<![CDATA[The study focused on the persistent forest and land fires in Sumatra and the conventional early warning systems that showed limitations in capturing the dynamics of the soil-atmosphere and tended to overestimate risk zones during the peak periods of fires. Existing multivariate systems also showed limitations when there was a lack of integration between atmospheric water demand and edaphic (soil) vulnerability, resulting in lower spatial efficiency in the risk area. This paper presented a new bivariate physics-based model called the VPD-LVI Integrated Fuzzy (VLIF) Model, which integrated Vapor Pressure Deficit (VPD) and Land Vulnerability Index (LVI). For the VLIF-Model, to ensure a high level of spatial detail, the model processed 2,614,056 spatiotemporal observations, representing a grid resolution of 0.1° over a two-year period (2023–2024). The model showed acceptable structural stability at k=3 (FPC = 0.6137), with distinct risk zoning contrasts. Furthermore, validation against 61,015 VIIRS thermal anomalies indicated the model's predictive reliability with an ROC-AUC of 0.8288 and a Brier Score of 0.0169, suggesting good discrimination and calibration. In the spatial efficiency analysis, 17.80% of the area categorized as 'high' risk captured 62.28% of actual thermal anomaly grids, achieving a lift factor of 3.50 and a recall of 88.91%. The VLIF-Model provided clearer contouring of peak fire zones and enabled a more targeted spatial risk intelligence layer for specific tropical peatland mitigation efforts.]]>
Copyrights © 2026
