洞庭湖流域高温热浪持续天数预测  

Prediction of duration of high temperature heat wave in the Dongting Lake Basin

作　　者：李丽蓉 黄一民[1] 张秋锦 郝丽婷 杨梓琪 Li Lirong;Huang Yimin;Zhang Qiujin;Hao Liting;Yang Ziqi(College of City and Tourism,Hengyang Normal University,Hengyang 421002,Hunan,China)

机构地区：[1]衡阳师范学院地理与旅游学院,湖南衡阳421002

出　　处：《地理科学》2024年第12期2247-2257,共11页Scientia Geographica Sinica

基　　金：湖南省自然科学基金项目(2024JJ8364、2024JJ6101);湖南省研究生科研创新项目(CX20231239)资助。

摘　　要：目前,高温热浪天气正在由极端事件逐渐成为常态,高温热浪给人体健康、自然生态环境以及社会经济系统带来了严重的不利影响,如何准确预测高温热浪的持续天数是一个亟待解决的问题。洞庭湖流域地处全球三大高温热浪频发区之一的长江中下游和淮河流域,本文利用1984-2020年洞庭湖流域地面气象台站观测资料和4大类高温热浪影响因子资料,通过全子集回归筛选影响流域高温热浪持续天数的显著因子,并采用全子集回归和BP神经网络算法分别建立洞庭湖流域高温热浪持续天数模型。结果表明:(1)洞庭湖流域高温热浪持续天数在20世纪60年代至20世纪70年代初表现为下降;20世纪70年初至20世纪90年中期变化不明显;20世纪90年中期以来呈现显著增加趋势。(2)流域高温热浪持续天数与气温、地表太阳辐射、南亚高压东伸脊点、植被生长状况、气溶胶、地面硬化和城市化呈显著正相关,与降水、相对湿度、西太平洋副热带高压西伸脊点、ENSO变化呈显著负相关。进一步利用全子集回归得出西太平洋副热带高压西伸脊点、ENSO变化、气溶胶、地面硬化和城市化是影响洞庭湖流域高温热浪持续天数的显著因子。(3)基于显著因子的同期观测数据建立高温热浪持续天数预测模型,结果显示BP神经网络模型预测效果优于全子集模型,可作为1984-2020年洞庭湖流域高温热浪持续天数的最佳模型。其中,BP神经网络模型训练集判定系数为0.90,均方根误差为3.03 d;测试集均方根误差为2.51 d,平均绝对百分比误差为1.58%。At present,heat waves(HWs)are gradually becoming the norm from extreme events.HWs have serious adverse effects on human health,natural ecological environment,and socio-economic systems.So,accurately predicting the duration of heatwaves is an urgent problem to be solved.The middle and lower reaches of the Yangtze River and the Huaihe River Basin are one of the three major frequent HWs areas in the world.Our study based on both the observed meteorological data from 1984 to 2020 in the Dongting Lake Basin,which located in the middle reaches of the Yangtze River,and four types of HWs influencing factors data(i.e.,global warming,large scale atmospheric circulation,human activity and land-atmosphere coupling),the significant factors affecting the duration of HWs in the Dongting Lake Basin were screened by full subset regression,and then the duration model of HWs in the Dongting Lake Basin was established by full subset regression and BP neural network algorithm.The results show that:1)The duration of HWs had a decreasing trend from 1960s to early 1970s,and remained stable from early 1970s to the mid-1990s,then shown a significant increase since the mid-1990s.2)The duration of HWs is positively correlated with air temperature,surface solar radiation,east extension ridge of South Asian High,vegetation growth,aerosol,ground hardening and urbanization,and negatively correlated with precipitation,relative humidity,west Pacific subtropical high west extension ridge and ENSO.The west Pacific subtropical high ridge,ENSO change,aerosol,ground hardening and urbanization were identified as the significant factors of the duration of HWs with the full subset regression model.3)A prediction model for the duration of HWs is established based on the observed data of significant factors during the same period.BP neural network model has better performance than the whole subset model.Thus it could be used as the model for the duration of HWs in the Dongting Lake Basin from 1984 to 2020.

关 键 词：高温热浪 持续天数 影响因子 全子集回归 BP神经网络算法 

分 类 号：P467[天文地球—大气科学及气象学]