Slashdot Mirror
How Weather Modeling Gets Better
Dr_Ish writes: Bob Henson over at Weather Underground has posted a fascinating discussion of the recent improvements made to the major weather models that are used to forecast hurricanes and the like. The post also included interesting links that explain more about the models. Quoting: "The latest version of the ECMWF model, introduced in May, has significant changes to model physics and the ways in which observations are brought into and used within the model. The overall improvements include better portrayal of clouds and precipitation, including a more accurate depiction of intense rainfall. The main effect of the model upgrade for tropical cyclones is slightly lower central pressure. During the first 3 days of a forecast, the ECMWF has tended to have a slight weak bias on tropical cyclones; the new version is closer to the mark."
For the chicks
I can understand why they get predictions about the future wrong, that bit is hard. What gets me is when the prediction for right now is at odds with what is actually happening. It seems like there is a significant delay between sensors on the ground taking a reading and the models being updated.
Apart from it being slightly comical when the guy on TV says sun is out but looking through the window I can see rain, it makes short term predictions useless. If they say it will rain this evening when I want to go out but the forecast for right now is wrong, what am I supposed to do with that information?
At best you get a vague prediction of the weather in the next few days, but the exact timing of events tends to vary quite a bit from what they say. It doesn't help that in the UK TV weather forecasts are delivered in the most confusing way possible, but fortunately we have the internet now.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
Weather often varies dramatically over small distances. It may be sunny at the studio and raining 1/2 a mile away where you are. Rain predictions are not 50% chance that a given spot will get an inch of rain. The prediction is that 50% of a large area will get a inch of rain. Big difference between those.
I just wish Weather Underground hadn't been taken over by The Weather Channel and, more to the point, their Web dev team. WU was an outstanding source for quick, concise reports on current weather, weather history, and news. Apparently they still post interesting content from time to time, but it just isn't worth my while to go slogging through the "new, improved UI" to get to it.
I still do pull it up for local conditions and radar; if I'm not in a hurry, it gives me the info I need, at least when I'm on my home 30/5 TWC connection. I had the misfortune of trying to use it through a throttled Hughes satellite connection the other weekend, and I finally gave up; there's so much AJAX crap going on that it took over a minute to load even part of the local conditions page, and the radar page simply wouldn't ever load. And this is with ads disabled.
I haven't been running public-facing Web projects lately, but if I go back to it, I'll insist that we test on something other than a fast LAN connection and giant screen. The current team at WU obviously doesn't do that.
I can understand why they get predictions about the future wrong, that bit is hard. What gets me is when the prediction for right now is at odds with what is actually happening. It seems like there is a significant delay between sensors on the ground taking a reading and the models being updated.
I'm not a Met, but I do have some understanding of the processes involved in model forecast runs.
Both the ECMWF and the US's GFS V2 (which, incidentally, is normally trounced by the ECMWF when it comes to predictive qualities) rely on a global network of real-time pressure and temperature sensors. Every six hours, that data is assembled into the databases needed for the ECMWF (and GFS, GEM, JAMSTEC etc) to run. They crunch the numbers and spit out a forecast, starting 6 hours out from the initialization conditions.
Apart from it being slightly comical when the guy on TV says sun is out but looking through the window I can see rain, it makes short term predictions useless. If they say it will rain this evening when I want to go out but the forecast for right now is wrong, what am I supposed to do with that information?
Back in the days when Pluto was still a planet and Bill Crosby wasn't creepy, the local TV stations actually employed a meteorologist to do their short-range forecasting (or had an agreement with the local airport or whatever). That guy (usually an older guy) would draw the maps and explain the weather, and usually get it right (with, of course, the obviously insane errors). Since that guy was doing local forecasts (and note the airport comment; planes need really good local weather to land, or at least they did; I'm not sure about today), it was actionable information.
Fast forward to today, and you see a TV personality showing off a narrow waist and ample cleavage whilst presenting the weather. They are reading off model forecasts that may or may not bear much resemblance to your reality. That crap is at best guidance only; and hardly actionable for this evening.
If you want to have actionable rain information, I've read about an app called Dark Sky (note: I have nothing to do with this app; it's feature set and how it works caught my attention). It uses the pressure and temperature information that comes from your phone (and app users around you) to build a local-to-you forecast. My iPhone 4S doesn't have the necessary sensors, so I've not really followed up with the app, but I have heard lots of good things. YMMV.
Isn't weather forecasting just solving PDEs?
Interesting to note that the 5-day forecast intensity error for 2014 is lower than 1, 2, 3, or 4 day error..
Contribute to civilization: ari.aynrand.org/donate
My experience has been that the predictions are usually quite accurate about what will happen, but is less precise about when. When they predict a front moving in, you can count on those thunderstorms, but don't always count on the storms starting exactly at 5pm.
The article mostly talks about predicting tropical storms, and for this the modelling of exactly what the path of the storm is going to be is critical. It does see that the predictions of storm tracks is getting better.
http://www.geoffreylandis.com
I live in Central Texas and I've had raging thunderstorms when the Weather Underground report insisted there was 0% chance of rain and the radar map showed nothing. At other times, the sky was clear when Weather Underground talked about "thunderstorm warnings" and showed red all over the radar. I don't trust the weather report at all.
dude, contribute to the conversation or go away.
your political agenda crap has no place in this thread.
What prediction of "right now" do you believe they have wrong? The prediction of "right now" that was done 6 hours ago and is seen within a half mile of where you are? Because here's a hint for you; the weather they give isn't the weather of what you can see over your house: it's the weather of your region.
"Apart from it being slightly comical when the guy on TV says sun is out but looking through the window I can see rain"
Yeah, right, this totally happened. Not made up at all. You have reams of evidence of this happening...
I can understand why they get predictions about the future wrong, that bit is hard. What gets me is when the prediction for right now is at odds with what is actually happening. It seems like there is a significant delay between sensors on the ground taking a reading and the models being updated.
Most regional models are run at least once an hour and make use of data from sensors up to a few minutes before the model is run. Even if a weather model was perfect, it would still produce incorrect results because models are always started with conditions that do not reflect the true state of the atmosphere. Weather models use other models or earlier runs to produce a best guess at the current state of the atmosphere. The atmosphere model is "adjusted" using the latest sensor, sounding, radar and satellite, data and then the model is run to predict the future. All weather models have to deal with the fact they only have an loose estimate of the state of each 3-D cell in the real atmosphere.
While we have lots of sensors close to the ground, we have very few up high in the atmosphere. The state of the upper atmosphere is rarely measured directly. Balloon soundings are done once or twice a day and only from a few places. The latest soundings have a huge effect on weather model results, so major corrections take about 12 hours if the sounding for an area was not representative of the local state of the atmosphere.
Apart from it being slightly comical when the guy on TV says sun is out but looking through the window I can see rain, it makes short term predictions useless. If they say it will rain this evening when I want to go out but the forecast for right now is wrong, what am I supposed to do with that information?
The best weather models usually get the features right, but are often off in timing. Frontal passages often happen hours from when they are predicted, but they do occur. Use radar and surface sensor data if you want to know current weather. Satellite data are usually less than 30 minutes old. Use model data to get a best guess on the future.
At best you get a vague prediction of the weather in the next few days, but the exact timing of events tends to vary quite a bit from what they say. It doesn't help that in the UK TV weather forecasts are delivered in the most confusing way possible, but fortunately we have the internet now.
Wide spread rain and snow events are the hardest to model or predict as the precipitation falls out in bands which form chaotically. Unless one precisely knows the starting position, one cannot predict the outcome over time. We will never have sensors every N meters in a 3-D grid over the earth so weather predictions will always be off.
Weather forecasts for airports is still, if not even moreso, required to be accurate and rigorous and there are MASSIVE penalties for getting it wrong, and massive amounts of money spent on getting it right. If anything the scenario you remember is even more exacting. Definitely not less.
Especially now that there are the ability of people to make their own weather forecasting service off the back of the observing network that government have to pay for (with everyone begrudging the expense). Government departments of meteorology have to do enough better to justify the expense that they have to charge for the network that the private companies doing the same work get for free.
Dr. Masters' article is about numerical models that are used for hurricane forecasting. There is nothing simple about hurricane forecasting.
1) The atmosphere is a chaotic system. Small errors in the initial state grow exponentially as the model is integrated forward. There aren't many in situ observations over the oceans, so the initial state of the model is generally less accurate there than over land. Forecasts produced will also have larger errors.
2) A wide range of models are discussed here ranging from global models (e.g., ECMWF, GFS, UkMet) to mesoscale models (e.g., HWRF). The global models are spectral models and don't have grid points, but the approximate resolution is around 20 km. The 2015 version of the HWRF, for comparison, is a gridded model with dx and dy of the inner nest set at 2 km (an upgrade from 3 km in past years). The HWRF also has an outer nest covering basically the entire basin and provides lateral boundary conditions for the vortex-following inner nest. There are huge differences in the models and, therefore, in the output they produce.
3) Hurricane tracks are generally driven by large-scale processes like high pressure systems over the Atlantic and Pacific and frontal systems like mid-latitude cyclones. These things are large enough that even the global models simulate them pretty well. Nonetheless, the sparseness of observations over the oceans can be a problem. Large-scale processes drive hurricane tracks.
4) Hurricane intensity is completely different. Large scale factors like moisture, sea surface temperatures, and vertical wind shear (change in wind speed and/or direction with height) do modulate hurricane intensity. However, the intensity is driven directly by the deep convection (basically thunderstorms, but with very little lightning and thunder) at the core of the storm. Although a hurricane has a huge cloud and precipitation shield, the eyewall storms are only on the order of 20 km across. The global models currently can't resolve this at all. Mesoscale models with coarser resolution like the NAM (12 km) can't resolve this. The high resolution models can resolve the eyewall convection, but additional resolution is definitely beneficial. It comes at a cost, though. Improving the horizontal resolution means adding grid points, which requires more computing power. Not only that, it requires a decrease in dt, which if too large will result in a CFL error and the model won't function. Lowering dt means even more computing power. Cutting dx and dy in half will multiply the number of gridpoints by 4 (2x2), but dt must also be cut in half. Assuming the number of vertical levels remains unchanged, it will require 8 times as much processing time and 4 times as much memory to run the model.
5) Observations are key to the models and more observations of the core of the storm will improve forecasts. But this region is also very hostile to observation systems. Manned flights are dangerous and expensive. They're also around 10,000 feet, so they have to use dropsondes to get data on what's going on below. Drones are a possibility for data collection. Assimilating radar data from these airborne platforms might help with analyzing the eyewall storms for the model's initial state. That said, getting the additional observations is difficult.
6) You still have to get the data into the model, and this isn't easy. Older schemes like a two-pass Barnes Analysis really aren't suitable for noisy data like radar observations. The trend is toward stochasic schemes like the Ensemble Kalman Filter (EnKF), which are more suited to this type of data. But work on using EnKF schemes to assimilate observations is still pretty early in development. It's used a lot in thunderstorm and tornado research to make use of mobile radar observations. But it's just now starting to get into operational use. It also requires more computing power because its an ensemble scheme, generally requiring at least 50 and preferably around 100 ensemble members.
There's a lot of good work going o
Reminds me of the term 'Social Engineering' (as well as others) and how it has been redefined to convey something other than its original meaning.
The terminology and trends are the same as when I did my thesis in numerical weather prediction 30+ years ago..... and the secret sauce, even then, was the sub-grid/harmonic scale physics for convection and clouds....
I am a meteorologist and I disagree with a lot of what you're saying. I'm not on TV. I'm a researcher.
Yes, a lot of forecasting now involves looking at numerical models. But you're getting a lot of things wrong here.
Yes, the ECMWF was once far better than the GFS. Over the past decade, the GFS has dramatically improved and the difference is much, much less. The GFS is criticized for its forecast of Hurricane Sandy eight days out, but there are plenty of times when the ECMWF is also dead wrong. The GFS isn't routinely trounced by the ECMWF. Not any longer.
A lot of shorter range forecasting is based on other models. The NAM runs out 84 hours, as opposed to 10 days or more for the GFS and ECMWF. It has a higher resolution and a different dynamical core. Sometimes it does really well, but it's best within about 48 hours. There are higher resolution, shorter term models like the RAP and HRRR that go out 18 and 15 hours, respectively. Sometimes they do a great job predicting when and where thunderstorms will develop. But I've also seen so many times when the HRRR predicts explosive thunderstorm development up and down the Plains while nothing happens. It's a useful model, but it's no substitute for looking at the observations and understanding the relevant meteorological processes. I'm fond of saying that you have to trust the science (like conceptual models and how the observations relate) and not simply read what the models say.
About TV meteorologists, I think they get a lot of unfair criticism. Most of the people on TV have meteorology degrees. Some are just broadcast meteorology degrees from schools like Mississippi State. I think the latter is of a lower quality when it comes to actually understanding the science. But there are also people who blow off the harder meteorology classes and manage to get a degree. That said, there are also plenty of on-air meteorologists who really do understand the science and do a good job forecasting. It's important to learn the conceptual models, but forecasting also requires experience and learning how to actually interpret the models and synthesize a forecast from observations, models, and the underlying science concepts. There's a lot of trial and error in learning to forecast well. One of my big criticisms of some schools that have large, prestigious meteorology programs is that their undergrad program really doesn't require students to learn how to forecast. They have extracurricular activities but those are optional. At the same time, there are smaller schools with lesser reputations that put a strong emphasis on forecasting and actually turn out better forecasters. Many of the people on TV also have other duties like reporting and writing content for the station's website so they don't get the time to focus on forecasting. They're also subject to the demands of management, which consists of people who don't understand weather but do understand ratings and profits. Meteorology often takes a back seat to other demands.
At the national level, I think the Weather Channel gets some very unfair criticism, including from meteorologists. Not all is unwarranted. They shouldn't be showing entertainment programs like Highway Through Hell and American Supernatural when there's dangerous weather going on. That's an indefensible decision made by the NBC executives for money, not by the meteorologists. Some of the programs aren't that good, but they have some that are of a good quality. Why Planes Crash is excellent. WxGeeks is outstanding. Storm Stories doesn't focus as much on the science, but I think it's a good reminder of how dangerous storms can be and that people need to take warnings seriously. It has its place. Sam Champion and Al Roker do nothing for me. They're not meteorologists. I don't think Jim Cantore has a meteorology degree, but he has plenty of forecasting experience and definitely has learned the science. Severe weather coverage with Cantore and Dr. Greg Forbes is the best coverage anywhere. Forbes was a professor at Penn State and definite
They need a model that is in a different league than this to get more accurate forecasting. To begin with, they must monitor the sun heat output in the direction of the earth to calculate the temperature impact on the earth and moon, accounting for objects inbetween. Then they need to monitor sea and earth temperature down to something like 10 m2 grid or better to get better accuracy. Then there is humidity, perspiration, percipitation and a whole lot of other factors including human-made factors. The earth core temperature of course also plays a role. There are so many factors not accounted for today, that it's no wonder forecasting is as bad as it is.
I don't know about the context, but it may well be true that the sun is out at the TV station while 100 km away where you are, there's rain. That would be my guess.
I'd imagine forecasting the weather in the UK might have less skill than in, say, the eastern US. A lot of your weather comes from the Atlantic, and there just aren't a lot of direct observations over the North Atlantic. The forecasts often shift in the US once storms move onshore from the Pacific. That's because there are a lot more observations over land than out over the water.
As for the models, it really depends on which model you're looking at. Normally within the span of an hour or two, forecasts are just based on extrapolating the motion of clouds on satellite and rain/snow on radar. Beyond that, models have a bigger role. In the US, short range models like the HRRR might incorporate new observations within an hour or two, because they run every hour. Longer range models like the NAM and GFS run every six hours, so it could take anywhere from 2-10 hours for new observations to make it into the model forecasts. I believe the two global models in the UK, the UKMet and ECMWF only run twice a day, so I think it could take as long as 6-18 hours for new data to make it into the model forecasts.
But I doubt that's actually the issue. Nobody looks at the models to tell them what's happening now. They look at surface observing stations, radar, and satellite. I think it's more likely that it's sunny at the TV station but you're a significant distance away and it's raining where you are. As for mid-range forecasts being inaccurate, that's probably because much of your weather comes off the Atlantic and there aren't nearly as many observations for the models to make a good forecast from.
Really?
I check 4 websites for the weather in my city [forecast.io, theweathernetwork.com, weather.gc.ca, accuweather.com], and they regularly will all have totally different weather. One will be 'sunny', one will be '5-10 cm of snow', one will be 'rain', and one will be 'overcast'.
It seems like they all get together, decide on the 4 most extreme possibilities, then randomly assign each one to a site.
Sleep your way to a whiter smile...date a dentist!
I think that's your problem. I'm not familiar with forecast.io and theweathernetwork.com. I am familiar with the last two. If I were you, I'd go with weather.gc.ca because Environment Canada actually does a pretty good job with forecasting. It's the equivalent of NOAA here in the US and they have pretty good forecasters. Accuweather is a pretty lousy operation based out of Pennsylvania. They've always come across to me as mostly being concerned with profit and not really being concerned about science. At best, we'll probably only be able to accurately forecast the weather out 21 days with any skill at all. Accuweather produces forecasts out, I believe, at least 30 days. That's way beyond what we're currently able to forecast and even beyond the theoretical limit. You can run a model out that far and get it to produce a forecast, but at that point any forecast is totally a work of fiction. I also suspect that most of what you're getting from them is a model, which may or may not be right, and not necessarily the product of a human forecaster who actually looks at the models and tries to decide which is best. I'm a meteorologist and I think Accuweather forecasts are garbage. Go with Environment Canada.
Part of the problem with UK forecasts is that they don't give the percentage chance, they just say it will it will not rain. I've always thought that was ridiculous.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
https://en.wikipedia.org/wiki/...
Too bad some idiots stole the domain from the original group back in 90's that wanted it as an archive.
I wasn't considering even a week in the future. These were next day forecasting.
One of them will nail it. I just don't know which one.
Sleep your way to a whiter smile...date a dentist!
They still are, a news article on "the other news site" just posted about their site (and others) serving malware ads. https://soylentnews.org/articl...
Weather often varies dramatically over small distances. It may be sunny at the studio and raining 1/2 a mile away where you are. Rain predictions are not 50% chance that a given spot will get an inch of rain. The prediction is that 50% of a large area will get a inch of rain. Big difference between those.
Not exactly. From the wikipedia Chance of Precipitation page:
if there is a 100% probability of rain covering one half of a city, and a 0% probability of rain on the other half of the city, the POP for the city would be 50%. A 50% chance of a rainstorm covering the entire city would also lead to a POP of 50%. The POP thus usually expresses a combination of degree of confidence and areal coverage.
Note that the Storm Prediction Center uses a different method when they give percentages for tornadoes and other severe weather. A 25% for tornadoes on a SPC convective outlook graphic means there is a 25% chance of a tornado within 25 miles of any given point in the area outlined on the map. As an aside, 25% using this method is pretty damn high, so if you ever see that indicated for your area you should really pay attention.
Now open source weather prediction is possible, these days. You can download the required weather initialisation files and the prediction engine. The only problem is noone has made open source for it. There is free RASP (see http://www.drjack.info and http://www.drjack.info/cgi-bin/rasp-forum.cgi and http://www.drjack.info/twiki/bin/view/RASPop/WebHome ) but it is not open and has maintenance issues. The hard part is validating and tweaking for a region.
A RASP operator (I can't even install it...).
My experience has been that the predictions are usually quite accurate about what will happen, but is less precise about when. When they predict a front moving in, you can count on those thunderstorms, but don't always count on the storms starting exactly at 5pm.
This depends heavily on where you live.
In places along the Front Range of the USA, the forecasts can be incredibly inaccurate. Presumably this represents chaos injected into the equations by the transition from high mountains to plains.
Residents learn to check the weather every day, and even then it will often be wrong.
From a gardener's perspective, it's incredibly frustrating to not know when one will get rain, or a frost, or hail, or how much will fall. This is why most people don't farm here.
I've been a weather geek for on about 35 years. Living in south Florida, the weather can be more than unpredictable. I've seen vast improvements due to advances in radar, satellite, and just sheer record keeping. I still love it, however, when the news says, "We don't know why we haven't had a hurricane this year", I always add, "Yet."