Borehole nuclear magnetic resonance (NMR) logging estimates how much hydrogen-bearing pore fluid contributes to a signal at successive depths and how quickly that signal decays after magnetic excitation. A useful borehole NMR log can illuminate water content, mobile-versus-less-mobile water fractions and the pore environment. It does not insert a tiny camera into the rock or directly measure water flowing through an aquifer. Reading the decay correctly means separating the observed magnetic signal from the empirical models used to infer hydraulic properties.
If you are looking at a borehole NMR T₂ distribution, begin by identifying the depth, the water-content units, the transverse relaxation-time bins and the tool’s effective measurement volume. A distribution may show strong short-T₂ signal and a weaker long-T₂ tail. The immediate question is how much water the recorded signal represents and how that signal is split across relaxation times. The later question—how freely the water can move through connected pores—needs field-specific calibration rather than an automatic label.
A deliberately simple example shows the translation. Imagine a calibrated spectrum with water-content contributions of 0.04, 0.08 and 0.06 cubic metres of water per cubic metre of sampled bulk rock in three T₂ groups. The total water content represented is 0.18 m³/m³, or 18% of the sampled volume. If a justified operational cutoff classifies the first 0.04 m³/m³ as less mobile, then 0.14 m³/m³ falls into the longer-time class. The arithmetic is straightforward; deciding whether that cutoff represents mobile groundwater under the site’s pore structure is a separate scientific claim.
What magnetic resonance observes
Hydrogen nuclei behave as tiny magnetic moments. A static magnetic field establishes a preferred orientation and precession frequency; radio-frequency pulses perturb the spin system, and the resulting magnetisation produces a measurable response that relaxes with time. Borehole NMR instruments use a designed magnetic field and pulse sequence to sample a volume around the well. The recorded decay amplitude depends on signal from hydrogen-bearing fluid and on acquisition sensitivity, environmental noise and calibration. The primary observable is a magnetic response, not a porosity percentage written directly by nature.
T₂ denotes transverse relaxation time: how rapidly the detectable transverse magnetisation loses coherence. In porous material it often reflects interactions between pore fluid and surfaces, fluid properties and magnetic-field inhomogeneity, among other influences. Shorter relaxation can be associated with water confined close to surfaces or smaller effective pores, while longer relaxation may represent less surface-dominated fluid. But the relationship is not a universal ruler. Temperature, mineral magnetism, fluid chemistry and surface relaxivity can shift relaxation without a corresponding change in nominal pore diameter.
An exponential decay is the simplest useful model
For a single idealised relaxation component, amplitude may be written A(t) = A₀ exp(−t/T₂). If T₂ is 40 milliseconds, amplitude at t = 40 ms is about 36.8% of the initial value; at 80 ms it is about 13.5%. These percentages describe exponential decay of magnetic signal, not the percentage of water that has physically left the pore. Real rocks generally contain many environments with different relaxation times, so a measured decay can be represented as a weighted sum or a continuous distribution of exponentials.
That multi-component inverse problem matters. The instrument measures a finite, noisy collection of echoes. Processing estimates a distribution over relaxation time that could have produced them. Regularisation and the selected number and spacing of T₂ bins influence how smooth or separated the result looks. A narrow peak in one inversion is not automatically proof that all water resides in a single sharply defined pore size. Readers should ask how the decay was acquired, how the inversion was constrained and whether changes in the spectrum are stable against reasonable processing choices.
Why echo time is not T₂
A pulse sequence commonly records a train of echoes separated by an acquisition setting known as echo spacing. Echo spacing in milliseconds is not the same thing as a fitted T₂ relaxation time. A tool with 2 ms echo spacing cannot perfectly recover every population that decays much faster than its earliest detectable echo. Some very short-T₂ water may be under-represented, and a porosity result may be biased low if the measurement misses fast-relaxing components. The detection window therefore belongs in the interpretation alongside the spectral bins.
Suppose one sample contains 0.06 m³/m³ of rapidly relaxing pore water that has nearly vanished before the first measured echo, plus 0.12 m³/m³ of slower relaxing water. A poorly suited echo spacing may faithfully report 0.12 m³/m³ of observable signal yet omit much of the first 0.06. Calling 12% the complete porosity without discussing acquisition loss would overstate certainty. A calibration, shorter-echo method or independent porosity comparison may help expose the missing component. The method’s ability to detect a component is part of its meaning.
T₁, T₂ and T₂* are different parameters
T₁ is longitudinal relaxation, describing how magnetisation returns toward equilibrium along the static magnetic field. T₂ is transverse relaxation, associated with loss of phase coherence after excitation. T₂* also reflects field inhomogeneity and can be shorter than the intrinsic T₂ under common conditions. Readers sometimes take a value from one pulse sequence and compare it directly with another because each is printed in milliseconds. The unit may match while the physical experiment differs. Keep the symbol, sequence and acquisition conditions attached to the number.
For borehole groundwater interpretation, T₂ distributions are often central, but some tool designs and workflows use T₁ information or multiple acquisition modes. That can improve fluid discrimination or reveal environmental effects. It also creates another opportunity for labels to mislead. A long T₁ is not a claim that water moves quickly, just as a long T₂ is not a measured hydraulic velocity. The relaxation parameter belongs to the nuclear spin system. Hydraulic mobility is a separate concept that must be inferred from pore connectivity and flow conditions.
The simplest integration: amplitude as a water-content indicator
When calibrated and acquired appropriately, the integrated NMR signal can be expressed as volumetric water content. A reading of 0.20 m³/m³ means an estimated 0.20 cubic metres of detected water per cubic metre of the instrument’s sampled bulk material. It does not mean the well holds 20% water along its entire length, nor does it guarantee that 20% of the rock’s pores are connected to a pumping well. In fully saturated, signal-accessible formations it may approximate water-filled porosity. In unsaturated or poorly detected conditions, water content and total pore volume can diverge.
Units prevent several common mistakes. Water content may be displayed as a fraction, percent bulk volume or porosity units. An NMR result of 0.18 m³/m³ corresponds to 18% bulk volume, not 0.18% and not 180%. If a plot stacks bins whose contributions sum to 0.18, the sum is the total represented signal fraction, not the mean relaxation time. To compute the geometric mean T₂, you need the log-weighted relaxation times as well as bin amplitudes. To estimate flow properties, you need an additional relationship and calibration.
Bound and mobile are operational classes
In applied NMR interpretation, a T₂ cutoff may divide relatively short-relaxation signal from longer-relaxation signal. Labels such as bound water, free-fluid index and mobile water are convenient when they have been calibrated to a particular material, pulse sequence and reference experiment. However, water is not literally tied to a mineral by an on-off switch at exactly one millisecond threshold. Bound water itself has several physical meanings, including surface-associated water and water retained by small pores under an imposed capillary condition.
A cutoff selected in one sandstone study should not be carried unchanged into a different clay-rich sediment, fractured bedrock or groundwater setting without evidence. Mineral surface relaxivity, pore size, chemistry and the purpose of the mobility classification can differ. A more defensible description says “signal assigned to the shorter-T₂ class by the stated cutoff” until direct hydraulic testing shows how that class relates to water that can move under a particular pressure gradient. This is not excessive caution. It prevents a processing choice from masquerading as a direct field observation.
A worked water-fraction classification
Consider an illustrative spectrum at a depth of 60 m. It contains signal equivalent to 4% bulk volume at about 5 ms, 8% at about 20 ms and 6% at about 100 ms. The integrated represented water content is 4 + 8 + 6 = 18%. If a local workflow uses a 10 ms cutoff, the 5 ms group is placed in the short-time class and the 20 and 100 ms groups form a 14% longer-time class. The longer-time share is 14/18, about 77.8% of the detected water signal. This is not a percentage of all groundwater available for extraction.
If a different justified cutoff of 30 ms were used, both the 5 ms and 20 ms groups would fall in the shorter-time class. The longer-time class would then be 6% bulk volume, only one third of the detected signal. Nothing about the measured three-bin spectrum changed. The classification changed because the chosen interpretation rule changed. This experiment teaches why the cutoff must be printed with reported bound or mobile fractions, especially when the result may feed a recharge, water-supply or contamination-transport decision.
Why a T₂ spectrum is not a pore-size photograph
Under a simplified surface-relaxation model, 1/T₂ may include a term proportional to surface-to-volume ratio multiplied by a surface relaxivity coefficient. For an ideal pore geometry, surface-to-volume ratio relates to pore size. However, the coefficient can vary among minerals and fluids, and real pores are connected, irregular and heterogeneous. Therefore a longer T₂ may suggest a larger effective pore environment under comparable chemistry and surface conditions, but it is not a direct micrometre-scale measurement of diameter.
Imagine two sites with identical geometric pores but different concentrations of paramagnetic minerals. Their T₂ distributions may differ because magnetic effects alter relaxation. Conversely, two sediments with different pore structures could show similar T₂ distributions because their surface properties differ. Translating a short-time peak as “clay” or a long-time peak as “coarse sand” without independent evidence collapses multiple causal mechanisms into one guess. The right practice is to combine lithology, gamma, resistivity, core descriptions and water tests while keeping signal-based inferences visibly conditional.
Borehole NMR versus a laboratory soil NMR experiment
A laboratory low-field NMR measurement may investigate a soil specimen under controlled saturation, sample geometry, temperature and pulse conditions. A borehole NMR tool interrogates an in situ radial volume around the well, through borehole fluid and potentially casing or annular material, with field-specific magnetic and noise constraints. These are not identical reader jobs. The previously published soil low-field NMR guide explains specimen-scale T₂ interpretation; this article is about depth-registration, tool sensitivity, water-bearing formations and field hydraulic inference.
Field measurements deserve their own quality questions. Is the borehole diameter compatible with tool requirements? Does casing interfere with excitation or detection? Has the instrument sampled beyond an annular filter pack? Is the well fluid unusually conductive or magnetically noisy? What is the vertical sampling interval and what is the radial depth of investigation? These questions do not disappear because the NMR inversion produces a smooth colourful spectrum. They explain why similar-looking laboratory and field figures can support different conclusions.
A radial measurement volume has practical boundaries
Borehole NMR tools are built around particular magnet and antenna geometries. Their sensitive regions may be annular volumes at a radius from the tool rather than all rock visible from the hole. If a borehole has a thick gravel pack or casing and the tool’s sensitivity does not fully reach the native formation, the response can over-represent material near the well. Conversely, a design with an investigation region extending beyond annular completion material may better target the aquifer. The exact geometry is manufacturer- and configuration-dependent, so generic radius claims should be checked against the specific tool.
Depth averaging introduces another limitation. A nominal half-metre sample interval is not the same as an infinitesimally thin slice of rock. The sensitive volume may include material above and below the nominal depth, and heterogeneous fractures can cross only part of it. A narrow permeable vein may contribute little integrated NMR signal while still carrying important groundwater flux. That is not a contradiction: NMR is sensitive to water volume and relaxation behaviour, whereas a flowmeter or packer test probes movement or hydraulic response under particular conditions.
The geometric mean relaxation time has a precise definition
A common descriptive statistic is the signal-weighted geometric mean T₂. For discrete positive-time bins, compute exp[(Σ aᵢ ln T₂ᵢ)/(Σ aᵢ)], where aᵢ is each bin’s signal-derived water contribution. In the illustrative 5, 20 and 100 ms groups with weights 0.04, 0.08 and 0.06, the weighted logarithmic mean produces approximately 25.13 ms. This is not the ordinary arithmetic mean, which would yield a different number and over-weight the long-time tail differently. Report whether a published mean is arithmetic, geometric or otherwise.
The calculation also requires compatible bin definitions. Are plotted values per logarithmic T₂ interval, per linear interval or already integrated contributions? Summing spectral heights without accounting for bin widths may give a biased water content or mean. Check whether the software exports integrated contributions directly. If data are in distribution density per logarithmic unit, the area under the plotted distribution requires multiplication by the appropriate logarithmic-bin width. This is the spectral equivalent of checking a denominator in a rate calculation. A correctly drawn axis can still hide an integration mistake.
NMR can estimate hydraulic conductivity but does not directly measure K
Hydraulic conductivity K relates water flux to hydraulic gradient under Darcy-type assumptions and is commonly expressed in metres per second. Borehole NMR may be used to estimate K with empirical models that combine water content and relaxation statistics. One family, often called the Schlumberger–Doll Research approach, uses a calibrated coefficient multiplied by a porosity term and a squared characteristic T₂ term. The exact exponent, cutoff and coefficient depend on the chosen formulation and calibration convention. No universal constant converts one T₂ value to a trustworthy K in every aquifer.
For groundwater work, this is a vital separation. A longer mean relaxation time may coincide with a more conductive pore network in a given calibrated setting, but connectivity, tortuosity, cementation, clay effects and fractures can change the relationship. An aquifer test may reveal high transmissivity in a narrow fracture that makes only a small contribution to the NMR water-volume signal. Conversely, a porous but poorly connected matrix can contain substantial water yet yield slowly. NMR estimates become most useful when their empirical relationship is checked against nearby direct hydraulic measurements.
The difference between permeability and conductivity
Intrinsic permeability k has units of square metres and refers to a material’s ability to transmit a fluid under a pressure gradient in Darcy’s law. Hydraulic conductivity K has units of metres per second and also depends on fluid density, gravity and dynamic viscosity. They are connected by K = kρg/μ for a single-phase fluid under the conventional form of Darcy’s law. If two tests use fluids with different viscosity, they may infer different hydraulic conductivities for the same intrinsic permeability. Confusing the two is a familiar cross-disciplinary translation error.
Suppose k is 1 × 10⁻¹² m², fluid density 1,000 kg/m³, gravitational acceleration 9.81 m/s² and viscosity 0.001 Pa·s. Then K = (1 × 10⁻¹²)(1,000)(9.81)/(0.001) = 9.81 × 10⁻⁶ m/s. The calculation checks dimensionally: m² × (kg/m³) × (m/s²) divided by (kg/(m·s)) yields m/s. The numerical result is a model example, not derived from an NMR spectrum. It shows why a report must state whether its final column is k or K and whether a fluid correction was applied.
When a model performs only within a factor of several
A field-calibrated NMR conductivity model may agree with independently measured K only within a factor of several, not within a few percent. This can still be valuable for mapping depth variations, but it is very different from saying the instrument directly measured K at each depth. A U.S. Geological Survey study of glacial aquifers evaluated an NMR K model against direct-push permeameter measurements and discussed calibration and uncertainty. The lesson is that the model must be evaluated against a relevant reference dataset, not just fitted until the plotted curves look attractive.
For a report, state the calibration sites, the range of K represented, whether the validation was independent, the error measure used and the performance outside the calibrated range. A factor-of-five spread corresponds to a much larger uncertainty than a plotted numerical label with three significant digits implies. If a model predicts 1 × 10⁻⁵ m/s, a factor-of-five comparison range spans roughly 2 × 10⁻⁶ to 5 × 10⁻⁵ m/s under a symmetric multiplicative framing. That range can change a water-management decision. The calculation’s apparent precision should not hide the model’s empirical uncertainty.
A first-pass interpreter’s depth sheet
At each selected depth, record the signal-derived volumetric water content, T₂ distribution, acquisition settings, short- and long-time fractions with explicit cutoff, quality flags and the source of any derived conductivity value. Then add the borehole completion details, radial investigation estimate, geological log and nearby hydraulic observations. This sheet prevents a compact NMR plot from acquiring impossible powers. The spectrum supports one set of physical inferences; the hydraulic model supports another, conditional on calibration; the well construction may determine whether either represents undisturbed formation.
For communication to non-specialists, say “the NMR signal indicates more detected water and relatively longer relaxation times at this depth” before concluding “this interval will yield more water.” The second statement may eventually be justified by field tests, but it does not follow automatically from the first. These distinctions mirror the vocabulary work in How English Works: Vocabulary and Meaning, where the meaning of a term is examined in context rather than treated as one detachable dictionary substitute.
Quality check: does the signal follow the instrument’s detection window?
Every NMR tool has limits set by pulse spacing, polarisation, signal-to-noise ratio and its sensitive volume. Before comparing two depths, determine whether those acquisition settings remained consistent. A late, noisy tail may be heavily influenced by regularisation; an early missing signal may reflect a physical component faster than the instrument can sample. Tool output may include signal-to-noise indicators, fit residuals and uncertainty estimates. A smooth coloured T₂ distribution does not prove that the inversion uniquely resolved every apparent peak, especially when the underlying decay contains only modest independent information.
Imagine the same specimen measured using two sequences, one with a short first echo and another that begins later. The later sequence can under-detect the rapidly relaxing component, making the inferred mobile fraction appear larger even though the geological water has not changed. If the resulting plots are compared as if the acquisition were identical, a processing difference becomes a false claim about groundwater mobility. The repair is to compare acquisition windows, reprocess on a compatible basis where possible and report the portion of relaxation space outside reliable sensitivity.
Quality check: magnetic minerals and conductive environments
Magnetic susceptibility and paramagnetic constituents can accelerate apparent transverse relaxation or create signal distortions. Formation and borehole fluid conductivity may affect radio-frequency penetration, tool response and noise. Temperature can influence fluid viscosity and relaxation, while salinity and dissolved ions can alter water’s NMR behaviour. None of these is automatically a fatal flaw. They are reasons to read the environment section of the method rather than declaring that a change in T₂ proves a change in pore throat size. The affected interval may require specialised corrections or calibration.
A practical comparison involves two lithologically distinct units with similar water content. The first is quartz-dominated and the second contains more magnetic or clay minerals. A shorter T₂ spectrum in the second unit may reflect smaller pores, surface interactions, magnetic effects or a combination. A conventional gamma log, resistivity log, mineralogical analysis and controlled laboratory sample can help distinguish hypotheses. But the final interpretation should state which explanation has supporting evidence and which remains a possibility. A plausible physical story is not the same as a uniquely identified mechanism.
What happens when a signal becomes too small
At low water content or poor coupling, NMR signal may approach the system’s detection threshold. An inversion can still produce nonnegative bins because of processing constraints, but those bins might not justify fine-grained claims about water mobility. Look for the instrument’s noise floor, uncertainty bands and retained echoes. If several neighbouring depths show nearly identical tiny peaks at the same T₂ values, ask whether the pattern is an artefact of regularisation. Under these conditions, “no confidently resolved signal” is more informative than a fabricated zero-water assertion.
A below-detection result is not equivalent to no water. Some water may be outside the sensitive volume or relax too rapidly for the chosen sequence. The appropriate question becomes whether the detection limit is low enough for the geological decision at hand. If a method is being used to screen layers for possible groundwater storage, conservative classification might require a second test rather than pronouncing every signal-poor layer impermeable. Measurement sensitivity sets the lower bound of what can be claimed, and understanding this boundary is part of translating the log accurately.
Why vertical sampling and radial volume change comparisons
In a heterogeneous bedrock well, the NMR tool may sample an annulus dominated by matrix porosity, while a narrow fracture conducts most flow. Its T₂ distribution may barely register the fracture’s small volume even when a flowmeter reveals substantial movement. Conversely, a thick porous layer can hold a large amount of water but contribute little flow because its pore connections are tortuous. The difference is not a weakness to be concealed. It is the distinction between storage-related evidence and rate-dependent hydraulic behaviour.
A depth plot may show NMR water content every half metre and a heat-pulse flowmeter response over a different effective vertical interval. Correlating a single sharp flow anomaly to a broad NMR water-content peak requires attention to support volume and registration. If the features do not align precisely, first test whether their instrument footprints differ. The geophysical log heat-pulse flowmeter guide is helpful for separating measured borehole flow from an inferred hydraulic contribution; neither replaces the other.
A hypothetical aquifer profile
Picture a 40-to-70 m sedimentary interval where calibrated NMR water content is roughly 0.25 m³/m³ but much of the T₂ signal lies in short-time bins. Beneath it, a coarse-grained interval shows 0.18 m³/m³ with a stronger long-time component. It is tempting to rank the first as the superior aquifer because 25% is larger than 18%. That conclusion may be wrong if the first material retains much of its water in small, poorly connected pores. The interpretation should compare both amount and relaxation distribution, then seek hydraulic evidence.
Now suppose interval tests independently show 1 × 10⁻⁷ m/s for the first unit and 1 × 10⁻⁴ m/s for the second. The more porous interval is less conductive by a factor of 1,000 under those test conditions. This numerical contrast is a constructed teaching scenario, not an empirical site result. It shows why a storage-related measurement and a transport-related measurement answer different questions. The NMR result remains useful because it describes available water signal and pore environment, but it never guaranteed hydraulic yield by itself.
Comparing an NMR-derived K model to field hydraulic tests
If calibration measurements cover K from 10⁻⁷ to 10⁻⁴ m/s, plotting both NMR-predicted and directly measured values on a logarithmic scale may reveal bias and scatter. A useful comparison reports how predictions perform across the range, whether high- and low-K units are systematically misestimated and how many reference measurements support the relationship. Agreement at one field point does not verify every depth, particularly if lithology or saturation changes. Avoid fitting and validating with exactly the same small dataset while presenting the residual as independent predictive accuracy.
Suppose a model predicts 2 × 10⁻⁵ m/s while a packer or permeameter test suggests 1 × 10⁻⁵ m/s. Their ratio is two, and their difference is 1 × 10⁻⁵ m/s. These are different error descriptions. If a second interval has a predicted value of 2 × 10⁻⁷ m/s and measured value of 1 × 10⁻⁷ m/s, it has the same ratio but a hundredfold smaller absolute difference. Ratio-based or log-error evaluation is often informative over a wide conductivity range, provided the convention is stated. Neither comparison removes the need to examine the direct-test assumptions.
The hydraulic model should show its units
An empirical conductivity expression may contain porosity, T₂ time and a calibration coefficient. Porosity is dimensionless, but a squared T₂ is time squared; the coefficient must carry units that produce the desired metres-per-second K. If T₂ is entered in milliseconds rather than seconds, a coefficient obtained under the other convention can give a wildly incorrect estimate. With a squared-time term, switching from milliseconds to seconds changes the numerical square by a factor of one million. The issue is not obscure mathematics; it is a mismatch between the data units and the empirically fitted model.
A practical method sheet states the exact formula, how the T₂ statistic is computed, the time unit, porosity convention and the empirical coefficient with its units. If the coefficient was calibrated using percent porosity values rather than fractions, that distinction also matters dramatically when porosity appears to a high power. Never transplant a fitted coefficient without reproducing its input conventions. Before publishing a numerical K, feed one known calibration row back through the equation and verify that the predicted result matches its published value. That reverse check catches many avoidable unit errors.
Why hydraulic conductivity and storativity differ
Hydraulic conductivity describes how easily water moves under a hydraulic gradient; storage properties describe how an aquifer releases or takes up water as head changes. NMR water content can be relevant to understanding available pore fluid, yet total detected water is not automatically drainable specific yield. Capillary retention, compressibility, confinement and saturation determine how water responds to a head change. A well with substantial NMR-detected water may release only a limited proportion during a short pumping test, or may draw on compressive storage in a confined setting.
An educational translation should separate three sentences: “The rock contains an estimated volume fraction of NMR-detected water”; “its relaxation distribution has a stated set of shorter and longer components”; and “its hydraulic release or transmission must be evaluated under a particular field test.” When those become one sentence claiming NMR directly measures aquifer yield, three different mechanisms have been collapsed. Explicitly naming the quantity makes the explanation shorter and more accurate. This principle transfers readily to reading energy, biology and economic datasets where a stock and a flow are regularly confused.
NMR and conventional neutron logs do not measure the same thing
A neutron log responds to hydrogen-related slowing or capture of emitted neutrons and is interpreted through tool calibrations and formation assumptions. A borehole NMR tool measures magnetic resonance from hydrogen-bearing fluids in a selected sensitive region and estimates relaxation distributions. Both can be related to water-filled porosity under appropriate conditions, but their physics, environmental sensitivities and information content differ. A close agreement between their bulk porosity estimates can improve confidence; a disagreement is an opportunity to investigate clay effects, hydrogen-bearing minerals, borehole influence or the detection window.
The existing neutron-log guide addresses hydrogen index and conventional neutron response. This page adds relaxation-time information and field-calibrated conductivity inference rather than rewording neutron porosity. When presenting a joint interpretation, avoid the vague phrase “both measure water so they should match exactly.” Their different measurement volumes and detection processes explain why modest disagreement is physically possible even when each instrument functions correctly.
NMR and density logs can support different porosity checks
A density tool measures gamma-ray response used to estimate bulk density, often followed by a matrix-fluid mixture calculation to infer porosity. NMR instead obtains signal associated with hydrogen-bearing pore fluids under its magnetic detection conditions. If density-derived porosity is 22% and NMR-detected water content is 18%, one question is whether some pore volume contains fluid not fully detected by the NMR sequence. Another concerns matrix density assumptions, borehole washouts or differences in sampled volume. It is premature to declare that exactly 4% is filled with air or immobile material without additional evidence.
Use the gamma-gamma density guide for the mass-density calculation and its environmental limitations. The independent-check workflow is to verify each instrument’s physics, apply appropriate depth registration, and compare under compatible saturation and lithology assumptions. A visible numerical difference is the beginning of a diagnostic question, not necessarily a failed instrument or proof of a single pore-fluid configuration.
Transfer test: distinguish a fraction of bulk rock from a fraction of detected water
A spectrum reports 0.24 m³/m³ total signal, including 0.09 m³/m³ in a short-T₂ class. What proportion of bulk material is assigned to the long-T₂ class, and what proportion of the detected water is that? The long-time bulk fraction is 0.24 − 0.09 = 0.15 m³/m³, or 15% bulk volume. Its share of the detected water is 0.15/0.24 = 0.625, or 62.5%. The two percentages have different denominators. Reporting only “62.5% mobile water” without clarifying the denominator and cutoff can mislead someone who reads it as 62.5% of rock volume.
The follow-up question is whether 0.15 m³/m³ must flow to the well during pumping. The answer is no. A relaxation cutoff provides an operational classification of observed water signal. Actual drainability and flow depend on connectivity, hydraulic gradients and timescale, and the cutoff requires site-specific validation. You can accurately report “15% bulk-volume signal assigned to the long-T₂ class,” then separately test whether it predicts behaviour. This is a powerful habit: write the numerator, denominator and physical test represented by every percentage.
Transfer test: calculate a T₂-weighted statistic carefully
A hypothetical three-bin spectrum gives 0.02 m³/m³ at 10 ms, 0.06 at 40 ms and 0.02 at 160 ms. The total detected fraction is 0.10. The signal-weighted logarithmic mean is [0.02 ln(10) + 0.06 ln(40) + 0.02 ln(160)]/0.10. Because the middle time receives 60% of the weight and the other values are symmetric in logarithmic distance around 40, the geometric mean is 40 ms. The ordinary weighted arithmetic mean would be 0.2 × 10 + 0.6 × 40 + 0.2 × 160 = 58 ms. Both are calculable, but they answer different statistical questions.
Why not choose 58 ms simply because it is larger? The reason is methodological rather than aesthetic: relaxation distributions span orders of magnitude, and a particular hydraulic model may have been calibrated using one explicitly defined statistic. Changing it without recalibrating alters the model inputs. A scientific report should state the statistic, not just “average T₂.” This small exercise transfers to any field where logarithmic scales, multiplicative variation or spectral densities are involved. Readers can defend the average only if they can reconstruct its weighting.
Transfer test: identify what an empty short-T₂ region means
A borehole spectrum begins at a first measurable echo around 2 ms and contains little signal in the first displayed bins. Does that prove the sample contains no tightly surface-associated water? No. It may indicate that the relevant signal decayed before the instrument could observe it, that magnetic conditions hindered detection or that the inversion lacked sensitivity to those components. A useful interpretation documents the detection window and considers complementary measurements. “Not detected” should remain distinct from “physically absent,” especially for fast-relaxing components.
The independent check is to inspect acquisition settings, model recovery tests if available and comparisons with laboratory or alternative porosity measurements. If the instrument has a known short-T₂ blind spot, constrain conclusions to the reliably sampled time range. Assigning an exact zero based on a smooth inversion plot ignores the method’s observation boundary. Technical readers do not require false certainty; they require a statement precise enough that another investigator can test it and understand what evidence was not collected.
Transfer test: compare storage and flow without a category error
A porosity-related NMR reading is 0.30 m³/m³ in layer A and 0.14 in layer B. A calibrated hydraulic investigation reports 5 × 10⁻⁸ m/s in A and 2 × 10⁻⁴ m/s in B. Which layer stores more detected water per sampled bulk volume, and which transmits more readily under the stated hydraulic model? A has the larger detected water fraction. B has the larger measured conductivity, by 4,000 times. These are not contradictory results because storage and movement depend on different aspects of pore architecture and fluid mechanics.
Neither single-layer value determines whole-well yield without thickness, connectivity, drawdown conditions and boundary effects. A thin high-K seam and a thick moderate-K aquifer may contribute differently to a test. Hydraulic transmissivity incorporates thickness when a representative K applies, but heterogeneous layers demand care. The reader should resist turning one depth sample into a production rate. This transfer problem tests whether you are reasoning with the physical quantity or merely ranking plotted numbers.
Frequently asked question: can NMR show permeability directly?
No. It produces magnetic resonance observations from pore-fluid hydrogen, from which water-content and relaxation properties can be estimated under calibration. Permeability or hydraulic conductivity is derived using a model that links those observables to pore connectivity and flow, usually requiring local testing. Some calibrations can be highly useful; none should be described as a direct measurement of flow through every pore. If a log displays an NMR-derived K curve, look for its formula, coefficient, calibration domain and validation error before using the values.
Frequently asked question: what does a long-T₂ peak mean?
Under comparable material, fluid and acquisition conditions, it often indicates a fluid environment with slower relaxation, which may be associated with larger effective pores or less surface-dominated water. But it does not uniquely identify grain size, connected permeability, drainable yield or water quality. A long-time peak could shift with mineralogy, fluid state, temperature or inversion choices. Read the peak together with its amplitude: a long relaxation component representing only 1% bulk volume is different from one representing 15%. The amount and the relaxation time are both part of the interpretation.
Frequently asked question: does a 20% NMR value mean 20% groundwater yield?
No. A 20% volumetric signal means an estimated fraction of the sampled material contributes detectable hydrogen-bearing fluid under the method’s assumptions. Pumpable yield depends on aquifer thickness, hydraulic conductivity, hydraulic gradients, boundary conditions, well design and sustainable drawdown. Water can be present without being rapidly extractable. For a public explanation, resist the tempting shortcut from “water in pores” to “water delivered by a pump.” A proper field investigation uses additional hydraulic tests for that prediction.
Frequently asked question: why do different inversion settings produce different peaks?
Finite noisy echo data do not uniquely define an arbitrarily detailed relaxation spectrum. Regularisation, non-negativity constraints, chosen time bins and smoothing affect the displayed distribution. A useful method report describes those choices and, for important anomalies, tests whether the principal integrated fractions survive reasonable alternatives. If a conclusion depends entirely on one narrow inversion peak that disappears under small processing changes, its evidence is fragile. Broader conclusions about total signal may remain comparatively stable even when the fine spectral shape is uncertain.
A final interpretation example with explicit limits
Suppose a fictional well interval has calibrated NMR signal equivalent to 0.19 m³/m³, mostly in longer T₂ classes than a neighbouring clay-rich layer. Density-derived porosity and a gamma-log lithology interpretation broadly support a coarser or less surface-dominated pore environment. A hydraulic model estimates K near 10⁻⁵ m/s, but local validation covers only nearby materials and predicts values with substantial scatter. The responsible statement is that the interval contains more detectable water signal with a comparatively long relaxation distribution and has a tentative calibrated conductivity estimate. It is not yet a proven high-yield production zone.
If an independent interval hydraulic test later agrees with that estimate within its uncertainty, the case strengthens. If it disagrees, the mismatch is scientifically useful: fracture-dominated flow, magnetic mineral effects, incorrect cutoff or scale differences may deserve investigation. Do not rewrite the original observation to fit the test after the fact. Retain both data, revisit the model and explain what changed in your understanding. Readers can then see the pathway from an NMR instrument to a defensible groundwater statement.
Learning route and contextual connections
The How to Translate Easily series index connects explanations built around translating measurements into meaningful claims. For conventional density, neutron and flow data use the neighbouring guides linked above, then compare their differing measurement volumes. To practise precise scientific language, the eduKate article How English Works Across School Subjects reinforces the distinction between observations, models and conclusions. Good technical translation keeps the units and assumptions intact, even when the explanation is short enough for a reader new to the field.
Institutional research and methodological sources
USGS Borehole Nuclear Magnetic Resonance overview explains the in situ method and target hydrogeologic properties. USGS-led field demonstration of slim-hole borehole NMR discusses tool dimensions, echo spacing, radial measurement, water fractions and comparison with other information. USGS assessment of NMR-based hydraulic-conductivity estimation reports calibration against direct measurements and its uncertainties. USGS fractured-bedrock NMR field study provides application context. All numerical exercises in this guide are invented solely to test the calculations; they are not observations from those investigations.