Evaluation of gridded precipitation products for streamflow modelling in GIN watershed, Sri Lanaka

Abstract

Watershed, Sri Lanka An accurate representation of spatial precipitation is significant for hydrological studies. Spatial precipitation is also the basic input for distributed hydrological models and the accuracy of spatial precipitation affects the performance of hydrological models. In many parts of the world, ground-based observation networks are inadequate to capture spatial precipitation because gauge stations cannot be set up anywhere as financial and geographical factors play a vital role in the establishment. To overcome those challenges two existing gridded precipitation data (TRMM and APHRODITE) are used to simulate discharge in the Gin watershed of Sri Lanka. The coefficient of determination improves to 0.78 and 0.65 respectively for TRMM and APHRODITE data after bias correction. While comparing two gridded precipitation data to observed data, the TRMM data shows superior to APHRODITE with the same value of daily and a monthly average rainfall of 11.15 mm and 339.29 mm respectively. The standard deviation shows 21.16 for daily and 167.72 for a monthly scale with the difference of 31.00 % and -0.06 % to observed the data set. The HEC-HMS model is used for generating streamflow from the two gridded and observed data against gauge data. From the other four-parameter (SCS Unit Hydrograph, Simple Canopy, SCS Method, Simple Surface, and Recession) soil moisture accounting parameter calculation was challenging as it has to be carefully determined. The three most sensitive parameters are soil percolation, tension zone storage, and impervious area while the groundwater storage two (GW2) is the least sensitive parameter. Model performance criteria such as RMSE, NSE, and PBIAS are carried out for calibration and validation. The observed data performed good in the simulation of streamflow compared to two gridded precipitation data with an NSE value of 0.70, RMSE Std Dev value of 0.50, and PBIAS of -8.40 % for calibration and NSE value of 0.66, RMSE Std Dev value of 0.66, and PBIAS of -2.34 % for validation. The result shows that the TRMM data is more suitable to be used for hydrological modelling for and water resources management in ungauged areas in Sri Lanka.