The University of Arizona

Climate Models Versus Weather Models: Different Approaches for Different Needs | CLIMAS


Climate Models Versus Weather Models: Different Approaches for Different Needs

Thursday, October 7, 2010

If you live in the Southwest, no doubt you know a little bit about our weather. It's interesting here. You likely know we have a monsoon during the summer, with awesome thunderstorms and flash floods, and our wettest winters and biggest floods happen when there's an El Niño event. We've also been hearing a lot about global climate change - and there's already a lot of evidence that it's happening here now. One big question is: how will our monsoon change in the future?

We have a few ways to look at this issue. We could look to the past hundred years or so, when we have temperature and rain gauge data, and try to make projections. We could reconstruct past climate during warm times to understand what might happen when temperatures are warmer in the future; we can do this using tree rings, cave formations, or lake sediments, for example. Or, we can use computer-based models to project what future climate might be like.

Now, I know that a lot of people don't trust models. Have you heard "scientists can't predict the weather in a few days; why should I trust them about fifty years out?" However...climate models are different than weather models. Each has uncertainty associated with it, but the uncertainties are different. 

Weather models are used to forecast day-to-day changes in weather, or rather to predict what will happen at a specific place and point in time in the near future, typically no more than five to seven days out.  Model-based weather forecasts generally less reliable beyond a week, because the atmosphere is an inherently chaotic system. Small changes in observed conditions, which are fed to the model regularly, can produce completely different weather predictions a week into the future because the atmosphere is so dynamic.

In contrast, climate models aren't trying to predict what is going to happen at a specific place and point in time. So they can’t produce a forecast for, say, March 15, 2077, or even tomorrow! Instead, climate models are used to determine how the average conditions will change - as in, will it be on average warmer or cooler, wetter or drier, in Tucson over the next 50 years? And this is information we need if we're going to plan for things like water shortages, or more frequent fires, or any of the other impacts on the Southwest that might result from the local effects of global climate change. To manage reservoir operations in Lake Mead, for example, the U.S. Bureau of Reclamation doesn't care about the exact conditions on March 15, 2077. They care whether over a 5 to 10 year period there will be enough water to meet the demands of Colorado River water users.

Using a suite - or "ensemble" of climate models can help sharpen the image somewhat. One model might project a 2.8°F change in 50 years (note: this is arbitrary). Another might say 3.1°F. By using 2, 4, 6, 10, or 15 of these models – all of which use slightly different approaches to represent atmospheric processes, we can begin to sharpen the focus of that predicted temperature change, precipitation change, or whatever variable is being looked at. Alternatively, an ensemble might consist of the same model run multiple times from a different starting point (say, using 1990 conditions, or 2010 conditions) – this will yield slightly different but realistic simulations of the future. For example, forecasts made by NOAA that indicate wetter or drier precipitation amounts, or hotter or cooler than normal temperatures for the next season, are largely based on an ensemble of future possibilities from a single model.

Ok, so that’s how climate models work. In my next post I’ll discuss how we can use a global climate model to project the North American monsoon into the future. Check back in a few days!