SPATIAL AUTOCORRELATION AND ASSESSMENT OF HABITAT ABUNDANCE RELATIONSHIPS IN LITTORAL-ZONE FISH

Spatial autocorrelation, wherein intersite similarity is correlated wi th distance between sites, is a characteristic of most ecological stud ies spanning a large environmental range. If data are spatially autoco rrelated, classical statistical techniques provide biased estimates of relationships between species attributes and environmental variables. We examined abundances of seven littoral fishes in 25 lakes that vari ed substantially in morphometry, chemistry, and elevation across centr al Ontario. Weak correlations were observed between abundances of part icular species and environmental variables before correcting for spati al autocorrelation, and we hypothesized that correlations reflected sp ecies' habitat preferences. However, spatial autocorrelation existed i n the abiotic and fish abundance datasets. Once large-scale geographic patterns (spatial autocorrelation) were removed using partial Mantel tests, correlations changed within and between datasets. A strong rela tionship emerged between abundances and lake elevation. By comparing p atterns within geographically corrected data with those without correc tion, we identified particular species that exhibited spatially autoco rrelated abundances. The geographic direction of spatial autocorrelati on provided additional insights into environmental factors also correl ating with species abundance. We recommend that ecologists examine bot h geographically corrected and noncorrected data when developing hypot heses to explain regional variation in species abundance.