86 TRANSACTIONS OF THE 
equation (1) and the 4 A. M. May 1st Denver normal 
temperature is obtained. 
The errors include the fact that one more year’s 
record might change the normals used and the graph of 
the equation might not coincide with the plot of this new 
data. These errors are very slight. In applying the 
method in the eastern part of the United States where 
cyclones cause large departures from the normals, 
longer records are required, the curve of annual tempera- 
ture change has a slightly different shape or amplitude 
and the ratio of maximum to mean for the day is 
smaller but approximately constant. 
One fourth of the earth’s surface is as dry as the 
Great Basin of Western U. S., where the precipitation is 12 
inches, the humidity 50%, and 80% of the days are with- 
out rain. These wide departures from normals are com- 
paratively rare and the above method is useful in long 
time temperature prediction. 
Using the above method it was found that in the 
arid west, the chances are one in six that the actual 
temperature will differ from the computed value by less 
than 2°F., two in five that it will be 5°F., one in four that 
it will be around 10°F., and 1 in 7 that it will be as much 
as .15°F. Cyclones and anti-cyclones are the main 
causes of these departures. By consultation with the 
Weather Bureau concerning the location of these “‘highs”’ 
and “lows,’”’ these differences may be very considerably 
reduced. 
