Crosshole radar tomography is a way to interpret electromagnetic measurements sent from one borehole and received in another. Instead of reading echoes that return to a single hole, the analyst reads first-arrival travel times and signal losses along many interwell paths, then estimates where radar waves travelled faster, slower, more strongly or more weakly. The resulting velocity or attenuation map is an inversion result, not a colour photograph of water or fractures.
To read a crosshole radar tomogram reliably, first verify transmitter and receiver coordinates and timing, calculate apparent path velocity from measured travel time, check attenuation against calibrated amplitudes and geometric spreading, then examine ray coverage, inversion smoothing and independent field evidence. A low-velocity zone may reflect higher effective permittivity and water content under a reasonable rock model, but it does not uniquely identify a flowing fracture. Likewise, increased attenuation can be consistent with more conductive fluid without proving a precise tracer concentration.
This guide explains the first-principles physics, the key axes and units, reproducible numerical examples, time-lapse saline-tracer interpretation, common mistakes and independent verification. Its specific reader question is: what do two-borehole radar travel-time and attenuation images legitimately say about the material and possible groundwater pathways between the holes? It complements, rather than duplicates, a single-hole reflection interpretation, which primarily asks where reflecting surfaces may lie around one borehole.
Understand the acquisition geometry before the colours
Imagine two accessible boreholes twenty metres apart. A radar transmitter is positioned at selected depths in borehole A, while a receiving antenna is positioned at selected depths in borehole B. The operators repeat measurements for many transmitter–receiver depth pairs. Each pair establishes one propagation geometry and one waveform, sometimes called a ray or path measurement. A simple presentation plots travel time and maximum received amplitude against transmitter depth and receiver depth. A tomographic image is then obtained by solving a separate inverse problem that combines the many observations.
A time-depth record and an inversion map are different products. In an original travel-time matrix, one axis might be transmitting-antenna depth and the other receiving-antenna depth, with a cell representing a first-arrival pick. A velocity tomogram, however, uses horizontal distance between boreholes on one axis and depth on the other. If the figure includes a colour bar labelled metres per nanosecond, it estimates speed; if labelled nepers per metre, it estimates attenuation; if labelled a dimensionless residual, it may display something else entirely. Read the legend and coordinates before interpreting a red or blue area.
The survey also needs deviation and collar data. Boreholes that look vertical on a schematic may incline, bend or have different elevations. A twenty-metre plan separation is not necessarily the length of every transmission path. If the transmitter is at forty metres measured depth and the receiver at fifty metres in parallel vertical holes twenty metres apart, the straight-line path length is √(20² + 10²) ≈ 22.36 metres, not twenty metres. The difference changes apparent velocity and therefore any permittivity inference.
Instrument configuration matters. Antenna centre depth may differ from cable-marked depth, time zero may include cables and electronics, and receiver coupling can vary with the borehole fluid and completion. A credible dataset records antenna frequency, bandwidth, depth positions, waveform sampling, source power, stack count, gain, time reference, well trajectory and logging sequence. A reader should not infer geological variability from a figure whose acquisition geometry or calibration is undocumented.
Physics: what the transmitted pulse is testing
In a low-loss approximation, radar velocity v is close to c divided by the square root of relative permittivity εr, with c ≈ 0.300 metres per nanosecond. Permittivity represents the medium’s electric polarisation response. Mineral composition, pore space, water saturation and temperature can affect the bulk value, so a propagation-speed difference may contain valuable geological information. But the relationship between water content and radar speed needs a petrophysical model and relevant calibration, especially in heterogeneous rock.
Attenuation describes reduction of signal amplitude as the pulse propagates. Intrinsic electromagnetic energy losses tend to increase with electrical conductivity and depend on frequency. Observed amplitude also changes through spreading, antenna radiation, coupling, waveform interference, scattering and receiver gain. An attenuation tomogram is therefore not a direct conductivity meter unless the processing and calibration address those other influences. A low-amplitude observation is a starting quantity; conductivity is a conditional interpretation.
Why do radar waves respond to saline tracer? Dissolved ions raise electrical conductivity in the water-filled portions of rock, so a sufficiently concentrated tracer can make transmitted radar signals weaker along paths interacting with affected fractures. The U.S. Geological Survey has studied crosshole attenuation-difference tomography using controlled saline tracer injections. The useful inference is the spatial and temporal pattern of electromagnetic change under known injection conditions. It is not a claim that the tomogram independently measures the exact chemistry of every pore.
The survey might use tens or hundreds of megahertz, with usable frequency chosen around investigation scale, electrical loss, borehole spacing and target size. High-frequency energy can give finer detail where the rock is resistive enough, but can fail to travel between holes in conductive conditions. A low-frequency configuration may provide more penetration with reduced spatial resolution. The acquisition design must be matched to the expected path losses across the full section rather than chosen on nominal resolution alone.
Direct arrivals versus the inverse model
A first arrival is the earliest defensible transmitted electromagnetic event at the receiver, not necessarily the waveform’s largest peak. In many conventional analyses its time is used to infer effective slowness along a propagation path. Slowness is the reciprocal of velocity: if speed is 0.100 metres per nanosecond, slowness is 10 nanoseconds per metre. That unit is particularly convenient because, for a straight ray in a homogeneous medium, travel time equals slowness multiplied by distance.
Each first-arrival time samples a path through the rock, not a single point. In the simplest model t = ∫s(x,z) dl along the ray, where s(x,z) is spatially varying slowness and dl is a small element of path length. An inversion looks for a slowness distribution that predicts the observed travel times within their uncertainty. Because many distributions can reproduce limited data, the analyst also imposes constraints such as smoothness, bounds or geological priors. Those choices help stabilise the calculation but also influence the image.
Attenuation can be handled in an analogous path-integral model after addressing source amplitude, coupling and geometric spreading. A cell on a velocity tomogram and a cell on an attenuation tomogram are not two direct measurements of that cell; they are estimates informed by multiple integrated paths. Their resolutions may differ because first-arrival picks are often more repeatable than absolute amplitudes. A sharply bounded red polygon in a report may be a display classification imposed on a smooth inversion rather than a geologically sharp boundary.
Worked first-arrival velocity example
Suppose the antenna centres are at the same elevation in vertical holes separated by 20.0 metres. The measured first-arrival time is 212 nanoseconds, and independently calibrated system delay is 12 nanoseconds. The corrected travel time is 200 nanoseconds. Under a straight-line path approximation, apparent radar velocity is distance divided by time: 20.0 metres ÷ 200 nanoseconds = 0.100 metres per nanosecond. Converting consistently, that is 100 metres per microsecond. The corresponding slowness is 10 nanoseconds per metre.
If the medium is approximately nonmagnetic and low loss, an effective relative permittivity estimate is (0.300 ÷ 0.100)² = 9.0. This is not automatically the volumetric water content and not a rock identity. Different mixtures of minerals, pores, water and temperature may produce similar effective permittivity. A confident water-content estimate would require an appropriate mixing law and independent information about mineral permittivity, porosity and saturation. The measurement supports a speed; the petrophysical interpretation adds assumptions.
Reverse the velocity calculation to check arithmetic. At 0.100 metres per nanosecond, a 20.0-metre path should take 200 nanoseconds after system calibration, and the raw 12-nanosecond delay restores 212 nanoseconds. If a report instead divides 20 metres by 212 nanoseconds without calibration, the estimated speed is about 0.0943 metres per nanosecond, more than five per cent lower. That could be misread as a geological change. An uncorrected instrument delay is therefore capable of inventing a physical anomaly.
Now consider a transmitter ten metres deeper than a receiver in the same pair of approximately vertical holes. The straight-line path length becomes √(20² + 10²) ≈ 22.36 metres. If the same medium speed of 0.100 metres per nanosecond applies, the corrected time should be about 223.6 nanoseconds. Dividing this time by the horizontal separation alone would give 0.0894 metres per nanosecond, an apparent low-speed zone caused by geometry. Raypath length and measured depth must be carried into the calculation.
What velocity colours do and do not mean
A velocity map often uses warm and cool colours to distinguish estimated fast and slow regions, but every palette is arbitrary. An area at 0.085 metres per nanosecond in a low-loss model has effective εr ≈ (0.300/0.085)² ≈ 12.5. A nearby region at 0.110 metres per nanosecond has εr ≈ 7.44. That contrast is compatible with changes in saturation, mineralogy, pores or lithology. It does not isolate which mechanism caused the difference or whether a fracture is open to flow.
An even bigger conceptual trap is to interpret velocity as water movement speed. The radar pulse might propagate at tens to hundreds of millions of metres per second, whereas groundwater movement is governed by pore geometry and hydraulic gradients at vastly smaller scales. A crosshole radar velocity map estimates electromagnetic propagation, not Darcy flux, seepage velocity or tracer migration velocity. Any translation from one to the other requires a separate hydrogeological model or time-lapse transport measurement.
When a report says a “low-velocity anomaly”, ask low relative to what baseline: a homogeneous starting model, adjacent cells, a depth-averaged reference, or a previous survey? A 0.010-metre-per-nanosecond change can be meaningful in one setup and within uncertainty in another. Provide absolute values, uncertainty, coverage and the reference model. A coloured anomaly without a measurable denominator or baseline is weaker evidence than the reader may suppose.
Why ray coverage shapes apparent detail
Crosshole rays tend to criss-cross the region between boreholes, but not every cell is equally sampled. Cells near well intersections may receive many paths with varied crossing angles, while sections outside the transmitter and receiver depth ranges can be almost unconstrained. An inversion may still fill the full rectangular image, because a regularisation rule supplies values where data do not. The coloured rectangle is not necessarily a rectangle of equal information.
A useful diagnostic is a ray-coverage plot, a sensitivity map or a point-spread/resolution assessment. If a purported thin fracture lies in a region where rays are nearly parallel, its apparent thickness or orientation may be dominated by smearing. If the analyst uses strong smoothing to stabilise the inverse problem, two real narrow zones can merge into one broad feature. With weak smoothing, noise can become patchy bright cells. Report how the selected regularisation strength affects conclusions, not just an impressive final colour panel.
Geometric symmetry can also hide ambiguity. A central low-speed patch in a smoothly regularised image may represent a real concentrated zone, a broader distributed property, or a compromise between several structures whose path integrals are similar. Measurements along additional depths or boreholes, and geological priors grounded in independent logs, can distinguish possibilities. Where they are unavailable, describe the image as a smooth effective-property model instead of claiming that individual pixels delineate fracture boundaries.
Separating field acquisition quality from geological inference
A first-arrival picker should use a consistent phase or onset convention and inspect neighbouring traces to avoid jumping between pulses. At low signal-to-noise ratios the earliest arrival may be hidden, so the detected onset occurs too late and creates a spurious low apparent velocity. Changing gain can expose a different pulse, while filtering may shift arrival time. Flag weak traces and perform an uncertainty-aware inversion instead of quietly treating all picks as equally precise.
A field quality check compares reciprocal or repeat measurements where practical. If transmitter and receiver roles can be exchanged with comparable equipment, travel times through the same physical path should be compatible once instrumental differences are calibrated. Similarly, a subset remeasured before and after logging can reveal clock drift, depth-position error or coupling changes. Such checks should be designed rather than retrospectively inferred from the finished inversion image.
A second check compares direct-wave amplitudes and baseline noise. Strong decay as transmitter–receiver distance increases is expected even in uniform rock because geometric spreading and attenuation change amplitude. An apparent high-attenuation area may be a zone of longer average paths or problematic coupling. Test whether the processing accounts for distance dependence, transmitter strength, receiver sensitivity and time-varying gain. Only then translate amplitude residuals into a geological attenuation claim.
Attenuation is a logarithmic path measurement, not a fracture percentage
In a simplified example, suppose a 20.0-metre path has been corrected for antenna calibration and geometric spreading, and its normalised amplitude falls from 1.00 to 0.25. The ratio is four to one. Logarithmic loss is ln(1.00/0.25) = ln(4) ≈ 1.386. Dividing that by 20.0 metres gives a path-average attenuation coefficient of approximately 0.0693 nepers per metre. It does not mean that 75 per cent of the path contains water, or that one quarter of the transmitted energy reached a particular fracture. The quantity describes attenuation under the stated model and correction assumptions.
In amplitude decibels the same ratio gives 20 log10(1.00/0.25) ≈ 12.04 decibels across the full path, equivalent to 0.602 decibels per metre. One neper corresponds to about 8.686 amplitude decibels, which provides an independent conversion check: 0.0693 × 8.686 ≈ 0.602. The twenty multiplier is for amplitude ratios under consistent impedance conditions; power ratios use ten. A report using the wrong logarithmic convention can produce a factor-of-two discrepancy while every graph still looks professional.
The calculation is deliberately idealised. Real crosshole amplitude changes with transmitter power, antenna radiation pattern, cable and receiver response, coupling to borehole fluid, geometric spreading, scattering, processing gain and instrument stability. If those influences have not been accounted for, the computed coefficient is an apparent loss statistic, not a trustworthy intrinsic formation property. A field reader should ask what was corrected and what remains mixed into the map, then use caution when converting attenuation to electrical conductivity or fluid salinity.
An attenuation map is built from many path-integrated logarithmic losses, not from a sensor stationed inside each image cell. For each measured path the inversion distributes loss among the cells that path crosses, subject to assumptions and regularisation. A zone of high reconstructed attenuation may be consistent with a conductive fracture, clay-rich rock or another loss mechanism; the map cannot uniquely identify the cause. When several cell patterns predict essentially the same path observations, a precise geological boundary would be overinterpretation.
Time-lapse tracer images: what changed, compared with what?
Time-lapse crosshole radar surveys compare repeated observations under deliberately documented conditions. A baseline run establishes travel times and amplitudes before a monitored hydrological change. A later run samples the same nominal antenna positions, sometimes during an authorised saline-tracer test. The analyst examines the change in travel time or logarithmic amplitude rather than interpreting every absolute value separately. Common-mode calibration and geometric effects may partially cancel, but cancellation is imperfect if temperature, tool coupling, source power or hole conditions change.
For an illustrative path, suppose a normalised baseline amplitude is 0.80 and a comparable later amplitude is 0.64. The ratio later/baseline is 0.80. The added logarithmic loss is −ln(0.80) ≈ 0.223. Across a 20.0-metre path, if one assumes uniform incremental attenuation, this corresponds to about 0.0112 nepers per metre. These numbers quantify a measurable electromagnetic difference. They do not give a universal sodium-chloride concentration, a groundwater flow rate or the total amount of tracer retained in the rock.
A concentration estimate needs a defensible petrophysical relationship among fluid electrical conductivity, temperature, salinity, porosity, saturation and bulk electromagnetic attenuation. Even with that relationship, a path integrates over heterogeneous rock and the tracer may bypass much of the imaged volume. A time-lapse attenuation anomaly can help locate where conductive-fluid changes affected propagation while leaving actual concentrations uncertain. Separate chemistry samples, hydraulic tests and transport models can sharpen interpretation, provided their sampling and assumptions are stated.
The USGS Mirror Lake research on saline-tracer attenuation-difference tomography shows the value of comparing baseline and tracer-condition surveys for understanding transmissive zones in fractured rock. It also highlights secondary porosity, scattering and calibration limits. This is methodological evidence, not advice for unqualified readers to inject salt into an aquifer. Field tracer experiments require scientific design, protection of water resources, lawful permissions and monitoring appropriate to the site’s sensitivity.
Travel-time differences require a separate interpretation
Suppose the corrected travel time of a 20-metre path changes from 200 to 215 nanoseconds between comparable surveys. Baseline effective slowness is 200 ÷ 20 = 10 nanoseconds per metre, while the later slowness is 10.75 nanoseconds per metre. Effective velocities are 0.100 and about 0.0930 metres per nanosecond. Under a low-loss model their corresponding relative permittivities are 9.0 and about 10.4. The result suggests a change in propagation properties if calibration and geometry remained stable, but cannot tell which physical cause changed them.
Do not interpret fifteen nanoseconds of additional electromagnetic travel time as fifteen nanoseconds of fluid travel, or convert the difference into a groundwater speed. Water flux and radar wave speed describe utterly different transport processes. A slower radar pulse may indicate a different dielectric environment, not water creeping through a fracture. If the same time shift appears on nearly every path, examine clock offsets or time-zero drift before invoking a geological transformation that uniformly affects the entire panel.
A good analysis shows residuals as well as final colours. For every transmitter–receiver pair, compute observed corrected travel time minus forward-model prediction. Plot residuals by depth pair and, where useful, against path length. Large systematic trends may indicate geometry or calibration errors. If a proposed anomaly improves a few paths but makes many others worse, its value is questionable. A model is judged by its ability to explain a diverse set of observations within known errors, not merely by producing a plausible contour.
Inversion: how the field data become a tomogram
In a straight-ray discretisation, divide the interwell area into cells, calculate how far each propagation path travels within each cell, and represent total travel time by the sum of cell slowness multiplied by corresponding path length. Written compactly, t ≈ Gs. The vector t contains measured corrected times, G describes ray lengths through cells, and s contains the unknown slownesses. This equation is a model of integrated travel time, not evidence that radar has directly sampled each individual cell.
For attenuation, replace travel-time observations with suitably calibrated logarithmic losses and estimate cell attenuation values using analogous path geometry. Actual inversion software may incorporate more complex forward physics, bending rays, constraints and nonlinear updates. Those improvements can be appropriate, yet they do not remove the need to document assumptions. A solver that converges has found a model consistent with its chosen objective and parameterisation, not necessarily the unique geological truth.
The inverse problem is usually nonunique or ill conditioned. Limited angular coverage can allow several cell-value patterns to predict nearly identical travel times. Smoothness regularisation discourages unstable variations, but may blur a thin real fracture into a wide patch. Weak regularisation can turn noise into apparent small-scale structure. A reader should ask how the penalty strength was chosen, how solutions vary over reasonable settings, and whether an independent constraint fixes a feature’s location or geometry.
A low-velocity band in a tomogram might therefore be a smeared representation of one sharp contact, several narrow fractures or a gradual lithological transition. The displayed colour boundaries may be interpolation and contour choices, not measured geological edges. Confidence should be lowest where ray paths are sparse, poorly crossing or confined to one direction. A useful report includes ray coverage, uncertainty, model resolution and a statement of what spatial scale can reasonably be distinguished.
Quantifying simple uncertainty rather than saying “accurate”
If a first-arrival pick on a 20.0-metre path is uncertain by ±3 nanoseconds, the corresponding simple slowness uncertainty from that pick is about ±3 ÷ 20 = ±0.15 nanoseconds per metre. Relative to a 200-nanosecond corrected time, this is roughly 1.5 per cent. Velocity uncertainty from timing alone is similarly around 1.5 per cent for small perturbations. But the nominal calculation ignores borehole position error, common-mode time-zero shifts, inaccurate path geometry and alternative first-arrival interpretations.
Suppose a hole trajectory survey places one antenna laterally 0.5 metres farther from the other than assumed, so the true path is 20.5 rather than 20.0 metres. With the same 200-nanosecond time, inferred speed changes from 0.100 to 0.1025 metres per nanosecond. That 2.5 per cent effect is larger than a one-nanosecond pick error. Underwater, deviated or irregular holes may require careful positional control. The magnitude of the geometric mistake depends on the actual configuration, so describe both nominal and plausible path length.
An independent reviewer should test robust picks by re-picking a selected set of waveforms without seeing the original interpretation. Compare the onset convention and the variability of picks across trace quality. If one analyst picks the first weak pulse and another picks the later strong pulse, averaging the two times is not an adequate repair. Investigate which event is physically credible and flag unreliable paths. An inversion that treats every timestamp as exact conceals this uncertainty.
Repeat surveys can check stability: acquire some depth pairs again, compare time and amplitude differences, and record environmental or acquisition changes. Reciprocal propagation measurements may provide another useful check when tool configuration permits comparable transmitter–receiver paths. If a large time-lapse anomaly is smaller than the measured survey-to-survey variability, it should not be called a detected tracer front. If it persists across repeats, calibrated gains and alternative inversions, the evidence becomes stronger.
Verify the geological interpretation with other measurements
Velocity and attenuation respond to multiple rock and fluid properties. A joint low-velocity, high-attenuation zone may be consistent with increased water content or conductive pore fluids, but clay minerals, pore geometry and temperature can produce similar effects. Check the natural and spectral gamma record for lithologic context and the induction conductivity guide for conductivity interpretation. These sensors interrogate different sampling volumes, so apparent agreement should be evaluated by depth and scale.
A heat-pulse flowmeter can identify intervals where water enters or leaves a borehole under known hydraulic conditions. A wall optical televiewer can show fractured surfaces where they intersect the hole. Combining these with crosshole anomalies can support a connected-fracture hypothesis. Yet an interwell electromagnetic contrast is not proof of hydraulic connectivity: confirmation may require pressure interference, packer testing, direct water samples or tracer breakthrough observations.
The important matching operation is not placing all images side by side and calling their colours similar. Align them to borehole measured depth, true vertical depth or other documented coordinates as appropriate; retain original observations and explain any depth correction. Look for a plausible physical link between each observable and the claim: gamma responds to radionuclide distributions, induction to electrical response, flowmeter to movement under a pressure condition, and radar to wave propagation. Agreement matters when those independent properties constrain the same explanation.
Translation and technical vocabulary: what each word commits you to
An image in this context is a computed estimate over a spatial grid; it is not a photograph. An attenuation anomaly is a departure from a specified reference, not automatically salt contamination. A tracer response is an observation under a controlled test; a tracer breakthrough usually needs evidence of arrival at a sampling point. Velocity may mean electromagnetic wave speed, fluid speed or a numerical inversion parameter depending on the sentence. Specify which quantity is intended rather than substituting familiar words uncritically.
For a useful wording repair, replace “Tomography proved fast groundwater flow through a fracture at the centre of the map” with “Crosshole travel-time and attenuation data support a modelled electromagnetic contrast in the central interwell region; its hydraulic significance remains to be tested.” If a controlled tracer experiment and independent breakthrough timing then support a transport pathway, state that extra evidence separately. The revised claim retains the important result while making uncertainty visible to a reader who must act on the report.
The eduKate context-stack explanation offers a general habit for technical translation: keep the actor, quantity, physical mechanism, measurement context and confidence level intact. The semantic-feature vocabulary guide explains how near-synonyms differ in what they imply. This matters especially when the original report crosses scientific, policy and public audiences, because “estimated”, “observed”, “consistent with” and “confirmed” are not interchangeable.
An independent transfer test using fresh numbers
Question one: two nearly vertical wells are 24.0 metres apart and a same-level transmission has a corrected first arrival of 240 nanoseconds. What is the straight-path apparent speed and slowness? Answer: 24.0 ÷ 240 = 0.100 metre per nanosecond, and 240 ÷ 24.0 = 10 nanoseconds per metre. With the low-loss approximation, effective relative permittivity is about 9.0. This result says nothing direct about Darcy flow or whether the rock contains one fracture or several, because the measured wave samples the entire path.
Question two: the transmitter is now 7.0 metres below the receiver while the wells remain 24.0 metres apart horizontally. What is the straight-line path length? Answer: √(24² + 7²) = √625 = 25.0 metres. If the rock speed remains 0.100 metre per nanosecond, predicted corrected time is 250 nanoseconds. A reader who divides 24 metres by 250 nanoseconds would infer 0.096 metre per nanosecond and could mistakenly label the interval anomalous. The geometry, not a geological change, explains the difference in this constructed case.
Question three: calibrated normalised amplitude changes from 0.90 to 0.72 along a repeated path. Compute the incremental logarithmic loss. Answer: later/baseline = 0.72/0.90 = 0.80, so −ln(0.80) ≈ 0.223. For a simple 25-metre uniform-increment model, divide by 25 to obtain 0.00893 nepers per metre. It is invalid to report that value as grams of salt per litre without a calibrated petrophysical relation, and equally invalid to treat it as a permeability measurement. The unit of the calculated coefficient identifies what was actually derived.
Question four: a smooth velocity tomogram shows an elongated low-speed zone but the ray-coverage map shows almost no crossing paths near one end. Can the analyst map the entire elongated feature as a confirmed fracture? No. The poorly sampled end may be determined mainly by interpolation or regularisation. One defensible answer is to map a constrained central anomaly, mark the extension as uncertain, and recommend a deeper or differently placed transmitter–receiver survey if that uncertainty matters to the decision. A visible contour is not the same thing as demonstrated resolving power.
Question five: a report states that the tracer arrived between boreholes ten minutes after injection because an attenuation map changed in a survey acquired ten minutes later. What extra evidence is missing? The survey’s actual timing and sampling intervals, baseline repeatability, spatial sensitivity, independent source and receiver conditions, any water-sample tracer measurements, and a transport model. “Detected by the next survey” is not automatically “arrived exactly ten minutes later.” The distinction is especially important when time resolution is coarse compared with expected flow variability.
How to audit a published crosshole radar figure
Begin with geometry: identify boreholes, collar coordinates, trajectories, depth datum, transmitter/receiver positions and sampled ranges. Check the original data quantity—travel time in nanoseconds, slowness in nanoseconds per metre, velocity in metres per nanosecond, or loss in nepers or decibels. Read the colour-bar units and reference values. Inspect how travel times were picked, how instrument delays were removed and how amplitudes were normalised. If these details are unavailable, label the figure an illustration rather than a basis for quantitative site decisions.
Then ask about inversion: which forward model was used, whether rays were straight or recalculated, what regularisation imposed, how data errors were weighted, how many observations were rejected, and what residuals remain. Review crossing-angle coverage, resolution tests, sensitivity to alternative starting models and whether the apparent anomaly stays where it is across plausible choices. A plot can be clear and useful even when uncertainty is large, provided the author does not hide what is poorly constrained.
Finally align the model with independent evidence. Geological core and imaging help constrain rock structure; water chemistry establishes what dissolved species were present where sampled; pressure and flow tests constrain hydraulic performance; tracer breakthrough helps assess transport. If an interpretation says a particular fracture carries water, identify which of these observations supports that extra hydraulic claim. The difference between a measured travel-time anomaly and a confirmed water pathway should remain explicit through the conclusions.
FAQs on crosshole radar interpretation
Does low radar velocity mean high permeability? No. Low electromagnetic velocity generally implies higher effective permittivity within the applicable physical model, but permeability depends on connected pore or fracture pathways and cannot be uniquely recovered from dielectric response. A clay-rich layer may have an electromagnetic signature associated with water while remaining hydraulically tight. Use hydraulic tests or an appropriately validated petrophysical model when permeability is the decision variable.
Can a crosshole attenuation map quantify tracer concentration? Sometimes attenuation changes can be related to conductivity or concentration with independent calibration and modelling, but a generic colour change is not a direct concentration measurement. Temperature, porosity, water saturation, scattering, geometric spreading and antenna coupling can affect signal. USGS research on saline-tracer experiments illustrates both the promise of attenuation-difference imaging and the need for careful calibration and field control.
Why do images show coloured zones outside strong ray coverage? Inverse algorithms often estimate values throughout a chosen grid. Regularisation, boundary assumptions and interpolation may supply values where few data are informative. Such zones should be masked, shaded, described as low confidence or excluded from site decisions. Do not treat equal display opacity as equal scientific confidence.
Why might travel-time and attenuation maps disagree? They respond to different aspects of wave propagation. Travel time is particularly sensitive to effective velocity and permittivity; amplitude loss is influenced by conductivity, scattering, coupling and other terms. Their data quality and inverse regularisation may differ. Disagreement is an invitation to examine physics, calibration and geology, not proof that one colourful image must be wrong.
What is the simplest independent calculation I should demand? Choose a well-characterised path, check its three-dimensional length from borehole coordinates, subtract the calibrated delay from the first-arrival time, and divide path length by corrected time. Then reproduce the reported apparent velocity and verify that the inferred permittivity uses consistent units. For attenuation, reproduce a calibrated amplitude ratio and its logarithmic conversion. These elementary calculations catch surprisingly consequential interpretive errors.
The decision rule that travels beyond one dataset
Translate a crosshole radar dataset in layers: an observed first arrival becomes an effective travel-time constraint; an amplitude ratio becomes a conditional attenuation constraint; a collection of constraints becomes an inversion model; a model anomaly becomes a geological hypothesis; and only independent checks may warrant a hydraulic or operational conclusion. Each arrow involves assumptions. Writing those assumptions where the reader can see them is more valuable than giving a single attractive map a false certainty.
Primary methods and related guides
The US Environmental Protection Agency borehole-radar methods reference covers the physics, frequency and penetration limits of radar measurements around and between boreholes. The USGS Borehole-radar methods for fractured rock differentiates reflection and crosshole applications. The USGS Mirror Lake crosshole tracer study documents attenuation-difference research, while the USGS nonlinear tomography study discusses reconstructing velocity and attenuation with inversion. Each source concerns methods and evidence, not a universal result for every field site.
For connected eduKate reading, compare the surface GPR reflection guide, the single-hole borehole radar guide, and why word-for-word technical translation can lose context.