If you live in Florida for any length of time, you learn something about hurricanes. I am far from an expert.
Updated at 6:46</b – Philosophy now seems inappropriate. The tracks have shifted, and it now appears Irma will move up the spine of Florida, spreading a 100 mile wide path of devastation.
Consider the scope of this storm
- There is an evacuation for everyone east of I-95 in Savannah
- 200 miles sw, schools are closed in Tampa and St Pete and there is a mandatory evacuation for people living in flood plains.
- Two nuclear power plants have announced they are shutting down.
- There is a mandatory evacuation in the Keys.
- There are traffic jams on both I-75 and I-95.
- Gas is running out – people want to leave but worry they will not find a second fillup.
Still becoming familiar with them brings certain things into high relief. The computer models built to model hurricanes are probably the most sophisticated in existence. They run on the fastest super-computers in the world, and they take hours to run. In a way they represent the best attempt to model reality in its complexity.
And yet those who rely on them are quick to point out how unreliable they are.
Here are the two best. And they suggest a disaster of epic proportions may be in the cards for Miami.
GFS, one of the two best models, showing a strong Category 5 near Miami. A small change in direction will have enormous consequences. pic.twitter.com/6yqh3ncZhW
— dcg1114 (@dcg1114) September 7, 2017
And here is the best, the Euro
The Euro, the best model, shows a category 4 making landfall. pic.twitter.com/UxGVdHpUbT
— dcg1114 (@dcg1114) September 7, 2017
And yet in its analysis, the National Hurricane Center doesn’t talk about how good its models are. It emphasizes uncertainty.
As a data nerd, I am struck by how the NHC highlights uncertainty. Political pundits do the opposite for the most part. pic.twitter.com/7Hfv8pnQoI
— dcg1114 (@dcg1114) September 7, 2017
One can argue, I suppose, that politics should be more predictable than Hurricanes. But I am not sure why would people think that. Political predictions suck – and they always have. What I find refreshing about the NHC is their honesty and humility. Over and over again they stress how much they DON’T know.
No similar instincts exist in political punditry. In fact, there is virtually no accountability. Sam Wang was wrong in 2004 and 2014, and yet people cited his absurd odds over and over again.
Virtually nobody was right about 2016 in any predictive way. But that doesn’t stop the same people from suggesting they can predict the future.
Few economists predicted the 2007 recession with great accuracy, and a number that did haven’t been right since.
Why is this important? Because to some extent these predictions define what we think is possible. Clinton cannot be beaten, so we should not try.
Trump cannot win, so we don’t really need to protect the blue wall.
The Great Depression won’t happen again, so we can loosen the rules on the banks.
So I look back and wonder at the cost of this absurd confidence.
It is good thing for all the NHC is immune.
I think a lot of what you have to say (on the political side) derives from the popular confusion of predictions and projections. We all know that a projection is based on existing data and nothing else, but a prediction while based on existing data may also attempt to model external factors that affect the data’s applicability. Yet, way too many attempt to use projections as predictions, and it is even easier to make ever-finer projections in this era of big data.
We assumed the blue wall because it existed the last 2 elections (though it almost didnt exist 3 elections back in 2004). We assume that the close districts in 2016 are the only ones to worry about in 2018, instead of recognizing that a different message can change the competitive districts (though that would need us to worry about issues instead of just adding up demographics).
In short, we need more thought to temper what crunching of big data gives us. For politics at least. And major realignment (which is desparately needed) doesn’t occur by projecting based on the last election.
Data wrongly used leads to failures of imagination.
You can see this in the markets. People systematically do not evaluate the risk of extreme events properly. People have gotten very rich buying margins way out of the money because of this.
The problem is that data and prediction are turned into barriers to change. I think that happened with respect to the the 2016 Primaries.
My take on this subject: Political predictions are based on variables that (1.) involve measuring instruments that are not as reliable and valid as meteorological instruments; and (2.) involve responses from human beings that are not as dependable as weather response variables.
Found this interesting article regarding Miami’s problems due to its topography. Disturbing read at this moment in time. Some commenters may have already read it, but I’m posting it for others. It was written 4 years ago.
http://www.rollingstone.com/politics/news/why-the-city-of-miami-is-doomed-to-drown-20130620
“But South Florida is uniquely screwed, in part because about 75 percent of the 5.5 million people in South Florida live along the coast. And unlike many cities, where the wealth congregates in the hills, southern Florida’s most valuable real estate is right on the water. The Organization for Economic Co-operation and Development lists Miami as the number-one most vulnerable city worldwide in terms of property damage, with more than $416 billion in assets at risk to storm-related flooding and sea-level rise.”
Yes. Helpful to know the difference between inferential and descriptive statistics.
Right now the track is your basic worst case. It lands over the Okefenokee swamp 30 miles west of Miami. The power is on the right, so this will drive storm surge and hit Miami with 170 MPH winds.
And there is no natural barrier for the wind.
But as was seen in Katrina, there is enough water in the Okefenokee to provide fuel for a hurricane, and it won’t weaken Irma. Irma then moves north and after TWELVE HOURS inland it still has 120 MPH winds.
Florida needs a miracle. An eye wall replacement cycle just before it makes land fall. Something.
I won’t mention that models show Jose pulling a loop in the Atlantic and hitting key west as a category 3.
Meant the Everglades above.
Assigning predictive performance grades for meteorological projections and political polling before the final outcomes are known is a good way to bias expectations.
Looks as if the US models have relied too much on the historical data of powerful storms that developed in the east Atlantic basin and too little on the unique characteristics of Irma. (So American.) But the European models have also been off. None had the right hand hook (north) as late (and far north) and as far west in their early models, and this hook is still a projection.
Something else (if Irma, Jose, and Katia aren’t enough):
NOAA G3 Watch: 7 through 9 September, 2017 due to CME Effects”
The variation in the data in the 2016 election was very broad; I commented on it at the time.
And what people don’t get is that a probability of 70% that one candidate will win the election means nothing but “luck” will fall 3 times for one candidate and 7 times for another if the same conditions are repeated (which, of course they aren’t in elections). Pollsters act like there is comparability between elections because most times it works good enough.
The National Hurricane projection cone widens as the forecast time increases. That big circle at the end is the expected area within which the eye will appear on that date and time. If the eye appears within that circle, the NHS has been successful in its projection.
What too many people cannot afford now that they could before National Hurricane Center forecasts were computerized is a “false alarm” evacuation in which an area is spared, people lose days of work, and businesses lose operating days. That cost of a false alarm was much lower in lost income years ago. And it affected much fewer people because the coasts had not been so developed into urban forms. Cheaply constructed cottages were easily rebuilt.
Evacuation is a mess. Left Tamapa area Thur. and it took 8 hours to drive to Tallahassee. There was no way to get a plane out of southern fl as all the tourists were in the air ports trying to get out. The airlines were busy cancelling flights as they don’t fly empty planes to evacuation areas. Currently just south of Montgomery Al and will wait out storm here. Wish me all the good stuff you can think of.
I am a firm believer in running from hurricanes. However, when you do that you should do so knowing that maybe 3 of 4 times it will turn out that you would have beed fine.
When she left it looked like it was going up the east coast. Now it looks like the west coast. She got an airbnb place in Tennessee and will spend the weekend relaxing.
There is no way to evacuate Florida. There are basically two highways out of the state. There are a couple of other paths (301, 41) but they are not interstates.
Evacuating millions of people within a few days isn’t easy anywhere. Back in Hurricane Donna’s day, the population of Florida was 5 million; today it’s 20.6 million.
Here’s the track modeling from two days ago:
While GFS has continuously had Irma tracking further east than the euro model, both have been too far east in comparison with where they now have it at T-24.