Utilizing Stream Network for Regional Flood Frequency Analysis in Ganga Basin

Abstract

Regional flood frequency analysis (RFFA) is used for estimating flood quantiles at ungauged sites or sites with low record length. RFFA estimates flood quantiles at ungauged sites using regional regression models. These regression models are mostly obtained using the ordinary least square regression (OLS), weightage least square regression (WLS) or generalized least square regression (GLS) methods. Literature shows that the GLS method gives better estimates as it accounts for cross correlation in the discharge among the gauged sites, which is related to the geographical distance between the gauging sites. However, the stream network that provides information on natural or physical connectivity between the gauging sites is ignored. In this study, the GLS based regression model is tested for estimating flood quantiles at ungauged sites by using the stream network distance. This methodology is applied to 53 sub-basins in the Ganga basin, India. The comparison of the performance measures using the geographic and the stream network distance shows improvement in the flood quantile estimates for the stream distance based regression model. This study shows the usefulness of stream network information for RFFA and suggests the need for future research to efficiently incorporate stream network information in flood quantile estimation.KeywordsRunoffStream networkFlood frequency analysisRegression equation