Evaluating Multi-Index Remote Sensing Frameworks for Drought Monitoring in Seasonally Dry Dipterocarp Deciduous Forests: A Case Study of Yok Don National Park, Vietnam

Abstract

This study examined the effectiveness of the normalized difference vegetation index (NDVI), land surface temperature (LST), and their integrated indices in capturing dry-season drought dynamics in the dipterocarp deciduous forest ecosystems of Yok Don National Park from 2014 to 2024. Results indicate that individual remote-sensing indices captured drought processes but reflected different dimensions of drought conditions, with NDVI indicating canopy greenness and LST detecting early thermal stress. Integrated indices partially mitigated the limitations of single indicators during confirmed drought years by showing stronger correlations with reference data. The effectiveness of the integrated indices varied significantly depending on the integration framework. The spatial-normalization–based integration of NDVI and LST showed consistently strong correlations with surface evapotranspiration (ET), with correlation coefficients of approximately 0.8 across years, and was less sensitive to land-cover heterogeneity. In contrast, the triangular NDVI–LST space integration exhibited variable relationships with ET due to its sensitivity to edge definition and phenological shifts in dipterocarp deciduous forests. Phase-dependent behavior was observed, with integrated indices maintaining more stable relationships with ET during drying periods, while NDVI showed greater sensitivity under wetter conditions. The results also indicate that VHI demonstrated a more robust and stable performance than TVDI, suggesting its suitability as a primary indicator in heterogeneous forest environments. These findings provide practical guidance for drought monitoring in heterogeneous forest landscapes by identifying adaptive integration strategies tailored to different hydroclimatic conditions.