Introduction to Numerical Weather Prediction Mike Leuthold Atmo 336
Model Components This flowchart, starting at the bottom, shows how the components of Numerical Weather Prediction (NWP) and how it fits into the forecast process. Initial Conditions-weather data Forecast Model-complex equations that duplicate atmospheric processes. Model Output-digital data, maps, tables, graphics Forecast-30% chance of rain,high of 65 degrees Graphics : www.comet.ucar.edu
Initial ConditionsWeather Data Many different sources of data are used to create the model's Initial conditions, which is the best guess of what the atmosphere looks like at the start of a model run. Errors are likely: Upper air stations are far apart, surface stations may report incorrect data, satellites do not measure atmosphere directly thus are subject to errors.
Forecast Model Internals Physical laws of motion and conservation of energy govern the atmosphere and these laws can be converted into mathematical equations. These equations, called the Primitive Equations, are the basis of any weather model. These equations then are translated into computer code. Due to their complexity, these equations can not be solved exactly by the computer in a reasonable amount of time, thus errors are introduced.
Forecast Model Internals A model is made up of an array of 3d grid points of various parameterstemperature, wind direction, wind speed, etc. These grid points can cover areas as large as the globe (global model) or as small as ~100KM (regional model). Distance between grid points can be from ~1km to 100km. Calculations are performed on every grid point for a specific point in time, then added to the previous value.
Forecast Model Internals A weather model also has data in the vertical dimension with the horizontal grid points arrays stacked up like uneven layers of cake. Vertical layers can range from 15 to over 90, depending on the configuration of the model. The more layers, the better the model can represent the atmosphere.
Model Physics and Parameterization Model Physics generally refers to all processes other than dynamics such as solar radiation, clouds, land/sea surface characteristics, and precipitation Models cannot resolve weather features and/or processes that occur within a single model grid box, but, the model must account for the effect.
Four Panel Global Forecast System Model Output
Forecasts National Weather Service model run lengths are from 5-14 days and vary in coverage from North America to the whole globe. Meteorologists use the output from models along with other sources of information to produce weather forecasts for public and private interests. Forecast for Tucson metro/marana/green Valley Area Forecast For: Tucson Metro Area Issued: March 13, 2008 03:20:46 MST Today: Breezy. Partly cloudy. Highs 81 to 86. South wind 5 to 10 mph increasing to west 15 to 25 mph and gusty this afternoon. Tonight: Partly cloudy. Lows 49 to 54. Southwest wind 5 to 15 mph. Gusts up to 30 mph this evening. Friday: Windy. Sunny in the morning then becoming partly cloudy. Areas of blowing dust in the afternoon. Highs 78 to 83. Southwest wind 10 to 15 mph with gusts to around 30 mph increasing to west 20 to 30 mph with higher gusts in the afternoon. Friday night: Breezy. Mostly clear. Lows 43 to 48. West wind 15 to 25 mph with gusts to around 40 mph becoming southwest 5 to 10 mph after midnight. Saturday: Breezy. Partly cloudy in the morning then becoming mostly cloudy. Highs 75 to 80. Southwest wind 10 to 15 mph increasing to 15 to 25 mph in the afternoon. Saturday night: Mostly cloudy. Lows 45 to 50. Sunday: Breezy. Mostly cloudy with a 20 percent chance of valley rain and mountain snow showers. Much colder with highs 55 to 60. Sunday night: Mostly cloudy. Chance of valley rain and mountain snow showers. Slight chance of thunderstorms in the evening. Lows 36 to 41. Chance of precipitation 40 percent. Monday: Mostly cloudy with a 30 percent chance of rain and snow showers. Highs 54 to 59. Monday night: Partly cloudy. Slight chance of rain and snow showers. Lows 34 to 39. Chance of precipitation 10 percent. Tuesday: Partly cloudy. Highs 68 to 73.
Regional Weather Models Much higher horizontal resolution-1 to 4KM instead of 10-100KM. Higher resolution enables some physics parameterizations to be removed thus better representation of actual physical processes. Resolves small scale features thus model is more accurate and produces better forecasts. Terrain is better resolved. This produces better forecasts in mountain and valleys, especially here in the SW during the summer thunderstorm season as thunderstorms almost always form over mountains.
Flagstaff Phoenix Tucson
August 28, 2006: Early Morning Heavy Rain Event
Rillito River at Swan August 2006
Other Applications: Cloud Forecast for Solar Power Generation
Climate Models Climate models operate on a global geographic scale and 100+ year time scale. Climate models are even more complex than standard forecast models. Computational requirements are large so resolution is reduced. Animation : www.gfdl.noaa.gov
Climate Processes
Climate Model Evolution Climate models have evolved from simplistic, low resolution models, to models that include much more complex physics and higher resolutions. Additional physical processes continue to be discovered that have an effect on climate forecasts and previous discoveries continue to be refined. Graphics :ipcc wg1.ucar.edu
Climate Model Evolution Resolution continues to increase as computer power increases. Recent work has been done to embed higher resolution regional models into global models in order to better model complex regional conditions. In spite of all the errors, global climate models can still give climate scientists general forecasts of global change.
How Hot? Graphics : www.wikipedia.org
How Accurate?
Comparison of Various Climate Models
Ensemble Forecasting Uses a set of slightly different initial conditions to make a series of forecasts. Also can use different models or physics schemes. Helps forecaster add confidence to the forecast.
Ensembles El Nino/La Nina Forecast