【摘要】 为了提高月降水气候趋势预测能力，利用国家气候中心海气耦合气候模式（BCC＿CSM1.1）环流预测和NCEP/NCAR再分析资料，建立基于多层感知器、支持向量回归和随机森林三种机器学习算法的广西月降水量气候预测模型，对比分析了2018年-2020年6月-8月降水预测试验效果。试验结果表明基于三种机器学习算法模型的预测准确率较BCC＿CSM1.1气候模式直接降水预测和逐步回归方法均有明显提升，其中多层感知器模型预测准确率最高。论文将机器学习方法运用于动力-统计相结合的月降水气候预测方法中，成为统计订正中的一环，实现了机器学习算法与气候动力深度融合，有效地提升了预测准确率。更多还原

【Abstract】 In order to improve the prediction ability of monthly precipitation and climate trends,three types of machine learning based on multi-layer perceptron,support vector regression and random forest are established by using the circulation prediction of the national climate center’s air-sea coupled climate model(BCC＿CSM1.1)and NCEP/NCAR reanalysis data. The algorithm based on the climate prediction model of monthly precipitation in Guangxi,is compared and analyzed the experimental results of precipitation prediction from June to August in 2018-2020. The experimental results show that the prediction accuracy of the models based on the three machine learning algorithms is significantly improved compared with the direct precipitation prediction and stepwise regression methods of the BCC＿CSM1.1 climate model,among which the multi-layer perceptron model has the highest prediction accuracy. In this paper,the machine learning method is applied to the monthly precipitation and climate prediction method combining dynamics and statistics,which has become a part of the statistical correction. It realizes the deep integration of machine learning algorithm and climate dynamics, and effectively improves the prediction accuracy.更多还原