SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Composition of animal communities can be shaped by both local and regional processes. Among others, dispersal of organisms links local and regional patterns and determines the similarity of communities at increasing spatial distances. Unique and shared spatial and environmental contributions to fish community composition were calculated for watercourse distances between 49 hydrologically connected lakes in the German lowland area. Variation partitioning indicated a dominant unique effect of spatial predictors on fish community composition, whereas the effects of lake morphometry and productivity were weaker. The spatial effect was attributable to an uneven occurrence of small, littoral fish species found even at the small spatial extension covered here (maximum spatial distance ˜550 km). Distance decay of community similarity was moderate, but significant, if all 31 fish species were considered, but the slope of the decay function became steeper if only 11 small-sized, primarily littoral species were included. These results suggest that fish in European lowland lakes can be considered a metacommunity with limited dispersal along watercourse connections in particular for small-sized species. The analysis supports that for an appropriate evaluation of spatial effects on fish community similarity, reliable estimates of local richness are required which include in particular also rare, small-sized species occurring primarily in littoral areas. Furthermore, watercourse distance is a more reliable approximation than Euclidean distance to the real spatial dimension of fish dispersal.

Metacommunity theory suggests that composition of animal communities is affected by both local interactions and regional processes (Leibold et al. 2004, Logue et al. 2011). Local interactions among species and between species and their environment are linked with regional processes via dispersal. Although the metacommunity concept is not new, recent studies have focused on the patterns and spatial scales at which organisms disperse and thus link local communities. Several traits are important in controlling dispersal range, a process which determines how strongly similarity of local communities decays with increasing distance (Nekola and White 1999). Above all, size of organisms and their dispersal mode interact in determining the steepness of this distance decay of similarity. In passive dispersers whose propagules are distributed by vectors, the dispersal rate is negatively correlated with the size of propagules (Hillebrand et al. 2001, De Bie et al. 2012). However, for active dispersers, dispersal range is assumed to be positively correlated with body size (Jenkins et al. 2007, Shurin et al. 2009), leading to slower distance decay in larger organisms. In aquatic systems, many organisms can disperse only through hydrological connections and hence should show strong effects of dispersal limitation (De Bie et al. 2012). Evidence for limited dispersal among aquatic taxa has been found for communities of fish, benthic macroinvertebrates and zooplankton which often exhibit a pronounced distance decay of similarity (Soininen et al. 2007, Shurin et al. 2009, Astorga et al. 2012).

However, simple correlations between community similarity and geographical distance such as distance decay are not entirely appropriate to elucidate the effect of dispersal limitation, because they cannot disentangle spatial and environmental drivers of similarity (Logue et al. 2011). An alternative statistical approach to tackle this problem is variation partitioning by which the unique effects of spatial and environmental predictors and the confounding effects from both predictor classes can be calculated (Griffith and Peres-Neto 2006, Peres-Neto et al. 2006). A unique spatial effect may reflect the amount of variation that is caused by dispersal limitation (Legendre and Legendre 1998). Recent studies using variation partitioning have demonstrated a relatively small, but statistically significant unique effect of spatial variables on similarity of aquatic communities (Beisner et al. 2006, Hajek et al. 2011, De Bie et al. 2012). However, the degree to which spatial distance is an important predictor of local community composition still remains to be determined.

In this study, we aimed to determine whether spatial distance contributes to similarity between fish communities in German lowland lakes. Local environmental determinants of fish community composition and abundance are well studied in this area, and encompass primarily effects of lake productivity (Mehner et al. 2005, Garcia et al. 2006) and morphometry (lake area and depth) (Diekmann et al. 2005, Mehner et al. 2007). However, spatial predictors of fish community composition have not yet been explored. Lakes in the European lowlands have been formed only after last glaciation about 12 000 yr ago, and hence dispersal limitation may still play an important role for local colonization of lakes and their fish community composition, as demonstrated for North-American postglacial areas (Magnuson et al. 1998, Knapp et al. 2001). On the other hand, spatial distances between the lakes within a single European ecoregion are usually small (< 1000 km) relative to the pan-European distances (up to 4000 km) at which strong effects of dispersal limitation on fish community composition have been found (Leprieur et al. 2009, Schleuter et al. 2012). Furthermore, the method to calculate spatial distances may modify the estimated strength of dispersal limitation. Distances between lakes are conventionally expressed by Euclidean distance between geographical coordinates as an approximation to potential dispersal ranges (Shurin et al. 2009, Astorga et al. 2012). However, because fish can actively disperse only through aquatic environments, watercourse distance may be a superior approximation of spatial scales at which fish can disperse (Olden et al. 2001, Beisner et al. 2006). Unfortunately, watercourse distances for a network of connected lakes have not yet been available from the European lowlands.

We addressed the interplay between local environmental factors and spatial descriptors on fish community composition of German lowland lakes using two approaches. First, we ran redundancy analyses and variation partitioning on a set of connected lakes to elucidate the contribution of spatial predictors to local fish community composition. To apply realistic spatial dimensions along which fish can disperse, we calculated the spatial matrix between the lakes along watercourse distances. We kept local morphometric and productivity-related descriptors separate to facilitate calculation of unique contributions from each of the predictor classes, and to elucidate spatial patterns of lake morphometry and productivity. Second, we calculated distance decay functions for subsets of fish species differing in size and primary habitat preference. We hypothesized that small-sized species would show stronger dispersal limitation than larger-sized species, suggesting that fish communities in German lowland lakes can in part be considered as metacommunity.

Methods

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Data were assembled for 49 natural lakes, all situated in north-east Germany in the European ecoregion ‘Central Plains’ with a maximum elevation < 200 m a.s.l. (Illies 1978) (Fig. 1). The regional species pool is identical for all lakes. The lakes have not been (legally) stocked with non-resident fish species in recent years. Details of the lakes and their fish assemblages have been published earlier (Mehner et al. 2005, Emmrich et al. 2011). Fish were caught by stratified random sampling with benthic and pelagic multi-mesh gillnets (12 or 11 mesh sizes between 5.0 and 55 mm, respectively, the 5.0 mm panel missing in the pelagic nets) according to the European standard (CEN 2005). In short, the total number of benthic nets applied per lake is defined by lake area and depth, and these nets were set randomly in all depth strata of the lake. In lakes deeper than 6 m maximum depth, additional pelagic nets were set over the total water depth at the deepest point of the lake (for more details, see Diekmann et al. 2005, Mehner et al. 2005). In addition, lake nearshore areas which are usually not effectively sampled by gillnetting, were sampled by electrofishing. Accordingly, the accumulated species presence–absence lists per lake were highly reliable and also included those small-sized species which occur predominantly in shallow littoral zones of lakes (Diekmann et al. 2005). Presence–absence data may strongly reflect dispersal limitation and its effect on species distribution and are less confounded by local predictors such as lake productivity than relative abundances of species (Declerck et al. 2011).

image

Figure 1. Geographical location of the lakes and their drainage systems (Elbe, Oder) studied in northeast Germany with 49 completely connected lakes (black dots).

Download figure to PowerPoint

A matrix of spatial distances between the lakes was assembled by determining the shortest watercourse distances from available maps in ArcGIS (ver. 10, ESRI 2011, Redland, CA, USA). Lakes were considered connected only if the dimension of the connecting watercourse was large enough (usually > 0.5 m wide and > 0.3 m deep) to allow dispersal of fish. Recent persistence of connectivity was checked by aerial photographs, GoogleEarth and local inspections. These lakes all belong to the Rivers Elbe or Oder drainages (Fig. 1) and are completely connected to each other, either via natural slow-flowing rivers or by two artificial waterways (channels) which connect these two drainages. Whereas the channels have been constructed only during the recent 200 yr, the connectivity within the watersheds is primarily the result of postglacial geographic processes. We cannot exclude, however, that a few of the currently connected 49 lakes have not been connected for some periods within the last 12 000 yr, for example by interruption of small connections during longer dry periods.

In addition, we collected data on six descriptors of environmental characteristics for the 49 connected lakes. To avoid overfitting, we limited the choice of variables to a defined subset for which the effect on fish communities in lowland lakes has been demonstrated (Mehner et al. 2005, 2007, Emmrich et al. 2011). Lake morphometry was described by lake area (arithmetic mean = 567 ha, range 50–11 300 ha), mean depth (mean = 7.4 m, 0.9–28.6 m) and drainage area (mean = 270 km², 1.9–7500 km²). Lake productivity was expressed by eplimnetic concentration of total phosphorus during summer (measured in unfiltered homogenized water samples, mean = 70 μg l−1, 17–330 μg l−1), transparency measured as Secchi depth during late summer sampling (mean = 2.53 m, 0.15–9.4 m), and pH (mean = 8.39, 7.6–9.3).

Statistics

Fish species presence/absence lists per lake were Hellinger transformed by taking the square root after standardizing the rows (species) to totals (Legendre and Gallagher 2001). After this transformation, the subsequent multivariate procedures such as redundancy analysis were based on the Hellinger distance which is more appropriate for community composition data than Euclidean distance, and provides unbiased estimates for variation partitioning (see below). All environmental predictors except pH were log10-transformed.

From the spatial distance matrix, we calculated Moran’s Eigenvector Maps (Borcard et al. 2004, Dray et al. 2006), based on centered truncated connectivity matrices (Griffith and Peres-Neto 2006). The resulting positive eigenvectors were used as linear spatial predictors of fish species composition. We tested for the effect of spatial distance (watercourse-based eigenvectors), lake morphometry (three predictors) or lake productivity (three predictors) on fish species composition by separate redundancy analyses (RDA) with subsequent significance tests at variable number of permutations (as conducted by the RDA function in R, see below). Furthermore, we tested by RDA whether gradients of lake morphometry or productivity were directly spatially structured.

Finally, we applied variation partitioning based on RDA (Peres-Neto et al. 2006). Variation partitioning decomposes total species composition into a unique morphometric component, a unique productivity-related component, a unique spatial component, components reflecting spatially structured morphometric or productivity variations, and the remaining residual variation. The magnitude of the unique spatial component reflects the amount by which fish dispersal may be limited by spatial distance (Legendre and Legendre 1998). Explained variation of unique and shared components were corrected for the number of predictor variables and sampling sites as expressed by the adjusted R², and their significance tested by 999 permutations (Peres-Neto et al. 2006). These adjusted R²-values were displayed by a Venn diagram.

Estimates of distance decay of community similarity over spatial distances based on (log)linear regressions are strongly influenced by study extent and grain size (overall spatial dimension and smallest homogeneous sample unit, respectively), and are therefore not directly comparable between differing datasets (Steinbauer et al. 2012). Alternatively, distance decay can be reliably estimated by generalized linear models (GLM) with binomial observation error and a loglink function (Millar et al. 2011). The equation takes the form

log(E[s]) =a–βd

with log(E[s]) being the expected similarity, d being the spatial distance, and a and β being the parameters of the loglinear model (Millar et al. 2011). Parameter estimates and their standard errors for a, β, and the similarity at zero distance (s0) and the halving distance (hd, km) were obtained by a leave-one-out jackknife procedure (Millar et al. 2011). The dependent variable was community similarity, expressed as (1–(Hellinger distance/sqrt(2))) to standardize similarity to the range from zero to one. The independent spatial variable was watercourse distance (km). The significance of the GLM parameters and differences of parameters between lake subsets were tested by t-tests.

All calculations were conducted in R ver. 2.15.1 (R Development Core Team) with the function PCNM of the package PCNM (Legendre et al. 2009) and the varpart and rda functions of vegan 2.0–4 (Oksanen et al. 2012). Loglinear decay functions and jackknife parameter estimates were fit by the glm function and a published R code (Supplement 1 in Millar et al. 2011).

Results

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

In total, 31 fish species were caught in the 49 lakes. The detailed species list has been provided earlier (Mehner et al. 2005). Species richness per lake varied between 9 and 27 (arithmetic mean = 15 species). Watercourse distances between the 49 lakes varied between 1 and 548 km (mean = 175 km).

Lake morphometry was a significant predictor of fish species composition (RDA, sum of predicted variance = 15.1%, F3,45= 2.66, p = 0.005 at 199 permutations). Lake area (loading to PC1 =–0.82) and lake depth (loading to PC2 =–0.80) substantially contributed to species composition, and arranged the fish communities along a gradient from dominance by coldwater species (for example, vendace Coregonus albula) in smaller and deep lakes to dominance of cyprinid species (primarily roach Rutilus rutilus, bream Abramis brama) in larger and shallow lakes. Likewise, productivity of the lakes predicted a significant proportion of fish species composition (RDA, sum of predicted variance = 11.4%, F3,45= 1.93, p = 0.01 at 199 permutations). TP concentration (loading to PC1 =–0.85) and Secchi depth (loading to PC1 = 0.93) split the fish communities between species of low-productivity clear lakes with oxygenated hypolimnion (vendace, and tench Tinca tinca) and those dominating in high-productivity turbid lakes (zander Sander lucioperca, catfish Silurus glanis, carp Cyprinus carpio). A modest effect by pH (loading to PC3 = 0.91) was seen in the fact that 9-spined stickleback Pungitius pungitius and catfish occurred more often at higher pH, whereas smelt Osmerus eperlanus and gudgeon Gobio gobio were found in lakes with lower pH.

The extracted six positive eigenvectors of the truncated watercourse distance matrix (truncation threshold = 348 km) predicted a significant proportion of fish species composition (RDA, sum of predicted variance = 21.4%, F5,43= 2.34, p = 0.01 at 199 permutations). The dominant first RDA axis separated large ubiquitous species such as perch Perca fluviatilis, bream, pike Esox lucius and roach (all loadings to RDA1 = 0.67) from smaller, primarily littoral species, for example Prussian carp Carassius gibelio (loading =–0.40), crucian carp Carassius carassius (–0.55), nine-spined stickleback (–0.47) and gudgeon (–0.29). These smaller species occurred more often in the lakes located in the southern and western part of the area, and became less frequent towards the north. Interestingly, the gradient of productivity of the 49 connected lakes was also directly spatially structured (RDA, sum of predicted variance = 28.4%, F5,43= 3.40, p = 0.005 at 199 permutations), whereas watercourse distances and lake morphometry were unrelated (RDA, F5,43= 1.82, p = 0.11 at 299 permutations).

In variation partitioning, variation of fish species composition was attributable to spatial, morphometric and productivity-related predictors (total adj. R²= 0.204, Fig. 2). Watercourse distances had the strongest unique effect (adj. R²= 0.085, Fig. 2), and alone predicted about 40% of the constrained variability in the species presence/absence matrix. The unique contribution of morphometric predictors (adj. R²= 0.042) was higher than the unique contribution of productivity (adj. R²= 0.025). All unique effects were significant (p < 0.05 at 999 permutations).

image

Figure 2. Venn diagram showing the effects of lake morphometry, productivity and spatial distance on fish species composition in lakes. The values indicate the adjusted R², as calculated from variation partitioning by RDA. Spatial distances were expressed as watercourse distances between 49 connected lakes.

Download figure to PowerPoint

Distance decay of fish community similarity with watercourse distance was moderate, but significant, in the 49 lakes (β= 0.00020, p = 0.029, Fig. 3, Table 1). If the decay was calculated only for 11 small-sized, primarily littoral species (Table 1), the slope was twice as steep (β= 0.00041, p = 0.042, Table 1). However, due to the strong scatter (Fig. 3), slopes of both decay functions did not differ significantly (t96= 0.98, p = 0.33). The slope of the distance decay function for the remaining 20 larger-sized species was not significant (β= 0.00019, p = 0.11, Table 1).

image

Figure 3. Scatter plot between community similarity (1–(Hellinger distance/sqrt(2)), based on presence–absence lists of in total 31 fish species) and watercourse spatial distance (km) of 49 connected lakes. The grey line represents the fit by a generalized linear model with binomial observation error and a log link function. For parameters of this equation, see Table 1.

Download figure to PowerPoint

Table 1. Parameter estimates for distance decay models of community similarity of fish in 49 lakes with watercourse distance, fitted by a generalized linear model with binomial observation error and a log link function. Parameters a and β represent intercept and slope of the loglinear model, s0 is the similarity at zero distance, and hd the halving distance (km). SE of parameters were estimated by a one-leave-out jackknife procedure (***p ≤ 0.001, *p ≤ 0.05, (*)p ≤ 0.10). Three subsets differed with respect to number of fish species included.
Subset a (SE)β (SE) S0 (SE) hd(km) (SE)
  1. #: 11 small, primarily littoral species: bitterling Rhodeus amarus (Bloch), Prussian carp Carassius gibelio (Bloch), crucian carp Carassius carassius (L.), gudgeon Gobio gobio (L.), sunbleak Leucaspius delineatus (Heckel), weather loach Misgurnus fossilis (L.), stone loach Cobitis taenia L., 3-spined stickleback Gasterosteus aculeatus L., 9-spined stickleback Pungitius pungitius (L.), smelt Osmerus eperlanus (L.), bleak Alburnus alburnus (L.).

All 31 species−0.560*** (0.0023)0.000200* (0.000089)0.571*** (0.013)3466 (*) (1883)
11 small species #−0.878*** (0.048)0.000414* (0.000198)0.415*** (0.020)1682 (*) (860)
20 large species−0.405*** (0.032)0.000190 (0.000116)0.667*** (0.022)3687 (2995)

Discussion

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Our results demonstrate that spatial watercourse distances among lakes contribute to variation of fish species composition in the lowland lakes in central Europe, because they predicted almost 40% of the total constrained variance in species occurrence within 49 connected lakes. Overall distance decay of community similarity between the lakes was moderate, but steeper when similarity was calculated among small-sized littoral species only. Our data suggest that spatial predictors cannot be neglected when explaining fish community similarity even at the small spatial distances of several hundreds of kilometres as found in our post-glacial lowland lakes, presumably attributable to limited dispersal of small-sized fish species.

Our results confirm that fish dispersal between lakes can be important for local community composition, and the degree of dispersal limitation is best reflected by the direct water connections between lakes (Jackson et al. 2001, Olden et al. 2001). There is only one study which compared the unique contribution of spatial variables on composition of lake communities between Euclidean and watercourse distances. In 18 lakes in Quebec (Canada), a significant effect of spatial patterning of lakes on fish communities was observed only when watercourse distances were considered, whereas spatial predictors were insignificant for assumed overland dispersal, as simulated by Euclidean distances between the lakes (Beisner et al. 2006). When applying Euclidean distances between the 49 lakes in our study for variation partitioning and distance decay, we found a similar weakening of spatial effects on fish community composition (results not shown). Therefore, dispersal of fish seems to be more limited if the longer waterway distances which reflect the true distances fish have to swim are considered relative to Euclidean distances which reflect theoretical dispersal via terrestrial systems (Olden et al. 2001).

Residual variation in the variation partitioning was high (˜80%), a feature that seems to be typical for analyses based on survey data (Beisner et al. 2006, Hajek et al. 2011, De Bie et al. 2012). Furthermore, community similarity at zero distance (s0), or for lakes in close proximity, was only intermediate (0.57), highlighting that spatial distance is only one of many potential factors which determine local fish species composition. Our results suggest that biotic interactions and effects of environmental variables not directly included in our analyses (e.g. water temperature) additionally modify local community structure and may have contributed to the high residual variation. The effect of lake morphometry (with potential indirect effects via water temperature) and productivity on fish community composition found in our study confirmed similar results from earlier studies (Diekmann et al. 2005, Mehner et al. 2005). However, environmental factors usually contributed only weakly to fish community similarity within the ecoregion ‘Central Plains’ (Mehner et al. 2007) and across the major river basins in Europe (Leprieur et al. 2009). Furthermore, for lake fish communities in the temperate European region, there is only weak evidence that local biotic interactions modify the presence/absence matrix of species (Tonn et al. 1990), for example because antagonistic interactions between fish predators and fish prey seem to have an only weak structuring impact in the German lakes studied here (Mehner 2010).

In contrast, spatial distance became a major factor for community similarity in European drainage basins (up to 80% of predicted variance in variation partitioning) if large spatial gradients covering also the major dispersal barriers (Pyrenees and Alps mountains) were considered (Leprieur et al. 2009). Climatic factors and geographic isolation also governed the functional diversity of fish communities in these European drainage basins (Schleuter et al. 2012). At the much smaller spatial scale as applied in our study which covered only two neighbored European river basins, the effect of dispersal limitation on community similarity could be expected to be relatively weak (Leprieur et al. 2009). Nevertheless, the spatial effects on the fish community similarity in the 49 connected lakes reflected ecologically meaningful patterns. The overall only moderate distance decay of community similarity along spatial distances was caused by the several ubiquitous species occurring in almost all lakes, whereas stronger dispersal limitation could be found for a few small-sized littoral species. Here, the inclusion of fish catches in littoral areas by electrofishing may have resulted in a more reliable estimation of local fish species richness. For active dispersers, dispersal range is assumed to be positively correlated with body size (Jenkins et al. 2007, Shurin et al. 2009), and hence small-sized fish species may contribute more to decreasing community similarity across spatial distances than large-sized species. Therefore, even if the explained variance of all redundancy analyses on fish community similarity did not exceed 20%, the calculations support that the similarity of fish communities between lakes results from the interplay of environmental and spatial effects.

Variation partitioning outcompetes spatial analyses of distance decay (Soininen et al. 2007, Shurin et al. 2009, Astorga et al. 2012) because the unique contributions from all included predictor groups can be disentangled. We have shown that spatial, morphometric and productivity-related predictors contributed significant, unique effects to community similarity. Furthermore, while we kept spatial, morphometric and productivity-related predictors separate in variation partitioning, the intricate relationships between spatial effects and local environmental predictors became obvious. According to the RDA of lake productivity and morphometry on spatial predictors, lakes at greater distances were more dissimilar with respect to their productivity, whereas lake morphometry and spatial distance were unrelated. These differing responses of local predictors make it clear that the spatial distribution of study objects is of fundamental importance for a mechanistic understanding of local biological structures (Legendre 1993).

We estimated the decay of community similarity with Euclidean distance by generalized linear models with binomial observation error and a log link function. In this approach, community similarity is modeled as binomial proportions, and the expected similarities are regarded as probabilities and fitted to the observed similarity values (Millar et al. 2011). Accordingly, the estimates of the decay function parameters are less biased by study extent and grain size than traditional (log)linear regressions, and the results can be compared across different sets of study objects (Steinbauer et al. 2012). Our comparison of the distance decay between sets of smaller and larger fish species demonstrates that there was a tendency toward steeper distance decay for the small-sized, littoral species. However, there was strong scatter in community similarity at similar geographical distance, and hence statistical comparison of slopes was not significant. The estimated slope of the distance decay for all species within the 49 connected lakes (β= 0.0002) was more than an order of magnitude lower than the slope of distance decay for native fish species between the European river basins (β= 0.006) (Leprieur et al. 2009). Nevertheless, the estimated slope and the halving distance of similarity of approximately 3500 km suggest that dispersal limitation may contribute to forming local fish communities even at small spatial scales, if local species richness is appropriately determined.

Fish are limited to dispersal via streams or channels between lakes. In a landscape of connected lakes, fish are operating more as a metacommunity compared to smaller organisms, such as bacteria, phytoplankton or zooplankton (Beisner et al. 2006, Shurin et al. 2009). This is counter-intuitive at the first sight because the large size of fish may facilitate active long-distance dispersal. However, dispersal ability interacts with local colonization rate (number of dispersing propagules) and extinction risk. The relatively large size of fish in comparison with other aquatic taxa results in relatively low population sizes (Cohen et al. 2003) and hence in low numbers of potential dispersers. Furthermore, low population sizes may subject fish to high local extinction risk, and sexual reproduction may result in Allee effects limiting spread and geographical range of species (Amarasekare 1998, Shurin et al. 2009). These factors explain why dispersal limitation was generally found to be stronger for fish than for aquatic invertebrate taxa (Shurin et al. 2009, De Bie et al. 2012).

Dispersal limitation is often seen as the opposite process to local selection to explain patterns of biological diversity (Ricklefs 2004, Vergnon et al. 2009). However, our results do not support that fish community similarity is entirely driven by spatial factors and their effects on dispersal of organisms, as predicted by the Neutral Theory of Biodiversity (Hubbell 2001). The contrasting species sorting model predicts that community similarity decreases with environmental distance, irrespective of geographic proximity, as a result of species differences along environmental gradients (Tilman 1982, Leibold et al. 2004). We found unique contributions from both local environmental and spatial predictors on presence of fish species, confirming that fish community composition in lakes reflects the interplay from both spatial and environmental factors and the overall spatial configuration of the lakes and their aquatic connectivity (Cottenie 2005). This result suggests that processes from both niche and neutral theories are needed to explain fish community structure (Gravel et al. 2006). The interplay of landscape connectivity and environmental gradients became obvious even at the small spatial scale we have studied here (˜550 km maximum distance).

Acknowledgements

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

We thank P. Peres-Neto and R. Millar for statistical advice, J. R. Post for stimulating discussions and comments on the final draft, and D. Opitz and N. Sommerwerk for help with GIS software.

References

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References