Element data were corrected for blank, drift, mass bias, and calcium matrix effects.

Boron isotopes were corrected using the standard-sample-standard bracket technique to correct drift and bias across each individual sample.

pKB* and δ11B borate were calculated using Branson (2024) CBSYST with implementation of Hain et al. (2015) speciation of Ca and Mg.

The MarChemSpec (Clegg & Turner, 2023) model uses measured inputs of salinity, temperature, DIC, TA, [B], [Mg], [Ca], [NO₃^-], [PO₄^-], and [SiO₄^4-]. Other major ion concentrations were estimated by applying LOESS smoothing regression (R Software version 4.3.1) to salinity versus [Cl^-], [SO₄^2-], and [Sr²⁺] data from a lagoon transect in Young et al., (2008) to account for minor non-linear variations at lower salinity. [Na] and [K] concentrations were estimated from the LOESS-derived [Cl^-] using [X] = [Cl]_estimated/[Cl]_Millero * [X]_Millero, where [Cl]_Millero = 536 mM and X is typical seawater [Na] = 456 mM or [K] = 10.2 mM (Millero et al., 2008). These estimates apply for mixed lagoon waters only. For endmember waters (well water, seawater, large spring), values from were taken directly from Young et al. (2008) and assumed to reflect expected concentrations in the identical sampling locations in this study. Charge imbalances from MarChemSpec calculations were corrected by adjusting [Na], chosen because it was not measured and because its seawater concentration dominates charge balance adjustments relative to [K]. [Na] absolute adjustment ranged from about -22 to +19 mM (< 10 % of [Na]). MarChemSpec calculated pH on the total scale using Pitzer constants.

Software:

MarChemSpec software (https://marchemspec.org/wp/software/) has been archived at Zenodo (Clegg & Turner, 2023, doi: 10.5281/ZENODO.8373046).

CBSYST software (https://github.com/oscarbranson/cbsyst) Version 0.4.9 has been archived at Zenodo (Branson et al., 2024, doi: 10.5281/ZENODO.14185784).

See the "Parameters" section for more details applicable to each data column in this dataset including how MarChemSpec and CBSYST software was used.

Note about aggregated data (means, standard deviations):
The averages are from the instruments themselves, not from true replicates. They typically take 4 to 10 readings and average the result, depending on the instrument. Since these are mass spectrometers counting individual ions, this averaging is necessary to reduce noise. For parameters calculated using CBSYST, mean and standard deviation are derived from 10,000 Monte Carlo realizations.