Land cover change caused by humans represents a major threat to the long term viability of natural areas. It is important to accurately classify and map existing natural areas so that this threat can be fully assessed within a given landscape. Availability of free orthorectified Landsat images through the U.S. Geological Survey provides a potentially valuable tool to evaluate human impacts to natural landscapes. We performed a supervised classification of multi-seasonal Landsat images to test the limits of using these images for mapping mixed landscapes at regional to local scales and to assess land cover changes within the Oak Openings region of Northwestern Ohio. Overall accuracy of our 15-class land cover map was 60% and 69% using traditional and fuzzy set analysis respectively. Overall map accuracy improved to 72% and 79% for traditional and fuzzy set analysis respectively using a more broadly defined 7-class land cover map. Accuracy of individual classes varied considerably, although classes made up of larger patches typically achieved greater accuracy. Human-dominated land cover classes currently occupy 73% of the Oak Openings region while < 3% of the region remains covered by native savannas, prairies, and barrens. Currently 10% of the region is permanently protected, including nearly all remnant savannas and wet prairies > 1 ha. Our findings highlight the utility of using Landsat images to evaluate mixed-use landscapes at regional scales but demonstrate the limitations of using these images at local scales.