Evaluating and mapping spatial drought in northeast Thailand: utilizing analytic hierarchy process and random forest algorithms

Drought is a recurring and costly natural disaster that affects all climate zones, posing significant challenges to agriculture and livestock. This study evaluates drought vulnerabilities in Northeast Thailand using geospatial datasets and analysis combined with a random forest (RF) approach. The Analytical Hierarchy Process (AHP) was employed to map different drought types based on 14 spatial criteria. The RF model, trained using geospatial datasets and the Standardized Precipitation Index (SPI) from 96 meteorological stations, demonstrated high accuracy and reliability. The results revealed that approximately 70% of the region is vulnerable to moderate to extreme drought conditions. The AHP model achieved an overall accuracy (OA) of 85.19%, aligning closely with the RF model, which had an OA of 88.89%. The RF algorithm effectively mapped spatial drought events with high accuracy, as confirmed by comparisons with rainfall data. These findings provide essential insights for drought mitigation strategies and regional planning.

干旱是一种周期性频发且损失惨重的自然灾害，可波及所有气候带，对农业与畜牧业造成显著威胁。本研究结合地理空间数据集与分析手段，采用随机森林（Random Forest, RF）模型，对泰国东北部的干旱脆弱性进行评估。研究采用层次分析法（Analytical Hierarchy Process, AHP），基于14项空间评价指标绘制不同干旱类型的空间分布图。基于地理空间数据集与96个气象站获取的标准化降水指数（Standardized Precipitation Index, SPI）训练得到的随机森林模型，展现出优异的精度与可靠性。研究结果显示，该区域约70%的范围面临中度至极端干旱的脆弱性风险。层次分析法模型的总体精度（Overall Accuracy, OA）为85.19%，与总体精度达88.89%的随机森林模型结果高度一致。经与降雨数据比对验证，随机森林算法可高精度地绘制空间干旱事件分布。本研究结果可为干旱减缓策略制定与区域规划提供关键参考依据。