Here comes an AI that can predict hurricane strength. Don't worry, NASA made it so it probably actually works

Scientists at NASA’s Jet Propulsion Laboratory are turning to machine-learning models to predict the severity of incoming hurricanes. The goal isn’t to forecast whether a storm will hit a particular region or not, it’s to assess how intense a storm might become over the course of 24 hours. Hurricanes are graded across five …

[Version 1.0]

Predictions are useful

But they are just predictions, I've been watching hurricanes for years and survived a couple of them, in general because their centers missed me, I drove through Katrina as it was coming ashore and arrived home just in time to get a tree through the bathroom ceiling ... so I made a cup of coffee and waited ... it sailed past and we spent a month cleaning up. The recent hurricane Laura has devastated Lake Charles, this was anticipated but the interesting thing is that Laura stayed at hurricane strength much further inland than any hurricane has done for years now - a swath of Louisiana will not get power or water until the end of the month if we're lucky. That was not predicted - I'm definitely in favour of improving predictions but we have to remember that they are just predictions based on updated dropsondes I guess, not actual facts.

Weather predictions are affected by what the general population thinks of them - if the prediction is for a very rainy day and we get a Derecho then everyone is pissed at the weather reporters and runs around screaming and wanting lawsuites, but if they are warned that at Derecho is coming and all we get is six inches of rain then everyone "knows" that they are smarter than the weather reporters and everyone is happy.

[Anonymous Coward]

Re: Predictions are useful

And we do have the alternate model of having a president that predicts hurricane movements with such precision that w change forecasts to accommodate him

[Charlie Clark]

Re: Predictions are useful

To be picky, they're forecasts about what might happen rather than predictions about what will. But, of course, as with most things statistical most (sic) of us can't reliably tell the difference.

Anyway, nice to see Watson getting some good press. As one of the pioneering ML systems it's taken some bashing but there has always been some very good technology in there and, just as important, some great people working on it.

[Version 1.0]

Re: Predictions are useful

The TropicalTidbits blog site discusses the possibilities of all storms in the gulf - it's well worth visiting and listening to the descriptions when a storm is likely to appear.

[Version 1.0]

Re: Predictions are useful

OK, I realize that six inches of rain is a lot in the UK but in Louisiana it's not a big deal in most areas. Our worst flood from rain was a few years ago when we got 48 inches in one week.

[Yet Another Anonymous coward]

Re: Predictions are useful

It's your own fault for not going metric - then you would only have around 1.2M

[Charlie Clark]

Re: Predictions are useful

The impact depends a lot on the geography and the usual rainfall patterns. Flat areas can absorb a lot of water by flooding extensively, but then take much longer to drain, especially if they are flood plains, and thus below the level of local rivers. In much of the western UK rain is so common (300 days per year) that it's difficult to get that much extra – though flash floods over the last few years have shown it is possible – and local systems can normally cope quite well with a reasonable excesses. But it's also just so much smaller than the US.

In the US hurricane alley you have flood plains combining with subsidence due to drainage, and rising and warming sea levels, which is why the Army Corps of Engineers has already had to revise down the viability of the post-Katrina works. At some point, some of those coastal settlements and flood plains will need to be abandoned.

[O RLY]

Re: Predictions are useful

"this was anticipated but the interesting thing is that Laura stayed at hurricane strength much further inland than any hurricane has done for years now - a swath of Louisiana will not get power or water until the end of the month if we're lucky. That was not predicted"

Wrong. That was predicted. See Advisory 24, among others, here: https://www.nhc.noaa.gov/archive/2020/LAURA_graphics.php?product=5day_cone_with_line

It is also common for major hurricanes to stay hurricane strength well ashore when moving as quickly as Laura. Major hurricanes making US landfall is what is not common, but when they do, they stay hurricanes for awhile. Camille in 1969 made it almost to Tennessee as a hurricane after coming ashore at the Alabama coast as a Cat 5.

[Ben Trabetere]

Re: Predictions are useful

>Camille in 1969 made it almost to Tennessee as a hurricane after coming ashore at the Alabama coast as a Cat 5.

Slight correction is in order - the eye for Camille passed over Pass Christian, MS. Katrina made landfall in the same area roughly 30-years later. Both Camille and Katrina touched the coast of Alabama and the Florida Panhandle hard, Bay St. Louis and Pass Christian took the full force.

Fun facts about Camille that rarely get mentioned

At Woodstock, The Band was finishing its set as Camille came ashore.

The official peak wind speed is 200mph, although some reports have it as "in excess of 200mph." The anemometers only went to 200mph, and they were destroyed in the landfall area.

The path took it through Kentucky, West Virginia, and Virginia. It entered the Atlantic Ocean and started to strengthen. It almost regained hurricane status.

Although Camille is classified as a Category 5 Hurricane, the Saffir–Simpson scale did not exist at the time.

Related to predicting hurricane strength, the local meteorologists still would be unable to forecast Camille, 50-years after the fact. Not even if Nash Roberts was whispering in their ear and they could use the interwebs.

[Pascal Monett]

The Earth is a complex system

And we are just beginning to understand how complex it is.

Now seems to be the right time to link to this.

Watch and be amazed. You might learn something.

[Valeyard]

oh no

According to Gus Gorman (legendary hacker) if you hack the weather satellites you can actually cause the storms instead

[Chris Gray 1]

in the news

Combining recent articles....

I wanna know what happens if you drop Tsar Bomba in the middle of a big hurricane. We demand proper research!

[herman]

Wow, Watson is finally doing something slightly useful?

[Tom Paine]

"AI"? What Reg said

(when Brian asked of he was on the JPF)

A lot of very clever meteorologists and modellers have spent a huge amount of time on the major models (and indeed the kinornones); ECMWF, GFS, UKMet, HWRF and the others that do a bit less well. The idea that ""AI"" will provide the magic pixie dust that can accurately predict RI 60h ahead or get track error down to 50 miles at 120 hours is ... Fantasy.

* GIYF

[Grunchy]

Chalk a win…?

Chalk a win for Watson? You mean, the Watson that defeated Ken Jennings?