Tropical Cyclones in Current Seasonal Forecast Models

This study assesses the skill of current seasonal forecast models in capturing observed characteristics of Northern Hemispheric Tropical Cyclones. In addition to delivering new scientific insights into their predictability on seasonal time-scales, this study also provides information about the potential usability of these forecasts by quantifying their reliability and robustness.

Using data from state-of-the-art seasonal forecast systems from five different centres (ECMWF, UK Met Office, DWD, CMCC, Météo-France), Tropical Cyclones are identified using a widely applied objective Tropical Cyclone tracking algorithm based on relative vorticity fields. Analyses are carried out for the three-months season from August to October and for three different ocean basins (Western North Pacific, Eastern North Pacific and North Atlantic). Besides the models ability to represent observed characteristics of Tropical Cyclones (e.g. lifetime, maximum intensity, genesis) their performance with regards to representing observed teleconnections, e.g. to ENSO is analysed. Special focus is put on quantifying the skill of each individual forecasting system as well as for the combined multi-model ensemble. This is conducted by using several deterministic and probabilistic skill scores, which also allows for quantifying the forecast reliability.

Early results are encouraging with positive and potentially useful skill for Tropical Cyclones over parts of the Northern Hemisphere in ECMWF-SEAS5. Especially over the North Atlantic basin, this model adequately captures the observed seasonal cycle as well as parts of the observed interannual variability. However, the level of skill seems to be partly sensitive to the choice of reference dataset, e.g. JRA55/ERA-Interim/IBTrACS.

.