Comparative study of semi-distributed and 2D-distributed hydrological models for streamflow prediction in a data scarce mountainous watershed

With the climatic and irregular topography favor, streamflow simulation in the study area is often challenging because of complex hydrological process and large uncertainty. To face these challenges, semi-distributed, and 2D-distributed hydrological models, SWAT and RRI, were proposed to carry out the streamflow simulation in this study. The present work aims to compare and evaluated the hydrological behavior of the selected models for a data-scarce mountainous watershed located in Myanmar. The models’ calibration (2001–2010) and validation (2011–2015) process are done in six gauging stations along the Chindwin River of the study area. In calibration, SWAT gave satisfactory monthly performances measured with the coefficient of determination (R) ranging between 0.65 and 0.78, Nash–Sutcliffe efficiency (NSE) between 0.5 and 0.65, whereas RRI obtained R was 0.65–0.7, NSE between 0.54 and 0.71, respectively. The predicted flow from RRI model gives a good agreement with the observed flow, is slightly better than the SWAT model in daily simulation from the comparative analysis. The variation of dry and wet seasonal estimated streamflow of the hydrological models relative to observed flow in calibration and validation was explored by Box plot with a whisker. Parameter sensitivity analysis results show CN2 and GW_DELAY are the high class parameters for SWAT. The manning “n” value and soil depth “d” parameters were also the most sensitive parameters in RRI model. The results from this study will helpful water resources managers and stakeholders involved in the water management of local river basins using models.