Horizontal-to-vertical spectral ratio, usually abbreviated HVSR or microtremor H/V, compares recorded horizontal and vertical ground-motion spectra to help characterise site resonance. If you are searching for how to read an HVSR curve, identify the fundamental frequency f0, judge peak reliability or relate an H/V peak to sediment thickness, begin with the actual recording components and frequency axis. An H/V peak is not a direct measurement of seismic amplification or a borehole depth.
The practical interpretation is narrower and more valuable: when conditions are suitable, a stable microtremor H/V peak may indicate a predominant resonance frequency associated with a subsurface velocity contrast. The curve also depends on the ambient wavefield, source direction, sensor installation, processing windows and local geology. A large peak is not automatically a large earthquake amplification factor, and a flat curve does not prove that the ground is safe from strong shaking.
This guide explains microtremor HVSR from three-component recording through Fourier spectra, horizontal-component combination, window selection, frequency smoothing, clear-peak assessment and constrained geological inference. It covers f0 versus period, simple quarter-wavelength approximations, independent shear-wave velocity information, multiple peaks, spatial variability and uncertainty. Worked measurements are expressly hypothetical so readers can check the arithmetic without mistaking them for an observed site.
Quick answer: what is an H/V spectral ratio?
A three-component seismometer records motions along two orthogonal horizontal directions and one vertical direction. For selected frequency windows, analysts estimate the Fourier amplitude or power characteristics of each component and calculate an H/V spectral ratio according to a declared processing convention. The result is a dimensionless curve against frequency, commonly in hertz.
A pronounced, repeatable peak can be useful evidence for a predominant site frequency in an environment where sediment overlies material with strongly different seismic impedance. The peak must be evaluated for robustness and possible non-geological disturbances. Reading the graph means understanding both its signal processing and the physical hypothesis that links resonance to subsurface layers.
Start with the axes before the geology
The horizontal axis usually displays frequency in hertz, sometimes on a logarithmic scale. The vertical axis gives a spectral ratio, often without units because the numerator and denominator represent the same kind of motion amplitude. A point at 2 Hz means two oscillation cycles per second; a ratio of four means the processed horizontal spectral quantity is four times the chosen vertical reference under that convention.
Do not call the ordinate metres per second, acceleration or an earthquake intensity scale. Nor should a number such as H/V equal to four be automatically translated into fourfold engineering amplification. The method’s physical relationship to amplification is more complicated, especially when noise wavefields have mixed compositions. State the measurement before proposing an interpretation.
Microtremor describes ambient motion, not an earthquake record
Microtremors are ambient vibrations produced by natural and human sources including ocean-generated waves, weather, transport and machinery. They can be measured without actively striking or vibrating the ground. Their wavefield may include different body-wave and surface-wave contributions, and its composition can vary by location and time.
A single-station HVSR method is attractive because it needs fewer instruments than an array survey, but this simplicity has boundaries. A source-dominated peak from machinery can mimic a stable frequency. Conversely, a genuine site resonance can be obscured when the wavefield does not excite it clearly. The reader should ask what sources were present during recording and whether the time windows represented ordinary conditions.
Understand the three components
Two horizontal records are often labelled north–south and east–west, though sensor orientation and site practice can differ. The vertical component records motion along the third axis. Researchers frequently combine the horizontal spectral estimates, for example using a geometric or quadratic mean, and divide by the vertical spectral estimate. The precise definition matters because it affects ratios.
If one report uses the square root of the sum of squared horizontal components divided by vertical, while another uses root-mean-square horizontal amplitude, the numerical ordinates can differ by a factor linked to normalisation. Both can be meaningful within a stated method. Before comparing peak amplitude between papers, inspect their horizontal-combination formulas and processing conventions.
A numerical horizontal-component example
Imagine that at 3 Hz a hypothetical processing window yields horizontal amplitude estimates H1 of 0.8 and H2 of 0.6 in consistent arbitrary spectral units, with vertical amplitude V equal to 0.25 in the same units. If horizontal amplitude is defined as the root mean square of the two horizontal squares, H equals the square root of (0.64 plus 0.36) divided by two, approximately 0.707.
The resulting ratio H/V is about 2.83. Under a root-sum-of-squares convention instead, horizontal amplitude would equal one and H/V would equal four. Neither calculation makes the other numerically wrong; they answer subtly different definitions. Comparing the curves without checking the convention can create a false disagreement about the magnitude of a spectral peak.
Spectral ratios depend on processing choices
A raw acceleration or velocity trace changes with time. Analysts generally select intervals, remove trends or means, apply tapers and compute spectral estimates. They may smooth frequency bins to reduce variance. Window length, overlap, filtering and treatment of short disturbances can alter both the appearance and repeatability of a peak.
A scientific interpretation should identify the algorithm, not merely show a final colourful graph. If an apparently perfect narrow spike appears only after aggressive smoothing or rejecting many windows, question its robustness. Equally, too little smoothing can leave a jagged curve that looks like several distinct resonances when random fluctuations are substantial. Processing choices affect what the reader sees, so they belong to the explanation.
Window length and low-frequency resolution
Resolving low frequencies requires enough recording duration. Very short windows cannot reliably characterise oscillations whose periods are long relative to the window. For example, a frequency of 0.2 Hz corresponds to a period of five seconds. A window containing only a few such cycles gives fragile spectral evidence even before considering source variability.
Longer windows can improve frequency resolution, but long segments may also contain changing noise sources. This creates a balance between enough cycles and a reasonably stationary wavefield. A report should state typical window length, total accepted duration and the frequency range actually supported by the recordings. A curve extending far beyond the reliable band is not proof of deep structure.
Smoothing can move attention without moving the site
HVSR curves are often smoothed, and the smoothing bandwidth can influence apparent peak width and height. A small irregular bump might become easier to see after smoothing, while two nearby narrow peaks might merge into one broad feature. The underlying soil has not changed; the graphical representation has.
When comparing surveys, inspect whether they used consistent smoothing methods. If a peak shifts markedly as reasonable processing choices change, its reported fundamental frequency requires a wider uncertainty interval. A stable estimate should survive sensible alternative choices rather than depend on one specially tuned plot. The point is not to reject smoothing but to recognise it as part of measurement interpretation.
What does the fundamental frequency f0 represent?
In favourable geological settings, the prominent HVSR frequency f0 approximates a fundamental resonance frequency associated with the sediment–bedrock velocity structure. The physical picture is a softer, slower material overlying a comparatively stiffer, faster substrate, creating wave interactions that can support resonant behaviour. The observed H/V spectrum does not reveal a unique profile by itself.
A distinct fundamental peak can guide further investigation, especially for site classification and estimating basin sediment thickness with suitable velocity information. The word fundamental must be supported rather than assigned automatically to the tallest spectral spike. Other peaks can arise from additional interfaces, wavefield features or anthropogenic sources.
Frequency and period are reciprocals
If a measured site-frequency estimate is f0 equal to 2 Hz, the corresponding period T0 is one divided by two seconds, or 0.5 seconds. If f0 equals 0.5 Hz, T0 equals two seconds. This relationship is exact because frequency is cycles per second and period is seconds per cycle, provided both describe the same oscillation.
A common reporting mistake is writing f0 equals two seconds or T0 equals 2 Hz. Those are different quantities. Another is assuming higher frequency means a deeper deposit. Under simple models, the opposite often follows for otherwise similar shear-wave velocity and impedance structure: thicker soft layers can have lower fundamental frequencies. But real profiles require more constraints than one trend.
Why the quarter-wavelength approximation is only a model
For a simplified uniform low-velocity layer of thickness H over a much stiffer half-space, a familiar estimate is f0 approximately equal to average shear-wave velocity divided by four times thickness. Rearranging yields H approximately equal to average shear-wave velocity divided by four f0. This is a simple one-dimensional resonance model, not a universal inversion formula.
The approximation assumes conditions that actual soils may violate: a strong impedance contrast, suitably simple layering and relevant wave behaviour. Multiple layers, lateral geology, basin geometry and nonuniform velocities can change the relationship. The formula is useful as a conceptual check and as one ingredient in a calibrated estimate; it should not be treated as direct measurement of bedrock depth from one peak.
Worked depth estimate with explicit uncertainty
Suppose a hypothetical site has a clear f0 near 2 Hz and independent evidence suggests an average Vs of 240 metres per second through the relevant sediment layer. The quarter-wavelength model gives H about 240 divided by eight, or 30 metres. This means an illustrative 30-metre estimate under the model, not an observed borehole contact.
If plausible average Vs ranges from 180 to 300 metres per second while f0 remains 2 Hz, the same model yields thickness estimates from 22.5 to 37.5 metres. The uncertainty in velocity propagates directly into the depth interpretation. Reporting ‘bedrock at exactly 30 m’ erases the strongest limitation of the calculation. Use conditional language and seek independent verification.
H/V amplitude is not necessarily earthquake amplification
The amplitude of a microtremor H/V peak reflects a combination of site response and wavefield composition. Earthquake site amplification concerns the response to seismic shaking under defined source, path and site conditions. Although HVSR can correlate with aspects of site resonance, its peak ratio is not a universal direct multiplier for earthquake ground-motion amplitudes.
The USGS review of MHVSR emphasises the difficulty of standardising physical interpretations because the ambient wavefield varies. The safe explanation distinguishes finding a predominant frequency from measuring absolute amplification. A site can have a useful H/V peak without the ratio establishing an engineering design factor.
A clear peak is a hypothesis that needs reliability checks
A peak should be prominent, repeat across accepted time windows and exceed the surrounding baseline in a way consistent with the method’s quality criteria. Its frequency should not jump erratically with small changes in processing or microphone-like noise conditions. Multiple processing settings, window groups or independent recordings can help evaluate repeatability.
Some guidance proposes numerical criteria for ‘clear peaks’ and ‘reliable curves’, but these should be applied with their original definitions and measurement context. Do not invent universal acceptance thresholds. Explain which document, spectral-band and sample duration were used. If criteria fail, report an uncertain or non-diagnostic peak rather than forcing an f0 value because a client expects one.
Why a flat H/V curve is not a safety certificate
A curve without a pronounced peak can arise in several situations, including weak subsurface impedance contrast, particular wavefield conditions, noise interference or complex layering. It does not establish the absence of soft deposits, liquefaction susceptibility or dangerous earthquake shaking. Those questions depend on further geotechnical and seismological evidence.
The correct interpretation is that the recording did not yield a robust characteristic peak under the stated processing and frequency range. It is an observation about the method’s diagnostic result. Claims about engineering safety require appropriate design codes, site investigations and hazards assessments beyond the scope of a single ambient-noise ratio.
Multiple peaks are not automatically multiple bedrock contacts
A microtremor curve may show two or more peaks. They can relate to different impedance contrasts, surface-wave modal effects, wavefield composition, local structures or transient machinery. Selecting the tallest one as the ‘true’ bedrock resonance without corroboration may be misleading.
Compare peak stability across windows and locations. If a 1-Hz peak is geographically coherent across a basin while a narrow 12-Hz spike occurs only near a pump, their plausible explanations differ. A second peak can be geologically significant, but its physical interpretation needs independent constraints. The number of peaks does not equal the number of strata.
Spatial surveys help distinguish local interference
A set of HVSR measurements along a transect can reveal frequencies that vary with geological setting. Repeated spatial patterns may support hypotheses about sediment thickness or lateral basin variation, especially when compared with boreholes or Vs profiles. An isolated outlier adjacent to a vibrating machine should be treated differently from a smooth trend reproduced by nearby stations.
However, a spatial map of f0 is not automatically a map of bedrock depth. Average shear-wave velocity and impedance can vary laterally alongside thickness. To infer depth, the method needs appropriate local velocity measurements or a justified regression developed from known contacts. Mapping one measured quantity and labelling it as another without calibration is an avoidable error.
Sensor coupling and orientation matter
A three-component instrument should be installed securely, levelled and oriented according to the acquisition protocol. Loose coupling, tilting, installation on an unstable surface or inconsistent orientation can alter the recorded components. The exact consequences vary with frequency and instrument design. A poor field setup cannot be repaired simply by applying stronger spectral smoothing later.
Field notes about sensor position, mounting, weather and nearby disturbance are therefore useful data. The report should explain whether the measurements were made on natural ground, pavement, a building floor or another surface. Responses from a structure can differ from free-field ground motion; the two should not be confused in a site-response conclusion.
Wind, traffic and machinery create specific risks
Wind can shake instruments or nearby vegetation and infrastructure. Traffic creates transient or persistent vibrations with directionality. Rotating machinery can produce narrow-frequency lines that look dramatic in an H/V spectrum. These sources are not merely nuisance noise; they can change the mixture of wave types reaching the sensor.
A good acquisition log records potential disturbances and repeats measurements if needed. Analysts should inspect component time series and spectrograms, not only an averaged curve. An HVSR peak that aligns with a stationary machine tone or disappears when traffic patterns change may not represent a stable geological resonance. The field context is part of the data, not decorative background.
The special challenge of two-dimensional basins
The quarter-wavelength picture is most straightforward for approximately horizontal layering. In a basin with steep edges or complex geometry, lateral wave propagation and resonance can affect observed frequencies and spectral amplitudes. A one-dimensional velocity model may fit some features while missing others.
A reliable geological interpretation examines whether the site sits near a basin edge, valley wall or other strong lateral boundary. Multiple recordings and complementary array data can help assess dimensionality. Describing a complicated two-dimensional response as a single uniform layer thickness gives the impression of precision without the underlying sensitivity.
The H/V method is not the same as ReMi
ReMi, or refraction microtremor, uses ambient vibrations recorded across a receiver array to estimate Rayleigh-wave dispersion from slowness–frequency information and then infer shear-wave velocity models. Single-station HVSR compares horizontal and vertical spectral content at one location. They address different parts of the site-characterisation problem and use different data.
The ReMi p–f and Vs guide explains why apparent slowness needs directional interpretation. An H/V peak should not be substituted for a complete ReMi dispersion curve. Their combination may constrain a site model, but each retains its own assumptions and limitations.
Compare HVSR with multichannel surface-wave analysis
Multichannel analysis of surface waves, or MASW, records wavefields at many receivers to extract frequency-dependent Rayleigh-wave phase velocities. The inferred dispersion curve provides constraints on shear-wave velocity structure, though inversion remains non-unique. HVSR instead describes a relative spectral property from three-component motion, commonly at one station. One method does not automatically generate the missing data of the other.
The MASW phase-velocity guide explains the relation between dispersion picks and inverted Vs profiles. A proposed subsurface model might be tested against both MASW dispersion and an HVSR peak. The agreement would strengthen a compatible geological interpretation, but only within each method’s resolution. A joint fit does not make an unmeasured contact directly observed.
Why an independent Vs profile is useful
A single HVSR f0 peak often cannot uniquely determine both average sediment shear-wave velocity and thickness. In the simple quarter-wavelength formula, raising velocity and thickness by the same factor leaves their ratio—and therefore the frequency—unchanged. Many different layer combinations can give similar resonance frequencies.
Independent Vs information from boreholes, downhole seismic measurements, crosshole tests, MASW or other carefully interpreted methods can reduce this trade-off. It does not eliminate every uncertainty: measured Vs may vary laterally and with depth, and wave types sample the ground differently. Nevertheless, combining compatible evidence is better than declaring a precise bedrock surface from one spectral peak.
Worked trade-off: two different models, one frequency
Consider model A with average Vs of 200 metres per second and layer thickness 25 metres. The simple relation gives f0 equal to 200 divided by 100, or 2 Hz. Model B with average Vs of 400 metres per second and thickness 50 metres also gives 400 divided by 200, or 2 Hz. Two different geological models therefore share the same simple resonance-frequency prediction.
An HVSR observation near 2 Hz cannot choose between them by itself. A downhole Vs profile, geological log or dispersion curve can help discriminate. This worked case demonstrates model non-uniqueness without relying on any invented field survey. A carefully written report should display the assumptions that allow a numerical answer, especially when turning frequency into metres.
Understanding frequency uncertainty
An HVSR peak might vary slightly among windows, nearby measurement points or reasonable smoothing choices. A responsible analysis examines that variation rather than picking a single number at excessive precision. For example, a hypothetical set of stable accepted estimates between 1.8 and 2.2 Hz may justify describing an approximately 2-Hz feature rather than asserting f0 equals 2.000 Hz.
The corresponding periods range from about 0.455 to 0.556 seconds. Under a fixed Vs assumption, the estimated thickness is inversely proportional to frequency. This means frequency uncertainty translates asymmetrically in some derived quantities. A report that provides only the central peak conceals this effect. Uncertainty should travel through subsequent calculations, not disappear when units change.
H/V curves need their processing history
Two analysts may record the same physical motion and produce curves with different amplitudes or peak widths because they select different time intervals, smooth the spectra differently or combine components using different conventions. This is not proof that one field site moved between analyses. It is a reason to preserve the processing settings alongside the reported result.
Useful reproducibility information includes sample rate, recording duration, accepted-window definition, spectral estimation, smoothing bandwidth, horizontal combination rule, peak-selection criteria and quality controls. Where the source files are available, readers can often test how robust the claimed f0 is. In technical education, explaining that trail is more useful than presenting a beautiful plot without its settings.
The peak can be broad for physical or analytical reasons
A broad spectral maximum may reflect a range of subsurface conditions, layered response, damping or mixed contributions to the ambient wavefield. It can also arise when smoothing or averaging combines several features. A broad peak is not necessarily poor science, but its central frequency may be less precise than a narrow and repeatable feature.
When describing a broad peak, report its relevant frequency range and the choices used to identify a preferred centre. Resist the temptation to call the left and right edges two distinct bedrock levels without supporting evidence. Consider whether nearby measurements show the same broad response and whether independent Vs constraints can explain it. The graph is evidence, not a geological cartoon.
How a narrow machinery line can masquerade as geology
Imagine an HVSR curve with a sharp 8-Hz spike that stays at almost exactly 8 Hz during a short recording. Such a line could be geologically meaningful, but it might also come from a rotating motor or other periodic source. Its extraordinary narrowness and source dependence deserve investigation, not automatic celebration as a high-quality resonance.
The analyst can examine raw time series, component spectra, spectrograms, time windows and repeat measurements at different periods. If the spike disappears when machinery stops or moves with a changing source, a geological interpretation weakens. If it persists across independent conditions and nearby positions consistent with geology, confidence can increase. Peak repeatability is important, but the kind of repeatability matters.
The direction of incoming waves may matter
HVSR depends on the composition of incoming vibrations. Strong waves travelling from one direction can alter the horizontal-to-vertical relationship, especially near two-dimensional structures or under highly anisotropic noise illumination. A single fixed orientation can therefore record a ratio influenced partly by directional sources and partly by the medium.
Rotating horizontal components or comparing directional estimates may help diagnose this sensitivity under suitable acquisition. The eduKate discussion of directionality and H/V wavefield interpretation gives the deeper geological case. The present guide’s main job is different: use the measured single-station curve to assess an f0 peak and its reliability, without confusing directionality with sediment thickness.
Site frequency and building frequency are different measurements
A soil site’s predominant resonance frequency is a property inferred from ground-motion observations under a model. A structure such as a building also has dynamic modes depending on its stiffness, mass and geometry. The two frequencies may interact in earthquake engineering questions, but a ground HVSR peak does not directly measure any building’s natural frequency.
Avoid comparing them without describing where instruments were placed and what was analysed. Recordings inside a building can be influenced by structural motion. Free-field soil measurements should be distinguished from measurements on floors or foundations. The practical value of HVSR is often identifying a site-related frequency that merits further investigation, not certifying a design or predicting damage alone.
Seismic hazard needs more than one peak
A seismic design question can involve expected shaking intensity, ground-motion spectra, local soil conditions, liquefaction potential, slope stability and structural response. A single H/V spectrum addresses only a piece of that information. It may provide a useful site-characterisation clue, but it cannot replace a geotechnical investigation or a code-compliant site response analysis.
When the intended use is safety-critical, keep the distinction explicit. A reader can use the graph to identify a plausible resonance band and formulate questions for qualified specialists. The graph should not be presented as a pass–fail hazard certificate. A narrow, reliable peak and a dangerous shaking scenario are not equivalent statements.
The difference between natural frequency and natural amplification
Frequency tells us where a response feature occurs along a time-scale axis. Amplification refers to how much some ground-motion quantity changes relative to a specified reference under defined excitation. These quantities have different units and require different measurement or modelling arrangements. A high frequency does not mean high amplitude, and a high H/V ordinate does not automatically supply absolute amplification.
A useful explanatory sentence might say, ‘The site shows a repeatable H/V peak around 2 Hz, consistent with a predominant resonance frequency under a strong-contrast model.’ It should not become, ‘Earthquake ground motion will be amplified by exactly five at 2 Hz,’ solely because an H/V ordinate happened to equal five. That leap erases the role of wavefield composition and reference motion.
How to build a cautious site-characterisation workflow
Begin with geology and the survey question. Record ambient three-component motion at suitable points with documented field conditions. Review raw traces for clipping, transient disturbances and obvious periodic machinery. Compute HVSR curves using a stated convention and evaluate their sensitivity to accepted windows and processing. Distinguish robust peaks from unstable features and artefacts.
Next compare frequency patterns spatially, inspect available borehole and Vs data, and test whether simplified resonance models make plausible predictions. Use complementary methods when thickness or subsurface layering matters. Finally, state measured f0, uncertainty, interpretation assumptions and unresolved alternatives. This sequence turns an H/V plot into defensible evidence rather than an unsupported subsurface picture.
Worked transfer: a different depth from the same frequency
Two hypothetical sites both show a reasonably clear peak at 1 Hz. Site A has a separately estimated average sediment Vs of 200 metres per second, while site B has 320 metres per second. Under the same simplified quarter-wavelength model, A gives a layer thickness near 50 metres and B gives 80 metres.
The numerical difference is caused by the independent Vs input. It does not prove either bedrock contact is truly at those depths; lateral structure and the assumed model may be wrong. The exercise illustrates why a frequency map cannot be converted to depth with one universal multiplier unless local conditions support that approximation. It also shows how a second measurement changes the useful interpretation of a first.
Worked transfer: when a peak disappears after window rejection
Imagine that an 11-Hz spike appears in the average curve only when three windows coinciding with a nearby machine’s operation are included. After those windows are removed, the rest of the record contains no similar peak. Is this a robust estimate of geological f0? No. It is evidence that the peak depends on a narrow disturbed interval.
A transparent report would state that the 11-Hz feature was associated with specific noise periods and was not retained as reliable site resonance. It could note whether any lower-frequency peaks remained. Rejecting the artefact is not failure; it is a positive quality-control finding. The reader learns more from an honest non-result than from a fabricated precise bedrock depth.
A note on local calibration curves
Some investigators estimate sediment thickness from empirical relationships between f0 and independently observed bedrock depths in a region. These can be useful where geology and shear-wave velocities are sufficiently consistent, but their coefficients should not be exported casually to unrelated basins. Local stratigraphy and changes in average velocity can shift the relationship.
Read the calibration dataset, the range of observed depths, the functional form and its uncertainty. A result extrapolated far outside that range can be weak even if the original regression had a strong fit. A regression predicts a target under its conditions; it does not convert a spectral peak into a measured depth. Distinguish a measured f0 from a calibrated thickness prediction in every table.
Peak quality is not a ranking of engineering safety
Quality criteria assess how reliably a curve supports a specified spectral interpretation. They do not determine whether a site is safe, suitable for foundations or compliant with a seismic code. A site can give a beautifully reproducible peak and still require extensive engineering investigation. Another site with no clear peak may nevertheless have serious hazards unrelated to HVSR.
This distinction should be carried into public communication. Use ‘well-resolved frequency peak’ rather than ‘good ground’ and ‘non-diagnostic H/V curve’ rather than ‘unsafe soil.’ Precise labels prevent a measurement-quality judgement from being misread as a land-use or structural design decision.
A quick checklist for reading a published curve
Identify the sensor location and whether the recording is free-field or structure-influenced. Read the horizontal frequency axis and confirm that the vertical quantity is a dimensionless ratio. Examine the combination formula for horizontal components. Find acquisition duration, window and smoothing choices, then ask whether the peak recurs across accepted windows and nearby sites.
Next identify any competing source explanations, the selected f0 and its uncertainty. If sediment thickness is estimated, locate the independent Vs evidence or local calibration. Check whether H/V amplitude was improperly equated to amplification. Only after these steps is it reasonable to summarise the geology. A reader following this order can detect many errors before they enter a report.
Common mistakes and how to repair them
A frequent mistake is treating an H/V amplitude of six as an absolute earthquake amplification factor. Repair it by describing the measured spectral ratio and seeking the appropriate site response analysis. Another is turning f0 into bedrock depth without Vs information. Repair it by stating a conditional model and reporting velocity uncertainty. A third is accepting a thin machinery line as soil resonance. Repair it through window inspection and repeat measurements.
Other mistakes include confusing period with frequency, comparing peaks computed with different horizontal-component conventions and treating a flat curve as proof of safe ground. Repair them by checking units, definitions and physical scope. A multi-peak plot should prompt a discussion of layers and wavefield effects, not an automatic count of geological contacts. These repairs preserve what the method genuinely teaches.
Frequently asked questions
What does HVSR stand for? Horizontal-to-vertical spectral ratio. Is f0 measured in hertz? Yes; its reciprocal is a period in seconds. Does an H/V peak show a precise bedrock depth? Not without independent constraints and a suitable model. Is the H/V ordinate a direct earthquake amplification factor? No. Does a flat curve prove that soil is safe? No.
Can machinery produce a false peak? Yes, under some recording conditions. Why use long recordings? They improve the chance of reliable low-frequency information, provided the signal can be treated appropriately. Does HVSR replace MASW or ReMi? No; they contain different measurements and assumptions. Can two different thicknesses produce the same f0? Yes, if velocity also differs. Is a high-quality spectral peak a hazard rating? No.
Original research and connected reading
The USGS review of the microtremor H/V method describes its physical foundations, acquisition challenges and uncertainty. The USGS application review explains the development of practical site characterisation. USGS research on estimating sediment thickness using H/V illustrates why local calibration methods can give different results.
To compare methods, see ReMi p–f plots and Vs profiles, MASW dispersion and Vs inversion, and cross-linguistic transfer of technical vocabulary. These are different reading routes, not claims that the underlying seismic methods are interchangeable.
Final lesson: identify a frequency before inventing a layer
Microtremor HVSR can provide a useful, inexpensive clue about site resonance, particularly in settings with a strong sediment-to-bedrock impedance contrast. The measured ratio must nevertheless be read through its recording, processing, wavefield and geological assumptions. A stable peak is evidence of a repeatable spectral feature, not an independently drilled boundary or a complete earthquake response model.
A good technical explanation moves from three-component recording to spectral ratios, from a robust f0 estimate to a clearly qualified physical interpretation, and from that interpretation to independent verification where needed. The distinction between measured frequency and inferred depth is what makes the method useful rather than misleading.