Integrating multiple earth observation data for monitoring forest cover change in cloud-prone, Western Rwanda.

Earth observations increasingly allow accurate assessments of forest cover, however consistent assessments are limited in developing countries including Rwanda. While the ruggedness and the remoteness of Western Rwanda makes remote sensing a cost-effective monitoring approach, the year-round cloud cover remain a challenge. The lack of consistent historical data has led to simplified narratives of forest change, attributing observed losses to high population density and the inflow of refugees following periods of wars. Without proper identification of drivers of forest cover change, conceived policies are unlikely to have impact. This research integrates cloud-free Landsat images with Google Earth images and field interviews, to document rates and drivers of forest cover change in Western Rwanda from 1986 to 2014. The TerrSet DecisionForest algorithm was used to produce forest cover maps from Landsat images. Google Earth images were used to validate observed changes and provide higher temporal information. Field interviews were used to document drivers of forest change and validate earth observation data. Results show 7% overall increase in forest area between 1986 and 2014; 90% of new forests are patchy monocultures of species valued for fuelwood and timber. Deforestation peaked between 1986 and 2006, and losses were equally high in the periods of 1986-1990 and 1995-2000, respectively corresponding to the establishment of pasturelands and the resettlement of refugees. This monitoring approach properly attributes drivers of forest cover change and highlights the need for revising current forest landscape restoration activities to improve landscape functioning and provision of diverse ecosystem services.