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."
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.
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
Yes, but it's not that simple.
The partial differential equations you're referring to are the Navier-Stokes equations, which have no known analytical solution. Instead, they're solved numerically with atmospheric data on a three dimensional grid. There are still a few problems here:
1) Although dx, dy, dz, and dt are pretty small these days, an order of magnitude lower than a couple of decades ago, we don't have in situ observations every dx, dy, and dz. You might have a few grid points in each county of the US now, perhaps more depending on the model. You probably don't have that many observations in most cases, and certainly not over the oceans where such observations are quite sparse. There are efforts to better assimilate observations into the models than what's been used in the past. For the most part, this has been a trend away from schemes such as a two pass Barnes Analysis and toward more complex and stochastic schemes such as the Ensemble Kalman Filter. Regardless, the first guess in any of these schemes is a forecast from a previous run of the model, which may or may not be a good approximation of the state of the atmosphere. The atmosphere is a chaotic system, so small errors in the initial state will grow greatly with time. Lewis Fry Richardson, one of the fathers of numerical weather prediction, put the theoretical limit on numerical weather forecasting with any skill at around three weeks.
2) There are processes that aren't directly simulated the model. These include surface processes like evapotranspiration and conduction of heat from the surface downward or into the lowest layer of air. Radiative transfer is another key process and it's affected by aerosols in the atmosphere. There are atmospheric circulations that occur on scales smaller than dx, dy, and dz such as subgrid turbulence and circulations in the lowest part of the atmosphere, the planetary boundary layer. Many of the global models such as the ECMWF, GFS, and UkMet have a coarse enough grid spacing that they can't resolve things like a thunderstorm. Microphysics, the types of hydrometeors like warm rain, ice crystals, snow, and graupel can't be directly simulated though the model dynamics, either. However, all of these things are very important to producing anything accurately resembling an accurate forecast. These are parameterized in the model instead of being explicitly resolved. The parameterizations have improved but they're still a somewhat coarse approximation.
There's a whole lot more involved than solving a few partial differential equations. If only numerical weather prediction were that simple...
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
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!
Open source weather prediction involves obtaining the data and running a numerical model.
There are a lot of acronyms for weather models in the comments such as the HWRF, HRRR, RAP, and NAM. All of the those models are actually various configurations of the Weather Research & Forecasting (WRF) model. Other models like the GFS, UkMet, and ECMWF are different and aren't based on the WRF. You can download WRF and compile it yourself. It's actually not that hard. The site is http://www2.mmm.ucar.edu/wrf/users/. WRF is in the public domain and so are the tools that you're most likely to need in order to run WRF.
WRF isn't usually a global model, so you'll need something to provide the initial and lateral boundary conditions for the model. That's usually data from another model. The data from the HRRR, HWRF, RAP, NAM, and GFS are all also in the public domain. The appropriate choice here depends on what you're planning to do with the model, but there's no shortage of public domain data.
The real limiting factor is the availability of computing resources. Current numerical models in their typical configurations can require hundreds of cores to run and each core needs several GB of RAM. High-quality numerical modeling is computationally expensive, which limits who is practically able to do this. But licensing, availability of software, and access to data aren't issues at all.
M-I-Z
kU still sucks!