AFWA's Joint Ensemble Forecast System Experiment

The Joint Ensemble Forecast System (JEFS) project was developed to evaluate how ensemble modeling and certainty information could be used to the benefit of DoD, as tremendous potential exists to increase mission success, save resources, and improve safety by exploiting that information.

The project emphasized two areas:

• Science and technology needed to create useful (i.e. forecasts deviating from climatology) and reliable (i.e. the observed frequency matches the forecast probability) forecast information

• Products and training needed for both forecasters and decision makers to exploit that forecast information

A Joint Global Ensemble (JGE) was designed by ingesting and post-processing one degree global ensemble data from the National Centers for Environmental Prediction (NCEP), Fleet Numerical Meteorology and Oceanography Center (FNMOC) and the Canadian Meteorological Centre (CMC). A Joint Mesoscale Ensemble (JME) was created by running ten independent models within the WRF framework with varied physics and initial/boundary conditions. Ten members of FNMOC's COAMPS model were to be included as part of this ensemble, but communication challenges prevented this from happening.

Objective verification showed that the global ensemble was both useful and quite reliable. This was attributed in large part to the fact that ensembles from multiple centers were included in its design. Subjective verification of high-impact events in the global ensemble showed that while it was able to predict the large scale features (e.g. upper waves, deep layer moisture) with usefulness, it was largely unable to simulate the finer scale details owing to its coarse resolution, leading to forecasts that remained reliable, but were not as useful.

Objective verification showed that the mesoscale ensemble was more useful than the global ensemble, but not as reliable as the global ensemble. The lack of reliability was attributed to only using NCEP ensemble members for initial conditions, a lack of data assimilation, and a limited number of ensemble members. Subjective verification of high impact events in the mesoscale ensemble showed that it could forecast useful probabilities of these “rare” weather events that the global ensemble could not, but that reliability was often less than desired more than 24 hours before the event.

Many forecast variables are more challenging to predict than others, specifically those that occur at scales smaller than the model resolution. Unfortunately, variables that impact the warfighter largely tend to fall into this category (e.g. clouds, icing, lightning, dust). Predicting these variables reliably and usefully requires not only a well designed ensemble, but also statistical and physical algorithms to account for the sub-grid scale uncertainties due to deficient models. Some success was achieved on this front in JEFS.

A product suite was developed that contained imagery for the meteorologist (e.g. upper level charts, meteograms) and the decision maker (e.g. probabilities of specific mission hindering phenomena like lightning or visibility). Feedback was quite positive from forecasters in the field who were often eager to leverage ensemble information to improve their forecasts. Many completed training developed at AFWA or at their OWS using local ensemble experts with relative ease. However, little progress was made passing certainty information on to decision makers, which is where the most benefit exists.