NO. 6 SOLAR RADIATION AND WEATHER CLAYTON 55 



Argentina, and Dr. Abbot was anxious to know whether such fore- 

 casts could be successfully made for other parts of the world. After 

 considering the matter, it was decided to try forecasts of temperature 

 for some particular point in the United States, and New York City 

 was selected because of its great importance as a commercial center, 

 and because its weather changes were known to be highly complex, 

 so that if weather forecasts could be successfully made for that point 

 they could probably be made for any part of the northern hemisphere. 



It was agreed that the annual change of temperature should be 

 eliminated by determining the normal maximum temperature for 

 each day in the year, and that forecasts of departures from the normal 

 maximum temperature of each day should be attempted. It was 

 further agreed that the forecasts should be verified by averaging the 

 temperature in three different classes. The days for which high 

 temperature was predicted were to be classed together, and the aver- 

 age departure of the observed temperature from the normal was to 

 be determined for that day and for two days preceding and following. 

 The days for which normal temperature was predicted were to be 

 treated in the same way, and also the days for which low temperature 

 was predicted. It was understood that in order to be successful the 

 temperature should average above normal when high temperature 

 was predicted and below normal when low temperature was pre- 

 dicted. It was further decided that the forecasts should be stated 

 numerically, and that forecasts of 5° or more above normal should 

 be forecasts of high temperature, forecasts of -1-4° to —4° should 

 be considered normal, and forecasts of 5° or more below normal 

 should be considered forecasts of low temperature. 



Forecasts were to be made for three, four, and five days ahead, 

 and also for 27 days ahead. After a preliminary test of shorter inter- 

 vals ahead for two months, forecasts were begun, in accordance with 

 the plan outlined, about December i, 1923. Also tests were made as 

 to the possibility of forecasting the mean temperature of the coming 

 week and month, making the forecasts three days ahead of the begin- 

 ning of the week or month. 



All these forecasts were verified by Dr. Abbot and Mr. Farmer at 

 the Smithsonian Institution, by means of data collected by them from 

 official sources, and were later checked by Mr. Eliot C. French and 

 myself in Canton, Massachusetts. 



