116 THE ANTARCTIC MANUAL. 
so that for temperate and tropical waters 20° C. would be a very 
suitable common temperature of reduction. 
In Arctic and Antarctic seas it might be well to reduce the 
observed densities to a common temperature about the mean tem- 
perature of the waters in their place. For this purpose 0° C. might 
be taken. It is only in the restricted areas ‘of the Arctic and Ant- 
arctic seas that the adoption of 0° C. as the common temperature of 
reduction has any justification, because it is only there that the 
prevalent temperature is such that the corrections wp to 0° C. would 
ever occur, and that very seldom. In all exact work corrections 
are to be avoided if possible. When they are unavoidable they 
are to be made as small as possible, and, if possible, they should 
balance each other. The standard temperature should be the 
mean of those most likely to occur, and we have seen that, as a 
general standard temperature for all the oceanic waters, 20° C. 
would be suitable. To adopt, as has been proposed, 0° C. or the 
extreme temperature in one direction, as the common temperature 
to which to reduce all observed densities, is contrary to the scientific 
canon. 
Table IX. gives in units of the fifth decimal place the difference 
between the density at 15°°56C. of an average sea water and its 
density at any other temperature between 0° C. and 30°C. It is 
compiled chiefly from Ditmar’s observations, and the average sea 
water had a density of 1°02600 at 15°-56C. We will apply it to 
the reduction of our numerical example. The density, that is, the 
weight of 1 c.c., or the specific gravity of the water referred to the 
density of distilled water at 4° C. as unity, is at 20°C. 1°02354. At 
20° C. the density of sea water is less than it is at 15°-56 C; there- 
fore, in order to find the density of the water at 15°56 C., we must 
add the tabular difference, namely, 0°00110 to 1°02354, and we 
have 1°02464 as the density at 15°°56, often represented by the 
symbol ,§,,.,.. If now we wish to know its density at 0°C., we 
have to add the tabular difference 0:00218 to the density at 
15°-56 C., namely, 1°02464, and we have 1°02682 as the density 
at 0°C. To reduce directly from density at 20°C. to density at 
0° C., we should have to add the sum of the differences, or 0°00328 
to 1°02354. The reader will see that in using 0°C. as a general 
standard temperature, the errors in the reduction are all in one 
direction, and densities are not observed below 0° C. which would 
enable us to introduce something to counterbalance it. Supposing 
that the temperature of the water in situ is 23°°4C., then the 
density im situ is 102464 — 0°00203 = 1°02261. If the tempera- 
