Statistical forecasting of tropical cyclone landfall activities over the North Indian Ocean rim countries

Abstract

This paper investigates the forecast potential of tropical cyclone (TC) landfall probabilities over the North Indian Ocean (NIO) rim countries using ocean-climate predictor variables in a new statistical seasonal forecast model. A Poisson regression model was used to predict the landfall probabilities for a period of 35 years of TC observations (1979–2013) from the Joint Typhoon Warning Center and the predictor variables include sea surface temperature, Southern Oscillation Index and ocean heat content. Poisson regression was found skilful in hindcasting of tropical cyclone landfall frequency for the NIO region with a correlation coefficient in the forecasted hindcast time series of 0.65 and 31% improvement above climatology. In the present study, genesis was modelled by kernel density estimation, tracks were fitted using a generalised additive model (GAM) approach with a Euler integration step, and landfall location was estimated using a country mask. This GAM model that is previously demonstrated very skilful to simulate TC landfall across the NIO rim countries by Wahiduzzaman et al. (2017, 2019) deems skilful for this study.