Drought risk assessment in Quang Tri Province, Vietnam using Landsat multi-temporal remote sensing data and machine learning algorithm

Drought is a major environmental challenge that significantly impacts agriculture, water resources and ecosystems, particularly in regions prone to arid conditions. This study aims to classify and monitor drought severity using multi-temporal remote sensing data, drought indices and machine learning techniques. Landsat satellite imageries from 2014 to 2024, collected at two-year intervals, are utilized to assess drought patterns in Quang Tri Province, Vietnam. Three key drought indices Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI) and Land Surface Water Index (LSWI) are computed to evaluate vegetation health, surface water content and soil moisture levels. The Random Forest algorithm on Google Earth Engine (GEE) is applied to classify drought into different severity levels based on spectral features extracted from satellite images. The results indicate a clear spatial and temporal variation in drought severity, with coastal areas consistently undergoing extreme drought, whereas central regions show increasing drought expansion over time. Western and southern areas remain relatively stable due to higher vegetation cover and water retention capacity. The study highlights the effectiveness of combining remote sensing data and machine learning in improving drought classification accuracy. The findings contribute to early-warning systems, water resource management and climate adaptation strategies, providing valuable insights for policymakers and land managers.