extratropical / tropical interactions in an GCM ideal predictability study

The role of interactions between the tropics and extra-tropics in regulating the growth of error in the tropical regions is studied in two Sets of idealized COLA T63 GCM predictability experiments. For each of 18 winters, ten pairs of simulations are run from observed winter initial conditions, with each pair differing only by a very small random perturbation applied globally (Set I) or in the deep tropics only (Set II). (Observed SST fields are specified). The growth of errors is assessed as a function of latitude and zonal wavenumber. The shape of the normalized error growth of 200mb u-wind, surface pressure and 850 mb temperature is quite distinct in the tropics compared to higher latitudes: the normalized tropical errors grow more quickly in short and medium range (up to 5 days) and thereafter grow more slowly (5 - 20 days). Surprisingly, however, the tropical errors continue to grow after the higher latitude errors have saturated. This effect is most distinct for planetary waves (zonal wavenumbers 1 - 3). The long tropical saturation times in a GCM which does not simulate a very robust Madden-Julian Oscillation implies a role for the extra-tropics in simulating tropical error growth.

The role of the planetary wave global structures found by Straus and Lindzen (2000) in the GCM results are assessed with wavenumber-frequency covariance calculations, and temporal wavenumber covariance analysis of the error fields. As for reanalyses, the GCM shows strong coherence between low frequency (periods of 30 - 60 days) upper level zonal wind fluctuations in the sub-tropics (35N) and in the tropics (15N) for planetary waves, although the variance in these scales is less than observed. The implied role of the extra-tropical / tropical coupling in the error growth experiments are being assessed by computing the coherence of the error covariances.

Comparison between the error growth in the two sets of experiments shows surprisingly little difference. In both experiments the (small) error in the first few days is confined to the tropics, consistent with very rapid short-time growth rates in the tropics. After a few days, mid-latitude error grows even in Set II because of error radiating from the tropics.

The role of resolution, damping, and the importance of organized tropical modes (MJO) in error growth will be discussed.