This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
RESEARCHPAPER
Rotifer species richness along analtitudinal gradient in the Alpsgeb_556 895..904
Ulrike Obertegger1*, Bertha Thaler2 and Giovanna Flaim1
1IASMA Research and Innovation Centre,
FEM, Environment and Natural Resources
Area, I-38010 San Michele all’Adige (TN),
Italy, 2Provincial Environmental Protection
Agency, Bozen, I-39055 Leifers (BZ), Italy
ABSTRACT
Aim Biodiversity patterns along altitudinal gradients are less studied in aquaticthan terrestrial systems, even though aquatic sites provide a more homogeneousenvironment independent of moisture constraints. We studied the altitudinalspecies richness pattern for planktonic rotifers in freshwater lakes and identified theenvironmental predictors for which altitude is a proxy.
Location Two hundred and eighteen lakes of Trentino–South Tyrol (Italy) in theeastern Alps; lakes covered 98% (range 65–2960 m above sea level) of the altitudinalgradient in the Alps.
Methods We performed: (1) linear regression between species richness and alti-tude to evaluate the general pattern, (2) multiple linear regression between speciesrichness and environmental predictors excluding altitude to identify the mostimportant predictors, and (3) linear regression between the residuals of the bestmodel of step (2) and altitude to investigate any additional explanatory power ofaltitude. Selection of environmental predictors was based on limnological impor-tance and non-parametric Spearman correlations. We applied ordinary leastsquares regression, generalized linear, and generalized least squares modelling toselect the most statistically appropriate model.
Results Rotifer species richness showed a monotonic decrease with altitude inde-pendent of scale effects. Species richness could be explained (R2 = 51%) by lake areaas a proxy for habitat diversity, reactive silica and total phosphorus as proxies forproductivity, water temperature as a proxy for energy, nitrate as a proxy for humaninfluence and north–south and east–west directions as covariates. These predictorscompletely accounted for the species richness–altitude pattern, and altitude had noadditional effect on species richness.
Main conclusions The linear decrease of species richness along the altitudinalgradient was related to the interplay of habitat diversity, productivity, heat contentand human influence. These factors are the same in terrestrial and aquatic habitats,but the greater environmental stability of aquatic systems seems to favour a linearpattern.
We studied rotifer species richness in relation to altitude in 218
lakes in the Trentino–South Tyrol region in the eastern Alps
(Italy), sampled during the summers of 1996–2008 (Fig. 1).
These lakes ranged from 65 to 2960 m a.s.l. with an average of
four lakes every 100 m, and covered 98.3% of the Alpine altitu-
dinal gradient (value according to Körner, 2007). The lakes
sampled spanned a wide range of environmental variables
(Table 1) and reflected the altitudinal distribution of lakes in the
study area (Fig. 2). The range of the north–south and east–west
direction was limited to less than 2°; therefore these spatial
parameters were not considered as gradients, but were impor-
tant for accounting for spatial autocorrelation (see statistical
modelling below). Species lists and environmental data were
either based on published data (45 lakes: IASMA, 1996–2000;
Salmaso & Naselli-Flores, 1999; Cantonati et al., 2002; Cantonati
& Lazzara, 2006) or provided directly by the authors. Zooplank-
ton samples were generally taken from the deepest part of the
lake with a plankton net (50 mm) and were preserved in 4%
formalin or 20% alcohol. The coordinates of the lakes’ centroids
were available on three different coordinate systems: the Univer-
sal Transverse Mercator (UTM) 32N/WGS84 (South Tyrol), the
Gauss Boaga/Rome40 W (Trentino), and the lat/long/WGS84
(Trentino); all the geographic positions were harmonized in the
lat/long/WGS84 system applying a datum-shift correction using
seven parameters specific to the study area.
Data treatment and statistical modelling
Species richness on a continuous altitudinal scale was calculated
as the number of rotifer taxa found in each lake. Species richness
at different altitudinal resolutions was calculated as the mean
Figure 1 Location of the 218 lakes sampled in Trentino–SouthTyrol (Italy). Continuous lines represent major rivers, filled circlesrepresent sampling sites.
Figure 2 Altitudinal distribution of lakes in Trentino–SouthTyrol for 100 m intervals. Black bars are mapped lakes (n = 667)and white bars are sampled lakes (n = 218).
Table 1 Minimum (min), maximum(max), median and mean values for themajor predictors of lakes sampled (n =218).
and rational quadratic) by the Akaike information criterion and
selected that with the lowest value. After including the appro-
priate correlation structure, we selected environmental predic-
tors to find the most parsimonious model; selection was based
on comparison of nested models by an ANOVA test. We also
performed a GLM with quasi-Poisson distribution and included
spatial covariates in the model because in the former GLS it was
difficult to find an appropriate correlation structure to model
spatial dependence. Thirdly, the residuals of the most parsimo-
nious model were regressed on altitude to investigate if altitude
had any additional explanatory power.
For every model, we reported: (1) the partial regression slope
as a measure for changes in a specific predictor while keeping the
other predictors constant, and (2) the relative influence of pre-
dictors according to the formula in Quinn & Keough (2002);
this parameter allowed us to investigate the hierarchy of influ-
ence on the response variable because its value is independent of
the magnitude of different measurement units of predictors.
Multicollinearity of predictors was investigated by the variance
inflation factor (VIF). For OLS models, we reported the stan-
dard R2, while for GLS models and GLM we reported the
pseudo-R2 according to the recommendations of Buse (1973).
All analyses were performed in R (R Development Core Team,
2005).
RESULTS
In the OLS regression, species richness decreased with altitude
(R2 = 0.33, P < 0.001) (Fig. 3, Table 1); in the GLS, species rich-
ness also decreased with altitude, and the inclusion of spatial
covariates and modelling of the optimal residual structure
improved the explanatory power of the model (pseudo-R2 =0.42, P < 0.001) with an increase in species richness in the
southern and eastern directions (Table 2, Fig. 4). Furthermore,
the monotonic decrease of species richness with altitude was
consistent across all the different altitudinal band widths
(Fig. 5).
For the multiple regression analysis of species richness depen-
dent on environmental predictors, we selected lake surface area
(area), total phosphorus (TP), reactive silica (Si), sulphate (SO4),
nitrate (NO3), conductivity, and surface water temperature
Figure 3 Linear regression of square-root transformed speciesrichness (sqrt(species richness)) dependent on altitude (OLS:r = –0.58; P < 0.001). The solid line is least squares regression;dashed lines are the 95% confidence interval.
Table 2 Results of regression analysisof species richness (square-roottransformed) dependent on altitudewithout spatial covariates and withspatial covariates (north-south (NS) andeast-west (EW)).
Estimate refers to partial regression slope, error to standard error, and influence to relative influence(for explanation see Materials and Methods). All predictors were statistically significant at P < 0.001.
Hessen et al., 2007). Besides source–sink and rescue effects that
can inflate the estimate of species richness (Jones et al., 2003;
Grytnes & McCain, 2007), aquatic studies that show nonlinear
species richness–altitude patterns (Nyman et al., 2005; Jan-
kowski & Weyhenmeyer, 2006; Li et al., 2009) may also be biased
by unconsidered broad latitudinal and longitudinal gradients. In
fact, Rahbek (2005) points out that in studies covering extensive
geographical ranges, species richness along altitudinal gradients
is influenced by the different impact of historical and ecological
mechanisms along large latitudinal and longitudinal gradients.
Our study, conducted over a narrow geographical extent,
demonstrated that rotifer species richness decreased linearly
with altitude on a continuous scale. While Rahbek (2005)
underlines that monotonic decreasing patterns are rare when
considering altitudinal bands, in our study a linear pattern also
prevailed within altitudinal bands of different widths. We
suggest that this consistency across different scales further cor-
roborates the goodness of fit of the linear shape. In addition, the
inclusion of spatial covariates in the species richness–altitude
relationship improved the fit of the model (R2 = 0.33 versus
pseudo-R2 = 0.42). In fact, models that include spatial autocor-
relation have a better predictive power than models without it
because autocorrelation accounts for variance in species rich-
ness data (Currie, 2007).
But what does the geographic variable ‘altitude’ actually stand
for in relation to rotifer species richness? Recent research has
shown that altitudinal gradients cannot be attributed to a simple
universal explanation (Rowe, 2009). Based on multivariate
regressions (GLS and GLM), we discussed the importance of
environmental predictors and focused on the GLM because it
was the most appropriate one with respect to spatial dependence
of data. When excluding altitude as an environmental predictor,
we could show that area, Si, temperature and TP had a positive
effect on species richness, while NO3 had a negative one. In
Figure 4 Rotifer species richness in sampled sites: the diameterof closed circles corresponds to different classes of speciesrichness.
Figure 5 Rotifer species richness at different altitudinalresolutions. The range of altitudinal bands differs: (a) 100 mrange; (b) 200 m range; (c) 300 m range; (d) 400 m range (forexplanations see Materials and Methods).
Estimate refers to partial regression slope, error to standard error, and influence to relative influence (for explanation see Materials and Methods). Speciesrichness is square-root transformed in GLS; all other variables except for temperature and spatial directions were log-transformed. Error refers tostandard error. The GLM is based on a quasi-Poisson distribution. All predictors were statistically significant at P < 0.01 except for conductivity (P < 0.05)and TP (P = 0.06).Area is lake surface area, Si is reactive silica, Temp is surface water temperature, TP is total phosphorus and Cond is conductivity.
Ulten) on metamorphic bedrock had higher than expected
sulphate (range 50–95 mg l-1 SO4) and conductivity (range 132–
233 mS cm-1) values (Fig. 6). While alpine lakes were once con-
sidered pristine environments, a recent review on European
mountain lakes shows them to be impacted by atmospheric
deposition and global warming (Battarbee et al., 2009). The
melting of ice and rock glaciers leads to an increase of conduc-
tivity due to the release of solutes and pollutants such as sul-
phate, nickel and manganese previously immobilized in the ice
(e.g. Thies et al., 2007). We suggest that these extraordinarily
high values of conductivity and sulphate for lakes located on
metamorphic bedrock could be an early warning sign of the
impact of melting ice on high-altitude aquatic systems.
By our model, we explained 51% of the variability of species
richness using environmental predictors and spatial covariates.
While environmental predictors of the species richness–altitude
pattern were related to specific key factors such as habitat diver-
sity, productivity, heat content and human influence, spatial
covariates were linked to the geologically heterogeneous terri-
tory. Our study underlined how including both environmental
and spatial predictors can enhance hypothesis-driven consider-
ations on species richness. Further unexplained variability of
species richness might be attributed to unconsidered factors
such as inter- and intraspecific competition for food sources
(Walz, 1995), hydrology (Soranno et al., 1999; Obertegger et al.,
2007) and dispersal (Swadling et al., 2000).
The linear decrease of species richness along the altitudinal
gradient was successfully captured by the interplay of habitat
diversity, productivity, heat content and human influence. But
why do terrestrial studies mainly show a hump-shaped pattern
of species richness while aquatic studies tend to show a linear
pattern? The inclusion of anthropogenically disturbed habitats
can cause a hump-shaped pattern (Nogués-Bravo et al., 2008),
and we argue that this may not completely apply to aquatic
systems. Recent improvements in water quality, especially in
temperate lowland lakes, while not limiting non-point sources,
have greatly reduced direct pollution (Søndergaard & Jeppesen,
2007). Moreover, precipitation (Gotelli et al., 2009) or water
availability (Grytnes & McCain, 2007) are very decisive in con-
junction with temperature for the unimodal species richness
pattern. However, aquatic systems – to be considered as such –
must retain their aquatic status, along with their main physical
properties related to water. Obviously, aquatic systems are also
subject to change, but in contrast to terrestrial ones, water pro-
vides a relatively stable set-up of environmental conditions in
which biotic interactions take place (Reynolds, 1998; Lampert
& Sommer, 2007).These characteristics of freshwater ecosys-
tems may ultimately distinguish them from their terrestrial
counterparts with regards to the species richness–altitude rela-
tionship. We argue that the factors determining species
richness–altitude patterns tend to be the same in terrestrial and
aquatic habitats, but the greater environmental stability of
Figure 6 Conductivity (mS cm-1) in sampled sites with theunderlying bedrock: the diameter of closed circles corresponds todifferent classes of conductivity. Lakes with unexpectedly highsulphate and conductivity values are circled.