Aboveground woody carbon density and change in tropical dry forests and savannas, with land cover change process fractions and raw annual biomass predictions, 2015-2024
Published by Environmental Information Data Centre
About this dataset
This dataset consists of four complementary geospatial products, and the model calibration data, that quantify woody carbon stocks and their dynamics across the world's dry tropical forests and savannas from 2015 to 2024. It includes a 100 m resolution map of time‑averaged aboveground woody carbon density (AGCD; MgC ha-1), a 100 m map of annual AGCD change expressed as a per‑pixel linear trend (MgC ha-1 yr-1), a 10 km four‑band raster of the proportion of each cell affected by major land cover change processes (deforestation, degradation, densification, extensification) over the period, and a 100 m ten‑band stack of annual aboveground biomass density (AGBD; Mg ha-1) predictions for 2015-2024 before any quality corrections, masking, temporal smoothing, or gap‑filling. The mapped domain follows the study's definition of the dry tropical forests and savanna biome, excluding very arid zones, Mediterranean/temperate climates, and evergreen closed‑canopy moist forests. The products were created by linking annual‑minimum L‑band radar backscatter from ALOS‑2 PALSAR‑2 ScanSAR (HV, γ0 in natural/linear units) to aboveground carbon estimates from a global network of large field plots using a generalized additive model with a log link. The ScanSAR scenes were speckle‑filtered and, for each year, the per‑pixel minimum HV backscatter was taken to reduce soil‑moisture effects, then aggregated to 100 m. Quality filters and temporal smoothing was applied to the time series, including continent‑specific quality masks, exclusion of pixel‑years where the surface soil moisture was high, interpolation across short gaps, and a five‑year Savitzky-Golay temporal smoothing. Annual change was derived from the linear trend of the AGCD time series for 2015 to 2024. Where ScanSAR coverage was sparse, a secondary model based on ALOS‑2 annual mosaic data was used to fill spatial gaps. Land cover change processes were defined from the AGCD time series, using a wooded threshold (>10 MgC ha-1) and a major‑change threshold (>20% over the period) and then aggregated to 10 km cells as proportions of each process within each cell. The raw AGBD stack is also shared without QA, smoothing, or gap‑filling. These data enable spatially explicit assessment of woody carbon stocks and dynamics in the dry tropics and support national reporting, monitoring of change processes, and assessment of regional drivers. This work is part of the SECO project, resolving the current and future carbon dynamics of the dry tropics funded by NERC (NE/T01279X/1). Full details about this dataset can be found at https://doi.org/10.5285/2fd38433-cf42-4dc2-afe9-a4a885486c73