1
$\begingroup$

I'm referring to an article I've read on Ars Technica about the differences in various weather models.
One thing that stood out for me was a mention of

Forecast models typically show their skill with three-, four-, and five-day forecasts. For simplicity's sake, we will focus on 120-hour forecasts. At this lead time, the average error of the European model with respect to Irma has been about 175km in its position forecast. The next best forecast is from the hurricane center, which is slightly more than 300km. An automated model, then, has so far beaten human forecasters at the National Hurricane Center (looking at all of this model data) by a wide margin.

A similar thing about models beating human forecasters was also mention in another article linked in this article and nowhere is there any other explanation for it, so:

Do humans seriously forecast extreme weather?
I get it, that meoteorologists get trained in physics, hydrodynamics etc, but I can't believe that humans try to compete with computers in crunching the numbers on differential equations. Or how else does 'human forecasting' work?

$\endgroup$
  • $\begingroup$ Too short to be an answer: Humans should be better than any one model because humans have all of the models at hand, plus information that various models do not take as inputs. $\endgroup$ – David Hammen Sep 10 '17 at 7:07
  • $\begingroup$ If you take humans as model users, then of course they should be better. Just the statement from the article seems to try to compare them directly as if they were equals, whereas what you say would imply a hierarchical relationship and thus they couldn't be compared. $\endgroup$ – AtmosphericPrisonEscape Sep 10 '17 at 9:16
  • $\begingroup$ You misread. Human weather forecasters have been using multiple computer models for decades. $\endgroup$ – David Hammen Sep 10 '17 at 10:03
  • 1
    $\begingroup$ What exactly is meant then, with the quote i gave? Particularly "An automated model, then, has so far beaten human forecasters at the National Hurricane Center [...] by a wide margin. " It seems hard to misread that. The author compares them directly. $\endgroup$ – AtmosphericPrisonEscape Sep 10 '17 at 10:07
  • 1
    $\begingroup$ It is humans, not machines, who make the official forecasts made by the National Hurricane Center. Those humans use their best judgment on if and how to combine the results of conflicting models. $\endgroup$ – David Hammen Sep 10 '17 at 14:30
1
$\begingroup$

The forecasts issued by the National Hurricane Center are made by people. Those people use every single tool they have at hand. This includes forecasts made by automated tools such as the European Centre for Medium-Range Weather Forecasts (the "European model"). It also includes forecasts made by several other models. The human-generated forecasts account for the skills of those various models, for the ages of the forecasts made by those models, and for data that are not used by those models.

When multiple models more or less agree, the human forecasters use techniques and their judgement to split the difference. There is no splitting the difference when models disagree massively, as currently is the case with Jose. The current forecast (advisory 20A at 0800 AST on 10 September 2017), depicted graphically below, appears to be exactly that made by the ECMRWF.

Forecast for Hurricane Jose, issued 10 September 2017 at 0800 AST. In this forecast, Jose starts heading away from land but makes a loop and starts heading back toward land near the end of the week.

$\endgroup$
  • $\begingroup$ To lend support to @David. There's an old computer science adage that is still applicable in all computer modelling whether it be weather, financial, economic or whatever modelling: GIGO - Garbage In, Garbage Out. The results of all computer modelling rely on the algorithms used, what is included in the model, what is excluded from the model & assumptions. All computer models are just tools that help humans make decisions. For one situation, the results from all models may agree & for another, the results may differ vastly. Artificial intelligence has not yet advanced to make such decisions. $\endgroup$ – Fred Sep 11 '17 at 3:20
  • $\begingroup$ I agree with @DavidHammen. To conclude what the quote in the question was saying: even though the hurricane center had the information of that particular model, by trusting its results only partly, they weren't able to make a better forecast than that model. $\endgroup$ – Communisty Sep 11 '17 at 7:18
  • $\begingroup$ US forecasters are also not always interested in 100% accuracy on path. They are most interested in minimizing loss of life. That means that given two relatively equal likely paths, they will choose the one that would put the most people at risk and call that the predicted path in order to have the most effective preparedness plan implemented. $\endgroup$ – dlb Sep 11 '17 at 19:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.