Meristem is a field instrument for plant professionals. In the offline modes, verdicts are computed against published standards rather than guessed. In the cloud modes, a vision model runs on your own AI key. This guide walks through getting started, every mode, and the resistograph module.
Sign up and start a 14-day free trial with full access to every mode. A card is required, with no charge for 14 days.
Snap or upload a photo, or enter a few field measurements. Works on any phone, online or off.
Every result shows the standard it used and the inputs behind it, so the number is defensible.
Save it to history and build a branded PDF, or email the report straight to a client.
For a full-screen, offline-capable experience, install Meristem to your device. On a phone, use Add to Home Screen from the browser menu. On desktop Chrome or Edge, use the install icon in the address bar.
Pick the modes you want from the chips at the top, all five or only the ones that fit the plant, take one photo, and tap Start scan. Meristem runs your selected modes in sequence, passing the identified species to every downstream mode, so diagnose, prune, risk, and grade all know what plant they are assessing. Results appear as they finish. Save any card on its own, or save the whole workup to history in one tap. A summary shows health, risk rating, prune grade, and nursery grade at a glance, and a full analysis report bundles every mode plus a care guide into one PDF. When you're done, tap Scan another at the top of the results to clear everything and capture the next plant in the same mode; scanning plant after plant is one tap between.
Names a plant from a photo.
Fill the frame with a single specimen and capture. If the top match is off, tap the correct one from the alternatives, or search and pick the exact species. Every result is checked against the GBIF taxonomy for the accepted name.
Standard: Cloud vision model, an on-device Florida species model, or the free Pl@ntNet path when configured.
Works up pests, diseases, and abiotic disorders.
Use AI photo for findings tied to visible evidence and a primary diagnosis, or describe the symptoms for the offline lookup, which returns ranked candidates with treatment notes.
Standard: Vision model plus a curated Florida knowledge base.
Grades a pruning cut.
Enter the cut type, collar position, stub length, live crown percentage removed, or palm frond status. Or let AI photo read those fields from the image. Either way the engine assigns the A to F grade from the rubric. The AI never grades directly.
Standard: ANSI A300 Part 1, ISA BMP, and UF/IFAS palm best practice.
Rates tree risk.
Enter likelihood of failure, likelihood of impact, and consequences. The two ISA TRAQ matrices compute a Low, Moderate, High, or Extreme rating for you. Trees only.
Standard: ISA TRAQ likelihood and consequence matrices.
Grades nursery stock.
Choose a palm or an ornamental or shade tree and enter the criteria. Eliminating factors make a plant a cull; otherwise the per-species leaf count (palms) or a six-criterion rubric (trees) sets the grade.
Standard: FDACS Florida Grades and Standards.
Meristem reads drilling-resistance bores from the IML RESI PD500 and keeps them on the trees they were taken on. The software measures each trace and proposes; the operator rules on every bore. Nothing in the module decides that a tree is decayed, sound or safe.
In the field, the Resistograph screen pulls the day’s measurements off the instrument over USB: in Chrome on Android with a USB-C or OTG cable, or in the Windows desktop app. iPhone and iPad expose no USB device API, so use them to review and report. At the desk, the console imports the .rgp files PD-Tools writes, repairing the unescaped inch mark the vendor software sometimes leaves in a label. Bringing a measurement in again refreshes its instrument data and keeps the tree, side, height and notes already set. Only read commands are implemented, so nothing here can alter or erase the instrument.
Each trace is read against a sound-wood baseline fitted at every depth, because resistance climbs along the bore as the shaft rubs in its own hole. Runs below 40% of the local sound level for at least 10 mm are proposed as candidate low-resistance regions, each with the settings that produced it and a plain reading of its shape. Depth into the trunk is reported separately from depth along the needle, corrected for tilt, so an angled root-flare bore does not overstate the wall. A candidate is not a finding: the operator accepts, adjusts or leaves each one and saves the review, and saving with nothing accepted records that the trace was examined and none of it judged a defect.
Every bore is checked for conditions that limit what it can show, each graded Blocking, Limits or Note and paired with the action to take. Examples: the needle never reached sound wood; the instrument recorded the run as stopped early, so the depth is a floor; the bore met no decay and ended short of the radius, so its wall is a floor rather than a measurement; no compass side is recorded, so the bore cannot be placed on the trunk; no trunk diameter; a bore deeper than the recorded trunk; a demonstration run; sound wood outside the 40 to 60% amplitude band IML recommend. Reading the feed and drill curves together adds the blunt-needle signature, local feed overlap (which IML associate with early white rot or reaction wood), cracks and compartmentalisation barriers.
Each proposed region is checked against the other bores on the same tree: those linked to it, or carrying the same tag on the same field day. A neighbour counts as the same side when its bearing is within 45 degrees and, where both heights are recorded, within 50 cm. Comparison is by radial depth. The region is reported as corroborated when a neighbour drops at the same depth, contradicted when a neighbour drilled past that depth and read sound, or alone when nothing on that side reaches it. A bore whose sound level disagrees with the others taken that day is flagged separately.
The tree page reads every linked bore together. A cross-section is drawn from the field day and level with the most sides bored, one bore per side, by interpolating between the bore lines. Between bores it infers rather than measures, so it is shown with its coverage, a confidence grade and the widest unmeasured arc; decay sitting in an unbored arc does not appear. Bores left out of the drawing are listed with the reason, and unreviewed bores in it are counted. Where a side has been bored at the same mark on two dates, the change in sound shell is reported.
A wall ratio (t/R) and Wagener loss are derived only from bores an operator has reviewed; an unreviewed bore shows why no ratio is given. A bore that met no decay and stopped short is shown as at least its depth, a floor. Wagener’s d³/D³ and IML’s remaining stability assume a round trunk with a centred cavity; the section modulus and sound area integrate the drawing and are only as good as its coverage. The figures are reported side by side because they disagree on an off-centre defect. The t/R 0.3 rule is contested and no threshold is applied.
A bore plan for the tree spaces positions at breast height by circumference, roughly one every 60 cm around the trunk, three at minimum and up to eight on a large trunk, and marks each position done when an existing bore at that side and height covers it. Without a diameter the plan asks for the trunk to be taped first. Feed-stage advice aims the curve at IML’s 40 to 60% band.
On a palm the drilling policy comes first: do not bore as a matter of routine, because the wound is permanent and most palm problems do not change wood density. Palms have no cambium and no annual rings, and a healthy trunk is dense at the periphery and soft toward the centre, so the wall, Wagener, remaining-stability and section-modulus figures are flagged as not applicable and no ring count is given. Bark reads as low as decay, and on live oak it runs one to three centimetres. Every bore therefore finds where the needle meets wood, holding half the sound level for 10 mm within the first 4 cm; the hold is long enough not to take the live oak cambium spike for wood. Nothing before that point is proposed, and the span is offered as the bark and entry zone, which can still be accepted if what was drilled was decay. On other species, a ring count from one radius is a minimum age at that height, and none is claimed on less than 30 cm of wood or where no dominant period shows.
The analysis supports reading a trace against an increment core taken beside the bore, with the depth axes aligned. A core is wood that can be looked at. Stain usually runs ahead of the density front, so firm, discoloured wood reading sound is expected and is not a disagreement. Only two outcomes count as one: the trace read sound through wood the core shows broken down or hollow, or the trace proposed a region against clean wood.
Every bore has a PDF report, previewed, downloaded or emailed from the bore page, with the trace, the instrument’s own figures, the candidate regions with their settings and the provenance of the file, and no verdict box. The tree page builds a tree report with the cross-section, the strength figures and their caveats, flags across the trunk, the bore plan and change over time. Reviewed bores travel into the tree’s risk assessment PDF as instrument evidence; unreviewed ones are counted there, not presented. The archive downloads as CSV, one row per bore with each derived figure beside its inputs, for the whole archive, one tree, or one tag on one day.
Limits: Drilling resistance measures density along one line, and strength is lost before density is. A trace that reads sound means the density along that line is unremarkable, not that the wood has kept its strength. Early soft rot, which embrittles wood without removing much of it, and vascular diseases that leave the wood intact do not show in a trace. The figures support a tree risk assessment and do not replace one.
Prune, Risk, and Grade are deterministic and need no key or connection. Diagnose has an offline symptom lookup, and Identify has an on-device model you download once and then use anywhere. When a cloud scan fails, the error offers a one-tap switch to the offline path. Assessments save on the device and sync to your history when a connection returns.
The cloud modes (Identify and Diagnose) use a bring-your-own key for Claude or GPT. It is stored only on your device and sent only to run a single scan. Meristem never holds, meters, or resells your key. Your scans and photos are private, and you can export or delete everything at any time.
Every result card builds a clean PDF report on the device, with no key or connection required. On a phone, reports go straight to the native share sheet; on desktop, a save dialog. Email a report straight to a client as an attachment. After an Auto run, a full analysis report bundles every mode for the plant into one document. Assess two or more items and a session report bundles them into a single log with a verdict and key finding per item. Capture your signature once in Settings and it drops into the certification block on every report you choose to sign.
A seed tag on a result means the values were transcribed from the source but not yet confirmed against the master document by a qualified professional. Confirm before relying on a seed verdict in the field. Meristem is decision support and does not replace a licensed inspection or your own field verification.
Start a 14-day free trial and run your first assessment in the field today.