Metaheuristic-optimized neuro-fuzzy models for meteorological drought prediction

Abstract Drought is a recurring natural hazard with significant impacts on ecosystems, agriculture, and water resources. Accurate forecasting is essential for timely mitigation strategies. This study presents an advanced hybrid modeling framework for meteorological drought prediction in Baden-Württemberg, Germany, based on the Standardized Precipitation Index (SPI). We integrate the Adaptive Neuro-Fuzzy Inference System (ANFIS) with three state-of-the-art metaheuristic optimization algorithms—Harris Hawks Optimization (HHO), Moth-Flame Optimization (MFO), and Multi-Verse Optimization (MVO)—to enhance predictive accuracy, stability, and interpretability. The proposed models optimize fuzzy inference system parameters, such as membership function centers and widths, to minimize prediction errors. SPI values at multiple time scales (SPI3–SPI12) are predicted using lagged SPI inputs, with training and testing phases designed to ensure robust performance evaluation. Comparative analysis demonstrates that the hybrid ANFIS–MVO model consistently outperforms the other hybrid variants, particularly for longer SPI time scales. Additional comparisons with ARIMA, SVR, and ANFIS–PSO further confirm the effectiveness of the proposed framework. The results highlight the potential of combining neuro-fuzzy systems with metaheuristic optimizers to improve regional drought forecasting, offering a promising tool for early warning systems and water resource management in climate-sensitive regions.