Lobelia Silhouettes | Jaret Arnold

TL;DR. A pressed herbarium specimen is a plant folded, broken and taped flat to fit a sheet. What a reader usually needs from it is its habit, the shape of the whole plant. Over 2018–19 I digitally restored Lobelia specimens in Photoshop and reduced each restored plant to a silhouette: 39 silhouettes across 27 species, every one scaled from the ruler photographed on its own sheet. Godden et al. 2025 credits them in its acknowledgements.

📄 The paper they appear in ⬇️ Download the set (CC BY 4.0)

Twenty-seven black plant silhouettes in a row on a common baseline, ordered left to right from shortest to tallest, from Lobelia feayana at 14 cm to Lobelia brevifolia at 134 cm, with a 50 cm scale bar. — **Every species in the set, at true scale**, ordered by height and resampled to a common scale using the **10 cm bar drawn into each file** — *L. feayana* at 14 cm to *L. brevifolia* at 134 cm. The plants never stood together, but their relative sizes are real.

This was undergraduate work at Kent State University with Dr. Andrea Case, and the earliest published piece of my involvement in the lab’s Lobelia project. The automated leaf-measurement work came several years later.

How they were made

Each silhouette starts as one digitized sheet, restored in Photoshop before it is reduced to an outline. The restoration puts the plant back into the shape it had before pressing:

Unfolding. Leaves pressed back over themselves, or folded to make the plant fit the sheet, are digitally unfolded.
Reattaching. Stems snapped in pressing or handling are rejoined.
Removing the furniture. Labels, tape, mounting strips, annotations and colour cards come out, so what remains is only the plant.

The whole restored plant, not a traced approximation of it, is then converted to a mask, and that mask is the silhouette. Two constraints separate this from illustration:

Nothing is added that was not already on the sheet. No compositing across specimens, no organ borrowed from a better example, no idealised representative. If a part was missing from that sheet, it is missing from its silhouette. Each one is a restoration of a particular plant, not a drawing of a species.

Scale comes from the sheet’s own ruler. Each silhouette carries a 10 cm bar drawn into the file, so true size is checkable from the images instead of taken on trust.

That is what makes the plate above possible. The files run from 30 to 85 pixels per centimetre, so it is built by reading each bar and resampling everything to one shared scale.

Explore the set

Every restored specimen, on one centimetre axis. Hover any plant for its species and measured size. Six reach past Lobelia sect. Lobelia as scoped in 2021: L. amoena, L. anatina, L. berlandieri, L. boykinii, L. fenestralis and L. laxiflora.

All 39 specimens at one shared scale, axes locked to equal aspect. Scroll sideways to reach L. brevifolia. The three L. cardinalis stand at 24, 47 and 52 cm — a spread between specimens, not a claim about the species.

How the set measures against real plants

Everything here is scaled from a ruler, so the obvious check is whether these plants come out the size real ones do. Restored heights against published stature, for the two species with a published reference:

species	restored	published
L. siphilitica	61 cm	46–122 cm, inside
L. cardinalis (×3)	24, 47, 52 cm	122–152 cm, far below

The disagreement is the informative half. A herbarium sheet holds what fits on a herbarium sheet. For a tall species the mounted specimen is often a portion of the plant, or a small individual. So a silhouette faithfully depicts the pressed specimen, not the species: read every height here as the height of what was mounted. The reference is horticultural and covers 2 of 27 species.

Where the specimens came from

Source sheets are filed by herbarium accession and state: LSU00039459Louisiana, USMS_000008523Mississippi, IND-0119684Mississippi. So every silhouette traces back to a physical specimen, and the collection localities are on record even though the photographs cannot be reproduced here.

Collection localities for 60 source sheets across 18 states. Accession records cover six species (brevifolia, canbyi, cardinalis, dortmanna, elongata, feayana), so the map shows those, not the whole set.

Where they ended up

The silhouettes are credited in the acknowledgements of the phylogenomic paper from the lab’s programme (Godden et al., 2025), where they form the morphology panel of the phylogeny figure: a band of plants on a shared ground line, each aligned over its own clade. That arrangement is what the true-scale rule was for. At the tips of a tree, one species reading knee-high beside another reading waist-high is the comparison being made, not styling.

Archived, and free to use

Arnold, J. (2026). Lobelia silhouettes: digitally restored whole-plant outlines from herbarium specimens (2018–2019). Zenodo. 10.5281/zenodo.21764522

The 39 originals, the plate, and a manifest of every measured height, all CC BY 4.0. Use them in a figure, a talk or a key; attribution is the only condition. The deposit holds no specimen photographs and no layered working files, both of which embed institutional imagery. What is released is the outline and the restoration work.

Both this and the automated leaf-measurement work end in a plant reduced to an outline, which is why they blur together. This one is reconstruction for depiction and keeps true size; that one is segmentation for measurement and normalises size away.

Status

Completed undergraduate work, recorded here rather than maintained. The silhouettes are my own work; the specimen images they were restored from belong to their holding institutions and are not reproduced here.