EXTENDED-RANGE FORECASTING BY WEATHER TYPES 
by the author durmg World War II [11]. Since then, 
additional work has been done upon them [1, 13]. In 
line with the character of this article, a few very general 
remarks will be included covering this approach. 
At times, particularly during meridional flow periods, 
certain systematic movements are apparent. Curiously 
enough, such motions frequently exhibit a westward 
displacement or retrogression, not so much of the in- 
dividual broad-seale crests and troughs aloft but of 
the amplitude or energy of the disturbance. Thus, 
following the development of a trough and crest pair, 
the next major trough or crest-trough pair upstream 
will increase in amplitude. The rate of the propagation 
is some seven degrees of longitude per day in winter 
[1]. Sometimes this upstream effect appears to be prop- 
agated mainly through the action of the subtropical 
easterlies; at other times the polar easterlies appear to 
be the controlling factor. Other evidence [13] indicates 
that systematic distortions of the westerlies develop 
slowly over a period of two or more weeks both up- 
stream and downstream from a major disturbance such 
as a blocking high. Recent theoretical studies [25, 29] 
of energy propagation indicate the possibility of a 
number of modes of energy dispersion and clearly 
indicate the possibility of interaction on a hemisphere- 
wide basis. This confirms much recently obtained 
synoptic evidence of the dependence of North American 
weather developments on events occurring elsewhere 
in the Northern Hemisphere. 
Considerations of the type just mentioned have been 
handled statistically through the use of the data ap- 
pearing in the forty-six-year weather-type catalogue. 
Naturally the first and simplest statistical aid to be 
developed from the catalogue was merely the deter- 
mination of the seasonal expectancies of the various 
types. Next came the expectancies for various types 
in one region following a given type. Later, the North 
American types were arranged in a certain order de- 
pending upon such items as the latitude of the Pacific 
anticyclone, strength and location of polar outbreaks, 
and natural statistical relationship, and numbers were 
assigned in accordance with this order. Then type- 
prediction regression formulas were developed [8] 
whereby future types could be predicted on the basis of 
the past sequence of types in one region. Success was 
limited. It is believed that the greatest shortcoming 
of these earlier regression formulas lay in their dis- 
regard of present and past types in other zones. More 
recently the type data for the five zones have been 
punched on International Business Machine sorting 
cards [1], and prediction schemes based upon present 
and past types in each of the five zones will be 
developed. 
The use of the type catalogue in analogue selection is 
of course obvious. It is highly efficient in this respect 
but selections must be made with extreme care in order 
to avoid the above-mentioned shortcomings of “thumb- 
print” analogues. In a theoretical study [8] of ‘‘thumb- 
print” analogue prediction schemes it has been shown 
that the weakness of this system compared to a gener- 
alized prediction scheme lies in the fact that it is not 
possible to consider direct and cross matches between 
839 
analogue elements with different time lags. The ana- 
logue scheme thus assumes that weather is a one-dimen- 
sional function of time. 
In statistical investigations using the type catalogue 
it has been found necessary to consider processes of 
much greater duration than those covered in the usual 
extended-period considerations. Some rather interest- 
ing results have been found in studies of seasonal char- 
acteristics. An unpublished study shows that in a 
normal season the extreme meridional flow types occur 
frequently in the autumn but that, as winter ap- 
proaches, the frequency of their occurrence drops off 
EXTREME ZONAL FLOW TYPES (WINTER) 
DAYS 
DAYS 
| 
! 
| 
| 
| 
| 
AL FLOW TYPES (WINTER) 
EXTREME 
MERIDIO 
| 
| 
| 
| 
| 
| 
| 
| 
l 
| 
| 
| 
| 
| 
| 
ANNUAL RELATIVE SUNSPOT NUMBERS 
40 
20 
{o) 
NEVIS) 1923 1927 1931 1935 1939 1943 
TO TO To To To To To 
1920 1924 1928 1932 1936 1940 1944 
YEAR 
Fie. 2.—Curves showing the frequency of meridional (mid- 
dle curve) and zonal (upper curve) flow types during the 
winter seasons 1919-20 through 1943-44. The lower curve gives 
the annual relative sunspot numbers during the same period. 
rapidly and the extreme zonal flow types (/) increase 
in frequency. Again, in the spring the frequency of the 
meridional flow types increases and the zonal flow 
type decreases. Apparently, cross-westerly heat trans- 
port and general lateral mixing are greatest in the spring 
and the autumn. In certain years there is a distinct 
tendency for the extreme meridional flow types to 
continue with high frequencies on into the winter 
season. Since winter meridional flow types exhibit 
stronger abnormalities than those of autumn and 
spring, this leads to quite exceptional weather (v7z., 
January, 1937 and 1949) and for this reason the phe- 
nomenon was studied in detail. The effect can best be 
