Real-time assessment of live forest fuel moisture content and flammability by using space-time universal kriging

Abstract

Despite the critical role that live fuel moisture content (LFMC) plays in shaping both fire occurrence and behaviour, integration of this factor in wildfire risk assessment remains constrained. Similarly, although flammability is a key factor, its cartographic representation at landscape level poses serious challenges, primarily due to the reliance on bench-scale laboratory experiments for obtaining data. This study aimed to evaluate the spatial and temporal dynamics of LFMC and fuel flammability quantified by the peak heat release rate (PHRR), within a fire-prone forest region in southern Spain. This vulnerable Mediterranean ecosystem is characterized by the prevalence of Pinus pinea L. forests and Cistus ladanifer L. shrublands. LFMC was assessed in fifteen field surveys spanning two fire seasons, across thirty-eight sampling plots, by spatio-temporal universal kriging (UK). Similarly, flammability was assessed in eight surveys, including one fire season, across eight sampling plots, by spatio-temporal universal cokriging (UCK). The auxiliary variables considered were temperature, seasonality, insolation and spectral indices derived from Sentinel-2 and MODIS satellite-derived data. The resulting models exhibited good accuracy, with RMSE values ranging from 11.78 % to 11.89 % for LFMC calibration and between 19.84 % and 20.15 % for the validation data set. Similarly, regarding flammability, RMSE values ranged from 24.08 % to 24.10 % for calibration and between 30.63 % and 30.66 % for validation. LFMC and flammability maps were generated. Temporal autocorrelation in the LFMC models had a significant impact on their performance, whereas PHRR demonstrated a stronger influence through spatial autocorrelation. These unprecedented findings are of great importance in fire behaviour analysis, as the concurrent use of LFMC and PHRR appears to yield diverse yet complementary insights. The use of these techniques, previously unexplored for this specific purpose, marks a significant advance in the field of forest fuel modelling and fire risk evaluation.