NO. 6 PREDICTION OF WASHINGTON TEMPERATURE I948 ABBOT 5 



in Washington temperature departures for the 9 years 1940 to 1948. 

 The upper part of the figure shows these results and their mean 

 graphically. It is satisfactory to see that the mean curve is nearly 

 smooth, with the correct phases for maximum and minima. The 

 range of the mean curve is i?63. This is certainly too low a value to 

 be a correct estimate of the amplitude of the periodicity. For, as 

 shown in my paper cited above, the phases of the periodicity are 

 subject to a sort of backlash effect, and they are apt to occur i, 2, or 

 rarely 3 days too early or too late. Hence where, as in my tabula- 

 tion just referred to, no allowance is made for these shifts of phase, 

 the real mean amplitude of the periodicity must be much greater than 

 computed. 



I have computed also the average range per year after the method 

 set forth in the test paper above. I give these results for the 9 years 

 in table 2. They are very closely all the same, and give a mean of 



Table 2. — Average amplitudes of the 6.6456-day temperature variations 

 in Washington, 1940-194S 



Years 1940 1941 1942 1943 1944 1945 1946 1947 1948 Mean 



Amplitudes . . . 5°8i 6?95 6?5S 6?56 6?58 6?6o 8?6o 6?oo 6?96 6?73 



6?73. For the reason mentioned above, I consider this value too 

 high an estimate of the ampHtude of the 6.6456-day periodicity in 

 Washington temperatures. But I do not think it as much too high 

 as the other one, i?63, is too low. However their mean, 4? 18 F., 

 may be nearly of the right order of magnitude. 



While yearly forecasts of dates of minimum and maximum tem- 

 peratures by this method may be out of phase by from i to 3 days, 

 they may be made with equal success for many years in advance, as 

 my earlier paper which covers 35 years shows. H the forecaster 

 should content himself, however, with forecasts only a month or two 

 in advance, and should take into account the phase prevailing when 

 the forecast was made, he might hope to be within i day of correct 

 phases for minima and maxima. See, for instance, in the lower part 

 of the figure, how closely the phases held true from February 5 to 

 March 31, 1946.^ The method, as I have said, is probably applicable 

 everywhere. 



* Note also by the curves of the upper part of the figure that in the years 

 1940, 1941, 194s, and 1948 the phases were prevailingly appearing 2 days or more 

 late. 



