Simulation of seasonal to decadal variations by CPC's new climate model

The Climate Prediction Center is currently undertaking a massive evaluation on the capacity of its new climate model in the long-lead climate prediction. One of the foci is to develop the probabilistic forecasts. Another is to evaluate the skill level of the model through massive hindcast experiments. The decadal variation capability is also under investigation. For these purposes, and numerous others, three major categories of integration have been considered. The model experiments were carried out with T42L28 resolution. They are: (1) An ensemble of 10 AMIP 50-year long regular integrations. For examining the impact of model changes, several other 50-year long runs were also conducted. (2) An ensemble of 10 6-month long hindcasts (in concert with the forecasts cited below) for every of the last 21 years (1979 to 1999). The observed SST was used for these hindcasts. (3) An ensemble of 20 6-month long forecasts for every new month as time goes along. Although the model still has systematic biases, the long-lead hindcasts do show substancial skills. When presenting in probabilistic forecast format, even a meager 20-member ensemble forecasts suggests useful dynamical guidance. A larger ensemble is expected to yield even more useful probabilistic forecasts. Also the wintertime forecasts are expected to yield much better results. In addition to short-term climate variability, the model also produces decadal variations ----- the super warming of the 80s and 90s. In general, the spread of climatology among AMIP runs is small enough to allow the study of variability on various sub-50-year time scales with one common climatology. The decadal variations is evaluated with the T850 and Z500 fields. The model shows clear decadal variations in both the lower and the middle troposphere. Further more, when validated with the NCEP/NCAR reanalyses, the simulations look remarkably good, especially during the latest decades when the re-analysis data are more trustworthy. The skill scores by anomaly correlation (AC) over the global extent are above 0.4 for each of the five decades considered. For the tropical Pacific sector, the AC scores are larger than 0.9, suggesting a strong association between the decadal variations with the tropical Pacific SST's decadal anomalies.