Large scale characterization of intertidal communities using a predictive model

This research presents a methodology that combines shoreline morphology, temperature, salinity, and current (tidal stream velocity) data with biological field sampling to develop kilometer scale intertidal biotopes in the southern Strait of Georgia, British Columbia, Canada. The abiotic component of the classification was based on a geomorphological shoreline classification acquired from low level aerial video, fetch (wave exposure) estimates, as well as salinity, temperature, and current velocity predicted from a hydrodynamic model. The biotic component of the classification was based on in situ field sampling of 39 sites using a published sampling methodology. The inter-relationships between community and habitat were first explored using the two-way indicator species analysis (Twinspan) to define communities and functional groups, and then with regression tree models to determine what combinations of abiotic variables were necessary to support the communities identified by Twinspan. This methodology is predictive, and can be used to quantitatively model the occurrence of biological communities based on abiotic data. The classification presented here is composed of two parts. The first part clustered site by species data collected at 39 sites in the study area using Twinspan. Seven community types were identified by Twinspan. Using an exploratory technique termed tree-based regression, these community types were used as response variables, while abiotic data were used as predictors. The regression tree model predicted seven intertidal biotopes with an accuracy of 72%. Binary decision tree rules produced from the model were used to attach biotopes to the remainder of the 1486 km of shoreline in the study area. Results indicate that this methodology is an effective technique for assessing the broad scale biological composition of large areas of shoreline. This method will be used to assist in the identification of representative, distinct, and rare communities at scales consistent with the identification of conservation importance for site protection and management. It is expected that this methodology will be applied to the remaining 29 000 km of shoreline on Canada's Pacific coast in the near future. (C) 1999 Elsevier Science B.V. All rights reserved.