Sensitivity analysis of the SWAT model to spatial distribution of precipitation in streamflow simulation in an Arctic watershed

• Published in: IOP Conference Series: Earth and Environmental Science
• Main Authors: Bui, Minh Tuan, Lu, Jinmei, Nie, Linmei
• Format: Article in Journal/Newspaper
• Language: English
• Published: IOP Publishing 2020
• Subjects: VDP::Technology: 500; VDP::Teknologi: 500; Arctic; Lundberg; Målselv; Norway; Northern Norway
• Online Access: https://hdl.handle.net/10037/20025; https://doi.org/10.1088/1755-1315/581/1/012025

This study approached a physically based, semi-distributed SWAT model to test the model sensitivity to the spatial distribution of precipitation. Ten scenarios of precipitation from five scattered rain gauges in an Arctic watershed Målselv in northern Norway were used as inputs to run the SWAT model. Streamflows were simulated. The model runs at monthly time interval based on the historical data of precipitation from 1979-2012. The study used statistical parameters, values of long-term average monthly streamflow and streamflow hydrograph between simulated and observed data for sensitivity analysis. The study found that the result of streamflow simulation is highly sensitive with spatial distribution of rain gauges input. For instance, the scenarios integrating rain gauge number 3, locating inside the watershed with lower precipitation amount than average of selected rain gauges, provided model unsatisfactory (statistical coefficient NSE < 0.5) in streamflow simulation. However, streamflow simulation is satisfactory (NSE: 0.5-0.6) at hydro-gauging station Lundberg far away from rain gauge 3. The hydrograph showed underestimated streamflow simulation in scenario 3,5,6-10 that integrated rain gauge 3, while scenario 1,2,4 that excluded rain gauge 3 showed reasonable agreement between simulated and observed flow. Underestimated streamflow was only found in scenario 3 and 5 at Lundberg. Moreover, the curves of average monthly streamflow showed that the simulated peak discharge in scenario 1,2,4 was performed better than the remaining scenarios.