FIELDNOTES — DATA DICTIONARY & METHODOLOGY

Source: BFRO reports compiled by Timothy Renner, distributed in the TidyTuesday 2022-09-13 snapshot.
https://github.com/rfordatascience/tidytuesday/tree/master/data/2022/2022-09-13
https://github.com/timothyrenner/bfro_sightings_data
https://www.bfro.net/GDB/default.asp

Scope: 5,021 report records. A historical snapshot, not a live feed. Full witness narratives are not reproduced in this website; each report links to the BFRO source.

FIELDS
id: integer BFRO report identifier (source column number).
state: U.S. state recorded in the source.
county: county label recorded in the source.
season: source's reported season, used without inference from dates.
date: source's ISO encounter date, or blank/null if missing. Some source dates may be approximate; do not assume day-level precision.
year: year extracted from date, or blank/null.
latitude, longitude: source coordinates in decimal degrees, or blank/null. These may be approximate and should not be used as exact field destinations.
classification: BFRO's Class A, B, or C label. Preserved as a source classification.

PREPARATION
Deduplicate by report ID, retaining the first row. Parse ISO dates; invalid/missing dates remain null. Parse coordinates and reject values outside valid latitude/longitude ranges. Do not impute or geocode missing values. Keep every unique report in totals and the downloadable archive. Sort newest dated reports first; unknown dates last.

ANALYSIS
All filters intersect. All years includes unknown dates; specific periods exclude them. Date unknown selects only reports without dates. Timeline aggregates by decade of encounter date. Season chart aggregates the source season field; unfamiliar or missing seasons are grouped as Unknown.
The sightings map plots only coordinates supported by an Albers U.S. projection. Its footnote reports excluded locations. The state map counts reports by state name, including those without coordinates. Darker colors reflect larger raw counts; colors rescale with the selection.
The search box searches report ID, county, and state within active filters. CSV export respects both filters and search. Search does not change the charts.
Percentages in field notes use the filtered report total as denominator. This analysis does not normalize by population, land area, time spent outdoors, or reporting effort. No predictive model or statistical significance claim is made.

LIMITATIONS
Voluntary reporting introduces selection and reporting bias. Publicity, population, outdoor activity, retrospective recall, and reporting access can change counts. No report in the archive does not mean no encounter occurred. Each row describes a report; raw counts describe the collection rather than a population census. Missingness can bias geographic and temporal patterns. The 2020s are incomplete in this snapshot, and the final encounter date is in 2021. A decrease in recent counts does not establish a decrease in encounters.

MAP & SOFTWARE
State boundaries: us-atlas 3.0.1 (U.S. Census-derived geometry).
Visualization: D3 7.9.0 (ISC), topojson-client 3.1.0 (ISC).
Hero forest: AI-generated fictional illustration, not a sighting photograph.

REPORT-ERA TIMELINES
The main page links to focused collections. Historical/context cards and numerical report records are separate collections and may overlap. The report pages sort encounters earliest first and show at most 12 cards per page. The archive is U.S.-only; Canadian and Mexican source material is displayed separately. Oral story time and documentation time are distinct when available. Unidentified books and unresolved accounts remain visible research leads without invented dates.