Wednesday, September 19, 2007

Weather 101: What Are Computer Models?

Some popular questions I get from time to time have to do with computer models. You may have heard of the GFS, the WRF, the GFDL, the HWRF, the NOGAPS, the UKMET, the ECMWF, the CMC, and more but what are they and which one is 'better' than the others?

First off, what are computer models? Computer models are simulated forecasts of what a computer thinks the weather will be like over a period of time. Think of them as simulations.

Each computer model will forecast weather for a different amount of time, based on the data it ingests, the mathematical equations that the computer running these forecasts needs to process, and the amount of detail or resolution that is required to run the forecast. Some computer models, such as the Global Forecast System (GFS) run forecasts out to 16 days while other models, such as the Rapid Update Cycle (RUC) will only run for 12 hours.

Here is a brief tutorial on what models are out there, how long they forecast for, and how often they update. These only cover the major 'free' computer models that one can access through a bevy of sites, including the models page here at Phillyweather.net which provides a number of different computer forecast model links.

MODEL Stands For How often/day How many hours Best used
RUC Rapid Update Cycle Hourly 12 Nowcasting / current weather
SREF Short Range Ensemble Forecast 4 times daily
(03, 09, 15, 21Z)
87 Short term
WRF Weather Research and Forecast 4 times daily
(00, 06, 12, 18z)
84 Short term
UKMET United Kingdom Meteorological 2 times daily 120 Comparison forecasting to other models

GEM (Canadian) Global Environmental Multiscale 2 times daily 144 Comparison forecasting to other models

EURO European Model 2 times daily 240 Weather pattern analysis

GFS Global Forecast System 4 times daily 384 Weather pattern analysis/comparison forecasting

NOGAPS Navy Operational Global Atmospheric Prediction System 2 times daily 144 Comparison forecasting to other models

Some other computer models that are not included here are the GFDL which is a hurricane model that is run based off of GFS input, the NAM (which has been effectively replaced by the WRF), and the MM5 which is a mesoscale model run in several different formats.

These computer models ingest observation data (temperature, wind direction and speed, humidity, pressure, and more) from the surface and the upper atmosphere and the supercomputers that these forecast models run from will generate forecasts a certain number of times per day, with two or four times being the most common number of times these forecasts are generated.

Whenever a storm is close at hand, some of the shorter range computer models are more useful. The RUC, which runs hourly, is useful to determine possible precipitation trends with storms. It is a higher resolution model and is only intended to forecast for 12 hours as its main purpose is to serve as a nowcasting tool. By the way, "nowcasting" is defined as short range forecasting that can utilize primarily radar, satellite, and surface observations in forecasting. The RUC is a useful side tool but should not be relied upon solely for a nowcast. The WRF model is another solid type of short-term model and can sometimes spot a change in 'forecasted' track for a storm before the GFS or other computer models do.

Regarding long range forecasts, some forecast models are better than others. For instance, the EURO and GFS are better at spotting mid and upper level atmospheric pattern changes than the NOGAPS or the Global models are. However, these other models can spot occasional storms before the GFS or EURO catch on to seeing them. In forecasting, you will typically see meteorologists rely upon the GFS or EURO for midrange and longer range forecasts as these computer models typically perform the best at seeing what changes can take place in the atmosphere.

Meteorologists use the different computer models out there and can make a forecast based partially on the forecasts these computer models make. However, the computer models are only guidance into what could happen and it is important to note that accuracy can sometimes vary widely. That's why good meteorologists will rely upon other tools such as surface observations in other parts of the country and climatology (past performance in similar situations) to come up with a forecast.