Controversial Application of SWAT Model to Kurau River Basin, Malaysia

Proceedings of The 3rd World Conference on Engineering and Technology

Year: 2022



Controversial Application of SWAT Model to Kurau River Basin, Malaysia

Rasha M. Fadhil, Md Kamal Rowshon, Koichi Unami, Aimrun Wayayok, Ahmad Fikri



Minimizing discrepancies between observed and predicted data series is one of the central topics of hydrological process modelling. There are two contrasting approaches of hydrological models: deterministic and stochastic. This paper discusses the controversy over deterministic models, mainly applied to practical problems. The Soil and Water Assessment Tool (SWAT) distributed watershed model hypothesizes the correctness of its model structure representing deterministic physical phenomena. SWAT coupled with Sequential Uncertainty Fitting 2 (SUFI-2) calibration and uncertainty tool, programmed in SWAT Calibration and Uncertainty Procedures (SWAT-CUP). In contrast, one of the stochastic models, is autoregressive with exogenous input (ARX) model with linear Markovian input-output relationship between rainfall and streamflow. Malaysia’s Kurau River Basin (KRB) is chosen as a study of predicting streamflow through both mentioned models. Common statistical indicators, coefficient of determination (R2), Nash-Sutcliffe Efficiency (NSE), and Percent Bias (PBIAS), are used for assessing the performance of both models. The outcomes of the sensitivity and uncertainty analysis of parameters using SWAT with SUFI-2 indicate that baseflow and percolation are notified as essential components for total discharge. However, the SWAT-CUP parameters designed for the environment of the North American continent are not suited for all regions. It can be observed that the SWAT model is deficient in capturing the baseflow due to the limitation of its ability to represent groundwater flows rigorously. Furthermore, the SWAT model needs modification to adapt to the peculiar hydrological environment of KRB. The impact of artificial viscosity on flattening and filtering peaks of the streamflow also needs evaluation.

keywords: ARX, Deterministic model, Statistical indicators, stochastic model, Streamflow prediction.