Predicting Plant Zonation in the Salt and Brackish Marshes of Georgia

Tidal marshes provide important economic and environmental functions. There are 1,745 km2 of tidal marsh habitat along the Georgia coast that can be separated into three distinct marsh types based on salinity levels. Tidal fresh, brackish, and salt marshes occur along a gradient that is maintained by tidal forces pushing inland and freshwater sources pushing towards the ocean. The first objective of this research was to characterize selected edaphic variables for salt and brackish marshes and determine whether or not there were significant differences between the two marsh types based on these parameters. Brackish marshes are positioned at higher elevations, with lower salinities, and higher redox potential and organic matter content in comparison to salt marshes. When species-specific differences were examined there were varying trends of significant difference among individual species. Overall, it is shown that determining species distribution using edaphic variables can be complicated by insignificant differences between species-specific habitats. The second objective of this research was to classify tidal marsh plants based on four edaphic variables and elevation using classification and regression trees (CART). Observations were separated by habitat type (salt or brackish) and classified based on the two variable groupings. Based on all observations combined, CART models for the edaphic-only and the edaphic-elevation model had overall classification errors of 34% and 32%, respectively, with salinity being the most important variable in classifying observations. Elevation did not greatly reduce classification errors. In the habitat-specific models, overall accuracy of the edaphic models were 19.74% and 19.72% for salt and brackish marshes, respectively. As with the CART based on all observations, elevation did not significantly increase the accuracy of edaphic models in the habitat-specific models (overall accuracies of 21.88% and 19.01% for salt and brackish, respectively). Water content was the most important variable in salt marshes and salinity was most important in classifying brackish marsh observations. For CART models, combining salt and brackish species together results in higher error rates (33%) than habitat-specific trees, (i.e. only salt marsh species or only brackish marsh species), indicating that habitat-specific analysis is preferable. INDEX WORDS: Brackish marsh, Salt marsh, CART, Elevation, Salinit