Borehole radar reflection logging is a method for reading echoes from fractures, contacts and cavities around a drilled hole, including features that never touch its wall. The essential interpretation is to separate what the instrument measures—wave travel time, amplitude, frequency and antenna position—from the geological hypothesis of where a reflector lies. A strong curved echo is not automatically a water-bearing fracture, and a weak echo is not automatically an absence of rock structure.
This guide explains how to read a single-hole borehole radar reflection section, convert a time pick into a defensible approximate distance, recognise the limits of omnidirectional antennas, and check a fracture interpretation against independent measurements. The governing first approximation is distance ≈ radar velocity × corrected two-way travel time ÷ 2. That simple expression is useful only after recognising the pulse’s origin, identifying a plausible reflection, choosing an appropriate propagation velocity and acknowledging geometry.
Unlike a surface ground-penetrating-radar profile, a single-hole instrument moves down a borehole and sends electromagnetic energy into surrounding rock. It can reveal a reflecting discontinuity beyond the borehole wall, whereas an optical or acoustic televiewer mainly describes the wall itself. The gain is a potentially larger investigation volume; the cost is interpretation ambiguity. A good report states both, then shows the checks that make its inferences reproducible.
The reader’s question: what does this radar picture actually demonstrate?
Imagine a vertical borehole in fractured crystalline rock. An unrolled optical wall image displays a few fractures that intersect the hole, but a geologist suspects another set extends several metres away. A borehole radar survey returns a sequence of traces, one at each antenna depth. On the section there is a repeated, sloping wavelet, a curved event and a strong band near the direct wave. Which one is geological? Which one is a reflection? Is the feature two metres or ten metres away, and does it connect to a pumping zone? These are different questions requiring different evidence.
Start with the data axes. In a conventional radargram, the vertical or horizontal display position of each trace may represent the radar antenna depth; the opposing axis represents elapsed time after the transmitted pulse. Software can reverse the axes, crop times, flatten gain, colour individual amplitudes, and interpolate between sampled positions. Before marking a feature, read the legend, polarity convention, depth datum, acquisition interval and whether time zero has been corrected. A photograph of the screen without these details is not enough for a distance calculation.
The term reflection is a measured wavefield event attributed to an electromagnetic contrast, not a direct picture of a crack. A lithologic boundary, water-filled fissure, cavity, conductive vein, casing boundary or isolated heterogeneity can reflect or scatter energy. The observable is a portion of the travelling electromagnetic pulse returning to the receiver. The subsurface category is a later inference. Keeping those two levels separate is the same discipline used in careful vocabulary translation: an apparent equivalent is not yet the same meaning in context.
First principles: why radio waves return from rock interfaces
A transmitting antenna launches an electromagnetic pulse with a finite frequency band. Its speed in a low-loss, nearly nonmagnetic geological medium depends mainly on relative dielectric permittivity, commonly written εr. For a simple approximation, velocity v ≈ c divided by the square root of εr, where c is about 0.300 metres per nanosecond in vacuum. If εr is 9, then v is about 0.100 metres per nanosecond. That is one tenth of a metre per nanosecond, or about 100 metres per microsecond, not 0.100 metres per second.
Why does permittivity matter? The electric field polarises minerals and pore fluids, delaying propagation relative to vacuum. Liquid water commonly has a much greater dielectric response than many dry rock-forming minerals at ordinary radar frequencies, so changes in saturation, porosity and fabric may change velocity and reflection strength. Electrical conductivity adds another influence: resistive rock may allow energy to travel far, while clay-rich, saline or otherwise conductive material can absorb or obscure signals. The medium therefore changes both the time of an echo and whether it survives to be recorded.
Reflection strength depends on a contrast in electromagnetic impedance across an interface, but field amplitude is also altered by geometric spreading, antenna patterns, coupling, scattering, absorption, transmission loss, processing gain and multiple reflections. A bright stripe can be an excellent reflector, an amplification choice or both. A weak stripe can signal small contrast, unfavourable geometry, destructive interference or loss. Interpret shape and continuity together with physical plausibility, rather than ranking fractures by screen brightness alone.
Frequency changes the trade-off. For an idealised 60-megahertz pulse and v = 0.120 metres per nanosecond, wavelength is v divided by frequency, about 2.0 metres. A rough quarter-wavelength scale is then 0.5 metres, but practical resolution also depends on pulse bandwidth, antenna response, noise and processing. Doubling nominal frequency does not automatically double trustworthy geological detail at the same range, because higher-frequency energy may attenuate more strongly. A field crew chooses antenna frequency around rock conductivity, target size, desired range and expected return strength.
What the tool records during a logging run
At each stop or continuous measurement interval, the instrument stores a received time series. Metadata should record transmitter and receiver separation, antenna types, centre depth, sampling interval, time window, acquisition gain, stack count and logging direction. In a real borehole the antenna is an extended object, not a mathematical point. The depths assigned to its centre, the relation of its transmitting and receiving parts, and tool eccentricity can alter the geometric conversion of echo time into radial offset.
A survey log may contain a first pulse that travelled directly from transmitter to receiver; after that appear ringing, wall interactions, internal tool responses and possible reflections from the rock. Some tools produce strong background bands at nearly constant delay on many adjacent traces. These can be fixed instrument artefacts or real laterally continuous responses. The key question is whether the event behaves with antenna depth as a plausible subsurface target would behave, and whether it remains coherent in unprocessed or differently processed displays.
Raw traces preserve polarity and timing information that aggressive processing may change. Background subtraction can remove horizontal instrumental bands but can also erase geological reflectors that remain nearly constant with depth. Time-varying gain can make late arrivals visible but distort relative amplitudes. Migration can place diffracted energy near a proposed reflector yet depends on the velocity field used. These operations do not invalidate the data; they require a record of settings and a comparison with an earlier, less processed stage.
Worked distance example: from corrected echo time to metres
Suppose an observed reflection arrives at 164 nanoseconds on a trace. The instrument’s calibrated system delay is 12 nanoseconds, so the corrected two-way reflection time is 164 − 12 = 152 nanoseconds. Assume an independently estimated effective radar velocity of 0.100 metres per nanosecond, and initially suppose the echo travelled outward and back along a nearly radial path. Estimated radial distance is 0.100 × 152 ÷ 2 = 7.60 metres. The numerator is metres because metres per nanosecond times nanoseconds cancels the time unit; dividing by two accounts for the outward and return legs.
This is a model-based distance, not the measured fracture aperture, not the depth along the borehole and not proof that a reflector is exactly 7.60 metres away. The echo might follow an oblique path, originate from a finite patch of a dipping plane, be refracted by variable dielectric properties or be a multiple. If velocity were actually 0.110 metres per nanosecond with the same corrected time, the estimate would be 8.36 metres; if 0.090, it would be 6.84 metres. An honest distance report carries that uncertainty, here roughly 6.8–8.4 metres for this isolated velocity range before geometry and picking errors.
A useful independent arithmetic check is to invert the calculation. For a target estimated at 7.60 metres and v = 0.100 metres per nanosecond, two-way time should be 2 × 7.60 ÷ 0.100 = 152 nanoseconds. Adding the 12-nanosecond system delay returns the displayed arrival of 164 nanoseconds. This reverse check catches a forgotten factor of two, a mixed microsecond/nanosecond conversion, or use of the uncorrected time. It cannot prove the target is geologically a fracture.
When the antenna’s motion turns one reflector into a curve
A single stationary discontinuity can be recorded at several antenna positions as the tool moves past it. For an ideal point reflector a perpendicular distance r from a straight borehole and a vertical offset z between antenna and reflector level give one-way path length √(r² + z²). Under a constant-velocity, collocated-transmitter/receiver simplification, two-way travel time t(z) = 2√(r² + z²)/v. The event is nearest in time when the antenna passes the reflector’s level and arrives later both above and below. Its plotted curve resembles a hyperbola.
For example, let the point be four metres from the borehole and v = 0.100 metres per nanosecond. At its elevation, the two-way time is 2 × 4 ÷ 0.100 = 80 nanoseconds. Three metres above it, path length is √(4² + 3²) = 5 metres; time is 100 nanoseconds. Six metres above it, path length is √(16 + 36) ≈ 7.21 metres and time is about 144 nanoseconds. The changing arrival time is geometrically meaningful. Yet a dipping plane can form a different coherent pattern, and tool offset or nonuniform speed modifies these simple curves.
Do not confuse a diffraction-like curve with a continuous planar reflector. Point-like heterogeneities often make compact curved events; an extended fracture plane can produce a sustained sloping reflection depending on incidence geometry. Several overlapping diffractors may create apparent planes, while intersecting planar features may break the idealised symmetry. Use a model with the actual borehole deviation and antenna layout before attributing dip or lateral continuity. If only one uncalibrated time slice exists, refuse precision that the acquisition geometry cannot support.
The azimuth problem: an echo is not a compass bearing
An omnidirectional receiving system is useful for detecting reflectors around a borehole but generally cannot determine their unique compass azimuth from a single undirectional trace. The same two-way time could correspond to a reflector north, south, east or west at a similar radial distance, subject to geometry. A directional antenna or sufficient additional geometry is needed to resolve orientation. Report a radial separation estimate without converting it into an invented map coordinate. A colourful radargram does not automatically contain azimuthal information.
Borehole inclination adds another layer. A survey depth is distance measured along the logged hole, while true vertical depth and geographic position differ in a deviated well. If the hole bends, a target’s relative path can change even when its geological surface is planar. A fracture orientation calculated from depth and apparent slope without applying a deviation survey may be systematically wrong. Check the tool orientation, measured-depth reference, magnetic corrections and any local coordinate conversion before drawing a map of predicted fracture traces.
Useful contextual evidence includes an optical televiewer image and an acoustic televiewer orientation log. Both primarily resolve structures at the wall; radar may follow features away from it. Correlation is valuable only if the matching depth, orientation and uncertainty are documented. A visually similar line at roughly the same depth is not enough to certify that two sensors observed the same fracture surface.
A repeatable picking protocol for real traces
Before asking what the echo means, build a reproducible pick table. For each selected event record borehole measured depth in metres, trace identifier, observed arrival in nanoseconds, the assumed time-zero correction, corrected arrival, amplitude or a clearly defined normalised measure, and a qualitative confidence flag. Record the window in which the pulse was identified, whether its positive or negative lobe was picked, and whether the same choice was made across adjacent traces. A reader should be able to reopen the trace and recover the pick to within the stated tolerance rather than reverse-engineer a marker from a presentation graphic.
Choose a consistent pulse feature. For a broadband pulse, the initial break, main positive maximum, main negative trough and envelope maximum can have different times. Tracking whichever lobe happens to look brightest changes the travel-time estimate. If the relevant reflection is strongly dispersive or overlaps another event, annotate it as uncertain. A transparent process is to compare neighbouring traces and require temporal continuity, sensible amplitude evolution and a physical geometry compatible with the feature. Do not claim sub-nanosecond precision merely because the sample grid supports it; repeatability matters more than display resolution.
A time-zero correction is not an arbitrary visual shift designed to make the result attractive. It can depend on instrument delays, cable lengths and transmitter–receiver configuration. Ideally it is measured by a calibration procedure whose date, geometry and recorded waveform are available. If 12 nanoseconds were removed in the worked example, say whether that correction arose from a measured system response or was adjusted using a reference reflector. A correction derived from the geological target itself cannot later be presented as independent evidence that the target’s distance was right.
The along-hole sampling interval also limits identification. Suppose the antenna stops every 0.25 metres, but a narrow echo changes substantially between stops. Interpolation may make its plotted curve look smooth, yet no measurements were acquired at the interpolated depths. Conversely, one metre spacing may miss an isolated steeply dipping fracture response. Report the actual sampling scheme, not the pixel density. If several traces were stacked to improve signal-to-noise ratio, give the stack count and explain whether motion during stacking could smear sharp features.
Building an honest uncertainty budget
An informative distance estimate separates time-pick uncertainty, time-zero uncertainty, velocity uncertainty and geometric-model uncertainty. Take corrected time t = 152 nanoseconds, estimated speed v = 0.100 metres per nanosecond, pick uncertainty ±4 nanoseconds, time-zero uncertainty ±2 nanoseconds and velocity uncertainty ±0.010 metres per nanosecond. With independent errors, time uncertainty is approximately √(4² + 2²) ≈ 4.47 nanoseconds. The corresponding distance term is (v/2) × 4.47 ≈ 0.224 metre. The speed term is (t/2) × 0.010 = 0.760 metre. Combining those two independent terms by root-sum-square gives about 0.79 metre; a geometry allowance remains additional, not magically covered.
This illustrative uncertainty calculation assumes that pick and calibration errors are independent, that velocity can be represented by one effective number, and that the radial two-leg path model is appropriate. In fractured or strongly heterogeneous rock none of those assumptions is guaranteed. If time zero and picked time share the same calibration error, covariance may matter. If uncertainty is dominated by a plausible alternative reflector interpretation, a statistical ±0.79 metre interval is misleading: report competing interpretations separately. A field report should not turn one nominal speed into an impression of survey-grade positioning.
For units, remember that a microsecond contains one thousand nanoseconds. A waveform arrival of 0.152 microsecond equals 152 nanoseconds, not 0.152 nanosecond. Multiplying 0.100 metres per nanosecond by 0.152 microsecond without first harmonising units creates a factor-of-one-thousand error. Translate every symbol into a quantity name before calculating: velocity is distance per time, t is two-way elapsed time, and the output is distance along the assumed path direction. This small habit prevents large field-report mistakes.
How attenuation changes what can be seen
Conductive pore fluids can make radar reflections disappear long before geometric range becomes limiting. An absence of returns at late time is therefore ambiguous. A saturated clay-rich interval may contain many fractures but yield a nearly blank section; an electrically resistive rock may permit clear echoes from farther away. The measurement has a detection threshold determined by source power, antenna coupling, wave spreading, attenuation, bandwidth, processing gain and background noise. The instrument cannot map features whose echoes never rise above that threshold.
A simplified amplitude model is A(r) = A0 times a geometrical factor times exp(−αr) for one-way propagation, with additional reflection coefficients and losses. For a reflection travelling outward and back, intrinsic attenuation acts along both legs. The coefficient α, measured in nepers per metre, is not itself a direct fracture concentration. If α were 0.05 nepers per metre for a homogeneous path, an idealised two-way 10-metre path would have an absorption factor exp(−0.05 × 20) ≈ 0.368 before spreading and reflection strength. Changing conductivity can therefore dominate what remains visible, even without a change in fracture geometry.
A receiver’s automatic gain or time-dependent gain may partly counter the falling amplitude. This improves visibility but destroys a simple interpretation of colour as reflectivity. Compare gain-corrected and less processed views, and ask whether a suspected feature is observable across several traces under multiple reasonable settings. A target that appears only after aggressive filtering may still be real, but its evidential weight differs from a coherent echo present in raw and minimally processed data. The processing history belongs in the interpretation, not in hidden file names.
US EPA’s borehole-radar reflection overview notes that useful range depends strongly on electrical conductivity and frequency; high-loss material can restrict penetration to very short distances, while resistive conditions permit much farther investigation. The correct reader lesson is not to copy a maximum range into every survey. Measure local signal loss and describe the portion of surrounding formation for which detection was actually credible. An echo at eight metres proves something was detected on that path, not that nothing exists at twelve metres elsewhere.
Fracture geometry versus groundwater function
A radar reflector can correspond to an open water-bearing feature, a mineral-filled or sealed fracture, a bedding contact, a cavity edge or another electromagnetic discontinuity. Groundwater transmissivity describes how readily water flows through a geological zone under a hydraulic gradient. Reflection amplitude does not measure transmissivity directly. A strong, continuous reflector could be tight; a narrow fracture producing a weak radar event could nevertheless carry substantial flow. This distinction is crucial for water-well design, remediation monitoring and any claim about aquifer connectivity.
Suppose a radar event projects toward a fracture intersecting a borehole at 48 metres measured depth. An ambient or pumped flowmeter profile shows flow entering at about the same depth. This makes the hypothesis of a hydraulically active structure more plausible, but several fracture segments may be close together. A packer test, pressure response, tracer breakthrough or independent cross-borehole evidence may still be required to establish continuity. Correlation strengthens a model; it does not turn two different measurements into the same observable.
Likewise, a visible wall aperture in a televiewer image is not a guaranteed hydraulic aperture. Rough walls, mineralisation, stress closure, clogging and partial intersection all affect flow. Use vocabulary carefully: fracture intersection is a geometrical observation, radar reflector is a wave phenomenon, transmissive interval is a hydrological interpretation, and connected flow path is a stronger claim requiring an appropriate transport or pressure test. Readers make better decisions when the language matches the strength of evidence.
A diagnostic sequence when an image is misleading
A useful failure check begins with a bright horizontal band repeated at nearly identical delay across almost the whole log. First compare the recorded direct wave and known cable/system artefacts; then view the raw data without automatic background removal. If the band remains constant through geological changes, a system-related origin becomes plausible. If it responds consistently to borehole position and is reproduced by another survey configuration, geological continuity remains an option. Never erase an inconvenient band until its origin has been investigated.
For a suspicious curved event, compare the curvature predicted by a point-scatterer travel-time curve using plausible velocities. If a fitted model requires a speed greater than vacuum or inconsistent with measured lithology, reject that fit, not physical reality. Reinspect timing units, depth scale, borehole geometry and overlapping arrivals. If several plausible models fit, document the ambiguity and what additional acquisition would distinguish them. For example a directional receiver, second borehole or different antenna frequency may reduce uncertainty more effectively than another round of colour-map adjustment.
If an image appears unusually sharp after migration, look at the migration velocity and whether it was independently calibrated. Incorrect velocity may move an event to the wrong depth or lateral offset, collapse a diffraction into an artificial point, or leave a genuine structure smeared. The right check is not whether one processed image is visually clean; it is whether the geological inference persists over a defensible range of velocities. A comparison panel or error envelope is more informative than a single immaculate picture.
The independent-check ladder
A first check tests basic arithmetic and measurement provenance: correct units, proper time-zero removal, meaningful signal-to-noise and repeatable picks. A second check tests geometric consistency: does the event’s slope or curvature agree with the proposed reflector position as the antenna moves? A third check tests independent sensors: does an imaging tool, core photograph, deviation survey or formation log support the rock feature? A fourth check asks the practical question: does hydraulic or geotechnical evidence support the role attributed to it? The checks should be labelled separately rather than collapsed into a blanket statement that results have been validated.
Consider two competing models for a coherent echo: model A is a steep fracture three to five metres from the hole, while model B is a lithologic interface oblique to the borehole. Both may reproduce one portion of a radargram. If an acoustic televiewer picks a steep fracture that trends away in a compatible direction, model A gains support. If spectral gamma, core and borehole density show a continuous lithologic contact at the projected depth, model B may gain support. The task is not to force the data to one story but to state which observations distinguish the stories.
A crosshole measurement is a particularly strong next step when the uncertainty concerns the location of a feature between wells. Single-hole radar asks where reflecting surfaces might lie around one borehole; a separate transmitter and receiver in two boreholes can constrain propagation through the space between them. But a crosshole inversion has its own sensitivity and resolution limits. Do not promise it will independently locate every fracture: the availability of ray paths, borehole spacing and electrical loss decide whether it can supply a meaningful check.
Vocabulary and translation: preserve the measurement while changing the words
Technical interpretation resembles translation across languages because a familiar word may hide a different quantity. In everyday speech depth can mean distance below ground, while in a log it may mean measured distance along a slanting well. Range can mean operational detection limit or the inferred distance to an individual target. Dip describes orientation of a plane, not simply the downward direction of a dark stripe on screen. Define the variable in context before substituting a more familiar term.
The eduKate guide to meaning, context and intent develops this broader habit: preserve the relation among evidence, claim and intended use rather than matching words alone. The semantic-features vocabulary guide gives readers a way to distinguish near neighbours such as detected, inferred, projected and confirmed. These distinctions matter when a technical report moves from a specialist to a public agency, drilling contractor, translator or student.
A second worked case: one reflector, three misleading distances
Suppose a feature is plotted at 118 nanoseconds after correction and speed is believed to be 0.120 metres per nanosecond. A radial-path estimate gives 0.120 × 118 ÷ 2 = 7.08 metres. Someone might report 14.16 metres by forgetting the return leg. Another might report 0.00708 metre by treating the nanosecond figure as though it were a microsecond figure. A third may use the 118 nanoseconds to estimate a borehole measured depth, even though the log depth is a separate survey coordinate. The same screen produces three wrong interpretations for three different reasons.
Now add a constant 10-nanosecond unrecognised delay. The true corrected arrival would be 108 nanoseconds, and the comparable radial estimate would be 6.48 metres, a 0.60-metre difference. This does not mean every 10-nanosecond uncertainty must be treated as a systematic calibration error; it shows why calibration metadata are part of evidence. Ask the reader to reproduce the conversion both ways and then discuss whether the reflector’s orientation, the speed model and antenna geometry could dominate even that corrected number.
Transfer test: apply the method without the author’s numbers
Exercise one: a radar reflection is picked at 210 nanoseconds, an independently measured system delay is 10 nanoseconds, and the effective velocity is 0.110 metres per nanosecond. Under the same simplified radial two-way-path assumption, estimate distance. Then give one reason the answer is not a mapped compass location. Exercise two: a reflector observed at the same antenna elevation has a corrected time of 160 nanoseconds and another at 120 nanoseconds. If both share velocity 0.100 metres per nanosecond, which is nearer? Exercise three: the field report writes a reflection range of 8.0 metres but omits antenna frequency, time-zero, velocity source and borehole deviation. What evidence is needed before relying on it?
Answers: in exercise one, corrected time is 210 − 10 = 200 nanoseconds and the radial distance is 0.110 × 200 ÷ 2 = 11.0 metres. Omnidirectional reception does not by itself identify a unique azimuth. In exercise two, distances are 8.0 metres and 6.0 metres respectively, so the 120-nanosecond reflector is nearer under the shared assumptions. In exercise three, request instrument and calibration metadata, the picked waveform, a documented velocity estimate, borehole trajectory and a justification for the geometry used. The missing fields are not bureaucratic trivia; each can alter the physical inference.
Exercise four introduces a more realistic choice. In one survey, the upper rock interval shows coherent reflections to long delays while the lower conductive interval is nearly blank. Does this prove the lower interval contains no fractures? No. The lower interval could attenuate radar energy strongly. A good answer proposes checking an independent conductivity or induction log, reassessing acquisition frequency and receiver sensitivity, and using wall imaging or hydraulic information. Absence of recorded echoes is a negative measurement only within the survey’s known ability to detect relevant targets.
Exercise five asks for a distinction in plain English. A report says, “The radar identified an open, hydraulically connected fracture 12 metres east of the borehole.” The survey used an omnidirectional single-hole antenna and had no independent flow test. Rewrite this as a defensible sentence. One possible answer is: “A coherent radar reflection is compatible with a subsurface discontinuity at an estimated radial distance near 12 metres under the assumed velocity; its azimuth, aperture and hydraulic connectivity remain unconfirmed.” The revised sentence is less dramatic but far more useful for designing the next field step.
What belongs in a competent reporting table
A reader should expect a table or downloadable data record containing the acquisition date, borehole identifier, datum, borehole diameter and material, antenna frequencies, transmitter–receiver arrangement, measured-depth range, step interval, time window, stack count, preprocessing steps, time-zero calibration, velocity model and picking convention. Each major interpreted event should include depth, picked corrected time, nominal distance, velocity assumption, uncertainty or plausible range, event class, alternative explanations and independent supporting or conflicting observations. If an item is unknown, write “not documented” rather than silently filling it with a typical value.
The interpretation map should distinguish borehole coordinates from inferred reflector locations. A single-hole undirectional radar result may be shown as an envelope or range around the borehole rather than as one falsely precise line on a plan. When directional data or additional holes refine the azimuth, document the geometry that justified narrowing the search region. If a site decision depends on whether a proposed excavation intersects a reflector, the report should make clear whether the reflector location is measured, modelled, bounded by evidence or merely suggested for follow-up investigation.
Raw data and processed data should remain traceable. Record instrument settings and processing software version when relevant, retain a minimally processed trace set, and keep independent logs in their original coordinate systems before co-registering them. A depth shift to align two pretty pictures can obscure a systematic datum problem. Compare like with like: measured depth against measured depth after documented corrections, two-way radar travel time against another radar quantity, and a hydraulic flow result against a hydraulic question. Technical credibility grows when a reviewer can trace a claim backward through these steps.
Boundaries and limitations that change the decision
The method works best when rock and fluid are sufficiently resistive for detectable propagation and when the desired structure has an electromagnetic contrast and size appropriate to the frequency band. Highly conductive conditions, excessive noise, borehole hardware, unfavourable antenna geometry and processing artefacts can shrink useful range. Radial echo time alone does not determine the three-dimensional coordinates of a reflector. A single fitted velocity may not describe stratified or strongly heterogeneous rock, and a reflection from a plane is not necessarily a point halfway along a straight ray.
A borehole radar reflection survey is therefore a way to narrow geological possibilities, not a substitute for geology, flow testing or ground truth. Choosing additional evidence should follow the risk: a drinking-water well might require proof of hydraulic pathways; a foundation investigation might require confirmation of cavity dimensions; a remediation study might require tracer or time-lapse tests; a research paper may need quantified inversion uncertainty and reproducible processing settings. The same radargram can support different decisions at different evidence thresholds, and those thresholds should be stated.
Even sophisticated migrated or inverted images are models rather than direct photographs. Regularisation, velocity assumptions, filtering, geometric coverage and noise all influence shape. A clear-looking boundary may be smoothed, misplaced or split. When the estimated reflector distance changes appreciably across plausible parameters, disclose the alternative locations. The reader does not need to distrust the technology; the reader needs to know the conditions under which its output can be trusted for a particular action.
Questions readers commonly ask
Can single-hole radar find a fracture that does not intersect the borehole? Yes, if the fracture or associated contrast produces a detectable return within useful range. That is a central advantage over wall-only optical imaging. Detection does not automatically reveal a unique compass direction, prove the fracture is open or show that groundwater moves through it. The radar image should be interpreted together with orientation logs, geological evidence and hydraulic measurements.
Why is distance half the velocity times travel time? The common reflection approximation counts a pulse travelling outward to a reflector and returning to the receiver. Multiplying the elapsed two-way time by velocity estimates the total path, and division by two estimates one leg. This shorthand assumes a simple geometry and effective speed; unequal transmitter–receiver positions, oblique reflectors and heterogeneous rock may require ray tracing or a calibrated forward model instead.
Can I compare amplitudes from two different logging runs? Only after checking antenna geometry, frequency, coupling, stacking, receiver gain, time-varying gain, depth registration, background processing and acquisition conditions. A louder echo may reflect a changed water or conductivity condition, a changed instrumental response or both. Time-lapse analysis becomes stronger when acquisition conditions and the calibration strategy make differences attributable to formation changes within known limits.
How can I tell a reflector from a processing artefact? There is no infallible one-pixel test. Look for a coherent event on adjacent raw or lightly processed traces, consistent travel-time variation with antenna position, response under alternative defensible processing choices, and correspondence with independent geology. Mark uncertain events explicitly. If the feature disappears when plausible settings change, do not quietly promote it into a confirmed mapped fracture.
Should I select the highest-frequency tool available? Not automatically. Higher frequency can improve ideal resolution but may sacrifice penetration through conductive or scattering material. The useful choice balances expected target size, investigation range, fluid conductivity, lithology, noise, hole geometry and equipment constraints. A low-frequency record that reliably detects a broad fracture zone may answer the practical question better than a finer image that sees only the first few decimetres.
Final method: claim only what the evidence carries
Read the coordinates and acquisition settings; locate the raw wavelet; apply a defensible time-zero correction; estimate velocity from a stated source; calculate an approximate distance with units and a return-path factor; test neighbouring traces for coherent geometry; check attenuation and processing sensitivity; examine orientation ambiguity; compare with independent geological and hydraulic evidence; and write a conclusion matched to the reliability of those steps. That is a usable method for translating radar traces into defensible geological language without pretending the subsurface has been photographed.
Primary methodological reading and connected eduKate guides
The US Environmental Protection Agency provides a detailed Borehole-Radar Reflection Logging technical reference on propagation, resolution, penetration and interpretation. The U.S. Geological Survey’s Borehole-radar methods for characterisation of fractured rock explains the role of single-hole and crosshole surveys. The USGS discussion of fracture and cavity mapping with borehole radar offers independent context about acquisition modes and limitations. These are methodological sources rather than guarantees for any particular site’s interpretation.
For related skills, compare the surface GPR radargram guide, the formation microresistivity imaging guide, and eduKate’s context-stack guide to preserving meaning. The transferable habit is to distinguish an observed quantity from an inferred mechanism and to use words that respect that distinction.