Mapping Africa Biomass with MODIS Imagery

Central Africa contains the second largest block of tropical forest remaining in the world, and is one of the largest carbon reservoirs on Earth. The carbon dynamics of the region differ substantially from other tropical forests because most deforestation and land use is associated with selective logging and small-scale landholders practicing traditional "slash-and-burn" agriculture. Despite estimates of 1-2 PgC/yr released to the atmosphere from tropical deforestation, the amount released from Central Africa is highly uncertain relative to the amounts released from other tropical forest areas. The uncertainty in carbon fluxes results from inadequate estimates of both rates of deforestation and standing stocks of carbon (forest biomass). Here we present new results mapping above-ground forest biomass for tropical Africa using machine learning techniques to integrate MODIS 1km spectral reflectance with forest inventory measurements to calibrate an empirical relationship. The derived forest biomass at each MODIS pixel shows the spatial distribution of forest biomass over the entire tropical forest region. The model has been tested in Uganda, Mali and part of Republic of Congo where field data were available. The regression tree model based on MODIS NBAR surface reflectance for Uganda, Mali and Republic of Congo explains 94 percent of the variance in above-ground biomass with a root mean square error (RMSE) of 27 Tons/ha. The approach shows promise for use of optical remote sensing data in mapping the spatial distribution of forest biomass across the region.