DEVELOPMENT OF PHYSICALLY BASED FLOOD FREQUENCY MODELS

Abstract

Flood frequency analysis is one of the most active areas of hydrological research. In the past, efforts have been mainly concentrated on the statistical analysis of available flood data. Statistical flood frequency methods require long term homogeneous series of flood characteristics such as peak discharge, volume, duration etc. Collection of flood data for a long period is tedious and expensive. Keeping this in view, attempts have been made in the past to develop physically based flood frequency models. These models use readily available rainfall data and catchment characteristics. The physically based flood frequency models or derived flood frequency distributions (DFFD) were first introduced by Eagleson (1972). The DFFD models have three components viz. (i) stochastic rainfall model, (ii) infiltration model and (iii) effective rainfall-runoff model. The stochastic rainfall model used by most of the researchers assumes bivariate exponential distribution of rainfall intensity and duration and these variables are considered to be independent of each other. The 0-index, Philip's infiltration equation and SCS curve number method have been tried as infiltration models. Kinematic wave (KW), geomorphologic instantaneous unit hydrograph (GIUH) and geomorphoclimatic instantaneous unit hydrograph (GcIUH) have been used as effective rainfall-runoff models. The physically based flood frequency models provide a potentially attractive and alternative solution to ungauged watersheds. The impact of watershed changes on flood magnitudes and frequencies can be studied through DFFD models. In the present study, existing DFFD models have been applied to five watersheds of Sub zone - 3C (India) and their performance evaluated. New DFFD ii models using bivariate exponential distribution for correlated and independent rainfall intensity and duration have also been developed. The new DFFD models developed in the study were also applied to three watersheds of U.S.A. Detailed at site/regional and regional flood frequency analysis for Sub zone - 3c has been carried out for comparing the performance of various DFFD models. It has been found in the present study that the parameters of stochastic rainfall model are most sensitive input to the DFFD models and therefore, should be estimated carefully. Out of the three infiltration models used, the parameter of SCS curve number model can be estimated quite easily with reasonable accuracy. The DFFD models based on this infiltration model perform better than the other models. GcIUH and KW theory based effective rainfall-runoff models perform equally well in DFFD models. The quantiles estimated by the DFFD models which consider rainfall intensities to be independent of their durations are higher than the flood quantiles estimated by the proposed model which accounts for the negative correlation between these variables. DFFD models for positively correlated case still need to be developed. Physically based flood frequency models are relatively new in the field of hydrology, and are under development stage. There is a need for application of these models to more watersheds having long term reliable rainfall and runoff data before recommending them for field use.