Evaluating Error in Using the National Vegetation

At the landscape scale, representation of reality using ecological community maps is limited by: how well the chosen classification system represents actual vegetation community composition; how effectively aerial photography captures the distinguishing features of each mapping unit within the classification; and how well these mapping units are delineated by photo-interpreters. Three errors deriving from these factors can be defined as classification system error, photo-limitation error, and mapper error. We evaluated the relative importance of these error types for ecological community map- ping in a 7283 ha area including the Lake Umbagog National Wildlife Refuge (LUNWR). We used the association level of the National Vegetation Classification System (NVC) to classify and map ecological communities through combined aerial-photo interpretation and fieldwork. Map accuracy assessment using an error matrix yielded an overall map accuracy of 46 ± 9%. Fuzzy set analysis and use of a "goodness-of-fit" table showed that classification system error accounted for 25% of the error, photo- limitations for 66% of the error, and mapper error for the remaining 9%. To improve map accuracy, classification system error can be reduced by: (1) refining class definitions to decrease ambiguity, (2) adding new classes to more adequately describe the complex of local vegetation patterns, or (3) using a higher level of classification within the NVC. Photo-limitation error can be reduced by: (1) defining mapping units by aggregating NVC associations into photo-interpretable groups, (2) utilizing aerial photographs with a higher resolution than the 1:15,840 scale photographs used in this study, or (3) mapping primarily using fieldwork.