Classifying agricultural land in an urban landscape with application to waterfowl conservation

This project evaluates technical considerations and human resources required to remotely sense agricultural lands and demonstrates how the results can be used for waterfowl conservation. Using a hierarchical decision tree and 3 agricultural classification schemes on Landsat 7 ETM data, the accuracy was calculated for several image transformation techniques. For an 8 class agricultural scheme, the Tasseled Cap transform had a higher overall accuracy (75.1% ± 1.6) than the normalized difference vegetation index (60.6 ± 1.8), second modified soil adjusted vegetation index (60.6 ± 1.8), or arctangent to the simple ratio (59.4% ± 1.8), and had comparable accuracy to the dataset using 84 data layers (77.6% ± 1.5). The decision tree classifier replaced the requirement of raster based classification software and reduced the financial cost by 25%. A classified agricultural map was combined with a species – habitat model for American wigeon to set conservation goals for agricultural lands