58 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. yj 



the forecast was made, and declined to near normal two days later. 

 For normal forecasts the mean temperature was slightly below normal 

 at the beginning of this interval, and slightly above at the end. For 

 forecasts of temperature below normal the temperature was slightly 

 above normal two days after the observations, and fell to a minimum 

 on the fourth to fifth day. There was a rise on the sixth day, but the 

 temperature did not return to normal until later. 



This curve demonstrates conclusively that on the average we 

 succeeded in making forecasts of the daily maximum temperature at 

 New York for four days ahead and that we were not aided in doing 

 so by prolonged departures from the mean on one side or the other 

 of the normal. Whether accurate forecasts could be made for such 

 an interval in advance by any other method I am not prepared to say, 

 but heretofore no one has made definite forecasts for such an interval 

 in advance and submitted them to a third party for verification by a 

 method which does not permit personal bias to influence the results. 



The averages in table 19 show that for three days in advance suc- 

 cessful forecasts were also made of the mean temperature of the 

 following week and month. 



The detailed forecasts for 27 days in advance based on the return 

 of similar conditions by a solar rotation were not successful, but 

 this by no means indicates that such forecasts will not be possible 

 when solar conditions and their relations to terrestrial phenomena 

 are better understood. 



In carrying out forecasts based on solar data at the present time, 

 serious difficulties beset the forecaster : 



( 1 ) The measurements of solar radiation are frequently prevented 

 by cloudiness, so that sometimes for two or three successive days 

 observations may be missing. At other times the accuracy of a 

 measurement is uncertain or doubtful, and at all times there is a 

 certain amount of error. 



(2) The effect on the earth's atmosphere of variations in solar 

 heat radiation differs with the season, and with different intensities 

 of solar radiation, so that it is necessary to determine and to keep in 

 mind the effect of both these variables. 



(3) Each new solar influence is superimposed on pre-existing 

 conditions. In some cases, if not in all, these conditions arise from 

 progressive movement of weather changes from one center of action 

 to another. In Argentina, for example, it was found that at certain 

 seasons an increase of solar radiation produces a fall of pressure and 

 a rise of temperature in northwestern Argentina and southwestern 



