Dost thou know the balancings of the clouds ...?
February 1, 2019 9:56 AM   Subscribe

"The polar vortex (previously) is dangerous, record-breaking, can’t-look-away weather. Yet this cold snap’s arrival was preceded by a marvel so spectacular that we hardly noticed it: It was correctly predicted. As early as a month ago, forecasters knew that colder-than-average weather would likely strike North America this month; a week ago, computer models spit out some of the same figures that appeared on thermometers today." Robinson Meyer in The Atlantic on how Modern Weather Forecasts Are Stunningly Accurate, and why that's an astonishing achievement.
posted by RedOrGreen (46 comments total) 35 users marked this as a favorite
 
This would have been a golden opportunity for one of those fancy data visualizations. Temperature, Wind, Rain forecast error bars as a function of time.

Even better if you could change location and see if which models work best for which areas. (I live in a river valley that gets pretty drastic 5 days changes regularly. The weekend prediction on Monday is often a crapshoot. But by Thursday, it's dead on)
posted by DigDoug at 10:08 AM on February 1 [3 favorites]


One thing that surprised me about moving to the Bay Area is that the weather forecasts around here seem to be much worse than the ones I'm used to. I don't know if it's because the relatively stable weather has made everyone lazy, or if it's the ocean being erratic, or what, but when they say it's going to rain tomorrow they're wrong. If they say it's going rain later this afternoon, they're still wrong.

Maybe it'll rain in forty minutes. Or possibly in the middle of the night. Or not at all.

...but whatever the forecast is, it's wrong.
posted by aramaic at 10:15 AM on February 1 [10 favorites]


But the zombie recieved idea that forecasts are little better than guessing persists
posted by thelonius at 10:19 AM on February 1 [1 favorite]


(not directed at aramaic - the Bay area has, I'm told, microclimates, and it is difficult indeed to forecast there).
posted by thelonius at 10:20 AM on February 1 [7 favorites]


The Portland OR forecast is reliably unreliable. Like aramaic I have long wondered what the reason is.
posted by Pembquist at 10:21 AM on February 1


Here in upstate NY, I've noticed a remarkably stable general forecast for 8 days at a time - yeah, the details change, it's 2F not 4F, but the trend lines and sharp weather fronts are accurately predicted. The high was 12F yesterday, the forecast is for a high of 37F tomorrow, and I can trust they have it right.

I guess chaotic eddies on the coast are still too hard to model, though.
posted by RedOrGreen at 10:30 AM on February 1 [2 favorites]


It's remarkable because weather is the canonical example of chaos theory in action -- everyone knows the butterfly effect. They say it fast in the article but don't emphasize it enough.

Forty years ago it looked like the only way to move forward was to have an infinitely fine grid of weather stations, or else smash all butterflies. Now they've gone and done it a third way.
posted by Quindar Beep at 10:30 AM on February 1 [3 favorites]


The Portland OR forecast is reliably unreliable. Like aramaic I have long wondered what the reason is.

What do you need a weather forecast for?

October-July Cool and damp, things not on fire
July-October hot and dry, things might be on fire
posted by Automocar at 10:35 AM on February 1 [13 favorites]




Being from New England, I always wondered what it was like to be a weatherperson in a place like San Diego. What exactly do they do? Um, it's going to be warm and sunny again today...
posted by Melismata at 10:42 AM on February 1 [1 favorite]


Improvements in weather forecasting occupied a chapter of Nate Silver's book on predictions.
posted by doctornemo at 10:43 AM on February 1


My understanding of why weather predictions on much of the west coast are so bad is that there are some NOAA sensors in the ocean that are out of commission, and have been for a long time. North American weather tends to arrive from the west, generally - sometimes northwest, sometimes southwest, but generally west. To the west of the coast is...a huge ocean. It's difficult to gather pertinent data that assists forecasts from such a vast expanse of saltwater, hence the sensors. The midwest and east coast have a continent of data to draw from and see what is forming up and how, thus giving rise to generally more accurate forecasts.
The Puget Sound area, like the Bay area, has a lot of microclimates as well, with funky convergence zones where it's a WHOLE LOT wetter, and rain shadow areas where it's a WHOLE LOT drier. So yeah - weather is fun, but hard to predict in some areas.
posted by dbmcd at 10:55 AM on February 1 [17 favorites]


It's funny, because weather forecasts somehow *feel* just as wrong as they always did, even though I know they are actually better than they used to be. I think it's because as they've gotten more accurate, they've also gotten more available and more precise, which creates new room for inaccuracy.

Like, 20 years ago, I could turn on the news and someone would tell me tomorrow's high and low temperature and whether it was likely to snow a little or a lot in Metropolitan Toronto with the general sense that things are always a little cooler by the lake than at the airport where the official temperatures are measured.

Now, I can turn on my phone and instantly be presented with a detailed, hour-by-hour forecast customized to the part of the city I'm in, including precipitation, wind chill, etc, etc. And then if I go outside and the snow doesn't start within the precise hour they predicted it would, I'm all 'Pffft, weather forecasting is such bullshit.'
posted by jacquilynne at 11:01 AM on February 1 [23 favorites]


I was just wondering on this last night - how it was predicted a week ago and then happened as predicted. I have been impressed.
posted by obliquity of the ecliptic at 11:43 AM on February 1 [1 favorite]


Computers and the advances in them are incredibly important for the enhancements in weather forecasting, which seems fairly obvious, but is hard to understate.

I work in a different forecasting field (transportation, mostly with a side order of land use) and read Silver's book, which is interesting because he comes at it from two other fields (politics and sports). The unique challenge of weather forecasting is there's so much of it, and it happens all the time.

In the transportation models we use, we can spend days testing a scenario, which is fine because it's a construction project that starts building 5 years from now and will take 10 years to do, so who cares about a few days. But a weather forecast, they need to get in the updated weather measurements from the entire world, clean all that data, then run the forecast for the entire world, and produce the results, and they can't take more than a few hours to do that, because otherwise the forecasts are useless. (And if I understand correctly, the standard is actually to do many model runs, which is why they say 30% chance of rain -- in around 30% of the scenarios, it rained, therefore they had to have a number of scenarios.)

I think the European model (which covers the world) uses 0.1 degrees, which seems like a small area -- and it is, there are almost 6.5 million grid cells at that resolution (not sure if they collapse the cells in the poles, where 0.1 degrees east-west is quite small indeed). And the weather is in 3D -- they have something like a thousand levels of altitude because the weather and temperature vary as you go higher in the sky, so now we're in the billions of cells, each of which has to be calculated every time step for as many steps as you want your forecast.

So 0.1 degrees seems small in terms of physical space, but it's still pretty big -- in San Francisco, the Legion of Honor and Transamerica Pyramid are both in the same 0.1 degree grid cell, as are much of the Presidio and the border with San Mateo county; basically the entire city, which has very different microclimates caused by all the famous hills. No wonder it's hard.

The flip side is, that because there's all this weather being made all the time, there's so much data to check forecasts and develop new models with. So when Nate Silver looks at Presidential elections, there's only 45 so far total, but realistically James Knox Polk vs. Henry Clay is meaningless as far as trying to say something about the current political era. Smartphones have been around for three (3) US presidential elections and the iPhone was still new back in 2008. You can go back three more elections so there's a whopping six, and you are now at the dawn of the internet as a mass medium. If you go back to JFK, the new technology transforming elections is fricking television, and you still only have fifteen observations to draw conclusions from.

Whereas weather, there's a week's worth of global observations being made every week, so there's tons of new out-of-sample data being generated all the time.
posted by Homeboy Trouble at 11:44 AM on February 1 [17 favorites]


I was at a conference once with a speaker from the Met Office, talking about how they talk about weather forecasts. They said that long-range are good but difficult to communicate effectively giving them a poor reputation. (80% chance that it will be a hotter than average summer doesn't mean it's going to be a scorching BBQ summer in the UK, but that's what the papers will write.) They did say that the five-day forecasts are now excellent, but didn't explain why. Now I know.
posted by plonkee at 11:46 AM on February 1


> aramaic: ...but whatever the forecast is, it's wrong.

...which, if true, would be as useful as a perfect forecast

But un-snarkily, it seems that for some reason, predicting rain is much much harder than predicting temperature. I'm guessing there's just something profoundly nonlinear about when and where things precipitate
posted by kleinsteradikaleminderheit at 11:55 AM on February 1


I always wondered what it was like to be a weatherperson in a place like San Diego. What exactly do they do? Um, it's going to be warm and sunny again today...

They made a movie about it in fact! I guess a lot of people had the same question
posted by Carillon at 12:34 PM on February 1 [4 favorites]


I remember reading a book by Solomon Diamond (World of Probabity: ...) published in the '60s, in which he claimed that the best weather forecasting available at the time was just barely better than predicting that whatever happened that day would happen again the next day.

Times really have changed.
posted by jamjam at 12:36 PM on February 1 [1 favorite]


My understanding of why weather predictions on much of the west coast are so bad is that there are some NOAA sensors in the ocean that are out of commission, and have been for a long time. North American weather tends to arrive from the west, generally - sometimes northwest, sometimes southwest, but generally west.

This was so incredibly frustrating to me when I lived in England. Great Britain is actually a pretty small island and the don't have anywhere near the radar coverage that North America does (probably due to the much lower threat of tornadoes). This meant that for seven years I couldn't use radar maps to plan my cycling and running around the rain areas I could see on radar maps. You could not see the big storms rolling across a continent. What they did have was satellite imagery which is pretty much just showed that there was cloud cover all the time. Useless.

It was such a relief to move back to Chicago and to get to see weather coming all the way across a continent before it got to us on radar.
posted by srboisvert at 12:42 PM on February 1 [4 favorites]


I always wondered what it was like to be a weatherperson in a place like San Diego. What exactly do they do? Um, it's going to be warm and sunny again today...

I don'r know about San Diego but I know LA and OC, and the local broadcast weather folks and tone is on the tip of the bleeding edge for hype, drama and fear whenever there's any kind of weather at all to report. This are the people who pioneered weather buzzwords of the "storm-pocalypse" variety as early as the 80s and 90s.

To be fair SoCal does actually get some gnarly weather. There's really heavy rain from monsoons and winter storms, which is why there's flood control channels everywhere. And the high winds and extreme fire danger from Santa Ana winds are notably dangerous, and then there's all the mudslides after burns if it gets heavy rain.

But yeah, when there's nothing to report the local broadcasts were pretty boring.
posted by loquacious at 12:46 PM on February 1 [3 favorites]


Just 11 years ago when I was studying aviation meteorology for my FAA ATP Dispatcher Certificate, our instructor for that module cautioned us that ANY forecast has less than a 50% chance of being accurate and that we should actually be impressed by how often forecasts are close to accurate.
posted by KingEdRa at 12:53 PM on February 1 [1 favorite]


As for forecasts in places like SF, Portland and Seattle - these places all have diverse microclimates and extremely chaotic weather flow patterns.

I was once biking through an intersection in Seattle and I experienced about four kinds of precipitation in the few seconds it took to cross the intersection. There was rain, blinding sunlight, sleet, hail, more sunlight, and snow and possibly graupel - aka puffy hail.

Each one of these has a distinct feel and texture on one's face on a bike at speed, and the variety was so bewildering it was like my senses got stunned and overwhelmed by the strobing effect, and I had to pull over for a moment to collect my wits because I started laughing so hard.

Heck, I can walk out my door and traverse a temperature range of up to 20+ degrees F between sunny patches or much cooler hollows and shady forested parts. I've been in the yard and have seen individual little puffy clouds the size of a car and within reach just wandering through while the rest of the air around is crystal clear, dry and not foggy at all.
posted by loquacious at 12:56 PM on February 1 [8 favorites]


Being from New England, I always wondered what it was like to be a weatherperson in a place like San Diego. What exactly do they do? Um, it's going to be warm and sunny again today...

REAL - they cover the weather for an insanely large land area in categories like beaches/mountains/plains/high desert etc. The forecast for each of those is very different. They don't just cover the weather for San Diego.

Which, speaking more generally now, is possibly why people deem forecasts to be inaccurate? The average local weather covers an entire MSA, so a 30% chance of rain over an area 100X100 sq miles is not going to seem particularly accurate to the average person. I don't deem our weather particularly accurate for that reason. They mostly get the big ideas right, but they can be off by multiple hours as fronts take time to pass through the MSA.
posted by The_Vegetables at 12:57 PM on February 1 [1 favorite]


I love weather maps, I love that weather prediction has gotten so accurate. It saves lives, allows farmers to plant/ harvest/ irrigate, etc. efficiently, and many other businesses are helped by this essential government function.

It's not always perfect, and people seize on They said we'd get 18" of snow and we got 6" but you can look at the animated maps and see that the storm veered left, and that's amazeballs.

I was just thinking how much I'd like to have a year's worth of animated weather maps playing. I find it endlessly fascinating. Here is a vortex animation from JPL.
posted by theora55 at 1:01 PM on February 1 [4 favorites]


also, weather nerds are usually nice, and weather nerdery is fun.
posted by theora55 at 1:01 PM on February 1 [4 favorites]


REAL - they cover the weather for an insanely large land area in categories like beaches/mountains/plains/high desert etc. The forecast for each of those is very different.

I meant to mention this as well. Growing up, the broadcasts on TV covered everything from the coast to the high deserts, so while it might be a balmy 70 in Santa Monica, it might be 100 in Chatsworth, 45 in Big Bear and 115 in Victorville.

They always had one extreme thing or another to talk about. There was almost always high winds somewhere, or a burn area in danger of mudslides. If they didn't talk about that they'd talk about air quality and smog alerts, because those go hand in hand with nice weather. If the weather was too warm then they'd talk about heat exposure. If they didn't have anything at all to talk about they'd talk about travel weather, or if everyone was obviously going to the beaches because it was a holiday or the weather was nice, they'd talk about beach safety, UV indexes and stuff like that.

Also they sometimes had names like Dallas Raines, which, come the fuck on script writers.
posted by loquacious at 1:07 PM on February 1


See, I have the opposite impression, theora, just mostly because meteorologists are, along with structural engineers, the most likely "scientists" to be trotted out to defend creationism or antivax or whatnot.
posted by Scattercat at 1:14 PM on February 1 [2 favorites]


Can concur that the standards of weather reporting in the UK are appalling. My local weather report will proudly insist that it's two degrees above freezing, which is lower than the low of three degrees which they haven't revised in light of this data, during a hard frost.
posted by Dysk at 1:22 PM on February 1


If I'm going to rag on local weathermen (not necessarily meteorologists), I'm also going to complain that their temperature measuring equipment is generally located in areas that are appropriate for meteorological reasons but don't necessarily correlate very well to local weather conditions. Like there are temperature gauges on every corner, weatherman don't tell us that we didn't quite top 100F on that balmy day in the summer. Just because your equipment didn't doesn't mean that tons of other places did.
posted by The_Vegetables at 1:30 PM on February 1


Here's a decent article about parts of the US that are especially tricky to forecast weather for and why. A lot of it comes down to topographical features that can make the difference between a spot-on forecast for half of a city and a total bust a few miles away. And summer thunderstorms can occur on such a small scale that it's impossible to say where exactly they'll pop up.

It's important to keep in mind that the polar vortex disruption was a planetary-scale weather event, and in some ways easier to forecast on a long time scale than predicting the weather for a particular location a few days in advance. Even so, the early predictions of the polar vortex being disrupted didn't give a clear picture of where exactly the cold air would go. I remember seeing projections as recently as last week that the vortex would be a few hundred miles further to the east.

On sort of an unrelated note it's interesting that the north-central US, which was the area most affected by the polar vortex disruption, is one of the easier parts of the country to forecast the weather this time of year. You're not really dealing with small-scale convection like thunderstorms popping up and the atmosphere in winter is generally much more stable than in the rest of the year. This doesn't apply to places downwind from the Great Lakes, however, where lake-effect snow bands are notoriously hard to predict.
posted by theory at 3:42 PM on February 1 [4 favorites]


I live on the coastal side of a ridge of hills and work in the valley just on the other side. I have a ten mile commute and regularly experience two entirely different weather patterns in a day. I've left home where there was barely a flake of snow and arrived at a freshly plowed parking lot and also have had to dig out my car and then arrived at work past green lawns and clear streets. You just have to deal with a probability range.
posted by Karmakaze at 4:07 PM on February 1 [1 favorite]


This meant that for seven years I couldn't use radar maps to plan my cycling and running around the rain areas I could see on radar maps. You could not see the big storms rolling across a continent.

As a young weather nerd, one of my favorite things in the summer was monitoring the radar as huge thunderstorm cells blossomed over the Dakotas in the late afternoon. By midnight, moving steadily eastward, they'd have ganged-up into a mesoscale convective complex and begin stalking across Minnesota, reaching the Twin Cities around 4am with rain bucketing down and vivid lightning. It was high drama.
posted by theory at 4:12 PM on February 1 [5 favorites]


I always wondered what it was like to be a weatherperson in a place like San Diego. What exactly do they do? Um, it's going to be warm and sunny again today...

I seem to recall this very sort of thing biting Steve Martin's weatherman character in "L.A. Story" right on the ass.
posted by Insert Clever Name Here at 6:42 PM on February 1 [3 favorites]


Planning for bike rides a few days in advance, I pay attention to the forecasts.
Temperatures: almost always within a few degrees F, excellent.
Rain: harder to predict, especially if the rain is spotty over the area.
Wind: often quite wrong, showing high gusts and strong winds that never show up.

Forecast discussion
I like to use the National Weather Service's "Forecast Discussion" to get an idea if the forecast is solid, or if different models are showing divergent results or timings. They also comment about which part of the local area will be affected.

It's on the weather.gov site: for example: Georgetown OH, linked from the city forecast page. Scroll down to the "Additional forecasts and Information" list. And I use the "Hourly Weather Forecast" link a lot -- showing temperatures, wind, clouds, rain, rainfall amounts, etc.

This discussion used to be difficult to read, showing the text in all caps with a lot of technical abbreviations within the text. So I started using wunderground.com's edited version, called "Scientific Discussion", that spelled out the abbreviations. I see that weather.gov is now more useful -- and it has easy popups for technical terms.

I still don't understand all the commentary, but I can get a good idea of their consensus.
posted by jjj606 at 7:37 PM on February 1 [2 favorites]


The 'dusting' of snow we were supposed to get in Northern Virginia today that lasted well into the afternoon is a nice contrast to this posting.
posted by 922257033c4a0f3cecdbd819a46d626999d1af4a at 7:41 PM on February 1


Step aside, radar techs and cloud-huggers: the groundhog has emerged and did not see his shadow. So now we know the weather through, what, mid-March? Nice.
posted by wenestvedt at 4:32 AM on February 2 [2 favorites]


One of my pet peeves: when people disparage forecasters, like "as if the weather people ever know what they're talking about [rolls eyes]".

On one hand, I know that it's just a conversational trope. But on the other...knock it off, man. Yes, forecasting is an inexact science. But the degree of accuracy that we do have is fucking astounding.

You think you could do any better? The atmosphere is a complex, chaotic, interdependent system that's literally the size of a planet. And yet, meteorologists can tell us that there's going to be a big thunderstorm on Tuesday, or that the incoming hurricane will probably land on your state, or that the freaking polar vortex is going to take a trip down to the Midwest next week.

I don't know how well it holds up – but one of the most illuminating books I've ever read (at age 14 or so) was James Gleick's Chaos: Making a New Science. Chaos theory and fractals were very 90s pop-science things, of course – but I think that complex systems remain underappreciated by the general public. People instinctively expect that a given input will result in a proportional output. But, on the real-world scales of social and economic systems, most systems are hugely interdependent and nonlinear.
posted by escape from the potato planet at 5:00 AM on February 2 [8 favorites]


For the West Coast folks complaining about your forecasts, dbmcd has it, with a side of microclimates. I have spent my entire career in the central states, and I can't count how many times I or a colleague has said, "the model should be better next cycle when the system gets onshore". West Coast forecasters don't have that luxury. Plus NWS headquarters and the main model developers are very DC/East Coast centric.

As for comments about getting paid to be wrong, I've heard them all, even from my family. It doesn't help that most people think if something doesn't happen at their house, it doesnt happen at all. Oh well, at least I'm not an economic forecaster. I once heard that it takes 3 months of number crunching to get the *current* state of the economy right, let alone the future.
posted by weathergal at 9:12 AM on February 2 [10 favorites]


IME the NWS Portland OR forecast is right far more often than it's wrong. There's an awful lot of confirmation bias and unreasonable standards among non-meteorologists.

Forecast: "Sunny, 68-70 degrees"
Reality: Partly overcast until about 10 AM, then sun the rest of the day, with a high of 67
Result: Loud crowing about how the forecast was just wrong-wrongity-wrong.

Even when the forecast turns out right, people will need to take the occasion to rehearse all the times when it wasn't, so the accurate forecast just doesn't count.

I don't know what the cure is, because I don't think people are going to give up this particular chew toy any time soon.

(Disclosure: I am not a meteorologist, but have an amateur's interest in the subject.)
posted by Weftage at 9:29 AM on February 2


I don't know what the cure is, because I don't think people are going to give up this particular chew toy any time soon.

Weather has always been a popular topic for phatic discourse, and maybe the enhancing theme that "they" are all idiots is just too appealing for people to give up.
posted by thelonius at 9:50 AM on February 2 [1 favorite]


Weather has always been a popular topic for phatic discourse, and maybe the enhancing theme that "they" are all idiots is just too appealing for people to give up.

Yeah, most people don't really track the actual weather reports all that closely, they half hear them or hear one from early in the week and don't hear the updates and then use that as a generalized basis of conversation. In some areas it has a routine built in where the pay off is in talking about how frequently the weather changes in the area they're in, the "If you don't like the weather, just wait X minutes" concept. It's taking a perverse pride in the alleged unpredictability of the weather as making the area more interesting in a sense. It has almost nothing to do with real meteorologist forecasts.
posted by gusottertrout at 12:16 PM on February 2


I wonder how Farmer's Almanac is doing prediction wise.

Prediction wise, Los Angeles Metro Area is usually pretty good even though it's West Coast. Because it's a Mediterranean Climate at the point where the arctic coming down through those northern bits meets the equatorial coming up from the southern bits. Things from the north die when they hit the warm water, Things from the south die when they hit the cold water. And that's how they track, so few real surprises. But yeah, it's mostly rainy-season mudslides and dry-windy season wildfires as far as actual weather happens. Unless you live up in the mountains or desert or the canyons or the hillsides.

I'd rather they work on the Earthquake prediction.
posted by zengargoyle at 12:36 PM on February 2


I feel like the meteorologists in Oklahoma are better than the meteorologists in most of the rest of the country. Two reasons for this:
1. Oklahoma is the Big Leagues. Tornadoes on the ground, and you need to be able to tell people where they are, and where they're going. There is no "gentle rain" in Oklahoma; you're going to be putting up severe thunderstorm watch boxes at the very least, and you are going to be on TV a lot.
2. As everyone knows, all of the nation's weather is produced at the National Severe Storms Lab in Norman, Oklahoma. So a lot of it is just proximity; things get weirder the further out you go. [/fake]
posted by Huffy Puffy at 12:47 PM on February 2 [2 favorites]


There's a rather lovely global map of weather data here. that gives a pretty good view of the polar vortex.

It always seems ridiculously sci fi to me.
posted by Just this guy, y'know at 3:02 PM on February 2 [4 favorites]


I've thought for awhile that the weather forecasting in Seattle is pretty good, as long as you account for the possibility of further of weather either stretching out (rain on one day becomes rain on two) or contracting (three sunny days turn into two sunny days). We'll get all the weather, but they can't tell you what's going to land on the weekend until Friday.
posted by Margalo Epps at 4:15 PM on February 18


« Older Langston Hughes Reads Langston Hughes   |   This Woman's Work: A Live Symphonic Tribute to... Newer »


You are not currently logged in. Log in or create a new account to post comments.