A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science

Thesis (M.Sc.) - Asian Institute of Technology, 1989

A Computer Assisted System for Flood Frequency Analysis, called CASFFA, was designed and implemented in this study. It consists of two components, the first, called DPU, is used for data preparation, and the second, called FFAS, carries out the required steps in flood frequency analysis on the data provided by the DPU. The developed system was used to analyse flood data in Thailand for both at-site and regional approaches. In the analysis, three popular distributions, the Generalized Extreme Value (GEV), the Log Logistic (LLG) and the Three Parameter Log Normal (LN3), were adopted for use along with three regional procedures, namely ; the Station Year Index, the Wallis and the Lettenmaier procedure. Three homogeneous regions, namely the Northern, Chi and Mun were identified by means of the Krusal - Wallis test. For the Northern region, the Lettenmaier regional procedure based on GEV fitted by the unbiased PWM estimators (GEV/PWM) is usually the best in terms of the standard deviation and the relative bias while the Wallis regional procedure used along with the LLG fitted by the plotting position estimators (LLG/PLOT) is the best in terms of the percent standard error. For both Chi and Mun regions, the Wallis regional procedure is found to be better in terms of the standard deviation when it is used along with the LLG/PLOT procedure and is better in terms of the percent standard error when it is used along with the LLG/PWM while the station year index regional procedure based on the LLG/ PWM is best in terms of the relative bias. It was also found that in most cases, the regional approach provides better estimation for the T- year event than does the at-site approach.