'24 Transactions of the South African Philosophical Society 



The two earliest years had one observation of cloud in the morning 

 and one in the afternoon ; but the three later had two in the morn- 

 ing and two in the afternoon." The Eoyal Observatory averages 

 are also probably fairly good, since they agree on the whole with the 

 Simonstown results. A change of method on the part of a single 

 observer appears in the sudden drop in the estimated percentage of 

 cloud at Graaff Eeinet in 1892. For the nine years earlier than 1902 

 the average appeared as 39 per cent., varying from 35 per cent, to 

 44 per cent. ; in the next seven it appeared as 16 per cent., varying 

 from 12 per cent, to 21 per cent. But a new observer in 1900 

 immediately doubled this last low estimate. Defects being under- 

 stood, we cannot always compare the cloudiness of one station with 

 another, but only, sometimes, the annual variation at the same 

 station. 



From Table 28 I have selected a number of w T hat seemed to be 

 the best of the cloud averages, and determined the constants as far 

 as the third harmonic term, putting in also the corresponding con- 

 stants for rainfall, together with the rainfall constants for Adelaide 

 and Alice Springs (South Australia), and Cordoba (Argentine). It is 

 interesting to compare the South African numbers in this Table with 

 those of Table 26, these being for stations, those for areas. The 

 rainfall constants given here for Kimberley differ somewhat from 

 those in the full formula computed above directly from Tables 1-12, 

 for various reasons. The last are obtained from twenty-four nearly 

 equal portions of a year. The values given in Table 29 (with the 

 exception of Cordoba) are computed from the twelve months of 

 various lengths reduced to thirty days. Thus the constants apply 

 to the rate per month, rather than to aliquot parts of a year. The 

 error introduced makes the times of the turning-points some four 

 days too early in the first term, and seven days in the second. For 

 comparative purposes only this is not of any great importance. In 

 the third and later terms, however, the error is likely to be more 

 serious. In the matter of the Cordoba numbers, the magnitudes 

 given in " el Clima de Cordoba" are reckoned from January 1st. 

 They are also determined from the average rainfall during each one- 

 twelfth of a year. Before quoting them here they have been altered 

 to count from the middle of January, so as to compare better with 

 the others. 



It appears from Table 29 that there is no very near approach to 

 agreement between the epochs of cloud and rain for the different 

 stations. At Durban, East London, and Kimberley the two elements 



* At Torquay, for the year 1899, the total percentage of (sunshine plus cloud) 

 uring daylight is 104-4. A. Chandler, Met. Rep.fdr 1899. 



