THE FORECAST PROBLEM 
sudden demand for accurate forecasts of this type, 
this is the one field of forecasting in which great 
improvement in forecasting skill has been demonstrated 
in the past twenty years. Although no verification 
figures for this type of forecasting are available, it is 
probable that the best forecasting organizations in this 
field perform in the 90-95 per cent range on a 50 per 
cent probability basis of verification. This performance 
does not represent any basic advance in weather fore- 
casting, but merely the happy results of a sudden con- 
centration of effort on the application of extrapolation 
techniques to very short-range developments. 
2. Forecasts of the ordinary daily range, for periods of 
from twelve to forty-eight hours in advance. Forecasts 
of this type usually express in some detail, for specific 
geographical areas, the expected day-to-day sequence 
of all the aspects of weather which materially affect 
human activity and well-being. The forecasts are dis- 
seminated by press, radio, wire, and special warning 
systems to the public and to many special interests, 
both civil and military, including aviation, agriculture, 
shipping, manufacturing, merchandising, and utilities. 
It is this type of foreecastmg on which the greatest 
effort has been concentrated since synoptic forecast 
methods were first introduced nearly a century ago and 
in which, paradoxically, basic progress has been almost 
nonexistent in recent years. Forecasting activity has 
been vastly expanded, but the quality of the forecasts 
has not been basically improved. The accuracy and skill 
of this type of forecast as practiced by the best fore- 
casters must diminish with increasing time range during 
the 12- to 48-hr period. The skill verification may be 
estimated to decrease through the range 90-70 per 
cent during the period. 
3. Extended daily forecasts, for periods imecluding 
from the second to the sixth or seventh day in advance. 
Forecasts of this type at the present time tend to 
assume a dual character, a combination of the ordinary 
daily forecasts on the one hand, and of the long-range 
forecasts on the other. In the first role they strive to 
extend to five or seven days in advance the same 
detailed forecast of the day-to-day sequence of weather 
that is offered by the ordinary daily forecasts up to two 
days in advance, while in the second role they specify 
the mean or average character of the 5- or 7-day period 
ahead in terms of the expected anomalies of tempera- 
ture, precipitation, and possibly other elements, such 
as sunshine. In the first role these forecasts are expected 
to serve the day-to-day planning of the entire life of the 
community in the same manner as, only more in advance 
than, the daily forecasts, while in the second role they 
are useful primarily to agriculture, irrigation, flood 
control and power interests, and to the fuel industry. 
Forecasts of the extended daily variety have been 
issued for many years, but the principal recent advances 
in extended weather forecasting have come in the de- 
velopment of statistical-synoptic techniques for the 
forecasting of large-scale weather patterns and mean 
weather conditions. The forecasting of the day-to-day 
weather sequence shows negligible skill beyond the 
fourth day in advance, probably decreasing slightly in 
733 
the 55-50 per cent range. It can be stated categorically 
that no claim to significant skill in forecasting the 
day-to-day sequence of weather more than five days in 
advance has been acceptably demonstrated, nor is it 
likely to be m the near future. The mean anomaly 
features of the more reliable extended weather fore- 
casts today verify in the 70-75 per cent range for 
temperature, but little better than 60 per cent for 
precipitation. Precipitation, relative to temperature, 
becomes increasingly difficult to forecast as the time 
range is extended. 
4. Long-range forecasts, for all periods extending 
beyond one week in advance. Forecasts of this type can 
deal justifiably only with mean conditions or period 
anomalies, usually of pressure, temperature, or precipi- 
tation. There is no legitimate basis whatsoever for 
extending forecasts of the day-to-day weather sequence 
more than one week in advance, frequent unverified 
claims to the contrary notwithstanding. Forecasts of 
the longer-period anomalies of temperature and pre- 
cipitation, whether for weeks, months, seasons, or whole 
years, can be of very considerable value, insofar as they 
have real merit, for a wide range of mterests, notably 
for planning effective agricultural production or mer- 
chandising campaigns, for the effective utilization of 
irrigation, hydroelectric and flood control facilities, and 
for long-range military planning. Unfortunately, how- 
ever, no forecasts of this type have demonstrated any 
clearly significant degree of skill in verification. Prob- 
ably under most favorable conditions the verification 
lies in the 55-60 per cent range. 
Forecasting Techniques or Aids—Their Merits and 
Limitations. The following definition and evaluation of 
weather-forecasting techniques as they are practiced 
today can be only approximate at best, because of the 
multiplicity of techniques which are in use, the infinite 
variety of combinations in which they are applied by 
different forecasters for different and even for the same 
type of forecast, and the lack of any real effort towards 
the objective evaluation of these numerous forecasting 
aids. In the following discussion an effort is made to 
classify the individual forecasting techniques or aids, 
insofar as they are distinctive, as to their essential 
character; to indicate something as to their usual appli- 
cation in the four principal types of forecasts (time 
range); and to comment qualitatively on their merits 
and limitations. 
Since the following classification of forecasting tech- 
niques is based on the essential character of the tech- 
nique, rather than on the forecast type (time range) 
to which it is applied, and since the practice of any 
type of forecasting frequently involves the combination 
of a variety of techniques, the grouping by this classifi- 
cation is suited primarily to an evaluation of the tech- 
niques individually rather than in combination. It is 
quite impossible in a general discussion of this kind 
even to enumerate the many combinations in which 
these forecasting techniques are applied, much less to 
evaluate these combinations. But in general the skill 
verification that is obtained by combining forecasting 
techniques is of the same general order as that of the 
