Incorporating spatial autocorrelation may invert observed patterns

• Published in: Diversity & distributions Vol. 13; no. 1; pp. 66 - 69
• Main Author: Kuehn, Ingolf
• Format: Journal Article
• Language: English
• Published: Oxford, UK Oxford, UK : Blackwell Publishing Ltd 2007; Blackwell Publishing Ltd; Blackwell Science; Blackwell
• Subjects: Altitude; Animal, plant and microbial ecology; Applied ecology; Autocorrelation; Autoregressive models; Biodiversity Letter; Biological and medical sciences; Ecological modeling; Environmental correlates; Fundamental and applied biological sciences. Psychology; General aspects; Germany; Least squares; linear models; Linear regression; Modeling; plant species richness; Plants; spatial autoregressive models; Spatial models; Species; Germany; Europe; plant species richness; Plant; Autoregressive model; Autocorrelation; Biodiversity; Environmental correlates; linear models; Linear model; Species richness; Germany; spatial autoregressive models
• ISSN: 1366-9516; 1472-4642
• DOI: 10.1111/j.1472-4642.2006.00293.x

Though still often neglected, spatial autocorrelation can be a serious issue in ecology because the presence of spatial autocorrelation may alter the parameter estimates and error probabilities of linear models. Here I re-analysed data from a previous study on the relationship between plant species richness and environmental correlates in Germany. While there was a positive relationship between native plant species richness and an altitudinal gradient when ignoring the presence of spatial autocorrelation, the use of a spatial simultaneous liner error model revealed a negative relationship. This most dramatic effect where the observed pattern was inverted may be explained by the environmental situation in Germany. There the highest altitudes are in the south and the lowlands in the north that result in some locally or regionally inverted patterns of the large-scale environmental gradients from the equator to the north. This study therefore shows the necessity to consider spatial autocorrelation in spatial analyses.