U.S. Hurricane Forecast Models

As is becoming more obvious every day, the US weather models are being outdone by the European model by a wide margin. The link below details some of the reasons. What gets me is the statement regarding a update to the GFS model.

“The loss of short- to medium-range [tropical cyclone] track and intensity forecast skill for the Atlantic basin in the proposed 2017 GFS is unacceptable to the National Hurricane Center.” Ultimately, the upgrade was initiated anyway.

The new model is less accurate than the old one. My question is - who makes such a decision in the face of the experts saying “No!” and why?

If mods think this is more suitable elsewhere…

A couple of quotes from the NOAA site:

I find this odd, but also important. I find it odd that they have only just got a coupled model for hurricanes working. But important in that this is what is needed to get accurate models. I suspect the problem they are having is that they have not got it tuned properly, and as it is a relatively new code and they need to work on it more.

The UK Met Bureau cost about three times the NOAA computer, and is at least three times the capacity. That said, the UK system is not run as a single lump.

Overall however, it is clear, the Europeans fund their weather at a much higher level, and they are more advanced. So the results are no surprise. I don’t doubt that NOAA will get their model sorted, and it will start to deliver on its promise. So to judge i now is premature.

For one thing, the upgrades were all done and being ran in test mode when the NHC raised the alarm. That means if management listened to them, they would have just wasted $25 million. Plus, the GFS has always lagged when it comes to hurricanes. I imagine they had two options, scrap the upgrade (or go back and spend money they didn’t have to fix it), or decided the improvements in other areas were worth it. I’m just guessing though.

That’s for the 2016 upgrade to the HWRF, not the 2017 GFS upgrade.

think of the future.

Coupled models are the chosen method of improving met (and oceanographic where I have some experience) models well in to the future. While other methods (say improvements to algorithms or faster computers running higher resolution models) might have been chosen, this was the method the community chose after a lot of consideration. Note that better algorithms and faster computers are still necessary, but coupling offers a clear path to significant new science and improvements.

There is a downside. Think of a car. You can improve a traditional car through small tweaks, or you can build yourself a hybrid. The hybrid offers greater long-term improvements at the cost of short-term complexity and problems.

You might ask why are we using a cutting-edge troublesome model to make public predictions? why not use the old models until the new models work as well/better? They did. Coupled models have been running for 10 years. The models you see now are ones that have been tested and shown to be better than the models of 10 years ago. (This is a simplification of a more complex story. As I indicated I am not a model expert and have not worked with met models. This story is based on my understanding of how things worked out in the oceanographic modeling community-which is quite similar to the met. modeling community.) As you noticed they may not yet be as good as today’s European models, but that wasn’t the design goal. As for why our older stand-alone models weren’t upgraded as the European ones were-I bet you can guess… $ NOAA and the research community have enough money to upgrade one set of models. Europe has enough money to upgrade both-for a while.

The NHC will be looking at this exhaustively during the off-season … every year they go through every forecast, every statement, every warning for the whole season and evaluate their effectiveness … and during this process they will look very closely at the situation the author of that piece brings up, as well as a few other “misses” the NHC has already had this year … certainly this new and not-quite-improved GFS model will be examined to a nauseating degree … I got to tell y’all that these folks down there are really busy at the moment and perhaps now is not the best time to be re-coding their models … just saying …

The NHC is using this European model, so please don’t think that the forecasts are somehow corrupted by some bogus software … whatever the problems, they are being minimized in the ensemble products which the NHC tends to rely on … public safety is not being compromised …

I read the following at Aljazeera. Does it agree with other analyses?

well, it simplifies part of the situation. The fuel for a hurricane is the total heat content of the upper ocean. That is usually calculated down to 100 meters (going from memory here). When the total heat content is above a threshold, the hurricane intensities. Usually the SST is a good measure of the total heat content, but remember it is based on measurements from the top 1/10 of a millimeter of the water column (IR based satellites), corrected to estimate the SST of the upper 1 meter. From that there are rules of thumb for estimating the total heat content. As you can imagine, if it was recently a cloudy or windy day, the measured SST will be lower than normal. The SST measurements try to account for such things, but they aren’t always correct. So talking about the SST is shorthand. The correct estimate, the ones the models use is total heat content.

I have no expertise regarding the dust. It certainly is a factor.

Katrina was a famous example of SST causing a problem. As Katrina approached the coast, it ran over an undetected subsurface warm core eddy from the Loop Current. There was no surface SST expression of the eddy. It was seen on altimetry measurements but at the time those weren’t used in the operational met models. That gave the hurricane a sharp boost just offshore which raised the wind speed to Cat 5 for a few hours. As it passed the eddy, the wind speed dropped back down. But the higher winds raised the storm surge bulge. That didn’t have time to relax before the storm came ashore. It came ashore in the right place and at the expected wind strength, but with an unexpected storm surge. That is one reason cars were washed out of the second level of parking garages along the coast.

I’m curious about where they got their information from …

Anyway, I yarded up the first mention of the tropical wave that eventually formed into Irma “NHC Graphical Outlook Archive - Aug 27th, 8pm EDT” … and here’s the portion of the discussion “Tropical Weather Discussion - 736 PM EDT Sun Aug 27 2017”

Emphasis mine

I’ll not bore everyone with anymore details but the Saharan dust and associated dry air figures prominently in these discussions about the Eastern Atlantic … so the quote form Aljazeera does agree with NHC analysis …

ETA: Solar powered electric fans pumping Saharan dust into the tropical flow …