Biogeography of the golden-striped salamander Chioglossa

ECOGRAPHY 24: 618–624. Copenhagen 2001
Biogeography of the golden-striped salamander Chioglossa
lusitanica: a field survey and spatial modelling approach
J. Teixeira, N. Ferrand and J. W. Arntzen
Teixeira, J., Ferrand, N. and Arntzen, J. W. 2001. Biogeography of the goldenstriped salamander Chioglossa lusitanica: a field survey and spatial modelling approach. – Ecography 24: 618 – 624.
Chioglossa lusitanica is a streamside salamander endemic to the northwestern corner
of the Iberian Peninsula. In an extensive field survey we found the species to be
widely distributed across northwestern Portugal. The distribution data were used to
identify macroenvironmental parameters related to the presence of the species in
Portugal, and to produce a descriptive distribution model for C. lusitanica. We
modelled the species distribution based on five environmental parameters (precipitation, relief, July temperature, number of frost months and hardness of water) selected
by logistic regression analysis. The model yielded a high score of correct classification
(95%). Excluding parameters unavailable for Spain the model included four parameters (precipitation, relief, July temperature and altitude) and had a 93% correct
classification score. Extrapolation of the model to Spain yielded a correct classification of 92% and served as a validation of the modelling technique in an independent
area. The models indicate the existence of several potentially suitable areas outside
the known distribution of the species, providing clues about the species historical
biogeography. The models, through the identification of highly suitable and stable
protection regions, may contribute to informed conservation planning.
J. Teixeira ( [email protected]), N. Ferrand and J. W. Arntzen, Centro de
Estudos de Ciência Animal, Campus Agrário de Vairão, PT-4485 -661 Vairão, Portugal.
Modelling wildlife distributions with multivariate statistics and Geographical Information Systems (GIS) is
increasingly used to identify suitable habitat and to
predict species’ potential distributions (Pereira and
Itami 1991, Austin et al. 1996, Corsi et al. 1999). The
golden-striped salamander Chioglossa lusitanica is endemic to the Iberian Peninsula. It lives alongside uncontaminated brooks, mostly in mountainous areas
(Teixeira et al. 1999) and has a distribution restricted to
regions with precipitation of over 1000 mm yr − 1 and
altitudes B1500 m (Arntzen 1981). Chioglossa lusitanica is the only extant representative of its genus and
possesses some unusual characteristics, including a protractile tongue, autotomy of tail, absence of functional
lungs and a peculiar morphology, with a thin body and
a long tail comprising ca 2/3 of the total length. Several
of these character states are shared with its sister-spe-
cies, the Salamandrid Mertensiella caucasica from the
Caucasus and with some Nearctic Plethodontid salamanders, revealing a remarkable convergent adaptation
to streamside habitats (Arntzen 1994a).
Our understanding of the species’ historical biogeography has relied on a poor fossil record and on reconstructions from genetic data. The presence of C. meinii
in the Upper Oligocene of France and Lower Miocene
of Germany (Estes and Hoffstetter 1976) indicates that
representatives of the genus once had a wide distribution across central Europe. Pleistocene glacial and interglacial periods (1.8 million yr – 11 000 yr) are
frequently implicated in the patterns of genetic variation observed in both animals and plants, presumably
resulting from range contractions and expansions
(Taberlet et al. 1998, Hewitt 1999). In C. lusitanica the
depth and spatial pattern of genetic variation suggest
Accepted 6 February 2001
Copyright © ECOGRAPHY 2001
ISSN 0906-7590
Printed in Ireland – all rights reserved
618
ECOGRAPHY 24:5 (2001)
isolation in the early Pleistocene of populations separated by the Mondego river valley in the south of its
present-day range (Alexandrino et al. 1997, 2000).
Despite some detailed ecological studies (Goux 1957,
Arntzen 1981, 1994a, b, 1995, Vences 1990) the distribution of the species was until recently (Sequeira et al.
1996) poorly documented. Accordingly, C. lusitanica
has the conservation status of ‘‘insufficiently known’’ in
Portugal (Anon. 1990), ‘‘rare’’ in Spain (Blanco and
González 1992) and ‘‘indeterminate’’ in the IUCN red
list (Anon. 1994). We here describe the biogeography of
C. lusitanica in Portugal using the complementary approaches of field surveying and spatial modelling. The
distribution data were used to identify macroenvironmental parameters related to the presence of the species
in Portugal, and to produce a descriptive distribution
model for C. lusitanica. The model was then applied to
predict the species probability of occurrence across the
Iberian Peninsula and to improve the understanding of
its historical biogeography.
Methods
Distribution
The field survey was conducted in Portugal from 1994
to 1998 on a UTM 10 ×10 km grid basis. In each grid
cell up to four potentially suitable habitats were surveyed for 45 min. Adults and other post-metamorphic
salamanders were searched for under stones, litter and
moss, while larvae were dipnetted from the backwater
sections of brooks. A preliminary survey was restricted
to the area circumscribed by 44 known records (Crespo
and Oliveira 1989). The full survey encompassed all
areas: 1) identified as ‘‘suitable’’ in preliminary analyses
(probability of occurrence \ 0.8 – including the mountains of Montesinho, Malcata, Aire e Candeeiros, Sintra and Monchique); 2) such that the documented range
was delimited by a border of grid cells in which the
species remained unrecorded. In total 374 localities
were inspected over 281 grid cells.
Environmental data
Twenty-one ecologically meaningful variables were selected for analysis (Table 1). An altitude map was taken
from the US Geological Survey (http://edcwww.cr.usgs.
gov/doc/edchome/ datasets /edcdata.html) and used to
produce a relief map using a set of filter operations
(Anon. 1997a). A vegetation index map (NDVI, Normalised Difference Vegetation Index) was obtained
courtesy of the Royal Dutch Meteorological Institute
(KNMI). Maps on the mean January and July temperature were digitised from the Portuguese Climate Atlas
(Anon. 1974). Information on the remaining 16
parameters was available in digital form at DGA
(Anon. 1995). Nonparametric correlation (rs) and hierarchical clustering were used to evaluate the level to
which information appeared redundant. Thus, for
parameters correlated at rs \ 0.8 a single variable was
Table 1. Environmental parameters used to model the distribution of C. lusitanica in Portugal.
Variable
Code
Selected for
analysis
Available for
Spain
G-test
Sources
Acidity of the soil (pH)
Altitude (m)
Chlorates content of subterranean water
(Cl− mg l−1)
Water drainage (mm yr−1)
Evapotranspiration (mm yr−1)
Frost days (number)
Frost months (number)
Hardness of subterranean water (CaCO3
mg l−1)
Humidity of the air (%)
Insolation (h)
Lithology
Vegetation index
ACID
ALTI
CHLO
YES
YES
NO
NO
YES
NO
482.5
128.0
471.1
Anon. (1995)
http://edcwww.cr.usgs.gov/
Anon. (1995)
DRAI
EVAP
FROD
FROM
HARD
NO
NO
YES
YES
YES
NO
YES
YES
NO
NO
947.4
798.3
8.8
11.3
443.2
Anon.
Anon.
Anon.
Anon.
Anon.
HUMI
INSO
LITH
NDVI
YES
YES
YES
YES
YES
YES
NO
YES
41.0
548.9
217.4
11.9
Precipitation days (d yr−1)
Annual total precipitation (mm yr−1)
Solar radiation (kcal cm−2)
Residual content of subterranean water
(mg l−1)
Slope (%)
Sulphates content of subterranean water
(SO4−2 mg l−1)
Annual mean temperature (°C)
January mean temperature (°C)
July mean temperature (°C)
PRED
PRET
RADI
RESI
NO
YES
NO
NO
YES
YES
YES
NO
553.7
891.3
612.9
552.7
Anon. (1992), Anon. (1995)
Anon. (1992), Anon. (1995)
Anon. (1995)
Royal Dutch Meterol. Inst.
(KNMI)
Anon. (1992), Anon. (1995)
Anon. (1992), Anon. (1995)
Anon. (1992), Anon. (1995)
Anon. (1995)
SLOP
SULP
YES
NO
YES
NO
438.1
448.48
produced from ALTI
Anon. (1995)
TEMP
TJAN
TJUL
YES
NO
YES
YES
YES
YES
399.7
181.14
574.6
Anon. (1992), Anon. (1995)
Anon. (1974), Anon. (1992)
Anon. (1974), Anon. (1992)
ECOGRAPHY 24:5 (2001)
(1995)
(1992), Anon. (1995)
(1992), Anon. (1995)
(1995)
(1995)
619
retained, selected on the basis of: 1) data availability for
Spain, 2) promise in terms of known salamander life
history, and 3) ease of use. Thirteen parameters were
retained for spatial modelling (Table 1). Variables were
introduced into a Geographical Information System
(GIS; Idrisi 2.007 – Eastman 1997; Ilwis 2.1 Anon.
1997a) as raster layers with 1 km spatial resolution.
Values for UTM grid cells were mean (or modal, for
the categorical variable LITH) pixel values. Cells covering B50% Portuguese terrestrial territory were excluded from the analysis.
Analysis and modelling
We choose to analyse the distribution of C. lusitanica
by logistic regression because this technique: 1) has
been shown to be a powerful analytical tool, capable of
analysing the effects of one or several independent
variables, discrete or continuous, over a dichotomous
variable (Hosmer and Lemeshow 1989), 2) relies on
fewer statistical assumptions than its alternatives and 3)
frequently produces robust results (Austin et al. 1996,
Brito et al. 1999). Logistic regression takes the form:
p(x)=(eg(x)/1+eg(x)), in which p(x) represents the probability of occurrence of the target species and g(x)=
b0 +b1x1 +b2x2 +…+bpxp, where b0 is a constant, and
b1, b2, … bp are the coefficients of the variables x1,
x2, … xp included in the equation (Hosmer and
Lemeshow 1989). Logistic regression analyses were performed with SPSS 8.0 for Windows (Anon. 1997b) by a
forward stepwise addition with Bonferroni’s correction
to the initial a=0.05 (Rice 1989). The variables were
presented in the final equations by order of entrance in
the models. Individual variables were tested for explanatory power with G-statistics in univariate logistic
regression analysis.
To build distribution models from presence/absence
data a preponderance of absence data may be desirable
because of the larger environmental variation expected
in areas where the target species is absent (Pereira and
Itami 1991). This procedure, however, frequently produces biased results (Hosmer and Lemeshow 1989) and
we equilibrated the impact of presences and absences
through a weighting procedure (Anon. 1997b). The
correct classification score (CCS) denotes the percentage of correctly described presences and absences at
p(x) =0.5. The behaviour of the models was further
described by the omission error rate (number of presences where absence was predicted/total number of
predictions (N)) and the commission error rate (number
of absences where presence was predicted/N). Validation of the model derived for Portugal was achieved by
an extrapolation for Spain, on the basis of nine environmental variables for which data were available
(Table 1, Anon. 1992). Because the knowledge of the
species distribution in Spain is not numerically well
620
documented, we used the range of the salamander
(Arntzen 1999) to evaluate model fit.
Results
Chioglosa lusitanica was observed in 202 Portuguese
UTM grid cells, including 198 grid cells out of 869 in
the analysis (Fig. 1). The salamander was not found in
any of the isolated areas identified as potentially suitable in the preliminary model (the mountains of Montesinho, Malcata, Aire e Candeeiros, Sintra and
Monchique).
The spatial model included five variables as follows:
g(x)=5.499+0.009PRET−0.687HARD−0.799FROM
+0.082SLOP− 0.594TJUL. CCS was 95% with no
marked difference for presences and absences. The
omission error rate was 0.012 and the commission error
rate 0.168. The map derived from the model (Fig. 2)
matches closely with the documented distribution of the
salamander (Fig. 1). Areas classified as potentially suitable outside its documented range included the mountains of Montesinho, Malcata, Aire e Candeeiros,
Sintra and Monchique, in line with the results from the
preliminary analysis. The model shows an east to west
spatial constriction corresponding with the Mondego
river basin.
The model produced with the nine environmental
variables for which data were available for Spain was
g(x)=−0.017+0.009PRET+0.131SLOP−0.004ALTI
−0.521TJUL. CCS of this model for Portugal was
93%. The extrapolation of the model to Spain yielded a
CCS of 92% (98% of the species presence area and 91%
of the species absent area) with omission and commission error rates of 0.002 and 1.091, respectively. The
predicted range covers the known range for Spain
entirely. The model also classifies large areas as suitable
for the species, where its presence has not been documented, including a zone from the Cantabrian Mts up
to and including the Pyrenees, as well as several isolated mountain ranges of the Central System (Figs 1
and 3).
Discussion
The field survey markedly increased the knowledge of
the distribution of C. lusitanica in Portugal. The main
results are that: 1) the species’ range extends in the
south to near the Tejo river, and 2) the distribution is
continuous at the selected spatial scale. The species was
not found in five grid cells listed by Malkmus (1995),
three of which correspond to old and imprecise records
(e.g., ‘‘Coimbra’’ – Bocage 1864) and thus they may
have been allocated to an incorrect UTM grid cell.
Following Arntzen (1981) we consider the old record
ECOGRAPHY 24:5 (2001)
Fig. 1. A – Range of C.
lusitanica on the Iberian
Peninsula (adapted from
Arntzen 1999), B –
distribution of C. lusitanica
in Portugal. Solid dots
correspond to presences,
open dots to absences, and
grey dots to literature citings
covering cells in which the
species was not observed by
us.
for Elvas (Boscá 1880) erroneous because local habitat
characteristics are clearly unfavourable. We failed to
find the species in Sintra Mts where it was introduced
ca 60 yr ago (Seabra 1943) and last seen in 1993 (G.-D.
Guex in Arntzen 1999).
The distribution of C. lusitanica is well described by
a limited number of environmental parameters, although it should be kept in mind that the good fit of a
model does not necessarily imply correct inference of
causation (James and McCulloch 1990). Areas determined as suitable for C. lusitanica are hilly or mountainous with a mild summer and winter climate, high
precipitation and soft water. This set of parameters to
some extent confirms the empirically described dependence of the species to humid mountain areas (Goux
1957, Arntzen 1981).
Teixeira (1999), using different modelling techniques
(overlap analysis, discriminant analysis and classification trees), obtained similar results: precipitation appeared in all the models; slope, mean July temperature
and a soil-related characteristic (hardness of water,
acidity of soil or lithology) were consistently included,
and altitude and the number of frost months were
occasionally incorporated. As measured by their performance in univariate analysis the variables contribute to
explain individually the distribution of the species in
ECOGRAPHY 24:5 (2001)
order of PRET, TJUL, INSO, ACID, HARD and
SLOP.
In view of these results, precipitation appears to be
the major explicative variable for the occurrence of
golden-striped salamander. This is not surprising because precipitation determines most of the ecological
requirements of the species: humidity of the air and soil
moisture, vegetation composition and permanency and
water flow of the brooks. The high relief of the terrain
is frequently referred to as a typical characteristic of C.
lusitanica habitats (Vences 1997, Teixeira et al. 1999),
being often associated with the presence of clean and
fast-running brooks with little human disturbance.
Temperature also seems to play an important role in
the distribution of the species since it is absent from the
regions with mean July’s temperature above 22.5°C. In
fact, Goux (1957) reports critical thermal values for C.
lusitanica in captivity as 28°C for adults and 25°C for
larvae. In contrast, Salamandra salamandra has critical
temperature values of 35°C for adults and juveniles
(Degani 1994), which appears to be reflected in a much
wider distribution including the southern Iberian Peninsula. Temperature, directly or indirectly, also affects
other important environmental variables such as soil
moisture and air humidity, permanency of brooks and
concentration of dissolved oxygen in the water. Water
621
hardness serves as an indicator of soil chemistry, with
high values in calcareous regions. The influence of this
parameter on the distribution of the species may be
related with the lower capacity of these substrates to
retain water. Indeed, Jaeger (1971) suggested that soil
moisture affected the distribution of some American
salamander species.
An intriguing situation occurs in Spain where the
model suggests a zone of suitable habitats from Galicia
to the eastern Pyrenees. This discrepancy with the
known distribution may in part be caused by the absence in the Spanish model of the parameters HARD
and FROM. The eastern limit of the C. lusitanica range
appears to coincide with a transition to calcareous soils
(i.e. the species is absent in regions with hard water,
Vences 1997). Alternatively, either our model does not
locally apply because the distribution of the species is
limited by different variables in different geographical
regions (Brown and Lomolino 1998), or northern Spain
provides suitable habitat that is not yet occupied by the
species.
Fig. 2. Probability of occurrence of C. lusitanica in Portugal
as given by the logistic regression analysis.
622
The genetic structure of C. lusitanica suggests that all
northern populations are derived from central Portugal
with a source located in between the rivers Mondego
and Douro, and inferential, that the northward dispersal has taken place from the Pleniglacial (18 000 BP,
Alexandrino et al. 2000). If C. lusitanica has a low
dispersal rate, it would be plausible that the species is
still expanding its range eastwards. Although it can
move substantial distances (\ 350 m overnight), dispersal occurs in a highly directional manner, parallel rather
than perpendicular to water courses (Arntzen 1994b), a
behaviour that may restrict colonisation from one water basin to the other. Competition with other species
expanding from a different glacial refugium could also
limit its dispersal, a biogeographical scenario that seems
to operate between Lacerta schreiberi and L. 6iridis in
the region (Marco 1997). However, the one Iberian
amphibian that has similar habitat characteristics –
Euproctus asper – and could perhaps outcompete C.
lusitanica is confined to the Pyrenees and thus the two
species do not co-occur (Montori et al. 1997).
Several areas to the south and east of the present-day
distribution of C. lusitanica are also indicated in the
model as potentially suitable for the species (e.g.,
Monchique, Montesinho). While the unrecorded presence of C. lusitanica cannot be ruled out, this is unlikely
given the extensive surveying in these areas (Loureiro et
al. 1996, Teixeira 1997, present study). This indicates
the presence of barriers to dispersal rendering theses
areas out of range. Schreiber’s green lizard Lacerta
schreiberi is an Iberian endemism that shares with C.
lusitanica an affinity for mountain brooks and has a
similar overall distribution pattern (Brito et al. 1996,
1998). In contrast with C. lusitanica it is found in
several isolated southern regions, suggesting that its
range was wider in the past. Nonetheless, C. lusitanica
may also have possessed a wider range but gone extinct
in those areas. The rediscovery and subsequent genetic
investigation of C. lusitanica from Sintra might be
specially informative in this respect. Genetic variation
across the range of C. lusitanica is substantial (Alexandrino et al. 1997, 2000), providing the possibility to
distinguish between introduced and native occurrences.
If introduced, it would show the survival of a population over six decades, confirming the general suitability
of the area for the species, as predicted from the model.
If native, it would indicate that C. lusitanica had a
wider distribution in the past that reached Sintra.
The logistic regression models proved helpful in the
study of the biogeography of the golden-striped salamander, in particular to describe its present-day distribution, to formulate biogeographical hypotheses to
explain its absence in apparently suitable areas and to
plan fieldwork surveys. The model may also contribute
to informed conservation planning, through the identification of highly suitable and stable protection regions, and by simulating the effects of environmental
change.
ECOGRAPHY 24:5 (2001)
Fig. 3. Probability of
occurrence of C. lusitanica in
Iberian Peninsula as given by
the logistic regression analysis
performed with the
environmental variables
available for Spain.
Acknowledgements – We thank Pedro Beja, José Carlos Brito,
James Harris and Steve Weiss for comments and João Alexandrino, Fernando Sequeira and others (too many to mention
them all) for assistance in the field. The work was supported
by the Instituto da Conservação da Natureza, the ‘‘LIFE’’
program of the European Union, and by the Fundação para a
Ciência e Tecnologia (PRAXIS XXI /BM/12745/97 and
SFRH/BD/3376/2000 grants to JT, and PRAXIS XXI/BCC/
11959/97 grant to JWA).
References
Alexandrino, J., Ferrand, N. and Arntzen, J. W. 1997. Genetic
variation in some populations of the golden-striped salamander, Chioglossa lusitanica (Amphibia: Urodela), in Portugal. – Biochem. Genet. 35: 371–381.
Alexandrino, J. et al. 2000. Genetic subdivision, glacial refugia
and postglacial recolonization in the golden-striped salamander, Chioglossa lusitanica Amphibia: Urodela. –
Molecul. Ecol. 9: 771–781.
Anon. 1974. Atlas Climatológico de Portugal. – Serviço Climatológico Nacional.
Anon. 1990. Livro Vermelho dos Vertebrados de Portugal.
Vol. I Mamı́feros, Aves, Répteis e Anfı́bios. – SNPRCN.
Anon. 1992. Atlas Nacional de España. Sección II. Grupo 9
Climatologı́a. – Instituto Geografico Nacional.
Anon. 1994. IUCN red list categories. – IUCN Species Survival Commission.
Anon. 1995. Atlas do Ambiente. Versão 1.1. – Direcção Geral
do Ambiente. Ministério do Ambiente e Recursos
Naturais.
Anon. 1997a. Ilwis 2.1 for Windows. Reference Guide. – ITC.
Anon. 1997b. SPSS for Windows. Release 8.0. – SPSS.
Arntzen, J. W. 1981. Ecological observations on Chioglossa
lusitanica. – Amphib.-Reptilia 1: 187– 203.
Arntzen, J. W. 1994a. Allometry and autotomy of the tail in
the golden-striped salamander, Chioglossa lusitanica. –
Amphib.-Reptilia 15: 267–274.
Arntzen, J. W. 1994b. Speedy salamanders: sedentariness and
migration of Chioglossa lusitanica. – Rev. Esp. Herpetol. 8:
81 – 86.
ECOGRAPHY 24:5 (2001)
Arntzen, J. W. 1995. Temporal and spatial distribution of the
golden-striped salamander Chioglossa lusitanica along two
mountain brooks in northern Portugal. – Herpetol. J. 5:
213 – 216.
Arntzen, J. W. 1999. Chioglossa lusitanica Bocage 1864 – der
Goldstreifensalamander. – In: Grossenbacher, K. and
Thiesmeier, B. (eds), Handbuch der Reptilien und Amphibien Europas. Akademische Verlagsgesllschaft, pp. 301–
321.
Austin, G. E. et al. 1996. Predicting the spatial distribution of
buzzard Buteo buteo nesting areas using Geographical Information Systems and Remote Sensing. – J. Appl. Ecol.
33: 1541 – 1550.
Blanco, J. C. and González, J. L. 1992. Libro rojo de los
Vertebrados de España. – ICONA.
Bocage, J. B. 1864. Notice sur un bratrachien nouveau du
Portugal (Chioglossa lusitanica, Nob.). – Proc. Zool. Soc.
Lond. 264– 265.
Boscá, E. 1880. Catalogue des reptiles et amphibiens de la
Péninsule Ibérique et des Iles Baleares. – Bull. Soc. Zool.
Fr. 5: 240 – 287.
Brito, J. C. et al. 1996. Distribution of Schreiber’s green lizard
Lacerta schreiberi in Portugal: a predictive model. – Herpetol. J. 6: 43 – 47.
Brito, J. C., Paulo, O. S. and Crespo, E. G. 1998. Distribution
and habitats of Schreiber’s green lizard Lacerta schreiberi
in Portugal. – Herpetol. J. 8: 187 – 194.
Brito, J. C., Crespo, E. G. and Paulo, O. S. 1999. Modelling
wildlife distribution: logistic multiple regression vs overlap
analysis. – Ecography 22: 151– 260.
Brown, J. H. and Lomolino, M. V. 1998. Biogeography. –
Sinauer.
Corsi, F., Duprè, E. and Boitani, L. 1999. A large-scale model
of wolf distribution in Italy for conservation planning. –
Conserv. Biol. 7: 150– 159.
Crespo, E. G. and Oliveira, M. E. 1989. Atlas da distribuição
dos Anfı́bios e Répteis de Portugal Continental. –
S.N.P.R.C.N.
Degani, G. 1994. Ecophysiology of Salamandra salamandra at
the southern limit of its distribution. – Mertensiella 4:
111 – 124.
Eastman, J. R. 1997. Idrisi for Windows. User’s guide version
2.0. – Clark Univ.
623
Estes, R. and Hoffstetter, R. 1976. Les urodéles du Miocéne
de la Grive-Saint-Alban Isère, France. – Bull. Mus. Nat.
Hist. Nat. 398: 299–343.
Goux, L. 1957. Contribuition a l’étude écologique et biogeographique de Chioglossa lusitanica Barb. Urodela, Salamandridae. – Bull. Soc. Zool. France 82: 361–377.
Hewitt, G. M. 1999. Post-glacial re-colonization of European
biota. – Biol. J. Linn. Soc. 68: 87–112.
Hosmer, D. W. and Lemeshow, S. 1989. Applied Logistic
Regression. – Wiley.
Jaeger, R. G. 1971. Moisture as a factor influencing the
distributions of two species of terrestrial salamanders. –
Oecologia 6: 191–207.
James, F. C. and McCulloch, C. E. 1990. Multivariate analysis
in ecology and systematics: panacea or pandora’s box? –
Annu. Rev. Ecol. Syst. 21: 129–166.
Loureiro, A. et al. 1996. Atlas Herpetológico da Serra da
Malcata: dados preliminares. – IV Congr. Luso-Espan.
Herpetol. Porto.
Malkmus, R. 1995. Die Amphibien und Reptilien Portugals,
Madeiras und der Azoren. – Westarp Wissenschaften.
Marco, A. 1997. Lacerta schreiberi Bedriaga, 1878. Lagarto
verdinegro. – In: Pleguezuelos, J. M. (ed.), Distribución y
biogeografı́a de los anfı́bios y reptiles en España y Portugal. Univ. de Granada and A.H.E., Granada, pp. 228– 230.
Montori, A. et al. 1997. Euproctus asper (Dugès, 1852). Tritón
pirenaico. – In: Pleguezuelos, J. M. (ed.), Distribución y
biogeografı́a de los anfı́bios y reptiles en España y Portugal. Univ. de Granada and A.H.E., Granada, pp. 106– 108.
Pereira, J. and Itami, R. 1991. GIS-based habitat modelling
using logistic multiple regression: a study of the Mt. Gra-
624
ham red squirrel. – Photog. Eng. Rem. Sens. 57-11: 1475–
1486.
Rice, W. R. 1989. Analyzing tables of statistical tests. –
Evolution 43: 223 – 225.
Seabra, A. F. 1943. Apontamentos sobre a fauna do Algarve
Vertebrados. – Mem. Estud. Mus. Zool. Univ. Coimbra
147: 1– 18.
Sequeira, F. et al. 1996. Distribución de Chioglossa lusitanica
(Bocage, 1864) en Portugal. – Bol. Asoc. Herpet. Españ. 7:
7 – 8.
Taberlet, P. et al. 1998. Comparative phylogeography and
postglacial colonization routes in Europe. – Molecul. Ecol.
7: 453 – 464.
Teixeira, J. 1997. Contribuição para o conhecimento da ecologia e distribuição da herpetofauna do Parque Natural de
Montesinho. – Degree thesis, Univ. of Porto.
Teixeira, J. 1999. Biogeografia de salamandra-lusitânica,
Chioglossa lusitanica. Utilização de Sistemas de Informação Geográfica na modelação da distribuição. – M.Sc.
thesis, Univ. of Porto.
Teixeira, J. et al. 1999. Bases para a Conservação da Salamandra-lusitânica Chioglossa lusitanica. – Estud. Biol. Conserv. Nat. ICN. 22.
Vences, M. 1990. Untersuchungen zur okologie, etologie und
geopgraphischen variation von Chioglossa lusitanica Boc.
1864. – Salamandra 26: 267– 297.
Vences, M. 1997. Chioglossa lusitanica Barboza du Bocage,
1864. Salamandra rabilarga. Salamandra-dourada. – In:
Pleguezuelos, J. M. (ed.), Distribución y biogeografı́a de
los anfı́bios y reptiles en España y Portugal. Univ. de
Granada and A.H.E., pp. 103– 105.
ECOGRAPHY 24:5 (2001)