Abstract:
|
The climate system is highly complex (nonlinear interactions between atmosphere and ocean, atmosphere and land, etc.). So computer models have been developed to represent this system which are also rather complex and primarily deterministic. When attempts are made to forecast future weather, climate, and even long-term climate change using such numerical models, a major obstacle is the difficulty in quantifying uncertainty. Techniques that have been used to quantify the uncertainty in weather and climate forecasts are reviewed, including subjective, statistical adjustment of model output, as well as the generation of multiple "ensembles." Presently, the climate change policy arena is particularly contentious, with much of the controversy centering on uncertainty. Unfortunately, the statistics community has not played much of a role yet. Statistical challenges are identified, such as devising stochastic approximations to deterministic computer models so that uncertainty analysis can be more routinely applied.
|