A deep low pressure system was predicted to develop in the Gulf of Alaska,
affecting the US and Canadian west coast around 99020612. This cyclone
was apparently associated with a very high degree of predictability: The
NCEP ensemble indicated its development first at 11.5
days lead time, and the ensemble mean mean sea level pressure
forecast did not change much after the initial time of 99012700 (10.5
days lead time), with the deepest predicted closed isobar of the low
somewhere between 972 and 964 mb at 9.5
days, 8.5 days, and shorter
lead times. Note the predominantly blue and even light blue colors on all
of these forecast maps, indicating that the ncep ensemble spread, nomalized
here by the average ensemble spread over the preceding 30-day period, was
much lower than usual at these forecast ranges over a large area including
and surrounding the low pressure system, suggesting an especially
high degree of predictability. Despite the extreme nature of this event,
all these forecasts, as expected, verified very well against the verifying
analysis that had a central pressure of 968 mb. The probabilistic quantitative
precipitation forecasts gave a high probability of heavy precipitation
on the west coast with 10 days and
shorter (9 days,
8
days, etc) lead times. Note also that the ecmwf ensemble
mean forecast at 10 days and
shorter lead times (9 days, 8
days, 7 days, etc) were very
similar to the ncep ensemble mean. It is also interesting to note
that, for example, in the ensemble forecasts started on 99012900 (NCEP)
and 99012812 (ECMWF) there is more difference between the two ensemble
means at very short lead time
(12 and 24 hours respectively, see area of low pressure wave approaching
the west coast of the US, marked by green and yellow colors as highly uncertain,
with a maximum difference in the ensemble means of 7 mb) than at a much
longer lead time
(204 and 216 hours, respectively, see the less than 4 mb difference near
the center of the predicted large-scale low pressure system, marked as
highly predictable by the light blue colors). The high degree of predictability
of the low pressure system is also highlighted by the fact that the estimated
uncertainty associated with the center of the low at
8.5 days lead
time is on the order of 2-3 mb (see light blue colors indicating
the actual, not the normalized ensemble spread in this figure) -
the same range of uncertainty that unpredictable systems are associated
with at
initial time - like the 2-3 mb spread
over the low pressure wave referenced above, approaching the west coast
in the same forecast at initial time . Though the large-scale low pressure
system in the Gulf of Alaska may be a weather event with exceptionally
high predictability, verification statistics indicate that the ensemble
can correctly identify 10-15% of all forecast cases as highly predictable.
In these cases the 12 days lead time forecasts are as reliable as the
"least predictable" 10-15% of cases at day 1. This is the power of
the ensemble: it lets us know in advance how much one can trust the forecasts.
On the 27th of January 1999, one could have made a 10.5 days lead time
forecast with the same (or higher) degree of confidence as sometimes 1-day
forecasts are associated with.