Grassfires are a recurring hazard in the Philippines, particularly during the dry Amihan season from January to March. In March 2025, a major incident occurred in Barangay Pundaquit, Zambales. The Philippine Space Agency (PhilSA) conducted a remote sensing-based damage assessment using medium-resolution Sentinel-2 and very-high-resolution BlackSky imagery. Sentinel-2 data enabled the derivation of burn extent and severity maps through the Differenced Normalized Burn Ratio (dNBR), while BlackSky imagery provided fine spatial detail for visual verification. This study evaluates the performance of Sentinel-2 Deep Resolution 3.0 (S2DR3) in generating finer-scale burn severity maps by comparing classified outputs from original 20-m imagery and its 1-m upscaled counterpart. The super-resolution process introduced noticeable shifts in classification, with the areal extent of High Severity burns reduced from 162.41 ha to 43.12 ha and a tendency toward lower severity categories. A pixel-level disagreement rate of 37.4% and low mIoU scores across sites indicated spatial inconsistencies and underestimation of intense burn areas—likely due to spectral smoothing. Nonetheless, the upscaled product showed improved delineation of Unburned zones, suggesting its value for resolving finer-scale heterogeneity. These findings highlight the trade-offs in using super-resolved imagery for post-fire assessments, the need for super-resolution model calibration for burnt areas, and the importance of context-aware validation.

This study is part of the 10^th Geomatics International Conference (GEOICON) on 23 July 2025 in Surabaya, Indonesia.

More information here.