The effects of data selection and thematic detail on the accuracy of high spatial resolution wetland classifications

Accurate wetland maps are of critical importance for preserving the ecosystem functions provided by these valuable landscape elements. Though extensive research into wetland mapping methods using remotely sensed data exists, questions remain as to the effects of data type and classification scheme on classification accuracy when high spatial resolution data are used. The goal of this research was to examine the effects on wetland mapping accuracy of varying input datasets and thematic detail in two physiographically different study areas using a decision tree classifier. The results indicate that: topographic data and derivatives significantly increase mapping accuracy over optical imagery alone, the source of the elevation data and the type of topographic derivatives used were not major factors, the inclusion of radar and leaf-offimagery did not improve mapping accuracy, and increasing thematic detail resulted in significantly lower mapping accuracies i.e., particularly in more diverse wetland areas.