NO. 4 



sun's variation and temperatures — ABBOT 



29 



a trial forecast of Washington temperature departures based on this 

 periodicity. 



I took the expected amplitude from the first two recurrences of the 

 period computed for October 1931, as plotted in figure 3. The phase 

 is unchanged throughout from that of early October. These data on 

 amplitudes really involve all the days from September 27 to November 

 I, but in different proportions. Six-tenths of the weight of the de- 

 termination rests on the data of September 27 to October 16. Ac- 



Fig. 12. — Thirty-six-day prediction of temperature departures at Washington, 

 based on previous phases and ampHtudes of the 6.6456-day period. Dotted curve 

 forecast, full curve the event, reduced by the factor 4/10. Arbitrary zero of 

 ordinates. 



cordingly my prediction from October 16 to November 18 is to a 

 slight degree governed as to amplitudes, not at all as to phases, by 

 what actually occurred as late as November i, but beyond that date 

 it is pure prediction in all respects. 



I give in table 9 all the numerical figures used in preparing the 

 forecast from the data plotted in figure 3, as above specified. It will 

 be seen how little work is involved. Figure 12 gives the forecast and 

 verification. The full curve is the graph of the actual departures from 

 normal temperature at Washington, as published by the United States 

 Weather Bureau, but reduced in scale by multiplying by the factor 

 4/10. The dotted curve shows the forecast of the departures caused 

 by the 6.6456-day periodicity, expressed in degrees Fahrenheit, The 

 prediction is based to give temperatures all above the zero, and the 



