Learning Lessons

Nice timeline over at Obsidian Wings on these events leading up to the Flooding of New Orleans. As everyone moves into CYA mode, this timeline helps set us all straight.

I would like to vent about a particular problem at the very top of the chain of events, namely the National Hurricane Center (NHC) forecasts of Katrina's path. It seems to me that NHC has settled into its own CYA position as a standard operating procedure: instead of issuing the best forecast possible on a hurricane's path, NHC waits to first get consensus from its suite of computer models, some of which are better than others at forecasting hurricane movements.

Delay at issuing the forecast means delay in political circles in issuing evacuation orders. It is a grave responsibility to issue a mandatory evacuation of a major city, and it shouldn't be the least bit surprising that Mayor Nagin hesitated at doing so. I don't think that NHC's delay at sounding the alarm should be excused, however.

About a year ago, I started doing hurricane path forecasts for friends who retired to Tampa, FL. I'm an air pollution meteorologist by trade, but it was fun to be doing weather forecasting again, to keep my skills sharp. I preferred to rely on the NOGAPS model for hurricane path forecasts: it's a baroclinic model that did very well with Ivan's path, and rarely blunders like some of the others.

I quickly noticed, particularly with last year's Hurricane Ivan, that NHC's preference for computer-model consensus meant delay when the models' opinions diverged. Each model's opinion was weighed equally, but some models were big-picture models, unsuited for fine-scale detail, while others were barotropic models, which simplify information regarding temperature gradients - information that is of crucial importance in the timing of the movement of the mid-latitude systems that often steer hurricanes. Baroclinic models like NOGAPS don't make these simplifications. Getting consensus from all models usually means delay, much like getting consensus from Congress on any problem-of-the day usually means delay.

The dirty little secret is that I don't think NHC minds delay that much: they would prefer to keep everyone in a state of mild apprehension about an approaching storm rather than risk an overly-specific forecast, and therefore lull others nearby that might be affected into unwarranted complacency. Chicken Little is NHC's CYA standard-mode-of-operation. And it's easy to see why this would be the case - it's horrible to be completely taken by surprise by a hurricane, and to be stuck with the guilt of having failed to make the alarm.

With Katrina, the NOGAPS model tipped off, or SHOULD have tipped off, meteorologists that Katrina was heading towards Biloxi as early as the morning of Friday, August 26th. Model consensus wasn't reached until Friday evening, however (I blogged about the process with posts here). It can make, and here it apparently did make, a huge difference to lose a weekday's daylight hours in preparation for the storm.

NHC needs to weed its basket of models used in making hurricane path forecasts. At a minimum, the barotropic models need to go.

As I say, I very much sympathize with Mayor Nagin if he delayed in ordering a mandatory evacuation. People can die in accidents in a mandatory evacuation. But I do think he should have gotten better weather forecasts: it might have provided an extra edge and saved some lives here.