82 



MEMOIRS OF THE NATIONAL ACADEMY OP SCIENCES. 



If 2^ is not constant during the day then dp + t S'p must be substituted for Sj) in the hist 

 equation. 



In § 1 there have been given values of the solar intensities obtained from the readings of the 

 conjugate thermometers. These, after the rejection of a few discordant observations, were used 

 in the above formulae for deriving values of A and p. The first step in the computation con- 

 sisted in the computation of the values of e for different hours of the day by means of the well- 

 known formuliB — 



e = sec « 



cos z = sin jc sin 5 + cos <p cos 8 cos h 



in which — 



z = sun's zenith distance, 



S = sun's declination, 



h = suu's hour angle, 



(p = latitude of the station. 

 A table was computed from which the values of e were obtained for each observation. 

 Approximate values of A and jt> were then obtained by trial computations with equation (4), 

 using a few of the observations on each day. The following were adopted : 



With these values^ equation (4) was used to give computed values of I, for comparison with 

 the observed values. The resulting residuals were used in equation (5) to give the corrections 

 SA and 3p by the method of least squares. It was found by the computation that the observa- 

 tions on April 28 were discordant, and also were not of suflicient range to give successful results 

 since they extend from 9.00 a. m. to 1.30 only. Observations taken at early or late hours are 

 necessary to give sufficiently large values of £ for the computation. .It was also found that the 

 assumed values of A and p on April 30, May 1, 2, 3 were not near enough to reduce the residuals 

 satisfactorily. Consequently a second approximation was mnde for these dates, using the values 

 of A and p derived from the first computation for the dates April 30 and May 2, but deriving new 

 trial-values for May 1 and 3. The coefficients ^^i^ and Ai, s^^/^-D were recomputed on the second 

 approximation for greater accuracy. The labor involved in these computations was very great; 

 the results obtained are given in the following tables : 



Table I. — -Values of A and p at each approximation. 



