LARGE-SCALE VERTICAL VELOCITY AND DIVERGENCE 
graphically. According to this figure, clear weather at 
the end of the trajectory was usually preceded by 
slightly negative vertical velocities, and cloudy weather 
or precipitation by vertical velocities averaging about 
+1 em sec. These values reflect average conditions 
only; the figure shows instances where the final weather 
was clear with upward vertical velocities between 1 
and 2 em sec“, and a few cases of final precipitation 
with slightly negative vertical velocities. Apparently, 
the relation is not perfect, largely due to errors in 
constructing trajectories, and errors in the vertical 
velocities. 
Vertical Velocities as a Forecast Tool 
The relation between vertical velocities and changes 
in cloudiness suggests the possibility of applying meas- 
ured vertical velocities to forecasting weather changes 
objectively. If the air has an upward component of 
motion, we might expect bad weather soon. One diffi- 
culty suggests itself immediately. If vertical velocities 
are computed by the adiabatic method, observations 
at the end of a 12-hr period are needed in order to 
compute average vertical velocities for that period. 
These vertical velocities are centered in the middle of 
the period and are therefore already 6 hr old at the 
time the analysis is started. A further lag is caused by 
transmission time, analysis and computation, so that 
the vertical velocities are at least 8 hr late when they 
are ready for application to “forecasting.” 
As Fig. 5 indicates, cloud changes are related to the 
average vertical motion along the air trajectory. If a 
24-hr forecast is desired, a 24-hr air trajectory must be 
forecast and the average vertical velocity along the 
trajectory must be estimated from quantities at the 
beginning of the trajectory. 
Since the forecasting method was to be objective, it 
was necessary to devise a method of forecasting trajec- 
tories objectively. It would require too much space 
here to describe the method finally applied; but it is 
clear that errors in the forecast trajectory will lead to 
inaccurate future positions of the air and hence to 
additional errors in the forecast. 
Experiments with different variables indicated that 
an estimate of the mean vertical velocity along the 
trajectory could be made from the initial vertical veloc- 
ity and the initial meridional velocity. The multiple 
correlation coefficient of mean vertical velocity on ini- 
tial vertical velocity and meridional wind components 
was 0.64 [15]. Clearly, an estimate of the mean vertical 
velocity based on these two variables is subject to 
considerable error. 
In practice it was possible to side-step the computa- 
tion of the average vertical velocity along the trajec- 
tory. Since the average vertical velocity is a function 
of initial vertical and meridional velocity components, 
and the change of weather is a function of the average 
vertical velocity along the trajectory, weather changes 
should be directly related to the initial meridional and 
vertical velocity components. Charts were constructed 
for tne first half of December 1945, which showed the 
final weather and cloudiness as a function of initial 
645 
weather and of the vertical and meridional components 
of velocity. In certain regions of these charts clear sky 
(cloudiness 0.4 or less) was predominant; in other re- 
gions overcast and precipitation predominated. A line 
was drawn which separated the sections of predomi- 
nantly clear sky from sections of sky mainly overcast. 
It was quite difficult to find dividing lines between 
overcast and precipitation on these charts. Such lines 
were drawn, however, under the assumption that pre- 
cipitation should occur with large upward vertical veloc- 
ities along the trajectories. 
Forecasts were made from these charts for the last 
half of December 1945. The total percentage of correct 
forecasts was 69.9 per cent. The chance score was 57.3 
per cent. 
A more severe test is the comparison of the per- 
centage of successful forecasts with those obtained 
from “low skill” forecasts. For example, if “clear” had 
been forecast all the time, the score would have been 
69.2 per cent; if no change of local weather had been 
forecast, the percentage of hits would have been 59.7 
per cent. The same forecasts were repeated by two 
graduate students who had considerable forecasting 
experience. These forecasters had no knowledge of the 
vertical velocities at the time. The two forecasters 
scored 63.4 per cent and 69.8 per cent correct, respec- 
tively. 
Altogether, the objective method of forecasting verti- 
cal velocity did not perform badly, in spite of the many 
sources of error. Later studies [16] showed the method 
considerably less successful. Moreover, Miller showed 
that equally good forecasts could be made by a very 
similar method, if the initial vertical velocities were 
not known at all, but the forecasts were based solely 
on the initial weather and meridional velocity compo- 
nent. Therefore the laborious computation of vertical 
velocities for forecasting purposes is possibly unneces- 
sary. Apparently, the meridional velocity component 
is a sufficiently good indicator of the direction of the 
vertical air motion. The relation between meridional 
motion aloft and weather is known to many meteorol- 
ogists, and has been incorporated, indirectly, into 
other objective forecasting methods. 
Some attempts were made to use vertical velocity 
patterns qualitatively. On daily vertical velocity charts 
the “unusual” features that did not conform to the 
simultaneous weather were noted. For example, an 
area of upward motion over an east coast wedge was 
judged significant. Such unusual features were fol- 
lowed sometimes, but not always, by unusual develop- 
ments; for example, unusual centers of upward motion 
were associated with cyclogenesis. Several forecast rules 
based on unusual vertical velocity patterns were sug- 
gested, but none of them proved reliable. Whether 
this was due to errors of the vertical velocities cannot 
yet be decided. 
Suggestions for Future Research 
Since most of the studies of vertical velocity were 
completed the network of radio and radar wind ob- 
servations has been greatly extended. As a result, de- 
