Abstracts
SEARCH Open Science Meeting
October 27, 2003
Seattle, Washington, USA
The Nature, Measurement, and Modeling of Feedbacks
Judith A. Curry1
1School of Earth and Atmospheric Sciences, Georgia Institute of Technology, ES&T Room 1168, Atlanta, GA, 30332, USA, Phone 404-894-3955, Fax 404-894-5638, curryja@eas.gatech.edu
This talk provides some thoughts on framing the feedback issues for SEARCH and useful strategies for addressing these issues. “Feedback” is a $10 word with a very specific meaning, but it is often used to denote a forcing (rather than feedback) or to refer to any physical process (a “feedback process"). Such inappropriate uses of “feedback” in science and implementation plans can lead to confusion, untestable hypotheses, unachievable objectives, and ineffective strategies.
The Science Plan for the U.S. Climate Change Research Program is used as a reference point for considering the issue of feedback in the context of climate. Several examples are presented in the context of SEARCH on using the concept of feedback to design observing systems, model feedbacks, and assess the impact of imperfect models with feedbacks on decision making. “Feedback” should not be used as justification for endless process studies; feedback should only be used to justify process studies that include consideration of an appropriate selection of variables that are related conceptually in a complete feedback loop. Design of a long-term monitoring network should ideally include consideration of a variety of variables that are linked in conceptual feedback loops and that can be assimilated into models for a more complete representation of the system.
To illustrate the difficulties in attempting to appropriately model feedbacks in a climate model, an example is presented that illustrates the impact of various choices in the parameterization of sea ice albedo. Climate models have a large number of degrees of freedom and multiple and interconnected feedback loops; as a result Monte Carlo (ensemble) prediction methods are needed to provide a quantitative measure of uncertainty. Coupling of submodels that interact nonlinearly adds considerable uncertainty to the model and most likely increases the need for larger model ensembles to provide useful predictions. This implies that there is a tradeoff in computing capacity to be considered in terms of increasing model complexity and adding additional subsystems, versus the size of the ensemble. Predictions based on ensembles can in principle provide far more useful information to decision makers than a single simulation that might provide a completely irrelevant picture of the future. A summary is given of recommendations regarding appropriate practices to utilize the concept of feedback for SEARCH.
Abstract Categories: Keynote
Back to main abstract page
To view the list of records returned in your search you can use your browser's
back button, or perform the same search again:
| 