DOXY-WA: a consolidated dataset of CTD and ARGO temperature/salinity/dissolved oxygen profiles and bottle data in the eastern tropical Atlantic

DOXY-WA v1 aggregates data taken from ARGO, World Ocean Data (WOD; CTD and OSD), GEOMAR cruises PANGEA (CTDs), and IRD cruises carried out between 2012 and 2023 (UPSEN2012, UPSEN2013, ECOAO2013, AWA2014, SCOPES2022). More details are provided in tables 1-2 below for PANGEA and ARGO data. The entire dataset was checked to remove duplicate profiles present in two distinct databases.  
The main purpose of assembling and publishing this dataset is to propose a ready-to-use data product to explore oxygen dynamics in the OMZ of the north eastern tropical ocean.  The dataset contains all valid data found over the domain -3oS-28oN and 35oW-10oW. A small number of profiles correspond to locations outside this domain. This is because ARGO floats with profiles inside the domain were systematically processed over their entire available mission time period including when they had profiles outside this domain.  
Given the subtleties of the oxygen vertical distribution the data processing strategy pays close attention to the vertical coordinate. All  original data are linearly interpolated onto regular pressure (dp = 2 dbar from surface to 750 db) and sigma-0 potential density (dsigma0= 0.01 kg/m3 from 23 to 28 kg m-3) grids with the exception of OSD and low resolution CTD from WOD23. For these data, each original data point is simply placed onto the pressure and sigma-0 level closest to where it was measured so that the corresponding DOXY-WA profiles are mostly filled with NaN values. The z coordinate is a priori more suitable to study the upper part of the water column where diapycnal processes are important while the density coordinate may be preferable to study interior ocean processes where potential density changes following fluid parcels are typically very slow.  
In addition to excluding all data with flags other than “good” or “probably good” data each profile/station was individually screened visually and compared with nearby observations both in z and sigma-0 coordinates. This led to the identification of additional bad data which were either led to limited linear interpolation or removal of the profile from the database (when the depth range of corrupted/suspicious data was too extensive).  Identification ancillary variables for each DOXY-WA profile makes it possible to return to the original data.   
Additional information is provided in the supporting information pdf document.