Integrated fire risk assessment using time-series satellite data and multi-criteria decision-making techniques

Abstract

Fire risk assessment is a vital aspect of forest management and strategic planning. This study develops an integrated fire risk model using time-series satellite data to identify key vegetation, anthropogenic, and climate-related factors. Analysis was conducted on the Google Earth Engine (GEE) platform, focusing on two forested regions in Kurdistan, Iran, with high-quality imagery from 2013 to 2024. Relevant indices derivable from satellite data were ranked and weighted using Shannon Entropy and the TOPSIS technique, based on six criteria: intensity, cumulative impact, cause-effect relationship, data validation, data availability, and recurrence in research studies. Additionally, index weights were determined through the Analytic Hierarchy Process (AHP) based on survey responses. These indices were mapped using the Weighted Linear Combination (WLC) approach. High-resolution imagery from Landsat-8, Sentinel-2, MODIS, and other sources was used to map indices including dNBR, NDVI, NDWI, Land Surface Temperature (LST), precipitation, slope, aspect, elevation, and distances to settlements, rivers, and roads. Results showed that recurrence in research studies, with the lowest entropy value of 0.9320, significantly influenced the selection of effective indices. Among eleven indices, dNBR ranked the highest in importance. Sensitivity analysis revealed that LST had the strongest influence on wildfire risk, highlighting its critical role under changing climate conditions. Receiver Operating Characteristics (ROC) were used to assess the accuracy of the fire risk model. For the Marivan and Sarvabad study areas, the AUC had a value of 0.93. The results of this study can assist decision-makers to prevent forest fires and minimize forest damage in Kurdistan province.