Statistical simulation of cyclonic storm surges in the coastal region of Bangladesh

Abstract

Very limited data on cyclonic storm surge heights from coastal areas of Bangladesh does not permit estimation of design surge heights by making frequency analysis of observed data. In this study, design surge heights of given return periods along with confidence intervals have been derived from synthetic data by performing a statistical simulation. In the simulation, cyclonic storm surges occurred randomly among three coastal regions. The regions are Ganges Tidal Plain (GTP), Meghna Deltaic Plain (MDP)and Chittagong Coastal Plain (CCP). The simulation model consists of three components: generation of annual maximumcyclonic wind speed, simulation of storm surge occurrence and genration of surge height, Annual maximumcyclonic wind speed data has been generated randomly from a . ,.~, Gumbel distribution. Parameters of the distribution have been estimated from observed data for the period 1960to 1991 using the maximumlikelihood procedure. The data is continuous for the period 19'7.1to 1991 but discontinuous for the period 1960 to 1972. The later data has oeen included in order to improve the reliability of the estimated parameters and the data has been treated as historical information. Estimators for the parameters of a Gumbel distribution from a sample with both systematic and historical data have been derived. A formula which is capable of predicting the maximumsurge height in the landfall area during a cyclonic storm has been selected. The formula requires maximum wind speed and length of the continental shelf as input data. A modification in the formula has been made in order to include the effect of pressure drop as well as to make it computationaly simple. It gives comparable results when verified against observed data and reported results from two dimensional numerical hYdrodynamic model studies of cyclonic storm surges in the Bay of Bengal. A methodology for determining the average maximumsurge height in a coastal region has been developed. Lognormal distribution is found suitable for fitting the distribution of surge height data while Gumbel distribution has been- found unsuitable. A methodology that is capable of treating zero values in a sample of surge data has been used in the frequency analysis of synthetic data using the lognormal distribution with the maximumlikelihood procedure.