基于敏感性分析的SWAT水文参数优化方法比较研究

发布时间：2018-09-19 18:22

【摘要】：SWAT （Soil and Water Assessment Tool）模型凭借其强大的水循环模拟能力在全世界范围内的应用已越来越广泛。它可以基于GIS平台建立的地形空间数据对流域汇水区进行自动划分，并形成水源地区域边界；结合模型输入的各种物理分布式文件，通过水文响应单元模拟研究区水文过程。这种分布式水文模型不仅对研究区实测资料依赖小，而且操作性强，对汇水区域划分的效率和质量也能够得以保证，尤其能够完成环境变化情景下的水文过程分析。长春作为典型的北方缺水性城市，新立城水库这个离长春最近的水源地的供水安全保障工作就显得尤为重要。本文建立了研究区月尺度的径流模拟SWAT模型，并检验了径流过程模拟的适用性和可靠性，并为新立城水库水资源保护提供了科学依据。本文以新立城水库汇水区为研究区，研究模型的结构组成和运行机理，分析了模型中与径流模拟相关的水文循环陆地产流阶段、汇流阶段和河道汇流演算阶段等3个阶段。通过收集研究区DEM高程图、土地利用/植物覆被图、土壤类型图以及逐日水文气象资料等信息，应用GIS技术进行空间数据处理并使用Matlab、pcpSWAT、dew02等软件创建了针对于研究区的土壤和气象数据库。并在此基础上运行SWAT模型，提取流域水系并确定了流域的汇流边界。通过对面积阈值6600ha的限制，将研究区划分为16个子流域。通过加载土地利用/植物覆被数据、土壤数据和气象数据，并设定土地利用和土壤类型的阈值分别为5%和10%，将研究区划分为93个HRUs。成功建立了新立城水库汇水区的SWAT基础数据模型。利用已建SWAT模型对研究区径流量进行以月为时间步长的模拟，并利用两种不同方法分别对模型进行敏感性分析以进行参数率定。采用伊通水文站2004—2008年的实测径流数据校正模型参数，2009-2010年实测径流数据对模型进行验证，然后分别对两种方法运行结果进行模型适用性指标评价。结果表明，利用SWAT自动校正模块在校核期和验证期的月径流量的判定系数R2分别为0.86和0.99；模型效率系数Ens分别为0.80和0.97；相对误差Re分别为20.0％和19.3％；利用SUFI-2算法在校核期和验证期的月径流量的判定系数R2分别为0.92和0.98；模型效率系数Ens分别为0.91和0.96；相对误差Re分别为6.0％和10.3％。两种方法对模型适用性评价的各项指标均满足研究要求，表明建立的SWAT模型可以应用于该流域。对比两种方法，无论从输出结果质量还是操作便捷程度及运行效率方面来看，，SUFI-2算法明显好于模型自带模块。
[Abstract]:With its powerful water cycle simulation ability, SWAT (Soil and Water Assessment Tool) model has been used more and more widely all over the world. It can automatically divide watershed catchment area based on topographic spatial data established by GIS platform and form regional boundary of water source area. Combined with various physical distributed files input by model hydrological response unit is used to simulate hydrological process in the study area. The distributed hydrological model is not only less dependent on the measured data, but also more operable, and the efficiency and quality of the division of the catchment area can be guaranteed, especially the hydrological process analysis under the environmental change scenario. Changchun is a typical water shortage city in the north of China. It is very important to ensure the water supply safety of Xinlicheng Reservoir, which is the nearest water source to Changchun. In this paper, a monthly scale SWAT model of runoff simulation in the study area is established, and the applicability and reliability of the simulation of runoff process are tested, which provides a scientific basis for the protection of water resources of Xinliteng Reservoir. Taking the catchment area of Xinlicheng Reservoir as the study area, this paper studies the structure composition and operation mechanism of the model, and analyzes three stages of hydrological cycle land runoff generation, confluence and channel flow calculation related to runoff simulation. By collecting information such as DEM elevation map, land use / plant cover map, soil type map and daily hydrometeorological data, The spatial data processing using GIS technology and the use of Matlab,pcpSWAT,dew02 and other software to create a soil and meteorological database for the study area. On this basis, the SWAT model is used to extract the drainage system and determine the confluence boundary of the watershed. By limiting the area threshold 6600ha, the study area is divided into 16 subbasins. By loading land use / plant cover data, soil data and meteorological data, and setting a threshold of 5% and 10% for land use and soil type, the study area was divided into 93 HRUs.. The SWAT basic data model in the catchment area of Xinlicheng Reservoir has been established successfully. The SWAT model is used to simulate the runoff in the study area with a monthly time step, and two different methods are used to analyze the sensitivity of the model to determine the parameters. The model parameters of Yitong Hydrologic Station from 2004-2008 to 2009-2010 were used to verify the model, and then the applicability of the two methods was evaluated. The results show that the determination coefficient R2 of monthly runoff is 0.86 and 0.99, the model efficiency coefficient Ens is 0.80 and 0.97, the relative error Re is 20.0% and 19.3%, respectively. Using SUFI-2 algorithm, the determination coefficient R2 of monthly runoff is 0.92 and 0.98, the Ens of model efficiency is 0.91 and 0.96, and the relative error Re is 6.0% and 10.3%, respectively. The two methods can meet the requirements of the study on the applicability of the model, which indicates that the established SWAT model can be applied to the basin. Compared with the two methods, the SUFI-2 algorithm is obviously better than the model module in terms of output quality, operation convenience and operation efficiency.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TV12

【参考文献】