【作者】 黄晓敏；

【导师】 王浩； 蒋云钟； 雷晓辉；

【作者基本信息】 东华大学 ， 环境科学与工程， 2014， 博士

【摘要】 松花江流域水土资源相对较丰富,是我国重要的粮食主产区之一。流域属寒温带大陆季风气候区,降雨量年内分布极不均匀,汛期降雨量约占年降雨量70%以上。该流域洪水峰高量大、历时长,保障防洪安全是该流域的首要任务。近年来国民经济发展、人口增多、生活水平提高,对防洪安全提出了更高要求。而随着经济社会发展,松花江流域的土地利用等下垫面覆被及河道状况发生了一定变化,同时流域内修建了大量水利工程,对洪水特性产生了重要影响。本文以第二松花江丰满以上流域作为研究区域,以包括土地利用变化和水利工程兴建在内的人类活动对洪水过程的影响研究为基础,以历史暴雨在现状下垫面下产生的洪水过程变化特性研究为重点,主要内容和成果如下：(1)流域径流变化规律及其影响因素分析研究了第二松花江丰满以上流域内径流量的年内分配、汛期径流量的年代际和年际变化特点。其中,年际变化规律中总结了汛期径流量和最大洪峰的多年变化趋势、突变变化以及丰枯周期性变换规律。分析了影响径流变化的自然因素和人类活动干扰。统计分析了气候因素、厄尔尼诺和拉尼娜现象以及太阳黑子等自然因素对径流的影响。统计了土地利用和水利工程等人类活动干扰因素的发展变化情况,为开展土地利用变化和水利工程兴建对径流影响分析提供了数据支撑。(2)考虑土地利用空间异质性的分布式水文模型研发对分布式水文模型EasyDHM进行了改进,提出了按土地利用类型进行空间离散的方法。将原EasyDHM内等高带(子流域内计算单元)单元,按土地利用类型细化为若干土地利用单元。每个等高带内有至多六个土地利用单元,分别为耕地、林地、草地、水域、居民地及未利用土地单元。每个土地利用单元的下垫面覆被特征采用独立的土地利用参数表征。土地利用参数具有明确的物理意义,在下垫面覆被方式改变后,可仅改变土地利用单元面积、直接移用参数值的方式。通过该模型分析了流域土地利用变化对产流的影响和各土地利用类型的产流贡献率。通过1980年和2000年下垫面覆被下的产流分析,耕地、林地、草地、水域、居民地和未利用土地的产流变化率分别为9.18%,0.64%,-1.63%,-5.03%,-3.08%和24.59%。各类土地利用类型的产流单位面积贡献率,由大到小依次为水域、居民地、未利用土地、耕地、林地、草地。基本呈现植被较好的土地利用类型比植被较差的土地利用类型产流少的趋势。(3)分布式水文模型水利工程群调蓄模块研发建立了能够描述不同规模缺资料水库群间主要水力联系的模块。该水利工程调蓄模块中,能够准确定位的大中型水库单独划分水库分区,只能大概定位的小型水库按其所在子流域打包为虚拟水库。不同规模水库间自小型至大型水库按上游到下游位置呈串联关系；同规模小型水库间假定调蓄能力相当,呈并联关系。无实测实时运行资料的水库均按调度图的方式进行蓄泄模拟,相同规模水库采用相同的通用调度图。包括调度图及虚拟水库位置参数在内的水利工程参数采用优化算法率定得到。采用嵌有水利工程群调蓄模块的分布式水文模型,分析水利工程对洪水的影响和进行现状下垫面下的历史暴雨洪水的重现。结果表明,考虑水利工程影响后的模拟精度有所提高：水库在大型洪水和小洪水都起到了很明显的削峰拦洪的作用,并且水库库容变化率对小洪水更加敏感。(4)参数敏感性信息引导搜索方向的优化技术研究受水文模型参数的敏感度不同启发,提出了一种以参数敏感性信息引导搜索方向的动态维度搜索算法,MDDS算法,即为解决高维参数优化问题而提出的DDS算法改进版本。该算法中敏感性分析独立且先于DDS算法运行。搜索策略是设置参数变异概率与参数敏度度呈正相关,模型输出响应较明显的敏感参数相比较敏感度弱的参数具有更多的变异机会,同时敏感度弱的参数仍有一定的变异机会；这种策略结合了敏感参数的变异倾向性和动态选择参数的随机性,在较低计算量下具有较高的搜索效率。测试结果表明,MDDS算法更适合具有高参数维度空间的分布式水文模型。在模型评价次数有限时,MDDS算法相对于DDS算法有更高的搜索效率和更好的稳定性。在与SCE算法和仅识别15个敏感参数DDS算法的性能比较中,MDDS算法均表现除了较高的搜索效率。(5)历史暴雨洪水重现系统研发系统采用Microsoft Visual Studio. NET为集成开发系统环境,以图形用户界面友好的C#语言和计算效率较高的Fortran语言分别作为系统集成和水文模型组件的开发语言。以Fortran语言编写动态链接库(DLL),在DLL中提供计算函数接口：在C#中调用该DLL中的计算函数,实现系统的运行计算。该系统采用混合编程技术、以Map Window和SQL Server作为可视化G1S软件平台和数据库管理系统,建立了一个集GIS技术、计算机技术、数据库技术和考虑人类活动的分布式水文模型等多项技术为一体的综合历史暴雨洪水重现系统。该系统应具有水文分析、水雨工情查询与分析、历史暴雨洪水重现功能及结果的展示与分析的功能。(6)流域内历史暴雨洪水重现规律研究将集成了考虑人类活动影响的洪水重现模型及高维参数识别技术的应用系统,应用到第二松花江丰满以上流域,重现历史洪水在现状下垫面下的演进过程,总结人类活动对洪水过程的影响规律。根据流域出口断面历史大洪水在现状下垫面条件下的重现结果,若历史暴雨在下垫面变化后的现状条件下发生,洪水的洪峰和洪量均有减小的现象；20世纪80年代前的洪水重现时,洪现时间有延迟现象。更多还原

【Abstract】 Songhua river basin, which is rich in water and soil resources, is one of the major grain-producing regions in China. This basin is located in cold temperate continental monsoon climatic zone. The precipitation concentrates in summer, holding70percents of the annual rainfall. The floods in this basin have the features of high peak, large volume and long lasting time, which make the flood control very important for the area safety. With the rapid development of national economy, the increasing population and improvement of the people’s living standard, higher requirement are proposed for flood control safety. With the economy development, the underling surface of the basin and the river condition has changed, and a large number of water conservancy project have been built, which both have important influence on the flood characteristic. The upstream part to Fengman reservoir in Second Songhua river catchment was selected in this paper as the study area to analyze the impacts of such human activities as the changes of land use/land cover and the construcion of water conservancy project on flood process and simulate the historical flood process on current underlying surface. The study is organized as the followings:(1) Temporal variation of observed runoff in flood season and its influencing factorsAnnual distribution, inter-decadal and inter-annual variations of the flood process were analyzed in this section. Among those, inter-annual variation were analyzed from three aspects, i.e. long-term variety of flood season runoff and maximum peak, abrupt change and periodic variations of high flow and low flow. The effects of natural factor and interferences of human activities on runoff were also analyzed. Runoff responses to natural factors, i.e. climatic factor, El Nino and LaNina, solar spot, were discussed firstly. And the changes of land cover and the development of water conservancy projects over the past dozens of years were described, which could provide data support for the following study about the impact of human activities on flood process. (2) The distributed hydrological model (DHM) considering the spatial variability of land useIn this section, the distributed hydrological model EasyDHM was modified, and a space discrete method to divide the study area according to the land use variation was proposed. Equal elevation bands in original EasyDHM were further divided into several land use units according to their land use types. There were at most six land use units in each equal elevation band, i.e. cultivated land, forestland, grassland, paddy field, residential area and unused land units. The underlying surface features of each land use type were characterized by a set of land use related parameters. As those land use related parameters have definite physical meaning, the parameter values could be directly adopted by only adjusting area value of corresponding land use unit if the land use patterns changed. The modified EasyDHM was employed in this study to analyze the impacts of land use changes on runoff generation and the runoff contribution per unit area for each land-use type. The study area showed9.18%,0.64%,-1.63%,-5.03%,-3.08%, and24.59%variations in runoff with changes in the areas of cultivated land, forest land, grassland, water surface, residential area, and unused land, respectively, during the20-year study period. The magnitude of the runoff per unit area for each land-use type was in the order water surface> residential area> unused land> cultivated land> woodland> grassland. Thus, well-vegetated areas (i.e., grassland and woodland areas) were likely to generate less runoff than areas with less vegetation cover (i.e., unused land and cultivated land) under the same rainfall conditions.(3) Development of cascading reservoirs operation module into the DHMA module for describing the most important aspects of reservoir dynamics with scare data availability was developed. Large and medium-sized reservoirs, which can be located accurately, were used to mark off their reservoir divisions. Small reservoirs, which cannot be located exactly, were packed as a virtual reservoir according to sub-catchments. In each sub-catchment, smaller reservoirs were located at the upstream of larger reservoirs. The method assumed that small reservoirs in each sub-catchment have similar storages capacity and they are shunt connected. The operation of all reservoirs without observation data was control by operation rule. The operation rule curces and other reservoir parameters could be identified by using optimization algorithms. The distributed hydrological model with reservoirs operation module was applied to analyze the influence of reservoirs on flood processes and simulate the historical flood occurrence on current underlying surface. The simulation results were improved after considering reservoirs’influences. Reservoirs played prominent parts by clipping peak, alternating peak and storing flood water in flood-control of rivers. Small floods were more sensitive to change of storage capacity than heavy floods. Historical floods before1990were simulated on current underlying surface, and the simulated flood peak and volume were less than that on historical underlying surface. In addition, occurrence of the flood peak delayed.(4) An optimization technology with parameter sensitivity guiding search directionMotivated by different sensitivity of hydrological model parameters, a modified dynamically dimensioned search (MDDS) algorithm with parameter sensitivity guiding search direction was proposed. The MDDS algorithm was developed in the context of solving the problem of high-dimensional parameter optimization. The sensitivity analysis preceded and was separated from DDS in the MDDS algorithm. In the modified MDDS algorithms, parameter mutation probability showed positive correlation with parameter’s sensitivity. The sensitive parameters that produced more obvious responses in the output model were given greater opportunities to mutate; however, the MDDS algorithm also allowed the insensitive parameters to participate in mutation. The MDDS strategy combined a preference for sensitive parameters with a use of random perturbations, as this yielded a higher probability of finding an improved solution. This evolution strategy enabled the high search efficiency of MDDS algorithm under a low amount of calculation. Test results showed that the MDDS algorithm is better suited to a distributed hydrological model that includes many parameters. When the computational budget is limited, the superiority of the MDDS consists in the greater average best function value and the greater stability of the algorithm; and the MDDS algorithm is more computationally efficient and robust than SCE and DDS identifying only the15most sensitive parameters in the context of distributed hydrological model calibration.(5) Application system for historical flood simulation on current underlying surfaceThe system was developed on the basis of the Microsoft Visual Studio.NET environment. C Sharp language, which was characterized by a friendly graphical interface, was used as system integration language. And Fortran language, which was characterized by the powerful ability of scientific computing, was used as development language of hydrological model component. The dynamic-link library (DLL) in which function interface was supported was coded in Fortran language, and was then called by C#language to realize the operation of system. Mixed programming technique, MapWindow platform and SQL Server were used in this system. So this system was a comprehensive simulation system based on a distributed hydrological model considering human activity disturbance, GIS platform, computer technology and database management system. This system integrates hydrologic analysis, water and rainfall information and engineering conditions query and analysis, historical flood simulation on current underlying surface, and simulation result display functions.(6) Study on historical flood simulation on current underlying surface The application system integrated the distributed hydrological model considering influence of human activities and high dimension parameter identification technology were used to simulate the historical flood on current underlying surface and summarize the influence law of human activities on floods. According to the historical flood simulation result in the outlet section of the basin, the simulation flood peak and volume show a decreasing trend, if they reoccurred on current underlying surface today. The simulation peak time are delayed for floods before1980.更多还原