Testing the annular mode autocorrelation timescale in simple Atmospheric General Circulation Models

Idealized general circulation models, such as a dry primitive equation model driven by the Held and Suarez (1994, hereafter HS) forcing, are widely used to study the atmospheric circulation. Owing to their relative simplicity, they allow for long integrations often beyond the reach of comprehensive models. In particular, such models have proved to be useful tools for studying the coupling between the troposphere and the stratosphere (e.g. Nishizawa and Yoden, 2004), and for testing theories of intraseasonal variability as it relates to the annular modes and North Atlantic Oscillation (e.g. Gerber and Vallis, 2007). Here, it is shown that much caution is needed when performing long-time integrations with idealized models, as the "memory" of the system, quantified as the autocorrelation timescale of the annular mode, is surprisingly sensitive to model numerics and resolution.

For this study, two dynamical cores, one pseudo-spectral and the other finite-volume, are used to solve the identical HS set of equations with different numerics at different vertical and horizontal resolutions. It is found that the timescale of the annular model is unrealistically large in both models at low resolutions, such as those that are often used in studies involving long integrations with idealized models. Moreover, the spectral model is particularly sensitive to vertical resolution at modest horizontal resolution, triangular truncation at wavenumber 42, which is generally considered sufficient for resolving the extratropical circulation. The model time scale, in fact, becomes worse as the vertical resolution is increased, as is often done in studies of tropospheric-stratospheric coupling. Furthermore, the mean state and synoptic variability in these unrealistic integrations is deceptively reasonable. This highlights the need to test a model's variability on intraseasonal timescales, in addition to current diagnostics which focus only on the mean state and synoptic variability.