ISSN:
1436-5065
Source:
Springer Online Journal Archives 1860-2000
Topics:
Geography
,
Physics
Notes:
Summary Recent research in dynamical extended-range prediction at the UK Meteorological Office (UKMO), based on 40-day integrations of a global 11-level general circulation model, is described. The forecast anomaly correlation scores, calculated with respect to some set of background atmospheric normals, contain a significant contribution due to differences between the normals and the true atmospheric climate for the years containing the experimental initialization dates. This contribution varies according to the choice of normals. The best set to use in practice are those which minimize the measured scores, since they are closest to the true climate. The model's own climatology is sufficiently realistic for it to be suitable for long-range forecasting. However, significant climate drift still occurs in all seasons, and empirical correction for this increases the model's skill substantially, although the use of dependent data exaggerates the improvement some-what. On average, the skill of winter and spring forecasts exceeds that of summer and autumn cases for days 1–15 and 6–20, for the domain 30–90° N. Although the mean skill remains well above zero throughout the forecast period, there are few cases of high skill, on the hemispheric scale, at extended-range. However, study of local skill, over a region centred on the UK, shows that the model's ability to forecast surface pressure anomalies compares favourably with that of the experimental long-range forecasts produced at UKMO using statistical forecasting techniques and medium-range dynamical predictions. The major improvement is for days 6–15. Based on the anomaly correlation score, encouraging results are obtained concerning the frequency with which the degree of local skill reaches a potentially useful level, and the prospects for predicting this skill in advance. However, further analysis using alternative skill scores is required to confirm these results.
Type of Medium:
Electronic Resource
URL:
http://dx.doi.org/10.1007/BF01027468
Permalink