We have found single model time lagged ensembles to be helpful in forecast activities at NWS Raleigh to ascertain model trends, gauge model consistency or volatility and to consider in determining forecast confidence. We access this data through Bufkit via locally constructed files of several previous model runs of a single model. The scripts used to create these files were generously provided by WFO Wilmington OH. Currently we have time lagged ensembles for the HRRR, RAP, NAM, and GFS.
An example of a GFS time lagged ensemble showing the guidance trend toward a more rapid arrival of cold air on 03 UTC Sunday 15 February, 2015 at Greensboro is shown
below. The model initialization time for each cycle is noted with all of the forecasts valid at 03 UTC Sunday. The more rapid arrival of the cold and even dry air is noted with the 850 hPa temperature becoming 12C colder during the evolution of the model cycles. This trend can also be seen in the partial thickness nomogram. It’s important to note that the forecasting axiom of “The trend is your friend” is true much of the time but not always so forecasters must use this data intelligently.