Geography Reference
In-Depth Information
Fig. 13.11
Anomaly correlation of 500-hPa height for 3-, 5-, and 7-day forecasts for the ECMWF
operational model as a function of year. Top and bottom of each band correspond to
Northern and Southern Hemispheres, respectively. Note that the difference in skill between
the two hemispheres has nearly disappeared in the past few years, due to the successful
assimilation of satellite data. (Courtesy of Dr. Adrian Simmons, updated from Simmons
and Hollingsworth, 2002.)
An indication of the skill of the present generation of global forecast models is
given in Fig. 13.11, in which forecast skill is plotted in terms of the
anomaly corre-
lation
of the 500-hPa geopotential height field, defined as the correlation between
observed and predicted deviations from climatology. Subjective evaluations sug-
gest that useful forecasts are obtained when the anomaly correlation is greater
than 0.6. Thus, over the past 20 years the range for which ECMWF forecasts show
useful skill has increased from 5 days to more than 7 days. There can, of course,
be quite wide variations in skill from one situation to another reflecting variations
in the degree of predictability from one atmospheric flow regime to another.
As mentioned earlier, it has become common for operational forecast centers to
assess the likely skill of medium-range predictions by carrying out several parallel
runs of a forecast model starting from slight perturbations of the initial conditions
determined by the data assimilation process. For example, in the ECMWF ensem-
ble prediction system as of 2003, 50 runs are made for each analysis time with a
T255, 40 level model. In each of the runs, the analysis is perturbed by an amount
that is within the uncertainty limits of the analysis. The perturbations are based on
the method of
singular vectors
, which is designed to identify the regions of the
atmosphere where small uncertainties are likely to magnify rapidly, and hence will
tend to have a large impact on the forecast.
