Dataset: Synthesis Product for Ocean Time Series (SPOTS)

The data from the 12 participating time-series programs were retrieved from data centers or directly obtained from the responsible principal investigator. In the latter case, merging, formatting, additional quality-control (QC), and archiving of existing data were carried out. Only bottle data for BGC EOVs (Biogeochemical Essential Ocean Variables) that had been measured by at least two of the participating programs were included in the pilot project, along with accompanying ancillary pressure, salinity, and temperature data. The product  was created by standardizing data format, units, header names, primary QC flags, times, locations, and fill values and subsequently merging the individual datasets of each time-series program into one file. Only data that received a WOCE quality flag 2 were included in the product. Existing data were altered as little as possible without interpolation or calculation of “missing” variables. Similarly, original station-, cast- and bottle numbers were kept or created artificially if non-existent to ensure consistency. The headers, units, and flags of the individual time-series datasets were standardized to conform with the WOCE exchange bottle data format (Swift and Diggs, 2008). 

The standardization process entailed unit conversions, most frequently from micromoles per liter (µmol L-1; nutrients and dissolved organic carbon (DOC)) or from micrograms per kilogram (µg kg-1; particulate matter) to micromoles per kilogram of seawater (µmol kg-1). The default procedure to convert from volumetric to gravimetric units was to use seawater density at in-situ salinity, reported laboratory temperature (otherwise assuming 20°C as laboratory conditions), and pressure of 1 atm (following recommendations from Liqing et al., 2022). For some time-series datasets, the combined concentration of nitrate and nitrite was reported. If explicit nitrite concentrations were provided, these were subtracted to obtain the nitrate values. If not, the combined concentration was renamed to nitrate assuming that the relative nitrite amount is negligible. For the HOT program specifically, low-level, high-sensitivity measurements of macronutrients (phosphate and nitrate) were available but not included in the pilot product. Particulate organic matter was derived by subtracting the particulate inorganic matter from the total particulate matter, if available. For particulate organic carbon (POC) and particulate organic nitrogen (PON), the factors 1/12.01 and 1/14.01 (inverse standard atomic masses) were used, respectively, for the unit conversion to micromoles per kilogram. For the HOT program, particulate carbon and nitrogen measurements correspond to total particle concentrations (PC and PN), but are here assumed to approximate POC and PON. If neither temperature nor pressure was provided, all corresponding data entries were excluded from the product. The potential density anomaly is the only calculated variable. Missing and excluded values were set to -999. 

The information on the applied methods of each time-series program, was evaluated against, ideally, published Best Practices (BPs), and otherwise known standard operating procedures (SOPs). “BP Flags” were assigned accordingly to each cruise of a time-series program (Lange et al., 2023). Precision (duplicate measurements) and accuracy (deviation from reference material) estimates, as provided by each time-series program’s primary quality-assurance procedure, were assigned to the bottle data. The temporal resolution of these estimates varies from estimates given for each cruise, i.e. on a cruise-to-cruise basis, to estimates given for longer time periods (covering multiple cruises) without recorded changes in applied methodology.

When additional quality control was necessary, various methods, packages, and tools were used: CANYON-B is a Bayesian neural network mapping that accurately reproduces GLODAPv2 bottle data and the biogeochemical relations contained therein (Bittig et al., 2018); AtlantOS Ocean Data QC software is an interactive tool for quality control of hydrographic cruise data (Velo et al., 2021); the work of Tanhua et al. (2010) provided rigorous quality control procedures to assure the highest possible quality and consistency. The QC was used to identify outliers and obvious errors, as well as to quantify systematic differences in reported values using crossover analysis (Lauvset & Tanhua, 2015).

Problems, Issues, Notes:
1st QC flags W bottle flags
2 = No problems noted
9 = encompasses WOCE flags 3, 4, 5, 7, 8, and 9 (samples not drawn, questionable or bad data) 

"BP flags" (Best Practices)
1 = Methods meet required and desired "Best-Practices"
2 = Methods meet required "Best-Practices"
3 = Methods do not meet "Best-Practices"