Assessment of Uncertainty in Grid-Based Rainfall-Runoff Model Based on Formal and Informal Likelihood Measures

Damage prevention from the local storms and typhoons in Korea, the development of a rainfall-runoff model reflecting local geological, meteorological and physical characteristics is necessary. The accuracy of the rainfall-runoff model is influenced by the various uncertainty factors that can occur in the modeling processes, including input data, model parameters, modeling simplification, and so on. Thus, the objectives of this study are (1) to estimate runoff for two rainfall events using Grid Rainfall-Runoff Model (GRM); (2) to quantify the uncertainty of the GRM model using the Generalized Likelihood Uncertainty Estimation (GLUE) method, and (3) to assess the uncertainty ranges of the GRM based on different likelihood functions. For this, two rainfall events were implemented to the GRM in the Cheongmicheon watershed, and informal likelihood functions (LNSE, LPBIAS, LRSR, and LLOG) based on the fitness indices (NSE, PBIAS, RSR, and Log-normal) were used for uncertainty analysis and quantification using GLUE method. As a result, the GRM parameters varied according to the different rainfall patterns even in the same watershed. In addition, among the GRM parameters, the CRC (Channel Roughness Coefficient) and CSHC (Correction factor for Soil Hydraulic Conductivity) characteristics are the most sensitive. Moreover, this study showed that the uncertainty range of the GRM model can be changed with the subjective selection of likelihood functions and thresholds. The GRM model is open source and has good accessibility. Especially, this open-source model allows various approaches to disaster prevention plans such as flood forecasting and flood insurance policies. In addition, if the parameter range of GRM is quantified and standardized at domestic watersheds, it is expected that the reliability of the rainfall-runoff simulation can be increased by the reduction of the uncertainty factors.