OBJECTIVE WEATHER FORECASTING 
tionships, and by the expression of the forecast itself in 
quantitative terms or at least in unequivocal terms. 
Fear has sometimes been expressed by forecasters that 
a result of the development of objective forecasting 
methods will be to supplant experienced forecasters by 
mechanical methods. It should be obvious, however, 
that the greater the reduction in the number of subjec- 
tive and uncertain decisions required in the process of 
preparing the forecast, the more time will be available 
to the forecaster either for studying the effect of new 
and untried variables and the value of new principles, 
or for interpreting the forecast for the exceedingly di- 
verse uses to which it is applied by the public. 
From the standpoint of discovering and understand- 
ing relationships which hold in the atmosphere, fore- 
casting investigations have been relatively ineffective 
because of their stress on lag relationships, and it seems 
clear that only a complete physical explanation of the 
atmosphere together with complete observational data 
will make it possible to produce perfect weather fore- 
casts. Practically, however, uncertainties exist which 
make the maximum attainable accuracy something less 
than perfection [22]. The forecasting problem is thus, 
in essence, one of estimating what is likely to occur with 
any given state of the atmosphere and its environment. 
More precisely, the problem is to state the probability 
that any specified weather event will occur within any 
specified time interval. 
The statistical or probability aspect of weather fore- 
casting was recognized as early as 1902 by Dines [7], 
who pointed out the impossibility of knowing exactly 
what weather is going to occur and suggested that the 
laws of chance should be applied. Hallenbeck [9] in 
1920 found an encouraging response from the public 
when he attempted the use of numerical probability 
statements as part of his agricultural forecasts. It seems 
to have been only recently, however, that this objective 
has been recognized by a large group of meteorologists 
and that attempts have been made to apply the meth- 
ods of mathematical statistics or to develop new meth- 
ods suitable for the estimation of forecast probabilities 
[5, 25]. Since the public generally has demanded cate- 
gorical forecasts, attempts to express the ‘‘chances” 
of a weather event occurring have usually been frowned 
upon by forecasters. Nearly every decision the fore- 
caster is called upon to make, however, involves weigh- 
ing the chance as indicated by one set of factors against 
the chance as indicated by one or more other sets. 
Objective forecasting studies have not often provided 
fmal, conclusive evidence of the chance of occurrence 
of the weather event in question, but such studies have 
reduced the uncertainty to quantitative and under- 
standable terms, and it is one purpose of such studies to 
determine the actual frequency with which any se- 
quence of events which it may be desired to specify can 
be expected to occur. 
Limitations of Objective Forecasting. Some of the 
objective forecasting methods which have been developed 
[1, 8,17] have demonstrated a rather clear superiority 
to forecasts made by conventional methods under rou- 
tine operational conditions. The question then arises, 
797 
why have forecasters generally not seized wholeheart- 
edly the opportunities offered by these methods of 
investigation for improving their own forecasts? There 
are a number of reasons, based partly upon misunder- 
standing of the methods and accomplishments of ob- 
jective forecasting, but based more largely on the limi- 
tations of objective forecasting under the present 
organization of forecasting services. 
One criticism often made is that objective methods do 
not take account of all of the pertinent variables nor of 
all the characteristics of the weather situation which 
help in judging the weather to come. This deficiency is 
certainly true of most, if not all, of the systems which 
have been published, but it is less a criticism of objec- 
tive forecasting as such than it is an admission that 
items are used in forecasting which the forecaster can- 
not express objectively or quantitatively. The extent 
to which present forecasting knowledge is real knowl- 
edge as distinct from lore may be judged in part by the 
extent to which forecasters have been able to incorpor- 
ate it into objective systems or have been able to write 
it down in a form which can readily be taught to 
students of forecasting. It is one of the goals of objec- 
tive forecasting studies to collect and systematize this 
knowledge so it will be available for study and subse- 
quent use by any forecaster, but major problems are 
encountered at this point. Even a relatively inexpe- 
rienced forecaster has an ability to find in a given 
weather situation features which appear to be signifi- 
cant for the forecast, but which are extremely difficult 
to express in any objective way. Methods must be 
devised to take into account these more tenuous con- 
cepts and features of the atmosphere and to separate 
the real relationships from the fictitious. 
The time required to work out a forecast by objective 
techniques is a limitation on the use of such techniques 
in many offices. The organization of forecasting services 
is generally centralized to reduce the work of plotting 
and analyzing the numerous maps and charts required 
for forecasting and to make it possible to utilize the 
best forecasters to cover the widest possible area. This 
generally results in the forecaster’s being faced with 
tight schedules and gives him a totally inadequate 
amount of time to spend in actual preparation of the 
forecast. In spite of the success of simplified objective 
methods in equalling the accuracy of conventional fore- 
casting procedures, it is suspected that forecasts which 
are improved to any great extent over present levels of 
accuracy will be produced, for the most part, only by 
methods which require a longer time to apply and 
which, therefore, cannot conveniently be used by fore- 
casters at present. 
One characteristic requirement of present-day fore- 
casts which are issued to the public is that a definite 
bias must frequently be introduced into the forecast, 
or in other words, the forecaster must not state exactly 
what he thinks is most likely to happen. For example, 
the bare mention of rainfall in a public forecast will 
often be construed by the public as a forecast of rain 
no matter how the forecaster qualifies the statement. 
If the occurrence or nonoccurrence of rain is an un- 
