How do you read a groundwater δ18O–δ2H isotope plot without turning one attractive diagonal line into a claim the data cannot support? Begin with the measured oxygen-18 and deuterium ratios, the reference scale used by the laboratory, the contemporary precipitation baseline and the physical question: whether the sample resembles local recharge, an evaporated source, or a mixture. A groundwater isotope plot is a comparison of water histories, not a direct map of underground flow.
On a graph of δ2H against δ18O, the familiar global meteoric water line is approximately δ2H = 8 × δ18O + 10, with both delta values in per mille relative to a specified water standard. The line summarizes a global precipitation relationship; it is not an acceptance boundary for every aquifer. Local meteoric lines, seasonal rainfall, evaporation, mixing and sampling context can shift the pattern. A low deuterium-excess value may support an evaporation hypothesis, but does not prove it alone.
This guide translates stable isotope results into a defensible interpretation, with independently worked calculations, field and laboratory checks, competing explanations, transfer questions and answers. Every number labelled illustrative is invented for teaching rather than presented as an observation. The method also shows why precise scientific vocabulary matters: source, mixture, evaporation, recharge and age are related concepts, but they are not interchangeable findings.
Start by defining the actual decision
The first decision is not which coloured trend line fits best. Write down the question an investigation can answer: Did the groundwater probably originate from local precipitation? Was an open-water body affected by evaporation before recharge? Do two sampled sources contribute to a mixed well? Has a water body changed between sampling campaigns? Each question needs different reference samples and independent field observations. A plot without a declared question makes almost any story seem reasonable.
Compare the target well with the same catchment’s precipitation, springs, rivers, reservoirs and neighbouring wells only when those are plausible endmembers. State sampling dates, screen intervals and hydrogeological boundaries. A deep confined well and a shallow seasonal well may respond to different recharge periods. If the monitoring design has no precipitation data, you can report an isotopic resemblance to reference waters, not claim a uniquely established recharge elevation or season.
What delta notation actually measures
The laboratory compares the heavy-to-light isotope ratio in a sample with a reference ratio: δ = 1000 × (Rsample/Rreference − 1). For oxygen, R usually means 18O/16O; for hydrogen, 2H/1H. The factor 1000 expresses a relative departure in per mille, written ‰. Thus δ18O = −8‰ means the sample’s isotope ratio is approximately 0.992 times the reference ratio. It does not mean eight percent oxygen-18 has vanished, and it is not a negative physical concentration.
Reference standards matter. Water isotope results are commonly reported relative to the VSMOW–SLAP scale, with laboratories using appropriate calibration standards and normalization. Mixing raw δ numbers from inconsistent calibrations risks producing false geographic signals. Read the laboratory certificate for scale, analytical uncertainty, sample identifier, date, matrix and any notes about evaporation or water contamination. An apparently minor reference-scale mistake can dwarf the spatial contrasts you are trying to explain.
Separate isotope ratios from isotope abundances and from tritium activity. δ18O and δ2H describe the relative stable-isotope composition of water; neither is itself a clock. Tritium radioactively decays and can sometimes inform age categories; stable-isotope resemblance can indicate processes and sources but cannot measure years since recharge by a universal conversion. Resist using the word ‘older’ merely because one groundwater sample is more negative than another.
First-principles fractionation during the water cycle
Water containing lighter isotopes generally participates somewhat more readily in phase changes than water containing heavier isotopes; equilibrium and kinetic isotope effects differ with conditions. Condensation, progressive rainout, evaporation and mixing therefore alter the distribution of δ18O and δ2H. Temperature, humidity, trajectories of moist air, rainfall amount, elevation and season help explain why precipitation is not isotopically identical everywhere. A geographic comparison is a test of competing water-cycle pathways rather than a simple latitude lookup.
Evaporation from a lake is especially instructive. Vapour leaving the liquid often differs from the water remaining, enriching the residual water in heavy isotopes under many circumstances. Kinetic fractionation under low relative humidity changes the relationship between hydrogen and oxygen signatures. The point on a two-isotope graph can move towards less-negative δ values along an evaporation trajectory whose slope is often lower than a meteoric-line slope. That observation is powerful only when plausible initial water and environmental conditions are known.
Groundwater is not automatically evaporated just because it is isotopically heavy. Warmer-season precipitation, coastal precipitation, low-altitude recharge, imported water, lake infiltration and mixing can all create relatively enriched signatures. Water–rock exchange at elevated temperatures may shift oxygen without producing the same hydrogen change. Check water chemistry, geology, source availability and sampling dates before attaching a causal label to a relative ratio.
Read the axes, sign and units before fitting a line
Many figures place δ18O on the horizontal axis and δ2H on the vertical axis. More negative values lie farther left or down. Some reports reverse the axes or put δD in place of δ2H; δD is a common name for deuterium delta, not a separate isotope. Look carefully at axis labels and whether both report ‰ relative to the same standard. A slope reported after reversing the axes is its reciprocal and should not be compared directly with a conventional δ2H-versus-δ18O slope.
If two hypothetical groundwater samples have δ18O values of −9.0‰ and −5.0‰, the latter is relatively enriched in oxygen-18. That says nothing by itself about whether its recharge was more recent, deeper or cleaner. A reported analytical uncertainty of ±0.1‰ for oxygen may make a 0.05‰ difference uninformative, whereas a 4.0‰ difference warrants geological and water-source investigation. Compare measurement uncertainties and natural temporal variability separately.
Global versus local meteoric water lines
The simple reference δ2H = 8δ18O + 10 is the conventional global meteoric water line associated with Craig’s early work on precipitation. It is a broad empirical orientation tool, not a global law applicable sample by sample. A local meteoric water line should be estimated from relevant precipitation observations using a defensible regression method and time window. Local slope and intercept can differ because precipitation formation conditions differ. Some local precipitation datasets show slopes near eight; others deviate appreciably.
The IAEA and WMO Global Network of Isotopes in Precipitation supplies context for rainfall isotope comparisons, but a distant station may not represent a mountain catchment or a particular rain season. Consider precipitation-amount weighting before describing the isotope composition of water actually available for recharge. A straight arithmetic average over storm samples gives a small storm the same influence as a large storm, often answering the wrong hydrological question.
Do not ‘correct’ a groundwater observation onto the global line merely to make it look meteoric. The distance from any reference line is itself a potential diagnostic, conditional on calibration and local baseline. Display the reference as a reference, then plot measured well values with uncertainty where appropriate. Preserve the original data and separately report any calculation. A beautiful graph cannot repair missing local precipitation samples.
Worked case: a sample that lies on the global line
Take an illustrative sample A with δ18O = −8.0‰ and δ2H = −54‰. The global reference predicts 8 × (−8.0) + 10 = −54‰. Therefore the sample falls on that line by this simple check. Its deuterium excess, d = δ2H − 8δ18O, equals −54 − (−64) = +10‰. Both calculations agree because d is the intercept of a slope-eight line passing through the sample.
The correct conclusion is modest: A is consistent with the global precipitation relationship. This single pair of values does not reveal a recharge year, elevation, aquifer residence time or guarantee that no evaporation occurred. A mixture of sources can also lie on the line. The independent check is to compare A with the local precipitation dataset and nearby plausible source waters sampled at a relevant period, then inspect the well’s screen depth and hydrochemistry.
Worked case: an enriched sample below the line
Illustrative sample B has δ18O = −5.0‰ and δ2H = −40‰. At −5.0‰ oxygen, the global line predicts δ2H = −30‰. B is therefore 10‰ below that reference in the hydrogen direction. Its deuterium excess is −40 − 8(−5) = 0‰. This is a numerical departure, not automatically a positive diagnosis of evaporation. The diagnostic wording is ‘relative enrichment with low d-excess, consistent with some evaporative pathways under appropriate conditions.’
To improve the claim, investigate whether a nearby lake, irrigation return flow or shallow unsaturated zone could impart an evaporative signature, and whether the local meteoric baseline is actually close to the global line. A locally lower precipitation intercept might explain part of the difference. Elevated chloride or electrical conductivity can support—but not uniquely prove—concentration by evaporation. If competing sources have different isotopic compositions, a simple mixing hypothesis also needs testing.
Deuterium excess is a derived contrast, not a magic switch
Define d-excess explicitly as d = δ2H − 8δ18O when using the conventional definition. The coefficient eight is fixed by that definition, not newly estimated for each set of wells. It highlights deviations from a slope-eight reference. Local d-excess can be influenced by moisture-source humidity, evaporation conditions, precipitation processes, land-surface evaporation and mixing. A small d value is evidence that demands a mechanism; it is not a universally calibrated percentage of evaporated water.
Compare three illustrative samples with δ18O and δ2H pairs: A (−8, −54), B (−5, −40), and C (−10, −70), all in per mille. Their d-excess values are +10, 0 and +10‰ respectively. A and C share the same d even though their absolute isotope signatures differ. B has a lower d despite being more enriched. A separate regression through A and C has slope eight, but adding B changes any fitted line. This is why one point and one slope answer different questions.
The term ‘excess’ does not mean an unsafe concentration or surplus mass of deuterium. It is a diagnostic offset in a coordinate system. Explain the result to readers as ‘the hydrogen isotope value is this many per mille above or below a standard slope-eight relationship for its oxygen value.’ This translation is more informative than announcing a low d-excess and assuming the reader knows what happened.
A local evaporation trajectory is not always straight
In a narrowly defined set of conditions, plotted waters exposed to evaporation may approximate a lower-slope line compared with the meteoric baseline. The slope depends on relative humidity, isotope composition of surrounding vapour, temperature and other conditions. A two-endmember mixing line can also look approximately straight. A fitted slope alone cannot discriminate evaporation from mixing unless you know where the potential initial waters lie and how the measurement sequence relates to hydrology.
Consider a reservoir sampled monthly and nearby shallow wells sampled at fixed intervals. If a sequence of reservoir observations traces a coherent enrichment path during dry periods and nearby wells follow later with similar changes, that temporal evidence is stronger than an isolated diagonal scatterplot. But a lag could reflect pumping, mixing or travel time. Measure lake stage, rainfall and conductivity alongside isotopes, and acknowledge that the isotope dataset alone rarely identifies a unique lag mechanism.
Mixing: calculate ratios before telling a mixing story
For water from two sources with isotope compositions A and B, a simple water-mass approximation uses δmix ≈ (1−f)δA + fδB, where f is the fraction from B. For water isotopes this linear approximation is generally excellent for many practical mixing cases when the sources and scale are comparable; exact calculations can use isotope ratios and isotope atom budgets. You cannot apply the same fraction indiscriminately to dissolved solutes where concentrations, reactions or precipitation alter the mass balance.
Suppose a hypothetical recharge source A is (δ18O, δ2H) = (−10‰, −70‰) and source B is (−4‰, −30‰). If the water fraction from B is f = 0.25, δ18Omix = 0.75(−10) + 0.25(−4) = −8.5‰. Similarly δ2Hmix = 0.75(−70) + 0.25(−30) = −60‰. The resulting d-excess is −60 − 8(−8.5) = +8‰. Both isotope coordinates support the declared fraction in this idealized example.
That computation does not mean every sample plotting between A and B is a genuine two-source mixture. Other sources could occupy the same space; endmembers may vary seasonally; selective evaporation may alter one before mixing; and sampling a long well screen may collect a blend of aquifer layers. Test the inferred fraction against independent conservative tracers and the water balance. If chloride disagrees, explain the disagreement before averaging the fractions.
Solve for a putative mixing fraction and test it
If an endmember A is δ18O = −10‰, B is −4‰ and a putative mixture is −8.5‰, the oxygen-derived B fraction is (−8.5 − (−10))/(−4 − (−10)) = 1.5/6 = 0.25. If the hydrogen values are A = −70‰, B = −30‰ and mixture = −60‰, the hydrogen-derived fraction is (−60+70)/(−30+70) = 10/40 = 0.25. Agreement is a useful internal consistency check, not proof that the sources were uniquely identified.
Now change only the observed hydrogen value to −52‰. It implies 18/40 = 0.45 from hydrogen while oxygen still implies 0.25. That inconsistency warns of a third source, variable endmembers, fractionation, data problems or the wrong conceptual model. Reporting an average of 35% would conceal a failed assumption. Prefer the honest conclusion that the candidate two-endmember model does not jointly explain the measurements.
Measurement uncertainty must travel through the calculation
Two-decimal-place isotope numbers can create false confidence. The laboratory uncertainty for δ18O and δ2H need not be identical, and temporal variation of an endmember can exceed instrument error. When calculating d = δ2H − 8δ18O, uncertainty in oxygen is multiplied by eight. Assuming independent random analytical errors for a simple illustration, the standard uncertainty is the square root of uH² + (8uO)². This is an uncertainty propagation statement, not a reason to ignore systematic calibration effects.
For illustrative uncertainties uH = 0.8‰ and uO = 0.1‰, the independent-error d uncertainty equals sqrt(0.8² + 0.8²) ≈ 1.13‰. A 0.3‰ change in d between two samples should not be narrated as a robust process change at that precision. Repeat analyses, standards and temporal replicates may be needed. Correlated calibration errors require covariance treatment rather than the simple independent formula. Document which uncertainty figure you are using.
Precipitation-weighted baselines and recharge weighting
To characterize a series of rainfall events, a precipitation-amount-weighted isotope mean is calculated as the sum of event amount times event δ divided by the total precipitation amount, on a common reference scale. This estimate is an input-weighted rain composition; it still may not equal recharge-weighted composition because only part of each rain event becomes groundwater. High-intensity storms, antecedent soil moisture, vegetation interception and evapotranspiration can change the fraction of rainfall infiltrating deeply.
Take two fictional storms: 10 mm of rain at δ18O = −2‰ and 90 mm at δ18O = −10‰. The unweighted mean is −6‰, but the rain-volume-weighted mean is (10×−2 + 90×−10)/(10+90) = −9.2‰. The difference is 3.2‰, far larger than many laboratory uncertainties. Before comparing groundwater with precipitation, identify whether the rainfall reference was event-averaged, volume-weighted, monthly averaged or recharge-weighted. These are not equivalent quantities.
Well construction can produce a misleading mixed sample
A long screened interval can connect or sample zones with distinct source histories. Pumps can preferentially draw from productive layers, and purge volumes, flow rates and stabilization criteria change the water entering a bottle. A sample that resembles an intermediate isotope composition may reflect mixing inside the well rather than broad mixing throughout the aquifer. Record screen top and bottom, static level, purge and pumping conditions, conductivity, temperature and sampling method.
When possible, compare discrete-depth samples, downhole flow information and hydraulic heads before asserting a plume or mixing boundary. The previous borehole flowmeter guide explains how transmissive intervals can alter which zone dominates a well sample. This contextual connection matters because an accurate isotope instrument can faithfully measure a water mixture produced by the sampling apparatus rather than by the regional aquifer.
Sampling field discipline before sending bottles
Use clean containers suitable for the laboratory’s isotope method, avoid unnecessary headspace and evaporative losses, close caps promptly, label bottles consistently and store them according to the laboratory’s instructions. Stable water-isotope bottles do not necessarily have the same collection protocol as dissolved noble-gas copper tubes or volatile organic samples. Confusing these protocols can compromise both analyses. Record site, date, time, operator, water source, screen interval and relevant weather.
Collect field duplicates and appropriate quality-control samples at a frequency justified by the investigation. Do not use a sample that was sitting open in warm weather to establish an evaporation signal for the underlying aquifer. An uncontrolled bottle can reproduce the very physical process you are attempting to infer in the field. Ask whether anomalous values correlate with handling batches, evaporation-prone shallow containers or laboratory standard changes.
When the seasonal story changes the conclusion
An annual isotope distribution hides dynamics. Some aquifers receive most recharge during a specific rainy season; others receive rare episodic pulses, snowmelt or river infiltration. A groundwater sample near the amount-weighted annual precipitation mean does not prove year-round recharge. A groundwater trend that changes after a wet season might reflect transport, mixing, changes in pumping or the opening of a preferential path. Water-level and chemistry time series can narrow the alternatives.
A strong design pairs the seasonal isotope record with rainfall amounts, local river or lake compositions and well hydraulic observations. Plot the samples with dates instead of hiding all temporal structure in one regression. A simple early-wet, late-wet and dry-period colour grouping can reveal whether a supposed universal aquifer line is actually a rotating mixture of source waters. Explain the temporal resolution required before asserting a response time.
Recharge elevation is not a one-step conversion
Precipitation often becomes isotopically depleted with altitude under particular regional conditions, but an altitude gradient is not universal or temporally invariant. To estimate recharge elevation, build or cite a local calibration relating rainfall or spring isotope composition to sampled elevation, and evaluate the plausible contributing altitude range. Recharge can occur upstream in a catchment at a very different elevation from the sampling well, and seasonal rainfall may confound the apparent gradient.
Suppose a locally validated hypothetical relationship changes δ18O by −0.2‰ for each 100 m of elevation. A difference of −1.0‰ might superficially suggest 500 m. But that inference assumes the same air-mass history, recharge season, water source, and endmember purity. If the calibration uncertainty or hydrological mixing is large, the nominal 500 m is not defensible as a precise origin point. Report a conditional elevation estimate with a supported range rather than a GPS-like location.
Salinity, heat and chemistry are independent checks
Electrical conductivity and major ions provide different information from water isotopes. Concentrating a water body by evaporation often increases the concentration of many dissolved solutes, but mineral dissolution, road salt, seawater intrusion and industrial inputs can change salinity without producing an evaporative isotope signature. Similarly, blending saline and fresh endmembers can produce isotopic and chemical patterns that look different from evaporation. A multi-tracer explanation should account for both.
Deep thermal water may have oxygen isotope shifts influenced by water–rock interaction, while its deuterium composition changes less strongly under some conditions. A simple low-slope water line could then have more than one cause. Include temperature, geology, chloride/bromide information, known anthropogenic inputs and mineral saturation context before claiming climatic evaporation. Beware the circular explanation in which chloride is treated as proof of evaporation and the isotope plot is interpreted solely because chloride was high.
How to describe the strongest defensible conclusion
Use three levels of language. Observation: ‘The well measured δ18O of −5‰ and δ2H of −40‰.’ Calculation: ‘Its conventional d-excess is 0‰ and it plots 10‰ below the global meteoric reference at that oxygen value.’ Interpretation: ‘Evaporation before recharge is one candidate explanation; local rain and alternative source mixing must be tested.’ Keeping the levels separate makes it possible for a colleague to disagree with the mechanism without disputing the measurement.
An unqualified statement such as ‘this is fossil groundwater evaporated twenty years ago’ combines unsupported age and process claims. Stable isotopes do not date water in calendar years by themselves. For age questions consult genuinely age-sensitive tracers and geological context. For hydraulic flux questions consult well tests or independent flow measurements. A strong article translates a specialist result into what can be said, what remains unknown and which additional observation would reduce uncertainty.
Create a reproducible plotting worksheet
Store each row with sample ID, date, source type, δ18O, δ2H, units, reference scale, analytical uncertainties and sampling metadata. Calculate the global-line predicted δ2H as eight times δ18O plus ten and d-excess as measured δ2H minus eight times δ18O. Include a column distinguishing measured data, derived values and inferred source labels. Export the plot with axis labels, reference line and date coding. Save the local-precipitation regression method separately.
Always record whether water sources were sampled on the same day. If not, state that the comparison is provisional. Plot field duplicates as separate points before combining them. An apparent linear trend across samples can be caused by a batch calibration shift or by assigning a single static river endmember to a seasonally varying river. Reproducibility includes being able to reconstruct every plotted point from the original laboratory record, not just sharing a polished picture.
An independent test of source mixing
Consider a new fictional field set: upstream spring S is (−11‰, −78‰); a reservoir R is (−5‰, −38‰); a well W is (−8‰, −58‰). Compute the reservoir fraction from oxygen: (−8+11)/(−5+11) = 3/6 = 0.50. From hydrogen: (−58+78)/(−38+78) = 20/40 = 0.50. The two calculated halves agree in this idealized example. That agreement establishes internal algebraic consistency, not a physical proof that no other sources matter.
Now add a hypothetical second well V = (−8‰, −52‰). Oxygen still suggests 0.50 reservoir water; hydrogen suggests (−52+78)/40 = 0.65. The candidate endmembers fail to explain both coordinates with one fraction. Ask whether endmembers shifted over time, whether a third source enters the well, whether evaporation occurred after mixing, or whether a sample was compromised. Do not invent a third-source proportion from two measurements without additional constraints.
Transfer problem one: on-line or below-line
Question: A hypothetical well reports δ18O = −7‰ and δ2H = −48‰. Calculate the global-line prediction and conventional d-excess. Determine whether it lies on, above or below that reference and state one thing the plot cannot tell you. Work with per mille throughout, not percentages. Pause before reading the answer and check both the sign of the product eight times minus seven and the difference from the measured hydrogen value.
Answer: The global-line prediction is 8×(−7)+10 = −46‰. The measured hydrogen value is −48‰, so the point is 2‰ below that reference. Its d-excess is −48−(−56) = +8‰. The plot alone cannot establish recharge age or a unique source. Because local rainfall may not sit on the global line, interpreting a 2‰ offset requires an appropriate local reference and uncertainty analysis.
Transfer problem two: fraction and contradictions
Question: Two reference waters have δ18O of −12‰ and −4‰. A supposed mixture is −10‰. What is the fraction of the second source? If hydrogen isotope endmembers are −86‰ and −30‰, what hydrogen value does that fraction predict? Finally, what would you do if the measured hydrogen value were −64‰ instead? Write the water fraction, not the heavy-isotope fraction, and use the same source order in both calculations.
Answer: The oxygen result is (−10+12)/(−4+12) = 2/8 = 0.25, or 25% of the second water. Predicted hydrogen is 0.75(−86)+0.25(−30)=−72‰. A measured value of −64‰ is eight per mille more positive than that prediction. Re-check sample IDs, uncertainty, endmember variability and the assumption of two-source conservative mixing. A line drawn through convenient points is not a substitute for explaining the inconsistent hydrogen budget.
Transfer problem three: weighting rainfall
Question: A 15 mm storm has δ2H = −20‰, while a 45 mm storm has δ2H = −80‰. Compute both the unweighted mean of storm values and the rain-volume-weighted mean. Explain which one is suitable for describing total sampled precipitation volume and whether either is necessarily the isotope value of the groundwater recharge. Include the denominator of the weighted calculation.
Answer: The unweighted mean is (−20−80)/2 = −50‰. The weighted mean is [15(−20)+45(−80)]/(15+45)=−3900/60 = −65‰. Volume weighting better describes the rain mixture across these two sampled events, assuming representative event isotope values. Neither automatically describes recharge because infiltration and evapotranspiration can weight the events differently. Additional recharge evidence is required.
What to do when the scatterplot has no clear line
With only three or four wells, a fitted line may be unstable. Plot individual uncertainty bars, identify field context and avoid sophisticated regression that implies a large independent sample. Wells may be spatially clustered or screened across connected depths; treating them as independent environmental realizations misstates the information content. A global-line comparison with explicit limitations may be more transparent than an apparently precise slope estimate.
If the plot contains multiple clouds, test spatial groups, hydrostratigraphic units and dates before declaring all outliers bad data. An outlying sample can reflect a real source, a laboratory issue, or sampling artefact. Preserve it while investigating. Rejecting observations merely because they spoil a desired evaporation slope is both poor measurement practice and a threat to the scientific credibility of the conclusion.
When this method is the wrong tool
Stable water-isotope results cannot directly supply hydraulic conductivity, transmissivity, well yield, recharge rate in millimetres per year or the exact age of a sampled parcel. Such inferences need independent hydraulic tests, water and solute budgets or radioactive and transient tracer models. They also cannot, on their own, prove drinking-water safety, microbial absence or compliance with water-quality standards. A normal-looking isotope point can coexist with substantial contamination.
Choose the method because it tests a source or process hypothesis, not because isotope analysis is technologically impressive. If the question is ‘how fast does water move?’, prioritize hydraulic gradients, transport constraints and residence-time evidence. If the question is ‘which of two waters is present?’, stable isotopes may be ideal, provided the endmembers are distinct and their variation is known. Match the instrument output to the inference.
Frequently asked questions: negative values and water age
Why are both numbers often negative? Because natural water can have a lower heavy-to-light isotope ratio than the calibration reference; negative delta is a relative scale value. Does a more negative isotope value mean colder recharge? It may be consistent with certain climate or elevation effects but is not definitive without a local calibration. Can I calculate an age from δ18O? Not as a universal conversion. Stable oxygen and hydrogen isotopes chiefly help trace sources and fractionation processes.
Is there one universally correct meteoric line? No. The conventional global equation is a useful comparator, while local precipitation conditions shape local relationships. Is d-excess an evaporation percentage? No. Can two-source mixing be checked from the two isotope coordinates? Yes, internal consistency is valuable, but the existence and composition of the endmembers must be independently justified.
Frequently asked questions: plotting and laboratory practice
Should I use δD or δ2H on the vertical axis? Either label is commonly understood if it refers to deuterium delta and its unit and reference are stated. Should I multiply δ18O by eight if I reverse the axes? The conventional global expression presumes δ2H vertically versus δ18O horizontally; a reversed plot requires appropriate algebra. What if lab duplicates differ more than expected? Investigate standards, labeling and sample handling before interpreting small offsets.
What should the results table include? At minimum sample ID, timestamp, source, location or well interval, isotope pair, uncertainty and reference scale; add relevant conductivity, water level, precipitation and source characterization if available. Why are extra water-chemistry measurements useful? Because they can disambiguate processes that create similar isotope plots. Can a spring isotope value represent all precipitation? Usually not without evidence of representative recharge and mixing.
A careful route into related ideas
Readers of The Gold Standard of Isotope Hydrology can connect this focused plotting method with a broader water-cycle framework. How Science Works: Hydrogeology explains the hydraulic system that determines which water a well samples. The groundwater pressure-logger interpretation guide addresses a different signal: changing hydraulic head, not stable-isotope composition.
Words carry model assumptions. The difference between ‘consistent with’ and ‘proves’ is as important here as the difference between a delta value and a concentration. The eduKate guide to lexical gaps provides another way to think about how specialist meaning is lost when a technical term is compressed into a casual one. The best translation of a graph includes its uncertainty and its alternative explanations.
Primary methods and reading sources
IAEA/WMO Global Network of Isotopes in Precipitation
USGS stable oxygen and hydrogen isotope interpretation
USGS stable isotope analysis and reporting method
USGS application of stable oxygen and hydrogen isotopes to groundwater