8 
2.4 Subdivision into reference and historical data 
As shown by Gouretski and Jancke (1999) based on a comprehensive data-set for the South 
Pacific, the accuracy of oceanographic measurements has changed considerably over the last 
decades, with earlier historical data exhibiting usually larger scattering and larger systematic 
biases. Therefore the selection of the historical data for the analysis is not exhaustive of the 
NODC CDROM data collection, but represents the data with the best quality indices as 
determined by a quality control procedure (Gouretski and Jancke,1999) briefly outlined in 
Section 3. Based on this data quality assessment a composite data set (Fig.1) was divided into 
two subsets, to avoid simultaneous treatment of data of substantially different quality: 
a) A reference data set, comprising high quality cruises occupied after 1970 (a total of 19867 
profiles distributed over 384 cruises). 
b) A historical data set, comprising older cruises occupied mostly before 1970 as well as 
profiles not included into the high-quality subset (a total of 1039668 profiles distributed over 
41757 cruises). 
WOCE data provided the basis for the high-quality reference data set. The reference data 
set includes also high-quality non-WOCE data. Though the total number of high-quality 
profiles is an order of magnitude less compared with the historical data, the high-quality 
profiles provide a reference against which the historical data are validated. The reference 
dataset was compiled for the study of systematic offsets in the hydrographic data and is 
described in details by Gouretski and Jancke (2001). 
2.5 Time- and spatial distribution of the data 
The depth coverage of the hydrocasts (e.g. the ratio of the last observed depth to the local 
bottom depth) is shown in Fig. 2. The coverage is better for the shallow regions of the 
ocean, whereas below about 1500 meters historical hydrocasts on average extend only to 
the ocean mid-depth. 
0.0 0.2 0.4 0.6 D.e 1.0 
Dee pest_S am ple_De pt h/ Botto m_De pt h 
Fig. 2: Average ratio of deepest sample depth to bottom depth versus sample depth.