The Research Of Swat Model Parameter Optimization And Runoff Response Of Land Use Change

The loess plateau is typically the most serious area in soil erosion and ecological environment in our country. Since the policy of returning farmland to forest in 1999, the ecological environment on the loess plateau area has been obviously improved. However, the increase of the vegetation cover to the area of runoff process and the influence of the distribution pattern is not clearly. So the study of the effect of land use change on runoff has important guiding significance.This article choose Hulu river watershed, the largest tributary of the north of Luohe River, as the study area, and build the distributed hydrological model to simulate the monthly runoff. The comparision of automatic optimization algorithms have been analyzed through the SWAT-CUP program and Matlab platform. The prediction of future land use in watershed has been done through CA-Markov model. And we analyzed the effect of land use change on annual average monthly runoff under the time-lapse scenarios and setting scenarios.The main results were as follows:(1) Pretreated the data such as DEM, soil, land use and hydrometeorological data through the GIS platform, the weather generator, SPAW software and the SWAT model was put into the SWAT model to caculate the runoff. The watershed was divided into 37 subbasins and 876 hydrological units. The preheating time was set 1 year and run SWAT model, so the distributed hydrological model were eatablished in the year of 1980~1990 and 2006~2012.(2) The model parameter sensitivity analysis and parameter optimization work were carried out by SWAT-CUP program. The model parameter sensitivity analysis combined one-at-time and the global methods, and selected 15 more sensitivity parameters. Then, we chose SUFI-2, GLUE, PSO algorithm to conduct the parameters optimization of the watershed according to the advantages and disadvantages of these methods. It can be concluded that PSO algorithm has higher simulation accuracy, the Nash-Sutcliffe efficiency coefficients(NS) in calibration and validation period were 0.75 and 0.61 respectively, and the determinatical coefficients(R2) were 0.82 and 0.80 respectively. Therefore, the PSO algorithm is more suitable for parameter optimization in Hulu River watershed compared to the other two algorithms in SWAT-CUP.(3) This study proposed a new method of parameter optimization parallel GA-PSO algorithm. The GA and PSO algorithm coupled the SWAT model in Matlab platform, and the parallel computing process was added in to increase the global optimized process, and greatly improved the solving speed. The Nash-Sutcliffe efficiency coefficients(NS) in calibration and validation period were 0.76 and 0.66 respectively, and the determinatical coefficients(R2) were 0.81 and 0.77 respectively. The simulation results were higher than the PSO algorithm, indicating that the parallel GA-PSO algorithm was more suitable to the parameters optimization, especially for the area of returning farmland to forest. However, the simulation value was lower than the observed value, showing that the algorithm was more suitable for the area of smaller peakflow.(4) The CA-Markov model was used to predict the future land use change according to the existing land use data of 2000, 2005, and 2010. The land use situation of 2010 was predicted based on the transfer matrix of 2000~2005 and the transformation rule defined by the suitability atlas. The Kappa coefficient is 0.87, and each land use type of pixel error was within 6% compared to the existing land use data of 2010, indicating that CA-Markov model can be used to predict future land use situation in the watershed. Later, the land use situation of 2015 and 2020 was predicted by CA-Markov model.(5) Using the existing land use data of 2000, 2005 and 2010, and the future land use data of 2015, 2020 to analysis annual average monthly runoff response to land use change. At the same time, three kinds of scenarios were set based on the national policy of returning farmland to forest and the data of land use in 2010 to analysis the impact of land use change on runoff. The annual average monthly runoff had a decreasing trend in the two situations. But runoff change is not obviously in the time-lapse scenario, so runoff change rate is also small. It was decreased in the normal season and dry season(October to May in the following year) in addition to March in 2005~2010, and the change rates were different in the wet season(June to September), it has decreased in some scenarios and increased the other scenarios. While the runoff had no significant changes in normal and dry season and had a ladder down trend in the wet season in the setting scenarios. It showed that vegetation coverage increases could reduce the annual average monthly runoff.