A Dynamically Adapting Forecast Cone Based on Ensemble Spread

Dynamically-based ensemble prediction systems (EPS) have gained considerable attention because they can provide the range of possible forecast outcomes and quantify the uncertainty in forecasts. In turn, this allows forecasters to convey clearer messages to the public on the range of forecast scenarios and displaying inherent uncertainty in weather forecasts, which can be difficult to do with deterministic forecasts. Although global ensemble prediction systems have demonstrated skill in their probabilistic track predictions, their information is not always fully utilized beyond looking at the mean forecast and standard deviation (i.e., spread). One way that these ensembles could be better employed is in the creation of the “TC Forecast Track Cone”, which is used by the National Hurricane Center (NHC) to convey the potential track forecast uncertainty. Currently, the forecast cone is based on the Official Forecast error averaged over the last five years; therefore, it does not reflect the confidence and uncertainty that NHC forecasters have in their forecast. Given the skill of global EPS, there is an opportunity to retain the current cone definition, but dynamically adjust in size based on the ensemble forecast uncertainty.

The focus of this study is to create a dynamical cone algorithm based on the TC track forecast spread in the Global Ensemble Forecasting System (GEFS) and/or ECMWF EPS. The algorithm stratifies forecasts into one of three categories based on the ensemble position spread at individual hours and computes the 66.67% error based on NHC forecasts (the current definition of the cone). Preliminary results using random sets of forecasts from the 2017-2023 seasons for training and testing indicates that the ensemble forecast spread can distinguish the size of the NHC official error, yielding distinct cone sizes that differ by ~80% from the small to large spread bin. Various example cases will be presented showing the dynamical and static cones.