Analyzing and modelling spatial distribution of summering lesser kestrel: The role of spatial autocorrelation

摘要: Abstract In modelling spatial distribution of species, ignoring autocorrelation (SA) and multicollinearity may lead to false ecological conclusions. Here we take into account both issues for examining the pattern abundance globally threatened lesser kestrel ( Falco naumanni ) during summer in a 38,400 ha area northwestern Spain where large premigratory aggregations species occur. Spatial was examined using Moran's correlogram, models were built including geographical coordinates autocovariate terms (which SA) generalized linear (GLM) hierarchical partitioning (HP) models. HP allow alleviate multicollinearity. A grid-based approach used by dividing study 24 contiguous 4 km × 4 km squares birds counted 2–3 visits per square (response variable). Environmental coarse-grained variables extracted from geographic information system (GIS) at three extents. correlogram showed that mean spatially autocorrelated up 4–8 km. The results GLM analyses roughly compatible. explained 80.0% variation same Lesser Kestrel not significantly landscape variables, but negatively related distance nearest communal roost breeding colony with more 10 pairs kestrel. An term added later improved their explanatory power (from 74.5 80.0%) model residuals, which longer autocorrelated, fulfilling thus statistical assumption independent errors. Findings suggest summering is, least, partially driven endogenous causes, such as conspecific attraction. Exogenous causes finer-scale (e.g. type crops food available) are yet likely needed kestrel-environment relationships.