Sunday, June 24, 2007

The National Academies | News | EPA Models Should Undergo 'Life-Cycle' Evaluation

The National Academies | News | EPA Models Should Undergo 'Life-Cycle' Evaluation:

"Although the computer-model results behind many U.S. Environmental Protection Agency regulations will always be constrained by computational limitations, assumptions, and data gaps, models can be improved through continuous evaluation, says a new National Research Council report. Peer review should be considered at each stage of a model's life, uncertainties should be communicated clearly, and all stakeholders should be ensured access to information about the models, the report adds."

Comment: On the one hand, computer models can be the best available tool to make predictions about environmental problems and to evaluate alternative courses of action. On the other hand, they are too often opaque and give inaccurate predictions.

The modelers have to get the theory right and get adequate data, and then get the model right. Validation is important, but so too is expert judgment. Indeed, the discipline imposed on modelers who know that they must subject their work to peer review is itself important.

No comments: