Improvement of SWAT2000 Modelling to Assess the Impact of Dams and Sluices on Streamflow in the Huai River Basin of China
Total Page:16
File Type:pdf, Size:1020Kb
HYDROLOGICAL PROCESSES Hydrol. Process. 24, 1455–1471 (2010) Published online 17 February 2010 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/hyp.7606 Improvement of SWAT2000 modelling to assess the impact of dams and sluices on streamflow in the Huai River basin of China Gangsheng Wang1,2*andJunXia1 1 Key Laboratory of Water Cycle and Related Land Surface Processes, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China 2 Department of Biological Systems Engineering, Washington State University, Pullman, WA 99164-6120, USA Abstract: Hydrological simulation and assessment in a dam–sluice regulated river basin are a complex and challenging issue. In this article, an improved SWAT2000 modelling system was developed that incorporated the Shuffled complex evolution (SCE- UA) optimization algorithm and the multi-site and multi-objective calibration strategy. The implication of multi-objective is different for different types of outlets, i.e. streamflow for an ordinary outlet, inflow for a sluice, and water storage for a reservoir. Model parameters were redefined to improve model simulations. The surface runoff lag time (SURLAG) was extended as a spatially distributed parameter, and a correction coefficient was introduced to modify the saturated hydraulic conductivity. The modelling system was then applied to the Huai River basin of China under various climatic conditions, including a very dry year (1999), a dry year (1981), an average year (1971), and wet year (1991). In all, 26 dams and 35 sluices were considered, among which about 20 dams/sluices were used for model calibration. The impact assessment primarily focused on the very dry year (1999). The results indicated that the released water from large reservoirs was blocked in the river channels by sluices located downstream. In the very dry year, the dam–sluice operations could result in an increase of the runoff volume during the non-flood season and a decrease in runoff during the flood season, but the changing magnitude during the non-flood season was much greater. An important conclusion of this case study is that the sluices in the Sha-Yin branch located in the north region and the dams in the southern mountainous region above the Wangjiaba Hydrological Station have played the most significant role in regulating the streamflow of the entire river basin. The methods addressed in this article can simulate hydrological regime in the river basins regulated by dams and sluices under different climatic conditions at the whole-watershed scale. Copyright 2010 John Wiley & Sons, Ltd. KEY WORDS dams and sluices; impact assessment; modelling; SCE-UA; SWAT2000 Received 30 March 2009; Accepted 21 December 2009 INTRODUCTION water system characterized by numerous river reaches Dams and sluices are important management means in and reservoirs. A river system controlled by many dams the exploitation and utilization of water resources (Gross and sluices are usually designed and planned to meet and Moglen, 2007; Lopez-Moreno et al., 2009). Today, with the need in flood control and water withdrawal. there are more than 45 000 large dams (more than 15 m These dams/sluices might be planned and constructed high) in the world, with more than half of these are by stages. Moreover, the planning might be conducted in China alone (ICOLD, 1998). Both dams and sluices from the point of view of a local region, and not the serve the objectives of regulating water artificially. There entire river basin. Hydrological modelling at the whole- are also some differences between dams and sluices that watershed scale is necessary and useful for the design- are addressed in this article. A dam and its associated ers/planers/administrators to evaluate the influence of reservoirs are usually constructed in mountainous regions local dams/sluices on water quantity/quality in other parts to form a huge water body with a large water surface, and or the entire watershed. Hence, a dam/sluice-controlled its primary function is to generate electric power besides river basin complicates steps taken to implement hydro- storage and flood control. A sluice, on the other hand, is logical modelling of the system. Regarding this kind of usually located in a river channel within an environment modelling, two important issues are needed to be con- characterized by relatively gentle terrain, and is primarily sidered carefully. One is the model development that used for flood control and water supply. considers the roles of dams and sluices, and the other Due to the influences of dams and sluices, the water is the model calibration and water resources applica- in a river is not a natural water system, but a complex tion. It is obvious that a lumped hydrological model is incapable of expressing the spatially distributed features of dams and sluices. Therefore, a suitable distributed * Correspondence to: Gangsheng Wang, Department of Biological Sys- tems Engineering, Washington State University, Pullman, WA 99164- model coupling dams and sluices represents a better 6120, USA. E-mail: gangsheng [email protected] choice. Copyright 2010 John Wiley & Sons, Ltd. 1456 G. WANG AND J. XIA The Soil and Water Assessment Tool (SWAT) is a rel- change the value of a given parameter by a fixed percent- atively ideal model to consider for basin scale water age of the initial value (Lenhart et al., 2002; White and resources applications, because of its reservoir mod- Chaubey, 2005), and the other was to vary it by a fixed ule (Neitsch et al., 2002a,c). SWAT has been widely percentage of the valid parameter range (Lenhart et al., used in a variety of investigations, such as hydrological 2002). In both sensitivity analysis approaches, the great- simulations and assessment, non-point pollution, climate est importance was attributed to soil parameters. In other change impact, parameter sensitivity, model calibration studies (Kannan et al., 2007; Wu and Johnston, 2007), and uncertainty analysis, model comparisons, and inter- sensitivity analyses were carried out by using those model faces of SWAT with other models (Borah and Bera, 2004; parameters identified from the instructions for the cal- Arnold and Fohrer, 2005; Gassman et al., 2007). ibration of the SWAT model, as given in the user’s In terms of hydrological simulations and assessment, manual (Neitsch et al., 2002a). A model-independent very few applications focus on influences of dams and non-linear parameter estimator, PEST (Parameter ESTi- sluices, especially the impacts of a large number of dams mation) (Doherty, 2004), has also been used to cali- and sluices. The following studies illustrate various ways brate SWAT, in which PEST implemented a particu- in which the model has been employed to investigate larly robust variant of the Gauss–Marquardt–Levenberg the impact of dams and sluices. Srinivasan et al. (1996) method of parameter estimation (Govender and Everson, applied SWAT to the Texas Gulf Basin, and 18 of the 2005; Wang and Melesse, 2005). Bekele and Nicklow 22 hydrologic unit areas (HUAs) were selected for mod- (2007) developed an automatic routine to calibrate daily elling. Average monthly results from two HUAs (Segiun streamflow and sediment concentration in SWAT using and Naches) were reported in this article, where two the Non-dominated Sorting Genetic Algorithm II (NSGA- reservoirs were considered in the Segiun river basin and II). NSGA-II is capable of incorporating multiple objec- one considered in the Naches river basin. The three reser- tives into the calibration process as well as employing voirs are located in the middle of these two adjacent parameterization to help reduce the number of calibra- HUAs, which are situated in the southeast of the Texas tion parameters. Muleta and Nicklow (2005) adapted a Gulf Basin. The results showed close agreement between Genetic Algorithm (GA) for single-objective evaluations, simulated and observed data, partially due to average and a Strength Pareto Evolutionary Algorithm for multi- monthly measured USGS streamflow data from the three objective optimization. The Shuffled complex evolution reservoir outlets that were used as input to the model. (SCE-UA) algorithm (Duan et al., 1992, 1994) has also Hotchkiss et al. (2000) incorporated complex operational been applied to calibrate SWAT in several cases (van rules into SWAT model, where six main stem dams on Griensven and Bauwens, 2003; Van Liew et al., 2005; the Missouri River were considered. Van Liew et al. Cao et al., 2006). (2003) used SWAT to investigate the impacts of retarding The aforementioned model calibrations were devel- structures on streamflow characteristics in southwestern oped in a progression from single-site to multi-site, and Oklahoma, where 13 flood retarding impoundment struc- tures were represented as uncontrolled reservoirs with from single-objective to multi-objective. However, few of principal and emergency spillways. In a case study of a these calibrations deal with hydrological simulations in a small watershed (17 km2) of India conducted by Mishra dam/sluice regulated river basin. As for this kind of spe- et al. (2007), SWAT was used to assess sediment trans- cific and complicated application, the concept of multi- port characterized by three on-stream sediment control site not only refers to multiple control points, but also structures called check dams. In these modelling studies refers to various types of outlets (hydrological station, involving dams, the reservoir module of SWAT becomes dam, or sluice) with