REAL GIS DATA PREPARATION — 2026-10-22

Open gis_analysis.gpkg in QGIS or ArcGIS. It contains the study boundary,
published wind-farm points, NPWS reserve polygons, filtered bird records,
1 km grid cells, and five 10 km analysis circles, one per selected farm.
The table grid_features_1km.csv has one row per cell (2376 cells).
observed_cell_month_features.csv has only observed cell/month combinations;
regional climate is joined by month and must not be treated as local 1 km data.

CRS: EPSG:7855, GDA2020 / MGA zone 55. Distances use cell centres and official
representative project points, not turbine footprints. Slope and height derive
from Copernicus DSM, bilinearly resampled to 50 m, then aggregated to 1 km.
The radius of 10 km is a demonstration analysis setting, not a validated impact
zone. Bird observations use 2020–2025 records with reported coordinate
uncertainty <= 1000 m. Empty cells are not absences, and observation counts are
not abundance. Some records may share survey events or be duplicated across
publishers with different identifiers. These are data-preparation outputs,
not a ready-made fair benchmark or causal before/after experiment.

Individual source licences and complete query metadata are in sources/provenance.json.
BirdLife and some iNaturalist records are CC BY-NC; this is a noncommercial
research demonstration. The NSW wind-service metadata did not state a licence;
do not assume the entire mixed bundle has a single open licence.

Quality checks are in data/quality_checks.json. Source checksums, filter rules,
limitations and full references are included in sources/provenance.json.
No model training, performance experiment or ecological-effect estimate ran.

INTERACTIVE GIS UPDATE — FIVE FARMS
All five selected farms are listed in wind_farm_summary.csv and have their own
distance fields and analysis buffers. farms_within_10km and
regional_farms_within_10km count the five farms. selected_pair_farms_within_10km
preserves the original Biala/Gullen pair count. nearest_other_farm_km is the
legacy distance excluding that pair; all_listed_farms_within_10km uses the
entire government source layer. Original pair-only results are explicitly
labelled pair_example in summary.json and are not presented as five-farm results.

The GeoPackage and wind_turbines_osm.geojson contain 202 mapped turbine
nodes, including surrounding context. The map displays 155 nodes linked
to the five named farms by explicit OpenStreetMap relation membership.
All mapped members of these five source relations are included, even outside
the bird/grid study area. Bird statistics remain clipped to that study area;
no new bird observations or predictions were fabricated. OSM membership is
source evidence, not independent verification of a complete current inventory.
farm_coverage.json contains per-farm IDs and membership-source links.

Original pair-focused static illustration PDFs are not included in this updated
data package. The original processing script in audit is historical; the
five_farm_processing.py update and source records document the current dataset.
The website shows all 155 mapped farm turbines and bird observation sites
directly at their source coordinates. Clicking a point opens the record without
a grid/list intermediate step. Co-located bird observations support previous
and next record navigation. Per-cell queries and filtered CSV export remain.
Satellite basemap is live;
processed terrain and all query records are stored locally.

TIME COMPARISON
wind_farm_timeline.csv contains government-source construction and operation
dates. Unknown construction dates remain missing. The map compares each bird
observation with the selected farm date: before is green, on/after is red.
Split-color points contain both periods at the same published coordinate.
Coordinates are never jittered. bird_observations_before_after.csv records
both study-region and farm-10-km counts. Observations cover 2020-2025 only;
missing before-period observations are not evidence of an absence of birds.
Counts describe records and do not establish a causal wind-farm effect.

MAP CHECKS
The satellite zoom limit is 18 at the five sampled farm locations. A network
failure does not lower this limit. Only confirmed provider coverage gaps can
limit the affected area, without changing other locations. Fullscreen control
has been removed. JSDOM checks cover direct point clicks, dates, filters,
zoom, source links and all 155 turbine associations. These are DOM functional
checks, not live browser rendering or visual QA.
