Wednesday, February 14, 2018

GEFS Plumes And Grid Points Used For Tucson

Question re this morning's post: Was there a consensus reached on the location or grid size used for the GEFS plumes? I enjoy looking at them to determine model-to-model variance but I don't find them to be representative of much more than that.

About a year ago we (myself and several readers) tried to determine where the 4 closest GFS grid points were relative to the airport. Presumably the four closest grid point forecasts are interpolated to yield the point forecasts for the airport. This proved to be a very frustrating exercise, since NWS technical documentation is sadly lacking important details, or much out-of-date. Mike Crimmins came up with the grid points shown above. However, when I examine them closely, they appear to be on the old GFS 27 km grid - current grid is 13 km. So I assume that the points that are interpolated to give the forecast for TUS are considerably closer to the airport than shown here. I just have been searching to try to find more detail - but alas just end up frustrated. It is important to remember that on a 13 km grid, the terrain is still grossly smoothed, so  that the real-world up and down slope flows are not accurately represented.

The GEFS plumes for QPF are the most unreliable forecasts, while other parameters (e.g., temperature, wind etc.) are more reliable. All of this makes it important to consider the 1.8 km grid forecasts from the Atmo WRF versions (even though these are deterministic forecasts driven by single model forecasts, as per the GFS version, the terrain effects are more accurately captured). The GEFS plume forecasts for the entire country can be found at:

A number of more progressive NWS offices use ensemble model forecasts to predict QPF amounts probabilistically. Typically these forecasts indicate the likelihood of different precipitation thresholds occurring - an example would be 20% probability of rainfall greater than inch, etc. etc.

Regardless, it remains unclear to me exactly how the GEFS plumes are derived for any given point. If someone can direct me to links that explain the details of the process, I would certainly appreciate it.


  1. Anonymous9:55 AM

    Thanks for that response!

  2. I don't know about the GEFS plumes product, but here's a quick comparison of raw precip data from 4 GEFS members over AZ for a 6 hour period yesterday evening:

    The grid lines are 1 degree.