A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering, School of Engineering and Technology

Thesis (M.Eng.) - Asian Institute of Technology, 1983

This study deals .with the analysis and simulation of the daily rainfall data of Thailand. Eleven rainfall stations, nine of which were having 27 years of record, one with 26 years of record, and the last with 25 years of record, were taken from the Electricity Generating Authority of Thailand (EGAT) . In the simulation, two first order Markov Chain Models are evaluated. Finally, a new mathematical model based on the method of fragments is developed and evaluated for daily data generation (Model 3) This model is further divided into two types. The first type uses Thomas-Fiering model, for producing the monthly amounts which sum up to the annual values. This is intended for relatively short historical records (say 10 years) . On the other hand, the second type directly simulates the annual rainfall values using either Jackson scheme or AR(l) model , which ever case is applicable. This second type is good for relatively long historical records (say more than 30 years) . It is observed in this study that all the models are applicable. As expected, they differ in some aspects such as accuracy , computer time and storage requirements, adequacy, simplicity, convenience and practicality. As an example, Model 3 (first type) requires the largest computer time consumption while Model 3 (second type) requires the least. With regard to simplicity and practicality, Model 3 ranks the first among all. For accuracy, Model 2 (gamma distribution) affords to be superior. Finally Model l is found to be the most complicated. Accordingly, all the computer programs needed in this generation are written with sufficient documentation to help any user in applying the models.