Numerical prediction of a cold-air outbreak: A case study with ensemble forecasts

The forecastability of a cold-air outbreak over eastern North America durin g January 1985 has been studied with ensemble forecasts from the NCAR Commu nity Climate Model version 2 run at T42 horizontal resolution. The cold-air outbreak case was characterized by a pool of very cold air (T < -35 degree s C at 850 mb) that moved southward into the central United States and inte nsified. The ensemble's 10 member forecasts were initialized at 0000 UTC 15 January 85, a few days before the cold-air pool began its southward moveme nt and reached its peak intensity. The ensemble members predicted the south ward passage of the cold air but faster and weaker than analyzed. The predi cted weakening of the cold-air pool was consistent with the model's systema tic error Quasi-Lagrangian diagnosis of the 850-mb temperature tendency bud get revealed that the analyzed intensification of the cold-air pool was due to residual rather than adiabatic effects. These residual effect, could ha ve been diabatic in origin but also attributable to observational errors. S imilar diagnoses applied to selected ensemble members indicated that diabat ic cooling, specifically longwave radiative cooling, contributed to the for ecast cooling of the cold-air pool by one ensemble member but was overwhelm ed by adiabatic warming in a weakening cold-air pool predicted by another e nsemble member. These results suggest that the forecast details of a cold-a ir outbreak may depend upon the subtle balance between diabatic and adiabat ic processes. Furthermore, forecasts constructed from ensemble predictions must account for model biases as well as information from the ensembles.