基于Copula的横断山区非平稳性气象干旱特征

Characteristics of Non-Stationary Meteorological Drought in Hengduan Mountains Based on Copula

为评估气候变化背景下横断山区的干旱时空演变特征规律,基于1961—2019年横断山区94个气象站逐月气象数据计算得到非平稳的标准化降水蒸散指数(NSPEI)表征气象干旱,并利用游程理论、单变量和多变量Copula方法,揭示干旱特征在不同重现期下的联合分布特征。结果表明:(1)横断山区发生干旱的频次在33%~80%,南部高于北部,对于不同干旱程度,轻旱比极旱高48%左右,重旱和极旱主要集中在南部,轻旱和中旱基本上在整个横断山区都发生过,但是依然南部高于北部。横断山区不同月尺度的干旱程度存在空间差异性。秋季和夏季的干旱程度小于冬季和春季,3月份的干旱程度要高于其他月份,干旱程度为20.97~38.10。(2)广义帕累托分布函数(GP)是干旱历时的拟合最优单变量分布函数,广义极值分布函数(GEV)是干旱烈度的拟合最优单变量分布函数,Gaussian函数是拟合效果最优的Copula分布函数。(3)干旱烈度的空间分布特征和干旱历时几乎一致,横断山区出现长历时、高烈度概率南部小于北部。大于10年一遇的干旱历时可以超过一年,大于50年一遇的干旱烈度可以达到100。空间分布上,横断山区北部呈现东西向分布,受到纬度影响,在中部和南部呈现南北向的分布主要受到纵向岭谷的地形制约。(4)干旱历时>10.07,干旱烈度>36.96的联合重现期>0.01,同现重现期<0.01,即联合重现期>预定重现期>同现重现期,联合重现期横断山区的中西部缘区和中南部的部分地区联合重现期高于横断山区的其他区域,同现重现期的空间分布与联合重现期相反。总体而言,从干旱频次上来看,横断山区总体上南部要高于北部,而干旱程度北部要高于南部,同时在不同的重现期下,北部干旱风险要高于南部,因此未来在干旱风险防范方面要关注到横断山区的北部地区。

In order to reflect the temporal and spatial evolution characteristics of drought in the Hengduan Mountains under the background of climate change,meteorological drought was expressed by NSPEI obtained based on the monthly meteorological data of 94 meteorological stations in the Hengduan Mountains from 1961 to 2019,and the joint distribution of drought characteristics under different return periods in the Hengduan Mountains were revealed by using the run theory,univariate and multivariate Copula methods.The results showed that:(1)the frequency of drought in the Hengduan Mountains was 33%to 80%,and the frequency of drought in the south was higher than that in the north;with respect to different drought levels,the frequency of mild drought was 48%higher than that of extreme drought,the severe and extreme droughts mainly concentrated in the south,and light drought and moderate drought occurred in the whole Hengduan Mountains,but the frequency was still higher in the south than that in the north;there was spatial difference in drought level at different monthly scales in Hengduan Mountains;the level of drought in autumn and summer was less than that in winter and spring,and the level in March was higher than other months,with the level of drought ranging from 20.97 to 38.10;(2)the generalized Pareto distribution function(GP)was the optimal fitting univariate distribution function of drought duration,the generalized extreme distribution function(GEV)was the optimal fitting univariate distribution function of drought severity,and the Gaussian function was the optimal fitting Copula distribution function;(3)the spatial distribution characteristics of drought severity was almost the same with drought duration.the probability of long duration and high severity in the south was less than that in the north;drought duration that occurred more than once in a decade could last more than a year,and droughts severity that occurred more than once in fifty years could last more than one hundred.The spatial distribution was influenced by latitude in the northern of Hengduan Mountains,and presented the east-west distribution,while the distribution in the center and south presented the north-south distribution,which was mainly restricted by the longitudinal mountains and valleys.(4)drought duration>10.07,drought intensity>36.96,the joint return period>0.01,co-occurrence return period<0.01,meant joint return period>predetermined return period>co-occurrence return period.The joint return period in the central and western margin of Hengduan Mountains was higher than that other areas,and the spatial distribution of the co-occurrence return period was opposite to the joint return period.In general,the drought frequency in southern part of Hengduan Mountain was higher than that in the northern part,and the drought level in the northern part was higher than that in the southern part.Meanwhile,the drought risk in the northern part was higher than that in the southern part under different return periods.Therefore,the attention should be paid to the prevention of drought risk in the northern part of Hengduan Mountain region in the future.