Enhancing Drought Monitoring in Taiwan’s Mountainous Terrain Using the Surface Water Availability and Temperature (SWAT)

Droughts pose a major global challenge, particularly in Taiwan, where critical industries such as semiconductor manufacturing are significantly impacted. The mountainous terrain, which constitutes 70% of Taiwan, complicates the estimation of Land Surface Temperature (LST) due to surface heterogeneity. Accurate drought estimations necessitate consistent LST retrieval methods. This study employs a Machine Learning (ML)-based normalization method linked to surface variables to enhance LST accuracy. We introduce the Surface Water Availability and Temperature (SWAT), integrating the improved LST, Normalized Difference Latent Heat Index (NDLI), and Normalized Difference Vegetation Index (NDVI). The SWAT, along with existing indices, was used to assess drought conditions in Taiwan from 2001 to 2023. These results were validated against satellite indicators such as the Crop Water Stress Index (CWSI) and Net Primary Productivity (NPP). Our findings reveal that the SWAT correlates strongly with the CWSI and NPP, indicating significantly higher sensitivity to drought status compared to existing indices. Additionally, the SWAT demonstrated high temporal consistency with the CWSI and NPP across most regions of Taiwan. Generally, the SWAT, supported by the ML-based LST normalization method, proves to be a robust index for monitoring drought conditions in mountainous regions.