Measuring Root System Architecture 🌱🪜¶
Root System Architecture (RSA) is the shape of a plant's root system — how long the roots are, how many tips and branches they have, and which way they grow. In Stage IV you watch roots respond to gravity; this page shows how to measure that response instead of just describing it, and turn a photo into numbers you can graph.
Great photos make great measurements
Root measurement only works on clean, high-contrast images. Grow roots against a plain background (black felt, or agar over a dark surface), keep the camera straight-on, and put the AstroCalibration marker flat in the frame — exactly as in Stage I: Scientific Photography. The marker does two jobs: it fixes the colours and it sets the scale (pixels → millimetres).
How it works, in four steps¶
Every root-measurement tool — whether it's a simple threshold or a trained neural network — follows the same idea:
- Calibrate. Use the marker to correct colour and learn how many pixels equal one cm.
- Segment. Separate the root from the background (find which pixels are "root").
- Skeletonize. Thin the root down to a one-pixel centre-line — its "wiring diagram".
- Measure. Walk the skeleton to add up length, count tips and branch points, and read the angle each root points.
You don't have to do these by hand. The AIR Root Tool does all four for you.
Use the AIR Root Tool¶
🛰️ Open the AstroRoot Tool
A free, browser-based tool (no install — works on a school Chromebook) that reads your seedling photos and returns the numbers below. Built on the open RootNav 2.0 root-tracing model. Your images stay on your device.
→ Launch AstroRoot · source & guide · dashboard
No sign-in, no upload — everything runs in the browser tab and your images stay on the device. It also comes with real NASA spaceflight demo images built in — ISS root plates from the ABRS time-lapse and the APEX-03 skewing-mutant experiment (Col-0, WS, spr1, sku5, flight vs ground) — so you can try the whole workflow before you take your own photos.
What the tool can do¶
Start simple, then add only the pieces your experiment needs.
Open a photo (or a demo image) and press Trace roots — the tool finds the roots, draws them on top, and reports the numbers. Choose the classical engine (works instantly) or the RootNav 2.0 AI model. Anything it gets wrong you can fix with the Manual trace editor, which has the tracing tools of RootNav 1, NeuronJ and SmartRoot all in the browser:
- Draw / Edit — drag, add or delete the dots along a root, and branch a lateral off a parent.
- Route — click a root's start and its tip and it draws the line along the root for you.
- Magnetic (live-wire) — the line snaps along the root as you move the mouse; click to set it (hold Shift for a straight line in tricky spots).
- Refine — snaps your dots neatly onto the centre of the root and measures its width (so you also get diameter, surface area and volume).
- Smooth — draws the roots as smooth curves.
When you're done, scroll to the bottom for a downloadable report — a tidy web page with your measurements, per-region and per-plant figures, and the traced image (great for a lab book, or print it to PDF).
Put an AstroCalibration marker in the shot and pick it under Marker → Auto-detect. The tool reads the card to turn pixels into millimetres and to colour-correct the photo — the same idea as PlantCV's colour-card detection.
Growing on gridded plates? Tick remove plate grid and the etched grid is filtered out before measuring, with a before/after preview. It's not always perfect — you can still fix the trace by hand.
If one image holds two genotypes (say one that skews and one that doesn't), click ✏️ Draw region, box each area and name it. Each region is measured separately, so you can compare — e.g. the skewing line's angle vs the straight one's.
Tell the tool how many plants are in the image and mark where each seed was sown (the root's starting point) — or let it auto-place them. You then get a row per plant, including its skew (how far its root drifts left/right from the seed). If you also drew genotype regions, each plant is automatically labelled with its genotype.
Photograph the plate every day, then in Batch load the whole folder. The regions and seeds you set once are applied to every frame, giving each plant's growth over time.
The dashboard collects every measurement you save: summary stats, charts, a searchable table, group summaries (compare genotypes), and a per-plant trajectory graph (each plant, and the genotype average, over the days). Download it all as a spreadsheet — or, if your teacher sets it up, pool the whole class's numbers together.
What you can measure¶
| Trait | What it tells you |
|---|---|
| Total root length | Overall root growth / vigour |
| Number of tips & branch points | How branched / bushy the root system is |
| Root angle from vertical | The gravitropic response — how roots reorient after the tray is turned 90° |
| Skew (per plant) | Which way, and how far, each root drifts from straight down — the trait that separates a skewing genotype from a straight one |
| Length / diameter / surface / volume, convex-hull, tortuosity | Full root-architecture traits when you import a traced RSML file (matching the archiDART trait set) |
| Root depth-distribution | How much root sits at each depth — a density-vs-depth profile you can chart and compare between plants |
| Estimated laterals (AI guess) | Laterals overlap in a photo, so the tool also estimates the number of laterals, their angle, and the "lateral fraction" from the root's shape. Treat these as a helpful estimate, not an exact count — the tool tells you so. The AI was trained on tidy single-root pictures, so on a crowded plate it knows it's out of its depth: instead of printing a wrong number it says the estimate isn't reliable and suggests you crop to one root first. |
Tie it back to gravitropism (Stage IV)¶
This is the measurement that makes Stage IV quantitative. Photograph your seedlings, rotate the gravity vector 90°, then re-photograph at set times. Measure each photo and compare the root angles before and after — that change is the gravitropic response, now a number you can graph and share.
Coder appendix — do it yourself in PlantCV
If you'd rather run the analysis in code, PlantCV does all four steps. The thresholds below are starting points — you will need to tune them to your own lighting and background.
from plantcv import plantcv as pcv
import numpy as np
# 1. Read + colour-correct with the AstroCalibration marker.
img, path, name = pcv.readimage(filename="seedling_day5.jpg")
img_cc = pcv.transform.auto_correct_color(rgb_img=img, color_chip_size=(10, 10))
px_per_cm = 60.0 # <-- calibrate from YOUR marker
# 2. Segment the root (try LAB 'l' or 'a'; tune the threshold to your images).
gray = pcv.rgb2gray_lab(rgb_img=img_cc, channel="l")
mask = pcv.threshold.binary(gray_img=gray, threshold=115, object_type="dark")
mask = pcv.fill(bin_img=mask, size=30)
# 3. Skeletonize + prune.
skeleton = pcv.morphology.skeletonize(mask=mask)
pruned, seg_img, seg_objs = pcv.morphology.prune(skel_img=skeleton, size=30, mask=mask)
# 4. Measure: length, tips, branch points, angles.
pcv.morphology.segment_path_length(segmented_img=seg_img, objects=seg_objs)
pcv.morphology.find_tips(skel_img=pruned)
pcv.morphology.find_branch_pts(skel_img=pruned)
pcv.morphology.segment_angle(segmented_img=seg_img, objects=seg_objs)
pcv.outputs.save_results(filename="rsa_day5.json", outformat="json")
Docs: morphology workflow · segment_path_length · skeletonize · auto_correct_color. Argument names change between PlantCV versions — check the docs for your version. For the leaf/canopy equivalent, see Green Area Index with PlantCV.