Integrating remotely sensed fuel variables into wildfire danger assessment for China

Abstract

As regulated by the ‘fire environment triangle’, three major forces are essential for understanding wildfire danger: (1) topography, (2) weather and (3) fuel. Within this concept, this study aimed to assess the wildfire danger for China based on a set of topography, weather and fuel variables. Among these variables, two remotely sensed key fuel variables, fuel moisture content (FMC) and foliage fuel load (FFL), were integrated into the assessment. These fuel variables were retrieved using radiative transfer models from the MODIS reflectance products. The random forest model identified the relationships between these variables and historical wildfires and then produced a daily updated and moderate-high spatial resolution (500 m) dataset of wildfire danger for China from 2001 to 2020. Results showed that this dataset performed well in assessing wildfire danger for China in terms of the ‘Area Under the Curve’ value, the fire density within each wildfire danger level, and the visualisation of spatial patterns. Further analysis showed that when the FMC and FFL were excluded from the assessment, the accuracy decreased, revealing the reasonability of the remotely sensed FMC and FFL in the assessment.