Multi-temporal Sentinel-2 consensus mapping framework for enhanced wildfire severity assessment in semi-arid East Africa

Abstract The study investigates the January 2023 wildfire event in Kitui County, Kenya, utilizing multi-temporal Sentinel-2 imagery acquired in December 2022 for (pre-fire) and February 2023(post-fire). A collection of relevant spectral indices and detection approaches including NBR, EVI, MSI, NDMI and CVA was employed to quantify the extent and severity of the burn. The analysis indicates 52.38% which is 1728,03 km2 of the 3299.20 km2 study area was impacted by the wildfire, with a concentration of moderate severity impacting estimated 70% of the burned extent. A consensus method combining eight different detection methods revealed high levels of agreement, with NDVI loss ( >0.25) showing the strongest correlation (95.62%) with the NBR-based reference. The spectral index comparisons exhibited notable re- ductions in vegetation greenness with (NDVI 0.464 to 0.199; EVI 0.409 to 0.182) and an increase in moisture stress (MSI 0.799 to 1.334) post-fire. The spatial pattern of burn severity depicted heterogeneity consistent with typical savanna fire dynamics. The multi-method assessment highlights the utility of integrating complementary spectral indices to enhance the accuracy of fire impact evaluations.