A newly integrated model for synthesizing multiple external forcing signals to predict the ridge line of the western North Pacific subtropical high in summer

Abstract

The western North Pacific subtropical high (WNPSH) is a critical atmospheric circulation system influencing weather and climate patterns in eastern China. Accurately predicting its activity is essential for anticipating and managing natural disasters such as droughts and floods in the region. In this study, we employ the decision tree (DT) as the fundamental model and devise a compatible framework to create an integrated model known as the Decision Tree Integration Model for Synthesizing Multiple External Forcing Signals (DTI-MS), to perform classification prediction of the WNPSH ridge line in summer. Within the construction of DTI-MS, the interpretability advantage of DT model is fully exploited. Three significant sources of external forcing signals, namely sea surface temperature in the Pacific, Tropical Indian Ocean and Atlantic Ocean, and snow depth across the Eurasian continent, are identified from distinct DT models, with clear physical meanings. Through the voting strategy, these signals are integrated, laying the groundwork for DTI-MS to synthesize multiple external forcing signals. Regarding prediction performance, DTI-MS achieves a high classification accuracy of 91% in the training set and 93% in the testing set, with recalls for the southerly and northerly positions of the WNPSH in the testing set also exceeding 80%. Compared with the commonly used machine learning models such as DT and Random Forests constructed through general modeling approaches, DTI-MS also demonstrates higher classification accuracy. Overall, DTI-MS not only shows promising potential in predicting the WNPSH ridge line, but its integration framework also provides a new paradigm for short-term climate prediction modeling.