|
There are different
techniques to highlight the larger, more predictable scales. The
most consistent way is to use the Ensemble Prediction System (see
ch. 9). But for non-EPS users there are plenty of useful
alternatives through different types of smoothing or averaging.
Spatial smoothing: The
average of the D+4, D+5 and D+6 forecasts from the same run might
serve as a useful complement to the proper D+5 forecast. Similarly
might the average of the current D+5, yesterday's D+6 and the D+7
from before yesterday highlight the large scale flow patterns.
For medium range
purposes the geographical area on display can be enlarged. This
will automatically suppress the impression of the smaller scales.
This is also in line with the speed by which atmospheric systems
influence each other downstream. Forecast beyond three days are
best understood when also the western part of the Atlantic and the
easternmost part of North America are included. For forecasts
beyond five days the whole of the North American continent and
easternmost Pacific ought to be included.
Temporal smoothing:
Mean temperatures are always more predictable than instantaneous
values. The same is true for maximum and minimum values. The reason
is that we have compromised with the time of the event to acquire
knowledge about the value.
The predictability will
increase with increasing time window, Rainfall can for example be
more skilfully forecast if accumulated over two days rather than 12
or 24 hours. The fact that the deterministic forecast has indicated
an event of wind speed >20 m/s becomes more significant if it is
attributed to a three day period than to the exact time provided by
that particular forecast.
The advantage of
condensing information by spatial or temporal filtering has, of
course, to be paid by the occasional risk of losing information
which, in hindsight, might have been important.
|
