12 
Depending on the spatial distribution of the data the size of the geographical box within 
which a mean density-parameter relation is assumed to be spatially invariant varied between 
111 km by 222 km and 333km by 666 km. 
Eight parameters were subjected to quality control: potential temperature, salinity, oxygen, 
silicate, nitrate, phosphate, ratio [NO]/[PO]. Since every density value (determined by T and 
S) is characterised also by the depth of the respective isopycnal, the depth of the T-S pair 
falling within a density bin was also checked as an independent parameter. The following 
explains the inclusion of the statistical check for the sample depth. 
Before the introduction of CTD-type probes, which measure pressure (depth) with an 
accuracy of about 0.1%, a common oceanographic practice was to estimate depth by means 
of wire length and angle between the bottles. The bottle trip depth was (sometimes) 
determined by means of pairs of protected/unprotected thermometers attached to some deep 
bottles. For a considerable number of historical casts unprotected thermometers were 
Nunberol observations f"10 
o 40 so 133 lee zoo 
lAjrrter ol observations I" 10^j 
o 33 40 eo so ioo 
- -1000 
J= 
- 40» 
Nunberol observations f10- 2 j 
0 33 40 60 30 100 
Ibiuber ol observations (‘10- 3 ) 
0 33 40 60 30 100 
- -1000 
-3300 
- <4000 
I 
I 
Fig. 7: Percentage of rejected observations (red) and a total number 
of observations (blue) versus depth. Depth bin width is 1 meter. 
not used or their number was insufficient to determine the depth of each sample. The 
frequency distribution of sample depth is depicted in Fig. 5 and demonstrates that