Methodology and data sources

Earth-Audit turns public satellite archives into a property-level record. Every available figure can be traced to a named source, observation date and resolution, so you can judge the evidence for yourself.

How the analysis runs

1. Location. Your point is resolved to coordinates and an area of interest is drawn around it. The dashboard states the area in hectares and the coordinate reference system used.
2. Scene selection. Cloud-free or low-cloud scenes are selected per year from the Landsat and Sentinel-2 archives. Where no usable scene exists, the year is reported as a gap, not interpolated.
3. NDVI. The normalised difference vegetation index is computed per pixel from the red and near-infrared bands, then averaged over the area of interest to produce a yearly trajectory.
4. Pixel change. A baseline window and a current window are compared pixel by pixel and classified as stable, loss, severe loss or regrowth. The result is reported in hectares as well as percentages.
5. Context. Fire detections, forest-loss records, climate anomalies, soil, biodiversity and landscape features are pulled from their own sources and shown with their own status.
6. Fingerprint. Every generated report is hashed with SHA-256. The hash is published so a third party can confirm the file they received is byte-for-byte the file we produced.

Limitations we do not hide

• Optical satellites cannot see through cloud. Persistently cloudy regions have sparser coverage.
• A 30 m Landsat pixel is a third of a hectare. Changes smaller than a pixel are not detectable at that resolution.
• NDVI reacts to seasonality and drought, not only to clearing. Comparable windows are used, but interpretation still needs local knowledge.
• JRC forest-loss data covers the tropical moist forest belt. Outside it, that indicator reads not available.
• We do not issue legal opinions, land-title conclusions or compliance certificates.

Sources