TerraPulse
AI-powered ground deformation and subsidence intelligence using Sentinel-1 InSAR
TerraPulse lets planners, engineers, and NGO workers draw an Area of Interest in QGIS,
select a time window, and receive an interpreted deformation map with risk attribution —
no SAR or InSAR expertise required. Powered by Sentinel-1 SBAS-InSAR, scikit-learn
classification, and optional LLM-generated plain-language reports.
Features:
- Draw or import AOI, select time window and processing mode
- Queries Sentinel-1 SLC stacks from Copernicus Data Space STAC API
- Runs InSAR processing locally via Docker/PyGMTSAR
- ML pixel classification: stable / linear / seasonal / accelerating / anomalous
- OSM infrastructure overlay with risk ranking (buildings, roads, pipelines)
- PDF/HTML report with optional AI-written narrative (Anthropic Claude)
- YAML provenance recipe + STAC 1.0 item for full reproducibility
- Persistent settings (CDSE credentials, API keys, output directory)
- Non-blocking background processing via QGIS task manager
Requirements:
- QGIS 3.34 or later
- Python 3.11 or later
- Docker Desktop (for local InSAR processing)
- Free Copernicus Data Space Ecosystem account (dataspace.copernicus.eu)