Temperature and salinity data from the 1991 cruise of the Norwegian research vessel JOHAN HJORT were obtained using a Neil Brown CTD with an underwater pressure sensor specified for a maximum pressure of 3000 dbar. Precision of the salinity data collected during this cruise was about +-0.002. In 1992, a Neil Brown CTD with an underwater pressure sensor specified for 6000 dbar was used. Precision of the 1992 salinity data was about +-0.003. In 1993 and 1994, a Seabird CTD with dual TC (temperature, conductivity) sensors was used. Precision of the 1993 and 1994 salinity measurements was about +-0.001. For calibration purposes bottle salinity samples from depths below 1000m were analysed on board JOHAN HJORT in 1991, 1992, and 1993 using a Portasal Model 8410 salinometer, and on shore in 1994 using an Autosal Model 8400 A salinometer.
Precision of the temperature measurements was +-0.001 C for all cruises. In 1991 and 1992, electronic SIS reversing thermometers were used to control the CTD (no correction was applied to the CTD data).
The first temperature and salinity data from the Greenland Sea date back to the beginning of the century. The quality of these measurements, however, is unclear. For this reason, we start our temperature time series in 1952 and the salinity time series in 1981.
We have to restrict the salinity time series to the 1980s, because the small changes in salinity (less than 0.01) are comparable to the noise level of most of the earlier data sets. At least one early survey however, the 1972 GEOSECS data, has the required data quality. Since the gap between 1972 (GEOSECS) and 1981 (TTO/NAS) is too large for an adequate interpolation, the GEOSECS salinities are only used as a reference for conditions in the early 1970s. In general, it is difficult to detect the small year to year changes in salinity that occur in the deep water even with modern techniques, and intercalibration uncertainties of a few ppm are difficult to exclude. However, the quality of the data is sufficiently high to show long-term trends in salinity during the 1980s and early 1990s.
In contrast, temperature data since 1952 have the precision and accuracy required to resolve the observed changes which are of the order of 0.1 C. The main source of the historical temperature and salinity data used in this study is the NODC data CD ROM (National Oceanographic Data Center [1991]). We used only the Nansen bottle cast data, because these are of the highest quality. Stations which did not reach a depth of at least 1500m were omitted from this study. Also omitted were data points below 1500m which fell off a mean profile constructed for each individual cruise by more than 2 sigma (1sigma : 0.01 C and 0.003 for temperature and salinity, respectively).