Diagnosing remote origins of forecast error: relaxation versus 4D‐Var data‐assimilation experiments

Abstract

AbstractTwo methods that can be used to diagnose possible remote origins for forecast error are compared. The idea is artificially to suppress the development of forecast error in certain parts of the globe (e.g. the Tropics) during the course of the integration and to analyze the influence that this has on forecast skill in remote regions (e.g. the extratropics). The first, computationally relatively cheap, method involves relaxing the European Centre for Medium‐Range Weather Forecasts (ECMWF) model towards analysis data during the forecast. The second, computationally much more expensive, method involves running the ECMWF 4D‐Var data‐assimilation system with assimilation of observations in certain regions only. The two methods are compared by studying the impact that forecast‐error reduction in the Tropics and the East Asian–Western North Pacific (EAWNP) region has on medium‐range forecast skill in remote regions. For both regions the two techniques yield similar results. Reduction of tropical forecast error leads to the improvement of medium‐range forecast skill in the Northern Hemisphere extratropics, especially over the North Pacific and the North Atlantic. Forecast‐error reduction in the EAWNP region is beneficial further downstream up to North America; the EAWNP region has little impact on medium‐range forecast skill over the North Atlantic and Europe. Copyright © 2011 Royal Meteorological Society