Exploration of Cross-Seasonal Forecasting Methods for Summer Precipitation in the Yangtze River Basin

Abstract This study develops a cross-seasonal summer precipitation prediction model for the Yangtze River basin of China by integrating a time-scale separation method. The proposed EEAE method combines the empirical cumulative distribution function (ECDF), ensemble empirical mode decomposition (EEMD), autoregressive integrated moving average (ARIMA), and empirical orthogonal function (EOF) iteration techniques. First, the ECDF corrects the summer precipitation predictions initialized in August by the BCC_CSM1.1m model, merging them with historical observational data to construct a new time series. The EEMD method then decomposes this series into multiple intrinsic mode functions (IMFs) exhibiting distinct periodic characteristics. For each IMF component, tailored prediction models are developed. IMF1, characterized by quasi-biennial oscillations, is predicted by summing the interannual increment with previous year values. IMF2, displaying 6–8-yr periodicities, is forecasted using the ARIMA method. Slow-varying IMFs with periods exceeding 15 years are corrected through EOF iteration of the model’s geopotential height field. The synthesized predictions by the EEAE model demonstrate superior skill compared to ECDF-only corrected forecasts, with consistent results across experiments with varying initialization times. This work highlights that integrating historical data with high-skill dynamical model outputs through time-scale separation can effectively enhance prediction accuracy and extend forecast lead time. Significance Statement This study advances seasonal precipitation forecasting by developing a novel multiscale prediction framework for the Yangtze River basin. Our key innovation lies in decomposing precipitation variability into distinct cyclical components and constructing tailored prediction models for each scale, effectively combining climate model outputs with historical observations. The approach enables skillful summer precipitation forecasts with a lead time of up to 10 months, significantly earlier than conventional methods, while improving prediction accuracy. This work demonstrates the potential of scale-aware modeling to extend forecast horizons and improve accuracy, offering a valuable reference for water resource management and climate adaptation strategies.