Mapping, Analyzing, and Visualizng 60 Years of Forest Cover Change in Northeast Kansas

Matthew D. Dunbar, M.A. Department of Geography, May 2005 University of Kansas Woody species can rapidly invade grassland habitats that are not burned, grazed, cultivated or mowed along the prairie-forest boundary, where it has been suggested that forest expansion has occurred within the last 100 years. Considering this dynamic environment, there is a need for detailed landscape-level analysis documenting the structural patterns of this forest cover change. For this study, aerial photography collected approximately every 10 years from 1940 through 2000 was used to map, analyze and visualize the changes in forest cover within two study areas located in the prairieforest ecotone of northeastern Kansas. The first study area was a 62 km landscape-sized region (southern half of the USGS Midland topographic quadrangle) heavily modified by human impacts, while the second smaller 2.58 km site (Fitch Natural History Reservation) within the larger study area provided a controlled local-scale environment devoid of human impacts, fire, and grazing. The results of the forest cover classification, using object-oriented segmentation and classification methods, show an overall classification accuracy across all photo years of 96% when compared with manual interpreters. The analysis of this classification indicates that forest cover has increased over the past 60 years both on the larger southern Midland quad area (40.7% increase) and within the Fitch Natural History Reservation (53.2% increase). Landscape metric analysis suggests that forest cover is becoming more cohesive as the average forest patch size is increasing in both study areas. In the southern Midland quad, forest cover is more prevalent and increasing at a faster rate on areas of greater slope. On the Fitch Reservation, the forest has been expanding more rapidly on less steep land, as areas of higher slope have already been forested and there are no limiting human factors to restrict growth on the flatter land. Finally, the flexibility of computer visualizations demonstrate their effectiveness for representing spatio-temporal land cover patterns in a clear, concise and effective manner. This investigation suggests numerous other avenues of research involving the object-oriented