Failure to find the relationship between dispersal and spatial autocorrelation in species abundance
Spatial autocorrelation in species abundances indicates a lack of independence between sample locations and causes problems in distribution modelling. Knowing the cause of such spatial autocorrelation is vital for selecting the best suited modelling methods. Most autocorrelation in distributions is caused by autocorrelation in the underlying environmental conditions. The aim of the study was to determine whether dispersal could be responsible for additional spatial autocorrelation. I used data from 107 species of the North American Breeding Bird Survey covering the conterminous United States to investigate this question. As is the case for most species, no direct information on the dispersal activity of the study species was available. Therefore, I derived dispersal indices from three ecological theories: the deviation from an abundance-occupancy relationship, the spatial exponent of Taylorâ€™s power law, and density dependence. Spatial autocorrelation was captured in conditional autoregressive regression models (CAR) and measured with a standardized version of the regression coefficient Ï, the extent of the optimized neighbourhood, and the additional variance explained in CAR models over traditional regression models. No consistent association between these measures of autocorrelation and the indices for dispersal was found. Results indicated that the indirect ecological indices for dispersal carried too much noise and too little information for successful analysis. Future research on the effects of dispersal on autocorrelation need to be based on improved indirect indicators or direct, empirical dispersal information.