1903 - 4 .] Prof. C. G. Knott on Ocean Temperatures , etc. 177 
temperature with increase of depth is very steadily maintained 
throughout the whole series of observations, and is best given by 
the means of all (see Table B, and the figure on page 181). 
Table C contains the differences of temperatures at correspond- 
ing depths at pairs of stations, at which the times of observations 
differed by approximately half a day. The precise difference in 
time in any case can be found from Table A. In all it will be 
seen that there are just nineteen pairs of stations available for the 
inquiry. If the waters to a depth of 100 metres were heated up 
during the day by direct solar radiation, and cooled off again during 
the night, these differences should all be positive. A glance shows 
that out of the nineteen there is one negative value at the surface, 
three at a depth of 2 metres, four at 5 metres depth, five at 10, 
four at 20, eight at 30, eight at 50, eleven at 70, and eleven at 
100. At depths greater than 20 metres there is no evidence of 
penetration of solar radiation. Even at 20 metres it is doubtful 
if we can find any evidence of direct daily heating. We may, 
however, take the means of the differences at each depth, and then 
test the sufficiency of the observations by calculating the probable 
errors in the usual way. The result is as follows : — 
Depth in 
Mean Daily Difference 
Probable 
Metres. 
of Temperature (C.). 
Error. 
0 
0-48 
± 0-057 
2 
0-37 
± 0-065 
5 
0-29 
± 0-062 
10 
0*22 
± 0-088 
20 
0*09 
± 0-12 
30 
- 0*08 
50 
- 0’09 
70 
- 0’18 
• • • 
100 
- 0*11 
The thermometers read to tenths of degrees, so that little value 
can be attached to the second decimal place. 
It would obviously be wasted labour to calculate the probable 
errors for the last four depths. At depth 20 metres the probable 
error is numerically greater than the mean ; so that we can say 
nothing definite as to the effect of solar radiation at this depth. 
The errors are so great that we may, without running any risk 
of introducing greater errors, combine these numbers by a linear 
formula, assuming that the difference of temperature t between 
morning and afternoon in the waters of the Mediterranean during 
PROC. ROY. SOC. EDIN. — YOL. XXV. 12 
