Precip Type Algorithm


Dan Cobb (MIC at NWS GRR) and Jeff Waldstreicher developed a top-down algorithm for determining precip-type, utilizing BUFKIT data files. A good description and full paper are available

Since I already generate Buffy profiles, I send the output to their perl script and post the results online. You'll see the 0Z/12Z NAM, 6Z/18Z NAM, GFS, and then from our local WRF (if the point is in our domain, and it's offline temporarily while I get nesting set up.)

Other stations currently in the mix (sub for "klan") include kord, kjxn, ktol, kfwa, kapn, ksbn, apx, dtx. Note that the WFO ID's do not have a 'k' in front.

If you want another location, holler at me, and as long as NCEP generates a BUFKIT sounding for that point I'll add it (and by default you'll get a faster source for your bufkit data.)

- Rob
That's the hope ;> If you click the explainer link, you can read the paper and watch a ~15 minute presentation on the topic.

The algorithm is theoretically better than saying "10:1" or using surface temps, which are the old ways of doing it. But it can't be any better than the model, so if the model is horribly wrong (obviously one of them is, for Toledo the NAM has 18", GFS has 1") then the output will be horribly wrong.
LAN is showing average ratios of 15:1 - quite a bit higher than I expected. Then again, the "ratio gradient" is probably lined up pretty tightly with the baroclinic zone. if the model is horribly wrong (obviously one of them is, for Toledo the NAM has 18", GFS has 1") then the output will be horribly wrong.

With all of the recent changes to the NAM, let's hope it's the GFS - I'd hate to think all of those changes were just a waste ;-)
Actually the NAM is going through a ton of upgrades in the near(?) future...

2b. Mesoscale Modeling Branch:
Geoff Dimego reported lots of work on solving problems with the NAM forecasts. The NAMX parallel is focusing on damping the external mode to improve balance and reduce incorrect adjustment away from the analysis state. The NAMY parallel is testing changes in shallow and deep convection, divergence damping, and diffusion. Work is ongoing to reformulate the horizontal diffusion to be effectively calculated on isobaric surfaces instead of sigma surfaces so that restricting its application in areas of steep terrain slope would not be necessary – a physically more sound and more universal fix than some of the summer crisis changes. Work is also being done to adopt the global model formulation for form drag and momentum drag, which are absent in the NMM. The Eta had form drag based on subgrid-scale terrain variance and wind direction, but NMM only had a simple dependence on terrain variance which was ultimately removed in the NAM WRF-NMM because it was causing near-surface winds to be too light over mountainous terrain. It is hoped that the global model treatment will improve trough speed and depth because mountain wave momentum drag had that effect in the global model. Also, new datasets need to be tested for implementation, including AIRS and MODIS data, along with an increase in the vertical resolution of the radiative transfer solver inside the assimilation.
Rob, I'm familiar with the Cobb technique. It actually works quite well to my knowledge. Just wondering if you know what it's showing roughly for this snow event down here in Norman.

Thanks Rob!

Wow, not counting the IP at the start 7.3". I wonder if the DGZ*omega toggle on BUFR is the Cobb technique? That would be interesting.

Last edited by a moderator:
As I understand it, the DGZ*omega is a first attempt at integrating it into BUFKIT. There is a small bug in the current release I'm using that puts snow into the accum bag when it should stay 0 (mixed.) That should be fixed next week.
For more precip type information, check out

For the NAM, we now run 5 different algorithms and then go with consensus at each point for a "dominant" precip type. On this page, for each of our station time series points, you can see what each of the algorithms is showing.

I can't add any more points on the maps, but the coverage here is pretty decent.