Submitted to: Journal of Geophysical Research
Publication Type: Peer reviewed journal
Publication Acceptance Date: 1/20/2012
Publication Date: 6/14/2012
Citation: Brown, P., Cordero, E., Mauget, S.A. 2012. Reproduction of 20th century inter- to multi-decadel surface temperature variablilty in radiatively forced coupled climate models. Journal of Geophysical Research. 117:1-15. Interpretive Summary: In the future sophisticated computer models may be used to predict decadal climate variation. However, before decision makers can confidently use decadal climate forecasts the computer models that make those predictions must be tested somehow. Here, a statistical and graphical method used previously to identify decadal climate variation in historical data records was adapted to test climate model’s abilities to reproduce observed temperature at 211 global grid locations during 1902-1999. The method samples annual temperature rankings over moving time windows, converts those samples to Mann-Whitney U statistics, and then normalizes the U statistics into Z statistics. The process is repeated using moving windows of varying duration to identify the most significant warm and cool regimes in the 98 year temperature records at each grid location. Those results were then compared with observed temperature records at each grid location to test the models ability to reproduce warm and cool periods during the 20th century. In this test of 12 climate models, 4 models were found to outperform many of the remaining models by a considerable margin. The dissimilarity in model performance was found to be tied to whether historical variability in solar forcing and the cooling effects of major volcanic eruptions were included in the computer simulations. These results suggest that future projections of temperature variation may be improved if they correctly account for these two important climate influences.
Technical Abstract: Coupled Model Intercomparison Project 3 simulations of surface temperature were evaluated over the period 1902-1999 to assess their ability to reproduce historical temperature variability at 211 global locations. Model performance was evaluated using the running Mann Whitney-Z method, a technique that highlights intra-to-multidecadal temperature regimes of arbitrary onset and duration in a time series. In addition to direct comparison with observation, a supplementary methodology was developed to score model performance relative to each models' own degree of unforced variability. These two metrics revealed distinct differences in model performance that were apparent despite of the existence of unforced variability. In particular, the top four models were found to outperform many of the remaining models by a considerable margin. Dissimilarity in model performance was found to be influenced by the differential inclusion of historical radiative boundary forcings. Notably, the three models that neglected solar and volcanic aerosol forcings scored worse than each of the models that did include these forcings. Based on these results, future projections of temperature variation may be improved if they are asymmetrically weighed toward the model/forcing combinations that preformed better in this analysis.