W. M. Connolley, J. C. King and S. A. Harangozo
British Antarctic Survey, High Cross, Madingley Road, Cambridge CB3 0ET, UK
Email: wmc@bas.ac.uk
In this paper we compare numerical weather prediction analyses from four major centres - the Australian Bureau of Meteorology (ABM), the European Centre for Medium-Range Weather Forecasts (ECMWF), the American National Meteorological Centre (NMC) and the UK Meteorological Office (UKMO). Two of the analysis series (ECMWF and NMC) are "reanalyses", ie the data have recently been processed through a consistent, modern analysis system. We have somewhat more than a decade available for ECMWF, NMC and the UKMO analyses. The ABM analyses are potentially very useful since they are available for a longer period, 1973-1993. There is, however, distinct evidence of non-homogeneity in the ABM series with a change occurring at about 1980. The may be caused by the ABM analysis method changing over time, but studies using available station data suggest that the change is real.
We compare 500 hPa height fields from the four analyses and find that ECMWF, NMC and UKMO are generally consistent with each other (measured by the correlation coefficient of the departure of each field from the long-term monthly mean) whereas the ABM analysis is less similar. Comparison with station data and the FROST intensive observing period confirms this finding. The degree of agreement between the analyses has remained roughly constant since 1986 which indicates that the UKMO analyses do not have spurious changes over time even though they have not been reanalysed. Before 1986 the degree of agreement declines. We examine particular periods of disagreement to investigate their cause and to try to determine which scheme is to be preferred.