Climate Models as Predictive Tools

One of the biggest challenges in addressing global climate issues is the lack of an experimental control. If we had a twin solar system with an exact duplicate of Earth, then we could experimentally determine the effects of doubling atmospheric carbon dioxide, for example. Unfortunately, we are limited to a single planet with which to experiment.

Since we cannot predict the outcome of anthropogenic (or mesopogenic) influences on climate with an experimental control, we turn to computer models as a tool for understanding the climate system. In general, climate models are mathematical descriptions of atmospheric and oceanic processes; they range in complexity (e.g., include soil processes, vegetation effects, etc.) and in resolution (what are the smallest details that can be resolved?) depending on the intended application. Some complex models strive to fully represent the climate system–and eventually be as useful as an experimental control for Earth. Yet this approach to climate modeling has two flaws:

1) No matter how good the model resolution, there will always be some process too small for the model to resolve. (In fact, the only way to escape this would be to resolve details down to the quantum level–and then the non-determinism of the quantum level would pose an entirely new problem!)

2) Even with a comprehensive model of the climate system, we do not know the initial conditions well enough. (In order to predict the behavior of the climate, we need to know it’s previous state.)

Many people seem to expect climate models to serve as a replacement for a control case, yet I doubt that this function for climate models will ever be feasible. Part of this perception stems from our use of models in weather prediction. Contrary to stand-up comedians, meteorologists have actually made tremendous strides over the past 30 years in the ability to provide accurate weather forecasts several days in advance. The weather system operates on a short enough time scale that these forecasts are possible and even useful in emergency response and disaster prevention.

The success of weather prediction, then, can create the perception that the same accuracy should be possible for climate prediction. Since climate operates on longer time scales than weather, the limited supply of climate data over time makes it difficult to achieve this degree of predictability. (If we had 100-200+ years of complete climate data, then we could make better long-term climate forecasts.)

This does not mean that models are useless in understanding climate trends, though. Though we may never realize a model capable of duplicating the climate system, models are certainly a crucial tool in understanding the underlying mechanisms.

Stay tuned for part 2: Climate Models as Diagnostic Tools