locean has been in charge of collecting sea water for the analysis of water isotopes on a series of cruises or ships of opportunity mostly in the equatorial atlantic, in the north atlantic, in the southern indian ocean, in the southern seas, nordic seas, and in the arctic. the locean data set of the oxygen and hydrogen isotope (δ18o and δd)of marine water covers the period 1998 to 2019, but the effort is ongoing. most data prior to 2010 (only δ18o) were analyzed using isotope ratio mass spectrometry (isoprime irms) coupled with a multiprep system (dual inlet method), whereas most data since 2010 (and a few earlier data) were obtained by cavity ring down spectrometry (crds) on a picarro crds l2130-i, or less commonly on a picarro crds l2120-i. occasionally, some data were also run by marion benetti on an isoprime irms coupled to a gasbench (dual inlet method) at the university of iceland (reykjavik). on the locean picarro crds, most samples were initially analyzed after distillation, but since 2016, they have often been analyzed using a wire mesh to limit the spreading of sea salt in the vaporizer. some of the samples on the crds were analyzed more than once on different days, when repeatability for the same sample was not sufficient or the daily run presented a too large drift.accuracy is best when samples are distilled, and for δd are better on the picarro crds l2130-i than on the picarro crds l2120-i. usually, we found that the reproducibility of the δ18o measurements is within ± 0.05 ‰ and of the δd measurements within ± 0.30 ‰, which should be considered an upper estimate of the error on the measurement on a picarro crds.the water samples were kept in darkened glass bottles (20 to 50 ml) with special caps, and were often (but not always) taped afterwards. once brought back in paris, the samples were often stored in a cold room (with temperature close to 4°c), in particular if they were not analyzed within the next three months. there is however the possibility that some samples have breathed during storage. we found it happening on a number of samples, more commonly when they were stored for more than 5 years before being analyzed. we also used during one cruise bottles with not well-sealed caps (m/v nuka arctica in april 2019), which were analyzed within 3 months, but for which close to one third of the samples had breathed. we have retained those analyses, but added a flag ‘3’ meaning probably bad, at least on d-excess (outside of regions where sea ice forms or melts, for the analyses done on the picarro crds, excessive evaporation is usually found with a d-excess criterium (which tends to be too low); for the irms analyses, it is mostly based when excessive scatter is found in the s- δ18o scatter plots or between successive data, in which case some outliers were flagged at ‘3’). in some cases when breathing happened, we found that d-excess can be used to produce a corrected estimate of δ18o and δd (benetti et al., 2016). when this method was used a flag ‘1’ is added, indicating ‘probably good’ data, and should be thought as not as accurate as the data with no ‘correction’, which are flagged ‘2’ or ‘0’.we have adjusted data to be on an absolute fresh-water scale based on the study of benetti et al. (2017), and on further tests with the different wire meshes used more recently. we have also checked the consistency of the runs in time, as there could have been changes in the internal standards used. on the isoprime irms, it was mostly done using different batches of ‘eau de paris’ (edp), whereas on the picarro crds, we used three internal standards kept in metal tanks with a slight overpressure of dry air). the internal standards have been calibrated using vsmow and gisp, and were also sent to other laboratories to evaluate whether they had drifted since the date of creation (as individual sub-standards have typically stored for more than 5-years). these comparisons are still not fully statisfactory to evaluate possible drifts in the sub-standards.individual files correspond to regional subsets of the whole dataset. the file names are based on two letters for the region (see below) followed by –wisotopes and a version number (-v0, …): example so-wisotopes-v0; the highest version number corresponds to the latest update of the regional data set. the region two letters are the followings: so: southern ocean including cruise station and surface data mostly from 2017 in the weddell sea (wapiti cruise jr160004, doi:10.17882/54012), as well as in the southern ocean south of 20°s si: oiso cruise station and surface data in the southern indian ocean (since 1998) (doi:10.18142/228) ea: 20°n-20°s cruise station and surface data (since 2005), in particular in the equatorial atlantic from french pirata (doi:10.18142/14) and egee cruises (doi:10.18142/95) na: 20°n-72°n station and surface data, mostly in the north[...]