|
 |
|
|
|
Computer Models
The advent of numerical models—software that can simulate a process mathematically—has transformed the study of the atmosphere over the last half-century. Countless models have been developed at NCAR and elsewhere to study weather, climate, and the processes behind them. While some models can be run on desktop computers, the most sophisticated and comprehensive models require the world's fastest supercomputers. Many try to predict the future; others aim to replicate the past. However, all of these models serve as a laboratory for some part of the atmosphere by simulating its evolution in cyberspace. Statistical models call on a large set of data about the past in order to calculate probabilities about the future. Other models are dynamical: they incorporate physical laws and track the atmosphere as it unfolds, regardless of its past behavior. Many models incorporate both statistical and dynamical processes. For example, a global climate model is too large in scale to depict individual thunderstorms. Instead, it approximates the numbers and locations of these storms by a statistical technique called parameterization, while the larger-scale atmosphere in the same model unfolds dynamically. All models require initialization, a starting point usually based on real-world data. Because observations are not collected on the precise mathematical grid used by the model, it can be a challenge to provide enough data to create the best-possible initial state. Many models also incorporate varied types of data, a process known as assimilation. One of NCAR's strategic initiatives is devoted to new techniques for data assimilation. NCAR supports a wide variety of specialized models, including those aimed at understanding
Two of the most common model types developed and used by NCAR scientists are general circulation models (GCMs) and mesoscale models. Climate studies use general circulation modelsGCMs are the key tool by which climate scientists analyze how the world's atmosphere will evolve in the coming decades to centuries. Global in scope, a GCM may simulate hundreds of years of climate, a few hours at a time, at points separated by perhaps 100–200 miles (160–320 km) horizontally. The entire task can take weeks of processing time on a supercomputer. The output is summarized in various ways—by months or by seasons, for instance. Data may also be transferred to a regional model that depicts the climate of a smaller area in more detail. The scale of a model—global or regional—presents tradeoffs in accuracy depending on what is being depicted. NCAR's flagship GCM is its Community Climate System Model (CCSM). A special version, PaleoCCSM, replicates ancient climates and compares them to our own. Other GCMs are maintained by major climate research centers around the world. Each GCM incorporates physical laws and characterizes uncertain elements in a different way. Together, these tools help scientists understand the behavior of climate and help policymakers decide how to deal with climate-change risks. Weather forecasting requires numerical or mesoscale modelsSince the 1950s, U.S. meteorologists have increasingly relied on numerical forecast models to foresee the movement of large-scale weather features. With each new development, forecast accuracy has improved markedly. A typical forecast model starts out with the current conditions from land and ship stations, weather balloons, aircraft, and satellites. It then calculates the evolution of key features, such as fronts and large-scale storms, over the next few days across its domain, usually a continent or hemisphere. Companion models use a statistical data base to estimate what sort of temperatures and precipitation might occur for a particular city. A human forecaster may adjust these figures based on her or his knowledge of local climatology and how the model handles certain situations. Because of weather's chaotic nature, errors or uncertainties in the starting point of a model can alter the results dramatically. One way to reduce the impact of such errors is through an ensemble of forecasts. In this technique, one model is run several times, each with a slightly different, intentionally varied set of starting points. Mesoscale (middle-scale) modeling focuses on areas larger than a city or county but smaller than a continent. Mesoscale models can provide higher-resolution output than a standard weather model, because of the focus on a limited area. Such a model might be run as often as once every three hours, with data points separated by only a few miles or kilometers. NCAR researchers are developing and refining a number of specialized models to advance weather and climate research. One example, a modeling framework suitable for both study and prediction, was developed by NCAR and five collaborators. The Weather Research and Forecasting Model (WRF) is bolstering understanding and prediction of mesoscale weather features and promoting closer ties between researchers and forecasters. The National Weather Service began using WRF output in October 2004. The version used NCAR researchers is called Advanced Research WRF (ARW). < previous | index | next >
Home |
Organization |
Research |
News |
Education |
Tools |
Libraries |
Search |
UCAR | NCAR | UOP | ©2008, UCAR |
Sponsored by
This document can be found at
|
|||||||||
