NO. 9 FORECAST OF U. S. PRECIPITATION — ABBOT 9 



to was indeed discovered by the study of over 30 years of daily "solar- 

 constant" observations of the Smithsonian Astrophysical Observatory. 

 But now that the harmonic family has been found in weather, no 

 observations of any kind are required. It is only necessary to employ 

 a long record of monthly mean values of precipitation, or temperature, 

 to make long-range predictions. These are approximately verified if 

 no unusual alterations of atmospheric conditions make the averages 

 from long records inapplicable. 



Sports. — As my forecasts depend on the assumption that the av- 

 erage conditions of the periods over a thousand months will be pro- 

 jected into the future, it is important not to include wild "sport" 

 values of precipitation in the thousand-month basis. Hence I have 

 diminished sporadic very high values to about two times normal, and 

 have raised sporadic drought values of less than 40 percent of normal 

 to exceed that limiting low value. These limits refer to 3-month 

 smoothed records. For most of my 32 stations these changes are very 

 rare. But in two or three of the desert stations possibly one value in 

 ten was changed to avoid spoiling the representative character of the 

 basis. The considerable measure of success of my forecasts is the 

 main defense of the method used to produce them. If the degree of 

 success is found to be valuable, no doubt those who in future will 

 use the method will greatly improve it by modifications dictated by 

 reason and experience. 



Backcasts. — Since my forecasts are made by adding the average 

 effects of 27 harmonic periods over an interval of about 1,000 months, 

 the 12 months of record for any one year can produce only about 1 

 percent of influence on the forecast for that year, even if those 12 

 months are among the thousand months employed as a basis. There- 

 fore all forecasts or backcasts are equally sound, whether they relate 

 to time before, within, or after the thousand months of record. 3 



The preceding paragraph is important. The forecasts for 32 cities 

 all extend from 1950 to 1967. The degree of similarity between the 

 forecasts and what happened up to 1958 is the index of their prob- 

 able agreement from 1959 to 1967. 



The 273-month period. — Daily solar-constant observations pro- 

 ceeded from 1920 to 1952 at Montezuma, Chile. This interval is not 

 long enough to determine the master period accurately. But the 

 10-1/9-month period in weather is a strong one and has long been 

 followed in Washington precipitation. I determined its amplitude for 

 several periods differing slightly from 10-1/9 months. For this pur- 



8 See discussion of backcasts at a later page. 



