14 
the number of observations at a depth level, spatial distribution of the data, data quality and 
natural variability. The percentage of rejected observations per one-degree latitude zone is 
given in the Fig. 8. 
There exists a clear decrease of percentages of rejected observations for each parameter with 
time (Fig. 9). This indicates improvements in observational instruments and methods. The data 
from the 1990s (the main operational phase of WOCE) are characterised by the lowest 
percentage of outlier. 
Year 
100 
1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 
I I I I I I I I I 
80 
60 
40 4 
20 
0 
100 
CO 
15 
T3 
T3 
ф 
'S 
ф 
'ф 
о 
ф 
О) 
со 
ф 
0. 
80 - 
60 
40 Н 
20 
0 
100 
80 
60 
40 Н 
20 
0 
Z-T-S CHECK 
TTTTTTTTTJTTTTTTTTTJTTTTT^^^r 
I I I I I I I I I 
OXYGEN QHECK 
MTjiiiimiTyiTHlitff|Tiiiiirf 
llllllllfjl 
- 40 
30 Ö 
1-20 Z 
10 
0 
NITRATE СНЕСИ;* 
IITplTnilll|lllTlllW|ITII 
100 
100 
100 
I I" 
1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 
Year 
Fig. 9: Percentage of rejected observations versus year (red) and the number of 
observations per year (blue). 
3.3 Systematic errors 
The problem of estimating systematic errors in the data was not properly addressed until 
recently. Identification of systematic errors was usually made at the stage of the subjective