Introduction
A sonic tomograph does not photograph the inside of a tree. It measures one thing only: how long a stress wave takes to travel from one sensor to another. Everything you see on screen — the blue, the green, the red — is a velocity map calculated from those travel times, with the space between sensors filled in by software. It is an estimate wearing the costume of a picture.
So I have settled on a simple way of describing what I expect from the instrument. It tells me whether something abnormal is there. It does not tell me what that abnormality is.
When a red region appears in a tomogram, I cannot say from the image alone whether it is a cavity, advanced decay, or a crack. I can say that sound travels slowly through it. And in my field practice in Tokyo, the instrument tends to read high — it shows more damage than I find.
I still use it, and not because it is precise. I use it because it lets me assess trees nobody will let me drill, and because it produces something I can show to people who are not arborists. In Japan, those two things decide whether the work happens at all.
In the resistograph article I wrote that the drill is measured along a single line and nowhere else. Sonic tomography sits at the opposite pole: you see the whole cross section, and most of what you see was never measured. That contrast is the reason to own both.
What follows is how the ArborSonic 3D (Fakopp Enterprise, Hungary) works, what numbers it produces, how I read them, and where I have watched people — including myself — read them wrong.
What an ArborSonic 3D Can Do
It works without drilling
Beyond driving sensors through the bark into the wood, the instrument does not damage the tree. No boring, no sampling, no radiation.
That reads like a line on a spec sheet. In Japan it means something else. Goshinboku (御神木, sacred trees), nationally designated natural monuments, municipally protected trees, temple and shrine grounds — trees where felling is out of the question and where even a 3 mm drill needle can be difficult to get agreement for. In Western tree risk assessment, removal remains a realistic option when risk is high. In Japan, a much larger share of the trees I assess are trees that will not be removed under almost any finding.
Under that constraint, being non-invasive stops being a nice property and becomes a condition of entry. There are sites where sonic tomography is the only window available.
None of this is unique to the ArborSonic 3D — every sonic tomograph is non-invasive. The point is that the property carries different weight here. Japan’s urban forest is ageing fast, as I described in the article on the postwar planting boom reaching its limit, and a large part of that ageing stock is culturally or legally protected.
It builds a 3D model — so non-specialists can see it
Measure several cross sections at different heights and the software stacks them into a three-dimensional model. That is where the “3D” in the name comes from.
I value this less as a diagnostic feature than as a tool for reaching agreement.
The conclusion of a tree assessment usually does not travel well. Tell a head priest, a neighbourhood association, or a municipal officer that “residual wall thickness is insufficient on the north side of the trunk at 1.2 m”, and nothing forms in their mind. Put a coloured three-dimensional model in front of them and the conversation moves. Being able to show a finding matters, in practice, about as much as producing it.
And the two properties compound. Trees that cannot be cut tend to be the trees with the most stakeholders, which means the most people to explain the finding to. Non-invasive assessment and visual explanation are needed on the same site, on the same day.
It works aloft
Measurement is not confined to the base of the trunk. With rope access you can take cross sections in the canopy. In mature trees, decay frequently enters through branch unions and old pruning wounds rather than from ground level, and ground-based assessment simply cannot reach those points.
I have written up a worked example of this at a Tokyo temple, where climbing diagnosis and the ArborSonic 3D were used together.
What the Instrument Actually Measures
One kind of data: travel time
The device acquires exactly one type of measurement — the time a stress wave takes to travel from one sensor to another. That is all.
The procedure runs like this:
- Sensors (transducers) are driven through the bark into the wood around the circumference.
- Each sensor position is measured and entered into the device — distances between sensors, or coordinates.
- Each sensor is struck in turn with a steel hammer.
- The device records the arrival time from the struck point to every other sensor, with microsecond precision.
- Velocity is derived from time and distance, the velocity distribution across the section is reconstructed, and the result is rendered as colour.
Fakopp gives the ArborSonic 3D a timing precision of ±2 microseconds. It supports 8 to 32 sensors. A single cross section takes 5 to 10 minutes — around 20 minutes including setup with ten sensors — the operating range is 0 to 40 °C, battery life is roughly two hours, and the unit weighs about 6 kg with case and ten sensors.
Sound moves quickly through sound wood and slowly through decayed or hollow wood, so slow regions are displayed as damage. That is the whole principle. The instrument is not looking at decay. It is looking at delay.
The geometry you enter becomes the assumption
This gets underestimated. The software treats the sensor coordinates you supply as ground truth and calculates the velocity distribution from them. The cross-sectional shape of the trunk is not a measurement result — it is a premise you provide.
So the further the trunk departs from a circle, and the more casually you capture its shape, the more the image distorts. On Japanese street trees and old temple trees, buttress roots, occluded wounds, fused stems, props and stakes are everyday conditions. Entering the geometry carefully buys you more accuracy than adding two more sensors.
Measured lines, painted surfaces
Velocity is genuinely measured only along the straight paths between sensors. With eight sensors, that is a fixed set of lines crossing the section. The smooth coloured field you see on screen is what an algorithm placed between those lines. It is not measurement.
So reading a tomogram means continuously asking one question: am I looking at a colour on a line, or a colour between lines?
Adding sensors makes the lines denser. A 2025 study in Forests on hoop pine (Araucaria cunninghamii) compared 8, 12, 16 and 20 sensors and found that while image resolution and completeness improved with sensor count, differences in average stress wave velocity were not statistically significant, and defective trees showed agreement of R² = 0.82–0.91 even with eight sensors. The authors concluded that eight sensors are sufficient in practice for detecting decay and cavities.
Sensor count governs how coarse the image looks. It does not govern whether you are right. Making the picture finer and reading it correctly are separate problems.
The Numbers, and How to Treat Them
The tomogram is striking as an image, but what ends up in the report is a number. And the numbers cannot be taken at face value.
Decay and cavity percentage — it changes with the species setting
This is the proportion of the cross-sectional area classified as damaged. The software isolates the low-velocity region from the measured distribution and calculates its share of the section.
Here is what matters, and what I have not seen written down in English anywhere: this number changes with the tree species you select. Measure the same tree, at the same height, with the sensors in the same positions, and a different species setting returns a different decay percentage.
Treat that as a warning, not a feature. Selecting the species is not a record-keeping field. It is an input to the assessment itself. Get it wrong and the number in your report is wrong with it.
Safety factor — where the species database enters
The ArborSonic 3D software requires species selection before measurement can proceed; the manual states that “tree species must be selected before you can move to the next page.” The chosen species calls two values out of the built-in species database — the drag coefficient of the crown and the yield strength of the trunk wood.
The safety factor is then computed as yield stress divided by maximum stress. Change the species and the numerator changes; change the drag coefficient and the maximum stress derived from wind loading changes with it.
Why species parameters matter: ArborSonic 3D and PiCUS
This is the comparison I almost never see made in English-language equipment reviews.
PiCUS also offers a species field. But there it is documentation only. The PiCUS 3 manual is explicit that the setting “will only affect the measuring process. It does not affect the calculation.”
So: in the ArborSonic 3D, species enters the calculation. In the PiCUS 3, as documented in its manual, it does not. It reaches the decay percentage and it reaches the safety factor.
In a country with Japan’s species diversity, and where the common street trees do not overlap much with the Western standard set, this is not a small distinction. Kusunoki (camphor), keyaki (zelkova), icho (ginkgo), enju (Japanese pagoda tree) — their wood properties, their crown architecture and the way they load in wind all differ from the oaks and maples that most equipment reviews implicitly assume. Being able to carry species properties into the assessment is a real advantage for an instrument used here.
Four Shapes Worth Recognising
The shape of a low-velocity region is a clue to its identity. It is not proof, but it is where you start before going back to the trunk.
1. A low-velocity region centred in the section
The classic form of heartwood decay or a central cavity: progression from the heart outward with sapwood still intact. When the residual wall is reasonably even around the circumference, the strength assessment is comparatively straightforward.
2. A low-velocity region weighted to one side
Suspect decay that entered through an old wound, an open cavity, a large removal cut, or damage at the base. There is usually an entry point visible somewhere, so walk the whole trunk and find it. Because the loss is asymmetric, the risk changes sharply with load direction.
3. A narrow low-velocity line running radially from the centre
Suspect a crack. A crack can be thin as a plane and still block the acoustic path completely, so it tends to appear as a fine linear slow zone. This is the easiest shape to misread: the area it occupies can be large while the material lost is small.
4. A low-velocity band following the circumference
Suspect ring shake. Separation along the growth rings is hard to distinguish from extensive decay, and circumferential separation behaves nothing like a cavity in bending. Confusing the two produces a badly wrong assessment.
None of these four is a diagnosis. Each is a decision about what to check next. Read the shape, go back to the trunk, and test it against what you can see. If it still will not resolve, go and take a line with the resistograph.
What the Tomogram Does Not Answer
Cavity, decay or crack — not from the image alone
Shape is a clue, but the image on its own cannot confirm what you are looking at. Cavities, advanced decay, cracks and ring shake all appear the same way: as regions where sound is slow.
You can, however, infer. Overlay the tomogram on what visual tree assessment gave you — open wounds, the species and position of fruiting bodies, sunken or raised bark, old injuries and pruning cuts, the quality of the sound under a mallet — and the identity narrows considerably. This is where the Japanese jumoku-i (樹木医, certified tree doctor) tradition of close visual reading and the Western VTA framework do the same job from different directions, which I compared in an earlier article on VTA and jumoku-i.
So the accurate statement is this. The image alone cannot distinguish them. Visual assessment, experience and the data together can produce a reasonable inference. This instrument does not deliver verdicts. It increases the material you have to judge with.
And the more material there is, the more the reader’s experience determines the outcome.
In my experience it reads high
Here my field impression diverges from the published literature, so I will give both.
Xiping Wang and colleagues at the US Forest Service compared tomograms with the cross sections of felled black cherry (Prunus serotina) and reported error running in both directions. For heartwood decay the tomogram underestimated: severe-decay colour covered 20% of one section where hardness mapping found 33.3% degraded. Where internal cracks were present it overestimated: acoustic shadow covered 40% of a section in which only 15.8% proved defective. The same study noted that small sapwood decay and insect holes in otherwise sound sections are not reliably detected.
My own impression, working on Japanese species, is that the instrument errs toward overestimation more often than not. At the same time, I find it does pick up very small cavities.
Those two observations are not in conflict. I think they are the same property seen from two sides. Sensitivity fine enough to catch a small defect cannot be separated from a tendency to flag things that are not defects. A sensitive instrument produces false positives.
Which leads to the operational conclusion: red on a tomogram is not a verdict of danger. It is an instruction about where to look next.
Where You Set the Pins Decides the Result
A small shift changes the data
Move the sensor positions slightly and the data changes.
There are two reasons. First, as described above, the device treats the entered coordinates as truth, so how you capture the trunk’s cross-sectional shape feeds directly into the image. Second, moving the pins moves the acoustic paths, which changes which defects those paths happen to cross.
Two people measuring the same tree will not necessarily produce the same image. This instrument depends on its operator.
Visual assessment chooses the pin positions
So what decides where the pins go? Not the machine. Mallet sound, fruiting bodies, bark abnormalities, the tree’s form and pruning history — findings from visual assessment decide it.
The structure is identical to the resistograph. The instrument confirms a hypothesis; it does not generate one. Practitioners who have built the skill of reading decay without any technology at all extract more from the machine than those who have not.
And that judgement about pin placement can only be built through experience. You can buy the instrument. You cannot buy knowing where to put it.
Trees you cannot walk around
On Japanese street trees this becomes a practical constraint. You cannot stand on the carriageway side. There are props and stakes. Buttress roots and fused stems protrude. The planting pit is too narrow to carry sensors all the way round.
Trees you cannot instrument evenly are not unusual. When that happens, the distortion in the image comes from site conditions, not from the accuracy of the device — and recognising a distortion as a distortion sits, again, on the experience side of the ledger.
Working It Together with a Resistograph
The order is fixed
In my practice the sequence runs:
- Visual assessment — mallet sound, fruiting bodies, bark abnormalities. This is where the hypothesis is formed.
- Sonic tomography — obtain a map of the whole cross section: roughly where something abnormal sits.
- Resistograph — go and take the truth along one chosen line on that map.
- Overall judgement.
Think of it as ordering within stage three of the four stages Tokyo sets out in writing: tree inspection, visual diagnosis, instrument diagnosis, overall judgement.
The map lets you choose the line. The line lets you check the map. An instrument that reads high and an instrument that is narrow but certain are at their strongest together.
Which to buy first
If you can only have one, I would buy the resistograph. Not because it is more accurate in general, but because it is measured rather than estimated. Narrow and certain is easier to build a judgement on than broad and inferred.
That is not an argument that sonic tomography is unnecessary. For a tree nobody will let you drill, tomography is the only option. For a tree whose condition has to be explained to a committee, tomography is the only option. And where both instruments are available, the assessment is clearly better than with either alone. Not finishing the job with a single instrument is the shortest route to a correct diagnosis, and the right one.
Frequently Asked Questions
Q: Is the red region on a tomogram the actual size of the cavity?
A: No. First, the image alone does not distinguish a cavity from decay or from a crack, though overlaying visual assessment findings allows a reasonable inference. As for size, my own field impression is that it reads high, while validation work on black cherry in North America reported underestimation for decay and overestimation where cracks were present. Read the coloured area as a location, not a dimension.
Q: Is the decay percentage an objective number?
A: Not entirely. On the ArborSonic 3D, the tree species setting affects the decay and cavity percentage. Measure the same section at the same sensor positions with a different species selected and the number changes. Species selection is an input to the assessment, not a record-keeping field.
Q: Which is more accurate, sonic tomography or a resistograph?
A: They are not on the same axis. The resistograph gives measured values along the line the needle travelled — narrow, but certain. Sonic tomography covers the whole cross section, but most of that coverage is interpolated estimate. Broad and vague versus narrow and certain; neither is superior.
Q: What is the difference between the ArborSonic 3D and PiCUS?
A: In practice the difference that matters most is species parameters. The ArborSonic 3D requires species selection before measurement and pulls the crown drag coefficient and trunk yield strength from its species database into the calculation. The PiCUS 3 manual states that its species setting affects only the measuring process and not the calculation.
Q: How many sensors do I need?
A: The ArborSonic 3D supports 8 to 32. Validation on hoop pine concluded that eight are sufficient in practice for detecting decay and cavities, with R² = 0.82–0.91 on defective trees. What matters more than the count is whether you can place them evenly around the trunk, and whether you have entered their positions accurately.
Q: Does it harm the tree?
A: Sensors are driven through the bark, so small insertion points remain, but there is no boring, no sampling and no radiation. It is treated as non-invasive in practice, which is why it can be used on protected and sacred trees where drilling would not be permitted.
Conclusion
A sonic tomograph is not a camera. It measures how long sound takes to arrive and draws a map of velocity from it. The purpose of a map is not to tell you the destination. It is to tell you where to look closely.
The identity of a red region cannot be read from the image. Most of the surface is interpolated. In my experience it reads high. If you cannot carry sensors evenly around the trunk, the image distorts. Shift the pins slightly and the result shifts. Change the species setting and the decay percentage changes.
None of that is a defect. It is the outline of the tool. Know the outline and the instrument works accurately. Miss it, and you are holding the most persuasive-looking and most misleading image in tree assessment.
All of which points at one thing. This machine does not deliver verdicts. It amplifies experience. Experience decides where the pins go. Experience narrows down what the red region is, by holding it against what the trunk shows. Experience recognises a distorted image as distorted. Two people with the same instrument will not extract the same amount of information from it.
And in Japan there are a great many trees nobody may cut, and a great many people to whom the finding must be explained. Being non-invasive and being able to show a three-dimensional model earn their value precisely on those sites. Knowing what your window distorts matters more than owning the window.
Field notes from Japan’s trees arrive a few times each month — subscribe via the form in the footer, or follow the work on LinkedIn.
