Soil X-ray micro-computed tomography, commonly called micro-CT or µCT, reconstructs a three-dimensional picture of internal soil structure using X-ray measurements taken from many directions. If you want to read soil micro-CT pore images, voxel size, pore segmentation, connected porosity or network connectivity graphs, the most important first question is which pore sizes the scan and segmentation actually resolved. A bright or dark voxel is a measurement of reconstructed X-ray attenuation, not automatically a labelled pore.
Micro-CT can reveal part of the architecture of air-filled pores, water-filled regions and mineral structures in prepared or intact soil cores. But a segmented pore map depends on imaging conditions, contrast, selected region, voxel resolution, filtering and thresholding. Pores smaller than the effective resolution may not appear. An apparently well-connected visible network does not prove that every smaller pore conducts water, while a broken image network does not prove that no water can flow.
This article explains soil micro-CT interpretation from the original physical core through scanning, reconstruction, grayscale images, region of interest, binary segmentation, voxel counting, connected components, tortuosity and porosity estimates. It includes original worked numerical examples, diagnostics, limitations and reading routes. All numerical values are hypothetical teaching data, not new field measurements. The aim is to preserve the boundary between image-derived geometry and independently tested soil hydraulic function.
Quick answer: what does a micro-CT soil pore map show?
Micro-CT collects X-ray projections as a specimen is examined from different angles. Reconstruction algorithms combine the measurements into a three-dimensional grid of voxels with grayscale values related to X-ray attenuation. In many dry soil scans, air-filled pore space may be distinguished from denser solid material through grayscale contrast. Software then labels portions of the image as pore or nonpore according to a segmentation method.
A reported ‘CT porosity’ is usually the volume fraction of the chosen region labelled pore under the image resolution and processing rules. It is not necessarily total physical porosity. Connectivity metrics describe paths among those labelled voxels under a specified neighbourhood and region boundary. They do not automatically equal hydraulic conductivity or gas diffusion measured in an actual soil.
What a voxel is and why it matters
A voxel is a volume element in a three-dimensional digital grid, analogous to a pixel with an additional dimension. If isotropic voxels have side length 30 micrometres, each voxel represents a cube 30 micrometres on each side in the reconstructed sample coordinate system. Voxels can contain portions of soil grains, water, air or other materials rather than a perfectly pure substance.
The voxel spacing is not necessarily the smallest feature that can be reliably resolved. Optical, detector, reconstruction and contrast limitations make effective spatial resolution more complicated. A very narrow pore throat occupying only part of one voxel might be blurred, misclassified or missed. When a report states ’30-micrometre resolution,’ determine whether it means voxel size, effective image resolution or minimum pore diameter reliably analysed.
A core is a physical sampling choice
Soil is heterogeneous across aggregates, roots, earthworm channels and horizons. A small core selected for high-resolution scanning may not contain the larger pores that influence flow through a whole field. Conversely, a large core scanned at coarser resolution can reveal macropore arrangement while missing fine intra-aggregate pores. The imaging question and sample scale must be matched.
Readers should identify core diameter, length, sampling depth, collection method and whether the sample was intact or disturbed. A report from an artificially packed cylinder cannot automatically establish the pore network of undisturbed soil. Statements about field properties require an appropriate sampling design and a bridge between image scale and landscape scale.
Why tomography requires multiple projection angles
An ordinary X-ray image compresses information along paths through the object. Tomography gathers many projections from different directions to reconstruct internal attenuation in three dimensions. If sampling angles, exposure or movement control are inadequate, reconstruction may introduce streaks, blur or other artefacts that complicate segmentation.
The resulting 3D array is not a direct photograph of empty spaces. It is a computational reconstruction of X-ray measurements, conditioned by the scanner, energy spectrum, sample movement and algorithm. Good technical explanation retains the reconstruction stage rather than pretending a segmented pore cylinder is a pure, unprocessed view of physical reality.
Attenuation depends on material and scanning conditions
X-ray attenuation depends on material composition, density and the photon energy spectrum. Mineral particles, liquid water and air may differ in attenuation, but mineralogical diversity creates overlapping grayscale intensities. A dense iron-rich grain might appear very bright under one reconstruction, while lighter minerals have less contrast. The meaning of bright and dark depends on the display convention.
Some software maps greater attenuation to larger grayscale numbers; others invert the display or rescale intensities for visualisation. Do not assume ‘black equals pore’ without checking the actual segmentation rules. A grayscale legend may be nonlinear or normalised across specimens. Quantitative comparisons need consistent acquisition and reconstruction settings.
Beam hardening can create false structure
Many laboratory X-ray sources emit a spectrum of photon energies rather than one fixed energy. Lower-energy photons may be absorbed more strongly, changing the spectrum along paths through the sample and creating beam-hardening artefacts. Images can show apparent variations in attenuation that are not true material differences. Ring artefacts and streaks can also appear.
Corrections during acquisition or reconstruction may reduce such problems, but can introduce their own choices. A reader should inspect whether artefacts are visible and whether comparisons between treatments used the same processing pipeline. A visually smooth image is not always more physically accurate than a noisier one. Its quantitative reliability depends on the reconstruction calibration and validation.
Region of interest defines the denominator
Researchers often crop a reconstructed core to a region of interest, abbreviated ROI, to avoid tube walls, damaged edges or unrepresentative sections. This choice determines the volume used when calculating imaged porosity and connectivity. Cropping can remove large pores or create artificial boundary connections, changing the network statistics.
A useful report states the ROI dimensions, whether a margin was excluded, and the actual number of voxels analysed. Comparing porosity among samples with different ROI volumes or positions can mislead. If one treatment’s ROI excludes a broken surface layer while another retains it, the analysis may reflect selection rather than treatment effects.
Grayscale segmentation converts continuous values into categories
A raw reconstructed image contains a range of grayscale values. Segmentation assigns each voxel to categories such as pore and solid, often using a threshold or more elaborate classifier. A simple global threshold treats all voxels below or above a chosen intensity as one phase. Its direction depends on imaging convention.
Threshold choice matters enormously near grain boundaries, mixed-phase voxels and noisy regions. An image threshold may change total detected pore volume, create or break bridges among pores, and alter apparent diameters. A final coloured pore network is therefore partly a computational interpretation. Always ask how segmentation was selected and whether its sensitivity was tested.
A worked pore-volume fraction
Imagine a cropped volume containing exactly 1,000,000 equal-sized voxels. After a defined binary segmentation, 180,000 voxels are classified as pore space. The CT-visible porosity estimate is 180,000 divided by 1,000,000, or 0.18, equal to 18% of the analysed ROI. The ratio is dimensionless because both numerator and denominator count equal-volume voxels.
This does not establish that the core’s true total porosity is exactly 18%. Smaller pores can remain unresolved, and mixed boundary voxels can be assigned imperfectly. The scientifically accurate label is imaged or segmented porosity under the specified scan and threshold. If a paper simply writes ‘soil porosity,’ locate its methods to determine what that number includes.
Voxel volume produces a physical scale
If a voxel has dimensions 30 by 30 by 30 micrometres, its volume is 27,000 cubic micrometres. One million such voxels represent 27 billion cubic micrometres, equal to 27 cubic millimetres because one cubic millimetre contains one billion cubic micrometres. Eighteen percent of that illustrative volume is 4.86 cubic millimetres of segmented pore space.
This arithmetic is correct only if the voxel dimensions and ROI count are as specified. The same voxel count at a different physical resolution represents a different real sample volume. A label such as ‘one million pore voxels’ is not a volume unless the voxel dimensions are known. Unit checking helps prevent seemingly precise 3D statistics from losing physical meaning.
Minimum detectable pore size is an operational concept
A pore several voxels wide is generally easier to identify reliably than a thin gap occupying a fraction of a voxel. Noise, partial-volume mixing and blur can make small pores difficult to classify. Different studies may set a minimum radius, diameter or number of voxels for reporting pores, depending on the analysis method and reliability requirements.
When reading a pore-size distribution, find the smallest analysed feature and the algorithm used to convert voxels into equivalent diameters. A graph that begins at 100 micrometres does not prove the soil lacks pores smaller than 100 micrometres. It says the reported analysis did not resolve or retain smaller pores under those conditions.
Partial-volume voxels complicate every boundary
A voxel intersecting a pore wall can contain both air and solid. Its measured attenuation may fall between typical pore and grain values. A binary classifier must place it in one category, even though the original physical space is mixed. When pores are narrow, boundary voxels can represent a large fraction of the pore’s apparent size.
This can bias diameter, surface area, volume and connectivity. The consequences are not confined to the image outline; a single disputed boundary voxel may connect two larger pore regions. Researchers should test whether their conclusions survive plausible threshold variation. Reproducibility requires reporting both the chosen threshold and the sensitivity of key results.
Noise reduction changes the apparent network
Image filtering can suppress random noise but may also smooth away narrow pore necks or merge neighbouring boundaries. The outcome depends on filter size, sequence and whether processing occurs before or after segmentation. A published study of how image processing alters soil pore descriptions examined these issues rather than assuming the prettiest segmentation was best.
Its methodological lesson is that pore estimates can depend on processing. Do not assume a smoothed image is automatically ground truth. A careful paper reports its filter, threshold selection and validation strategy. The question is whether the segmentation approximates the physical target at the available resolution, not whether the image looks aesthetically convincing.
Threshold sensitivity can be checked quantitatively
Suppose a fictional soil image gives 18% visible porosity with one threshold, 16% with a slightly more restrictive threshold and 20% with a slightly more permissive threshold. These three numbers describe how one data set responds to processing choices. They are not three independent biological measurements.
If a treatment difference is only one percentage point, but reasonable threshold changes move the estimate by two points, the interpretation should be cautious. The relevant uncertainty is not limited to repeatability of the scanner. Segmentation decisions may dominate a small effect. The analysis should quantify or at least transparently discuss such sensitivity before claiming a robust treatment response.
Connected porosity is a subset of classified pore space
A pore voxel can be present without belonging to a network that spans the ROI. Connected porosity asks how much classified pore volume belongs to a connected set according to a particular adjacency rule. Some definitions count the largest connected component; others count a component connecting opposing boundaries. These definitions are not equivalent.
Suppose an ROI has 18% segmented porosity, and 12% of its total volume belongs to a network spanning from top to bottom. The spanning fraction is 12 percentage points of the total ROI, or two-thirds of the segmented pore volume. The denominator must be named. A result saying ‘67% connected’ is ambiguous without specifying whether it is relative to all pore voxels or all core voxels.
Neighbourhood rules affect digital connectivity
In a three-dimensional voxel grid, connectivity may be defined through face-sharing neighbours alone or by allowing edge and corner contacts. A pair of voxels touching at a corner is connected under some digital rules and disconnected under others. This apparently technical choice can alter whether a network percolates across the ROI.
Read whether the algorithm uses 6-, 18- or 26-neighbour connectivity in a cubic grid. The appropriate convention depends on the analysis and the physical interpretation. A diagonal point contact may not represent a substantial fluid pathway. Reporting the rule makes the result reproducible and prevents readers from mistaking a software convention for an experimentally measured throat.
A large connected component is not hydraulic conductivity
A visible pathway spanning a soil core suggests geometrical continuity at the imaged scale. But actual water or gas movement depends on pore throat sizes, wetting state, pressure gradients, viscosity, tortuosity and other transport properties. Two soils with equal connected porosity can have different conductivity because their narrowest controlling constrictions differ.
A study of connectivity across X-ray CT scales discusses how resolution and sample size influence network characterisation. Readers should not translate a connectivity percentage directly into a conductivity value unless a suitable physical model has been validated against independent flow observations.
Connectivity depends on the selected boundaries
A top-to-bottom pore network may not connect left to right through the same ROI. Rotating the coordinate frame or cropping the soil cylinder can alter apparent spanning connectivity. Natural soils can be anisotropic, with preferentially vertical channels or horizontal structural features. The direction of interest should match the intended physical question.
For infiltration, vertical connection may be relevant under some conditions; for lateral drainage, horizontal paths may matter. But geometry alone does not settle transport. State the direction, boundary conditions and pore-size cutoff before interpreting a connectivity statistic as a soil function.
Why throat width is different from cavity diameter
A pore body may be wide while the entrance connecting it to a neighbouring pore is narrow. Two-dimensional slices often emphasise a large dark cavity, which can tempt readers to interpret it as a high-capacity flow path. Yet water and air transport may be controlled by the narrowest connected necks. A network graph that assigns one diameter to an entire object can hide this constriction unless the image-processing method measures local throat geometry.
To interpret an apparent pore diameter, identify whether it comes from a local thickness algorithm, equivalent-sphere volume, maximal inscribed sphere, watershed segmentation or another definition. Those methods may produce different histograms from the same image. None is the literal diameter of every irregular natural pore. A useful paper gives the mathematical definition, resolution and any excluded boundary objects instead of treating pore size as a unique property revealed by the scanner.
Watershed segmentation can split one connected space into many objects
A watershed method can separate a segmented void into pore bodies based on distance maps and local maxima. It is useful for constructing a network with pore bodies linked through narrower throats. However, noisy images can generate too many local maxima, causing over-segmentation, while aggressive smoothing can merge distinct cavities. The number of pore bodies is therefore partly algorithm-dependent.
Check the marker-selection rules, minimal object sizes and throat definitions when a study compares pore counts across treatments. Suppose two scans have the same binary pore volume but different watershed parameters. Their reported pore-body count may differ even if nothing physical changed. A reproducible report either holds the analysis constant or quantifies the sensitivity of counts to reasonable parameter choices.
Equivalent-sphere diameter is an interpretation, not a ruler
One common summary assigns an irregular segmented pore an equivalent-sphere diameter: the diameter of a sphere having the same volume. For an illustrative pore volume of 4,188,790 cubic micrometres, the formula d = 2 × [3V/(4π)]^(1/3) gives roughly 200 micrometres. That calculation is appropriate for a volume-equivalent number, provided the source volume and units are correct.
It does not establish that the real pore is spherical, has a 200-micrometre entry throat or would conduct water like a cylindrical tube of that width. A long narrow crack and a rounded cavity can have the same equivalent volume but different transport behaviour. Readers should name the diameter convention whenever comparing micro-CT results with capillary or intrusion-based methods.
Skeletons reveal a different version of connectivity
Some analyses reduce a connected binary network to a one-voxel-wide skeleton or graph. Nodes represent junctions or selected pore bodies, while edges approximate routes between them. A graph can reveal branching, dead ends and spanning structures that are difficult to interpret in a visually dense rendering. It is a topological abstraction of segmented geometry, not a direct map of molecules moving through soil.
Skeletonisation can introduce short spurious branches from rough segmentation boundaries. Analysts may prune these according to defined rules, changing the degree distribution and apparent network complexity. When a study reports coordination number or average branch length, inspect its pruning and graph-construction conventions. A more branched graph is not automatically a more permeable soil if key throats remain narrow or disconnected under wet conditions.
Euler characteristic can summarise topological structure
Topology describes connectivity and enclosed features independently of some metric details. The Euler characteristic is one family of summaries used for three-dimensional binary structures, combining counts of connected components, loops and cavities under a specified convention. An Euler number can therefore provide information beyond ordinary segmented pore volume.
It should not be presented as a direct count of plant-accessible pores or a measure of flux. Its value depends on how the binary volume was constructed, which neighbours count as connected and how the boundary of the selected region was handled. A topology metric becomes useful when those choices are explicitly reported and when the ecological interpretation is consistent with the geometric object actually analysed.
Tortuosity has multiple definitions
A pathway connecting the top and bottom of a soil core may be longer than the straight-line sample length. Geometric tortuosity is often discussed as the path length divided by straight-line distance, which is at least one for an admissible path. Transport tortuosity, however, may be defined through effective diffusion or conductivity models and is not necessarily numerically identical to the geometric path-length ratio.
For an illustrative path 18 millimetres long crossing a core 12 millimetres thick, the simple geometric ratio is 1.5. That number does not tell us the flow rate. Constrictions, saturation and pore-wall effects also influence transport. Check whether a reported tortuosity came from skeleton geometry, random-walk simulation, a transport equation or an empirical fit before comparing publications.
Anisotropy means direction changes the answer
Soil structure can have preferential orientation: vertical root channels, horizontal layering, cracks along one axis or compaction patterns. Connected porosity measured across the height of a core may differ from connected porosity measured laterally. This anisotropy matters for infiltration, drainage and gas movement. It can also arise from how a cylindrical core was extracted or trimmed.
A good comparison states its coordinate frame and boundary conditions. If a paper reports one scalar connectivity number but studies directional flow, determine whether that statistic is aligned with the physical direction of interest. Rotating an image volume can change a face-to-face percolation result without changing the underlying solid material. Readers need the direction as well as the percentage.
Percolation is a threshold phenomenon, not a guaranteed flow rate
A network percolates when connected pore space spans the designated region under the method’s connectivity rule. A small geometrical change can create or break the first spanning pathway. This makes percolation useful for investigating structural transitions, but the existence of a spanning digital path does not establish its hydraulic conductance. A tiny bottleneck may dominate actual transport.
Research on connectivity and percolation of structural pore networks in cultivated silt loam shows how finite sample size and pore network organisation matter. Any quoted threshold from that system is tied to its scale and analysis. Do not promote one experimental percolation percentage to a universal threshold for every soil type, imaging resolution or transport process.
Finite sample size changes the connected fraction
A very small core may miss rare large channels, while a larger core scanned at a coarser voxel size may miss smaller pore connections. This is a fundamental scale trade-off in micro-CT. The field of view and spatial resolution are linked through scanner geometry and detector constraints. An analysis can gain representative volume while losing visibility of fine throats, or gain fine detail while sampling less soil.
When comparing two instruments, check both voxel size and analysed physical volume. Identical segmented percentages at different scales may refer to different populations of pores. If a study presents a representative elementary volume analysis, inspect how the estimated property changes as subvolumes increase. A stable result within one core is helpful but does not by itself prove representativeness across a heterogeneous field.
Edge effects can distort a segmented network
Soil pores intersect the boundary of a selected region. Image processing may treat a boundary-touching void as a complete object, truncate it or discard it according to the metric. These choices influence apparent pore-size distributions, component counts and pathway lengths. Even when the voxel classification is accurate inside the region, cropping can change topological measures.
A careful analysis explains how it handles boundary objects and whether an interior margin was excluded to reduce artefacts. If the question concerns face-to-face connectivity, boundary intersections are essential rather than irrelevant noise. The treatment of edges should match the metric’s purpose. Do not compare pore-body distributions from studies with incompatible rules as though they were direct measurements of the same complete structures.
Multiple threshold values reveal numerical fragility
Segmentation often converts a continuous grayscale histogram into pore and solid classes. A single threshold can look authoritative when shown without alternatives, even though small changes in the threshold may change the connected network sharply. A proper sensitivity exercise repeats key statistics across a plausible range of threshold settings supported by image contrast and independent checks.
Suppose an illustrative scan yields 9% pore volume at one threshold and 13% at another. The four-percentage-point difference is an uncertainty related to processing, not a biological treatment effect. Report both the range and why a threshold was chosen. If connected porosity changes from almost zero to several percent across nearby thresholds, be especially cautious with physical interpretations about drainage or habitat.
Reconstruction artefacts can look like soil anatomy
Ring artefacts, streaks, beam hardening, motion and detector calibration errors can create patterns that resemble cracks or pore boundaries. These structures may recur in reconstructed slices despite not being physical voids. Preprocessing can reduce artefacts, but aggressive correction can also erase or alter genuine fine features. The analyst must distinguish correction from invention.
Ask whether the raw projection quality was checked and whether the reconstruction parameters were held constant across the comparison. A new dark stripe appearing only after one filtering choice is not convincing evidence of biological pore formation. A good diagnostic traces a suspicious feature through neighbouring slices, reconstruction alternatives and the sample’s physical context before assigning ecological meaning.
Water-filled pores can have a different appearance from air-filled pores
X-ray contrast depends on material attenuation, so pore segmentation in a dry soil is not necessarily transferable to a wet soil or a contrast-enhanced experiment. Water, air, organic material and minerals can have overlapping grayscale responses under particular scanning settings. A threshold calibrated for an air-filled pore network may misclassify water-filled volumes.
If a wet soil image is used to infer water distribution, check the contrast conditions and segmentation classes. Contrast agents can aid discrimination but may alter transport or chemistry and introduce their own assumptions. The presence of a dark or bright voxel is not enough to prove its contents; it must be interpreted through the experimental calibration and material properties.
Registered time-lapse images show change only after alignment
Researchers can scan a soil core repeatedly during drying, wetting or physical loading. Time-lapse micro-CT can reveal changes in segmented geometry, but movement of the sample relative to the detector can mimic pore shifts. Image registration aligns scans in a consistent coordinate system before comparing voxels. Without adequate registration, a stable particle boundary may look like a change.
A valid time comparison also considers changing image contrast due to moisture and scanner settings. Distinguish actual displacement or structure change from differences in segmentation at two times. If the paper claims new cracks appeared, seek adjacent-slice evidence, stable registration landmarks and a consistent physical thresholding strategy. A before-and-after rendering is suggestive but not self-validating.
Pore maps do not count microorganisms
Micro-CT can identify geometrical spaces that may be relevant as habitats, but most bacteria and many microbial interactions occur at scales finer than routine whole-core soil tomography. A voxel labelled pore does not tell us whether it contains living microbes, which species are present or whether oxygen and nutrients are available. Those require complementary molecular, chemical or imaging methods.
The soil NanoSIMS isotope-image guide explains a different measurement scale and why atom-level signal cannot be equated with a pore network. Link the two only through a research design that connects samples and resolutions. Similar-looking spatial maps do not imply identical biological information.
Use porosity as a ratio with a declared denominator
Imagine 4,000,000 voxels remain after selecting an ROI and 480,000 are classified as pores. The segmented porosity is 480,000 divided by 4,000,000, or 0.12: twelve percent of the selected voxel volume. If each voxel represents the same physical volume, no separate voxel-size factor is needed in that ratio. The numerator and denominator already share the same unit.
This percentage does not mean twelve percent of all soil pores were measured, or that the soil’s physical total porosity is twelve percent. Unresolved pores and excluded regions may contribute uncounted volume. State ‘twelve percent resolved image porosity in this region under this segmentation’ to prevent a geometric ratio from becoming an unsupported whole-soil property.
Compare pore volume rather than simply pore count
A scan may contain many tiny pore objects yet most classified pore volume may belong to a few large structures. Pore count and volume distribution answer different questions. For example, 900 small components could occupy less total space than one large root channel. A graph of counts by diameter does not tell the same story as a histogram weighted by pore volume.
Check whether a pore-size distribution is number-weighted, area-weighted, volume-weighted or network-throat-weighted. This choice can reverse which size category appears dominant. A student should rewrite ‘most pores are small’ as ‘most detected pore objects by count fall in the smallest reported bin’ when that is the measurement. That sentence is more precise without being difficult to read.
Geometric evidence and simulated flow must be distinguished
Some studies solve fluid-flow equations through a binary micro-CT volume. The computed result depends on boundary conditions, assumed fluid properties, pore geometry, numerical resolution and whether the segmented network faithfully represents the conducting space. Simulation is a useful extension of image analysis, but it does not create an independent measurement of the real sample.
A convincing permeability claim compares the simulated value with laboratory measurements when feasible and reports discrepancies. It should not present a numerical model output as observed hydraulic conductivity unless an actual flow test was performed. A high-resolution image can support model construction; validation remains a separate task.
Geometrical connectivity and transport accessibility differ
Not all topologically connected pores conduct a fluid at a given water potential. A network can be air-connected while water occupies films or smaller pores, and a water-connected pathway can be disrupted by drainage. Whether a throat is open depends on capillary conditions, surface properties and phase distribution, not just whether it appears dark in a reconstructed slice.
This distinction makes image-based soil science especially dependent on moisture state. A porosity map acquired after drying can be relevant to structure yet insufficient to predict infiltration under wetter field conditions. Readers should combine the geometry with water retention, pressure or conductivity data when asking a flow question.
Relating micro-CT to a water retention curve
A soil water retention curve connects water content to matric potential under specified conditions. Micro-CT instead describes imaged pore structure at a selected state and resolution. The two can inform one another, but their reported pore sizes may be derived through different physical assumptions. Capillary-equivalent size from a retention model is not automatically an imaged sphere diameter.
For the functional background, read soil water retention curves, field capacity and wilting point. An integration requires matching sample condition and recognising hysteresis. A pore-body histogram from micro-CT cannot simply replace a measured drainage curve.
Why pore throat data need another method
Very narrow throats may fall below the effective micro-CT resolution while controlling access to larger cavities. Other methods such as mercury intrusion porosimetry use pressure to infer accessible entry sizes, subject to their own assumptions and safety constraints. A cross-method comparison is useful because the observable is different: geometric space in an image versus pressure-governed intrusion into an accessible network.
Do not expect perfect agreement between these methods. An ink-bottle pore can have a wide body and a narrow entrance, so an imaging pore-body diameter and an intrusion-equivalent throat diameter need not match. The correct scientific response is to identify the two operational definitions before calling one study wrong.
Example: a pore threshold that changes a scientific headline
Suppose one scan is segmented under two justified thresholds. Under the first, ten percent of the ROI is pore and only three percent belongs to a top-to-bottom spanning component. Under the second, twelve percent is pore and six percent spans. The total classified pore fraction increases by two percentage points, but the spanning fraction doubles.
This illustrative calculation shows why connectivity can be disproportionately sensitive to a few narrow links. A headline saying ‘The soil’s drainage capacity doubled’ would be unjustified because no water-flow rate was measured. The correct conclusion is that the image-derived spanning pore fraction doubled under the chosen processing alternatives. The distinction is essential to honest interpretation.
Example: resolution changes the apparent result
Imagine scanning the same physical region using fine and coarse voxel spacing. At fine spacing, a narrow connection is resolved, while coarse reconstruction blends the throat with surrounding mineral pixels and labels it solid. The connected component may split in the coarse image even though the real pore connection still exists. A reader who ignores resolution could conclude that the soil structure changed.
Thus comparisons require a shared spatial resolution or a documented method for harmonisation. Downsampling a high-resolution scan can test sensitivity, but it cannot restore features missing from the coarse acquisition. Claims about treatment effects should be stronger than the differences produced by plausible image-processing choices.
How to write a robust figure caption
A helpful caption says what the physical core was, the voxel size and analysed ROI, which materials the colours represent, how segmentation was obtained and what connectivity convention was applied. It identifies whether the statistic describes total classified pore volume, only spanning pores or another subset. A good caption also states the relevant imaging resolution limit.
A caption need not become an entire methods section, but it should stop the reader from treating a processing colour as a direct biological property. For example, ‘Blue marks voxels classified as pore under the stated threshold, not measured water flow’ is a compact and valuable distinction. Proper labels often prevent more errors than decorative visual complexity.
Common interpretation errors and their repairs
The first error is treating voxel spacing as guaranteed minimum pore resolution. Repair it by examining effective resolution and contrast. The second is calling grayscale values intrinsic mineral identities. Repair it through calibration, material context and segmentation validation. The third is treating connected voxel paths as measured hydraulic conductivity. Repair it by distinguishing geometry from actual or simulated transport.
The fourth is forgetting the ROI denominator. Repair it by reporting the analysed physical region and excluded boundaries. The fifth is assuming all connectivity algorithms use the same neighbour rule. Repair it by stating six-, eighteen- or twenty-six-neighbour conventions. The sixth is assuming a pore-count histogram represents volume. Repair it by checking each bin’s weighting.
Practice question: can corner contact carry water?
A segmented volume classifies two cubic pore voxels as touching only at one corner. Under a twenty-six-neighbour rule they are connected; under a six-neighbour face-contact rule they are not. Does the corner touch prove a physically open hydraulic throat? No. The digital rule describes an adjacency choice, while fluid access depends on actual subvoxel geometry and boundary conditions.
The improved report says the chosen digital connectivity rule changes the classification of this pair. If the connectivity result matters, a researcher should examine resolution, physical calibration or an appropriate transport calculation. This answer keeps software topology separate from hydraulic reality.
Practice question: calculate porosity after cropping
A full image contains six million voxels but the analyst excludes a damaged outer shell, leaving a region of four million. Within that region 640,000 voxels are labelled pore. What is the image porosity? Divide 640,000 by four million to obtain 0.16, or sixteen percent. Using the full six-million voxel count would instead yield about 10.7%, which answers a different question.
Neither ratio is total soil porosity unless the relevant unobserved spaces and sample representativeness have been addressed. The exercise shows why the denominator needs to accompany every percentage in a scientific sentence. A plot without that denominator cannot communicate a reliable quantitative comparison.
Practice question: diagnose a false flow conclusion
Two soil cores have the same imaged porosity and both contain a top-to-bottom spanning void. A report claims they must have the same saturated hydraulic conductivity. Why is that invalid? The pores can differ in throat size, path length, branching and wall conditions, and sub-resolution pores may contribute differently. Even the same overall geometry may conduct differently if the fluid distribution or boundary conditions differ.
A useful follow-up would compare transport simulation under explicit assumptions with measured hydraulic conductivity. The image establishes a part of the geometrical premise; it does not determine the complete physical outcome. A good student answer names at least one missing transport parameter and proposes a measurement rather than merely saying the conclusion is uncertain.
Frequently asked questions about soil micro-CT
Does a smaller voxel guarantee every smaller pore is visible? No, effective resolution also depends on contrast and reconstruction. Is micro-CT always nondestructive? The scan may preserve the sample physically, but preparation, drying, contrast treatment or repeated exposure can affect the investigated system. Can micro-CT directly identify a microbial species? Not from ordinary pore grayscale segmentation. Is high image porosity the same as high permeability? No.
Are pore size and pore throat size identical? Not generally. Can a segmentation threshold change connectivity? Yes, sometimes sharply. Do all image analyses define six-neighbour connectivity? No; conventions vary. Are watershed pore counts direct observations? They are algorithm-derived object counts. Does one core represent an entire field? Only to the extent supported by sampling and scale analysis. Can simulation replace real flow measurement? Not without validation for the intended claim.
Sources and connected reading
The Geoderma study of X-ray pore-network connectivity and percolation examines scale-sensitive connectivity in a cultivated soil. A micro-CT study of mercury-intruded compacted clay illustrates differences between observed pore geometry and intrusion methods. These studies concern their own samples and assumptions; neither establishes one universal pore threshold or conductivity rule.
For the surrounding measurement system, compare soil clay XRD basal reflections, soil Vis–NIR spectroscopy and SOC calibration and soil water retention curves. Each interrogates a different material or hydraulic property; none should be confused with image-derived connectivity.
Final principle: separate imaged space from hydraulic function
Soil micro-CT is powerful because it gives a spatial view of structures that bulk methods average together. Its reconstructed grayscale volume, segmentation and network model each add interpretation. Good reporting follows the chain from physical soil core through scanning conditions and voxel scale to pore classification, connected components and explicitly bounded hypotheses about soil function.
A careful reader can explain why a reported connected pore fraction depends on the ROI, threshold, neighbours, resolution and orientation. They can also identify what is still missing before claiming a flow rate or ecological habitat. That is the practical translation skill: preserve the exact geometry the instrument and software revealed, while refusing to turn it into a measurement the experiment never performed.