Spatial Assessment of Wildfires Susceptibility in Santa Cruz (Bolivia) Using Random Forest

Wildfires are expected to increase in the near future, mainly because of climate changes and land use management. One most vulnerable areas world is forest central-South America, including Bolivia. Despite that this country highly prone wildfires, literature rather limited here. To fill gap, we implemented a dataset burned area occurred department Santa Cruz period 2010–2019, digital spatial data describing predisposing factors (i.e., topography, cover, ecoregions). The main goal was develop model, based on Random Forest, which probabilistic outputs allowed elaborate wildfires susceptibility maps. overall accuracy finally estimated by using 5-fold cross-validation. In addition, last three years observations acted as testing dataset, allowing evaluate predictive performance model. quantitative assessment variables revealed “flooded savanna” “shrub or herbaceous flooded, fresh/saline/brakish water” respectively ecoregions cover classes with highest probability predicting wildfires. This study contributes development validation an innovative mapping tool for fire risk assessment, implementable at regional scale different globe.