Pasture degradation is of increasing global concern as it enforces erosion processes and impacts the carbon storage capacities of the soil. Reliable methods for pasture degradation mapping are thus of great use to provide important information for sustainable landscape planning. Our research focusses on the Guapi-Macacu watershed (Rio de Janeiro (RJ), Brazil) as part of the biodiversity hotspot Mata Atlantica. The area is characterized by strong forest fragmentation and pasture degradation. We investigate the suitability of RapidEye and Landsat 5 TM data in comparison to a high-resolution image composite product based on RapidEye and downscaled Landsat 5 TM SWIR bands for land cover classification and pasture degradation mapping. Land cover classification results improved significantly for the image composite product (overall accuracy (OAA) 89%) compared to the application of RapidEye (OAA 87%) or Landsat 5 TM (OAA 85%) data alone. Pasture degradation was mapped using degradation class thresholds derived from field data and vegetation cover fractions on a per pixel basis and modelled using multiple endmember spectral mixture analysis (MESMA). The pasture degradation map based on the image composite achieved an overall accuracy of 77.5%, compared to 75% (RapidEye) and 61% (Landsat 5 TM). We further tested the relationship between degradation and slope class and concluded that more than 90% of the pastures on slopes ¿ 10 degrees show signs of degradation, whereby on above 20 slopes the portion of moderate to strong degradation is above 57%.