Predicting Regeneration of the Brazilian Amazon from the Synergy of Using a Process-based Model and EO data

Across the world, particularly in the tropics, the extent of forest clearance has been widespread. At present, few studies have been undertaken and little is known on the long-term effect of land use history following clearance, on forest recovery, a significant sink for atmospheric CO2. This study aimed at quantifying the capacity of regenerating forests in the Brazilian Legal Amazon (BLA) to recover carbon using a combination of Earth Observation (EO) data and the 3-PG forest growth model. As a case study, three sites were selected within the BLA, representative of different clearance histories on which extensive deforestation has occurred. Land use history for these areas was obtained from Landsat images time-series (1970’s - present) analysis. The tree species diversity was calculated from forest inventory data collected during three field campaigns conducted in 1993, 1995 and 2014. This dataset was used to interpret the reflectance trajectories of these areas, which were used to identify alternative pathways of succession. An advanced sensitivity analysis (SA) based on the Gaussian process emulator was also carried out on the 3-PG model to identify its most sensitive model inputs when applying it to a mixed tropical rainforest. A parameter set for mixed tropical forests was identified using a Monte-Carlo simulation and by comparing simulated outputs to field data. These parameters were then used within 3-PG to provide yearly estimates of carbon sequestrated. Results of our study showed that sites with a higher land use intensity accumulated biomass at a significantly slower rate than those used less intensively. Accumulation rates predicted from the model closely matched those calculated from the forest inventory data gathered at each site. SA results demonstrated scientifically credible behavior of the model and allowed identification of the most responsive model inputs and interactions. Findings also illustrated the suitability and potential of 3-PG process based model when combined with EO data as a way to forecast the productivity of mixed secondary successional forest in Brazil. Development of these methodologies has applications to other tropical ecosystems that have experienced a similar history of disturbance and can provide invaluable information for future land-use planning and REDD+ monitoring.