Document Type : Original Article

Authors

• Sadaf Piri ¹
• Mohammad Ansari ghojghar ²

¹ Department of Reclamation of arid and mountainous regions Engineering, Faculty of Natural Resources, University of Tehran, Iran

² Assistant professor, Department of reclamation of arid and mountainous regions Engineering, Faculty of Natural Resources, University of Tehran, Karaj, Iran.

Abstract

Meteorological drought forecasting is essential for sustainable water resources management and climate risk mitigation in arid and semi-arid regions. This study evaluates the performance of hybrid machine learning and time-series approaches for meteorological drought forecasting in Khuzestan Province, Iran. Monthly precipitation data collected from eight synoptic stations during the 1989–2020 period were used to calculate the Standardized Precipitation Index (SPI) at 1-, 3-, 6-, and 12-month time scales. The proposed hybrid framework integrates ARMA-based temporal dependency analysis, GPR nonlinear learning capability, and AF-based adaptive optimization to improve drought forecasting accuracy. Several standalone and hybrid forecasting models were evaluated using the correlation coefficient (R), root mean square error (RMSE), Nash–Sutcliffe efficiency (NS), and mean absolute error (MAE) indices. The results demonstrated that longer temporal scales, particularly SPI-12 and SPI-6, provided more reliable forecasting performance compared with shorter-term indices. Among the investigated models, the proposed triple hybrid framework achieved the highest predictive accuracy across most stations and SPI time scales. For example, under SPI-12 conditions at Bostan station, the proposed framework achieved R, RMSE, NS, and MAE values of 0.922, 0.177, 0.949, and 0.145, respectively, compared with 0.896, 0.215, 0.909, and 0.185 obtained by the ARMA-AF model. The findings highlight the potential of structured hybrid forecasting frameworks for improving meteorological drought prediction reliability in climatically complex regions.

Keywords

• Autoregressive Moving Average Model
• Gaussian Process Regression
• Machine Learning Model
• Prediction Models

Main Subjects

• Drought