Statistical corrections to the NMC medium range 700 mb height forecasts

This paper examines four statistical procedures for correcting the NMC medium-range forecast (MRF) errors in the 700-mb heights over the Northern Hemisphere. The tests were designed to include normal operational constraints, such as frequent changes in the forecast model and the lack of long-term forecast histories on which to base estimates of systematic errors and climate drifts. It was found that the two statistical procedures which provided a substantial reduction in mean square error were a procedure based on simple regression correction at each grid point and one based on lagged-average correction with one day lag at each grid point.