Spatial autocorrelation and red herrings in geographical ecology

摘要: Aim Spatial autocorrelation in ecological data can inflate Type I errors statistical analyses. There has also been a recent claim that spatial generates 'red herrings', such virtually all past analyses are flawed. We consider the origins of this phenomenon, implications for macro-scale patterns species diversity and set out clarification problems generated by its presence. Location To illustrate issues involved, we analyse richness birds western/central Europe, north Africa Middle East. Methods correlograms five environmental variables were using Moran's coefficients. Multiple regression, both ordinary least-squares (OLS) generalized least squares (GLS) assuming structure residuals, used to identify strongest predictors richness. Autocorrelation residuals obtained after stepwise OLS regression undertaken, ranks full GLS models compared. Results Bird is characterized quadratic north-south gradient. usually had positive up c. 1600 km. Including successively model reduced non-detectable levels, indicating explained data. In principle, if not autocorrelated then special case GLS. However, our comparison between including revealed de-emphasized with strong long-distance clinal structures, giving more importance acting at smaller geographical scales. Conclusion Although should always be investigated, it does necessarily generate bias. Rather, useful tool investigate mechanisms operating on different Claims do take into account flawed without foundation.