Capturing Trees

This is what you do every day in the field. Open the app, walk to a tree, take a photo. The app handles everything else automatically — GPS, species, DBH, height, condition, recommendations.

Quick Start

• Open an address (the property you're inventorying).
• Tap the green + button at the bottom right.
• Take a photo. Wait a few seconds for AI analysis.
• Edit any field if the AI is wrong. Save.
• Repeat for the next tree.

What gets captured automatically

When you take the photo, the app captures:

• GPS coordinates — pulled from your phone's location services. Accurate to ~3-5 meters in the open, less under heavy canopy.
• Photo with EXIF — full-resolution image, timestamped.
• Capture timestamp — used later to group trees into "visits" (a 30-minute window of capture activity at a property).

What the AI infers from the photo:

• Species — common name + scientific name (e.g. Acer saccharum — Sugar Maple).
• DBH (diameter at breast height) — estimated from visual cues. For an exact measurement, use a DBH tape and edit the field.
• Height — estimated. Use a clinometer if you need exact.
• Crown spread.
• Condition — overall health rating (Good / Fair / Poor / Dead).
• Visible defects — co-dominant stems, included bark, deadwood, cavities, lean, soil heave.
• Recommendations — pruning, removal, monitoring, treatment.

Editing the capture

After the AI returns its guess, the tree detail screen shows every field. Tap any value to edit:

1. Species

   Tap the species name. A search box appears. Type the common or scientific name to filter; pick from the list. Or type a custom species if it's not in the database.

2. Measurements

   DBH, height, and crown spread are number inputs. Type the actual value — units are inches and feet (US) by default.

3. Condition + defects

   Select the condition rating from the dropdown. Defects are a multi-select: tap any that apply.

4. Notes

   Free-form text. Use this for anything the structured fields don't capture — easements, access constraints, client preferences ("don't touch this one — sentimental").

Multiple photos per tree

You can take more than one photo of the same tree. Useful when:

• One photo shows the trunk and another shows the canopy
• You want a "before" markup of a defect plus a wider context shot
• The first photo had bad lighting

1. From the tree detail screen, tap the photo

   The photo viewer opens.

2. Tap the camera icon to add another photo

   The capture screen opens. Take the new photo. It's added to the carousel.

3. Swipe between photos in the carousel

   The current photo (the one displayed) is what the AI re-analyzes when you tap Re-analyze. It's also what appears in the PDF report.

Voice notes per tree

Hold-to-talk on the microphone button records a voice memo for the tree. Whisper transcribes it; Claude extracts structured fields (e.g. "co-dominant stems with included bark, recommend cabling" → defects: [co-dominant-stems, included-bark], recommendations: cabling).

A dedicated Voice Notes doc lands in Tranche 2 — for now, the short version: hold the mic in the tree detail screen, talk, release. Whisper transcribes; Claude classifies into structured fields.

Pin numbers

Each tree gets a unique pin number per address (1, 2, 3, ...). Pin numbers are how Arboris (the AI assistant) refers to trees in conversations: "Tree #4 at 12 Maple St has co-dominant stems."

Pin numbers are NOT global — Tree #4 at one address has nothing to do with Tree #4 at another. The address scopes them.

Common Questions

What if GPS is wrong (heavy canopy, urban canyons)? Tap the location field on the tree detail screen and drag the pin to the correct spot on the satellite map.

Can I capture without taking a photo? No. The MVP requires a photo per tree because the AI needs visual input to infer species + condition. A "manual entry" mode is on the roadmap.

What about trees in groups (clumps, hedges, screens)? Capture them as separate trees (each gets its own pin). For very dense screens (e.g. a 30-tree boxwood hedge), you can capture one representative tree and note the count in the notes field — but each tree in the report is a single row.

How long does AI analysis take? 3-8 seconds in good network conditions. The capture is saved immediately on the device; AI fields populate when the response returns. You can keep capturing while previous analyses are still running.

Related

• Photos and Markup — annotate photos with arrows, text, shapes
• Generating Reports — what the captured data looks like in a PDF
• Hazard Rating (TRAQ) — when capture turns into a formal risk assessment