Satellite-Based Forest Clearing Detection in the Brazilian Amazon: FORMA, DETER, and PRODES

This issue brief introduces a system called FORest Monitoring for Action (FORMA), which provides near-real-time information on new forest clearing in the humid tropical forests of Asia, Africa, and Latin America. To assess FORMA's performance, its spatial and temporal accuracy are compared against PRODES (Projeto de Monitoramento do Desmatamento na Amazônia Legal por Satélite) and DETER (Sistema de Detecção do Desmatamento em Tempo Real na Amazônia), two well-established systems that monitor forest clearing in the Brazilian Amazon. The assessment focuses on a rapidly deforesting area that overlaps Manicoré and Novo Aripuanã municipalities, located in the southeastern portion of the Brazilian state of Amazonas. FORMA and DETER are both designed to identify new forest clearing at medium spatial resolution, while PRODES identifies cleared areas with higher precision on an annual basis. The assessment examines the relative spatial accuracy of FORMA and DETER against the higher-resolution annual data provided by PRODES, as well as the spatial and temporal correlation of information provided by FORMA and DETER. In terms of spatial accuracy, forest clearing areas identified by FORMA and PRODES are highly correlated. This result is consistent and stable when several statistical methods are applied. In the comparison of identified clearing areas, FORMA's correlation with PRODES is significantly higher than DETER's. FORMA and DETER appear to be complements rather than substitutes, with a combination of DETER and FORMA identifying PRODES cleared areas more accurately than FORMA alone. In terms of temporal accuracy, FORMA-identified areas of new clearing consistently appear in cleared areas identified by PRODES in its next annual review. Temporal identifications of cleared areas by FORMA and DETER are highly correlated. In the identification of newly cleared areas, FORMA actually leads DETER by about half a year. The advent of systems like FORMA has the potential to provide rapid identification of new clearing, thereby dramatically strengthening efforts to conserve and sustainably manage these important ecosystems. FORMA is a near-real-time forest clearing alert system that uses a cloud computing algorithm to combine frequently updated satellite imagery with complementary information on factors that affect forest cover loss, such as fires and precipitation. The system generates twice-monthly "alerts" for humid tropical forests that identify 500 x 500 meter areas where new, large-scale clearing is likely to have occurred. FORMA is designed for quick identification of new forest clearing, allowing for rapid response and prioritization of scarce financial and human resources dedicated to forest conservation or sustainable forest management. FORMA uses an automated statistical algorithm that relates spatially formatted data on forest clearing to information on vegetation reflectance, active fires, and rainfall. The algorithm employs parallel processing in a distributed server system that enables rapid analysis of very large data sets. It employs statistical techniques to achieve the best fit to data on forest clearing published by Hansen et al. (2008a) in the Proceedings of the National Academy of Sciences. These data identify 500 x 500 meter areas in humid tropical forests where clearing at "agro-industrial scale" was highly likely during the period 2000-05. FORMA's parameters are refitted to the Hansen data for each of the terrestrial ecoregions defined by WWF International for the pantropics, which span 89 countries. On a twice-monthly basis, the system estimates the probability that large-scale clearing has occurred since 2005 for each 500 x 500 meter area in humid tropical forests.