A pixel-level carbon trajectory index for monitoring land-use-driven carbon dynamics: contrasting pathways across four decades in the Cerrado, Pantanal, and Atlantic Forest

Authors:

Abstract
Land-use-driven carbon loss is a major but poorly quantified component of the global carbon cycle, particularly in tropical regions lacking continuous, validated satellite-based biomass records before the mid-2000s. We developed and validated a pixel-level Carbon Trajectory Index (CTI) — an indicator of land-use-driven aboveground carbon change — across the Cerrado, Pantanal, and Atlantic Forest biomes of Mato Grosso do Sul, Brazil (1985–2024), reconstructed using a field-calibrated, stock-difference accounting framework applied to a 40-year MapBiomas land-cover time series. We evaluated the reconstruction against five independent external validation lines and tested its sensitivity to the calibration source using an alternative, remote-sensing-derived parameterization. Statewide carbon stock declined 51.2% (240.5 to 117.4 million Mg C), with the sharpest relative loss in the Cerrado and the most protected trajectory in the Atlantic Forest. Validation revealed a consistent asymmetry: metrics of carbon dynamics — fire-driven loss (ρ = 0.985) and total disturbance-related loss (ρ = 0.65) — validated strongly against independent satellite records, while absolute stock validated only moderately (R² < 0.5) against three independent biomass products. A sensitivity analysis confirmed that field- and satellite-calibrated parameterizations, despite diverging by up to 76% in absolute stock, converged on statewide decline estimates within 4.4 percentage points, indicating that the reconstruction's temporal pattern is comparatively robust even where its absolute level is not. Spatial clustering analysis (Getis-Ord Gi*, Moran's I) showed that the Cerrado and Pantanal did not differ in median carbon change yet diverged sharply in spatial organization of loss, with pixel-level trajectory type predicting local spatial context more strongly than biome identity. These results demonstrate that spatially explicit, validated carbon trajectory reconstruction can support fire-carbon monitoring and reveal land-use-driven spatial patterns invisible to aggregate or stock-only approaches. By coupling this stock-based index with a trajectory typology and spatial-clustering diagnostics, the framework offers a transferable, field-validated indicator for land-use carbon monitoring in other data-limited tropical regions.