Cross‐equatorial error propagation: A numerical simulation

The numerical simulations of Baumhefner (1971, 1972) and Miyakoda and Umscheid (1973) have shown that a “wall” placed at or near the equator has a serious effect on Northern Hemisphere forecasts after 10–14 days. In practice, however, numerical models have available some information from the Southern Hemisphere. The question is posed, “How much information from the Southern Hemisphere is necessary to yield a forecast for the Northern Hemisphere which is more accurate than that obtained by integrating over the Northern Hemisphere alone?” A simple numerical experiment demonstrates that a global model in which only the largest scales of the Southern Hemisphere are known at initial time yields a more accurate forecast for the Northern Hemisphere than a hemispheric model.