Skip to content

Imaginary Maine parks for Lab 10

ME_ImaginaryParks.geojson contains four fixed imaginary park polygons and 1,500 occurrence points representing 67 species: 17 observed species names and 50 invented species. It preserves all 846 real in-park records from amphibians.geojson and adds 654 simulated points. Coordinates are WGS84 longitude/latitude (EPSG:4326).

Park Strategy Area (km²) Records Species
SL — Alder Basin Preserve single_large 1,000 819 55
SS1 — Fern Hollow Woods several_small 400 252 42
SS2 — Coldwater Ridge Preserve several_small 300 192 49
SS3 — Spruce Lantern Reserve several_small 300 237 47
Three small parks pooled several_small 1,000 681 60

Park boundaries and locations are unchanged. Every occurrence has its park_id, park_name, and scenario assignment already stored.

Empirical range and abundance templates

  1. Load all 14,577 real occurrences in amphibians.geojson. Retain the 846 in-park records with their coordinates, GBIF IDs, species names, and other source attributes unchanged.
  2. For every observed species, calculate the convex hull of all its source points and buffer it by 50 km in EPSG:26919 (NAD83 / UTM zone 19N). One-point hulls become circles and two-point hulls become buffered lines.
  3. For each invented species, sample one observed donor species uniformly with replacement. Donors must have a buffered hull intersecting at least one park. The invented species inherits the donor's buffered geographic range and statewide record count together. This samples the empirical species-level distribution, including rare species, without imposing a separate lognormal model or weighting common donors twice.
  4. Reserve one point per invented species, ensuring that all 50 appear. Allocate the remaining 604 points proportionally among real and invented species using observed record counts and inherited donor counts as weights. Use largest-remainder rounding to make the final total exactly 1,500.
  5. Sample uniformly within each species' buffered range intersected with the four parks. Choose among park pieces in proportion to eligible area and sample uniformly inside the chosen piece. Invented species use their donor's range; simulated records of real species use their own range.

The abundance weights come from the observations, while the minimum of one record per invented species guarantees representation at this small sample size. Final species totals also include the original local observations. Some observed species receive zero added points after rounding. No park richness targets are imposed. Range overlap, geometry and sampling determine which parks share species.

Seed 597 makes the bootstrap and point sampling reproducible. Invented names use the genera Imaginarana, Fictotriton, and Sylvophantoma. The added points are an instructional simulation, not new observations or a habitat-suitability model.

The FeatureCollection's simulation metadata records the donor for each invented species, abundance weights, source and added counts, eligible areas, CRS, buffer distance, and seed. Buffered hulls are not added as features to this file; it still contains only four park polygons and the occurrence points.

Provenance

  • feature_type: park or occurrence.
  • synthetic: false for the 846 source occurrences, true for added points.
  • imaginary_species: true only for the 50 invented taxa. Simulated points using an observed species name have synthetic=true, imaginary_species=false.
  • donor_species: the observed range/abundance template for each added point; for simulated records of an observed taxon, this is its own species name.
  • record_source: amphibians.geojson or convex_hull_buffer_simulation.
  • Synthetic occurrenceID values start with SYNTH-ME-; synthetic gbifID values are unique negative placeholders, not real GBIF keys.
  • Synthetic year and coordinateUncertaintyInMeters are null because there is no survey event or measured coordinate uncertainty.

The fictional field refers to the park design. Park properties include area_km2, center_lon, and center_lat.

Rebuild and experiment

To regenerate records while keeping the current parks fixed:

python scripts/augment_imaginary_parks.py

The function always reloads real records from amphibians.geojson; existing occurrences in the input park file are ignored. Rerunning replaces the previous augmentation rather than adding to it.

from scripts.augment_imaginary_parks import augment_imaginary_parks

features = augment_imaginary_parks(
    total_records=1500,
    imaginary_species_count=50,
    buffer_km=50,
)

This returns a GeoDataFrame without saving. Supply output_path to save an experiment separately. Other parameters are input_path (the park file), source_path (real occurrences), and seed. Set imaginary_species_count=0 to use observed species only. Up to 50 invented species names are available.

The requested total must accommodate all original in-park records plus one record for every invented species. Invalid totals or a positive allocation without eligible park area raise ValueError before saving.

The park-building function remains available for separate placement experiments:

from scripts.build_imaginary_parks import build_imaginary_parks

centers = {
    "SL": (-70.40, 43.95),
    "SS1": (-70.65, 43.48),
    "SS2": (-69.95, 44.35),
    "SS3": (-70.70, 44.30),
}
features = build_imaginary_parks(centers, output_path="park_experiment.geojson")

Centers are (longitude, latitude) in degrees. Areas and shapes stay fixed. The builder checks validity, area, non-overlap, and containment inside Maine and the sample bounds. The Maine boundary source is documented in ../labs/ME_ShapeFiles/README.md.