 2013. Unattended processing of Olsson 2013] shipborne hyperspectral reflectance measurements. Remote Sens. Environ. 135: 202–212. [doi: 10.1016/j.rse.2013.04.001] 2013 [Simis et al. 2014] Stefan Simis, Jenni Attila, Mikko Kervinen: Automated hyperspectral remote sensing from ships-of-opportunity in the Baltic Sea: progress, system performance, and new services; presentation, 6th FerryBox Workshop 2014, Tallinn, Estonia (http://ferryscope.org/wpcontent/uploads/2014/10/SIMIS_rflex_baltic_FB2014.pdf) 09.09.2014 [Simis et al. 2015] Stefan Simis, Linhai Li, Mariano Bresciani, Claudia Giardino, Lin Li, Mark Matthews: Remote sensing of sun-stimulated fluorescence from phycobilipigments, oral presentation, Association for the Sciences of Limnology and Oceangraphy (ASLO), Aquatic Sciences Meeting, Granada 27.02.2015 [Kutser et al. 2015] Tiit Kutser, Stefan Simis, Martin Boettcher, Kari Kallio, Jenni Attila, Carsten Brockmann, Birgot Paavel, Martin Ligi, Mikko Kervinen, Seppo Kaitala: Improving the performance of remote sensing products in optically complex waters, poster presentation, Sentinel-3 for Science Meeting, Venice 04.06.2015 [Simis et al. 2015b] Stefan Simis, Jenni Attila, Mikko Kervinen, Philipp Grötsch: Phytoplankton products derived from automated along-track hyperspectral reflectance: do we need satellites? Seminar, University of Cape Town 05.06.2015 - 14 v1.0, 30.06.2019 Brockmann Consult GmbH FerryScope Annual Report 1 - v1.0, 30.06.2015 15 -