A120-0003
A new long-range island scale tropical cyclone outlook for southwest Pacific nations and territories

Friday, 11 December 2020
Poster
Andrew Magee, University of Newcastle, Centre for Water, Climate and Land (CWCL), Callaghan, NSW, Australia, Andrew Lorrey, NIWA National Institute of Water and Atmospheric Research, Auckland, New Zealand, Anthony Kiem, Centre for Water, Climate and Land (CWCL), University of Newcastle, Callaghan, NSW, Australia and Kim Colyvas, University of Newcastle, Newcastle, NSW, Australia
Abstract:
Southwest Pacific (SWP) tropical cyclones (TCs) account for up to 76% of natural disasters in the region, impacting people, infrastructure and economies across 15 island and atoll nations and territories. The spatio-temporal variability and erratic nature of SWP TCs exacerbates the vulnerability of the SWP to these extreme events. Producing skilful statistical TC outlooks remains a significant challenge for the region. Current statistically-derived operational outlooks typically have poor spatial resolution (not at the island/atoll scale), low skill, and many only consider the influence of El Niño-Southern Oscillation (ENSO) in modulating TC activity.

To address these issues, we have developed the Long-Range Tropical Cyclone Outlook for the Southwest Pacific (TCO-SP). Using generalised linear modelling, assuming a Poisson regression, we demonstrate that skilful TC outlooks can be derived for island and regional-scale locations across the SWP, up to four months before the start of the SWP TC season (November-April). Using up to 14 predictors which modulate the spatio-temporal characteristics of SWP TCs, hundreds of possible model covariate combinations are tested. For each location and time period, the best performing models are selected and undergo extensive three-way cross-validation.

Monthly updates are provided between July and January, offering continuous refinement of pre-season and in-season TC counts. For example, the New Caledonia TC outlook model (pre-season July outlook) is significantly correlated with IBTrACS up to r= 0.78, with a skill score of 60%. Between 1970-2019, the model has a seasonal strike rate of 34% (i.e. correctly predicted TC counts for 17 TC seasons out of 50) and a seasonal strike rate (±1 TC) of 84% (42 TC seasons out of 50). Other island/regional-scale models demonstrate similar skill.

TCO-SP, a new island-nation scale TC outlook has important implications for disaster risk reduction and will assist local authorities to prepare for the coming season’s cyclone activity. TCO-SP produces both deterministic and probabilistic outlooks between July and January and will be used to complement existing TC outlooks for the region. Outlooks are available at www.tcoutlook.com.

<< Previous Abstract | Next Abstract >>