【作者】 李敏；

【导师】 冯平；

【作者基本信息】 天津大学 ， 水文学及水资源， 2019， 博士

【摘要】 在变化环境下,受到气候变化与人类活动的影响,水循环过程中的降水、径流等水文系列特征发生了一定程度的变化,导致基于一致性假设的传统水文分析方法受到质疑。因此,在不断变化的环境中开展径流非一致性分析计算问题研究,具有重要的科学意义和实际的应用价值。本文以受气候变化和人类活动影响较为显著的潘家口水库流域为例,探讨区域变化环境下径流非一致性分析计算的典型问题,主要研究内容及主要成果如下:(1)环境变化下时间序列模型参数的贝叶斯分析。利用时间序列模型对1961～2010年的潘家口入库径流序列进行了模拟,考虑到环境变化对径流序列的影响,分别采用了极大似然法、贝叶斯方法以及考虑径流序列突变点的贝叶斯方法,进行了模型参数的估算,并比较了三种模型的模拟以及预测效果。结果表明,1961～2010年的潘家口入库径流序列为非平稳序列,且1979年为最可能变异点。三种模型中,考虑径流序列突变点且基于贝叶斯方法进行参数估计的时间序列模型,具有最高的样本拟合精度和预测精度。进一步证实了变化环境下,潘家口入库径流序列发生了突变。(2)气候变化对径流频率非一致性影响及不确定性分析。主要分析了未来气候变化对潘家口水库流域入库径流的非一致性影响,并量化了非一致性径流频率分析中的各种不确定性来源。基于Generalized Additive Models for Location,Scale and Shape(GAMLSS)模型分别以时间和降水为潘家口水库流域入库径流的协变量,建立并优选出最优的非一致性模型,把通过统计降尺度方法SDSM模拟得到的未来降水序列代入最优模型的参数中,得到未来径流频率曲线。为了分析不同的全球气候模式GCMs(MIROC-ESM-CHEM、Can ESM2、BCC-CSM1.1)、不同的排放情景RCPs(RCP2.6、RCP4.5、RCP8.5)以及GAMLSS模型参数的不确定对模拟得到的径流的影响,采用方差分析方法量化了这三种不确定性因素及其相互作用对径流的影响。结果表明,非一致的GAMLSS模型对径流序列的拟合效果优于一致的GAMLSS模型,且以降水为协变量的非一致GAMLSS模型又要优于以时间为协变量的GAMLSS模型。GCMs和GAMLSS模型参数对径流不确定性的影响最大,分别占总不确定性源的14%和83%,而RCPs以及不确定性因素之间的相互作用的影响,分别占2%和1%。进一步分析表明,降水序列的波动是导致非一致GAMLSS模型统计参数的成为主要不确定性来源的主要因素,再次证明了将降水序列作为径流频率分析中的协变量的合理性。(3)气候变化对不同季节的径流相关结构的影响。主要研究了未来(2031～2080)气候变化对潘家口水库流域春、夏、秋径流序列及其相关关系的影响。对研究区不同季节的降水与径流的相关性以及不同季节之间的径流的相关性进行了检验,基于不同季节的降水与径流的相关性建立并优选拟合了不同季节径流序列的一致边缘分布,以及修正后的三种排放情景RCPs下的降水序列为协变量的非一致边缘分布。基于不同季节径流间的相关性,以优选出的不同季节径流的边缘分布,建立了具有常数(一致性结构)和时变相关参数(非一致性结构)的Copula模型,比较了在不同的情况下具有一致性结构和非一致性结构的Copula模型的区别。结果表明,用于拟合春、夏和秋季径流序列的非一致性边缘分布优于一致性的边缘分布,证明了春、夏和秋季径流序列具有非一致性。在相同的条件下,具有非一致性结构的Copula模型优于一致性结构的Copula模型,说明了春-夏季与夏-秋径流序列的相关关系为非一致性,且相关关系主要受到夏季降水的影响。在不同的排放场景、不同的时期以及不同的边际概率下,基于一致性结构和非一致性结构的Copula模型对应的联合概率也均不同,说明了春-夏季与夏-秋径流序列的相关关系均受到以上条件的影响。(4)气候变化与人类活动对水文干旱的影响。考虑了大尺度气候模式和人类活动对水文干旱的影响,以提高变化环境下水文干旱预测模型的准确度。通过遥相关检验并筛选流域内各水文站点与径流相关的气候因子,并计算了对径流产生影响的人类活动因子。基于多变量正态分布,把气候预报因子与人类活动因子代入水文干旱预测模型,计算了不同干旱等级间的转换概率。利用蒙特卡洛模拟方法验证了气候因子与人类活动因子对转换概率的影响,评估并比较了考虑气候预报因子与人类活动因子和不考虑两种因子的模型的预测精度。结果表明,考虑气候因子与人类活动因子的水文干旱预测模型能够明显地提高预测精度。更多还原

【Abstract】 Under the changing environment,influenced by climate change and human activities,hydrological features such as precipitation and runoff have changed to a certain extent in the process of water cycle,which leads to the questioning of the traditional hydrological analysis method based on the stationarity assumption.Therefore,it is of great scientific significance and practical application value to carry out the analysis and calculation of runoff non-stationarity in the changing environment.Taking Panjiakou Reservoir Basin as an example,which is significantly affected by climate change and human activities,this paper discussed the issues of non-stationarity of runoff under changing environment.The main research contents and achievements are as follows:(1)Bayesian analysis of time series model parameters under changing environment.The time series model was used to simulate the runoff series of Panjiakou Reservoir from 1961 to 2010.Considering the impact of environmental changes on the runoff series,the maximum likelihood method,Bayesian method and Bayesian method considering the change point in runoff series were used to estimate the model parameters,and the performance of the three models was compared.The results showed that the runoff series of Panjiakou reservoir from 1961 to 2010 is non-stationary and 1979 is the most probable change point.Among the three models,the time series model considering the change point of runoff sequence and based on bayesian method for parameter estimation has highest sample fitting accuracy and prediction accuracy,which further confirms that the runoff sequence of Panjiakou Reservoir has non-stationary characteristics under the changing environment.(2)Non-stationary analysis and uncertainty of runoff frequency under the changing environment.The main purpose is to analyze the non-stationary impact of future climate change on the runoff of Panjiakou Reservoir Basin,and quantify the sources of uncertainty in the frequency analysis of non-stationary runoff series.Based on Generalized Additive Models for Location,Scale and Shape(GAMLSS),time and precipitation were taken as covariates of runoff series of Panjiakou Reservoir,and the optimal non-stationarity model was established and optimized.The future precipitation series simulated through statistical downscaling method were substituted into the parameters of the optimal model to obtain the future runoff frequency curve.In order to analyze the effects of the uncertainties of GCMs(MIROC-ESM-CHEM,Can ESM2,BCC-CSM1.1),RCPs(RCP2.6,RCP4.5,RCP8.5)and GAMLSS model parameters on the simulated runoff,the variance analysis method was used to quantify the three uncertainties and their mutual effects.The results showed that the performance of non-stationary GAMLSS model is better than that of stationary GAMLSS model,and the non-stationary GAMLSS model with precipitation as covariate is better than that of time as covariate.The parameters of GCMs and GAMLSS models have the greatest impact on runoff uncertainty,accounting for 14%and 83% of the total sources of uncertainty,while the RCPs and the interaction among the uncertainty accounts for 2% and 1% respectively.Further analysis showed that the fluctuation of precipitation is the main factor leading to the uncertainty of non-stationarity GAMLSS model parameters,which proves again the rationality of using precipitation series as a covariable in runoff frequency analysis.(3)Effects of climate change on runoff related structures in different seasons.The main purpose is to study the impact of future(2031～2080)climate change on the spring,summer and autumn runoff series and their correlation in Panjiakou Reservoir Basin.The correlation between rainfall and runoff and the correlation between runoff in different seasons in the study area were tested.The uniform marginal distribution fitting different seasonal runoff series and the nonstationary marginal distribution taking the precipitation series under the three revised emission scenarios RCPs as covariates were established and optimized.Based on the correlation between runoff in different seasons,the Copula model with constant(stationary structure)and time-varying correlation parameters(non-stationary structure)were established respectively based on the optimize the marginal distribution of runoff in different seasons.The two types of models were compared under different conditions.The results showed that the runoff series in spring,summer and autumn fitted by the non-stationary marginal distributions are better than that fitted by the stationary marginal distribution,which proves that the runoff series in spring,summer and autumn are non-stationary.Moreover,under the same conditions,the Copula models with non-stationary structure were superior to the stationary Copula model with stationary structure,which indicated that the correlation between spring-summer and summer-autumn runoff series are non-stationary.Under different emission scenarios,different periods and different marginal probabilities,the Copula models with on stationary and nonstationary structures were also performed differently,which indicated that the correlation between spring-summer and summer-autumn runoff series are affected by the above conditions.(4)Effects of climate change and human activities on hydrological drought.The main purpose is to consider the impact of large-scale climate models and human activities on hydrological drought in order to improve the accuracy of hydrological drought prediction models under changing environments.The climatic factors related to runoff at hydrological stations in the basin are screened by remote correlation test,and the human activity factors which influence runoff during human activity period are constructed.Based on the multivariate normal distribution,the selected climatic and human activity factors were substituted into the hydrological drought prediction model and calculated.The Monte Carlo simulation was applied to verify the influence of climate factors and human activity factors on the transformation probabilities.The prediction accuracy of the models considering climate prediction factors and human activity factors and not considering the two factors were evaluated and compared.The results showed that the hydrological drought prediction model considering climate factors and human activity factors can significantly improve the prediction accuracy.更多还原