Comment on New GFS Forecast Model
Complacency, comfort, and reliance can breed a false sense of security. That is nature’s rule. Why do you think game wardens do not like people feeding wild animals? Because they become dependent on food being handed to them and lose the skills natural to them to make them stronger and better. Because man has become so smart, perhaps we lose sight of this sometimes. We need challenges. Here’s a simple way to put it: Unto all good there must be opposition.
When it comes to computer modeling, we are so reliant on it that we expect it to be the answer, not the tool. Because of that, we love when it shows something we like and we hate it when it doesn’t. The worst is when it shows something only to take it away. People go out of their minds.
Since I started looking at the European model (ECMWF) back in the ‘80s, I have adopted an attitude that if it doesn’t see a storm, be skeptical of what the U.S. GFS model — which is in the process of being upgraded, criticism over its level of accuracy notwithstanding — shows. I am sure everyone can point out times when the U.S. model scores a coup, but by and large the Euro does a better job. All of us pay for U.S. modeling, which is developed by way of money for research. And where does that come from? Taxpayers. How much do we pay? Well, if we look at the entire modeling suite and add up how many years and how much money was put into it, we can get an approximation.
But let’s put it in perspective. It’s hypothetical, but I think you will get the point how lucky we actually are and how cost efficient it is. By some rough calculations, the modeling part of your taxes is perhaps a dollar or two a year. While not perfect, U.S. models are certainly worth a dollar or two a year.
But are they the best?
The idea above says that for the good of the country we all pay a couple of dollars a year to see U.S. models. But to see the Euro, at least the longer-term stuff, it cost much, much more. There are a lot of companies in the U.S. that pay to get the Euro. Now, if it was not better, why would they do that?
Obviously, the market of free thought says it has enough merit to warrant paying a certain price for. The market had decided it is superior to the GFS model, but part of the frustration with the model given its tiny cost is that we have lost our reliance to some extent and have grown comfortable with outsourcing our answers to the very tools that we are supposed to use to get the answer.
We have to understand that models are not the answer. There are so many events that can happen even a week out that can send the best model spinning into a frenzy of futility.
It is certainly good that our government tries to advance computer modeling. This is the United States of America. We are supposed to lead. And it has to be a thorn in NOAA’s side that so many people love the European over the GFS. But I think the adjustments being made miss a factor here.
“Systemic” is a word I have been using since the 1990s when studying the error patterns of U.S. models.
I believe the GFS cannot handle heat, which causes a feedback problem.
Let me go to one glaring example of where it has a problem with what is a huge source region for the planetary weather: the Madden–Julian Oscillation (MJO), which, according to Wikipedia, “is characterized by an eastward progression of large regions of both enhanced and suppressed tropical rainfall, observed mainly over the Indian and Pacific Ocean.”
Think about the amount of energy this very important feature releases into the atmosphere given the very nature of where it occurs. We showed before how the increase in convection in the eastern Indian Ocean in early December set off the warm latter half of December for the U.S. But the major amplitude of the wave also set into motion stratospheric warming that ballooned over the North Pole and set the stage for the late January “Coldmegeddon” outbreak. The sensible weather over the U.S. had its origin in an event that preceded the reaction by two to six weeks, but the reaction was very real both ways (warm and cold). It was no accident.
Here’s the huge problem. Now that you can see what the MJO can do, what happens when we have a model that can’t even forecast it? This is where the secret to the systemic error lies, in my opinion.
Here was the GFS model 15 days ago:
It’s running wild into phase 7.
Here was the Euro for the same time period:
Fifteen days later, the MJO followed almost verbatim the Euro idea. It did not go wild through phase 7.
Here’s the point: If the model has so much trouble with this, it hint as to what the systemic error is — a basic flaw in its ability to handle heat and convective feedback. That is an ugly flaw. But the question then becomes, do you spend even more money correcting the flaw in the foundation, or do you a) just keep upgrading around it and hope that overwhelms the flaw, or b) simply say the Euro is better?
If you want to find the answer, the first thing I would look at is why the GFS does what it does with the MJO. Start with the tropics. There is so much energy there, which is the key.
Let’s sum this up.
U.S. models are well worth what we pay for them. They are pretty good. We are likely disappointed because they update so often we can always look at them and see many examples of an outstretched hand with free candy that is yanked away on subsequent runs. And there is always the “danger” it scores the coup.
NOAA, by and large, continues to move the ball forward. And NOAA, by and large, is a huge benefit to all of us that love the weather. The price we pay is tiny compared to the knowledge gained. If you spend a lot of time researching the past, what NOAA has assembled in that aspect is enough. It is greatly valued by this forecaster.
Systemic changes take more than just upgrading the system. It is like saying I have to work my calves harder in bodybuilding. If you don’t have the genetics, they are going nowhere. A deep dive into the GFS model and comparison vs. the Euro at a foundational level is needed. If the problem with the MJO is solved, I suspect the GFS will close the gap. That is where I would go and try to see what I could find.
Finally, you can’t always get what you want, but if you try sometimes you get what you need. Maybe we all need to calm down about the models.
Joe Bastardi, a pioneer in extreme weather and long-range forecasting, is a contributor to The Patriot Post on environmental issues. He is the author of “The Climate Chronicles: Inconvenient Revelations You Won’t Hear From Al Gore — and Others.”