Partner review build Prepared for review by tribal natural resources staff. Not for public distribution. Every layer shows habitat capability, never occurrence.

Traditional plants · habitat capability

Bagaaniminzh (Hazelnut) habitat screen

Where basin shores sit next to settings that look biophysically capable of supporting Bagaaniminzh (Hazelnut, Corylus americana) — forest edges, openings, and thickets. Capability only: this map does not show where the plant is known to grow. Region: Beltrami, Cass, Clearwater, Hubbard, and Itasca counties.

31,750

1 km land hexagons across Beltrami, Cass, Clearwater, Hubbard, and Itasca counties

9,425

hexagons with a present-day DNR Native Plant Community survey over at least 15% of their area (94,731 survey polygons)

0.72

AUC of the 1850s screen against present-day survey positives for this guild (2,368 of 9,425 surveyed hexagons; 0.5 is chance)

A diagnosis, not a score

Every hexagon gets a category and a named limiting factor rather than a bare number. Capability uses a weighted geometric mean, so a single fatal factor sinks the result instead of being averaged away — and the limiting factor reported is, by construction, the one doing the most damage.

A detected barrier never lowers capability. It changes the category and names the obstacle, because a capable site with a ditched outlet is a restoration project with a known intervention, not a bad site.

What this map does not show

This shows capability — where habitat could support the guild — not occurrence. Precise locations of active stands and gathering sites are community knowledge, not ours to publish or sharpen. No rare-plant element-occurrence records or gathering-site data are used. The DNR Native Plant Community survey describes communities, not where any plant is.

Each 1 km hexagon is scored on the plant communities mapped inside it: the present-day DNR survey where one covers at least 15% of the hexagon, the 1850s Marschner map otherwise. The Evidence control lets you look at either source alone. Treat this as a screening aid for where to look, not a statement about any particular parcel.

Method & limitations

  • Present-day communities. MN DNR Native Plant Community polygons (Minnesota Biological Survey and DNR unit surveys), generalized to about 20 m and intersected with the hexagon grid. Each guild weights community classes from the DNR field-guide species tables; the weights are provisional and are the first thing we would ask a partner to correct. Survey coverage is uneven — public and surveyed land is well covered, much private land is not — so an unsurveyed hexagon falls back to the 1850s map rather than being scored low.
  • Reference condition. Marschner's reconstruction of Public Land Survey notes (1847–1907), intersected with each hexagon.
  • Validation. The 1850s screen is tested against present-day survey positives — hexagons where strongly weighted NPC classes cover at least a quarter of the area. The survey is never an input to that screen, so the AUC shown above is an independent check of whether the historical map still predicts where the community is mapped today.
  • Soils and wetlands. USDA SSURGO map units (Web Soil Survey county packages, attributes from Soil Data Access) are classed on their dominant component and surface horizon — drainage, hydric rating, organic (Histosol) soils, calcareous or acidic reaction, sand content — and summarised as area shares per hexagon; the Minnesota updated National Wetlands Inventory supplies the wetland share and its plant communities. Each guild scores the soil condition its literature describes (sugar maple on well-drained loam, trailing arbutus on acidic sand, the fen orchid on wet calcareous or rich organic ground) with provisional bands.
  • Lidar terrain. Where etl_lidar.py has covered a hexagon it carries median slope and relief from the Minnesota Gen2 0.5 m DEM at 10 m; only the fen orchid guild scores slope so far. Hexagons without terrain are scored without it, never with a zero.
  • Near-infrared imagery. Where etl_nir.py has covered a hexagon it carries Copernicus Sentinel-2 summaries at 20 m: median greenness (NDVI) in a July–August leaf-on composite and in a snow-free leaf-off composite (late April to early May, late October to mid-November), each pooled over three years and cloud-masked; the share of land that stays green in both (evergreen canopy) or is green only in leaf-on (deciduous canopy); and how uneven the leaf-on canopy is. The pine and hardwood guilds score the canopy share their literature describes, the edge shrubs score canopy heterogeneity, all with provisional bands and weights held low until the survey cross-check earns more. Grass and crops read as green in spring and are not yet masked out. No image or pixel is published, only these shares.
  • Screening only. No entry on this map should be treated as a restoration recommendation without a site visit, and any work on or near tribal land or treaty-reserved harvesting waters belongs to the nations whose waters they are.