IPCC AR4 Ensemble Statistics

IPCC AR4 Ensemble Statistics

The 20c3m experiments are forced with historical green house gas forcing as well as the time varying ozone, sulfate, volcanic aerosols, and solar output for the 1900-2000 period. This analysis uses 20th century experiments from 10 models. The B1, A1B and A2 experiments are forced with a predicted green house gas forcing scenario for the 2000-2100 period. There are 10 models for the A1B scenario, 9 for the B1 scenario and 8 for the A2 scenario. Click here for details on what models are included in each scenario.

For models with multiple realizations of the same experiment (ensembles), these ensemble members are first averaged together to get a mean for that model before computing the multi-model (ensemble) mean. It is desirable to perform multiple realizations because small differences in the climate system, for example slight displacements in high and/or low pressure systems, grow with time so that after a few weeks the atmospheric state will differ considerably between simulations. (This is sometimes referred to as the butterfly effect and why one can't make specific weather forecasts beyond about 10 days.) By performing a number of ensembles runs and then averaging across them suppresses the internal variability to highlight the change in climate due to greenhouse gases or other forcings.

The observations are constructed in the manner described by Fritch et al. (2002) using daily station data from the Global Historical Climate Network and interpolating the resulting extremes' indices to a T42 grid using inverse distance weighting.

The variables are based on daily values over the course of the annual cycle; they are not broken out by season.

Output from each model is either on or interpolated to a common T42 grid (~2.8 x 2.8 degrees) to facilitate the model inter-comparison. More details on the IPCC scenarios can be found here.


Problems? Contact PSL Web and Data Group