Seasonal forecasting of tropical cyclones in the North Indian Ocean region: the role of El Niño-Southern Oscillation

Abstract

In this study, we have investigated the contribution of El Nino-Southern Oscillation (ENSO) to the North Indian Ocean (NIO) tropical cyclone (TC) activity and seasonal predictability. A statistical seasonal prediction model was developed for the NIO region tropical cyclone genesis, trajectories and landfalls using the Southern Oscillation index (SOI: as a metric of ENSO) as a predictor. The forecast model utilised kernel density estimation (KDE), a generalised additive model (GAM), Euler integration, and a country mask. TCs from the Joint Typhoon Warning Centre were analysed over the 35-year period from 1979 to 2013. KDE was used to model the distribution of cyclone genesis points and the cyclone tracks were estimated using the GAM, with velocities fit as smooth functions of location according to ENSO phase and TC season. The best predictor lead time scales for TC forecast potential were assessed from 1 to 6 months. We found that the SOI (as a proxy for ENSO) is a good predictor of TC behaviour 2-months in advance (70% skill). Two hindcast validation methods were applied to assess the reliability of the model. The model was found to be skillful in hindcasting NIO region TC activity for the pre and post monsoon season. The distribution of TC genesis, movement and landfall probabilities over the study period, as well as the hindcast probabilities of TC landfall during ENSO events, matched well against observations over most of the study domain. Overall, we found that the phase of ENSO has the potential to improve NIO region TC seasonal forecast skill by about 15% over climatological persistence.