NO. 5 LONG-RANGE TEMPERATURE FORECAST — ABBOT 1 5 



were representative, not of absolute Fahrenheit degrees of tempera- 

 ture, but of percentages of the normals A and B of table 3. One 

 might regard them as flowing functions, i+X and i+Y, times the 

 normal temperatures A and B, where X and Y are flowing variables, 

 each ranging from —80 to +150 ten-thousandths. 



Taking the mean 1+ X and the mean i+Y, used as multipliers of 

 the normals A and B, these normals could be transferred into new 

 normals suited to the atmospheric conditions prevailing from 1950 

 through 1959. By assumption these new normals were usable in the 

 forecasts from i960 through 1967. But, as stated above, it was found 

 that neither in scale nor in level did the new normals from X agree 

 with the new normals from Y. To make the 2 values fairly compa- 

 rable with the Obs values, a scale correction to 2 was first determined. 

 This was done as stated above by plotting 2 and Obs from 1950 

 through 1959, and obtaining the mean ratio of amplitudes of some 

 half-dozen principal obviously common features of the two curves. 

 After this adjustment of scale a level correction to 2 was computed. 

 This was the ratio of the sums of scale-corrected 2 to the unchanged 

 Obs for the years 1950 through 1959. Applying it, the two variables 

 became justly comparable. As thus reconciled, the new final normal 

 absolute temperatures were tabulated, and then they were reduced to 

 ordinary Fahrenheit by subtracting 459.7 . In this form the new 

 normals, which are assumed to be suited to the atmospheric conditions 

 1950 through 1967, are given in table 4. 



Subtracting the new normals from the two reconciled columns of 2 

 and Obs we obtained the monthly march of Obs from 1950 through 

 1959, and that of 2 from 1950 through 1967. The march 2 after 

 i960 represents the forecast of chief interest. But the comparative 

 marches of 2 and Obs 1950 through 1959 gives the evidence on which 

 a judgment of the probable value of the forecast i960 through 1967 

 principally depends. 



For convenient general views of the predicted march of tempera- 

 ture, I give in table 6 4-month mean values of the forecasted de- 

 partures from normal 1950 through 1967, together with 4-month 

 means of the observed departures from normal 1950 through 1959. 

 From a comparison of these 10 years a judgment may be formed of 

 the worth of the forecast after 1959. 



As further evidence of the value of the forecast I give in table 7 

 the average discrepancy between forecast and event 1950 through 1959 

 and the average algebraic mean difference to show that the corrections 

 for scale and level combined closely reconciled forecast and event. 



