A Rapid Land Cover Classification Method for Use in Urban Watershed Analysis

Abstract

Because of the profound effect of urban development on aquatic systems, characterizing the land cover of a region is critical for a variety of resource management applications. In the Pacific Northwest, this characterization has been used most commonly to correlate the intensity of human activity with observed stream or wetland conditions, in order to predict the health of the stream system or to guide the allocation of mitigation efforts. Unfortunately, there is little consistency or quality control in how land-cover data are collected and analyzed. We have developed an alternative approach using Landsat satellite imagery to produce the same general type of land-cover characterization as has currently found widespread acceptance and use across the region. However, our methodology does so in a way that achieves maximum utility and consistency for a particular group of users -- individuals and agencies needing to assess watershed conditions in the urban, and urbanizing, parts of western Washington. The classes of land cover produced have been chosen to reflect the categories that can be readily distinguished in the satellite data and to have important differences in their associated runoff and watershed characteristics. The advantages of such an approach are obvious. The algorithm is developed only once; after completion, it can be applied rapidly to any other selected area through GIS software. It does not depend on the discretion of individual operators and so the results are reproducible. These advantages have not been lost on public agencies, but those agencies are not equipped to pursue such efforts systematically, given project-related geographic boundaries, time constraints, staff turnover, and the difficulty of inter-agency communication. With suitable testing and documentation, the release of these data layers through the University of Washington may encourage agencies across the region to adopt a uniform methodology, resulting in a degree of uniformity in data collection, analysis, and reporting of these data that is currently unavailable.