Abstract
A large ensemble of multidecadal atmospheric general circulation model (AGCM) simulations is examined to determine a quantity central to the model’s potential predictability—the fraction of the simulated monthly air temperature (T2M) variability that is tied to the imposed sea surface temperature (SST) boundary conditions as opposed to the background atmospheric noise. Combining this information with ensemble simulation data from other AGCMs in turn allows an intermodel comparison of two separate quantities: model potential T2M predictability and the underlying (in the absence of noise) teleconnections between SSTs and continental T2M. To a large extent, the models tend to agree with each other regarding both—they all show, for example, the expected highest predictability in the tropics as well as low potential predictability in central Asia, and, particularly for DJF and MAM, they show very similar ocean–land teleconnections throughout the Americas. However, the models do show some differences in teleconnections, indicating room for model improvement that could, in principle, lead to benefits for long-term prediction. Importantly, by combining the model results with observational temperature data, we provide a new estimation of real-world predictability, a property of Nature that is not directly observable. The latter results suggest, with caveats, that for monthly T2M in the tropics, AGCMs tend to overestimate the ratio of predictable variance to noise-derived variance.
© 2025 American Meteorological Society. This published article is licensed under the terms of the default AMS reuse license. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).
