EXTENDED-RANGE FORECASTING 
THE EXTENSION OF FORECASTS TO LONGER 
PERIODS 
The basis of the preparation of most medium-range 
forecasts is essentially synoptic in nature, although 
there is a substantially greater infiltration of statistical 
treatment than in short-range forecasting. As longer 
periods are considered, there appears to be greater and 
greater reliance upon statistical techniques at the ex- 
pense of synoptic techniques. Yet the possible exten- 
sion of synoptic techniques to forecasting problems for 
periods longer than a week has been explored very little. 
In view of the fact that some statistically significant 
success has been achieved in applying synoptic methods 
to monthly mean maps (described below) it would 
appear that meteorologists, particularly dynamic me- 
teorologists, might be able to make rewarding contribu- 
tions by directing some of their efforts to this problem. 
As stated earlier, this article is concerned only with 
those long-range forecasting methods which can demon- 
strate skill in the rigid statistical sense over reasonably 
long periods of time. This means consistently predict- 
ing occurrences with greater success than the climatic 
probabilities. This criterion, while harsh, is after all 
the only rigid means of differentiating between a suc- 
cessfully operating forecasting method and a theory. 
When the field of long-range forecasting methods is 
scanned through this revealing eyepiece, there appears 
surprisingly little that can be discussed. In part, this 
reflects the unavailability of rigid statistical verifica- 
tions of forecasts, but more probably it mdicates the 
overwhelming complexity of the long-range forecast 
problem. 
Public claims of success in long-range forecasting 
should be backed by statistically sound verification. It 
is high time that professional societies bring the spot- 
light upon those whose claims cannot be substantiated. 
Meteorology, no less than medicine, can ill afford to 
harbor quacks. This is not meant to detract from the 
painstaking work of those who have spent years, if 
not decades, in the quest for forecasting tools. Their 
work must be judged on other grounds than its imme- 
diate application to forecasting problems. 
Before closing this article the author feels compelled 
to dwell a bit further on the question of the extension of 
synoptic-statistical techniques to longer-range forecast- 
ing. In 1942 he began an experiment in the preparation 
of thirty-day forecasts, the aim of which was to explore 
the possibility of extending mean circulation methods 
then used in five-day forecasting to monthly periods. 
Of course, it was soon recognized that the normal state 
of the circulation and its rate of change become in- 
creasingly important m such work. In essence the prob- 
lem was found to be similar to the shorter (five-day) 
problem. At the start two highly pertinent questions 
arose: 
1. Do the thirty-day mean circulation patterns de- 
termine the average weather conditions (anomalies) for 
the month? The answer is an unqualified yes. The scien- 
tific basis for this answer has been briefly discussed and 
is expanded upon elsewhere [7, 10, 13]. 
2. Is there a rational continuity to the great centers 
811 
of action when treated ona thirty-day mean basis? Here 
again, experience gained during the last decade, in 
part reported on im the literature [8, 12, 13], indicates 
that there is a rational continuity when thirty-day 
charts are studied carefully with the aid of kinematical 
and physical techniques. This continuity is discussed 
twice monthly in a routine fashion by workers in the 
Extended Forecast Section of the U.S. Weather Bureau 
with as much conviction as are the movements of 
cyclones and anticyclones at daily map discussions. 
The detailed method for projecting the centers of 
action into the future of the next thirty days is naturally 
too involved to be discussed here, but its basis may be 
briefly stated this way: 
1. The past and current rate of evolution of mid- 
tropospheric flow patterns is assessed with the aid of 
the tendency method described earlier. These methods 
automatically introduce the effect of normals changing 
with time. 
2. After kinematic displacements of the major fea- 
tures of the patterns (ridges, troughs, and centers) are 
made, it must be decided if these kinematic displace- 
ments are in harmony with vorticity (thus wave-length) 
principles. In other words, within certain speeds of the 
mid-troposphere westerlies the atmosphere prefers a 
certain spacing between ridges and troughs. If kine- 
matically computed displacements introduce no severe 
contradictions into these physically harmonious upper- 
air patterns, the forecast is relatively straightforward. 
If contradictions arise—as is often the case—certain 
large-scale features, whose immediate behavior is more 
clearly outlined in their profiles and tendency fields, 
must be chosen as anchor points and the adjacent por- 
tions of the forecast molded into harmony with these. 
Meteorologists can, and usually do, immediately raise 
embarrassing questions regarding this technique. In 
the first place, what reasons are there for applying 
physical reasoning (vorticity principles), designed es- 
sentially for instantaneous daily maps, to maps cover- 
ing periods as long as thirty days? No satisfactory 
answer to this question can be given at present. How- 
ever, it should be emphasized that the thirty-day mean 
is not composed of randomly distributed data. The 
daily circulations of which it is composed are themselves 
serially correlated and, moreover, during one month 
there are generally repetitive circulation types. Thus 
in a given area a circulation, which on a daily map in- 
troduces a certain field of vorticity attemptimg to 
modify its surroundings, recurs again and again during 
the month to perform a similar modifying function. Is 
it surprising, then, that when the mean chart is con- 
sidered from a vorticity standpoint it, too, seems to 
behave as a physical entity? Another consideration in- 
volves an analogy to the statistical theory of turbulence. 
Here the equations of motion for laminar flow can be 
applied to an average fluid flow on which are superim- 
posed random eddies, provided certain frictional stress 
terms are added. In the work with thirty-day means, 
these additional terms are, of course, not evaluated or 
known, but their omission may not be as serious as at 
first thought. It should be pointed out that they are 
