Credibility of Regional Climate Model Projections of Future Climate: Issues and Challenges
Linda Mearns is an expert in regional climate change and its impacts. She draws on computer simulations to develop scenarios for decision makers who are preparing for the societal and environmental impacts of climate change. Mearns also investigates the effects of climate change on agriculture and on human health, particularly in relationship to the frequency of extreme weather events such as heat waves.
The issue of credibility of regional climate model simulations of the future has been discussed for a long time, but concern about how to properly quantify the relative credibility of different simulations has become of considerable interest over the past decade. One can discuss credibility from the point of view of climate models in general (e.g., what does it mean to have reliable projections of future climate change) or from the point of view of a particular ensemble of models (e.g., should we have more confidence in certain ensemble members (as opposed to others) based on some evaluation metric). It is also related to the concept of ‘added value’ regarding regional climate model simulations. This is a an important issue in quantifying uncertainty, since differential credibility often leads to differential weighting of climate model results, and differential weighting is often used when producing PDFs of regional climate change (e.g., for seasonal temperature, precipitation, or both simultaneously). Basically there are three traditional positions that can be taken: 1) continue with the ‘ultra-democracy’ approach, (i.e., one model one vote), which has been the standard approach in the IPCC, 2) eliminate simulations that are believed to be too poor in quality to contribute any useful information about future climate; or 3) differentially weight the various simulations based on some metric of simulation quality. In the extreme, method 3 can become indistinguishable from 2 if some models are down-weighted to the point of essentially exerting no influence over the final measures of uncertainty. In this talk I will discuss these various methods in the climate literature as well as suggest an additional approach to establishing credibility that is more climate-process based and provide examples of this from the North American Regional Climate change Assessment Program (NARCCAP).
Linda Mearns has served as lead or co-convening lead author on several Intergovernmental Panel on Climate Change reports and shared in the 2007 Nobel Peace prize. Mearns is director of the Weather and Climate Impacts Assessment Science Program and a senior scientist in NCAR’s Institute for Mathematics Applied to Geosciences.