University of Birjand-Research Group of Drought and Climate Change Journal of Drought and Climate change Research 3092-6076 4 1 2026 05 22 Modeling and Optimization of Water Allocation Using a Hybrid LSTM–HHOA Approach Under Climate Change Conditions Modeling and Optimization of Water Allocation Using a Hybrid LSTM–HHOA Approach Under Climate Change Conditions 131 158 3696 10.22077/jdcr.2025.9958.1165 FA Parisa-Sadat Ashofteh Associate Professor, Department of Civil Engineering, University of Qom, Qom, Iran 0000-0003-2561-0510 Leila Shakarami Ph.D. Student, Department of Civil Engineering, University of Qom, Qom, Iran Journal Article 2025 08 22 The present study aimed to optimize water allocation from the Qarnaku Dam for the future period (2070–2099) and enhance the resilience of the water supply system. Among CMIP5 climate models, GFDL-ESM2G was selected as the most suitable option for temperature, and NorESM1-M was selected for precipitation. Subsequently, monthly runoff was simulated using Long Short-Term Memory (LSTM) networks, yielding satisfactory accuracy. Agricultural water demand was estimated using the Cropwat model, indicating a 20% increase in future net irrigation requirements. The performance of the Horse Herd Optimization Algorithm (HHOA) and Genetic Programming (GP) was evaluated using the Rastrigin and Rosenbrock benchmark functions, revealing superior results for HHOA. Further analysis showed that HHOA reduced water deficits by up to 22% and improved reservoir resilience from 67% to 79%. Additionally, the LSTM model achieved a Nash–Sutcliffe Efficiency (NSE) coefficient of 0.85, confirming its reliability in simulating monthly runoff. Overall, the integration of advanced metaheuristic techniques with intelligent runoff prediction models provides an effective framework for reservoir management under climate change conditions. The present study aimed to optimize water allocation from the Qarnaku Dam for the future period (2070–2099) and enhance the resilience of the water supply system. Among CMIP5 climate models, GFDL-ESM2G was selected as the most suitable option for temperature, and NorESM1-M was selected for precipitation. Subsequently, monthly runoff was simulated using Long Short-Term Memory (LSTM) networks, yielding satisfactory accuracy. Agricultural water demand was estimated using the Cropwat model, indicating a 20% increase in future net irrigation requirements. The performance of the Horse Herd Optimization Algorithm (HHOA) and Genetic Programming (GP) was evaluated using the Rastrigin and Rosenbrock benchmark functions, revealing superior results for HHOA. Further analysis showed that HHOA reduced water deficits by up to 22% and improved reservoir resilience from 67% to 79%. Additionally, the LSTM model achieved a Nash–Sutcliffe Efficiency (NSE) coefficient of 0.85, confirming its reliability in simulating monthly runoff. Overall, the integration of advanced metaheuristic techniques with intelligent runoff prediction models provides an effective framework for reservoir management under climate change conditions. https://jdcr.birjand.ac.ir/article_3696_95c9d994f8d75d4d60f8bb8f25902339.pdf