Hydrological evaluation of gridded climate datasets in a texas urban watershed using soil and water assessment tool and artificial neural network

Abstract

Precipitation is a vital component of the hydrologic cycle, and successful hydrological modeling largely depends on the quality of precipitation input. Gridded precipitation datasets are gaining popularity as a convenient alternative for hydrological modeling. However, many of the gridded precipitation data have not been adequately assessed across a range of conditions. This study compared three gridded precipitation datasets, Tropical Rainfall Measuring Mission (TRMM), Climate Forecast System Reanalysis (CFSR), and Parameter-elevation Relationships on Independent Slopes Model (PRISM). This study used the conventional gauge observation as reference data and evaluated the suitability of the three sources of gridded rainfall data to drive rainfall-runoff simulations. The Soil and Water Assessment Tool (SWAT) and Artificial Neural Network (ANN) were used to create daily streamflow simulations in the Leon Creek Watershed (LCW) in San Antonio, Texas, with the TRMM, CFSR, PRISM, and gauge rainfall data used as inputs. A direct comparison of the gridded data sources showed that the TRMM data underestimates the volume of rainfall, while PRISM data most closely matches the volume of rainfall when compared to the gauge rainfall observations. The hydrological simulation results showed that the PRISM and TRMM rainfall data driven models had preferable results to the CFSR and gauge driven models, in terms of both graphical comparison and goodness-of-fit indicator values. Additionally, no significant discrepancy was found between SWAT and ANN simulation results when the same precipitation data source was used, while SWAT and ANN simulation results varied in an identical pattern when different precipitation data sources were applied.