It is currently Jan 17, 2019 6:53 am 

Reply to topic  [ 1 post ] 

Previous topic | Next topic 

  Print view

 Post subject: Weather Models
Post Posted: Mar 24, 2010 7:26 pm 
Oh my, Big Time poster!
Oh my, Big Time poster!
User avatar
Joined: Feb 19, 2007 5:52 pm
Posts: 2926
Location: Conifer Mountain
Updated December, 6, 2018

I have been asked by a few people about the various weather models that us weather geeks use to predict the weather. So I thought I would provide a brief explanation as to what weather models are, and the different models in use.

The term weather model refers to a set of computer programs that ingest weather observations (the initial conditions), and then solve the "equations of motion" to predict a future state of the atmosphere. The first attempts at solving the equations of motion began at MIT in the 1950's, using grad students with slide rules to solve the equations with input data. the problem was that it took 2 days to produce a 12 hour forecast. As computers became more powerful, they were used to solve the equations with numerical approximations, because math theory cannot exactly solve second order partial differential equations which exist in the equations of motion. This is the main reason that forecasts are never 100% accurate. So numerical weather prediction was born, and today the fastest and largest computing centers in the world are used to run weather models. The observations come from airports, aircraft, radiosonde balloons, satellites, and NEXRAD radars. All of this initial data is fed into the computers and they run algorithms to predict the weather. The output from models is done at equally spaced grid points horizontally, and at unequally spaced pressure levels in the vertical. The more dense the horizontal and vertical resolution, usually the more accurate the model is, but the physics involved in the model also plays a big role in how it solves the weather puzzle.

Here is a list of the predominant weather models:

GFS - Global Forecast System. This is a global model run by the U.S. National Weather Service (NWS) every 6 hours. The forecast output is available at 3 hour time intervals, extending out to 384 hours. The horizontal resolution is grid points every 22 km with 90 vertical levels.

NAM - North American Model. This is a regional model covering most of North America and run by the NWS. The current NAM is based on the WRF (Weather Research Forecast) model. The model is run every 6 hours and the output is avaiable at 1 hour time intervals extending out to 84 hours. The horizontal resolution is grid points every 5 km, and it uses 52 pressure levels in the vertical.

WRF - Weather Research & Forecast. This is another regional model covering North America run by the NWS. This model was developed by academic and research institutions and is hence readily available for private companies to utilize and adapt to other regions. This model is run every 6 hours and the output is available at 1 hour time intervals, extending out to 84 hours. The horizontal resolution is 10 km, and there are 68 vertical pressure levels.

RAP - Rapid Update Model, formerly known as the RUC. This is a high resolution regional model run by the NWS for the U.S. This model runs every hour and has output at 15 min time increments, and extends out to only 18 hours. The horizontal resolution is every 11 km, with 50 vertical pressure levels.

HRRR - High Resolution Rapid Refresh model is part of the RAP model suite. The HRRR uses a 3 km horizontal grid domain with enhanced physics and ingests NEXRAD radar for model guidance. The HRRR models is run hourly with output available at 15 minute intervals, out to 18 hours in the future for most hours, but avaialble out to 30 hours in the future at 0000, 0600, 1200 and 1800 UTC. This model is especially valuable in short range convection forecasting.

ECMWF - European Center for Medium Range Weather Forecasting. This modeling center in Reading, UK is a European consortium to provide the best medium range weather forecasts. This is a global model run every 12 hours, with output available at 3 hour time intervals, and extends out to 360 hours. The horizontal resolution is 9 km with over 140 vertical levels. This model also has a 50 member ensemble prediction system (EPS). The ECMWF is accepted as the best medium range weather model available, and the EPS can provide probabilistic output.

UKMO - United Kingdom Met Office. This is a global model run every 6 hours, with output available at 6 hour time intervals, and extends out to 384 hours. The UKMO and GFS model are back ups for each other, in case the one center has a problem, the other provides the global forecast. The resolution of the UK model are nearly the same as the GFS model.

EPS - Ensemble Prediction System. Each model has several members of the model suite that are run at the same time. Each member may have slightly different physics or use a different variational analysis scheme to initialize the model. Hence, each member can provide a slightly to largely different solution. Meteorologists use a EPS to evaluate the range of possible solutions in a particular model solution. If the EPS member solutions are close together, there is more confidence in the model forecast. If the member solutions differ greatly, there is more uncertainty in the model solution.

Uses, Strengths and Weaknesses
The GFS, ECMWF and UKMO models are global and hence provide a worldwide forecast. they are used by airlines and corporate flight departments to compute optimized flight plans, as well as by shipping companies to provide optimized ocean routes. These models all have their strengths in a global solution, there are no boundaries to the model area. The ECMWF has the reputation of being the best medium range forecast model in existence. In my opinion the GFS and UKMO do almost as well in most scenarios. The GFS tends to be overly wet, meaning it produces too much precipitation, and it tends to move systems at a fast pace, sometimes too fast, especially when a system is intensifying or becoming a cut off low. In general I like to use the GFS for winter forecasting beyond 48 hours. Within 48 hours, I like to compare the GFS with the NAM and RAP model output.

The NAM and RAP models are used primarily by the NWS for operational forecasting. They have a higher resolution, more frequent updates and time intervals, and according to NCEP (National Center for Environmental Prediction) better physics. The modelers tend to spend the most time tweaking these models to better predict small scale weather events, such as severe weather and strong winter storms. They have better terrain in the model, and hence tend to do a much better job of thunderstorm forecasting, especially in mountainous terrain. This winter they also seem to be doing better for orographic snows as well. These models can sometimes pick up localized weather events that the global models can't resolve. However, because these are regional models, they have boundaries, and all regional models can suffer from boundary conditions, meaning the boundary points are not initialized properly.

HRRR - High resolution Rapid Refresh model suite. This model suite has been developed by NCEP to provide a high resolution rapidly updating model for use by the NWS. The RAP model is the basis for this suite, and higher resolution models run for severe weather events have been developed and are in evaluation by the NWS local offices. Horizontal resolution is down to 3 km with 150 vertical pressure levels, 3 arc-second terrain and very good physics that include ground moisture. It is also being used to provide the first guess field for the NAM and WRF models. New techniques and physics tried in this model will make it into the GFS, WRF and NAM models when it has been verified by operational testers, so the HRRR provides a proving ground for the other regional models. This is by far the best model to look at for forecasting within 24 hours and for near term (nowcasting)forecasting.

In general, operational forecasters will look at various model output in order to see how well the different model output agrees. If the models are not in good agreement, then there is less confidence in the forecast. As one gains experience with the models, you begin to see their trends, weaknesses and strengths in situations. An experienced forecaster knows how the different models tend to perform in various scenarios. A typical forecast is usually a blend of different model output, unless a forecaster really trusts a certain model in a specific situation.

So, hopefully now you understand weather models a little better, at least now you don't think they are pieces of plastic we put together with glue when we're bored staring at the sky :) Any decent forecaster will use model output, as well as looking at many other things like the upper air pattern, surface pattern, satellite imagery etc... You only become a good forecaster with experience in a region. You learn as you go and apply lessons learned. As long as you understand there will always be busted forecasts. I will be sexist here and say the atmosphere must be female, as just when you get comfortable and think you know exactly what it will do, it does exactly the opposite. Remember that the next time SpazCat or myself busts a forecast, as it is inevitable. We just hope we get it right more often than not.

"Climatology is what you expect, Weather is what you get" - R. Heinlein.

"Sunshine is delicious, rain is refreshing, wind braces us up, snow is exhilarating; there is really no such thing as bad weather, only different kinds of good weather" - John Ruskin

Reply with quote  
Display posts from previous:  Sort by  
Reply to topic  [ 1 post ] 

You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:

Who is online

In total there are 18 users online :: 1 registered, 0 hidden and 17 guests (based on users active over the past 60 minutes)
Most users ever online was 2823 on Mar 26, 2012 6:26 pm

Users browsing this forum: yakbard and 17 guests

Powered by phpBB © 2000-2012 phpBB Group

This website copyright © 1994-2019 by is a member of the Platte Canyon Area Chamber of Commerce