一种区域降水预测的时空集成模型  

A spatiotemporal integrated model for regional precipitation prediction

作　　者：马宇红 薛生倩 王小小 路金叶 MA Yu-hong;XUE Sheng-qian;WANG Xiao-xiaO;LU Jin-ye(College of Mathematics and Statistics,Northwest Normal University,Lanzhou 730070,Gansu,China;Editorial Department of the University Journal,Northwest Normal University,Lanzhou 730070,Gansu,China)

机构地区：[1]西北师范大学数学与统计学院,甘肃兰州730070 [2]西北师范大学学报编辑部,甘肃兰州730070

出　　处：《西北师范大学学报（自然科学版）》2024年第3期105-114,共10页Journal of Northwest Normal University(Natural Science)

基　　金：国家自然科学基金资助项目(51368055)。

摘　　要：提出了一种融合多种深度学习方法的时空预测集成模型LSTM-ARIMA-SARIMA-BP-CNN(简称LASBC模型),其中ARIMA模型捕捉降水时间序列的近邻性,SARIMA模型捕捉周期性,BP神经网络揭示降水分布的时空相关性,CNN挖掘气象、地理因素对降水的影响,最后通过LSTM网络对4个模型的预测结果进行融合,提高预测精度.以中国西部地区12个城市1985年1月至2021年12月的月降水量数据为主,应用LASBC模型对12个城市的月降水量进行预测,结果显示:LASBC模型的预测精度显著提高;基于预测月降水量,给出了6个主要城市未来10年气候偏干旱或湿润的月份及降水量;未来10年,我国西北干旱区年均降水基本保持稳定,高寒冻土区略有增加,西南湿润区增长趋势明显.A spatiotemporal prediction model LSTM-ARIMA-SARIMA-BP-CNN(LASBC)for regional precipitation is proposed,which integrates multiple deep learning methods.Among them,ARIMA model captures the time-neighbor correlation of precipitation time series,SARIMA model captures the periodic trend of precipitation time series,BP neural network reveals the spatiotemporal correlation of precipitation distribution,CNN network excavates the influence of meteorological and geographical factors on precipitation,and finally,LSTM network is used tOfuses the prediction results of the four models sOas tOimprove the prediction accuracy.Based on the monthly precipitation data of major cities in western China from January 1985 tODecember 2021,the LASBC model is used tOforecast the monthly precipitation of 12 cities in western China.The results show that the prediction accuracy of the LASBC model is greatly improved;the predicting months and precipitations for six major cities with a slightly arid or humid climate in the next 10 years are given based on the predicted monthly precipitation;in the next ten years the annual average precipitation in the northwest arid region has remained stable,with a slight increase in the high-altitude permafrost region and a significant growth trend in the southwest humid region.

关 键 词：降水分布 时空预测 机器学习 集成模型 时空相关性 

分 类 号：TP391.42[自动化与计算机技术—计算机应用技术] R742.5[自动化与计算机技术—计算机科学与技术]