Model QPF Performance 4/14/12

[Mike Hardiman]

My memory might be a little blurry -- but did anyone else notice how abyssmal the NAM/GFS (maybe others) handled the QPF forecasts for Saturday over the Plains. If I recall correctly, several models/runs showed NO initiation over most of the Plains until late Saturday night at best, despite indicating relatively little CINH and modest/weak large scale forcing. If nothing else, it seems things initiated earlier than suggested -- with widespread WAA elevated storms over Western Kansas before Noon becoming surface based in time. Where I live in the Desert Southwest, Model QPF skill is typically very poor, but this seems unusual over the Plains, especially for a Day 1-2 forecast.

 

[Tim Supinie]

Well, the NAM and GFS are going to rely on convective parameterizations for their QPF for these sorts of events. There's problem #1, because first of all, the NAM's convective parameterization is designed for tropical environments, and therefore has trouble breaking out precipitation in capped environments sometimes. I'm not sure if the GFS's scheme has any similar issues.

Now that we've covered the "will there be rain?" question, now let's cover the "how much?" question. The GFS QPF really shouldn't be trusted for anything much more than location because the grid spacing works out to about 40 km, and we know that precipitation amounts vary on scales that are a lot smaller than that. The 12 km grid spacing of the NAM is a bit better in this respect. But both come back to the fact that the model isn't explicitly predicting the amount of rain falling out of the cloud, so you have to be careful trusting the model QPF, especially the GFS.