REPORT OF THE ACTING SECRETARY. 77 



naturally smaller, owing to the high altitude and clear sky, but the results for 

 the two stations in general agree within 2 or 3 per cent, and thus within 

 the probable error of the Washington observations. 



Considering the great difference in altitude of the stations, this tends 

 strongly to increase confidence in the soundness of the method of " solar con- 

 stant " determination employed. 



B. Absolute value of the " solar constant." 



Preliminary comparisons between the secondary pyrheliometers heretofore 

 used and the standard pyrheliometers now perfected indicate that the prelimi- 

 nary scale of values of the "solar constant" heretofore published in these 

 reports is probably very nearly correct. During the observations in 1905 on 

 Mount Wilson the solar radiation was generally below the mean value of about 

 2.12 calories as observed at Washington, and averaged about 2.03 calories. 



C. Estimation of the " solar constant " by the pyrheUometer alone. 



As stated in last year's report, Mr. Fowle has found that a fairly close value 

 of the solar radiation outside our atmosphere may be obtained on good days 

 in Washington by making an exponential treatment of high and low sun obser- 

 vations obtained with the pyrheliometer alone and adding 14 per cent to the 

 value reached outside the atmosi)here. As was to be expected, the empirical 

 correcting factor is found to be smaller than 14 per cent when the Mount Wil- 

 son observations are treated in this manner, and it is also more variable — 

 ranging from 8 to 12 per cent — probably depending on the humidity of the air. 

 Some ten years ago Prof. K. Angstrom made observations at high altitudes 

 on the island of Teneriffe with his pyrheliometer. This instrument reads 

 lower than our standard pyrheliometer, so that a correction must be added to 

 his observations to make them comparable with ours. Applying a correction 

 for this cause and reducing Angstrom's observations by the exponential method 

 and adding 10 per cent (the mean of the corrections determined on Mount Wil- 

 son) we obtain for mean solar distance 2.25 calories for the solar constant 

 during Angstrom's Teneriffe experiments of 1895, a value about as much higher 

 than the mean given under B as the Mount Wilson mean value is below it. This 

 value is well below the highest values found by spectrobolometric work at 

 Washington within recent years. Thus the best modern pyrheliometry is not 

 discorded with the view that the " solar constant " is close to 2.12 calories. 



D. Evidence of solar variability. 



In general, the solar constant determinations obtained on Mount Wilson have 

 internal evidence of being of very great weight. For example, the exponen- 

 tial plots of the kind published as Plate VI in my report for the year ending 

 June 30, 1903, often have eight or ten points distributed over a range of air 

 mass three times that which the sun shines through in the zenith, all lying 

 within 1 per cent of the representative straight line. So far as can be seen it 

 appears that the Mount Wilson determinations of solar radiation outside our 

 atmosphere are usually accurate relatively to 1 per cent or better. The obser- 

 vations between June 1 and August 5, 1905, are on the average higher than those 

 between August 5 and October 27. But fluctuations of from 5 to 8 per cent 

 occur repeatedly, and often in periods of about ten days. 



Washington observations confirmed the results of September and October, 

 but were lacking in July and August owing to bad observing weather. In De- 

 cember excellent days were had in Washington, indicatmg low values, but in 



