Modelling the extreme value of river flow data in West Sabah using bayesian approach

Cheong, Ri Ying (2018) Modelling the extreme value of river flow data in West Sabah using bayesian approach. Masters thesis, Universiti Malaysia Sabah.

[img] Text
24 PAGES.pdf

Download (704kB)
[img] Text
FULLTEXT.pdf
Restricted to Registered users only

Download (4MB)

Abstract

This study aimed to model the annual maximum series data of river flow in several sites in Sabah with small sample size to Generalized Extreme Value (GEV) distribution. In flood frequency analysis, annual maximum river flow is always used as an indicator. Uncertainty in the model and prediction process leads to inaccurate estimation of extreme events. Therefore, Bayesian approach is suggested to cope with the uncertainty involved. Maximum likelihood estimation (MLE) was treated as the standard parameter estimation method due to wide application in extreme value analysis. A stationary model which holds all the parameters constant is compared to two non-stationary models which consists of linear time dependent to location parameter as well as both location and scale parameters. A simulation study among probability weighted moment (PWM), MLE and Bayesian Markov Chain Monte Carlo (MCMC) were conducted to determine the best parameter estimation of GEV distribution. The performances were compared using root mean square error as well as bias. The results showed that Bayesian MCMC was better than PWM and MLE in estimating GEV parameters especially with small sample size. Likelihood ratio test showed that the annual maximum river flow data over the homogeneous region followed the distribution with common shape parameter. Hence, the quantile estimation at 10-, 100-, 1000-year return period were obtained using new single model with Bayesian MCMC as the parameter estimation. This method was believed to consider the parameter uncertainty and provide a more reliable return level estimates. Most of the stations were found to exceed the maximum level once every 100 years.

Item Type: Thesis (Masters)
Keyword: River flow, Flood frequency analysis, Bayesian approach, Maximum likelihood estimation, Non-stationary models
Subjects: G Geography. Anthropology. Recreation > GB Physical geography > GB3-5030 Physical geography > GB651-2998 Hydrology. Water > GB980-2998 Ground and surface waters > GB1201-1598 Rivers. Stream measurements
Department: FACULTY > Faculty of Science and Natural Resources
Depositing User: DG MASNIAH AHMAD -
Date Deposited: 12 Mar 2025 14:43
Last Modified: 12 Mar 2025 14:43
URI: https://eprints.ums.edu.my/id/eprint/43125

Actions (login required)

View Item View Item