首页|基于逐步多元线性回归和随机森林模型预测黄河流域极端气温事件

基于逐步多元线性回归和随机森林模型预测黄河流域极端气温事件

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
全球变暖背景下,极端气候事件频发,且对黄河流域等地区的经济发展及人民生活造成严重危害.基于 1961-2020 年黄河流域 80 个站点的日气温数据提取了 6 个逐月极端气温指数(ETI).利用多重共线性分析去除有相依性的环流指数,并考虑滞后性进行Pearson相关分析,筛选出各ETI的关键环流指数及最佳滞后时间;之后基于最佳滞后时间下的关键环流指数建立逐步多元线性回归(SMLR)和随机森林(RF)模型.对模型进行精度评价,探究环流指数在单站点及整个流域的重要性,并预测了 2022 年 11 月的 6 个ETI值.结果表明:黄河流域 ETI中最高气温(TXx)、暖昼天数(TX90p)、酷热天数(TD30)和最低气温(TNn)呈波动上升趋势,而霜冻天数(FD0)和冷夜天数(TN10p)呈下降趋势;极端高温事件的强度和发生频率的空间分布特征与极端低温事件基本相反.以靖远站TXx为例,各关键环流指数对TXx具有不同程度的影响(0.10<rmax<0.89),rmax对应的最佳滞后时间主要为 5、6、11、12 个月.SMLR和RF模型对黄河流域各ETI的预测能力都较好,验证期的决定系数(R2)范围分别为 0.53～0.95 和 0.64～0.95;除对TXx的模拟效果稍弱外,其他 5 个ETI的RF模型模拟效果均优于SMLR模型.太平洋区极涡强度指数(PPVI)是影响黄河流域TXx、TNn、TX90p和FD0 的最重要环流因子,北非—北大西洋—北美副高脊线位置指数(NANRP)对TN10p和TD30 的影响最大.预测的 2022 年 11 月ETI的空间分布特征与多年平均情况基本相似.研究结果为黄河流域极端气温事件预报提供了参考.
Prediction of extreme temperature events in the Yellow River Basin of China using the SMLR and RF methods
Under the background of global warming,extreme weather events occur frequently,and cause serious harm to the economic development and people's lives in the Yellow River Basin and other regions.Therefore,it is necessary to explore its response to the atmospheric circulation and predict extreme temperature events in the Yellow River basin.Based on the daily temperature data of 80 stations in the Yellow River Basin from 1961 to 2020,the six monthly extreme temperature indexes(ETI)were calculated.Multi-collinearity analysis was used to remove the dependent circulation indexes,and the Pearson correlation analysis was conducted considering the lag.The key circulation indexes of each ETI were selected,and the optimal lag time of the circulation index to each ETI was determined according to the maximum value of Pearson correlation coefficient(r).Then,stepwise multiple linear regression(SMLR)and random forest(RF)models were established based on the selected key circulation indexes with specific lag time to evaluate the accuracy and explore the variable importance of circulation indexes at a single station and the whole basin.Six ETI of 80 stations in the Yellow River Basin in November 2022 were predicted.The results showed that:the TXx,TX90p,TD30 and TNn in the ETI of the Yellow River Basin showed a fluctuating upward trend,while the FD0 and TN10p showed a downward trend.Spatial distribution characteristics of extreme temperature warm indexes and cold indexes were basically opposite.Taking TXx of Jingyuan station as an example,each key circulation index had different degrees of influence on TXx(0.10<rmax<0.89),and the lag time corresponding to rmax was mainly concentrated in 5,6,11 and 12 months.Both SMLR and RF models had good predictive ability for ETI in the Yellow River Basin,with R2 of 0.53～0.95 and 0.64～0.95 in the validation period,respectively.Except for TXx,RF model had better simulation effect on the other five ETI than SMLR model.For the Yellow River Basin,the Pacific Polar Vortex Intensity Index(PPVI)was the most important predictor of TXx,TNn,TX90p and FD0,and the North African-North Atlantic-North American Subtropical High Ridge Position Index(NANRP)had the greatest influence on TN10p and TD30.The predicted extreme temperature indexes in November 2022 were basically similar to the multi-year average in spatial distribution.The results provide a reference for the prediction of extreme temperature events in the Yellow River Basin.

extreme temperature indexcirculation indexrandom forest modelstepwise multiple linear regression modelYellow River Basin

陈俊清、李毅、王斌、杨雪宁、刘峰贵

展开 >

西北农林科技大学 水利与建筑工程学院,教育部旱区农业水土工程重点实验室,陕西 杨凌 712100

水利部黄河流域水治理与水安全重点实验室(筹),河南 郑州 450003

澳大利亚新南威尔士州初级产业部,澳大利亚 新南威尔士州 NSW 2650

青海师范大学 地理科学学院,青海 西宁 810016

展开 >

极端气温指数 环流指数 随机森林模型 逐步多元线性回归模型 黄河流域

水利部黄河流域水治理与水安全重点实验室(筹)研究基金国家自然科学基金项目国家自然科学基金项目中国科学院地球环境研究所黄土与第四纪地质国家重点实验室开放基金项目

2022-SYSJJ-015207911452350410451SKLLOG2125

2024

10.13577/j.jnd.2024.0107

自然灾害学报
中国地震局工程力学所 中国灾害防御协会

自然灾害学报

CSTPCD北大核心
影响因子：0.862
ISSN：1004-4574
年,卷(期)：2024.33(1)
  • 1
  • 33