MEPS 478:15-25 (2013)  -  DOI: https://doi.org/10.3354/meps10267

ABSTRACT: The organisation of benthic communities across coral reefs is underpinned by spatially structured ecological processes and neighbourhood interactions such as larval dispersal, migration, competition and the spread of disease. These give rise to spatial autocorrelation in reef communities. This paper demonstrates how the measurement of spatial autocorrelation can profitably be incorporated into studies of coral reef ecology through a series of 5 simple statistical exercises: for the generation of maps depicting the strength of spatial relationships between ecological communities, as an indicator of optimal dimensions for sampling ecological communities on coral reefs, as a diagnostic tool for model misspecification, as an indicator of a spatial process underpinning the distribution of an observed community pattern and as a surrogate for missing variables in a model. The benefits of incorporating spatial autocorrelation include (1) quantifying the extent and pattern of autocorrelation across reefs, (2) signifying the presence of redundant information in field datasets, (3) indexing the nature and degree to which fundamental assumptions of classic (i.e. non-spatial) statistical techniques are violated, (4) indicating the nature (spatial versus non-spatial) of an observable pattern to be modelled, and (5) offering an opportunity to partition out and utilise the spatially structured component of model error as a surrogate for a missing variable. Collectively, the statistical exercises presented here provide a persuasive case for the measurement and interrogation of spatial autocorrelation in studies of coral reef ecology.

KEY WORDS: Spatial statistics · Autocorrelation · Moran · Geary