基于GAMLSS模型的三峡库区主汛期降雨非一致性分析
高 斌1,肖伟华1,鲁 帆1,崔 豪1,侯保灯1,张星娜2,杨 恒1,3

(1.中国水利水电科学研究院 流域水循环模拟与调控国家重点实验室,北京 100038; 2.北京交通大学 计算机与信息技术学院,北京 100044; 3.中国长江三峡集团有限公司 科学技术研究院,北京 100038)

主汛期降雨; 三峡库区; 非一致性; GAMLSS模型; 气温

Analysis of Rainfall Non-stationary in the Main Flood Season in the Three Gorges Reservoir Area Based on GAMLSS Model
GAO Bin1,XIAO Weihua1,LU Fan1,CUI Hao1,HOU Baodeng1,ZHANG Xingna2,YANG Heng1,3

(1.State Key Laboratory of Simulation and Regulation of Water Cycle in River Basin,China Institute of Water Resources and Hydropower Research,Beijing 100038,China; 2.School of Computer and Information Technology,Beijing Jiaotong University,Beijing 100044,China; 3.Institute of Science and Technology,China Three Gorges Corporation,Beijing 100038,China)

rainfall during the main flood season; Three Gorges Reservoir area; non-stationary; GAMLSS model; temperature

备注

采用广义可加模型(GAMLSS),以时间t和4种气温形式(年平均气温Tave、年平均最高气温Tmax、年平均最低气温Tmin和年平均温差ΔT)为协变量对三峡库区12个站点主汛期降雨量(Pmfs)的非一致性进行了研究,为变化环境下防洪、抗旱、水利工程安全评估等提供依据。结果如下:(1)以t为协变量拟合时,50%的站点的Pmfs呈现出非一致性特征。其中,库首秭归地区和库尾沙坪坝地区Pmfs的均值不断增大(秭归地区增速达29.6 mm/10 a),库尾江津地区Pmfs的方差不断增大,这些地区可能在未来引发洪涝、滑坡、干旱等自然灾害。(2)以气温为协变量拟合时,库区各站的Pmfs均呈现出非一致性,模型拟合效果总体优于前者且具有物理意义。库首以Tave为协变量时,模型拟合效果最好; 库腹、库尾总体上以ΔT为协变量时,拟合效果最好。此外,库首秭归地区Pmfs在1998年左右发生了均值突变,均值由从原来的480 mm突增至580 mm左右,可能是由于人类活动或气候变化导致Tave变化引起的; 库尾的江津地区Pmfs的方差逐渐增大,风险同以t为协变量的拟合结果。
The GAMLSS model was used to study the non-stationary of rainfall during the main flood season(Pmfs)at 12 stations in the Three Gorges Reservoir area; time t and four types of temperature(annual average temperature Tave,annual average maximum temperature Tmax,annual average minimum temperature Tmin and annual average temperature difference ΔT)were taken as the covariates so as to provide the basis for flood control,drought resistance,and water conservancy project safety assessment under changing circumstances. The results are shown as follows.(1)With t being taken as a covariate for fitting,50% of the sites shows inconsistent characteristics. The mean value of Pmfs in Zigui area at the head of the reservoir and Shapingba area at the end of the reservoir are increasing(the growth rate in Zigui area reaches 29.6 mm/decade),while the variance of Pmfs in the Jiangjin area at the end of the reservoir is constantly increasing,which indicates that these areas may cause natural disasters such as floods,landslides,and droughts in the future.(2)While the temperature is used as the covariate to fit,Pmfs of each site in the reservoir area shows non-stationary,and the fitting effect of the model is better than that of the former on the whole and is of physical significance at the same time. While Tave is used as the covariate at the head of the reservoir,the model fitting effect is the best; while ΔT is generally taken as the covariate at the middle and the end of the reservoir,the fitting effect is the best. In addition,the mean value of Pmfs in Zigui area at the head of the reservoir had a sudden change in the mean value around 1998. The mean value changed abruptly from 480 mm to about 580 mm,which might be due to the changes of Tave caused by human activities or climate change; the variance of Pmfs in Jiangjin area at the end of the reservoir gradually increases,and the risk is the same as the fitting result with t as the covariate.