Comparison of SWAT and GSSHA for High Time Resolution Prediction of Stream Flow and Sediment Concentration in a Small Agricultural Watershed

Ratino Sith,Kazuo Nadaoka

doi:10.3390/hydrology4020027

Ratino Sith, Kazuo Nadaoka

Open Access

https://doi.org/10.3390/hydrology4020027

Copy DOI

Journal: Hydrology	Publication Date: May 4, 2017
Citations: 15	License type: CC BY 4.0

Affiliation: Tokyo Institute of Technology

Abstract

In this study, two hydrologic models, the Gridded Surface Subsurface Hydrologic Analysis (GSSHA) and the Soil and Water Assessment Tool (SWAT), were applied to predict stream flow and suspended sediment concentration (SSC) in a small agricultural watershed in Ishigaki Island, Japan, in which the typical time scale of flood event was several hours. The performances of these two models were compared in order to select the right model for the study watershed. Both models were calibrated and validated against hourly stream flow and SSC for half-month periods of 15 to 31 May 2011 and 17 March to 7 April 2013, respectively. The results showed that both models successfully estimated hourly stream flow and SSC in a satisfactory way. For the short-term simulations, the GSSHA model performed slightly better in simulating stream flow as compared to SWAT during both calibration and validation periods. GSSHA only gave better accuracy when predicting SSC during calibration, while SWAT performed slightly better during validation. For long-term simulations, both models yielded comparable results for long-term stream flow and SSC with acceptable agreement. However, SWAT predicted the overall variation of long-term SSC better than GSSHA.

Highlights

In recent years, a number of physically-based and distributed watershed models have been developed and applied [1]
Simulation of Baseflow and Stream Flow. Both Gridded Surface Subsurface Hydrologic Analysis (GSSHA) and Soil and Water Assessment Tool (SWAT) models were calibrated against observed stream flow and suspended sediment concentration on hourly temporal resolution from 15 to 31 May 2011 during which four significant flood events were observed
Both models performed well in estimating baseflow in the study watershed with very good statistical agreement (R2 = 0.93, Nash–Sutcliffe Efficiency (NSE) = 0.90 for GSSHA and R2 = 0.90, NSE = 0.88 for SWAT) during calibration and (R2 = 0.89, NSE = 0.88 for GSSHA and R2 = 0.92, NSE = 0.88 for SWAT) during validation

Summary

Introduction

A number of physically-based and distributed watershed models have been developed and applied [1]. The watershed models commonly use three approaches (lumped, semi or quasi-distributed, and fully distributed approaches) to simulate hydrological responses. To select an appropriate model for a specific watershed, a comparison study is obviously needed. Various researchers have reviewed and compared the performances between modeling approaches in investigating hydrological processes [2,3,4,5,6]. This study was carried out to compare a fully distributed GSSHA) [7] and a semi-distributed continuous model (Soil and Water Assessment Tool, SWAT) [8] for estimating hourly stream flow and sediment concentration. There is no previous comparison study between these two models, especially with the sub-daily configuration. GSSHA was compared with a lumped model, namely, Hydrologic Engineering Center-Hydrologic

Objectives

Methods

Results

Conclusion