PALEOBEARD FIELD DATABASE — PUBLIC DATA README
================================================

This directory holds the public projection of the Paleobeard field fossil
database. The LOCAL field CMS is authoritative; these files are generated.
Do not edit them by hand — the next export overwrites them.

FILES
  occurrences.json        manifest of partitions + facet counts. This is an
                          INDEX, not the database: browsers fetch only the
                          partition chunks that can match the active filters.
  occurrences/*.json      record chunks partitioned by public locality and
                          year, at most 200 records each.
  formations.json, members.json, localities.json, sections.json,
  layers.json, taxa.json  the public stratigraphic framework.
  statistics.json         honest counts derived from the published records.
  samples.json            separately published sampling events and counts.
  downloads/              public CSV, JSON and Darwin Core-compatible CSV.

PRIVACY
  Only records I explicitly publish appear here after a site rebuild.
  Exact GPS coordinates, GPS accuracy, elevation, property records,
  private locality names, sub-locality detail (quarry area / bench /
  trench / grid), private notes and original photograph filenames are
  NEVER exported, in any format. Public photographs are sanitized
  derivatives with metadata (EXIF) removed. Public locality detail is
  limited to the approved public name and region. If you need finer
  resolution for research, contact me; access is granted, not
  scraped.

RECORD COUNTS vs SPECIMEN COUNTS
  statistics.json separates record_count (documented occurrence records)
  from specimen_count_sum (numeric counts whose unit is specimens).
  counts_by_unit keeps specimens, fragments and pieces separate. An
  occurrence documenting 38 specimens is ONE record and 38 specimens.
  Sampling events remain separate and may overlap occurrence records;
  never add the two datasets. Positive measured area supports per-sample,
  per-unit density only for recorded numeric counts, not qualitative labels.
  Preservation proportions use records with known preservation as their
  denominator; size means use only records with measurements.

OBSERVED vs DERIVED vs IDENTIFIED vs INTERPRETED
  observed   — what the collector recorded in the field.
  identified — taxonomic determinations, with confidence, kept as history.
  derived    — context (formation/member/layer) resolved from the section.
  interpreted — later scientific interpretation, clearly labelled as such.
  Interpretations are never presented as field observations. "unknown" is
  a valid, honest value and means exactly that.

METHOD BIAS
  Counts reflect where and when collecting actually happened, not the
  abundance of fossils in the rock. Effort, exposure and weather bias
  everything. "Rare" lists mean "infrequently documented in this dataset",
  NOT geologically rare.

NULLS AND UNKNOWN
  Missing values are empty in CSV, null/absent in JSON. We never guess
  layers, taxonomy or counts to fill a gap.

DARWIN CORE MAPPING (downloads/…-dwc.csv)
  A Darwin Core-compatible projection, not a certified DwC-Archive.
  occurrenceID -> id, basisOfRecord -> literal HumanObservation,
  eventDate -> date, recordedBy -> collector, scientificName from
  genus/species when curated (identification text otherwise),
  individualCount -> count only when count_unit is specimens;
  organismQuantity/organismQuantityType -> numeric count and its unit.
  Confidence never invents a cf. qualifier. Latest identification history
  supplies identifiedBy/dateIdentified. formation/member/lowestStratigraphicZone ->
  stratigraphy, locality -> PUBLIC locality name only.
  Coordinates are intentionally absent; informationWithheld says so.

CITATION AND LICENSING
  No licence is asserted here. Cite Paleobeard and the record IDs; contact
  me before republication. Licensing is my decision
  and is not guessed by the export.

Generated 2026-10-02T22:50:06Z (schema 1).