摘要
为解决自然灾害发生频率评估中的偏差问题,利用内蒙古中部强沙尘区的植被返青期和春季大风事件资料,构建了单一类型和多类型混合的Copula函数模型,对比了拟合精度并进行了相关结构的分析,通过计算极端灾害事件对应的各变量的尾部阈值,评估了变量的尾部风险。结果表明:混合Copula函数比单一Copula函数更适合于构建两个特征变量的联合分布模型;两个特征变量具有非对称的尾部关系,且极端下尾有较强的正相关关系,出现极端低值时,两者具有更高的相关性;植被返青期早于90d和春季大风事件小于9次,以及植被返青期晚于223d和春季大风事件大于100次时,两个特征变量有较高的正相关关系,当两者同时取得尾部阈值时,发生强沙尘暴可能性更高。因此利用混合Copula函数能够有效提高灾害特征变量相关结构模型的拟合优度,改善灾害发生频率评估精度。
Aiming at solving the deviation problem of natural disaster frequency assessment,based on the vegetation green up date( G) and spring strong wind event( D) data in central Inner Mongolia,this study established the single as well as mixed Copula model,compared the fitting results of both models,conducted corresponding dependence structure analysis,calculated the tail thresholds of variables that correspond to the extreme events and evaluate the risk of variables in the tails. The results show that: mixed Copula is more suitable for building the joint distribution model of G and D than single Copula; G and D show an asymmetrical pattern in the upper and lower tails. G and D have a strong positive correlation in the lower tail,and the correlations between the two variables become significantly stronger when the extreme low values occur; there is a strong positive correlation between the two variables when G < 90 day and D < 9 or G > 223 day and D > 100. Severe dust storms are more likely to occur when the two variables reach the thresholds. Therefore,the goodness of fit of dependence structure model and the accuracy of frequency assessment can be improved by using mixed Copula function.
引文
[1]王钦秀.一般线性回归模型———参数估计、假设检验、回归预测[J].数量经济技术经济研究,1988(2):39-44.
[2]叶明华,汪荣明,丁越,等.基于Copula相依函数的安徽省气温与降雨量相关性研究[J].长江流域资源与环境,2017,26(1):110-117.
[3] YI W,LIAO S. Statistical properties of parametric estimators for Markov chain vectors based on copula models[J]. Journal of Statistical Planning&Inference,2010,140(6):1465-1480.
[4]淳伟德,付君实,赵如波.基于混合Copula函数的金融市场非线性极端风险传染研究[J].预测,2015(4):53-58.
[5]孙志宾.混合Copula模型在中国股市的应用[J].数学的实践与认识,2007,37(20):14-18.
[6]林宇,陈王,王一鸣,等.典型事实、混合Copula函数与金融市场相依结构研究[J].中国管理科学,2015,23(4):20-29.
[7] Serinaldi F. Copula-based mixed models for bivariate rainfall data:an empirical study in regression perspective[J]. Stochastic Environmental Research&Risk Assessment,2009,23(5):695-696.
[8]操颖,方兆本.基于混合copula函数的沪深股市与香港股市一体化趋势分析[J].中国科学技术大学学报,2014,44(6):508-515.
[9]吴吉林,孟纹羽.时变混合Copula模型的非参数估计及应用研究[J].数量经济技术经济研究,2013(8):124-136.
[10] GAO J,AN Z,LIU B. A dependent stress-strength interference model based on mixed copula function[J]. Journal of Mechanical Science&Technology,2016,30(10):4443-4446.
[11] Nguyen C,Bhatti M I,KomorníkováM,et al. Gold price and stock markets nexus under mixed-copulas[J]. Economic Modelling,2016,58:283-292.
[12]张尧庭.连接函数(copula)技术与金融风险分析[J].统计研究,2002,19(4):48-51.
[13] HU L. Dependence patterns across financial markets:a mixed copula approach[J]. Applied Financial Economics,2006,16(10):717-729.
[14]刘建华.基于混合Copula篮式信用违约互换定价研究[D].大连:大连理工大学,2013.
[15] FENG J,LI N,ZHANG Z,et al. The dual effect of vegetation green-up date and strong wind on the return period of spring dust storms[J]. Science of The Total Environment,2017,592:729-737.
[16] CHEN J,Jonsson P,Tamura M,et al. A simple method for reconstructing a high-quality NDVI time-series data set based on the Savitzky-Golay filter[J]. Remote Sensing of Environment,2004,91(3):332–344.
[17] Tucker C J,Grant D M,Dykstra J D. NASA’s global orthorectified landsat data set[J]. Photogrammetric Engineering&Remote Sensing,2004,70(3):313–322.
[18] Piao S,FANG J,ZHOU L,et al. Variations in satellite-derived phenology in China’s temperate vegetation[J]. Global Change Biology,2006,12(4):672-685
[19]卢锦玲,於慧敏.基于混合Copula的风光功率相关结构分析[J].太阳能学报,2017,38(11):3188-3194.
[20] FENG J,LI N,ZHANG Z,et al. How to apply the dependence structure analysis to extreme temperature and precipitation for disaster risk assessment[J]. Theoretical&Applied Climatology,2018,133:297-305.