eCite Digital Repository

A statistical seasonal forecast model of North Indian Ocean tropical cyclones using the quasi-biennial oscillation


Wahiduzzaman, M and Oliver, ECJ and Klotzbach, PJ and Wotherspoon, SJ and Holbrook, NJ, A statistical seasonal forecast model of North Indian Ocean tropical cyclones using the quasi-biennial oscillation, International Journal of Climatology, 39, (2) pp. 934-952. ISSN 0899-8418 (2019) [Refereed Article]

Copyright Statement

Copyright 2018 Royal Meteorological Society

DOI: doi:10.1002/joc.5853


Previous studies have shown that the skill of seasonal forecasts of tropical cyclone (TC) activity over the North Indian Ocean (NIO) tends to be poor. This paper investigates the forecast potential of TC formation, trajectories and points of landfall in the NIO region using an index of the stratospheric quasi‐biennial oscillation (QBO) as the predictor variable in a new statistical seasonal forecast model. Genesis was modelled by kernel density estimation, tracks were fitted using a generalized additive model (GAM) approach with an Euler integration step, and landfall location was estimated using a country mask. The model was trained on 30 years of TC observations (1980–2009) from the Joint Typhoon Warning Center and the QBO index at lags from 0 to 6 months. Over this time period, and within each season and QBO phase, the kernel density estimator modelled the distribution of genesis points, and the cyclone trajectories were then fit by the GAM along the observed cyclone tracks as smooth functions of location. Trajectories were simulated from randomly selected genesis points in the kernel density estimates. Ensembles of cyclone paths were traced, taking account of random innovations every 6‐hr along the GAM‐fitted velocity fields, to determine the points of landfall. Lead–lag analysis was used to assess the best predictor timescales for TC forecast potential. We found that the best model utilized the QBO index with a 3‐month lead. Two hindcast validation methods were applied. First, leave‐one‐out cross‐validation was performed where the country of landfall was decided by the majority vote of the simulated tracks. Second, the distances between the landfall locations in the observations and simulations were calculated. Application of seasonal forecast analysis further indicated that including information on the state of the QBO has the potential to improve the skill of TC seasonal forecasts in the NIO region.

Item Details

Item Type:Refereed Article
Keywords:landfall, North Indian Ocean, quasi‐biennial oscillation, statistical modelling, tropical cyclone genesis, tropical cyclone trajectories
Research Division:Earth Sciences
Research Group:Oceanography
Research Field:Physical oceanography
Objective Division:Environmental Policy, Climate Change and Natural Hazards
Objective Group:Understanding climate change
Objective Field:Climate variability (excl. social impacts)
UTAS Author:Wahiduzzaman, M (Mr Mohammad Wahiduzzaman)
UTAS Author:Oliver, ECJ (Dr Eric Oliver)
UTAS Author:Wotherspoon, SJ (Dr Simon Wotherspoon)
UTAS Author:Holbrook, NJ (Professor Neil Holbrook)
ID Code:128876
Year Published:2019 (online first 2018)
Web of Science® Times Cited:2
Deposited By:Oceans and Cryosphere
Deposited On:2018-10-19
Last Modified:2019-07-02

Repository Staff Only: item control page