What a coral growth image really measures
A coral skeleton can preserve a history of growth, but a bright band in an X-ray image is not automatically an annual climate measurement. Reading coral skeletal CT density bands means connecting a physical specimen, an imaging system, a chronology, a density calibration and a biological interpretation. Each connection has its own evidence and limitations. A beautiful image may show a clear pattern without providing enough information to assign an exact year or absolute mass of mineral.
This guide explains how to read linear extension, skeletal density and calcification as separate quantities. We will use hypothetical numbers with checked units, diagnose artefacts, and distinguish an individual colony’s growth record from the carbonate budget of an entire reef. Numerical exercises are deliberately invented rather than represented as observations of real coral.
A defensible reading starts by asking what was directly observed: an image of skeletal structure. The next step asks how a growth path and successive boundaries were identified. Only then can calibrated image values and extension measurements be transformed into a calcification estimate. A climate explanation requires another step beyond that, using external conditions and a research design that can evaluate competing causes.
Understand the sampled skeleton
Corals build complex three-dimensional mineral skeletons. In suitable massive forms, dense and less-dense regions may form bands that researchers can trace through cores. Banding is not guaranteed to be annual in every species, site or circumstance. Some structures split, merge or become difficult to recognise after stress or damage. Scientific chronology needs independent justification rather than a rule that every visible stripe equals one year.
The growth direction is also not necessarily a perfectly straight line. A cylindrical drill core can intersect inclined growth structures. A two-dimensional projection may compress or overlap geometry, while computed tomography can resolve selected structures in a volume. None of these methods automatically identifies the biologically correct path. Orientation, colony identity, core location, preservation and preparation records are essential for interpreting distance.
Before measuring an annual interval, record which end is old, which is young, and whether the analysed structure connects the intended growth surfaces. Reverse display order can reverse a supposed history. A study reporting temporal change should show how its band sequence maps to calendar years, and should mark uncertain or missing sections rather than silently filling them.
From attenuation to an image
X-ray contrast arises because materials attenuate radiation differently. Denser regions can exhibit different apparent intensity from porous regions, but image intensity also depends on scanner settings and reconstruction. Conventional radiography compresses information along a projection path. CT reconstructs a three-dimensional image from many projections, allowing the analyst to inspect internal structure as voxels and virtual slices.
A CT voxel is a small volume element, not an independent coral colony. Its numeric value may reflect attenuation or a reconstructed image index; without a suitable conversion it is not automatically grams of skeletal material per cubic centimetre. Terms such as brightness, CT number, attenuation, relative density and calibrated bulk skeletal density must not be used interchangeably.
Artefacts include beam hardening, rings, partial-volume effects and changes in reconstruction. Thickness and geometry can influence apparent contrast. The best processing record therefore states scanner conditions, reconstruction settings, calibration materials and quality-control tests. If those are unavailable, use cautious language about relative image patterns instead of assuming absolute density.
Define extension, density and calcification
Linear extension is a length per time, often centimetres per year. Skeletal density is mineralised mass per unit skeletal volume, commonly grams per cubic centimetre. Their product is a simplified skeletal calcification metric with units grams per square centimetre per year. The dimensional cancellation is informative: cm yr−1 multiplied by g cm−3 yields g cm−2 yr−1.
That metric is not total coral mass gained per year, because whole-colony surface area has not been supplied. It is also not automatically net reef carbonate accumulation. Reefs contain multiple producers, erosion, dissolution, detrital transport and sediment processes. A coral-core estimate concerns a selected organism and growth geometry; extrapolation to reef area needs additional ecological measurement.
A paper that calls all three outputs ‘coral growth’ without identifying which quantity is plotted can mislead the reader. Density can increase while extension decreases, leaving calcification stable or falling. Accurate translation preserves the three distinct measurements and their units before discussing causes.
A complete arithmetic example
Imagine a hypothetical coral with annual extension 1.20 cm yr−1 and calibrated density 1.40 g cm−3. Their product is 1.68 g cm−2 yr−1. The calculation is elementary but the meaning is specific: deposition per unit skeletal area along the defined growth axis. We should not replace the final unit with simply grams per year.
If extension is expressed as twelve millimetres per year, convert to 1.2 centimetres per year before multiplying by a density expressed per cubic centimetre. Using the number twelve without conversion produces a tenfold error. A scientific sentence can be grammatically elegant yet dimensionally wrong if its units are silently altered.
All teaching values here are invented. They demonstrate the method, not a typical global coral rate or a forecast. Real estimates need documented band chronology, physical calibration, geometry and uncertainty. The arithmetic should be reproducible from the reported inputs, but reproducible arithmetic does not repair a flawed measurement.
Recognising annual boundaries
Alternating density bands may help establish annual growth intervals when their periodicity is validated for the organism and environmental setting. Analysts choose boundaries using images and anatomical knowledge; different readers may disagree where bands are faint or discontinuous. An explicit boundary annotation improves reproducibility more than asserting that software automatically finds ‘the years’.
A time series requires decisions about ambiguous regions. A damaged growth zone, an extra subannual band or a missing interval could alter age assignment. Where independent dates or growth markers exist, use them to test the chronology. Otherwise report candidate intervals with appropriate caution rather than an exact calendar alignment not supported by evidence.
The oldest preserved region of a core is not necessarily the colony’s first mineral. The drill may miss the earliest growth and the outer margin may be damaged. An absolute age inferred from the number of visible bands needs to account for these endpoints and the underlying periodicity assumption.
Measuring a growth path
Two successive band surfaces can be separated by different distances depending on where and how the measurement is made. A straight line across an oblique slice can differ from a path following local growth architecture. When geometry is curved, three-dimensional tracing can provide a more defensible extension path, but the algorithm’s anatomical assumptions still need scrutiny.
For example, an apparent band separation of eight millimetres on one projected line and ten millimetres along a validated growth trajectory differs by 25% relative to the shorter value. More decimal places cannot resolve a wrong choice of direction. A useful method description shows the selected paths and explains how atypical or damaged regions were handled.
If the study averages many paths, clarify the averaging level: local positions, annual bands, separate cores or separate colonies. Measurements nested within one specimen are not independent biological replicates. This becomes important when interpreting standard errors or claims about an entire coral population.
Calibrating image intensity to density
An image cannot supply absolute skeletal density merely because brighter pixels look mineral-rich. Reference materials with known physical properties and an instrument-specific relation are required. Analysts may fit a conversion curve between a calibrated CT index and density and then apply it to defined skeletal volumes. The fit must be checked across the intensity range relevant to the specimen.
Suppose an invented calibration is density = 0.50 + 0.0020 × image index. An index of 450 yields 1.40 g cm−3, while 500 yields 1.50. The example is a mathematical demonstration, not a transferable equation. Actual scanner coefficients must be determined from standards, geometry and validated processing for that system.
A regression’s coefficient of determination is not the entire calibration audit. Inspect residuals, outlying standards, check samples and sensitivity to scan settings. Shared systematic error can bias many years of a coral record simultaneously. It is possible to have tightly clustered voxel readings but an incorrectly calibrated density scale.
Band-average density is not peak brightness
A year’s skeleton contains regions of different porosity and mineral density. An annual estimate typically summarises an explicitly defined volume or region between successive growth boundaries, according to the chosen method. It should not simply use the brightest thin stripe and call that the whole year’s mean density. The selected image mask and treatment of voids matter.
If two annual regions have mean calibrated densities of 1.40 and 1.55 g cm−3, their contrast is 0.15 g cm−3. That contrast is an observation about the processed skeletal material. It does not identify a unique environmental cause. Skeletal architecture, growth form, mineralisation, bioerosion and imaging corrections may all influence the relationship.
Where a system uses millions of voxels, their number is not the number of independent annual density measurements. The sampled colony, annual interval and anatomical region are the relevant levels of interpretation. Preserving those levels helps prevent misleading statistical precision.
Why density and extension can move in opposite directions
Consider three invented annual records. Year one has extension 1.0 cm yr−1 and density 1.50 g cm−3, giving calcification 1.50 g cm−2 yr−1. Year two has extension 1.3 and density 1.40, giving 1.82. Year three has extension 0.9 and density 1.60, giving 1.44. The highest density year is not the highest calcification year.
From the first to the second year, density falls by about 6.7% while extension rises by 30%. The product rises by approximately 21.3%. From the second to the third, density increases but extension falls enough that calcification declines. A summary that discusses only density would misdescribe the biological growth metric.
The components can also be correlated through measurement. Changing a selected band boundary may alter the distance assigned to an annual interval and the volume from which density is calculated. Quality control should therefore evaluate both components and not only the final product.
An uncertainty calculation
If two inputs have small and independent relative standard uncertainties, the relative uncertainty of their product is approximately the square root of their squared relative uncertainties. Extension of 1.20 ± 0.06 cm yr−1 and density of 1.40 ± 0.07 g cm−3 each have a 5% relative uncertainty. Their product, 1.68 g cm−2 yr−1, has an approximate relative uncertainty of 7.1%, or roughly 0.12 g cm−2 yr−1.
This simplified propagation omits correlated errors and systematic effects. Density calibration could shift an entire set of annual estimates. Uncertain band placement could affect extension and density together. The covariance between them may require an explicit term. Reporting only a repeated-pixel standard deviation would not capture these uncertainties.
Where uncertainties cannot be quantified, describe their sources and test whether plausible alternative image boundaries change the conclusion. Scientific candour is preferable to six decimal places that imply precision unsupported by the scanner and chronology.
Replication and statistical independence
One core can yield hundreds of CT slices and many annual observations, but they remain nested within a biological specimen. Repeating the scan tests instrument or processing repeatability. Taking multiple paths through a core tests within-core geometry. Sampling different colonies tests biological variation. These activities answer different questions, and their sample sizes should not be combined into a single impressive but meaningless number.
If ten years from one colony show a decline, that may be a compelling individual history. It does not prove every coral on a reef declined. Annual values can be temporally correlated, while colonies can share environmental exposure. Analyses should respect hierarchy rather than treat all voxels, slices, years and specimens as interchangeable independent units.
A reader should be able to find the number of colonies, the number of cores, the years analysed, the number of scans and the number of replicate paths. Those numbers improve interpretability more than an unexplained ‘n’ in a figure caption.
Missing material and bioerosion
Internal cavities, damaged sections and ambiguous skeletal boundaries can disrupt the record. CT can reveal features that a simple projection hides, but visible damage is not automatically dated. Material removed after initial deposition may affect what remains of a historical growth band. A preserved core is evidence of retained structure, not always a perfect record of every depositional event.
Excluding a region can be scientifically justified, but the mask, rationale and sensitivity of the result should be documented. If missing material compromises a particular year, it can be more honest to report an unresolved interval. Forcing every image region into a complete annual table makes the output look tidy while hiding a significant uncertainty.
Claims about bioerosion should also be distinguished from calcification. Both processes matter in reef development, but their measurement methods and spatial denominators differ. An individual coral’s skeleton cannot be used to claim the entire reef’s carbonate balance without supplementary evidence.
From measurement to environmental explanation
Ocean temperature, light, nutrients, carbonate chemistry, disturbances and biological condition can influence coral growth. Yet one density or calcification change cannot uniquely identify its cause. A statistical association with a warm year may support a hypothesis, but alternative processes and methodological changes must be tested. Correlation should not be silently rewritten as exclusive causation.
Before comparing old and new scans, verify calibration consistency and imaging workflows. Otherwise a scanner upgrade could masquerade as a biological trend. Cross-study comparisons need compatible processing and an understanding of species and colony differences. A common colour map is not a common physical calibration.
A strong report writes separate statements for the measured skeletal outcome, the environmental variable and the proposed mechanism. Readers can then see whether the study established an association, a mechanistic explanation, or a causal estimate under a particular design.
The difference between a coral core and a reef
Skeletal calcification on a core growth axis is often measured in g cm−2 yr−1. A reef carbonate budget considers production and removal over a defined seabed area, including multiple species, bioerosion and transport. Matching mass-per-area-time units do not guarantee matching physical denominators. The sampled skeletal area is not automatically one square metre of reef footprint.
A purely numerical conversion from 1.5 g cm−2 yr−1 to 15,000 g m−2 yr−1 is dimensionally correct. Calling that net reef accumulation would still be wrong without information about coral cover, geometry, other carbonate producers and losses. A rigorous translation must preserve both the unit and the surface to which the unit refers.
A separate eduKateSG guide discusses reef-wide carbonate budgets. This article focuses on the measurement chain within coral skeletal CT and X-ray studies. The distinction is valuable because researchers can correctly calculate a core metric while still drawing an unsupported ecosystem-scale conclusion.
Documentation that makes an image auditable
A reproducible workflow records sample collection, identity, orientation, preservation, imaging, reconstruction, standards, artefact correction, band annotation, growth-path selection, segmentation, density conversion, extension measurement, product calculation, uncertainty and exclusions. It should preserve raw and processed files where permitted and make clear which observations were included in each annual estimate.
The 2026 USGS Rotating X-ray Computed Tomography Coral-Core Archive describes standardised procedures intended to improve comparisons of density, extension and calcification among stored cores. Such a method provides an example of careful calibration and provenance; it is not a universal correction formula that can be applied blindly to every scanner and coral species.
The simplest trustworthy sentence is sometimes longer than the dramatic alternative: ‘The selected coral core shows a decline in estimated annual skeletal calcification under a defined calibrated CT workflow, with specified chronology and measurement limitations.’ It informs the reader without claiming to have measured the entire reef or isolated one environmental cause.
Guided transfer: a corrected annual calculation
Problem one. An analyst records annual linear extension of 12 millimetres and density of 1.5 grams per cubic centimetre, then reports calcification of 18 grams per square centimetre per year. The arithmetic has ignored the length conversion. Twelve millimetres equal 1.2 centimetres, so the correct product is 1.8 g cm−2 yr−1. The error is a tenfold inflation caused by mixing units, not a disagreement about coral biology. Show both input units before multiplying, and retain the full output denominator in any translated caption.
Problem two. Two consecutive annual zones have the same calibrated mean density, 1.6 g cm−3, but extensions of 0.8 and 1.2 cm yr−1. Their respective calcification values are 1.28 and 1.92 g cm−2 yr−1. The second is 50% higher than the first despite unchanged density. The scientific explanation must therefore discuss extension, not invent a densification process. This example shows why density curves and calcification curves cannot be used as synonyms even when plotted together.
Problem three. A coral grows 1.0 cm in one year at density 1.50 g cm−3, then 1.3 cm at density 1.40 in the next year. Calcification changes from 1.50 to 1.82 g cm−2 yr−1, an increase of approximately 21.3%. Density fell by about 6.7%, but the extension increase more than offsets it. An article that says lower density means less mineral growth per unit skeletal area would misread the product. State both component changes before interpreting the resulting deposition metric.
Problem four. A plot shows a calibrated density of 1.40 g cm−3 at image index 450 under an invented relation, density = 0.50 + 0.0020 times the index. At index 500 the model gives 1.50. Those numbers are correct for the hypothetical equation, but the equation must not be applied to a different scanner. A proper density model needs standards from the relevant imaging workflow and tests for artefacts. Calculating confidently with the wrong calibration is not improved by reporting additional decimal places.
Problem five. An image contains ten alternating light and dark regions and a manuscript calls the coral ten years old. This confuses visible image features with independently demonstrated annual periodicity and ignores the possibility that earliest material is missing. Repair the wording to ten observed candidate band intervals and describe what validates an annual interpretation. If independent dates exist, cite them. If not, keep the chronology conditional rather than forcing the image into a calendar that the evidence cannot support.
Problem six. A dense band splits into two narrow stripes midway across the core. One analyst counts two years; another counts one. Both cannot be resolved simply by selecting the visually tidier answer. Re-examine three-dimensional continuity, anatomical growth direction, neighbouring bands and external chronological evidence. Record the alternative boundary assignments and test their consequences for calculated extension and density. A disputed band should not secretly become an exact date in a public-facing summary.
Problem seven. A core has a valid annual boundary at each end, separated by 8 mm on a flat projection but 10 mm along the validated curved growth path. The path distance is 25% greater relative to 8 mm. Depending on growth anatomy, using the short projected ruler could systematically underestimate extension. Document which path corresponds to skeletal accretion and measure it consistently. A three-dimensional model is a tool for anatomical measurement, not an automatic guarantee that every path it draws is correct.
Problem eight. A researcher calculates annual density from only the brightest pixels of a high-density stripe. The selected pixels may not represent the bulk volume deposited in the year, which includes porous and dense microstructures. Define the complete region between agreed annual boundaries and summarise the appropriate voxel values using the validated method. Describe treatment of cavities and segmentation. A peak brightness value and a calibrated annual mean density answer different questions.
Problem nine. A published image has attractive colours and fine contrast but no reference standard for absolute density. It may still support identification of bands or relative brightness patterns. It cannot establish an absolute g cm−3 estimate merely because its colours look like those in a calibrated archive. Explain which measurement is available, and do not borrow numerical density conversion factors from another scanner. Absolute calcification should be withheld if its essential density input cannot be justified.
Problem ten. The extension estimate is 1.20 ± 0.06 cm yr−1 and density is 1.40 ± 0.07 g cm−3. Under a simplified independent small-error model, their relative uncertainties are both 5%, yielding approximately 7.1% for the product. The calcification estimate is 1.68 ± 0.12 g cm−2 yr−1. The example does not include uncertainties from correlation, systematic calibration, band periodicity or site selection. Treat it as a dimensional lesson rather than a complete field-specific budget.
Problem eleven. Two scanners generate different brightness values for the same reference material. A researcher merges their uncalibrated annual density series and identifies an abrupt biological regime shift at the date of scanner change. That coincidence is a methodological warning. Harmonise using valid standards and inspect residual bias before interpreting the change. The apparent trend could partly or entirely result from instrument settings, reconstruction or geometry instead of actual skeletal differences.
Problem twelve. A density calibration has a very high r-squared but systematically underestimates the highest-density check material. The reported fit statistic does not remove the systematic error. Inspect residuals against image intensity and verify the predicted densities of independent standards across the range occupied by the coral. An apparently excellent average fit can bias the densest annual zones and distort a long-term trend. The appropriate response is further calibration assessment, not merely showing a better-looking regression line.
Problem thirteen. The same coral core is scanned three times and yields nearly identical annual density curves. The agreement supports technical repeatability, but it does not turn one coral colony into three independent organisms. Report one biological specimen and three scans. To estimate natural population variability, sample additional colonies under a defensible design. The unit of replication determines which confidence claims are possible, and a large image dataset does not substitute for biological sampling.
Problem fourteen. A CT archive contains a million voxels from one core, and a summary describes its sample size as a million corals. This is a category mistake. Voxels are elements of a reconstructed image, not independent colonies or even independent annual density observations. Correct the report to identify the number of colonies, cores, years and image elements separately. The scanner may provide high spatial detail while the ecological scope remains limited to one specimen.
Problem fifteen. An internal cavity crosses two adjacent annual regions, and the automated algorithm includes the void as zero-density skeletal material. The effect on annual means depends on the method’s definition of bulk skeletal density and handling of bioerosion. Inspect its segmentation rules and test the impact of masking versus inclusion. Neither method can be called universally correct outside its stated scientific purpose. Document what has been measured and how the void was treated.
Problem sixteen. A coral core has a missing section just before a suspected stress event. The manuscript assigns the nearest visible boundary to a specific year and concludes that the event caused the gap. Missing structure creates chronological ambiguity, and the environmental explanation is another inferential step. Mark the unresolved interval and seek independent dating or contextual evidence. A visually persuasive alignment is not proof of cause, particularly when damaged material may have altered the apparent band sequence.
Problem seventeen. A data table reports skeletal calcification at 1.5 g cm−2 yr−1; the translation changes the unit to g m−2 yr−1 while retaining the value 1.5. One square metre contains ten thousand square centimetres, so the numerical value would need to become 15,000 for a pure unit conversion. But that converted value still refers to the equivalent skeletal surface, not net production per square metre of reef floor. Both dimensional and ecological denominators must be preserved.
Problem eighteen. A four-year sequence contains calcification estimates 1.7, 1.6, 1.5 and 1.4 g cm−2 yr−1. A consistent downward direction appears, but four serial observations in one core are not four independent reefs. Before claiming a regional trend, examine longer chronology, uncertainty, multiple specimens and environmental context. The correct descriptive claim is that the estimated series for the sampled interval declined. Generalisation requires a design that can support it.
Problem nineteen. A researcher notes that warm-water years correspond to lower calcification and writes that temperature alone caused the decline. Temperature may be relevant, yet light, nutrients, acidity, organism health, disturbance and measurement methodology can covary. The repaired text identifies the association, explains the possible mechanism and states which alternatives were tested. Causal wording must match design strength; otherwise a plausible hypothesis is being presented as established mechanism.
Problem twenty. A CT image is smoothed before annual band selection. The processing may improve readability while shifting the apparent position of a narrow boundary or combining small features. Publish the smoothing parameters and retain the original image for comparison. If the annual estimate changes appreciably when processing varies within reasonable limits, report that sensitivity. A smooth plot can communicate data effectively, but smoothing must not silently create precision or remove difficult features.
Problem twenty-one. An author treats a single colony’s calcification as the whole reef’s net carbonate budget. Coral skeletal growth and ecosystem net accumulation have different spatial domains. The reef also contains other calcifiers, bioerosion, dissolution and sediment transport. The correction is to name the colony-specific metric and avoid scaling it to seabed area without separate coverage and process measurements. Numerical agreement between units does not establish physical equivalence of denominators.
Problem twenty-two. A study reports density to six decimal places after calibration against uncertain standards and ambiguous boundaries. That presentation implies more precision than the method may justify. Report enough digits for reproducibility without falsely suggesting the uncertainty has vanished. If a derived result is 1.683421 but meaningful uncertainty is around 0.12, a rounded estimate near 1.68 with its uncertainty is generally more transparent than displaying all computational digits.
Problem twenty-three. Two analysts draw slightly different annual boundaries and obtain distinct extension estimates. Rather than averaging their answers without explanation, investigate anatomical rules, record the alternate boundaries and assess how much calcification changes. Interobserver comparisons are valuable quality-control information. If the uncertainty materially changes the ecological conclusion, disclose it. The goal is not to make all operators agree at any cost, but to reveal how interpretation affects measurement.
Problem twenty-four. A translated figure refers to ‘mass production’ but the underlying measurement is skeletal density in g cm−3. Density is mass divided by volume, not a production rate. A production-related proxy requires extension or a suitable time-resolved mass-change model. Correct the figure label and preserve the unit. Readers can then distinguish a dense but slowly extending interval from one with greater annual skeletal deposition.
Independent problems and answer checks
Check A. You are given extension 0.75 cm yr−1 and density 1.8 g cm−3. The product is 1.35 g cm−2 yr−1. Explain why the result has an area denominator and why you cannot infer the colony’s total annual mineral mass without geometry. The answer is dimensional: length per time multiplied by mass per volume gives mass per area per time, while the physical area over which to integrate has not been supplied.
Check B. You are given a coral core with 15 accepted candidate annual intervals and no independent proof that all bands are annual. May you report a 15-year calendar record? Not unconditionally. The correct response is to report the observed number of intervals and the tentative periodicity, then identify the evidence needed for confident age assignment. An unvalidated count cannot supply exact calendar endpoints merely by being repeated in a polished chart.
Check C. Annual extension changes from 1.1 to 0.9 cm yr−1 while density increases from 1.4 to 1.6 g cm−3. Initial calcification is 1.54 and later calcification is 1.44 g cm−2 yr−1. Despite densification, the product declines. State that the two components respond differently and avoid describing the density increase as proof of faster skeletal production.
Check D. The standardised CT workflow achieves a useful calibration fit, but one year’s growth zone lies in a heavily beam-hardened region. May the density for that year automatically be trusted? No. Calibration and artefact diagnostics need to cover the relevant geometry. The correct answer is to assess whether correction and uncertainty are adequate for that region, and potentially exclude or flag the annual estimate.
Check E. A group compares one coral from a sheltered lagoon with one from a wave-exposed slope. Different growth metrics could reflect biological and environmental differences, but two specimens are not a controlled population-level experiment. Explain the sampling restrictions and identify what additional replicated colonies and environmental observations would improve inference. The purpose is to separate an informative comparison from an overextended conclusion.
Frequently asked questions
Can any coral X-ray band be used as a year? No. Periodicity, continuity and anatomical interpretation need validation. Missing or split bands can change chronology, and a core may not contain the colony’s full life history. Reports should show the band-assignment method and explicitly identify uncertain intervals rather than treating visual repetition as automatic annual dating.
Can extension be measured without absolute density calibration? Sometimes, if band boundaries, growth direction and scale are established. The inability to infer absolute density does not eliminate all geometric information. However, the standard extension-times-density calcification estimate cannot be completed faithfully without a suitable density input.
Is CT automatically superior to a conventional X-ray? CT provides volumetric information that may improve structural interpretation, but it also requires calibrated reconstruction and careful handling of artefacts. A well-characterised radiograph can answer a limited extension question, while poorly calibrated CT can produce unreliable density. Choose methods by validated quantities rather than by technological prestige.
Do calcification changes prove ocean warming or acidification? Not by themselves. Skeletal outcomes can respond to multiple conditions, and associations need careful environmental analysis and sampling. A strong conclusion states the measured trend, the plausible mechanisms, which alternatives were examined and any unresolved uncertainty.
Connected methods
For the broader ecosystem accounting distinction, consult eduKateSG’s coral reef carbonate budget and bioerosion guide. For another example of separating instrument signal from a physical estimate, see the radar altimeter waveform and significant wave height guide. The linked methods have different equations and should not be used as substitutes for coral imaging calibration.
First-party and specialist sources
Source 1. U.S. Geological Survey (March 3, 2026), USGS Rotating X-ray Computed Tomography Coral-Core Archive: scope, access and standardisation, https://www.usgs.gov/publications/usgs-rotating-x-ray-computed-tomography-rxct-coral-core-archive-scope-access-and . The methodological discussion addresses the calibration of skeletal density, annual extension and calcification from CT.
Source 2. DeCarlo and colleagues (2025), CoralCT: a platform for transparent and collaborative analyses of growth parameters in coral skeletal cores, https://doi.org/10.1002/lom3.10661 . The method discusses annual regions, voxel averaging and the distinction between grayscale and calibrated density.
Source 3. Earth Observatory of Singapore at Nanyang Technological University (2021), Core-CT: quantitative analysis of sediment and coral cores, https://www.earthobservatory.sg/research/publications/core-ct-a-matlab-application-for-the-quantitative-analysis-of-sediment-and-coral-cores-from-x-ray-computed-tomography-ct . The source describes structure segmentation and applications to coral growth bands and cavities.
Source 4. NOAA Ocean Acidification Program (2021), measurements of blue coral linear extension, density and calcification, https://oceanacidification.noaa.gov/oap_pubs/linear-extension-skeletal-density-and-calcification-rates-of-the-blue-coral-heliopora-coerulea/ . Species and morphology need to be identified when transferring methods or interpreting numerical ranges.