Slashdot Mirror
NASA's Hurricane Model Resolution Increases Nearly 10-Fold Since Katrina
zdburke writes: Thanks to improvements in satellites and on-the-ground computing power, NASA's ability to model hurricane data has come a long way in the ten years since Katrina devastated New Orleans. Their blog notes, "Today's models have up to ten times the resolution than those during Hurricane Katrina and allow for a more accurate look inside the hurricane. Imagine going from video game figures made of large chunky blocks to detailed human characters that visibly show beads of sweat on their forehead." Gizmodo covered the post too and added some technical details, noting that, "the supercomputer has more than 45,000 processor cores and runs at 1.995 petfalops."
Nice and all, but the model resolution was never the problem. E.g. Katrina was going to hit whatever model we had, with not much warning. The problem was the response.
Violence is the last refuge of the incompetent. Polar Scope Align for iOS
The path of a hurricane is somewhat unpredictable (been known to turn 90 degrees for no apparent reason).
The bigger issue which is harder to address is making homes that can "largely survive" being hit by a hurricane. The biggest issue is the junk flying around due to the strong winds (and storm surge if you are near the water). Once a building starts to disintegrate it provides the wind with ammunition for taking out other buildings.
In Australia when a cyclone is heading towards your community and potentially make land fall within 48 hours there is a whole pile of things that kick in for preparation (food, water, fuel, tie down and clean up - most people will be sent home by work during this period). At about six hours it is a case of bunker down and wait for it to go overhead.
Better prediction will reduce the amount of communities put on alert and associated disruption but unlikely to reduce the damage in affected areas (for that you need better building codes and people willing to take appropriate measures).
What the hell is a petfalop?
HAHAHA!
Hooray beer!
Why are typos so much funnier when drunk?
"Helping to keep you two steps ahead of the Thought Police!"
It's been estimated that if you had temperature, pressure, and humidity readings for every cubic foot of atmosphere, you could only predict weather to about a month...if you had the computer power, which you wouldn't.
That's the problem with the famous butterfly effect. The tiniest deviation of a single molecule changes the weather patterns months down the road. The microscopic movements dictate large scale events down the road. The error scientists made was assuming small changes dissolved into statistical irrelevance, and that statistics of large masses of air drove particular day-to-day weather patterns. They drive climate, not weather (though particularly severe weather events can shift climate to other strange attractor basins, too. A bad volcano or two and a summer where snow doesn't clear and an ice age can come on in a year or two.)
Weather is the pattern of glitter on the water, not a statistical analysis of the pattern.
(-1: Post disagrees with my already-settled worldview) is not a valid mod option.
building a home that can handle the winds from a Category 1 storm isn't that hard.
It is when the primary criteria is "build it as cheaply as we can get away with and not have to bribe the building inspector". It's embarrassing the crap being slapped together today, especially to a former remodeler. When you step into a multi-million dollar house and notice that the counter tops aren't even level, the floor trim and cove molding rely on caulk and plastic wood to come together, and the ceiling is so wavy that the chandelier base plate doesn't even touch in places you know damn well that there aren't hurricane braces on the roof joists and the wall framing isn't anchored to the floor joists.
"Think about how stupid the average person is. Now, realise that half of them are dumber than that." - George Carlin
The school buses didn't belong to the school district, much less the city. These free-market idiots who believe in privatizing everything to make it more expensive and less efficient had ensured that there were no school buses available to move people. Nagin was an idiot, but that was one failure that can't be laid at his feet.
More disturbing to me was that Cuba had sent a ship full of doctors and Venezuela had sent a tanker full of fuel for hospital generators, and both were turned back by the Navy. Most of the hospitals stayed staffed by nurses and candy stripers (the doctors could afford to evacuate) until the All Clear, and the generators ran out of fuel until Halliburton trucks could get to them (even domestic trucks of donated fuel were turned back because only authorized vendors selling at elevated prices were allowed in).
"Think about how stupid the average person is. Now, realise that half of them are dumber than that." - George Carlin
Actually we can predict pretty damn well. It depends on what you compare to: perfection, or what we could predict thirty years ago, or seventy-five years ago.
We're living in an era of rapid improvements in weather forecasting in terms of accuracy, precision and scope. Back in the 70s there was a perennial science fair project in which the student compared the accuracy of tomorrow's weather forecast to simply assuming that tomorrow would be like today. The answer back then was, it was about equally accurate. Today would be a totally different story. The forecasts we get for three days out is better than the forecast we used to get for tomorrow back in the 70s; people just haven't updated their thinking.
It's not surprising when you realize that the difference is satellite tracking, meteorological data networks, and incomprehensibly more powerful computers. Today's smart phones are roughly as powerful as the supercomputers of the 1980s.
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
If one is approaching, be sure to stock up on plenty of munchies, beer and weed. Use a bunch of car batteries through an inverter to avoid the clanking, rattling generator.
What else is there to know? I mean, besides not building matchstick homes so close to the open seas where there are hurricanes? And certainly not below sea level! What were they thinking? Trying to save a few pennies? You don't need high resolution to know that doing stuff on the cheap is pretty risky business. Though I'm sure the pictures look nice.
“He’s not deformed, he’s just drunk!”
I think that is another accuracy be precision issue. The storm clips are likely in place, because they are usually a specific inspection item. Being cheap means using all the lumber on the truck, and not rejecting the wavy boards. Building square, straight, and plumb buildings with dimensional lumber is hard. It is much easier to do with TJIs and Glulam, but tolerances are typically over 1/8" in framing.
I'll bite.
With a higher resolution you can begin to resolve some of the dynamics of the eye wall, which gives you a much better idea about how the storm's intensity will develop.
There was a very good wunderground blog post about this last week with a /. story about it.
~.~
I'm a peripheral visionary.
I'm going to agree with the AC above me and say yes, increasing the resolution will generally result in a better forecast. There are two areas that generally can improve weather models: better initial conditions for the model and higher resolution.
Hurricane tracks are primarily influenced by large areas of high and low pressure and the accompanying upper-level winds. Even a coarse model with grid points every 50 km will probably produce a reasonably accurate representation of large features. Adding to the resolution might help a bit, but it's not likely to improve the forecasting of hurricane tracks that much.
Hurricane intensity, however, is driven by the storms in the inner core of the hurricane. They typically form a ring around the eye of the storm and are referred to as the eyewall. They're basically thunderstorms but without a lot of lightning, typically less than 10 flashes per hour. Inside those storms, large amounts of water vapor condense, and when this happens, latent heat is released. The latent heat comes from the energy that's no longer being used to keep the molecules of water apart once the state changes from gas to liquid. This heat warms the air, causing the column of air to expand upward. When this happens, you get high pressure at the top of the hurricane, and air spirals out from there. Air is flowing out of the core of the storm at the upper levels, so there's less air to press down on the surface. This causes the surface pressure to fall and the hurricane intensifies. These storms are small, on the order of 20 km across, but they're really important to predicting hurricane intensity. If the model has its grid points spaced 50 km apart, you may not have any grid points to resolve the storms. One of the newer hurricane models, the HWRF, now goes down to a grid point spacing of 2 km. There are a lot more grid points from which to represent the storms in the inner core of the hurricane. That allows much better prediction of hurricane intensity.
For a few decades, hurricane track forecasting had improved quite a bit as the global models could better predict the high and low pressure systems that drive hurricane tracks. During a lot of that time, predictions of hurricane intensity didn't get a lot better. Only within the past decade or so have we seen bigger improvements in forecasting hurricane intensity. That's largely a result of more computing power and models that are able to directly simulate what's going on in the inner core of the hurricane.
I do agree with the AC that it's not as simple as changing a couple of numbers in a configuration file to get a higher resolution and automatically getting a better forecast. There is something of an art to modeling. Even with a really high resolution grid, there are still processes that can't directly be simulated or are on a smaller scale than the grid. We still have to parameterize those. But improving the resolution is generally a good thing.
M-I-Z
kU still sucks!
The resolution of the model is the spacing between horizontal grid points. Vertical levels are treated separately. A typical model now uses on the order of 50-100 vertical levels, which has increased substantially in recent years. Models tend to use sigma or eta levels, which are a vertical coordinate system that is terrain following, at least in the lower levels of the atmosphere. The grids are also vertically stretched, resulting in more grid points in the lower levels of the atmosphere. This is necessary because of the complexity in simulating the planetary boundary layer and interactions with the surface. Also, shrinking the distance between grid points requires a corresponding decrease in the model time step. If you cut the distance between grid points in half, typically you'll also need to cut your time step in half, too.
M-I-Z
kU still sucks!
Actually, no, that's not correct at all. Even as Katrina was crossing south Florida, three day forecasts were pretty far off. The forecasts certainly didn't call for it to become a major hurricane. Here are some forecast graphics for you:
http://www.nhc.noaa.gov/archive/2005/graphics/AT12/10.AL1205W5.GIF
http://www.nhc.noaa.gov/archive/2005/graphics/AT12/10.AL1205I.GIF
The three day forecast called for a category 2 hurricane moving in the general direction of Panama City, FL. The rapid intensification of a number of hurricanes that year wasn't predicted well. You can animate the graphics for yourself, if you'd like. Here's a link: http://www.nhc.noaa.gov/archive/2005/KATRINA_graphics.shtml.
The rapid intensification of Rita wasn't forecast well at all. Forecasts predicted it hitting south Texas for awhile rather than near the Texas/Louisiana border. Rita was forecast to become a major hurricane, but not as strong as it became.
The forecast track of Wilma was actually really good. However, the rapid intensification from a minimal category 1 hurricane to a 175 mph category 5 hurricane in 24 hours wasn't forecast at all.
All the NHC forecasts for 2005 are archived at http://www.nhc.noaa.gov/archive/2005/. If you look at the strongest hurricanes that year, you'll see that the rapid intensification cycles were pretty much not in the official forecasts at all. The track forecasts weren't great, but generally were just within the margin of error.
The hurricane forecasts in 2005 left a lot to be desired. Pretty much nobody knew several days before those storms how bad they were going to be.
M-I-Z
kU still sucks!
Nothing excuses Ray Nagin's incompetence once hurricane watches and warnings were issued for Louisiana. Although Nagin did encourage evacuation, he also promised residents who didn't evacuate that, "we will take care of you." The city didn't keep that promise and it certainly gave people incentive to stay when they might have otherwise evacuated. Furthermore, the hurricane wasn't actually what killed people in New Orleans. The levees broke because they weren't properly designed and maintained, despite ample warning. It was known back in 1965 that the levees and flood walls were inadequate and improvements were estimated to be complete by 1978. By 2005, the work was still incomplete with an estimated finish in 2015. Much of the levee and flood wall design issues can be blamed on the Army Corps of Engineers. Had the levees been designed and maintained properly so they didn't break, almost nobody would have died in New Orleans. There's plenty of blame to go around and not all of it belongs with people who chose to ride out an indirect hit from what had then weakened to a category 3 hurricane.
However, it's absolutely false that the forecast models in 2005 were adequate. The models weren't capable of resolving the storms at the inner core of hurricanes and couldn't provide forecasters with guidance about the rapid intensification of Katrina, Rita, and Wilma. The grid spacing of models in 2005 was simply incapable of resolving the inner core of hurricanes, which is essential to accurately predicting intensity. Instead, intensity forecasts relied on statistical models that do well in a wide variety of situations, but not during those rapid intensification cycles. The intensity forecasts for those storms was particularly awful and the forecast tracks of Katrina and Rita left a lot to be desired.
M-I-Z
kU still sucks!
NHC forecasters consider a very large number of models when making hurricane forecasts. Many of these aren't run by NOAA. Here's an old list: http://www.nhc.noaa.gov/modelsummary.shtml. NHC has used the NOGAPS model for a long time. NOGAPS is a global model developed and run by the US Navy. If a NASA model can make useful predictions, NHC forecasters will certainly use it.
M-I-Z
kU still sucks!
It is nice to know a disaster will strike, but it is another story to handle it properly. Is there any progress here, or will we see again medias calling theives the people that seek survival by taking food from closed supermarkets?
Thanks, very informative post. Spatial resolution has a major effect on the accuracy on any FEA model. Taken to the extreme, if the block size in a FEA simulation is comparable in size to the Gulf of Mexico, then you won't see hurricanes in your model. One consideration is that the size of a hurricane does not necessarily indicate storm intensity. Cyclone Tracy that hit Darwin in the 70's was an unusually small cyclone, IIRC less than 50km across, but the winds were amongst the strongest ever recorded in Darwin, strong enough to flatten it. I believe Darwin is now a Mecca for storm modellers hunting data, during the wet season a tropical storm conveniently develops over Darwin bay every afternoon, it's like clockwork, locals have a name for it, IIRC it's called "Harry".
To those who are bringing up the "chaos" issue, yes we will never get accurate forecasts more than about a month in advance. Since Tracy hit Darwin in the 70's accuracy has improved from 3 days at best, to 10 days at best, that's extremely useful for the military and commercial logisticts. Predicting cyclone tracks is notoriously difficult but it seems these days that tracks for the next 5 days are usually pretty accurate.Track predictions for any number of days above 1 in the 70's were almost useless. New Orleans had at least 3 days warning, everyone was told a massive storm surge was expected, hell I was 10,000km away in Australia and heard about the expected surge days before it hit. New Orleans was not a natural disaster, it was the "worse than useless" response from authorities that was the problem. Darwin suffered the same "lack of response" problem in the 70's, but to be fair it was a very remote place back then.
And did you exchange a walk on part in the war for a lead role in a cage? - Pink Floyd.