Seawater DMS and DMSP Measurements, Chromatograms and Flux

The reading problem: what does a seawater DMS number mean?

A report giving seawater dimethyl sulfide, or DMS, as 8 nanomoles per litre is describing a dissolved volatile sulfur compound, not the concentration of every reduced sulfur molecule in the sample. The number has a sampling time, an analytical method, a calibration, a detection threshold and an operational definition. Reading it well means reconstructing those conditions before deciding whether eight is high, low or changing. A reader who skips the method may unintentionally compare samples that contain different combinations of dissolved DMS, cell-associated sulfur and compounds converted during preparation.

Dimethylsulfoniopropionate, abbreviated DMSP, is a different chemical species. Some phytoplankton and other organisms produce it; microbial pathways can transform it into DMS or into other sulfur products. The presence of a large DMSP pool therefore does not imply that an equivalent DMS pool exists at the same moment. The distinction is like separating inventory from throughput: DMSP tells us something about precursor availability, while measured DMS reflects production, consumption, exchange and mixing acting together.

Because DMS can escape to air, both how a sample is collected and how quickly it is handled can change the result. The practical question is not simply whether a chromatogram has a convincing peak. It is whether that peak can be traced back to the original seawater without uncontrolled losses, contamination or conversion of precursors. This guide walks from chemistry to sampling, separation, calculations and uncertainty, and then uses hypothetical data to show what the evidence supports and what it leaves unresolved.

First principles: a molecule, a pool and a rate are not interchangeable

DMS is a neutral volatile molecule with the formula (CH3)2S. Its sulfur atom can pass through multiple transformations in seawater. DMSP contains a sulfonium group and is often reported in dissolved, particulate and total operational fractions. A dissolved DMS concentration typically uses a molar concentration unit such as nmol L−1, while a production rate adds a time denominator such as nmol L−1 h−1. A flux across the sea surface needs an area and time denominator, often µmol m−2 d−1. Those three outputs answer different questions.

A concentration is a snapshot of the pool within the sampled water. A rate describes how quickly a process moves material into or out of that pool. A flux describes transfer across a specified boundary. If DMS production and loss are both rapid, concentration may remain almost unchanged. Conversely, water movement can alter concentration without much local biological production. Interpreting a positive concentration gradient as a measured production rate therefore confuses state with process.

Suppose two bottles each contain ten nanomoles of DMS per litre. One has abundant DMSP and strong biological turnover; the other has little precursor and slow exchange. Their DMS concentrations agree but their futures may differ. A useful reading connects the measurement to the question being asked: dissolved inventory, microbial transformation, air–sea transfer or chemical source attribution. Write the intended quantity in words before selecting equations, or a technically correct conversion may end up calculating the wrong thing.

Why marine researchers measure this sulfur compound

DMS matters because it connects microbial sulfur metabolism with atmospheric sulfur exchange. The ocean surface contains organisms, particles and dissolved organic compounds that produce and consume sulfur-containing chemicals. Some DMS escapes to the atmosphere and participates in subsequent oxidation chemistry. However, it would be misleading to claim that a single DMS observation predicts clouds, regional climate or the complete direction of climate feedback. Those questions need additional atmospheric processes, transport and observation constraints.

The NOAA Global Surface Seawater DMS Database illustrates the importance of metadata. Its existence does not mean seawater contains one universal DMS background; measurements vary with location, season, water mass, ecosystem state and sampling method. A meaningful comparison asks whether the two observations refer to comparable depths, analytical fractions, time windows and calibration standards. A regional average is a statistical description of a defined dataset, not a substitute for reading the actual measurements.

For a student encountering ocean chemistry for the first time, a useful analogy is a busy railway station. The number of passengers on a platform is the standing inventory. Arrivals and departures are opposing fluxes. A busy station can maintain a nearly constant crowd, while a closed station may briefly hold the same number. DMS measurements behave similarly: concentration is what is present, while pathways and exchanges determine how that amount changes.

Sampling must preserve the meaning of dissolved

Water taken from a Niskin bottle, surface intake or another sampler is already a modified environment. Pressure, temperature, turbulence, bubbles and contact with tubing can influence volatile compounds. The first priority is to define a sampling path that limits gas exchange and avoids crushing cells when the aim is dissolved DMS. Record sampling depth, bottle ID, elapsed time, temperature, volume, headspace, filtration method if used and any chemical treatment. These details establish whether later variation can reasonably be attributed to the ocean.

Several tempting shortcuts are analytically dangerous. Pouring through open air, vigorously shaking a sample or leaving a large air pocket can favour DMS loss. Aggressive filtration may rupture phytoplankton and release intracellular precursors; depending on the procedure, DMS may then form during processing. Delayed handling creates an interval in which biological production and consumption continue. A sample label that says only ‘surface water’ cannot explain these changes.

Operational definitions should be stated as part of the result. ‘Dissolved’ may mean material that passed a defined filter, but that filtration itself can alter volatile sulfur chemistry. A whole-water analysis may incorporate transformations from particulate material during handling. The right method depends on the scientific question and its validated protocol. Do not retroactively call a whole-water result a dissolved concentration, and do not assume all laboratories implement identical size cutoffs or preservation steps.

Purge-and-trap analysis: what the instrument actually sees

In purge-and-trap analysis, an inert carrier gas is passed through or over a measured seawater aliquot to transfer volatile DMS from liquid to gas. A sorbent or cryogenic trap may concentrate the analyte before it is released into a gas chromatograph. Chromatographic separation reduces interference from other volatile molecules. The detector then produces a response associated with the DMS peak. Depending on the method, detection may use sulfur-sensitive systems or mass spectrometry; the exact instrument configuration matters for selectivity.

The gas chromatogram does not itself contain a direct seawater concentration. It records a detector signal over retention time. A peak must be assigned to DMS using validated retention behaviour and, where appropriate, mass-spectral evidence, standards and blanks. Next, a calibration function connects response with quantity. Extraction efficiency, trapped fraction, detector drift and the injected sample volume all influence that relationship. An attractive peak shape cannot repair incorrect calibration or a sample taken after significant volatile loss.

Purge-and-trap should also be distinguished from direct headspace injection. In a headspace method, a defined gas phase reaches equilibrium with water and the gas fraction is analysed; the dissolved amount is recovered through an appropriate gas–liquid mass balance. In purge-and-trap, the intention may be to extract a large fraction of the volatile analyte. These methods can both be valid, but their equations, recovery checks and handling sensitivities are different. Never apply a headspace equilibrium equation to a purge-extracted result without method-specific justification.

DMSP is a precursor family, not an alternative name for DMS

DMSP analyses commonly distinguish dissolved DMSP, particulate DMSP and total DMSP. This separation can be operational rather than absolute because fractionation itself changes the sample. For example, if a validated total measurement is thirty nanomoles per litre and a comparable dissolved measurement is eight, the difference of twenty-two nanomoles per litre is an estimated particulate-associated fraction. It is not a claim that twenty-two nanomoles of DMS gas must appear if the sample is left standing.

Certain assays convert DMSP chemically into DMS and then measure the generated DMS. The numerical relationship relies on reaction stoichiometry, conversion yield, blank correction and any DMS already present. Treating the final DMS signal as naturally occurring DMS without subtracting initial DMS would merge two analytically distinct pools. The sample preparation record must show whether an aliquot was analysed untreated, chemically treated, separated into fractions or frozen under a validated protocol.

Consider a laboratory worksheet containing DMS, total DMSP and dissolved DMSP in adjacent columns. A translator, data analyst or reader must preserve each qualifier. Replacing ‘total DMSP’ with ‘total DMS’ changes the compound. Replacing ‘particulate’ with ‘solid sulfur’ loses the operational meaning. Such distinctions are not pedantry: they decide which process interpretations and budgets are valid.

Calibration translates detector output into a measured amount

A trustworthy calibration contains standards whose concentrations or injected masses are known through preparation and verification. The detector response should be evaluated over an appropriate working range rather than assumed linear at every possible signal. If the regression is response = 50 times concentration in nM plus 20 signal units, a sample response of 470 gives (470 − 20)/50 = 9 nM before any method-specific dilution or recovery correction. The intercept matters because simply dividing 470 by 50 would overstate the result.

Suppose the analysed aliquot was diluted one part sample with one part suitable diluent and the method showed that the procedure did not change DMS recovery. Then the original concentration would be 18 nM, not nine. But dilution of a volatile sample can create extra handling losses and is not automatically equivalent to diluting a nonvolatile salt. The analyst should confirm that the validation covers this treatment; otherwise the arithmetic alone is insufficient evidence.

A calibration check is different from a blank. A blank tests contamination or background, while a check standard tests response at a known level. A continuing check can reveal drift after several samples. A matrix spike can reveal recovery problems in seawater relative to prepared solutions. A field blank can expose contamination introduced by the sampler or containers. A reader who sees only a beautiful calibration R-squared yet no blank, recovery or drift evidence should treat the concentration with appropriate caution.

Unit discipline: convert the measured quantity before interpreting it

Nanomoles per litre, written nmol L−1 or nM, refer to amount of substance per liquid volume. They are not interchangeable with nanomoles per kilogram without seawater density. A concentration of eight nmol L−1 in a five-millilitre aliquot corresponds to 8 nmol/L times 0.005 L = 0.040 nmol, or forty picomoles. The small amount is entirely consistent with a nanomolar concentration; the volume determines how much analyte was placed in the system.

A concentration of eight nmol L−1 is eight micromoles per cubic metre, because one cubic metre holds one thousand litres and one micromole holds one thousand nanomoles. This equality is especially convenient when matching a dissolved concentration gradient with a gas transfer coefficient expressed in metres per day. It does not mean the two units look identical on a label. The written conversion protects the denominator and makes a flux calculation independently checkable.

For a mass concentration, use the molecular mass of DMS, approximately 62.13 grams per mole, with the correct powers of ten. Eight nanomoles per litre multiplied by 62.13 grams per mole gives about 0.497 micrograms per litre. The figure describes DMS mass, not elemental sulfur mass; sulfur alone would use a different molar mass. When comparing older publications, note whether reported sulfur has been expressed as compound or as elemental sulfur equivalents.

Worked example: a calibrated, blank-corrected surface sample

Imagine a teaching dataset, not an observed cruise record. A set of standards indicates a linear response over the relevant range. The instrument-reported equivalent concentration for one sample is 8.6 nM, while its associated process blank is 0.6 nM. Under a method that allows concentration blank subtraction, the corrected result is 8.0 nM. State which blank was used and whether it went through the same processing steps. A solvent-only blank does not necessarily capture contamination from collection bottles.

Now imagine replicate corrected values of 7.8, 8.0 and 8.2 nM. Their arithmetic mean is 8.0 nM and their sample standard deviation is 0.2 nM. This demonstrates repeatability of three processed aliquots under those conditions; it is not a full uncertainty estimate for collection, preservation, extraction and chemical specificity. A method can be repeatable and biased if every bottle loses the same proportion of DMS before analysis.

The blank is 0.6/8.6, or roughly seven percent of the uncorrected value. That is large enough to report rather than hide. If the sample instead produced 0.5 nM with the same nominal blank of 0.6 nM, a negative subtraction would not prove negative DMS in seawater. It would indicate the measurement lies at or below the method’s useful background resolution. The defensible report would use its validated blank and detection-limit policy rather than forcing a biologically impossible sign.

Worked example: the difference between a pool and a time trend

Imagine that a closed bottle begins at 12 nM DMS and, after twenty minutes, measures nine nM. The arithmetic slope is (9 − 12) nM divided by one third of an hour, or −9 nM per hour. This describes the net change in that bottle over the measured interval. It does not by itself identify bacterial DMS consumption, because gas exchange with headspace, photochemistry, leaks and sample manipulation could also alter the measured pool.

To support a process claim, researchers design incubations with time-zero controls, appropriate treatment and dark or light comparisons when relevant, and replicate bottles whose integrity is checked. If a killed or otherwise validated control loses a similar amount of DMS, that observation warns against attributing the entire change to living metabolism. Yet control manipulations can themselves perturb chemical reactions, so even a difference between treatments must be interpreted in light of the control’s limitations.

Rates also depend on which interval is chosen. A change from 12 to nine nM over twenty minutes does not guarantee another three-nanomole decline in the next twenty minutes. Microbial activity, substrate supply, oxygen, temperature and mixing can change. If the relationship is nonlinear, a straight line through only the endpoints may be a poor summary. Plot concentrations with time, replicate spread and method blanks before giving one number the status of a process constant.

How to read an air–sea DMS flux estimate

One simplified gas transfer expression is F = k(Cwater − Cequilibrium), with k a transfer velocity and concentrations expressed in compatible volume units. A positive value under an explicitly stated sign convention means flux from water to air. The equilibrium concentration depends on atmospheric conditions, temperature, salinity and the gas’s solubility relation. A concentration measured in a bottle is not itself an air–sea flux; the boundary-exchange model supplies additional assumptions.

Consider a deliberately simplified example: Cwater = 10 nM, Cequilibrium = 2 nM and k = 2 metres per day. The gradient is eight nM, numerically eight micromoles per cubic metre. Multiplying by two metres per day gives sixteen micromoles per square metre per day. This works because the cubic metres in the concentration denominator cancel one metre from transfer velocity. It is an arithmetic demonstration, not a field flux estimate or a recommended k for any wind condition.

A real flux calculation also needs a justified wind and sea-state parameterisation, gas solubility, surface temperature, atmospheric concentration and treatment of surfactants, waves and spatial variability. The gas transfer coefficient may be adjusted to a reference Schmidt number and then back to the gas under study. Do not lift an oxygen or CO2 transfer value into a DMS calculation without checking the formulation. Uncertainty in the physical exchange model can be substantial even when the DMS assay is precise.

Recovery checks are more informative than a single attractive chromatogram

An extraction method may fail differently at low and high DMS levels. Spike recovery testing assesses whether an added known amount can be recovered through the full workflow; it should include a suitable matrix and the same container, timing and transfer steps as ordinary samples. If a method recovers only eighty percent of a spike, applying a universal correction factor is not always justified. The cause might be concentration-dependent adsorption, evaporation, saturation or incompatible sample preparation.

Instrument carryover is another distinct source of bias. A strongly enriched sample may leave residual material in tubing or a trap that appears in the next low-concentration sample. Alternating high and low standards can reveal this problem more clearly than repeatedly measuring samples from the same range. A blank immediately after a large signal may be essential. An analyst should also monitor retention-time stability, detector response, pressure and trap performance, rather than assuming chromatography eliminates all uncertainty.

Method blank, field blank, reagent blank, instrument blank and spike check therefore answer different questions. Their names must not be translated as though they are synonyms. Ask which steps each control experienced, when it was run and how its value was used. A concentration produced without a traceable quality-control chain might be useful for preliminary exploration, but it should not be presented as having the same evidential status as a fully validated measurement.

Why temperature, light and sample delay change interpretation

Temperature affects gas–liquid partitioning, biological activity and sometimes analytical extraction efficiency. Light can influence photochemical pathways and living organisms. Storage time permits chemical and biological changes unless a validated preservation strategy controls them. Therefore a laboratory record should distinguish collection temperature from equilibration temperature, storage temperature and instrument temperature. A single field called ‘temperature’ can hide several physically different variables.

Suppose two sample bottles were collected simultaneously, but one was processed immediately and the other sat in sunlight for an hour. Even if both were eventually measured at the same laboratory temperature, they no longer represent identical handling histories. Their difference cannot be confidently assigned to a difference in the original ocean water. The experiment would need dedicated time-delay or handling controls before converting that observed contrast into a defensible production or loss narrative.

Rapid processing is often helpful for a volatile, reactive analyte, but ‘faster’ is not a substitute for a validated method. Some laboratory protocols deliberately impose controlled equilibration or chemical treatment. The correct measure is whether timing and conditions match the procedure tested for the target analyte. In a published dataset, note whether timing is documented well enough to support comparisons among stations, day and night samples or historical campaigns.

When a detection limit is the scientifically important result

At very low concentrations, the question may be whether a peak can be distinguished from noise and process background. Detection and quantification thresholds should come from the validated method, not an arbitrary number chosen after seeing an inconvenient sample. A reading below a quantification limit can still convey information, but it should not be treated as an exact zero. Analysts may preserve the numerical instrument signal internally while reporting a censored value under a documented rule.

For example, if the method’s validated quantification limit is 0.5 nM and a processed sample yields a nominal 0.3 nM, writing 0.300 nM suggests unsupported precision. It is safer to state that it lies below 0.5 nM under the method’s reporting convention, along with whether the analyte was detected above the detection threshold. This distinction becomes critical when computing averages or comparing periods containing many low signals.

Replacing every nondetect with zero, half the detection limit or the full limit is not universally correct. Each convention can bias the summary differently, particularly when limits vary between samples. A scientifically responsible analysis records how censored data were handled and explores sensitivity to plausible alternatives. The same logic applies when a data pipeline automatically rounds small values into apparent zeros, because that conversion may create a false impression of an absent sulfur pool.

Uncertainty is a chain, not merely the last decimal place

Consider three independent relative uncertainty components in a hypothetical concentration estimate: five percent from calibration, three percent from repeat measurement and four percent from extraction recovery. A simple root-sum-of-squares calculation gives the square root of 25 + 9 + 16 percent-squared, about 7.1 percent combined. At eight nM, that corresponds to approximately 0.57 nM. This demonstration assumes the components are independent, random and properly characterised; systematic handling losses are not erased by that calculation.

The distinction between precision and accuracy is central. A cluster of replicate values can have a tiny standard deviation but still be consistently wrong because the calibration standard was degraded. Conversely, a well-calibrated measurement may vary across samples because water masses truly differ. When reporting an uncertainty interval, state whether it covers analytical repeatability, full method uncertainty, sampling variability or model uncertainty. These are not interchangeable.

Correlated errors deserve special attention. If all samples use the same incorrect gas standard, each may share a common calibration bias. Averaging hundreds of samples will not make that bias disappear. If two analyses share a systematic error, subtracting them may reduce or amplify the impact depending on the relationship. The reader should identify which uncertainties are independent, shared or unquantified before using a DMS difference as evidence of environmental change.

DMSP pathway interpretation: formation is only one possible outcome

DMSP may undergo cleavage pathways that generate DMS, but organisms also use other DMSP transformation routes. The DMS formed may then be consumed, oxidised or ventilated. In an open water mass, mixing adds another term. A concentration budget can be written conceptually as rate of change equals sources minus sinks plus transport. It is a framework for choosing measurements, not evidence that all its terms have been determined from a single DMS bottle.

Suppose total DMSP doubles while measured DMS stays constant. One tempting conclusion is that DMS production failed. Another is that DMS turnover accelerated enough to balance additional production. A third is that the DMSP pool changed in organisms that did not contribute much to DMS at that time. The observation alone cannot decide among these stories. A well-designed process study might pair concentration measurements with isotopic tracers, incubations, molecular evidence and transport estimates.

A frequent translation error is to turn a pathway arrow in a schematic into a certainty statement. ‘Can produce’ is not the same as ‘produces all of’; ‘potential production’ is not measured production; and ‘associated with’ does not establish causality. This matters particularly in science writing intended for non-specialists. Correct wording does not weaken the finding. It tells readers which links are supported by data and which remain hypotheses.

Building a clean, reusable data record

A data table becomes interpretable when every variable has a definition, unit and provenance. Useful fields include station, position, UTC timestamp, depth, sample identification, water temperature, salinity, DMS concentration, fraction definition, analytical method, instrument, blank treatment, calibration record and quality flag. For DMSP, record whether the result is dissolved, particulate or total. A raw peak area should not be stored in the same numerical column as a processed concentration without a clear distinction.

Keep raw analytical responses separate from corrected values, and document the operations that connect them. A workflow may need columns for observed equivalent concentration, procedural blank, dilution or recovery treatment, final reported value and detection status. This permits later audit without requiring readers to reverse-engineer a spreadsheet formula. Units belong in metadata or column names consistently. Empty values should be distinguishable from measured zeros and values below detection.

Metadata also protects cross-language meaning. A translator should retain acronyms with definitions, use unambiguous decimal punctuation, preserve superscript minus signs and avoid changing nmol L−1 into nmol kg−1. If an instrument parameter is uncertain, it is better to annotate the uncertainty than to silently choose a familiar term. Effective translation of a scientific dataset preserves the operational chain through which observations became numbers.

Transfer problem one: choose the defensible interpretation

An imaginary dataset reports seawater DMS of six nM in a freshly handled sample and nine nM in a bottle held for thirty minutes. Which statement follows: production increased in the sea, DMS concentration increased during handling, or DMSP cleavage definitely occurred? Only the middle statement directly follows, assuming both analyses are valid and the aliquots began comparable. A handling effect is visible, but its chemical cause has not been isolated. To identify it, add controlled timing and treatment comparisons.

Transfer problem two: total DMSP is 42 nM and dissolved DMSP is 15 nM, both measured with comparable validated procedures. The difference is 27 nM of operational particulate-associated DMSP. It is not 27 nM measured DMS. If the ‘dissolved’ preparation broke cells, even this difference might be biased. State the arithmetic, the fraction definitions and the relevant uncertainty before interpreting the biological inventory.

Transfer problem three: a surface sample is 15 nM DMS, the equilibrium concentration at the specified conditions is three nM and a justified gas transfer coefficient is 1.5 m d−1. The gradient is twelve µmol m−3 and the illustrative upward flux is eighteen µmol m−2 d−1. Ask which assumption dominates the result: the concentration contrast or the chosen gas exchange coefficient? Without a quantified uncertainty, we cannot select a universal answer; both can matter, and the transfer model may be the largest unknown.

Transfer problem two: audit a claimed comparison between cruises

Two studies give DMS surface means of four and seven nM. The second cruise used a different intake depth and a new method with a higher quantification limit. It is tempting to publish a seventy-five-percent rise, computed from (7 − 4)/4, as though it were a long-term environmental trend. The numerical change is correct for the reported means, but the causal inference is not established. Before comparing, examine location, season, depth, data selection, method change and uncertainty in both estimates.

Suppose the first study reported only samples above a 1 nM threshold and the second reported all values down to 0.2 nM. The dataset composition and censoring behaviour could materially affect their summary statistics. More importantly, the two sampling frames might not represent the same ocean population. A fair comparison could stratify by season and water mass, align reporting conventions or use overlap samples and interlaboratory tests. Sometimes the right answer is that the studies are not directly comparable.

The repair is to write a comparison as a chain of propositions. First: which measured quantities were included? Second: what mathematical difference was calculated? Third: can the analytical conditions support its interpretation? Fourth: which environmental variables were controlled? This chain separates a numerical difference from an explanatory claim, preserving both useful information and intellectual honesty.

Frequently asked questions: what the numbers can and cannot say

Is a high DMS concentration proof of more phytoplankton? No. Phytoplankton community structure, DMSP physiology, bacterial production and consumption, mixing and ventilation can each affect DMS. DMS may be associated with biological activity without functioning as a universal biomass proxy. A separate chlorophyll or cell-count measurement answers a different question and can help interpret the sulfur result when the measurements are temporally and spatially aligned.

Can a single DMS sample tell us whether the ocean is releasing more sulfur to the air? Not by itself. Flux needs an atmospheric equilibrium term and a validated exchange estimate under the relevant conditions. Can a larger DMSP concentration be called a larger DMS production rate? No. Pool size lacks a time denominator and does not specify transformation efficiency. Does a low chromatographic peak always mean low seawater DMS? No; low response may reflect poor recovery or sample loss unless calibration and controls exclude those possibilities.

Why do published DMS concentrations from different methods disagree? They may sample different fractions, use different handling protocols, calibration ranges and reporting conventions, or genuinely encounter different water. Is a global database a substitute for quality control? No; it helps organise evidence and identify patterns, but users must read the contributing methods and metadata. The strongest conclusion is the most specific one that the measurement chain can actually support.

A practical reading checklist before accepting an interpretation

First identify the substance: DMS, DMSP, DMSO or another sulfur species. Then identify the operational fraction and unit. Follow the sample from collection through transport, preparation, extraction, chromatography and calibration. Check whether blanks, drift standards, spike recoveries and censoring rules were recorded. Only then ask what comparison or equation the author performs. If a step is missing, the reader should mark the inference as conditional rather than automatically rejecting all the data.

Second ask what kind of quantity is being discussed. A bottle value is concentration. A series over time can describe net change but needs controls for mechanistic attribution. An atmospheric emission estimate is a flux that relies on transfer physics. A DMSP-to-DMS statement refers to a biological or chemical transformation that needs separate experimental support. Confusing these categories is one of the fastest ways to make a specialist paper appear clearer while actually changing its meaning.

Finally check the conclusion’s reach. A local sample is not a global climate response; a peak is not an ecosystem mechanism; a confident decimal is not a complete uncertainty statement. The reader’s task is to preserve what was measured, what was calculated and what was inferred. When those layers are kept separate, sophisticated sulfur chemistry becomes understandable without pretending that every ocean process has already been resolved.

Five more applied audits: comparative uncertainty, metadata, sampling and calibration

The first extended problem concerns a sample measured twice using two instruments. One instrument gives 10.4 nM with a stated combined standard uncertainty of 0.8 nM; the second gives 9.6 nM with 0.5 nM, and their errors are assumed independent for the exercise. The difference is 0.8 nM. A simple uncertainty for that difference is the square root of 0.8 squared plus 0.5 squared, approximately 0.94 nM. A difference smaller than one combined standard uncertainty does not provide strong evidence of a systematic disagreement. Yet the calculation is only illustrative: if both instruments used the same biased standard, they might agree precisely and still both be wrong. That is why laboratory intercomparison requires independent standards and attention to shared sources of error.

The second problem is about missing metadata. A table lists DMS as 12 nM but has no collection-to-analysis delay, blank or operational fraction. A competent reader can retain the number as reported while assigning it a limited evidential role. They should not add invented collection conditions, nor should they quietly interpret the value as a biologically stable dissolved pool. An honest annotation might state that the reported concentration is unverified with respect to sample-handling effects. This is not an accusation against the researchers; it is recognition that reproducibility needs details. If a later methods appendix supplies the missing conditions, the evaluation can be revisited. The distinction matters for secondary data reuse, where uncertainty is sometimes introduced by the reuse process rather than by the original measurement.

The third problem asks whether an ocean average can be reconstructed from unequal sampling coverage. Suppose a cruise measured twenty nearshore samples and five offshore samples. The raw mean of all twenty-five values describes that collected set. It is not necessarily an unbiased estimate of the ocean area’s mean unless the sampling design and weighting justify it. A coastal station visited repeatedly may dominate an arithmetic mean, while a vast offshore region is represented once. A defensible regional summary might stratify spatially, weight by area, or explicitly report the sampling distribution. No weighting formula can compensate for entirely absent seasons or water masses without modelling assumptions. Readers should distinguish what the sample mean establishes from the population claim being made.

The fourth problem is a translation audit. A draft caption says ‘DMSP oxidation released DMS at 3 nM.’ This combines a proposed biochemical mechanism, a direction of reaction and a concentration rather than a rate. An editor should ask whether DMSP cleavage was actually demonstrated, whether the experiment measured DMS concentration or a formation rate, and whether ‘oxidation’ is the right pathway label. A safer caption may report the measured concentration directly while describing the biochemical hypothesis separately. If a rate was measured, write its time denominator explicitly. This discipline is especially useful in multidisciplinary teams where chemical, biological and language specialists may use the same short acronym but mean different parts of the system.

The fifth problem tests decision-making under a calibration failure. A batch contains twenty samples, and the continuing calibration check after sample ten was outside the laboratory’s acceptance limits. It would be unsafe to silently average all twenty as though instrument response remained validated throughout. The analyst should examine earlier checks, possible drift timing, retained standards and whether reanalysis is possible. Samples before the failed check may also be affected, depending on when the instrument changed. A documented data qualification or reprocessing decision is preferable to a convenient blanket adjustment. Quality-control failure is information about the evidence, and the reader needs that information to judge whether small environmental differences are real.