Monday, September 11, 2017
Four Timely Lessons In Probability From Hurricane Irma's Strike On Florida
One of the earlier "cones of probability" - based on data available at that time, had Irma plowing up the "gut" of Florida. The most likely path then had Miami and the FL east coast getting the brunt of the storm. Too many didn't process it was a 'likely path' - not engraved in stone.
A Category 5 hurricane originally projected to hit Miami causing hundreds of thousands to evacuate then changing track to put Florida west coast cities under threat - and ordered evacuations. People hunkered in the Miami shelters, meanwhile, and simmering over being there when "it wasn't needed any longer". Dopes like Rush Limbaugh confecting conspiracy ideations about the "hysterical" media coverage and then tying it to advertisers' making money on bottled water.
All of which points to the need for a number of lessons to do with changing probabilities as applied to stochastic phenomena like hurricanes.
One of the most confounding, clueless comments read on a major media site appeared in The Financial Times three days ago, posted by someone using the pseudonym "Bajan Boy". He posted his comment after Irma had altered its course to a strike on the Florida west coast, and noted "It is nuts and hysterical to move or evacuate four days ahead of a hurricane when it can change its track at any time." He then referenced living on the island of Barbados - where I also lived - and not being disposed to moving, given "We know a hurricane can change its direction ever so slightly so what's the point?".
And indeed, that occurred in 1979 when the monster Cat 5 Hurricane David bore down on the island with 170 mph winds. Most of us on the island prepared for the worst, realizing in a matter of hours our homes would be reduced to splinters and giant storm surges would sweep those of us near the sea (like wifey and me) out to perdition. Almost by serendipity, at the last instant, the brute storm veered ever so slightly north of the island and plowed into the tiny neighboring island nation of Dominica instead, causing untold damage.
"Bajan Boy's" point might appear sound but in fact, is not. Let's first observe he's writing as one confined to an island 22 miles by 14 miles - for a total of 166 square miles. There really is no place to evacuate to on such a tiny "rock". although the smarter Bajans - say near the sea- might head toward higher altitude terrain like Mount Hillaby - some 1500 ft. above sea level.
Miami and the Florida gold coast has no such high terrain available and indeed exists at barely 4 feet above sea level. This fact makes the city prone to flooding and the perils of storm surge - as we beheld with the scenes of flooded Brickell Avenue yesterday. (See "lesson two" below) Hence, given the many tens of thousands at risk - and with the capacity and space/distance to move - it made rational sense to do so. Let me put that this way: it made sense to follow the orders of local authorities (as well as Governor Rick Scott) to escape harm's way - as it was currently defined in the most probable scenario.
That made use of the map shown in the graphic with the cone of probability including the most probable path for the eye right up the interior of the state - but with the vicious northeast quadrant striking Miami most directly. This projection indicated the most devastating effects for Miami and surrounding areas of Miami-Dade (most prone to flooding) to evacuate. That the storm subsequently altered its path to the west (though still inside the cone) didn't make that evac decision hysterical at all. It made it rational to do at the time. Had Miami been struck as the path originally indicated and hundreds or thousands perished from a deadly 10' - 12' storm surge, authorities would never have heard the end of it.
But this is why we have the first lesson to do with Irma's strike at Florida: A hurricane's path will always be a probabilistic one, not deterministic. In other words, it is not like being able to predict the precise path of a baseball thrown at an angle at such and such a speed, from a given height. When one throws the ball its path is immediately under the control of gravity and the Newtonian laws of motion. The ball is a definite object, not a mass of whirling molecules, and so obeys those deterministic laws. The same doesn't apply to a giant storm composed of swirling air and water molecules..
This sounds obvious on its face, but the fact so many people in Florida carped and complained about the changed path shows it wasn't processed or appreciated. E.g. one guy writing in The Miami Herald: "I evacuated to the west coast and now see I am under the gun! Why do we pay these weathermen so much anyway?" Well, because they are able to give us early warnings of the approach of a storm and where (in general) it will strike, e.g. FLA, never mind it is to within + 5 percent confidence levels for specific regions. The comment writer then, ought to have known that the probability cone allowed for a small chance that Irma would move to the west and take the path we've now seen - striking Key West, Naples, Fort Myers, Tampa and even Orlando.
Let me also hasten to note how much more improved the Irma forecasts were compared to those for Superstorm Sandy five years ago. In fact, U.S. meteorologists were unable to predict Sandy's path with any degree of accuracy. Meanwhile, Europeans (using their "European model") predicted with confidence it would turn toward New Jersey and Staten Island, NY.
Why the difference between now and then? Well, because we now have a vastly more sophisticated tool, the "Finite Volume on a Cube-Sphere" (FV3) which helps to forecast hurricane tracks more accurately because more relevant data can be integrated within. Indeed, it now compares with the European model.
As a first test, 5 days before Harvey hit Texas, NOAA's Geophysical Fluid Dynamics Laboratory used the FV3 to predict that the storm would develop a second eyewall and produce extreme rainfall across the region. Both predictions as well as the path of the storm were spot on. Because residents and public official relied on these forecasts the death toll was remarkably low for such a Category 4 storm. (Latest reports indicate 60 died in Hurricane Harvey compared to 1,833 in Hurricane Katrina).
This brings us to the second lesson to be learned from Irma's impact on Florida: The ordered hurricane evacuations were not done randomly. They were based, for example in Miami, on areas most susceptible to storm surge - the deadliest aspect of most hurricanes. The storm surge map for Miami-Dade county is shown below:
The areas shown in red (Zone "A") are the most susceptible to storm surge flooding and were logically the priority locations targeted for evacuation. Similar maps apply to other Florida cities, for example Fort Lauderdale, Fort Myers, Naples, Tampa.
Thus, Rush Limbaugh's claim that the ordered evacuations were triggered by "hysteria" is nonsense. To get a glimpse of how the 5-6 foot predicted storm surge adversely affected Miami (even though Irma was churning up the west coast) see the video at this link:
At the same time, we've since learned that the projections for very high storm surges didn't materialize in other places. For example, the storm surge was predicted to be 10 to 15 feet at Naples (projections made by the National Hurricane Center) but actually topped out at 4- 5 feet. The same applies to Marco Island and Fort Myers. Each experienced storm surge significantly less than had been predicted.
This leads to our third lesson for Irma and hurricanes in general, which is analogous to the first: The prediction of storm surges is not an exact science, depending as it does on multiple variable parameters. NBC Meteorologist Bill Karins explained it thusly yesterday afternoon after the first tidal gauge readings came in a lot less than predicted:
"The storm gauge showed a surge at Naples of about five to six feet and it's now dropping. Remember they said two to three hours after the eye passed the highest level (ten to fifteen feet) would be reached. I stared at the tidal gauge to make sure it was accurately reporting and the surge was indeed dropping. What does that mean? That means it saved billions of dollars in damage and it prevented the loss of lives that the ten to fifteen foot surge would have seen had it happened.
Storm surge predicting and flood predicting is not like predicting the high of 75 and sunny the next day. There's a lot more than goes into it. Consider: the storm now is not as strong as was expected, it's further inland than was expected. So the parameters that they put into the computer to give the storm surge were likely too high. And don't forget they always said it was 'a possible storm surge of ten to fifteen feet'."
What's the takeaway from this lesson? It is to always allow for the maximum threat - say maximal storm surge - but never to expect it is a 100 percent, absolute exact forecast. And certainly no sane person - despite any inconvenience - should gripe when the worst doesn't come to pass. He ought to instead be grateful that it didn't.
This leads to the fourth lesson: Irma was so large - able to fit two 'Floridas' inside it - that every area of the state was affected. This again we saw played out in multiple scenes yesterday - from the Keys, to Miami (where two construction cranes were snapped by the winds and massive flooding occurred), to Naples, to Fort Myers and Jacksonville early this morning.
As The LA Times put it:
"So broad and punishing was the storm’s reach that no corner of Florida, the country’s fourth most-populous state, was unaffected.....Sunday’s dizzying sequence of stormy weather saw dual landfalls by the hurricane over a span of little more than six hours. After striking the Keys in midmorning, the eye of the storm moved over Marco Island, south of Naples. And soon after came the floodwaters, with water levels in Naples increasing 7 feet in just 90 minutes....As the storm’s trajectory took it north, water was sucked from part of Tampa Bay, exposing a muddy expanse that would normally be underwater — a frightening portent of flooding to come when that water, and more, comes rushing back."
This lesson - taken to heart - would have meant residents on both coasts needed to hunker down or make shelter arrangements- wherever they were, OR wherever they were going to. Those people who fled Miami for the west coast then, needed to locate potential shelters soon after they arrived, in case the hurricane track altered to the west, as it did. By the same token, those who remained in Miami in shelters, then carped about being cooped up, and squawking "It's not that big a deal now Irma has moved west so I'm leavin'!" needed to get a grip. On seeing the flooding from the storm surge in Miami, crashing construction cranes and downed power lines - their lives were probably saved.
Again, like those complaining that the altered track suddenly putting them in jeopardy those complaining about being "stuck in a shelter for nothing" processed the lessons of Irma inaccurately. In the latter case, just because a Cat 5 storm didn't strike with full fury and drown hundreds of people - collapsing high rise condos - didn't mean there was no impactful event. Or that ordered evacuations weren't really needed. Hindsight is always 20-20 and especially so for events that are probabilistic from the get go.
Weather Channel hurricane specialist Bill Norcross put the general issue to do with the FL east to west coast veer in the proper context (WSJ today, p. A4):
"Everybody is going to say 'Well, they forecast it to go east and it went west'. In actuality, they forecast it to come to Florida and it came to Florida."
This sensibly acknowledges that everything that was within the cone of probability was indeed affected, whether greater or less. The changes (east to west) of the main path within the cone don't alter the fact that those changes were covered by the overall probability. Yes....it'd be awesome if the specific track changes could also be accurately predicted. But, we're not there yet with the models we have, even the advanced FV3.
Here's the real worry for people who found themselves complaining for one reason or other about Irma's eventual strike: You should be more concerned about Trump's proposed cuts to NOAA's budget. Those plans include slashing the agency's funding by one fifth which will impact all its forecast functions - from tornadoes to hurricanes. In the latter case the budget for weather satellites alone (those that give us satellite views of storms) is scheduled to be cut by 17 percent.
In the words of Mick Mulvaney - the hack turkey budget chief - trying to justify it:
"We're simply trying to get things in order so we can look at the folks who pay the taxes and say 'Look, yeah we want to do some climate science but we're not going to do some of the crazy stuff that previous administrations did."
Yeah, like accurately forecasting hurricane paths or at least providing the most confident cones of probability for their paths. Pardon me while I insist my tax money be used for such "crazy stuff"!