Estimation of Flash Flood Guidance Considering Uncertainties of Rainfall-Runoff Models

Nowdays, the flash flood is characterized as flood leading to damage caused by heavy rainfall for short duration, steep slope and impervious layer. Furthermore, in Korea, the flood damage risk by flash flood is increasing due to climate change. So, we could use the Flash Flood Guidance(FFG) which means the amount of effective rainfall needed for flash flood occurrence. The FFG is calculated from rainfall-runoff relationship curve that is derived from rainfall-runoff model and threshold runoff. Therefore, the FFG can be changed largely by the rainfall-runoff models and their parameters. Here we used four different rainfall-runoff models. Also, the FFG can show different value by Threshold Source Area / Minimum Source Channel Length for calculating parameters of GcIUH, and by section of stream for calculating bank-full discharge. In this study, we analyzed the uncertainties on rainfall-runoff relationship derived from rainfall-runoff models and their parameters, threshold runoff, and existence and nonexistence of stream section. As a result, the derived confidence intervals for rainfall-runoff relationship curves show different shape and extent of interval. We performed the rainfall analysis as follows: First, we distributed probabilistic rainfall by 1hr, 3hr and 6hr using Huff method. But we can not draw rainfall-runoff relationship curve through one probabilistic rainfall. Therefore, we divided that into ten and derived rainfall-runoff relationship using each distributed probabilistic rainfall by rainfall duration. Then, we suggested the confidence interval on average of sampled rainfall through Monte Carlo simulation for uncertainty analysis of rainfall-runoff relationship curve. Next, we gave weighting factor to each four model and combined one. So, we can suggest one rainfall-runoff relationship curve from four rainfall-runoff models. And we compared FFG by change for value of CSA(Critical Source Area)/MSCL(Minimum Source Channel Length), and existence and nonexistence of stream section data. When stream cross section data was inputted, FFG value is larger than not inputted. Also, FFG value was more influenced by CSA than MSCL. Therefore, on flash flood forecasting, it's important process to select to proper model and optimal parameters for estimating FFG.