A PROCEDURE OF SHORT-RANGE WEATHER FORECASTING 
casting centers are needed to assist the local forecaster 
in issuing certain special forecasts. For example, we re- 
fer to such formal U.S. Weather Bureau organizations 
of the local forecast offices under special forecasting 
centers as the Corn, Wheat, Cotton, and Livestock 
Weather Service, the Hurricane-Warning Service, the 
Airways Forecasting Service, the Flood-Warning Ser- 
vice, the Fruit-Frost Service, and the Fire Weather 
Service. In addition we may mention the application 
of special forecasts to such enterprises as the shipping 
of perishable goods, the distribution of seasonal mer- 
chandise (winter clothing, ice cream, power, light, and 
fuel), the cinematic industry, the conservation of water, 
and irrigation. 
The geographical extent of the forecast localities, of 
roughly a 40-km radius, generally should be somewhat 
irregular and should be subject to occasional changes 
because of the fluctuating administrative and profes- 
sional functions of the forecast localities. Wherever 
there is no local forecast office to service new require- 
ments, a nearby local forecast office is appointed to 
fulfill the service. Such an appointment is based on 
several factors in addition to the geographic one, for 
instance on the size and ability of the administrative 
and professional staff to handle the new task. 
Summing up, we may state that map plotting, analy- 
sis, and prognosis—for reasons of economy—call for 
a certain centralization of analytic and prognostic work, 
especially of upper-air and routine work. On the other 
hand, the necessity of detailed forecasts (of different 
types) valid for restricted areas and localities leads to 
a demand for a decentralization of prognostic and fore- 
casting work. The most practical compromise between 
these two tendencies will depend on the topography 
of the region, on the geographical latitude, and on the 
types of forecasts desired [41]. 
The Accuracy of Forecasts. Fronts, precipitation 
areas, etc., linearly extrapolated 24 hr to 36 hr ahead, 
may well have space errors from two to four times as 
large as do their analyzed positions, which can be fixed 
with an accuracy of from about one-half to one-fourth 
of the distance between adjacent synoptic stations, that 
is, with an analyzed accuracy of about 50 km. Depend- 
ing on their velocity of propagation as well as on the 
errors in their analyzed positions, the maccuracies of 
the arrival time of a fast-running disturbance moving, 
say, with a constant speed of 100 km hr may there- 
fore be of the order of magnitude of 2 hr. Further in- 
accuracies—the so-called ‘“‘weather surprises’’—are in- 
volved in the extrapolation of fast-running disturbances 
coming from a region where, owing to a sparse network, 
their positions have been analyzed inaccurately. 
The value of a forecast depends upon its minutiae 
as well as upon its accuracy; a forecast proved correct 
according to purely formal methods is not necessarily 
the most useful forecast (for more details see [10, 18, 38, 
52, 78]). Although it is not easy to verify a short-range 
weather forecast and to judge fairly the efficiency of a 
forecaster, statistics of success may be of some interest 
in weather services where the methods of analysis and 
prognosis are to a certain extent standardized. (For a 
769 
survey of literature on the verification of short-range 
weather forecasts, see [54] and the discussion elsewhere 
in this Compendium.!) 
THE PROGNOSIS OF THE SURFACE MAP 
The Geometrical Extrapolation of the Surface Map. 
The first stage in the order of operations for obtaining 
the t -+ 24> prognostic surface map is the geometrical 
extrapolation of the analyzed surface maps for finding 
the to + 3" prognostic surface map. In a sequence of 
surface maps carefully analyzed 3 hr apart,’ the fore- 
caster can determine directly the motion and develop- 
ment of their generally nonphysical features, such as 
isobars, pressure centers, ridges, troughs, and fronts 
with their cloud and precipitation areas: By super- 
imposing on the tracing table the maps for f) — 3" and 
to, average values are found for the 3-hr period ending 
at to. Though less reliable than such direct determination 
of velocity, the kinematical formulas independently in- 
troduced by Dedebant during 1927-82 [19, 20], Angervo 
during 1928-30 [2, 3, 4], Wagemann in 1932 [81], and 
Petterssen in 1933 [58] are nevertheless of some use for 
determining, from the reported 3-hr pressure tendencies, 
the motion and development of features and systems 
which have just entered the analytic region. Valid only 
wherever there are no discontinuities in the space and 
time derivatives of pressure, these speed formulas should 
preferably be applied to sea-level systems without 
marked frontal structure. Thus, for determining the 
tangential speed of a frontal cyclone, Petterssen [60, 
pp. 398, 411] has recommended that the nonfrontal 
trough formula be based upon the pressure and tendency 
profiles poleward of its center where there are no fronts. 
The deepening and filling of sea-level pressure centers 
and cols, as determined either from the application of 
numerical formulas to the last surface map or from the 
last two available surface maps, must be corrected for 
an appreciable diurnal variation before being used for 
extrapolation. 
By means of the average values thus found for the 
3-hr period ending at the time fo of the last map, the 
forecaster extrapolates linearly and in a merely formal 
way the position, 3 hr ahead, of the map lineaments 
and the deepening and filling of sea-level pressure 
systems. A correction, which is next performed as part 
of stage one, consists in using the sequence of 3-hr maps 
from tf) — 24" to t in order to take into consideration 
higher-order terms in the extrapolation. Moreover, if 
the sea-level allobaric centers (preferably for 12 hr) 
move with approximately the same velocity as the 
pressure center, their past movement may be used to 
determine more surely the past movement of the pres- 
sure system. If this is not the case, the future velocity 
changes of the sea-level pressure systems are sometimes 
indicated by the recent movement of the corresponding 
sea-level allobaric centers. 
As stage two, a rough sketch of the sea-level prognostic 
surface for ty + 24" is made by applying these geometrical 
1. Consult “Verification of Weather Forecasts” by R. A. 
Allen and G. W. Brier, pp. 841-848. 
2. Prior to 1930-35 the maps were prepared at 6-hr intervals. 
