مشخصات پژوهش

Traditional forecasting models decompose time series into multiple subseries but typically simulate all components using a single modeling approach. In this study, we proposed a novel methodology to disaggregate precipitation time series into distinct high- and low-frequency components and simulate them separately for different climates. We tested the proposed method on monthly and annual precipitation, representing arid, semiarid, and humid climates in Iran. The proposed approach demonstrated superior performance compared to single model approaches. Implementing separate modeling of the components improved forecast accuracy by 39–44%. A climate-specific analysis revealed distinct performance patterns. Arid climate stations achieved high precision scores (precision = 0.83), while semi-arid stations showed a moderate accuracy (precision = 0.495). Humid climate stations maintained robust performance (precision = 0.74). The proposed two-step temporal decomposition approach can effectively address the non-stationary and random characteristics of precipitation time series, thereby improving forecast accuracy