Refraction microtremor, commonly called ReMi, is a passive seismic-array method that uses ambient ground vibrations to estimate a shear-wave velocity profile. If you are trying to understand a ReMi p–f plot, slowness versus frequency, the lower-velocity dispersion envelope or an inverted Vs model, begin with the array direction and which part of the noise wavefield the analyst selected. A bright patch on a slowness–frequency image is not automatically a soil layer.
The central interpretation challenge is directional. Ambient waves may reach a linear array from many angles, so the measured propagation along the line can have an apparent phase velocity higher than the true velocity associated with the wave. ReMi processing exploits the distribution of wavefield power in slowness–frequency space and seeks a defensible dispersion boundary before inverting a layered-earth model. The final Vs profile is a model conditional on those picks, not a direct borehole measurement.
This guide explains ReMi acquisition, ambient microtremors, p–f spectral-ratio images, reciprocal velocity units, directional bias, fundamental versus higher Rayleigh modes, inversion uncertainty, Vs30 and comparison with active MASW. Worked values are fictional and checkable with simple arithmetic. The aim is to help readers understand what the array recorded, what was inferred and when a seemingly impressive depth claim exceeds the data.
Quick answer: what does ReMi measure?
ReMi records ambient seismic motion at multiple surface receivers, transforms the data into a frequency–slowness representation and interprets a surface-wave dispersion relationship. The picked relationship is inverted to estimate shear-wave velocity varying with depth. It is called a refraction microtremor method, but its central dispersion interpretation should not be confused with ordinary first-arrival seismic refraction through body waves.
The original John Louie 2001 ReMi paper developed a slowness–frequency approach for estimating shallow shear velocity using common multichannel seismic recording equipment and ambient noise. Its case studies reported performance under specific array and ground conditions. Those historical demonstrations are not promises that every present-day site can be investigated to one fixed depth or accuracy.
Ambient vibrations can carry useful information
Microtremors are small ground vibrations produced by natural phenomena and human activity. Traffic, industrial machinery, ocean-generated motion and other sources may contribute at different frequencies and times. A passive survey uses these signals rather than a specifically triggered hammer impact as its principal energy source.
This can be valuable when large controlled sources are impractical. However, the absence of a controlled trigger changes interpretation. The direction, coherence and spectrum of the noise sources matter, and the analyst may not know their locations. ‘Passive’ means the instrument listens to available waves, not that the data automatically represent an isotropic or perfectly diffuse source field.
Why a linear array is useful
ReMi commonly places many receivers at known positions along a line. The simultaneous records allow comparison of wave motion across space as well as time. If a coherent wave travels along the line, its progression produces a measurable time or phase trend across neighbouring geophones.
The array is a geometric measurement device. Its spacing, length, sensor coupling and orientation influence which wave components can be distinguished. A line installed beside a road may receive highly directional vehicle-generated energy, while another location may have noise arriving from several azimuths. Both can be analysed, but the source distribution affects which features of the p–f image are trustworthy.
Refraction in the name can mislead
In conventional seismic refraction, analysts often pick first arrivals and model body-wave travel through layers with different wave velocities. ReMi instead uses ambient recordings to recover a dispersion relationship for surface-wave energy, generally relating frequency and phase velocity. The appearance of ‘refraction’ in the method name should not be used to infer that its results are equivalent to a refraction travel-time section.
This difference also explains why a ReMi profile may report Vs without the survey generating clear S-wave first arrivals in the same way as crosshole seismic testing. The information comes through a different wavefield and inverse problem. In technical writing, define the observed quantity rather than relying on the historical name of the method.
What is slowness?
Slowness is the reciprocal of velocity. If a wave has phase velocity c, its slowness p is 1 divided by c under the stated propagation convention. A velocity of 250 metres per second corresponds to p = 0.004 second per metre, equivalently 4 seconds per kilometre.
Slowness is sometimes plotted in seconds per kilometre because the numbers are convenient. A slowness value of 2 seconds per kilometre corresponds to 500 metres per second, not 2 metres per second. Reversing the reciprocal relation or misreading the axis unit is one of the fastest ways to create a wrong velocity interpretation.
Worked example: reading a p–f axis
Suppose a ReMi spectrum has a coherent feature at five hertz and a slowness of four seconds per kilometre. Convert the slowness to seconds per metre: 4/1,000 = 0.004 second per metre. The corresponding apparent phase velocity is 1/0.004 = 250 metres per second.
If another feature lies at two seconds per kilometre, its apparent velocity is 500 metres per second. The feature plotted at the numerically smaller slowness therefore represents the numerically greater speed. Any explanation that calls the upper or lower coloured ridge ‘faster’ must first check the actual orientation and units of the displayed p axis.
Apparent velocity can exceed true phase velocity
Imagine a plane wave arriving at an angle to a straight line of receivers. Only the component of its propagation across the line contributes to the observed phase delay from one receiver to another. Under a simple geometric model, the apparent phase speed along the line equals the true phase speed divided by the cosine of the arrival angle to that line.
A wave travelling exactly along the array has an apparent speed equal to the true speed. If the wave arrives obliquely, the apparent speed can be larger. This direction-dependent effect is central to interpreting a passive one-dimensional array. A very fast p–f feature does not necessarily mean extremely stiff soil; it may partly reflect source azimuth.
A numerical example of source direction
Assume a surface wave has true phase velocity 250 metres per second. If it travels at 60 degrees to the receiver line, the cosine is 0.5. Under the idealised plane-wave model, the along-line apparent speed becomes 250 divided by 0.5, or 500 metres per second.
The corresponding along-line slowness is 0.002 second per metre instead of the true 0.004. The same underground material can therefore create differently positioned apparent features depending on wave arrival direction. A ReMi analyst does not simply pick the brightest or fastest energy and label it the earth’s velocity.
What is on a ReMi p–f image?
A p–f plot displays slowness against frequency and usually uses colour or intensity to represent a processing measure related to coherent wavefield power or spectral ratios. Some implementations show a broad region of energy because waves of different arrival directions contribute different apparent slownesses. Other signals can arise from body waves, noise and higher modes.
The image is a transformation of recorded traces. It does not display the depth of a physical interface on either axis. The colours are not mapped soil categories and the highest-power pixel is not necessarily the desired true Rayleigh-wave dispersion point. Readers should find the processing definition, scale convention and chosen envelope before describing the physics.
Why a lower phase-velocity envelope is sought
Because oblique propagation can inflate the apparent velocity along a receiver line, the lower boundary of coherent apparent phase velocities can help constrain the true surface-wave speed under suitable wavefield conditions. In slowness coordinates, lower velocity corresponds to larger positive slowness. The method’s detailed picking criterion depends on the transformed data and signal strength.
Do not confuse ‘lower-velocity envelope’ with the lowest value on the vertical plot regardless of how the axis is oriented. Read whether the report displays slowness increasing upward or downward. The physical question is where coherent surface-wave energy supports the slower phase-velocity boundary, not which coloured band appears visually lowest on a page.
The envelope still needs energy travelling near the array direction
A reliable lower apparent-velocity boundary requires usable wave energy propagating with a sufficient component along the array. If most recorded noise arrives from angles far off the array, the true minimum may not be adequately sampled. The observed apparent velocities can then be biased high, leading to an overly fast inferred subsurface model.
This limitation means that increasing the number of minutes recorded does not necessarily guarantee a better result. A longer recording can help if it captures a greater variety of source directions or more coherent waves. But if source geometry remains unfavourable, additional samples may repeat the same directional bias. Acquisition and interpretation must consider the actual wavefield.
Why a busy road can help and hinder
Vehicle traffic may generate surface-wave energy with useful frequency content and directional coherence. A linear array aligned with a source path can record arrivals that make the slowness envelope easier to interpret. But isolated vehicle events or changing source positions can also create nonstationary spectra and directional artefacts.
The phrase ‘urban noise is useful’ is therefore conditional. A survey should record environmental context and inspect multiple time windows rather than treating every traffic source as ideal. A quiet site might need different acquisition arrangements, while an extremely noisy site may have plenty of energy but poor modal separation. Noise quantity and information quality are different concepts.
Receiver spacing can create aliases
A receiver line samples the wavefield at discrete positions. When spacing is coarse relative to the wavelengths and apparent slownesses of interest, multiple propagation patterns may become indistinguishable under sampling. Spatial aliasing can create misleading power ridges or false apparent slowness features.
A good processing report should state receiver positions and the credible frequency–slowness region. The fact that a plotting program displays energy everywhere does not mean every pixel is resolved. Like time sampling in an audio recording, spatial sampling limits what patterns can be reconstructed. Quality control begins with geometry before a dispersion envelope is picked.
Array length influences low-frequency interpretation
A longer spread can help constrain wave behaviour at larger wavelengths, assuming the available ambient sources excite relevant modes coherently. It also extends across more ground, potentially increasing lateral heterogeneity. The resulting one-dimensional model may average material over a broad aperture rather than identify local conditions directly under one geophone.
Do not treat an array length as a guaranteed maximum investigation depth. The usable wavelength range, source direction, low-frequency resolution and ground model all matter. Claims about a particular depth should be supported by sensitivity or inversion evidence, not borrowed from an unrelated field study with different equipment and soil conditions.
Recording time is a quality variable
Passive measurements can vary substantially through the day as traffic, wind or machinery activity changes. Several shorter windows may be more informative than one long record dominated by a transient source. Analysts often examine consistency among intervals, rejecting sections contaminated by instrument faults or unusual disturbances.
A record’s duration is not a direct proxy for depth resolution. If lower-frequency wave energy is absent, an extra hour may or may not improve the spectrum. The useful question is whether the selected windows contain reproducible coherent features suitable for phase-velocity interpretation. A good report says how windows were screened and combined.
What does the frequency axis mean?
Frequency is measured in hertz, cycles per second. Different surface-wave frequencies can carry information about different length scales because wavelength equals phase velocity divided by frequency. If a hypothetical wave has c = 250 m/s at f = 5 Hz, its wavelength is 50 metres. At 25 Hz with c = 200 m/s, the wavelength is eight metres.
This arithmetic does not say that the five-hertz component measured soil exactly fifty metres deep. Depth sensitivity depends on wave modes and layered elastic structure. Frequency is an observation coordinate; depth appears only after a model links the wave behaviour to the subsurface.
Wavelength is not equal to penetration depth
A common shortcut assigns an investigation depth as a fixed fraction of maximum wavelength. Such rules can be useful early design heuristics under certain assumptions, but they are not universal measurements. A high-contrast interface, a shallow velocity inversion or a different mode can change sensitivity significantly.
The report should justify depth using observed dispersion and modelling rather than an unqualified shortcut. Readers should ask what would happen to the predicted dispersion if a deep layer were made faster or slower. If the observable curve barely changes, that layer is poorly constrained regardless of how far the software extrapolates its plotted profile.
Rayleigh waves and higher modes
Layered ground can support a fundamental Rayleigh mode and higher modes. Higher-mode energy may appear at different apparent phase velocities for the same frequency. A ReMi image can therefore show multiple coherent features even when the instruments work correctly.
The analyst needs to identify which feature corresponds to the intended mode or incorporate several modes with an appropriate model. A bright higher-mode band should not be used as the fundamental dispersion curve without justification. A systematic mode-picking error can produce a convincing but biased Vs inversion, particularly at sites with sharp velocity contrasts.
Why the power maximum is not always the right pick
A peak in p–f power can reflect the direction and strength of a source, not only the earth’s true modal phase velocity. A dominant wave arriving obliquely may have high apparent speed and large signal intensity. Selecting its maximum would confuse wavefield amplitude with a property of the ground.
The original ReMi method explicitly considered how the envelope of coherent velocities can be separated from interfering waves. A transparent report should show the selected boundary and explain alternative ridges. When the envelope is uncertain, the reader should see that uncertainty before the profile is accepted as precise.
ReMi is a passive measurement with a modelled result
After a dispersion relation is picked from the p–f image, a layered elastic model is used to predict Rayleigh-wave behaviour. An inversion or interactive forward-modelling procedure seeks a Vs profile consistent with that dispersion. This sequence is important: the method observes ambient wave propagation and infers an underground model.
No receiver directly reports the shear-wave velocity at every depth in the final figure. The layer thicknesses and velocities are estimated parameters, often conditioned on starting assumptions. The resulting profile may be useful, but it should be interpreted as one solution consistent with the chosen observations and modelling framework.
A good fit does not establish a unique earth model
Two different Vs profiles can predict dispersion curves that are nearly indistinguishable over a limited frequency range. A shallow gradient and a two-layer step, for instance, may produce similar observations if the data are not sensitive to their differences. More fitted layers can decrease numerical misfit without making the added layer boundaries physically real.
A trustworthy interpretation considers acceptable alternative models and independent borehole or other field evidence. Readers should examine the frequency range of the fit and which depths remain unconstrained. A smooth, sharply coloured velocity section can conceal broad uncertainty in layer thickness or deeper Vs.
The first-principles lesson: preserve measurement and inference
The ReMi dataset contains recorded vibration traces and derived p–f information. A selected dispersion envelope is an interpretation of those traces. A Vs profile is a further model-based inference. These are three different levels of evidence, and the strength of the last depends on every earlier choice.
This distinction also affects public communication. A result described as ‘measured Vs of 400 m/s at eighty metres’ may be too strong if that value came only from a poorly constrained model. A more accurate report says the model estimated that layer velocity under specified assumptions, and explains the evidence or uncertainty supporting the estimate.
Model sensitivity is the real depth test
A ReMi model may display velocities at considerable depth, but a plotted parameter is not automatically constrained by the observed dispersion. To assess depth sensitivity, vary deeper layer velocities and thicknesses and calculate how much the predicted dispersion changes within the reliable picked frequency band. If a large change in the deep model barely alters the measured-range prediction, the data provide little independent information about that depth.
An analyst should therefore report the range of acceptable profiles or other meaningful uncertainty where available. Assigning one exact deep Vs value solely because the inversion software returned it overstates what was observed. The useful result may be a range of average velocities or a limit on possible contrast rather than a sharply resolved geological boundary.
An example of reciprocal slowness and wavelength
Suppose the interpreted ReMi envelope contains a point at 5 Hz and 3.5 seconds per kilometre. Convert to 0.0035 second per metre, then invert to obtain phase velocity about 285.7 metres per second. Its wavelength is phase velocity divided by frequency, or approximately 57.1 metres.
A second point at 10 Hz and 5 seconds per kilometre represents 200 metres per second and a wavelength of 20 metres. The lower-frequency feature is associated with a longer wavelength in this illustrative comparison. It may be sensitive to deeper structure under suitable models, but the numbers do not independently locate an interface at 57.1 or 20 metres.
Why the units on a p–f diagram matter
Some figures use seconds per metre, others seconds per kilometre or milliseconds per metre. These may be equivalent after conversion, but copying a value without its unit changes the implied velocity by factors of a thousand. A tick labelled 4 in a plot with units s/km is not a velocity of 4 m/s.
Before interpreting any curve, rewrite at least two points as pairs of frequency and phase velocity in consistent SI units. This one check detects reversed axes, reciprocal errors and some software export problems. It also reveals whether the analyst mistakenly treated an apparent phase velocity as a body-wave arrival speed.
A spectral boundary is not a depth boundary
The ReMi p–f plot uses frequency and slowness, neither of which is geological depth. An analyst who draws an underground boundary directly across a region of spectral energy is making an unjustified visual substitution. A depth interface appears only as a parameter in an earth model fitted to the dispersion.
Readers should insist on the intermediate reasoning: which energy supported the dispersion pick, which mode was assumed, what model was used and how its prediction matched the observed curve. The chain is particularly important when an attractive colour image gives the impression of direct subsurface photography. The analysis can be strong without making that claim.
Comparing time windows reveals source sensitivity
Ambient wavefields change over time, especially in urban environments. One window may be dominated by a passing vehicle travelling along the array, another by distant traffic at an oblique angle. If both are interpreted separately, their p–f images may show different apparent velocities even when the soil has not changed.
A robust survey can compare windows, inspect direction effects and identify stable envelope features. If results vary strongly with transient sources, the profile should carry that caveat. Averaging all windows indiscriminately may blend incompatible signals rather than remove noise. Quality control seeks physically consistent propagation information, not simply the largest possible dataset.
Traffic direction can produce falsely fast ridges
Consider the earlier example of a 250 m/s true surface wave arriving at sixty degrees to the line, giving an apparent along-line velocity of 500 m/s. If a loud vehicle supplies most of the wave energy from that direction, the brightest ridge may lie near the higher apparent velocity. This does not make the subsurface twice as fast.
An appropriate envelope interpretation looks for coherent energy at lower apparent velocities when waves approach more nearly along the array. If that energy is absent, the true dispersion might be poorly bounded. The consequence is not to force a slower curve into the image. It is to state uncertainty and consider a different array direction or complementary observation.
Why a second array orientation can help
Where site access permits, changing the array azimuth can alter the relationship between dominant ambient-wave directions and the receiver line. A feature that changes strongly with orientation may reflect source geometry or lateral heterogeneity rather than a stable one-dimensional earth response. Multiple orientations can help test that ambiguity.
A second direction is not a universal cure. If ground properties vary horizontally, the arrays may genuinely sample different structures. A scientific report distinguishes changed source incidence from changed ground location. The decision to combine results should be justified by geometry and observed coherence, not by the desire for one smooth average profile.
Quality of a spectral ratio needs explanation
ReMi processing is often described using a power-slowness or slowness–frequency spectral-ratio display. The detailed quantity depends on implementation, normalisation and windowing. A larger displayed ratio is evidence of a particular processing contrast, not directly greater soil stiffness or a larger shear-wave velocity.
Ask whether the report describes what the colour scale measures and how the frequency band was chosen. A black-and-white plot and a rainbow plot can encode exactly the same physical result. The reliable information is in the transform, slowness conversion and selected coherent feature, not the emotional impression made by bright colours.
Distinguish body waves, air waves and surface waves
Ambient seismic recordings may contain compressional body waves, acoustic waves in air, Rayleigh waves, higher surface-wave modes and unrelated instrumental signals. They do not all share the same dispersion physics. Some can appear as coherent trends across a receiver array and resemble the desired energy at first glance.
A reader should inspect the analyst’s method for distinguishing modes and rejecting irrelevant features. Comparing apparent velocities with plausible physical ranges and analysing coherence can help, but an arbitrary threshold is not sufficient proof. The original method’s value lies partly in exploiting multichannel behaviour to select the relevant surface-wave boundary amid competing signals.
Fundamental and higher modes require separate reasoning
The fundamental Rayleigh mode is commonly used for near-surface Vs inversion, but layered sites can show strong higher modes, especially where there are large velocity contrasts. Higher-mode energy can be physically real yet incompatible with a fundamental-only forward model. A clean inversion of the wrong branch can yield biased layer speeds.
A responsible interpretation tests whether the selected envelope corresponds to the mode used in modelling. In some cases multimode analysis can add information. The key is to avoid silently treating any coherent ridge as the same mode. A low inversion misfit cannot resolve a mistaken mode identification, because the misfit calculation assumes the model was posed correctly.
An inversion curve should be plotted beside its data
A final report is easier to assess when it presents picked dispersion points and the predicted curve from the selected Vs model. This comparison shows where the model fits well and where disagreement remains. It also reveals which frequency intervals were omitted or poorly measured.
One overall fit statistic is not enough. Large low-frequency errors may be especially consequential for deeper model inference, while high-frequency mismatches can indicate shallow layer or mode issues. Readers should examine both the shape and uncertainty of the picks. A smooth curve that passes through a few arbitrary points may convey more precision than the data warrant.
Borehole comparison is a test, not a conversion formula
A downhole or crosshole seismic investigation may provide independent shear-wave velocities at known depths. Comparing them with a ReMi inversion can help detect unrealistic layering or large biases. However, ReMi profiles are spatially averaged and modelled, while borehole measurements sample their own local wave paths.
The USGS blind ReMi and MASW comparison against borehole logs reported encouraging average velocity agreement for its studied sites, but it did not establish a universal accuracy percentage for every soil or array. Differences should prompt a location- and method-specific investigation rather than a blanket rejection of either dataset.
What the original ReMi paper actually established
The 2001 Louie study on refraction microtremor arrays demonstrated that ambient recordings on sufficiently long linear arrays could constrain useful shear-velocity information in its test settings. It discussed the slowness–frequency transformation and comparisons against other seismic observations. The reported depths and performance reflect those examples, not a default capability of all deployments.
When citing a method paper, separate a demonstrated case from the guarantees a service provider might later claim. The original experiment supports the feasibility and theory; current field suitability still depends on geometry, noise sources, processing and ground conditions. A citation is not a replacement for a new project’s quality assessment.
How ReMi differs from active MASW
An active MASW survey records surface waves from a controlled source and commonly constructs frequency–phase-velocity dispersion images through multichannel processing. ReMi principally exploits ambient waves and a slowness–frequency approach designed to handle uncertain source directions. Both may invert Rayleigh-wave dispersion to estimate Vs, but their source and picking problems differ.
Read the MASW dispersion and inversion guide for fundamental and higher modes, array aperture and model nonuniqueness. Calling the methods interchangeable just because both produce Vs profiles would hide an important source of uncertainty. Their agreement is useful corroboration only after those differences are considered.
Why neither passive nor active is automatically better
A controlled source may generate clean short-wavelength energy but struggle to excite longer wavelengths needed for some deeper targets. Ambient noise may provide useful longer-period waves yet have uncertain incidence and variable source distribution. The best method depends on the site’s wavefield, geometry and required depth or resolution.
A combined approach may strengthen interpretation by broadening frequency information, but merely merging curves can be dangerous if modes or apparent velocities were misidentified. Any joint inversion should state how data types were reconciled and weighted. An engineer chooses based on information quality, not a simplistic active-versus-passive ranking.
Vs30 still requires a valid depth model
A ReMi Vs profile may be used to calculate the travel-time-weighted average shear velocity in the upper thirty metres. For layers with thickness hi and shear-wave speed Vi, Vs30 equals 30 divided by the sum of hi/Vi. It is not a simple average of the velocities listed in a table.
Take an illustrative model with fifteen metres at 180 m/s and fifteen metres at 360 m/s. The travel time is 15/180 plus 15/360, or 0.125 second. Vs30 equals 30 divided by 0.125, giving 240 m/s. Whether the result is reliable depends on the resolution and completeness of the model across all thirty metres.
Extrapolated layers need clear labels
If the ambient data do not meaningfully constrain the full thirty-metre interval, the Vs30 calculation may require information from boreholes or a stated extrapolation. An analyst should indicate which intervals were independently constrained and which were assumed. A displayed line extending to thirty metres is not itself proof of a valid Vs30 estimate.
Seismic site classifications also depend on the relevant building code, edition and engineering context. A global educational article cannot supply one universal legal threshold table. It can show the averaging mathematics while directing professionals to the applicable design code and project-specific geotechnical evidence.
From an estimated Vs to Gmax
Under appropriate elastic assumptions, low-strain shear modulus Gmax equals bulk mass density multiplied by Vs squared. Suppose a model yields Vs = 240 m/s and a measured or justified mass density of 1,850 kg/m³. The corresponding Gmax is 1,850 times 240 squared, or 106,560,000 pascals, approximately 106.6 MPa.
This is a model-derived small-strain stiffness, not an independently measured static shear strength or a direct prediction of foundation settlement. The uncertainty in the inverted velocity and assumed density both carry into the modulus. A confident engineering report names those inputs rather than listing a precision that the wavefield cannot support.
A ten percent velocity error has a larger modulus effect
Because Gmax uses Vs squared, a ten percent upward bias in modelled Vs produces a 21 percent upward bias in Gmax if density is fixed. A ten percent lower Vs gives a modulus factor of 0.81. This nonlinear sensitivity makes modal and directional mistakes more consequential when the result is used to estimate stiffness.
The example also shows why reporting Gmax alone can conceal the source of error. A reviewer should first inspect the p–f envelope and inversion, then density, before assessing stiffness uncertainty. The derived property is only as defensible as the measurements and model from which it came.
ReMi cannot identify soil type from colour alone
A ReMi Vs profile may show faster or slower intervals, but shear-wave velocity is influenced by density, stress, cementation, saturation and material structure. Different soil types can exhibit overlapping velocities. A transition in modelled Vs does not necessarily coincide with an exact lithological boundary.
Borehole descriptions, grain-size tests, groundwater observations and other geotechnical data can help assign geology. The report should distinguish a seismic velocity contrast from an interpreted geologic contact. A colour boundary may be an interpolation or model parameter, not a sample recovered from that depth.
Do not use one ReMi profile as a liquefaction verdict
Shear-wave velocity can inform seismic site response and some liquefaction assessment frameworks. But liquefaction evaluation also depends on shaking demand, effective stress, groundwater, soil susceptibility and calibrated empirical or mechanistic relations. ReMi does not directly test those conditions merely by producing a Vs profile.
A low-velocity layer can deserve attention without proving failure during earthquake shaking. A high velocity alone does not guarantee that all construction performance criteria are met. The useful role of ReMi is to add a modelled elastic velocity constraint, which should be integrated with the broader site investigation rather than used as a stand-alone design approval.
Reading a p–f figure step by step
First find the frequency axis, slowness units and colour definition. Next identify array direction, receiver spacing and ambient recording conditions. Then locate coherent energy and the interpretation envelope, making sure the chosen branch is described in terms of apparent velocity, source direction and Rayleigh mode.
After that, compare the dispersion pick with its predicted curve and uncertainty. Finally inspect the Vs-versus-depth model, known geological information and any extrapolated layers. In this order the reader sees which steps were measured and which were inferred. Reversing the process—starting with the final ground profile—makes it easy to overlook why the result may be ambiguous.
A diagnostic example of false stiffness
Imagine a p–f spectrum dominated by waves arriving obliquely to the array. The main ridge corresponds to apparent speeds around 500 m/s even though a suitably directed wave would reveal 250 m/s. An analyst mistakenly selects the brightest ridge as the true fundamental phase speed and obtains a very stiff inverted ground profile.
The next step should be to revisit wave direction, lower-velocity envelope evidence and mode identification. If no reliable envelope exists, the result needs explicit uncertainty or additional measurements. Merely changing the borehole log to agree with the fast inversion would turn a possible wavefield artefact into a false geological narrative.
Common mistakes and corrective habits
Mistake one is to read slowness as velocity. Repair it by taking the reciprocal and checking units. Mistake two is interpreting the dominant apparent phase velocity as the true value without considering source direction. Repair it by assessing the coherent lower-velocity boundary and array azimuth. Mistake three is treating every high-power feature as the fundamental Rayleigh mode. Repair it through mode-aware interpretation.
Mistake four is calling the final Vs model a directly measured borehole profile. Repair it by naming the inversion assumptions. Mistake five is equating wavelength to unique investigation depth. Repair it with sensitivity analysis. Mistake six is reporting code site class or liquefaction status without appropriate project-specific checks. Repair it by separating the seismic measurement from design decisions.
Transfer exercise: three numbers, three meanings
A hypothetical p–f pick has slowness 4 seconds per kilometre at 10 Hz. The apparent phase velocity is 250 m/s, because four seconds per kilometre equals 0.004 second per metre. The wavelength associated with that component is 25 metres. Neither number identifies a layer boundary at 25 metres without an inverse model.
Now a layered inversion assigns Vs = 300 m/s to one interval. Does that contradict the 250 m/s phase velocity at the selected frequency? Not necessarily; a Rayleigh-wave observation depends on more than one layer and is not the same as a layer’s Vs. Finally, if density is 2,000 kg/m³, estimated Gmax for the 300 m/s layer is 180 MPa. This is small-strain stiffness, not ultimate shear strength.
Technical vocabulary and translation across disciplines
Slowness sounds like a general judgement of poor performance, yet here it has a precise mathematical meaning: reciprocal velocity. Apparent speed is a projection-dependent measurement, not necessarily the actual wave speed through the ground. An envelope is a picked boundary of spectral energy, not a literal container. Inversion is model fitting, not automatically the reversal of a physical process.
These distinctions matter when reports move between languages, engineering teams and non-specialist readers. The verified cross-linguistic transfer guide helps explain how familiar words can become false friends in technical contexts. A good explanation preserves equations, measurement units and operational definitions rather than merely replacing each English noun with a dictionary equivalent.
Frequently asked questions
Does ReMi require a large controlled seismic source? No; its defining use is ambient vibration. Does ‘refraction’ mean it uses first-arrival body-wave inversion? Not in its usual dispersion interpretation. Is slowness a velocity? No, it is reciprocal velocity. Can apparent speed exceed true speed? Yes, because of oblique wave incidence along an array. Does the most colourful ridge always give the right answer? No.
Can traffic help a survey? Sometimes, when it provides suitable coherent wave energy. Does more recording time guarantee deeper penetration? No. Can different Vs models fit the same dispersion? Yes. Does ReMi directly measure Gmax? No; modulus requires a velocity estimate and density under an elastic model. Is a calculated Vs30 independent of the chosen inversion? No. Can one profile establish liquefaction? No.
Original methodology and further study
The original Louie 2001 ReMi paper describes the method’s slowness–frequency transformation and inversion approach. A USGS-hosted blind ReMi/MASW comparison evaluates results against independent borehole velocities in a defined setting. A study of ReMi at a rock site illustrates how propagation direction and frequency-dependent uncertainty influence interpretation. These reports are evidence about methods and case studies, not universal performance guarantees.
For practical comparison, consult MASW dispersion and inversion, crosshole seismic velocities, downhole interval Vs and Vs30 and cross-linguistic meaning. Each link resolves a different interpretation problem.
The enduring lesson: the receiver line sees projections
ReMi is useful because it transforms ambient vibrations into information about Rayleigh-wave dispersion and possible subsurface shear-wave velocities. Its difficulty is that a linear array receives projections of waves from directions the survey may not control. The chosen slowness envelope is therefore a scientifically meaningful interpretation step, not a decorative part of the image.
A strong report makes the chain visible: array geometry, ambient wavefield, p–f representation, apparent velocities, selected dispersion, inversion and uncertainty. A reader who preserves that chain can use the inferred Vs profile wisely without turning a passive surface measurement into an imaginary direct record from deep inside the earth.