GENERAL ASPECTS OF EXTENDED-RANGE FORECASTING 
By JEROME NAMIAS 
U. S. Weather Bureau, Washington, D. C. 
Absolute definition of the term ‘‘extended-range fore- 
casting” is not possible and probably will not be for 
many years. As understood today it vaguely refers to 
those predictions which go beyond the limit imposed 
by ordinary standardized forecast techniques. The latter 
forecasts, disseminated by meteorological services the 
world over, generally cover periods of 24, 36, or at most 
48 hr. With this in mind, “extended-range forecasting”’ 
presumably embraces any atmospheric prediction in- 
volving any time period beyond 48 hr and up to the 
geological epochs including the ice ages. The field of 
extended-range forecasting thereby becomes most in- 
viting to people both within and without professional 
meteorology, particularly in view of the tremendous po- 
tentialities it offers in the direction of improving the lot 
of man. 
Considering the extent and complexities of the prob- 
lems posed by the vast spectrum of long-range weather 
forecasts, meteorologists have made small and slow 
progress. This spectrum may be divided arbitrarily 
into the broad bands: medium range (covering periods 
from three or four days to a week), monthly, seasonal, 
annual, decennial, centurial, and so on to millennial. 
In spite of the fact that many workers have spent 
arduous years attempting to find solutions applicable 
to these periods, no system has evolved which even re- 
motely approaches reliability. Unfortunately, the same 
comment can be made in regard to the short-range 
forecasts covering as little as 24 hr in advance. This 
sad state of affairs reflects the lack of basic knowledge 
of the workings of the atmosphere and points up the 
need for periods of stocktaking as exemplified by this 
Compendium. 
But if objectives appear distant, it does not mean that 
progress has not been made. Indeed, medium-range 
forecasts covering periods from a few days to a week 
are already proving economically valuable in many 
countries. Forecasts of general weather conditions for 
periods a month in advance have shown promise. There 
are even some optimistic meteorologists who believe 
that seasonal, annual, or even decennial forecasts of 
some degree of reliability may be possible in the not- 
too-distant future. For the most part, this optimism 
cannot be traced to statistically significant results in 
the prediction of weather phenomena. If we look at the 
problem through the harshly revealing magnifying glass 
of statistics, it appears that, of the spectrum of ex- 
tended-range forecasts cited, only the medium-range 
and perhaps monthly forecasts have demonstrated some 
degree of skill. The term skill is used here in its rigid 
statistical sense, that is, that the predictions are correct 
more often than are predictions made from many years 
of climatic records, or that forecasts must in the long 
run show a higher proportion of successes than the 
climatic probability of occurrence of the event being 
forecast. 
The use of this yardstick makes it possible to restrict 
this article to only those two forms of extended-range 
forecasting, medium range and monthly, which in the 
knowledge of the author have met this criterion over a 
period sufficiently long to be considered significant. It 
must be admitted that this manner of selection may be 
somewhat unfair to those who claim ability to forecast 
for decades, centuries, and longer, for adequate me- 
teorological records are as yet unavailable to verify these 
predictions. 
THE MEDIUM-RANGE FORECAST PROBLEM 
The historical development of modern short-range 
weather forecasting has been essentially inductive in 
nature. A multitude of observations taken at small 
space and time intervals are organized into an analysis 
designed to give physical meaning to the observations. 
This analysis is then compared with earlier analyses, 
generally in order to ascertain changes in circulation 
and weather over larger intervals of time and space, 
and by this method, essentially kinematic, a prediction 
is made possible. To the extent that the weather is de- 
termined by features of the weather map of roughly 
the dimensions of the cyclone and the anticyclone, the 
forecasts are successful. But the day-to-day movement 
of cyclones and anticyclones is often of roughly the same 
order of magnitude as their dimensions (their diame- 
ters). This rapid motion of weather systems, often 
coupled with erratic path, speed, and development, 
points up the difficulty of attempting to make forecasts 
for more than a day or two by using standard forecast 
methods which still rely, for the greatest part, on 
methods of interpolation and extrapolation. 
Perhaps the first real forward step in crystallizing the 
problem of extended-range forecasting came with the 
discovery of the great “centers of action” —a term first 
applied by L. Teisserenc de Bort. These centers are 
brought to light by the statistical averaging of daily 
pressure maps during one month (or season) for many 
years. But if the averaging is carried out only for one 
month of one year (or even for only one week), one also 
obtains clear-cut large-scale features of the general cir- 
culation usually many times the size of ordinary mi- 
gratory cyclones and anticyclones of the weather map. 
Since the time of Teisserene de Bort (1855-1913) it has 
been recognized that it is the position and intensity of 
these centers of action that govern, or at least are as- 
sociated with, the longer-period characteristics of the 
weather. The use of “‘govern” has been and still is ob- 
jectionable to that diminishing group of meteorologists 
who view the centers of action only as a statistical 
average of randomly behaving cyclones and anticy- 
802 
