EXTENDED-RANGE FORECASTING 
based upon the field of thermal vorticity indicated by 
the thickness lines. Further synoptic studies are being 
carried on by the British group to lend support to these 
criteria, and it seems possible that the vexing problems 
of trough development or disappearance may find par- 
tial solution in this work. 
In the practical routine of extended-range forecast- 
ing in the U.S. Weather Bureau, the thermal field of 
the upper air is given careful consideration—not only 
in the thickness patterns themselves but also in charts 
showing departures from normal of both the thickness 
patterns and the temperatures at 700 mb. These ther- 
mal fields are also used as adjuncts m determining deep- 
ening or filling of individual long-wave systems. 
These purely physical methods of prognosis are natur- 
ally far from perfected. For this reason it is necessary to 
use them in conjunction with the kinematic procedures 
described earlier. Just as with short-range forecasting, 
the evaluation of time derivatives becomes important. 
In a sense, to the extent that the tendencies are accu- 
rate, the kinematic methods represent an integration of 
physical processes. The arbitrary division into physical 
and kinematical tools is made only because it is pos- 
sible to make partially successful extended-range fore- 
casts with these kmematic methods even though the 
user has no idea of the physical processes involved. 
Experience indicates that the highest skills are ob- 
tained by judicious combination of physical and kine- 
matical tools. A specific example of such coordination 
may clarify this point. Let us say that there are two 
pronounced troughs in the upper-level westerlies whose 
wave length is about equal to the stationary wave 
length, but that the tendency fields suggest strongly 
that the troughs are spreading apart—perhaps the 
western one is retrograding while the eastern one is 
progressing eastward. Kinematic displacement compu- 
tations would therefore suggest the incipience of a new 
trough, and the forecaster would be alerted to examine 
other evidence for further clues, such as computed 
vorticity paths, thermal fields, and cyclogenetic char- 
acteristics of the latest weather map. 
The methods described above are meant to apply 
primarily to the construction of five-day (or longer- 
period) prognostic mean upper-level charts. From such 
upper-level prognoses the associated mean sea-level 
charts may be prepared by methods of differential 
analysis. 
In the past ten years it has become increasingly clear 
to the author that the most successful extended-range 
forecasting procedures must involve not only the current 
state (daily or mean) of the atmosphere, but also its 
manner and rate of evolution. Methods of extended- 
range forecasting which do not incorporate in some 
manner the evolution must be very limited in scope and 
success. 
Here we come to one of the principal dilemmas of the 
modern weather forecaster. How can he coordinate the 
tremendous mass of data in space and time necessary 
to incorporate evolution in a three-dimensional, glob- 
ally interactive atmosphere? Yet he must do this to 
arrive at a regional prognosis which is internally con- 
809 
sistent with other hemispheric features. The principal 
hope for the solution of this apparently superhuman 
task seems to lie in the development of new electronic 
high-speed computing machines from which may 
emerge at least a first approximation to the prediction. 
THE RELATIONSHIP BETWEEN CIRCULATION 
AND WEATHER 
The methods of prediction described above aim at 
forecasting features of the atmospheric circulation gen- 
erally represented by pressure patterns at one or more 
levels. The tacit assumption is made that weather 
forecasts are by-products of the circulation prognosis. 
Precisely this assumption underlies the preparation and 
transmission of short-range prognoses as practiced 
around the world. While this assumption is to a con- 
siderable extent justified, controlled experiments indi- 
cate that even if forecasters were provided with perfect 
prognostic charts (analyses rather than prognoses) they 
would be unable to make perfect forecasts. In fact, in 
many situations forecasts of weather from such charts, 
whether a short- or long-range (mean) prognosis, leave 
much to be desired. For this reason it would appear that 
a disproportionate amount of research may be currently 
placed on improving circulation prognoses while very 
little is being done to interpret circulations in terms of 
associated weather. 
This problem, equally important in longer-range 
work, appears to have received more attention than it 
has in short-range forecasting. For example, Baur [1] 
has made extensive statistical studies of the weather 
regimes associated with various large-scale circulation 
patterns over Hurope, and similar studies for America 
have been carried on at the California Institute of 
Technology [5]. One of the most systematic studies 
relating pressure patterns to weather phenomena has 
been in progress for several years at the U. S. Weather 
Bureau. Here an attempt has been made to associate 
temperature anomalies and total precipitation observed 
over five-day and monthly periods to mean pressure 
patterns at sea level and, more especially, at 700 mb. 
The essence of the most promising methods for finding 
these relationships appears to lie primarily in the large- 
scale contour patterns of the lower troposphere. 
Studies for different seasons and for many points over 
the United States indicate that regardless of how mean 
maps are made up (that is, the sequence and distribu- 
tion of day-to-day patterns within the period), they 
determine to a large extent the average temperature 
and precipitation anomalies® for the period considered. 
This statement often comes as a surprise to many 
meteorologists in spite of the fact that it seems to have 
been recognized almost a century ago by the pioneers 
who discovered and studied the great centers of action. 
In part, the physical reality of the mean as it relates to 
average weather is due to the fact that events in time 
are serially correlated. Not only this persistence but 
5. The term ‘“‘anomaly,’’ as used in this article, refers to 
departure from a long-term average over time and has nothing 
to do with the departures from means computed for latitude 
circles. 
