Abstract - Monitoring Landscape Change with Landsat Classifications Land managers and policy makers need accurate information to guide land use decision-making processes. Historically, there have been several sources of statistical inventory information available. Unfortunately, there are often large discrepancies between figures in these surveys and no means to tie the statistics to specific geographic locations. Satellite remote sensing offers a promising alternative: accurate land cover classifications tied to geographic space and time which have the potential to be compared to previous classifications. Landsat classifications provide complete (pixel by pixel) enumeration of land cover without sampling error, although they are subject to classification error. A major objective of our research is to determine how to minimize classification error so these products can be used to efficiently generate accurate landscape change statistics and maps. Classification of 1991 and 1998 Landsat Thematic Mapper imagery have yielded promising results with overall accuracies of 89% and 92% for five, level I classes (agricultural cropland, forests, water, wetlands, urban). Land use change statistics and maps have been derived from the 1991 and 1998 classifications and compared to other survey statistics for the seven-county Twin Cities Metropolitan Area. For a copy of this paper, please contact: Kali Sawaya.