Probability Distributions for U.S. Climate Change Using Multi-Model Ensembles
Abstract
Projections of future climate change vary considerable among different atmosphere ocean general circulation models (AOGCMs) and climate forcing scenarios, and thus understanding of future climate change and its consequences is highly dependent upon the range of models and scenarios taken into consideration. To compensate for this limitation, a number of authors have proposed using multi-model ensembles to develop mean or probabilistic projections of future climate conditions. Here, a simple climate model (MAGICC/SCENGEN) was used to project future seasonal and annual changes in coterminous U.S. temperature and precipitation in 2025, 2050, and 2100 using seven AOGCMs (CSIRO, CSM, ECHM4, GFDL, HADCM2, HADCM3, PCM) and the Intergovernmental Panel on Climate Change's six SRES marker scenarios. Model results were used to calculate cumulative probability distributions for temperature and precipitation changes. Different weighting schemes were applied to the AOGCM results reflecting different assumptions about the relative likelihood of different models and forcing scenarios. EQUAL results were unweighted, while SENS and REA results were weighted by climate sensitivity and model performance, respectively. For each of these assumptions, additional results were also generated using weighted forcing scenarios (SCENARIO), for a total of six probability distributions for each season and time period. Average median temperature and precipitation changes in 2100 among the probability distributions were +3.4° C (1.6-6.6° C) and +2.4% (-1.3-10%), respectively. Greater warming was projected for June, July, and August (JJA) relative to other seasons, and modest decreases in precipitation were projected for JJA while modest increases were projected for other seasons. The EQUAL and REA distributions were quite similar, while REA distributions were significantly constrained in comparison. Weighting of forcing scenarios reduced the upper 95% confidence limit for temperature and precipitation changes under the EQUAL and REA distributions, but had little effect on the REA distributions. Although these methods do not represent a completely unbiased sample of possible climate outcomes, multi-model ensembles from simple climate models may be useful for better incorporation of climate uncertainty into impact assessment and other applications.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2004
- Bibcode:
- 2004AGUFMGC51D1077P
- Keywords:
-
- 6309 Decision making under uncertainty;
- 3354 Precipitation (1854);
- 1610 Atmosphere (0315;
- 0325);
- 1620 Climate dynamics (3309);
- 1630 Impact phenomena