The perpetuity and improvement of REDD+ projects for curbing deforestation require rigorous impact evaluations of the effectiveness of existing on-the-ground interventions. Today, a number of global and regional remote sensing (RS) products are publicly available for detecting changes in forest cover worldwide. In this study, we assess the suitability of using these readily available products to evaluate the impact of REDD+ local projects targeting smallholders (owning plots of less than 100 ha) in the Brazilian Amazon. Firstly, we reconstruct forest loss for the period between 2008 and 2017 of 17,066 farms located in the Transamazonian region, using data derived from two landcover change datasets: Global Forest Change (GFC) and Amazon Deforestation Monitoring Project (PRODES). Secondly, we evaluate the consistency between the two sources of data. Lastly, we estimate the long-term impact of a REDDÅ project using both RS products. Results suggest that the deforestation estimates from the two data-sets are statistically different and that GFC detects systematically higher rates of deforestation than PRODES. However, we estimate that an average of about 2 ha of forest were saved on each participating farm during the first years of the program regardless the source of data. These results suggest that these products may not be suitable for accurately monitoring and measuring deforestation at the farm-level, but they can be a useful source of data on impact assessment of forest conservation projects.
