At which scales does landscape structure influence the spatial distribution of elephants in the Western Ghats (India)?

UMR AMAP, Equipe 3International audienceIn spatial ecology, detailed covariance analyses are useful for investigating the influences of landscape properties on fauna and/or flora species. Such ecological influences usually operate at multiple scales, involving biological levels from individual to group, population or community and spatial units from field to farms and regions. The aim of this work was to analyze possible multiscale influences of some landscape properties on elephant distribution in the Western Ghats, India, by applying a recent and simple mathematical method to quantify such ecological relationships across space and scales. This method combines a moving window with various correlation indices to investigate the relationship between two mapped variables. Maps of landscape heterogeneity (quantified here at all locations of the landscape with a modified Shannon index) and Asian elephant presence (a two-dimensional presence probability) were significantly correlated. This correlation systematically decreased with increasing scales (i.e. sizes of the reference moving window). Yet, this global relationship includes both positive and negative correlations located at distinct places. We documented a positive feedback (reinforcement) because elephants appeared to seek greater habitat heterogeneity, in heterogeneous areas, such as along the interface between natural and a human-disturbed habitat or in the natural part of the studied landscape. In parallel, we observed a negative feedback (compensation) making elephants seeking more homogeneous places in some relatively heterogeneous zones. Such negative feedbacks corresponded to higher than average probabilities of elephant presence. Finally, when elephant density varied according to landscape heterogeneity (corresponding to significant correlations), it pointed towards swamps and grasslands, but not towards semi-evergreen or secondary forests (as it may have been expected). Land cover information appeared to be less relevant than an integrated heterogeneity index computed at all scales