Densely populated village ecosystems in subsistence agriculture regions cov- er a global area equivalent to two-thirds of that of tropical rainforests. Measuring long- term anthropogenic changes in these regions presents methodological challenges for ecol- ogists, because ecosystem processes must be measured and compared under preindustrial vs. contemporary conditions within highly heterogeneous anthropogenic landscapes. In this study, we use landscape classification and observational uncertainty analysis to stratify, estimate, and compare changes in landscape structure caused by the transition from tra- ditional to modern management within a single village in China's Tai Lake Region. Con- temporary data were gathered on-site during 1993-1996 using aerial photography, field surveying, local knowledge, and household surveys, while traditional period estimates, ;1930, were obtained using interviews, back estimation, and historical sources. A hier- archical landscape classification scheme was used to stratify village landscapes into 35 fine-scale landscape components with relatively homogeneous ecosystem processes. Monte Carlo simulation and data quality indexing were used to calculate and compare village and component areas and their changes. Using this approach, we observed significant long- term declines in the proportion of village area covered by paddy (212%), fallow, and perennial areas (28%), and increases in areas under buildings and infrastructure (17%). Aquatic and wetland areas increased by nearly 40% from 1930 to 1994. Significant declines in fallow and perennial vegetation and increases in constructed and heavily trafficked areas indicate overall increases in human disturbance. Our methods for observational uncertainty analysis, anthropogenic landscape classification, and the linking of imagery with field, household, and other local data are powerful tools for detecting and monitoring long-term ecological changes within anthropogenic landscapes.