The CP ensemble was run at 4.4 km resolution using the latest tropical configuration of the CP version of the Unified Model (UM). Forecasts were run twice per day out to five days, and were nested inside the 18 members of the latest Met Office global model ensemble (MOGREPS-G) run.
A subjective assessment of these real-time forecasts suggests that the CP ensemble was consistently providing added value, both in terms of forecast track and intensity, over MOGREPS-G. For Hurricane Irma, for example, the forecast initialised five days prior to Florida landfall predicted that the hurricane was more likely to make landfall on the west side of Florida rather than on the east side. This was consistent with the equivalent MOGREPS-G forecast. However, the timing of landfall with the CP ensemble was close to the observed time, whereas the MOGREPS-G forecast predicted landfall twelve to twenty-four hours later, due to a southerly track bias over Cuba not seen in the CP ensemble. For Hurricane Maria, the CP ensemble captured the rapid deepening phase in forecasts initialised up to 36 hours before it began. The rapid deepening took Maria from being a category 1 to category 5 storm in little more than 12 hours and, in contrast to the CP ensemble, there was no indication of this deepening in the MOGREPS-G forecasts.
Motivated by these encouraging initial real-time results, a more systematic objective assessment of the performance of the CP ensemble is now underway and will be presented. An assessment of the additional qualitative and quantitative information that the CP ensemble provides on other features such as storm size, storm structure and the precipitation distribution will also be presented.
Finally, ongoing work and future plans to further develop and improve the performance of the CP ensemble will be outlined.