| Summary: | Surface samples have been collected in the North Atlantic in the past one hundred years for determining the ocean salinity and its temperature. A large share of the data we have used were collected by merchant vessels or weather ships of European countries and to a large extent are listed in reports, in particular in the Bulletin Hydrographique. We investigate whether these data are relevant for determining low frequency fluctuations of the sea surface salinity. We find many crossings in the 1920s for which salinity is anomalously high compared with the climatology or with other crossings collected on the same ship line. These anomalies are indicative of a contamination of the sample. By examining hydrographic data, reports and recent experience in collection and storage in sea water, we can attribute these large errors to unclean buckets where salt crystals dissolve into the sample and to breathing of the samples during the storage. Each of these stages contributes in estimating a too large salinity and adds to the scatter of the measurements. To further investigate these errors we compare the surface salinity and temperature for each monitoring program with nearby hydrographic casts, mostly in the eastern Atlantic. We find large differences between the various monitoring programs of different periods, and we use comparisons to empirically correct the data. Unfortunately, the number of comparisons is often too small resulting in a large uncertainty in these corrections, in particular before 1914 and for the UK and German monitoring programs before 1939 which exhibit the largest average bias in the 1920s. Despite this, we find that surface samples provide a useful complement to the hydrographic station data for investigating low-frequency variability of upper ocean waters. In the two areas where we did construct these time series: the Faeroe-Shetland Channel and the eastern Atlantic near 50-degrees-N, the surface data critically reduce the aliasing caused by insufficient sampling by the hydrographic casts. Both ...
|
|---|