TL;DR Quickstart¶

Run this first

Use run_senbygg_improved.py. It wraps the improved candidate detector, optional Sentinel-2/coherence fusion, output writing, and preflight checks.

1. Install¶

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

The stack includes openeo, rasterio, geopandas, xarray, h5netcdf, h5py, shapely, scipy, pandas, numpy, and matplotlib. GDAL/PROJ must be available.

2. Configure `run_senbygg_improved.py`¶

AREA_NAME = "kvamskogen"
BBOX = {"west": 5.900, "east": 6.162, "south": 60.360, "north": 60.412}
HISTORY_START = "2018-01-01"
MONITOR_START = "2021-01-01"
MONITOR_END = "2024-12-31"
DOWNLOAD_IF_MISSING = True
MIN_AREA_M2 = 150.0
USE_S2 = True
USE_COH = True

Use at least one full annual cycle of history; two years or more is recommended.

3. Choose candidates¶

Leave CANDIDATES_GPKG = None to use local FKB-Tiltak snapshots, or set it to a custom GeoPackage of valid candidate polygons.

4. Run and review¶

python run_senbygg_improved.py

Open results/senbygg_improved/<area>_detected.gpkg in QGIS and style by flagged, source, change_date, coh_flagged, and s2_flagged.