Abstract
An observational and modeling study is made of tropical-extratropical interactions on time scales relevant to medium and extended range forecasting. First, an empirical orthogonal function (EOF) analysis is made of outgoing longwave radiation (OLR) in the tropics over seven winters. Having removed the seasonal cycle and interannual variability, the two leading EOFs describe the 30–60 day oscillation. A composite of extratropical 500 mb geopotential height correlated simultaneously with this mode of tropical variability is constructed. In its two phase-quadrature components, this composite has significant projection onto the Pacific/North American teleconnection pattern and onto the North Atlantic oscillation pattern, respectively.
The 500 mb height composite is compared with the Simmons, Wallace and Branstator (SWB) mode of barotropic instability, which has similar periodicity and similar spatial structure in both its phase-quadrature components. A simple theoretical analysis shows that the SWB mode can be strongly excited by a periodic forcing in the tropics whose spatial structure resembles the oscillation in convective activity described by the first two EOFs of OLR. This is confirmed in a barotropic model integration, which is forced using the observed EOFs of OLR. The model response in the extratropics compares well with the observed composite oscillation in 500 mb height.
In the final phase of this study, the ECMWF model has been integrated over four wintertime 20-day periods. For each period, five integrations have been performed; a control forecast, an integration in which the tropics are relaxed towards the verifying analysis, an integration in which the tropics are relaxed towards the initial analysis, an integration in which the extratropics are relaxed towards the verifying analysis and finally an integration in which the extratropics are relaxed towards the initial analysis. The four initial dates were chosen on the basis that in the succeeding 20 days, observed OLR and extratropical height provided a reasonable realization of each separate quarter of the composite oscillation.
It was found that in the extratropics, skill scores in the range of 11–20 days were noticeably improved, particularly over the Pacific/North American region (consistent with expectations from the data analysis). The mean geopotential height error in the extratropics; i.e., the error averaged over the four experiments, was also reduced (mainly in the Pacific area) when the model tropical fields were relaxed towards the verifying analysis. Indeed, maps showing the time evolution of geopotential height from the first 5 days of the forecast were generally correlated with the differences between the integrations with tropics relaxed to the verifying analysis and to the initial analysis indicating a link between tropical and extratropical low-frequency variability.
The impact of the extratropics on the tropics was also studied where it was shown that the largest response was on the nondivergent component of the wind over the tropical east Pacific. Tropical skill scores and model systematic error in upper tropospheric streamfunction were significantly improved with the extratropics relaxed to the verifying analysis. By contrast, extratropical relaxation had a much smaller impact on the divergent component of the tropical wind.
