Numerical Modeling of the Weather and Climate

Data collected by balloons, ships, satellites, and other methods, help us understand the processes which control our glo bal environment and determine which natural events can be predicted. Scientists use this data to create computer models that simulate the weather and climate of Earth's environment. As is seen on TV weather shows, winter storms move from the Pacific Ocean, across the Rockies, into the Eastern US, then out over the Atla ntic. In summer, hurricanes move from the warm oceans into the continent. To study climate, which can be thought of as the average weather, we must make computa tions for the entire globe, including land as well as oceans.

The weather can be defined by five key variables:

• temperature
• wind magnitude
• wind direction
• water vapor (humidity)
• pressure

We can write mathematical equations that reveal how different weather variables relate to each other and to external fac tors such as solar energy.

One of the primary applications of the modern high-performance computer is numerical weather prediction. Beginning in the 1950s, technology evolved to the point where weather equations could be solved on a computer. However, the task was not simple. One complication was that since weather is a global process, a global picture of the atmosphere is required. While obtaining the surface picture (with 75% ocean cover) is hard, organizing global coverage of observations through the height of the atmosphere is far more complex.

For more than 100 years, data in the upper atmosphere have been obtained by weather balloons carrying an instrument package called a radiosonde. Such balloons are launched from hundreds of stations around the world twice a day. The U.N. World Meteorologic Organization fosters the necessary global cooperation. In spite of this, the data from the upper atmosphere remains sparse. More recently, weather satellites have provided valuable help; they look "down" into the atmosphere and provide useful information. However, even with all of these different ways to observe, we still do not have a complete picture of the atmosphere.

By "feeding" observations of the atmosphere into the equations that describe how the atmosphere works, one can complete the set of observations. The computer solves the equation and the computer winds carry information from observed regions to unobserved regions. This takes advantage of the fact that weather often moves in organized systems. Perhaps more importantly, the models can provide us with estimates of things we cannot observe, like the evaporation and vertical transport term in the equation. In essence, we constrain the part of the equation we observe and then let the computer fill in the rest. This is data assimilation.