Meteorological Drought Risk Monitoring and Zoning Using Random Forest Model

Drought is a natural phenomenon that can occur in any climate and its harmful effects are much wider and deeper than other natural disasters. This natural phenomenon is considered one of the most harmful and economically, socially, physically and environmentally harmful natural disasters. This study was conducted with the aim of predicting and preparing a hazard map, identifying sensitive areas for sustainable development that reduces risks and enhances resilience. In this study, the layers of precipitation, temperature, climate, normalized vegetation difference index and morphometric indices including topographic wetness index, mass balance index, geographical location index and wind shelter index were used to prepare a drought zoning map in Ardabil province using the random forest model. Variable importance analysis using the permutation index showed that precipitation, temperature, climate are the most important factors affecting drought in Ardabil province, respectively. The study of the drought sensitivity zoning map of this province using the above hybrid model showed that about 50 percent of the study area is in the category of areas with high and very high sensitivity to drought events. The results of the evaluation of the ROC curve with an area under the curve above 0.9 for this model indicate the high accuracy of this model in presenting results.