An evaluation of UKMO one-month ensemble forecasts of MSLP in the Southern Hemisphere |
| |
| Authors: | R I C C Francis A B Mullan J A Renwick |
| |
| Institution: | (1) National Institute of Water and Atmospheric Research Ltd. Wellington, New Zealand, NZ |
| |
| Abstract: | Summary ?We evaluate United Kingdom Meteorological Office (UKMO) one-month ensemble forecasts of mean sea-level pressure (MSLP) in
the southern hemisphere (SH) to 60° S, with a special focus on their utility near New Zealand (NZ). There are 105 9-member
ensembles, at approximately two-week intervals, between 1995 and 1999. Each forecast is averaged over two successive 15-day
periods and verified against the NCEP/NCAR reanalysis data set. Compared to climatology, the skill of the ensemble mean is
slightly positive in days 1–15, and slightly negative in days 16–30. Skill near NZ is slightly lower than the SH averages.
For SH-scale circulation patterns (as seen in the first few principal components), skill is greater than for most individual
grid points, but is still negligible or negative in days 16–30. Moderate skill-spread correlations (ρ ≈−0.5) were found for
some skill scores. The way that skill varies with season and the Southern Oscillation Index is consistent with other research
but not statistically significant for this small data set. Probabilistic forecasts of low and high pressures have skill similar
to that of the ensemble mean. The ensemble spread is generally too small, in that the analysis lies within the ensemble less
often than the theoretically optimum value of 80% of the time. Measured as a fraction of the natural variability, the spread
increases substantially with time and latitude: it is less than 0.5 near the equator in days 1–15, and takes values near 1
only at higher latitudes during days 16–30. The initial sequential structure of the ensembles (a consequence of the use of
time lags in their genesis) is still apparent in days 1–15 but has disappeared by days 16–30. Three potential alternatives
to the ensemble mean were all found to have less skill than it.
Received June 17, 2001; revised July 4, 2002; accepted November 22, 2002
Published online March 17, 2003 |
| |
| Keywords: | |
| 本文献已被 SpringerLink 等数据库收录！ |
|
