Mexico City Land Subsidence
67 SLC scenes; 2,211 candidate pairs reduced to 78 valid interferograms; −35.5 to −2.9 cm/yr, consistent with the peer-reviewed −39.1 cm/yr peak.
Overview
Most Earth observation projects lose weeks to finding, downloading, correcting and formatting data. I build the infrastructure that turns that into one command: federated search, verified downloads, processing and cloud-native outputs, all reproducible.
What you get
- Automated acquisition across public and commercial providers, with one interface
- Analysis-ready imagery: corrected, cloud-masked, reprojected and mosaicked
- STAC 1.0 catalogues and cloud-native outputs: COG, GeoParquet, FlatGeobuf, PMTiles
- Web-map delivery: tile pyramids and PMTiles from very large rasters
- Data-quality checks and automatic repair of common vector and raster defects
- Scheduled YAML pipelines with run history, tests and CI
Who this helps
Research groups
Who need the same data preparation to run identically for every study area and student.
Platforms & start-ups
Ingesting imagery from several providers without writing a connector for each.
Agencies with mixed archives
Moving from folders of GeoTIFFs to catalogued, cloud-native data.
Web-mapping teams
Serving large orthomosaics and indices fast, without a tile server.
How I approach it
-
01
Inventory the sources
One search across 22 verified providers, including Copernicus Data Space, USGS, NASA Earthdata, ASF, Planetary Computer, Element 84, OpenTopography, Planet and Maxar, plus a generic OGC WMS / WFS / WCS connector.
-
02
Standardise the metadata
Results are normalised to STAC 1.0 with real footprint geometry and export to GeoJSON, GeoParquet or CSV, so every scene is traceable.
-
03
Download data you can trust
Parallel, resumable, band-selective and checksum-verified downloads, with credentials stored encrypted (Fernet) or in the system keyring.
-
04
Process to analysis-ready
Atmospheric correction (DOS, Sen2Cor, 6S), cloud masking and filling, reprojection, pan-sharpening, mosaicking, 17 spectral indices, and SAR calibration and despeckling.
-
05
Publish cloud-native
Cloud Optimized GeoTIFF, GeoParquet, FlatGeobuf and PMTiles; raster2pm converts very large GeoTIFFs to PMTiles block by block.
-
06
Automate and guard quality
YAML pipelines with scheduling and run history; geo-repair and pygeoproof for validation and drift detection; pytest on GitHub Actions with Codecov coverage.
Evidence from my work
- 22Verified data providers behind one API
- 17Spectral indices built in
- 64 GBOrthomosaic converted to PMTiles with raster2pm
- 200+Tests in the geoviz toolkit
pygeofetch is published on PyPI, CI-tested, documented on Read the Docs and archived on Zenodo (doi:10.5281/zenodo.22087230).
Selected work
View case study
Obuasi Vegetation Decline
Boundary-clipped USGS Landsat NDVI trend: 32.0% of the district in measurable decline (9.2% strong, 22.8% moderate), concentrated around mining.
What sets my work apart
- 1
Provider-agnostic by design
Switch between Copernicus, USGS, NASA or commercial imagery without rewriting the pipeline.
- 2
Verified, not just downloaded
Checksums, resumable transfers and a doctor command for install and connectivity diagnostics.
- 3
Production habits from day one
Typed APIs, test suites, CI coverage and documentation, as in any software product.