Concurrent burned-area detections are associated with persistent agro-pastoral conversion after first forest loss in the Peruvian Amazon

Cerna Cueva A.F. · Remote Sensing Applications: Society and Environment (2026) · DOI: 10.1016/j.rsase.2026.102308

Resumen

Concurrent burned-area detections may distinguish forest losses followed by persistent conversion. This observational study tested whether agreement between MapBiomas Fire and the Global Annual Burned Area Mapping (GABAM) product was associated with persistent anthropogenic conversion in the Peruvian Amazon and whether destinations were predominantly agro-pastoral. A retrospective cohort sampled a 0.01° lattice. Hansen Global Forest Change defined first loss during 2014–2021, MapBiomas Peru Collection 3 defined baseline and post-loss cover, and persistent conversion required anthropogenic cover for three consecutive post-loss years. Risks were standardised by loss year, spatial block and baseline forest class, with block-bootstrap uncertainty. In the broad dual-versus-neither comparison (n = 5911), conversion risks were 72.3% and 38.7% (risk ratio [RR] 1.87, 95% confidence interval [CI] 1.69–2.18). Among locations still classified as forest in the loss year (n = 4055), risks were 45.2% and 23.0% (RR 1.96, 95% CI 1.60–2.69). Coarsened control for loss extent and prior anthropogenic frontier yielded RR 1.45, whereas the full-cohort model estimated RR 1.98. In the broad cohort, more than 99% of persistent dual-detection conversions ended as mosaic or pasture. Complementary VIIRS observations occurred within 750 m in the loss year for 83.3% of strict-cohort dual detections and 42.4% of neither detections. Dual detections identify an associated trajectory, not an isolated causal effect. Thermal corroboration does not validate individual burned pixels or subsequent land use.

Instituciones: Universidad Nacional Agraria de la Selva