【作者】 田烨；

【导师】 许月萍； Martijn J.Booij；

【作者基本信息】 浙江大学 ， 水工结构工程， 2013， 博士

【摘要】 近百年来,人口剧增、温室气体排放量显著增高,全球平均温度普遍升高,气候变化以及对人类的影响已受到社会各界的普遍关注。水循环作为气候系统中的重要部分,也受到气候变化的影响。近年来,一些地区的极端气候事件,如暴雨、干旱等灾害发生越来越频繁,且强度越来越大。有的极端事件的降雨量和径流量甚至是历史上罕见的。研究气候变化对极端事件的影响对防灾减灾和水资源管理有着极其重要的意义。目前研究气候变化对水文过程的影响主要工具有GCM、RCM、水文模型等。一般的步骤是首先通过GCM得到未来不同情景模式下全球气候变量的变化情况。但通常GCM的输出数据精度太低,不能满足流域尺度的水文研究的较高分辨率的输入要求。如果直接使用会产生较大误差。因此,通过降尺度的方法,包括统计降尺度和动力降尺度（RCM）,来得到更加精细的气候变量数据。然后将气候变量数据输入到水文模型中,对未来该地区的径流做出预估。但是在这个过程中,每个环节都有不确定性,而且这种不确定性是无法完全消除的。因此,如果水资源管理者和决策者想要做出全面稳健的决策,量化气候变化各环节要素对径流的不确定性是必须的。本文首先分析了浙江省历史极端降雨的变化趋势,并分析其成因,按照其降雨成因划分为了梅雨季和台风季。随后选取了浙江省金华江流域为研究区域，构建了气候变化下极端水文事件预测的不确定性分析框架,分析了未来多重不确定因素影响下气候变化对极端径流的影响,包括排放情景不确定性、GCM不确定性、水文模型结构的不确定性和水文参数的不确定性。论文的主要工作和成果如下：（1）用Mann-Kendall法研究了浙江省内18个气象站的七个降雨有关的极端气候指标的历史变化趋势,并分析其成因,按照其降雨成因划分为了梅雨季和台风季,比较了降雨较集中两季的变化趋势及其空间分布。结果显示,浙江省的降雨趋势变化具有自东向西变化的空间分布特点。东部降雨呈上升趋势而西部呈下降趋势。浙江省大部分地区的大雨雨量呈上升趋势。尽管极端雨城的上升情况不是很普遍,但是一些地区的极端雨量上升趋势很明显。降雨强度在大部分地区呈上升趋势,在沿海地区梅雨季和台风季的降雨强度上升显著。（2）研究降雨与极端径流的关系,构建了三个水文模型,利用GLUE法,对水文模型的参数进行不确定性分析,量化水文模型参数对极端径流的影响。同时,评价水文模型结构的不确定性。研究结果表明极端最大径流的不确定性随着径流量增大而增大。本文所用的三个模型中,HBV模型模拟的极端最大径流的不确定性最大,新安江模型模拟的极端最大径流不确定性最小。水文模型对极端最小径流的模拟值比实测值偏低,其中新安江模型对实测值低估最多。水文模型参数的不确定性大于水文模型结构的不确定性。（3）在水文模型结构和水文模型参数不确定性的基础上,本文考虑了全球气候变化下不同情景对极端径流的影响不确定性。采用全球气候模式HadCM3在IPCC第四次报告的A1B、A2和B2三种未来排放情景模式下的输出,并用区域气候模式PRECIS将其降尺度到分辨率为50km×50km的网格中。用DBS法和线性偏差纠正法对PRECIS模型2011-2040年的输出进行了偏差纠正,然后用偏差纠正后的气象数据驱动水文模型。研究结果表明在使用区域气候模式的数据之前进行偏差纠正能使结果与实际更加吻合。A1B、A2和B2情景下极端最大径流有所减小。极端径流的不确定性最主要来源于模型参数,然后是模型结构,最后是排放的情景模式。（4）最后,本文引入了最新的IPCC第五次报告中发布的典型浓度路径RCP2.6、RCP4.5、RCP6.0和RCP8.5作为未来的情景模式,并且考虑了三个不同的大气环流模式BCC_CSM1₁、HadG EM2-ES和GISS_E2_R对极端最大径流的影响,采用LARS-WG天气发生器生成更长的时间序列,利用GR4J、HBV和新安江模型计算了预测期和基基准期的极端最大径流。研究结果表明,未来四种浓度路径下日最低温的上升幅度要大于日最高温的上升幅度。三个GCM中,HADGEM2-ES模型中不同浓度路径的对极端最大径流影响的不确定性最大,其次是GISS_E2_R模型,BCC_CSM1₁模型下不同的浓度路径对未来极端径流的不确定性影响较弱。HADGEM2-ES模型中RCP4.5和RCP2.6浓度路径下极端最大径流有上升的迹象,RCP6.0和RCP8.5下呈下降的趋势。GISS E2R模型中RCP4.5浓度路径下的极端最大径流有下降的迹象。同一浓度路径下不同GCM引起的极端最大径流的不确定性最大的为RCP4.5。本研究中各不确定性来源对极端最大径流的影响由大到小依次为：水文模型参数不确定性>GCM不确定性>浓度路径的不确定性>水文模型结构不确定性。更多还原

【Abstract】 The recent century has witnessed the enormous increase of population, greenhouse gas emission and global mean temperature. The global climate change and its impact on human beings have drawn wide attention from all walks of life. Water cycle as a part of the climate system has been affected by the climate change as well. In recent years, the extreme climate events like rainstorm and drought have taken place with a higher intensity and frequency in some regions. Some extreme precipitation and discharges could be scarcely found even in the historical record. The research on the impact of the climate change on extreme events is crucial for the damage prevention and water resources management.Up to now, the commonly used tools to study the impact of the climate changes on the hydrological processes are GCMs, RCMs, hydrobgical models and so on. There are several steps in assessing the impact. Firstly the climate changes under different scenarios are obtained by the GCMs. Then, the output from the GCMs are downscaled through downscaling methods, including the statistical downscaling and dynamic downscaling, to satisfy the requirement of the resolution for the river basin studies. After that, the outputs from the GCMs are used as the input for the hydrological models to make the estimation of the river discharges. However, the uncertainties are involved in every step above and they could not be eliminated thoroughly. Therefore how to quantify the uncertainties in every step of climate change impact analysis and deal with the uncertainties in the river discharges are the basis for policy makers to make the robust decisions.In the thesis, the trend of the historical precipitation is calculated and the rainfall period is divided into two seasons, the plum season and the typhoon season based on the cause of rainfall. In order to study the impact of climate change on the discharges, the Jinhua river basin was chosen as the study area and the uncertainty resources like emission scenarios, GCMs, parameters and structure of hydrological models are considered. The major work and conclusion of the thesis are like follows:（1） a selection of seven extreme indices is used to analyze the trend of precipitation extremes of18meteorological stations located in Zhejiang Province, east China using the Mann-Kendall test. Then the precipitation trends in the plum season （from May to July） and typhoon season （from August to October） are studied separately. The results show that the precipitation trend varies from east to west. There is a positive trend in the east and a negative one in the west. The largest part of Zhejiang Province shows a positive trend in heavy precipitation. Although the upward trend of extreme precipitation is not prevailing, the range of increase in specific areas is apparent. Precipitation intensity exhibits an upward trend in most areas. Precipitation intensity in both plum and typhoon seasons has increased too, especially for the coastal stations.（2） three different rainfall-runoff models, namely GR4J, HBV and Xinanjiang, are applied to Jinhua River basin, eastern China. The Generalised Likelihood Uncertainty Estimation （GLUE） approach is used for estimating the uncertainty of the three models due to parameter values. Uncertainty in simulated extreme flows is evaluated by means of the annual maximum discharge （MHQ） and mean annual7-day minimum discharge （MAM7）. The results show that the uncertainty in high flows increases with the discharge magnitude. The parameter uncertainty in high flows is the largest in the HBV model and smallest in the Xinanjiang model. Low flows are mostly underestimated by all models with optimum parameter sets. The uncertainty originating from parameters is larger than uncertainty due to model structure.（3） uncertainties in extreme high flows originating from greenhouse gas emission scenarios, hydrological model structures and their parameters for the Jinhua River basin, China are assessed. The baseline （1961-1990） climate and future （2011-2040） climate for A1B, A2and B2scenarios were downscaled from the General Circulation Model （GCM） using the PRECIS Regional Climate Model with a spatial resolution of50km×50km. Bias correction methods are applied to the temperature and precipitation of PRECIS. The bias corrected precipitation and temperature are used as input for three hydrological models （GR4J, HBV and Xinanjiang） to simulate extreme high flows. It is found that bias correction before the use of the RCM Data for assessment study could improve the results, which are of a higher degree of consistency with the observation. Under scenario A1B, A2and B2the extreme high flows decreased in the future. The order of the uncertainty range from high to low are hydro togical model parameters, hydro logical model structure and the emission scenarios.（4） the representative concentration projection （RCP）2.6, RCP4.5, RCP6.0and RCP8.5in the5th IPCC report are used as the future emission scenarios. Meanwhile the impacts of the GCMs on the extreme flows are considered. The weather generator LARS-WG is used for the output of three GCMs namely BCC_CSM1₁, HadGEM2-ES and GISS_E2_R to generate a longer series. Hydro logical models GR4J, HBV and Xinanjiang are applied to simulate the river discharges in the past and the future. The results show that there are increase in the temperature, the range of the increase is larger for the daily lowest temperature than that for the daily highest temperature. Among the three GCMs, the uncertainty of the extreme flows from the emission scenarios is largest for the HADGEM2-ES, followed by the GISS_E2_Rand the smallest is for the BCC_CSM1₁. The extreme high flows would increase under the RCP4.5and RCP6.0by HADGEM2-ES. However, it would decrease under the RCP6.0and RCP8.5. The extreme high flows would decrease as well under RCP4.5by GISS_E2_R. the uncertainty of the extreme flows from the GCMs is the largest for the RCP4.5. the extent of the impact of uncertainty resources on the extreme flows from high to low are:hydrological parameters, GCMs, RCPs and hydrological model structures.更多还原