EGI OpenIR  > 研究系统  > 空间对地观测与系统模拟研究室
气候波动下开都河流域的水文响应模拟及其不确定性分析
田霖
学位类型硕士
导师刘铁 ; 包安明
2017-05-01
学位授予单位中国科学院大学
学位授予地点新疆乌鲁木齐
学位专业理学硕士
关键词水文模型 再分析数据 气候波动 不确定性 hydrologic model reanalysis data climate change uncertainty
摘要The socio-economic and ecological problems caused by climate change haveattracted worldwide attention. The arid region of central Asia is a typical ecologicallyfragile area, and its water resources are sensitive to climate change response. In thispaper, the Kaidu river basin in the typical mountainous region of Central Asia wastaken as the study area. Firstly, the applicability of CFSR reanalysis data in themodeling of arid mountainous hydrologic process was discussed. LOCI, Gamma andgamLOCI methods were used to correct for the bias of CFSR precipitation data, andthe daily runoff simulation results of the three methods were compared. Secondly,driven by the GCMs data and the statistical downscaling method, the hydrologicalmodel of the Kaidu river basin was used to simulate the hydrological response in theRCP4.5 and RCP8.5 scenarios. Finally, four kinds of GCMs were used as the rainfallinput data, and three kinds of hydrological models to simulate the runoff. Theuncertainty among the input data, model structure and model parameters wasanalyzed in the whole technical route. The main conclusions are as follows:(1) Compared with the observation, the overestimation for light rainfall andunderestimation for heavy rainfall of CFSR dataset were significant in Kaikong riverbasin. CFSR dataset is likely to underestimate heavy rainfall in high-altitudemountainous areas and to overestimate light rainfall in plains. There are too manysmall precipitation events for CFSR data both in high-altitude mountainous areas andplain areas. The deviation and correlation between CFSR precipitation data and theobservation are related to elevation and season. In general, the correlation of highaltitude mountainous area is better than that in plain area. The correlation in springand autumn is better than that in summer and winter. The deviation of high-altitudemountainous area and plain area is higher than that of middle-altitude area.Based on the results of three bias correction methods, the LOCI method cancorrect the wet day and wet rainfall intensity of the precipitation data, but it willdistort the extreme value of the precipitation change. The Gamma method can keepthe extreme value of the climate change signal, but cannot correct the wet dayfrequency and wet day precipitation intensity. The gamLOCI method synthesizes theadvantages of the LOCI and Gamma method, not only the wet-day frequency andprecipitation intensity of CFSR data could be corrected, but the extreme waspreserved. Three correction methods could equally and effectively improve thereliability of CFSR reanalysis data in plain area, while the correction effect withgamLOCI method was the best in mountainous region. Using the raw CFSR datasetas the precipitation input data of hydrological model, the simulation result of runoff istoo large, and the relative deviation is greater than 40%, and the spring flood peak getin advance and increased. Therefore, there exists obvious bias in reanalysisprecipitation datasets in Central Asia, the appropriate bias correction method has to beconsidered before applying these datasets in hydrological model of the aridmountainous area.The three bias correction methods, LOCI, Gamma and gamLOCI, have differentcharacteristics, but the results of runoff simulation show little difference. All of them can give a satisfied performance.(2)under background of global warming, the annual precipitation andtemperature in arid mountainous areas showed increasing trend, in agreement with thechange from "warm and dry" to "warm and humid‖. But the increase with months ofthe year was not consistent. Precipitation increases in both winter and summer, whilethat is not significant in spring and autumn. Temperatures is increasing all the year,but it increases faster in winter and summer, and slower in spring and autumn. Ingeneral, the volatility and uncertainty of precipitation in arid mountainous areas willincrease and precipitation will be more concentrated in the summer in the RCP8.5scenario. The distribution of precipitation in each month will be even more averagedin the RCP4.5 scenario.Affected by climate warming, the surface runoff, groundwater recharge andsnowfall will be reduced in the dry mountainous areas of Central Asia in future, whileprecipitation, actual evapotranspiration and potential evapotranspiration will increasesignificantly. By 2035,the runoff will increase in spring and autumn, and it willdecrease in winter and summer. The runoff during all seasons of the year will increaseby 2070. The water resources in Kaidu river basin is getting richer from 2016 to2023.The amount of annual runoff of Kaidu river basin is about 2 to 3.5 billion cubicmeters in the period from 2051 to 2070. The amount of annual runoff in RCP8.5scenario ranges between 1 and 3.5 billion cubic meters. And the annual runoff of2051-2070 in RCP4.5 scenario is around 1.5 billion cubic meters. However, thefluctuation of the water resources amount in RCP4.5 scenario is more smaller thanthat in RCP8.0 scenario.The amount of snowmelt in the high mountainous areas of Kaidu river basin,which is vulnerable to climate change, is obviously increasing. While the amount ofsnowmelt in plains and low-altitude valleys of mountains is decreasing. This isbecause the temperature increases in the basin, and the snow line rises as the result.(3) The uncertainty of the bias correction methods in the process of hydrologicalresponse simulation to climate change is the smallest, and the input data, the modelstructure and the uncertainty of the model parameters mater indeed. The uncertaintyof the input data is relatively larger in wet seasons of the year than that in dry seasons;the uncertainty of model structure is large both in dry seasons and wet seasons. Butthe uncertainty of model structure in the flood formation and uplift stage is larger thanthat in flood regression stage.
其他摘要气候变暖引起的社会经济和生态环境问题已受到全球广泛关注。亚洲中部干旱区是典型的生态脆弱区,其水资源系统脆弱,对气候变化响应敏感。本文以天山南坡典型山区流域开都河为研究区,首先探讨了 CFSR 再分析数据在缺资料的干旱山区水文模型建模中的适用性,采用 LOCI、Gamma 分布映射和gamLOCI 三种方法对 CFSR 再分析降水数据进行偏差校正,比较了三种方法应用于 SWAT 模型日径流模拟的结果差异。其次, 采用经过统计降尺度处理的 GCMs数据驱动率定好的水文模型,模拟了在 RCP4.5 和 RCP8.5 情景下开都河流域到本世纪中叶的水文响应和水资源变化趋势。最后,采用了四种 GCMs 模式作为降雨输入数据和三种开都河流域的水文模型模拟径流,分析了整个技术路线中输入数据、模型结构和模型参数的不确定性。主要结论如下:(1)在干旱区山区,CFSR 降水数据相比于站点观测数据主要表现出低估强降水、高估弱降水的特征。对于高海拔山区的强降水月份,CFSR 数据表现出低估的特征;对于平原地区和其他月份整体表现出高估特征,特别是在低海拔的山前平原区,高估幅度最大。无论是在高海拔山区还是平原区,CFSR 数据都存在过多的小降水事件。CFSR 降水数据与站点观测数据的偏差、相关性与高程、季节相关。总体而言,高海拔山区相关性优于平原地区,春秋两季相关性优于夏冬两季,高海拔山区和平原地区的偏差大于中海拔地区,夏秋两季的偏差小于春冬两季。综合分析三种偏差校正方法的校正结果,LOCI 法能校正降水数据的湿日频率和湿日降水强度,但会扭曲降水变化中的极值;Gamma 分布映射法能保留降水信号中的极值,但不能校正湿日频率和湿日降水强度。gamLOCI 法综合了 LOCI 法和 Gamma 分布映射法的优点,对于高海拔山区的 CFSR 降水数据,采用 gamLOCI 法的校正效果最优,而在平原地区三种方法没有显著差别。采用原始 CFSR 降水数据驱动水文模型模拟的径流量偏大,相对偏差大于 40%,且春季洪峰提前, 洪峰峰值提高。所以 CFSR 降水数据在西北干旱半干旱区偏差大、精度低,并不能直接作为干旱山区水文模型的输入数据,采用偏差校正方法进行订正是必要的。LOCI、Gamma 分布映射和 gamLOCI 法校正后的降水数据驱动水文模型得到的径流模拟效果都有很大提升,三种校正方法各有特点但径流模拟结果差别不大,都能获得较为满意的校正效果。(2)在全球气候变暖背景下干旱山区年降水量和温度整体呈增加趋势,即从“暖干”向“暖湿”逐渐转型,但年内各月份增加幅度却并不一致。降水在冬季和夏季增加幅度较大,在春季和秋季增加幅度并不明显。全年各个月份温度都在升高,但是气温在冬季和夏季增温速率较快,在春季和秋季增温速率较慢。总体而言,在 RCP8.5 情景下干旱山区的降水量在年内的差异性和波动性都将增加,降水会更集中在夏季,而在 RCP4.5 情景下年内各季节的降水分配趋于平均。中亚干旱山区未来地表径流补给、地下水补给、降雪将减少,而降水、实际蒸散量、潜在蒸散量将大幅增加。到 2035 年,春季和秋季径流将增加,冬季和夏季径流会减少。到 2070 年,全年各季节段的径流都呈现出减少的趋势。2016-2023 年开都河流域水资源量较为丰富,年径流量在 20~35 亿立方米。但是到 2051-2070 年,RCP8.5 情景下流域年径流量在 10~35 亿立方米,且水资源量的年际波动较大,而 RCP4.5 情景下 2051-2070 年的年平均径流量为 10~15亿立方米,年际变化较平稳。未来开都河流域内周边高海拔山区的融雪量增速明显,而在低海拔的山前平原区和峡谷区融雪量呈减少的趋势。这可能是由于流域内温度升高,雪线上升造成的。(3)气候波动的水文响应模拟过程中偏差校正方法的不确定性最小,输入数据、模型结构和模型参数的不确定性都是不容忽视的。输入数据的不确定性在年内的丰水期比较大,在枯水期比较小,模型结构的不确定性在丰水期和枯水期都比较大。但是模型结构的不确定性在径流形成和增涨阶段的不确定性要大于径流消退阶段。
学科领域地图学与地理信息系统
语种中文
文献类型学位论文
条目标识符http://ir.xjlas.org/handle/365004/14898
专题研究系统_空间对地观测与系统模拟研究室
作者单位中国科学院新疆生态与地理研究所
第一作者单位中国科学院新疆生态与地理研究所
推荐引用方式
GB/T 7714
田霖. 气候波动下开都河流域的水文响应模拟及其不确定性分析[D]. 新疆乌鲁木齐. 中国科学院大学,2017.
条目包含的文件
条目无相关文件。
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[田霖]的文章
百度学术
百度学术中相似的文章
[田霖]的文章
必应学术
必应学术中相似的文章
[田霖]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。