|Multi-scale analysis of species-environment relationships|Thrush, S.F.; Hewitt, J.E.; Herman, P.; Ysebaert, T. (2005). Multi-scale analysis of species-environment relationships. Mar. Ecol. Prog. Ser. 302: 13-26. dx.doi.org/10.3354/meps302013
In: Marine Ecology Progress Series. Inter-Research: Oldendorf/Luhe. ISSN 0171-8630, more
Regression models · Model assessment · Scale · Spatial extent · Ecological extent · Species–environment relationships
|Authors|| || Top |
- Thrush, S.F.
- Hewitt, J.E.
- Herman, P., more
- Ysebaert, T., more
Species-environment models are important tools for ecology and conservation, but ecologists generally lack knowledge of how spatial extent, habitat and environmental heterogeneity interact to affect relationships. This study investigates the effect of the spatial scale when predicting the response of maximum density and probability of occurrence of macrobenthic taxa to changes in sediment mud content. Three spatial scales were used, 2 km (an area within an estuary), 11 km (a single estuary) and 500 km (multiple estuaries), with the range of mud content being similar at the 2 largest spatial extents. Models from each scale were compared using model fit and the shape of the response curve. Different effects of scale were found for different model types and for different taxa, varying from no-scale dependence to markedly different responses at different scales. However, for most taxa, mud content was an important factor and qualitative responses could be predicted. Four species, the bivalves Austrovenus stutchburyi and Macomona liliana and the polychaetes Heteromastus filiformis and Nereidae, exhibited flat response curves, indicative of small changes in density over a broad range of mud content. Further analyses for these species indicated that general responses derived from a number of estuaries may reflect direct, local effects of mud content, but also ‘mediated’ effects, i.e. effects of other variables that correlate with mud content within an estuary, but have different correlations with mud content across estuaries. These results suggest that models should be validated against changes in environmental heterogeneity and spatial extent. Factors operating across scales may compound biases generated by unmeasured variables, but the use of models reflecting different aspects of spatial distribution (e.g. presence/absence, maximum density), incorporation of natural history information and the testing of hypotheses concerning the relationships developed at different scales may help determine the limits of extrapolation in the application and interpretation of species–environment models.