Detecting wetland change: A rule-based approach using NWI and SPOT-XS data

Spectral similarities among wetlands and agricultural fields and forests ca n create difficulties in satellite image classification. Utilizing ancillar y information, such as National Wetlands Inventory (NWI) data available fro m the U.S. Fish and Wildlife Service,to define wetlands provides a practica l solution to this problem. However, NWI data are at least a decade old for many areas. In this study, data derived from SPOT-XS imagery were used to create a rule-based model to detect wetland change and update NWI data. Fir st, the pixel vector modulus was calculated (sqrt(b(1)(2) + b(2)(2) + b(3)( 2)/N)), and a coarse land-cover layer was developed from SPOT-XS imagery. S econd, combinations of the mean modulus value, the majority land-cover valu e, and the NWI system class for each polygon were used to develop logic rul es to indicate areas of potential wetland change. Logic rules and change-de tection accuracy varied according to wetland type. At least 8 percent of th e wetlands in the study area had undergone land-cover change since the mid- 1980s; however, most such wetlands were small (<1 ha).