Marine Carbonate Chemistry

Turn two independent seawater carbonate measurements into a reproducible speciation table, mineral saturation estimates, and a record of the calculation assumptions. Targets PyCO2SYS 1.8.3.4, tested with Python 3.13 and NumPy 2.5.1. The PyCO2SYS v2 beta uses a different implementation; do not mix its examples with this pin.

When to use

• Analyze bottle samples, shipboard carbonate measurements, or acidification experiments.
• Calculate total-scale pH, seawater pCO2/fCO2, carbonate ion, aragonite/calcite saturation state, or the Revelle factor from a valid measured pair.
• Convert a system determined at laboratory conditions to specified ocean conditions.
• Quantify how stated measurement uncertainties affect the calculated results.

This workflow concerns seawater carbonate equilibria. Freshwater, porewaters with substantial uncharacterized alkalinity, brines outside the selected calibration range, and reaction/transport models require additional chemistry and validation. Do not infer an air-sea flux or atmospheric carbon removal from a carbonate equilibrium alone.

Establish the measurement contract

Before running a solver, identify the two measured variables, their units, quality flags, and their temperature/pressure basis. Retain a separate source table containing station, depth, timestamps, methods, reference materials, and original QC codes, joined by sample ID. Do not turn missing values or rejected measurements into zero.

TA and DIC remain constant during the solver's temperature/pressure conversion for a closed sample. pH and gas parameters change. Two inputs measured at different conditions cannot simply share one temperature value. Establish a consistent measurement basis first. Temperature correction does not repair sample changes caused by gas exchange, biology, evaporation, or mineral dissolution/precipitation.

Use two independent carbonate parameters. pCO2 plus fCO2 is not an independent pair. Three or more measurements enable an overdetermination check: solve independent pairs and compare predicted versus measured third parameters, including their uncertainty. Do not average inconsistent solutions to hide a calibration or scale mismatch.

Install

Create a dedicated environment in the user's working directory:

uv venv --python 3.13 .venv
uv pip install --python .venv/bin/python "PyCO2SYS==1.8.3.4" "numpy==2.5.1"

On Windows the environment's interpreter is .venv/Scripts/python.exe. The commands below use the POSIX interpreter path. Set the shell variable SKILL_DIR to this installed skill's directory. Keep inputs and generated outputs in the working directory.

工作流程

1. Prepare paired measurements. Use the schema in references/input-and-results.md. Resolve units and quality flags before creating the input file. Supply phosphate and silicate explicitly; zero is an assumption to justify, not a missing-data code.
2. Choose equilibrium constants. Read references/chemistry-decisions.md for pH scales, carbonic-acid constants, borate, saturation interpretation, and uncertainty limits. Match the study's validated convention and report it. The helper supports carbonic-acid options 10 and 15; other systems require a separately verified direct PyCO2SYS call.
3. Solve with scripts/solve_carbonate.py. It validates the full input table, solves the pair, checks finite outputs and DIC species balance, then writes carbonate.csv and provenance.json into a new output directory.
4. Review flags and consistency. Inspect input and output calibration-range flags, carbonate balance, measured-third-parameter residuals when available, and controls. A successful solve does not validate the sample, constants, or measurement method.
5. Report at the intended conditions. Results ending _out describe the supplied output temperature/pressure. Unsuffixed results describe input conditions. Include parameter pair, pH scale, units, constants, nutrient assumptions, uncertainty scope, software versions, and excluded/flagged samples with the result table.

Worked example: closed-sample condition correction

The following values are synthetic, not field observations. Save this as samples.csv in a working directory. The two samples differ only in DIC; the second represents a fixed-alkalinity CO2-addition comparison. Their measurements are at 25 C and 0 dbar; results are also requested at 10 C and 1000 dbar.

sample_id,par1,par2,salinity,temperature,pressure,total_phosphate,total_silicate,temperature_out,pressure_out,u_par1,u_par2
baseline,2300,2000,35,25,0,0,0,10,1000,2,2
added_co2,2300,2100,35,25,0,0,0,10,1000,2,2

Run from that working directory:

.venv/bin/python "$SKILL_DIR/scripts/solve_carbonate.py" samples.csv \
  --par1-type alkalinity --par2-type dic --k-carbonic 10 \
  --output-dir carbonate-results

For the baseline, the tested version gives input-condition total pH 8.045886, pCO2 396.958 microatm, and aragonite saturation 3.386201. At the specified output conditions, total pH is 8.241241 and aragonite saturation 2.605691. These rounded values are regression checks for this exact setup, not universal seawater benchmarks. With independent 2 micromol/kg uncertainties in TA and DIC only, u_pH_total is about 0.004580. This excludes equilibrium-constant and other input uncertainty.

For TA + measured pH, use --par2-type ph --ph-scale total only if the source explicitly identifies total-scale pH; replace par2 and u_par2 with the measured pH and its absolute standard uncertainty. A column named merely pH is insufficient to establish its scale.

Uncertainty and interpretation

Optional u_ input columns contain absolute one-standard-deviation uncertainties. They propagate to total pH, pCO2, and aragonite saturation at each requested condition. The helper assumes independent errors and treats unlisted inputs/constants as exact. For covariance, constants uncertainty, or strongly nonlinear uncertainty, follow the decision guide and validate a tailored propagation instead of calling these outputs a complete uncertainty budget.

Omega < 1 indicates thermodynamic undersaturation with respect to the named mineral. It does not establish a dissolution rate or an organism's response. A lower pH across unmatched samples is not by itself evidence of an anthropogenic acidification trend.

Sources and validation boundary

• PyCO2SYS v1 arguments, units, settings, and result keys
• Uncertainty propagation
• Upstream validation
• Humphreys et al. (2022), PyCO2SYS v1.8

Repository tests exercise the pinned solver, independent-pair round trips, carbon balance, pH-scale equivalence, condition correction, uncertainty quadrature, CSV errors, and the worked example. They establish software behavior, not independent field-data validation.