Guidance for Adjusting for the Forecast Error

When dealing with ensembles to make a forecast, if the ensemble mean is accepted as the most likely outcome, ofter forecasters will take the ensemble member closest in appearance to the ensemble mean and make use of its data. However, what should be done when the ensemble mean is not expected to verify?

Two possible methods are discussed below. These two should not be considered exhaustive.

• Use dynamical and physical reasoning based on the existing output and the expectation of a stronger short-wave disturbance. For example, possible adjustments to the NWP and ensemble forecasts might consider
• Greater diabatic heating from more precipitation, resulting in even further deepening of the disturbance and more ridging downstream of the disturbance which will result in a deeper storm and further west track than the operational models and ensembles indicate.
• May influence development and intensity of jet streaks, jet-streak induced indirect circulations, and related cold air damming along the Appalachians
• Use the ensemble that most resembles the adjusted outcome as forecast guidance
  • Example: SREF member Eta-P1 from 09 UTC 06 January 2002

  SLP, prec., snow., freezing rain for Eta P1 from 12 UTC 6 Jan. - 9 UTC 7 Jan. 2002

  • We compare the 18 UTC 06 January 2002 forecast versus surface map by clicking on the graphic button below.

    Comparison of 18 UTC 6 January SREF Eta p1 forecast to observations

    In this graphic, we see that

    • The low pressure area is:
      • Double-centered
      • Deepening rapidly in the observations (10-hPa/3-hr pressure falls at Cape Hatteras!)
      • Not deep enough in the forecast
      • Too far north and east for the coastal low, too far south and west for the inland low.

      There is also a pronounced coastal front developing, with 30° and more temperature contrast across it in NC. Cold air damming is evident inland over the Mid-Atlantic Piedmont.

    • The precipitation is:
      • Forecast to have advanced too far north and east
      • Forecast to be maximum over the Piedmont from MD to NC, where little or no precipitation was actually occurring at 18 UTC
      • Snow in areas that were forecast to be rain (WV, eastern KY). Moderate snow is already falling in Johnstown, PA and Morgantown, WV, and snow has started in State College, PA. Given the strength of the storm center and the eastward position of the rain- snow line, one might expect that the storm will produce more precipitation and be colder than forecast in this particular ensemble. This would put a significant risk of winter storm category precipitation in areas that could expect to experience snow, mainly in inland areas of the Mid-Atlantic.