Fire danger and wildfire occurrence in Botswana’s savannas: evaluating fire-danger indices and early-warning thresholds using 2001–2024 observations

Abstract. Fire danger indices are important tools for operational fire management, yet their performance in African savannas remains poorly evaluated. We assessed eight widely used indices, Fine Fuel Moisture Code (FFMC), Forest Fire Weather Index (FWI), McArthur Forest Fire Danger Index (FFDI), Initial Spread Index (ISI), Drought Moisture Code (DMC), Buildup Index (BUI), Drought Code (DC), and Keetch–Byram Drought Index (KBDI) against 24 years (2001–2024) of Moderate Resolution Imaging Spectroradiometer (MODIS) derived fire occurrence data across Botswana. Index values were calculated from ERA5 Land reanalysis data and evaluated using Spearman rank correlation, logistic regression with cross-validated area under the receiver operating characteristic curve (AUC) and percentile-based odds ratios across annual, dry-season, and wet-season periods. Fine fuel moisture indices consistently outperformed deep drought codes. FFMC showed the strongest discrimination (AUC = 0.821; P90 OR = 9.08), followed by FFDI, FWI, and ISI, whereas DC and KBDI ranked lowest. Fire occurrence odds were approximately seven to nine times higher at or above the 90th percentile than below the threshold, with the steepest increases occurring above approximately the 70th–75th percentile range. No index was significantly correlated with seasonal fire totals (ρ < 0.22, p > 0.05), indicating that these indices are better suited to assessing daily fire danger than cumulative seasonal activity. The findings support a fire danger monitoring framework for Botswana in which FFMC serves as the primary daily indicator, supported by FFDI and FWI, with alert thresholds anchored to the 75^th and 90^th percentiles of a locally derived climatology.