Effective management of benthic habitats is important for maintaining heathy and functional aquatic ecosystems. To provide managers with the best possible information, characterizing benthic habitats at the community level is essential; yet, acquiring the data sets needed to achieve this task is resource intensive and, at times, prohibitively expensive. Thus, thoughtful assessments of which data to collect and utilize in benthic habitat characterization studies are needed. Environmental data sets commonly used to characterize benthic habitats include a range of variables from water depth and sediment grain size to seabed features identified by sonar backscatter. The objective of this study was to identify the most useful environmental variables for characterizing infaunal benthic habitats and to determine how to best utilize these variables in analyses (e.g., by comparing continuous vs. categorical explanatory variables). The modeling approach used multivariate regression tree and redundancy analysis along with a critical cross-validation step for model evaluation. Results indicated that models with more than ~7 environmental predictors overfitted the data sets analyzed and that categorizing continuous predictors into categorical ones influenced the proportion of infaunal community variation explained by each model. Habitats identified and characterized on the basis of sonar backscatter explained more of the infaunal community variation than any model that used a combination of other environmental variables (e.g., water depth & sediment grain size) or those constructed using categorical habitat classes from existing classification schemes. We therefore recommend maximizing the potential of sonar-derived variables for characterizing infaunal benthic habitats in nearshore, soft-sediment ecosystems.
