The processing workflow described below to produce this dataset (Grace, 2026; doi:10.5061/dryad.wh70rxx2q) contains all the files described below associated with this geochemical project; raw data, analytical order and parameters, quality assurance and control, sample and standard-specific data, and subsample-specific means.

The instrument measured samples following the analytical order and analytical parameters described by files in the raw_data > analytical_order_and_parameters folder, with specific information for the first cruise and, in aggregate, the second and third cruises:

Raw data output from the instrument are included in the "runs" folder:

raw_data > runs

Unprocessed intensities versus time (counts per second; CPS) for measured trace elements (analytes) output from spectrometer for each analytical session (run), organized by cruise (e.g., at4224_runs; tn391_at5004_runs).

These data were processed using IOLITE (v.4) trace elemental reduction software (Paton et al., 2011) following Longerich et al., (1996) with MACS-3 as the calibration material. When used in IOLITE, these files require a "laser log" to be read; a laser log has been included for each run. See the README file for a full organizational scheme that associates each log with each run. This software output sample-specific continuous trace elemental data in units of CPS, X:Ca CPS (elemental ratio to calcium), and ppm (calibrated values), included in the "sample_and_standard_specific_data" folder:

raw_data > sample_and_standard_specific_data

Sample-specific continuous data for samples from each cruise. Each sheet is a sample (whole valve). Data are in units of: i) CPS, ii) X:Ca CPS, and iii) ppm (calibrated).

The 'timeseries' datasets contain multiple sheets; each sheet is data from the transect of one valve, from the start of the valve (at or near the umbo) to the edge of the shell across the maximum growth axis.

The 'whole_valve_means_macs3' datasets contain mean values for whole valve samples and standards, averaged across each transect. Also included are standard errors (2x) and levels of detection.

The 'whole_valve_means_all_calib' dataset contains the same values as the 'whole_valve_means_macs3' datasets, but for all calibration materials (e.g., NIST612, GSE-1G, etc.). This dataset was used to determine that MACS-3 was the best calibration material for these samples.

Whole-valve data were used primarily for quality assurance and quality control procedures, which included i) comparison of magnitude of signal versus level of detection and ii) assessment of sample % residual standard deviation (% RSD; precision).

Additionally, measurement of standard reference materials NIST 612, GSE-1G, ECRM 752, and MACS3 facilitated assessment of standard % RSD and recoveries (accepted versus measured values). Approximately every 15 samples were bracketed by standard reference materials ECRM-752-NP, NIST 612, and USGS GSE-1G to evaluate instrument precision (residual standard deviation; % RSD; ideally below 15%) and accuracy (ratio of average analyte concentration to NIST-certified values or GeoReM Preferred Values; ideally close to 1). Datasets containing values for measurements of standards are above in the "sample_and_standard_specific_data" folder and in the "qa_qc" folder:

raw_data > qa_qc

The file 'magnitudes_rsd_recoveries' includes a comparison of the magnitude of signal versus level of detection, sample % RSD, and the standard % RSD for MACS-3, as well as recoveries for standards calibrated using MACS-3.

The file 'reference_material_values' contains the GeoReM reference material values for MACS-3, ECRM 752, and NIST 612 in units of ppm, with uncertainty.

Average % RSD for reference materials was below 15%, and recovery across analytes was close to 1 for MACS-3, which was therefore used as the primary calibration standard. Precision and other quality assurance metrics were best for ²⁵Mg, ⁵⁵Mn, ⁶⁰Ni, ⁸⁸Sr, and ¹³⁸Ba.

The MATLAB program 'WindowProgram' (processing_data_and_tools > WindowProgram) was used to smooth the original data ('origppm') for each whole-valve sample using 7-point moving median and mean windows, and output the 'reduced' dataset.

Because larval shells have up to three shell growth regions (i.e., larval, prodissoconch growth regions I and II; settled, juvenile dissoconch), it was necessary to isolate geochemical measurements within each shell growth region. To this end, MATLAB was used to set the start point (along the laser transect) for prodissoconch I for small to medium-length valves following SEM-based growth region length measurements for the species (Arellano et al., 2014), although we applied more conservative intervals. That is, for most valves, prodissoconch I was automatically terminated at 75 µm for G. childressi and 100 µm for B. heckerae, and prodissoconch II was set to start at 150 µm for G. childressi and 175 µm for B. heckerae. For particularly long valves, prodissoconch I was sometimes not selected at all or instead was terminated at 50 µm for G. childressi and 75 µm for B. heckerae; likewise, prodissoconch II was set to start around 125 µm for G. childressi and 150 µm for B. heckerae. Growth region intervals were manually set for valves presenting issues such as laser ablation burn-through or unusual elemental time series using RegionPicker, a custom-built MATLAB application for comparing valve images alongside geochemical time series. Additionally, an elemental transition zone approximately 30-50 µm long was apparent in time series between each growth region for most valves, possibly reflecting distinct but gradual shifts in growth rate, physiology, metabolism, or change in other vital effects. Thus, an automatic offset to each growth region was applied to avoid overlapping regions. The concentration of tin (Sn¹¹⁸) was used as a proxy for tape ablation (shoot-through) despite poor precision and accuracy for tin. Intervals with very high tin concentrations (e.g., > 4 log X:Ca) were typically excluded regardless of visible holes in the valve, as were intervals with low tin concentrations that did overlap visible holes. Intervals absent of tin were typically not excluded despite visible holes, as this could indicate post-ablation valve collapse in lieu of shoot-through during ablation.

processing_data_and_tools > RegionPicker

The 'valve_images' folder contains z-stacked mosaic images of each valve after ablation. These images are seen alongside geochemical time series within the RegionPicker program.

The 'RegionPicker_screenshots' folder contains screenshots of each valve sample in its final processed state in the RegionPicker application.

The 'RegionPicker_output' spreadsheet contains metadata for each sample processed in RegionPicker, including the intervals along the transects designated by the user for later extraction (of the geochemical time series).

MATLAB was used to extract shell geochemical data associated with defined intervals (i.e., rather than using data from entire transects or entire growth regions) using IntervalExtractor.mlx:

processing_data_and_tools > IntervalExtractor

The 'combinedoutput' spreadsheet (and .mat file) is simply a renamed version of the RegionPicker_output spreadsheet. The IntervalExtractor program output 'subsamples_extracted', a dataset of mean elemental values in ppm for each subsample (growth region-specific dataset). These data were then organized as follows:

raw_data > subsample_specific_means

The dataset 'means_by_run' contains multiple sheets; each sheet contains those subsamples that were measured during a given analytical session (run).

The dataset ' means_agg' aggregates all subsamples in one sheet (columns B-P). These values were used in the conversion of elemental ratios to mmol:mol X:Ca ratios.

Before conversion of elemental ratios to mmol:mol X:Ca ratios, values below level of detection (LOD) and outliers were identified:

processing_data_and_tools > data_cleaning

The file 'below_lod_and_outlier_removal' contains six spreadsheets associated with a step of outlier identification and removal of values below LOD. First, subsample values for indivdual elements below LOD were identified; second, values below LOD were removed; third, prodissoconch I shell growth region values above 3 standard deviations + mean were identified as outliers; fourth, this same criteria was applied to prodissoconch II values; fifth, this same criteria was applied to dissoconch values. Finally, all values below LOD and those identified as outliers (> 3 standard deviations + mean) were removed (see the 'outliers_removed' sheet).

The file 'subsamples_conversion_to_ratios' contains four spreadsheets associated with the process of converting subsample elemental values in ppm to ratios to calcium in units of mmol:mol. First, the ratio of trace elemental values in ppm to bulk calcium in ppm was calculated. Second, the relative atomic mass for each trace element was used to convert the units of that ratio to mmol:ppm (i.e., Mg24 in mmol:bulk calcium in ppm). Third, the natural abundance of each trace element was used to calculate concentrations of bulk elements (e.g., Mg instead of Mg24) in mmol. Fourth, the conversion factor of bulk Ca in ppm to Ca in mol was used to convert these ratios to bulk elemental concentrations in mmol relative to bulk calcium in mol (i.e., Mg:Ca mmol:mol).

This produced the following dataset:

processed_data_preanalysis

The dataset 'subsamples_xca_mmolmol' is the highest-level, reduced and cleaned dataset input to the statistical analysis software PRIMER, but is not itself prepared for statistical analysis (i.e., it has not been normalised, transformed, grouped by shell growth region, etc.).