A hybrid drought index prediction framework under climate change scenarios by equipping index fusion with an individual artificial intelligence method

Due to the multidimensional nature of drought and the limitations of single-variable indices in capturing combined meteorological, hydrological, and socioeconomicimpacts, predicting hybrid droughts is valuable insights for water resource management. This study proposes an Index-fusion-based framework for hybrid drought modeling in the Marun basin, Iran, during the historical (2002–2021) and future (2022–2040) periods under climate change scenarios (SSP3.70 and SSP5.85). Multiple standard drought indices based on precipitation, inflow, reservoir storage, and discharge data are combined using Shannon entropy (SE), Copula functions (C), and Principal Component Analysis (PCA), respectively, to propose three hybrid drought indices including Shannon Entropy-Hybrid Drought Index (SEHDI), Copula-Hybrid Drought Index (CHDI), and PCA-Hybrid drought index (PCAHDI). After evaluating optimal time scales of univariate indices, the CHDI is chosen for its comprehensive and smoothed representation of drought evolution and better reflect of drought duration and severity. Then, three Individual Artificial Intelligence (IAI) models, including Multi-Layer Perceptron (MLP), Support Vector Regression (SVR), and Random Forest (RF), are used, and the results are applied to predict the HDIs. Here, the MLP and RF models are also used to fuse the IAI modeling to improve predictive accuracy. The results show that, the MLP fusion-based model performswell for CHDI under SSP3.70, and RF fusion-based model performswell for CHDI under SSP5.85. When fusion-based models are used, based on the error metrics, the accuracy of HDI index estimations is significantly improved to 2.4% under SSP3.70. The results suggest an increased risk of drought under the high-emission SSP5.85 scenario, but a decreased risk of drought under the more moderate SSP3.70 scenario, highlighting the usefulness of the proposed framework for drought monitoring and water management planning.