156 
Philippine Journal of Science 
1920 
is 79.4 per cent to 89.0 per cent transmitted J to anyone looking 
down at the lower haze from a mountain, it seems as if most 
of the absorption must occur in the last 2 kilometers. Looking 
from Santo Tomas northwest over the China Sea on the morning 
mentioned, cumulus clouds were forming low over the water, and 
some of these were growing large enough to push their summits 
through the haze. The contrast in brightness between these 
white summits and the ruddy banks in the haze was extra- 
ordinary. 
Suppose that 85 per cent is transmitted vertically by the whole 
atmosphere, and that one-third of the 15 per cent absorption oc- 
curs in the haze layer ; the latter I will assume of uniform absorb- 
ing power and negligible refraction. Then the haze layer trans- 
mits w=94.45 per cent of any light normally incident upon it, 
and horizontal light, from incidence to tangency, is 0.5 per cent 
transmitted. Light passing in and out clear through the haze 
layer and tangent to the earth’s surface is 0.003 per cent trans- 
mitted. The effect of refraction is to lengthen the path of ab- 
sorption and so diminish the proportion transmitted. 
Attempts to observe the passage of the pale dawn led to dis- 
cordant results, as is natural, considering the artificial illumina- 
tion of the Manila sky and the uncertain relation of this phase 
to the Zodiacal Light. But I found that the appearance of the 
first tint of blue at the zenith gave solar depressions which were 
variable, to be sure, but for similar conditions of the sky were 
fairly concordant; and this is quite surely a terrestrial phe- 
nomenon. 
The only apparatus used was a watch, whose error was reg- 
ularly determined at noon by the fall of the time^ball of the 
Manila Observatory. This ball is dropped at 12 noon on signal 
from the standard clocks of the observatory, which are regularly 
compared with the transits of stars. In 1916 the maximum 
error in starting the time ball is reported ® as 0.5 second, and 
the average error as less than 0.2 second, which shows what 
precision to expect in general. The going of this watch was 
’ Kayser, Handbuch der Spectroscopie 3 (ed. 1905) 341, gives a table 
of nine such values. More modern data would alter the following com- 
putations considerably, but they may serve for illustrations. 
* Annual Report of the Weather Bureau for 1916. My own star observa- 
tions, entirely independent of the Observatory, justify confidence in this 
time ball. 
