Thursday, March 07, 2013

The Standard Self-Reflection & Where Meteorology Goes On

Yesterday's storm verified well from a wind standpoint, reasonably well from a coastal flooding standpoint as we received moderate tidal flooding along the Shore last night and early this morning. However, from a precipitation standpoint it verified horribly.

With this storm, the global models (GFS and EURO) performed better than the mesoscale models (NAM and SREF) in terms of their handling on the storm.  For Philadelphia, the 24 hours ending at 1 AM produced a whopping 0.13" of precipitation.  The "best" performing model for Philly was the GFS, which produced 0.50" of precipitation out of the storm, with the Euro coming in second at 0.59".  Neither did very well but they both performed better than the aggressive NAM, which spit out 1.16" over 24 hours.

Washington's best handled model (and that's a stretch because they all wanted to produce healthy to excessive snowfall) was the Euro, at 1.09" versus 1.00" in liquid reality.  Granted, the snow equation in DC is where the models really screwed the pooch as the GFS and NAM both wanted to produce anywhere from 6 to 12 inches of snow in various model runs, with the Euro gradually nudging itself into the six inch category for DC in Tuesday night's run.

Credit to Ray Martin (NWS) for data assembly.

Up in NYC, the Euro was less awful although all three models wanted to spit out a decent amount of precipitation...again, the NAM the most aggressive of the three.

A few lessons out of yesterday's "storm" --

1) The NAM.  It sucked last month with the February storm that, in a model run a day out, tried to bury Bucks County with 10" of snow.  It kept trying to produce six to eight inches of snow across South Jersey and Delaware -- even into last night's computer model run.  Neither of those occurred. There are some benefits to the NAM in the summer with thunderstorm development (it can sniff out locations where storms may fire up) and in the winter with temperature profiles around here (sniffing out cold damming, warm pushes up the coast).  Other than that, it's a garbage computer model.  We didn't rely on it for this storm. You shouldn't either...and neither should the meteorological community for the most part.  I mentioned the idea of having the model de-funded over twitter yesterday and using its resources to improve the GFS.  I still think that having one strong US computer model would serve forecasters and the public better than having two models that struggle with various issues.  Putting the NAM's bandwidth and financial resources into strengthening the GFS won't make the GFS superior but it may help improve it further.

If you need a reminder of its "wonder", here's the snowfall forecast it projected for the region on Tuesday afternoon.

2) The state of computer modeling overall.  Earlier today, Matt referenced Cliff Mass' points about putting more money into modeling and improving our weather prediction.  I agree with it at the surface but it's fair to point out that the European model whiffed on this storm for many locations as well.  The science to get weather right is still far removed from where we need it to be. For Washington, this busted forecast is arguably worse for them than the March 5th, 2001 storm was for us in the sense that it busted as the storm occurred, not in the hours leading up to it.  At least in the Bolaris storm, the models trended away at the last minute from a snowstorm of the century.  Here, they did not in Washington's case.  Modeling even into yesterday morning was projecting six inches or more of snow down there.  When busts like this occur, it's clear the science is not where it needs to be.  Resources should be put into improving weather modeling...whether it's combining the NAM's bandwidth into the GFS or more money...or both...but having two imperfect models isn't going to cut it.

3) Getting a good understanding of WHY things went wrong on the precipitation and snowfall side. The coming days will provide plenty of armchair quarterbacking from the meteorological community.  For us, the strength of the blocking high over Quebec shearing the northern edge of the storm apart and preventing some of the steadier and heavier precipitation from moving in is one of the more likely explanation for this storm failing to launch.  For DC, the intensity of the precipitation wasn't as heavy as predicted at key times, combined with the lack of cooling in temperatures at the surface to finish the transition from rain to snow.  That extra bit of warmth hurt DC as well.  There are other reasons that will likely surface but those two start the conversation in my opinion.

For a snowlover, this storm was looked upon as a failure in Washington, Baltimore, the Philly metro, and for the NAM computer model.  However, not everything failed.  For the Shore and Beaches, where winds gusted to 71 mph in the Delaware Bay near Lewes and to 67 mph on land, the combination of wind and coastal surge resulted in moderate tidal flooding.  It was bad down there and the calling card for the storm for those guys were exactly these two factors.  Parts of Virginia and West Virginia got that foot plus of snow.  Parts of the Appalachians in Pennsylvania picked up ten inches of snow.  This is where the models got it right for the most part.

In the end, the science of getting it right is still a ways off and that's what makes meteorology fun, harrowing, and a pain in the rear all in one.  This storm serves as a great example of how we need a perfect set of ingredients in many cases to get the desired outcome...and if one or two of those ingredients are missing, you're going to get the result you don't want.  It also serves as a reminder to the meteorological community that the science isn't where it needs to be...and that we need to improve it to improve our outcomes.