Development and validation of a composite drought index CDI-SE for the water-receiving area of China's Tao River diversion project

Amid global warming, compound meteorological droughts—driven by the synergistic effects of declining precipitation and increasing evapotranspiration—are occurring with greater frequency, posing serious challenges to water resource management. Conventional single-factor indices such as the SPEI and EDDI fall short in fully capturing the synergistic disaster-causing mechanisms of such compound droughts. In this study, the water-receiving area of China's Tao River Diversion Project was selected as the research region. A novel compound drought index (CDI-SE) was developed by integrating SPEI and EDDI using an empirical Copula function. Based on monthly and annual data from 1985 to 2022, the performance of CDI-SE was comprehensively evaluated through a multi-dimensional validation framework. This included trend analysis (Mann–Kendall test, Sen's slope estimator), change-point detection (Lee–Heghinian method), periodicity decomposition (CEEMD), run theory, and machine learning models (Gradient Boosting, Random Forest, SVM, and Linear Regression).The results indicate that: (1)Temporally, CDI-SE exhibits high consistency with benchmark indices in identifying trends, detecting change points, and recognizing extreme drought events; (2) In terms of periodic characteristics, it integrates the advantages of SPEI and EDDI, achieving a balanced response across multiple timescales from monthly to interdecadal; (3) Spatially, it reflects a more coherent drought distribution pattern and effectively avoids the spatial bias inherent in single-factor indices. Pearson correlation analysis, drought event consistency test, and machine learning validation all confirm the rationality of the CDI-SE construction, while independent validation based on NDVI preliminarily verifies the reliability of CDI-SE. The CDI-SE index proposed in this study features a clear physical basis and strong comprehensive characterization capability, which can serve as a scientific tool for the accurate monitoring, mechanism analysis and risk assessment of regional compound droughts. • A novel composite drought index CDI-SE was developed by integrating SPEI and EDDI via an empirical Copula function, capturing the synergistic effects of reduced precipitation and increased evapotranspiration. • CDI-SE effectively integrates the strengths of SPEI and EDDI, demonstrating high consistency in trend detection, change-point identification, and multi-scale periodic decomposition across monthly to interdecadal timescales. • Spatially, CDI-SE exhibits a more coherent drought distribution pattern with lower spatial bias compared to single-factor indices, as confirmed by run theory and coefficient of variation analysis. • The reliability of CDI-SE was preliminarily validated through multiple approaches including machine learning models and independent NDVI-based verification, with the 1-year lagged Spearman correlation reaching 0.448 ( p < 0.05).