Markov Chain Monte Carlo simulation and regression approach guided by El Niño–Southern Oscillation to model the tropical cyclone occurrence over the Bay of Bengal

Abstract

Tropical cyclone (TC) is one of the most devastating weather systems that causes enormous loss of life and property in the coastal regions of Bay of Bengal (BoB). Statistical forecasting of TC occurrence can help decision-makers and inhabitants in shoreline zones to take necessary planning and actions in advance. In this study, we have investigated the impact of El Nino–Southern Oscillation (ENSO) on the frequency of TC over the BoB by using 100 years TC and Southern Oscillation Index data. The frequency of TC is approximated through observation and Markov Chain Monte Carlo (MCMC) simulation. Two-sample Student’s t test has been applied for examining the statistical significance where the results are significant at 5% level for all cyclonic disturbances. The monthly and seasonal distribution show this feature more distinctly. The total annual frequency of depressions and cyclonic storms in El Nino and La Nina conditions does not differ much, but the monthly/seasonal distribution shows high differences for certain months and seasons. The simulated frequency of TC landfall using MCMC matches well with the observation. The proposed methodology is illustrated through a case study in BoB rim countries-Bangladesh, India, Sri Lanka and Myanmar. Poisson and Bayesian regression have also been used to predict the probabilities of TC frequency over the BoB. Both the regression approaches show 10 and 32% improvement than climatology for the forecast and cross-validation skill respectively. We have also analyzed TC impact over Bangladesh as a case study. Possible links of the variation of TC activities with the largescale geographical distribution of sea surface temperature, vertical wind shear, vorticity, moisture and relative humidity are also explored.