Abstract: Humans have altered the water cycle with methods such as construction of dams. Large dams and reservoirs can significantly modify the natural flow within channels, as well as the distribution of water in the basin. This kind of changes will evidently affect the runoff simulation which is of high importance in many kinds of hydrological research issues. The Soil and Water Assessment Tool (SWAT), one of the most popular hydrological models over the world, was designed for use on rural basins, and only take simple human activities into account. The algorithm in the reservoirs module of SWAT was designed for those dams just with flood control needs, and it unavoidably made the model not suitable enough for such basins with highly regulated reservoirs, especially in the application at daily time step. To consider the operations of reservoirs in runoff simulation, a new algorithm for reservoirs with predefined complex operational rules was integrated into SWAT in this paper, and the rules adopted were designed for the benefits including water supply, electricity generation, and flood control. The paper was aiming at providing users of SWAT model with a simple and convenient way to deal with reservoirs in runoff simulation and improving the runoff simulation of SWAT model in regions with highly regulated reservoirs at both daily and monthly time steps. The performance of the modified model was tested at daily and monthly time steps over the East River (Dongjiang) Basin in China. After comparing the outputs of runoff simulation before and after modification, the results showed that the modified SWAT model could solve the problem that the simulated outflow of reservoirs had been too low in non-flood seasons and excessive in flood seasons before. Moreover it could work out a better output in runoff simulation at both monthly and daily time steps. The improvement of runoff simulation at the stations hugely impacted by dams was more remarkable. In addition, there is also an impressive result in the simulation of the water level in reservoirs. In terms of parameters, the sensitivities of SOL_K (Saturated hydraulic conductivity), SOL_AWC (Available water capacity of the soil layer) and GWQMN (Threshold depth of water in the shallow aquifer required for return flow to occur) became lower after modification. In details, at the monthly time step, the coefficient of determination (R2) was increased by 0.06~0.22 in calibration period and 0.1~0.19 in validation period , while the Nash-Sutcliff Efficiency (NS) was increased by 0.06~0.52 in calibration period and 0.12~1.22 in validation period after modification; at the daily time step, R2 was increased by 0.04~0.16 in calibration period and 0.11~0.15 in validation period, while NS was increased by 0.04~0.46 in calibration period and 0.15~1.21 in validation period after modification. Although the accuracy of the model is limited by the error in simulating the process upstream of the reservoirs, the improvement of the runoff simulation in the East River Basin is still significant after modification. Thus, it is a valuable reference for hydrological simulation in such watersheds that are hugely affected by reservoirs operations.

Key words: SWAT model, reservoir regulation, hydrological simulation, the East River Basin