19 
Particularly large biases for some historical cruises are indicated in Fig. 10. For example, most 
of the data from the 1930s were obtained during the expeditions of the R/V Discovery in the 
Southern Ocean. These data are highly biased in oxygen by 0.5- 0.6 ml/l, and in silicate (15-50 
|jmol/kg). Another problem with historical data is a large scattering of individual profile offsets 
within the cruise. We decided therefore to apply individual profile bias corrections instead of a 
cruise average correction. 
The spatial distribution of adjusted profiles is shown in Fig. 11 for all parameters. The coverage 
of available high-quality cruises made it possible to calculate biases for a total of 71776 
historical profiles.