748 
vein declares: “The ordinary forecasting procedure 
makes no use of any mathematical or quantitative 
physical methods of diagnosis and prognosis.” He had 
pointed out earlier that ‘‘the successful introduction of 
such quantitative physical methods into practical daily 
and longer range weather forecasting constitutes the 
ultimate goal of a large part of present meteorological 
research. But, up to the present time, the mathematical 
computation of local changes of weather elements on 
the basis of physical principles, as distinct from quanti- 
tative kinematical extrapolation, has been effectively 
applied to only a few cases of short-range forecasts at 
the expense of much time and effort. Really scientific 
weather forecasting is far from being an accomplish- 
ment of the present.” 
Although efforts to apply dynamic and thermody- 
namic principles to weather forecasting contiue at an 
accelerating pace and some progress is being made, it 
must be admitted that forecasting is still largely an art 
and only in a relatively small part a science. It is diff- 
cult at times to explain past and present weather satis- 
factorily in terms of physical principles and almost im- 
possible in terms of mathematical formulas. Indeed, 
only two such formulas are in general use in forecasting 
offices at the present time. 
The principal change in forecasting techniques during 
the past three decades has been the elevation of air- 
mass characteristics to an importance equal to that of 
pressure distribution. That is, with the vast imcrease in 
upper-air observational data, the forecaster now has a 
greatly improved three-dimensional synoptic weather 
picture. However, it should be noted that in the second 
phase of the forecast problem—the preparation of the 
surface and upper-air prognostic charts—the emphasis 
remains on the pressure distribution, although a suc- 
cessful forecast of this characteristic of the weather map 
far from guarantees a successful weather forecast. As 
emphasized by Houghton [29] and others, there is no 
unique correspondence, but only a statistical relation- 
ship between the pressure field and the weather. While 
Houghton advocates further studies of this relation- 
ship, he points out the greater importance of the field of 
motion. 
The principal progress in forecasting in the past two 
decades has been evidenced by a significant improve- 
ment in the accuracy of forecasts up to 8-12 hr, in the 
development of certain specialized types of forecasting, 
and in the detail currently attempted with some suc- 
cess. For periods from 12 to 48 hr, forecast accuracy has 
improved no more than 3 per cent. In regions of rapid 
and frequent weather changes, such as in most of the 
United States, the accuracy of weather forecasts di- 
minishes 3 to 4 per cent for every additional 6 hr of the 
forecast period. In view of our better knowledge of at- 
mospheric dynamics and thermodynamics and the 
vastly increased observational data, it must be admitted 
that the accuracy of short-term forecastmg has not 
increased proportionately. 
With the development of new electronic computing 
devices and recent work by Charney and Eliassen [14], 
there is currently renewed interest in the classical at- 
WEATHER FORECASTING 
tempt of Richardson [44] to forecast the weather math- 
ematically. There is, apparently, good reason for op- 
timism that some practicable forecast aids may be 
forthcoming from this source. At the moment, however, 
the possibility of forecasting the weather wholly on an 
objective basis seems very remote. 
The great amount of work completed and under way 
on the upper air encourages the belief that the develop- 
ment of practical applications of dynamics and thermo- 
dynamics, useful in forecasting, will continue. But for 
some time to come it seems likely that, in the words of 
A. H. R. Goldie, “forecasting will continue to be a 
combination of physical reasoning with the practical 
experience of the synoptic charts.” 
PROCESSING THE OBSERVATIONAL DATA 
Observations. In accordance with international agree- 
ment, a basic surface observational network consists of 
stations spaced approximately 140 mi apart and denser 
national and secondary networks. Observations from 
the ternational network are taken over large sections 
of the world at approximately 00, 06, 12, and 18Z and 
distributed to weather centrals and to district and local 
forecast offices by means of teletype, radio, and other 
communication facilities. Observations from the na- 
tional network are distributed to the analysis centrals 
and the district and local forecast centers and from the 
secondary networks to adjacent centers and to such 
other forecastmg offices and centrals as may be re- 
quired. In most areas, reports are transmitted, dis- 
tributed, and plotted as rapidly as possible. The spac- 
ing of pilot-balloon and radiosonde observations 
depends largely upon the budget of the various weather 
services. 
Charts and Graphs. The composition of the several 
types of weather observations as well as the station 
model for plottmg data on the basic surface weather 
chart may be found in publications of the national 
weather services. The basic surface chart usually rep- 
resents a compromise between the desire for both as 
large a scale and as large an area as possible. If possible, 
it should include all areas whence weather in one form 
or another can reach the forecast district within 48-60 
hr. A Lambert conformal conic projection (standard 
parallels at 30° and 60°, and scale 1:7,500,000) is rec- 
ommended for middle latitudes. Larger-scale charts 
may be required for local, airway, and other specialized 
forecasting. 
Pilot-Balloon Charts. These charts contain observa- 
tions of wind velocity and direction for specific levels 
from stations spaced about 150-175 mi apart in ac- 
cessible areas. The levels normally include surface, 
2000, 4000, 6000, 8000, 10,000, 12,000, 15,000, 20,000, 
and 25,000 ft above sea level. The charts are prepared 
at six-hourly intervals. Some stations attempt a stream- 
line and trough-and-ridge-line analysis at the 0900Z° 
and 2100Z observational periods between the regular 
radiosonde observations. 
Thermodynamic Diagrams. Radiosonde observations 
are taken at 0300Z and 1500Z at stations 300-500 mi 
apart although this spacing is insufficient for detailed 
