Predicting macrofaunal species distributions in estuarine gradients using logistic regression and classification systems

There is a growing need to predict ecological responses to long-term habitat change. However, statistical models for marine soft-substratum ecosystems are limited, and consequently there is a need for the development of such models. In order to assess the utility of statistical modelling approaches for predicting likely changes in species distributions under varying environmental conditions, we tested the utility of logistic modelling and classification approaches. We successfully developed models relating the presence/absence of common intertidal macrofauna to changing environmental variables such as sediment characteristics, depth/elevation, tidal currents and wind-wave (i.e. wind-generated wave activity) disturbance. The final model for each species contained between 1 and 6 variables, where the percentage correctly predicted was moderate to high, ranging from 59 to 97%. We were also able to identify relationships between higher level variables such as estuary type, basin morphometry and catchment-draining processes in driving macrobenthic community composition; however, we were unable to fully test the utility of the classification approach due to differences in the scale at which the macrobenthic data was collected and the scale of the higher level physical variables. These models were developed and tested using data that covered a range of environmental conditions in 5 estuaries in New Zealand. Such broad-scale statistical models play a critical role in our understanding of the likely effects of large-scale habitat change. However, a greater understanding of the fine-scale mechanistic controls on species distributions such as life-history characteristics, density information and biotic interactions would potentially lead to the development of more sensitive models.