Seeds and cuttings I collect around North Carolina, what they turned out to be, and how to grow them for a NC Piedmont native-plant nursery.
Hydrophyllum virginianum · Hydrophyllaceae (Boraginaceae under APG)
Nothing above ground in August. April on a base-rich river terrace is the whole game, and there is no Person County record to work from.
Community photos of Hydrophyllum virginianum to study before I head out: leaf, flower or bract, ripe fruit, bark, whole plant.
iNaturalist NC, queried 2026-08-23: 240 records, 120 research grade. Person 0, Caswell 0, Granville 0, Durham 0, Orange 0, Chatham 0; Alamance 6, Randolph 1, Wake 4. GBIF herbarium: 131 NC sheets, Rockingham 31 (inflated by duplicate sheets from about three collection events), Moore 7, Lee 6, Wake 4, Alamance 3, Randolph 2, and zero from Person, Caswell, Granville, Durham, Orange or Chatham. Durham and Orange carry 322,107 plant observations between them, so that zero is real and the Eno, Duke Forest and Occoneechee should be treated as absent. Person and Caswell have only about 2,564 and 2,315 observations, so those zeros are weak: unsurveyed, not proven absent. Every NC iNat pin is coordinate-obscured by a stale geoprivacy setting, randomized 15–22 km, so herbarium label text is the only source of real localities
Best bet: Fall rhizome division. Dig wide and lift the mass, since stems snap if pulled, split so each piece has a leaf bud and roots, wash the soil off the roots for jumping worm, and pot into a shaded irrigated bed Oct–Nov. A fall division is a flowering-size plant the next spring. No published take rate exists; the nearest proxy is Iowa State's transplant establishment rating of 85–100% survival at two years. For seed, sow within days of collection into a deep shaded flat, surface-sow or barely cover 2–3 mm, mulch with about an inch of leaf litter, and never let it dry
Native to the NC Piedmont and hardy here, USDA 4a–8b, so 7b/8a sits near the warm edge. Heat isn't the limit, moisture is. It needs shade, 2–6 hours of sun at most, and reliable irrigation. Tolerates clay if the clay stays moist, and wants pH 6.0–8.0 or better still circumneutral 6.8–7.2, so lime lightly and compost heavily on acid Piedmont clay running 5.2–5.8. Expect midsummer dieback on anything unirrigated, which is a sales problem (empty pots in Jul–Aug) rather than a killer. Rhizomatous and aggressive, so give it its own bed with edging. Wild collection should stay light: Piedmont populations are small and local, formerly NCNHP watch-listed, and there is no record anywhere near Roxboro
Each look-alike with the one character that settles it. Tap a photo to zoom.
Read the leaves on the flowering stem, not the rosette. Canada waterleaf's principal stem leaves are palmately lobed, maple-shaped, about as wide as long, with blunt lobe tips; virginianum stem leaves stay pinnately divided into 5–7 acute segments, longer than wide. Backup: in canadense the peduncle is shorter than the petiole of the leaf below it, so the clusters hide under the top leaves, while virginianum holds the cyme clear. 615 NC records, only 1 across the nine central-Piedmont counties.
Run a hand up the stem. macrophyllum is densely covered in coarse spreading white hairs and feels rough; virginianum is smooth or has only short hairs lying flat. Backup: 9–13 leaf segments against 5–7. NC counties are mountains only, so the single Person County iNat pin (obs 41075005, still needs_id) contradicts every voucher-based list.
The style. Phacelia styles divide about halfway down, Hydrophyllum styles divide only at the tip, and both are long enough to see the fork without a lens. Backup: bipinnatifida opens a flat 1.3 cm blue-purple corolla with a white center where waterleaf keeps a 6–12 mm bell whose lobes barely part, and it's pubescent throughout and blooms Apr–May. 87 records in the nine counties, so this is the likeliest false positive on a Uwharrie or Eno slope.
Size and flower count. covillei is a tiny annual under a foot with flowers under 1 cm in loose 2–6 flowered cymes and stamens not conspicuously out; virginianum is 30–80 cm with 8–20 flowers per dense cyme and stamens out 4–9 mm. Backup: rounded buttercup-like terminal lobe, no white blotching, blooms Mar–Apr. NCNHP SR-T, S3, G3, and 140 of its 151 NC records sit inside these nine counties. Do not collect it.
Hold a leaflet to the light and follow the veins. In Cicuta the veins run to the notches between the marginal teeth; in Hydrophyllum they run to the tooth tips. Backup: Cicuta's stem is hollow, purple-striped or mottled, cross-partitioned at the nodes, over short tuberous roots, with twice or thrice divided leaves and flowers in umbels. Waterleaf stems are solid and unspotted over fibrous roots and rhizomes, leaves once-divided, flowers in cymes. 132 records in the nine counties.
The stem is hollow with distinct purple blotches and the crushed plant smells foul and mousy; waterleaf stems are solid, unblotched and odorless. Backup: leaves three to four times pinnately dissected to parsley fineness, and it wants disturbed ground, roadsides, fencerows and ditch banks rather than closed rich forest. 27 records in the nine counties and confirmed spreading in Orange.
The leaf is simple and undivided, kidney to heart-shaped, glossy, with an entire to bluntly scalloped margin; waterleaf's leaf is cut to the midrib into 5–7 toothed segments. Backup: glossy yellow buttercup flowers on single stalks, tubers and bulbils at the base, and the whole plant is gone by late spring. 2,499 records in the nine counties, more than every native look-alike here combined, and documented on Ellerbe Creek and the Eno at Penny's Bend. A dense carpet on Piedmont bottomland in March is this until proven otherwise.
Palmate against pinnate. Geranium segments all radiate from a single point at the top of the petiole; waterleaf segments sit in pairs along a central axis with a terminal segment. Hold the leaf up: one origin point, or a row. Backup: pink-lilac five-petaled saucer flowers about an inch across with stamens not protruding, then the long beaked cranesbill fruit. 393 records in the nine counties.
Trout lily leaves are simple and unlobed, two smooth strap blades with no teeth and no divisions, mottled brown-purple; waterleaf leaves are cut clear to the midrib into sharply toothed segments and blotched whitish-grey. Backup: no true stem, the two leaves rise straight from the ground beside a nodding yellow flower. 1,649 records in the nine counties.
Two things to settle before anything else, and both change what this guide is for.
First, timing. Late August is close to the worst week of the year for this plant. It's aestivating, died back to the rhizome. Flowering ended around mid-May and the seed dropped somewhere between June and July, so there is no 2026 seed left. A hike this week finds nothing above ground, or brown collapse that can't be taken to species. Mid-September through October gets new basal rosettes, which is good enough for reconnaissance and a GPS pin but not for a clean ID. April 2027 is the real trip.
Second, range. On current evidence this isn't a Person County plant. Zero iNaturalist records and zero herbarium sheets from Person, Caswell, Granville, Durham, Orange or Chatham. The Durham and Orange zeros are real absence — 322,107 plant observations between them across the Eno, Duke Forest and Occoneechee, and not one waterleaf. The Person and Caswell zeros are weak, about 2,500 observations each, so unsurveyed rather than proven absent. If I hunt it here I'd work the lower Hyco and Mayo Creek bottoms and the bluffs above them in April, and understand that a find would be a county record: photographs, a location note, herbarium-grade images.
The landform is specific. Base-rich alluvium: a low river terrace, a seepy toe slope, and the bluff base behind it, roughly the first hundred meters off the bank of a river or large creek. Circumneutral loam, deep leaf litter, closed hardwood canopy. Every NC Piedmont station sits on Triassic basin, calcareous or metavolcanic substrate. Heavy garlic mustard is a negative indicator; populations decline under it.
In April a colony reads as a green carpet with pale dots floating above it, because the peduncles are 10–15 cm and hold the flower clusters clear of the leaves. In March the young leaves carry the whitish water-stain blotches that name the plant. Both of those disappear by midsummer.
Seed window is genuinely unsettled and I'm not going to pretend otherwise. Vascular Plants of NC says fruits July to August. LBJ says capsules mature 8–10 weeks after flowering. Iowa State and Larkspur say late June to early July, PFAF says July–August, and both are northern. Working plan for the Piedmont, where flowering is 4–6 weeks ahead of the NC mountains and about a month ahead of Midwest guides: start checking capsules roughly five weeks after peak bloom, so mid-May, and re-check every three or four days. They split and drop fast, and the plant is senescing while they ripen.
Clip whole cymes while the capsules are tan-brown and just starting to split. Each round hairy capsule is 4–5 mm and holds two to four brown wrinkled seeds at about 2,500 per ounce.
Seed handling is the real constraint. This is not orthodox seed. Cullina and the NCBG manual both say it won't tolerate dry storage: sow outdoors immediately after collection or hold in damp sphagnum. Prairie Moon does sell dry-handled seed stored cold, so some drying is survivable, but nobody publishes a viability half-life. I'd assume one season. Sow within days of collection and never let the flat dry out.
Germination has three published positions: 60 days cold-moist at 33–40°F; 90 days at 40°F then a shift to 70°F; or deep simple epicotyl dormancy, two to three months warm to push the radicle out in late fall, then cold, with the shoot arriving the following spring. Sowing fresh in summer satisfies all three, which is the resolution and why it beats the refrigerator. Emergence comes the following spring. There is no published germination percentage, no days-to-germinate figure and no viability duration for this species; Prairie Moon flags it as possibly difficult and slow.
Fall rhizome division is the reliable route and the one I'd actually plan around. Dig wide and lift the mass, because the stems snap if pulled. Split so each piece carries a leaf bud and roots, wash the soil off for jumping worm, and pot into a shaded irrigated bed in October or November. A fall division is a flowering-size plant the next spring. No published take rate, but Iowa State rates transplant establishment at 85–100% survival at two years and PFAF calls division very easy.
One safety note that runs backwards from most plants. Virginia waterleaf is edible — Shawnee salad, John's cabbage, young leaves and flower buds raw — which is exactly why a rosette misidentification here matters. Never eat a waterleaf rosette I haven't seen flower. Cicuta and Conium share the ground and the season and both will kill. Bloodroot and white baneberry are in the same woods too.
Best-documented target on the list is the Haw River upstream of the Union Bridge paddle access, Old Greensboro Highway / SR 1005 on the Alamance–Orange line. An NCU voucher from 2020-09-13 (sterile), two NCU sheets from spring 1952 labeled west bank of the Haw 12 miles west of Chapel Hill, plus iNaturalist records from 2022-09-10, 2023-04-21, two on 2025-10-04 and one on 2026-04-17. That corridor covers both the April flowering window and the September–October rosette window, and it's the same trail system I already work at Saxapahaw.
Closest in raw distance is the Smith River at Eden and Matrimony Creek in Rockingham County, about an hour west and in the same Dan River drainage. The labels read "rich woods above field on E bank of Smith River just S of N.C. 14, in Leaksville" (1966) and "along Smith River, shale slope" (1968). Mayo River State Park sits in the same county, drainage and geology with no record of its own, which makes it worth a look.
After that: Umstead State Park in Wake, Richland Creek and the Buckeye Trail. Caraway Creek and Caraway Mountain in Randolph, the closest documented material to Uwharrie though north of the forest proper. The Deep River Triassic basin in Moore and Lee, repeatedly recollected, all April dates.
Every NC iNaturalist pin for this species is coordinate-obscured, randomized inside a cell 15–22 km wide, left over from a retired state rarity status. County assignment uses the true coordinates so the counts hold, but there's no usable waypoint. Herbarium label text is the only route to a real locality, which is why the list above is written out of label transcripts rather than map pins.