Xinjiang is located in the mid-latitude region of Eurasia in northwestern China. Precipitation is predominantly concentrated in northern Xinjiang, while southern Xinjiang remains comparatively arid. Summer precipitation accounts for 54.4 % of the annual total. Changes in summer precipitation have a significant impact on the ecological environment and economic development of Xinjiang. Therefore, accurate and reliable precipitation forecasts are critically important for Xinjiang’s adaptation to climate change, regional water resources management, and the sustainable development of the core area of the Silk Road Economic Belt. The Central Asian Climate Team at the Desert Meteorology Institute has focused on the challenge of summer precipitation prediction in Xinjiang. Utilizing observational precipitation data from 95 meteorological stations in XJ and 130 climate indices, the SHAP (SHapley Additive exPlanations) method was applied in combination with an extreme tree model to quantify the contributions of variables towards precipitation. Artificial neural networks, support vector machines, and extreme gradient boosting were considered to predict summer precipitation. The study found: (1) the ANN model demonstrated robust performance during both training and prediction periods. For Northern and Southern XJ, MAE and RMSE values of the ANN model were 15.34 (20.40) and 23.21 (30.01), respectively, surpassing the predictive accuracy of other ML models. (2) NINO B SSTA, PSHI, WPSHI, and MEI were identified as the four most important predictor variables for summer precipitation in NXJ. The NINO B SSTA was the predominant climatic variable linked to a reduction in summer precipitation in NXJ. Additionally, a weakened PSHI or WPSHI reduced summer precipitation in NXJ. (3) Sea surface temperatures across the tropical Indian Ocean and tropical Pacific Ocean were closely linked to summer precipitation anomalies in Xinjiang.

Figure1 Technical Routine

Our research provides valuable insights into predicting summer precipitation in Xinjiang, enabling other researchers to utilize the key predictive variables identified in this study without the need for additional variable selection. Additionally, we investigated the physical mechanisms between these key predictors and precipitation, which has deepened our understanding of the processes governing summer precipitation in Xinjiang. The related results were published under the title “Prediction of Summer Precipitation Via Machine Learning with Key Climate Variables: A Case Study in Xinjiang, China” in the SCI Q1 Journal of Hydrology: Regional Studies. The corresponding author of the paper is Dr. Junqiang Yao from the Institute of Desert Meteorology, and the first author is Chenzhi Ma, a master's student at Xinjiang University. This work was jointly funded by the "Tianshan Talents" of Xinjiang for Science and Technology Youth Top-notch Talent Support Program, Eco-Hydraulic Engineering Research Center for Cold and Arid Regions, Xinjiang Uygur Autonomous Region (Academician Expert Workstation), and Third Xinjiang Scientific Expedition Program.