Reduction of systematic forecast errors in the ECMWF model through the introduction of an envelope orography

Abstract Day 1 forecast errors in the ECMWF model geopotential height fields during wintertime show a distinctive and highly reproducible signature with negative biases over the major mountain barriers. These biases are largest on days when the 500mb flow over the mountains is strong and they tend to shift position from day to day so as to remain close to the region where the jet stream crosses the mountain barrier. These systematic errors evolve through the forecast interval until by day 10 they assume an equivalent barotropic structure which strongly resembles the upper‐level stationary wave pattern, but has the opposite sign, which indicates that the model does not have sufficient (presumably orographic) forcing to maintain the stationary waves. The time evolution of the systematic error pattern is investigated by means of a series of experiments with a barotropic model, in which the observed mean wintertime 300 mb flow is perturbed by a steady forcing derived from the day 1 forecast error pattern. The day 10 error patterns, as simulated by the barotropic model, bear a strong similarity to the observed ones. The forcing in the vicinity of the northern Rockies makes a particularly large contribution to the simulated day 10 error pattern whereas that in the Himalaya region appears to be relatively less important. The impact of an enhanced orography upon the climate of the ECMWF model is investigated by carrying out a pair of extended integrations out to 50 days. The control run is based on the conventional average‐type orography used for operational forecasting; the other run is based on an ‘envelope orography’ constructed by adding to the conventional, grid‐square averaged, orography an increment proportional to the standard deviation of the sub‐grid scale variance. Results for this one, rather short pair of integrations suggest that the envelope orography may be capable of yielding a more realistic simulation of the observed wintertime flow pattern, particularly with respect to features in the Pacific and western North American sectors. Certain aspects of the zonally averaged circulation are also more realistic in this envelope simulation. A series of 21 successive ten‐day forecast integrations has also been carried out with this envelope orography from February 1982 data. Results indicate that the introduction of the envelope orography results in a slight degradation of the short‐range forecasts together with a distinct improvement of the forecast beyond four days. The beneficial impact towards the end of the forecast interval appears to be large enough to increase the forecast usefulness by perhaps as much as half a day.