Nodes
30
connected typekeys
Munich,
cooled.
A Sentinel-2/3 pipeline that maps urban heat across Munich. Ingest gigabyte-scale raster scenes, reproject with GDAL, mask cloud, compute thermal indices per pixel, aggregate by neighbourhood, ship a signed GeoJSON map a city planner can act on.
Spatial primitives
17
GdalIngest · Reproject · TileEnvelope · …
Source
Sentinel2 / 3
Copernicus L2A · free
Output
GeoJSON
evidence-packed · signed
A city decision
from free satellite data.
Every European city has to answer the same question: which streets, which neighbourhoods, which roof areas get hot enough to threaten the people who live there? The data exists. It is published, for free, by ESA. The hard part is turning gigabyte scenes into a neighbourhood-level map a planner can actually use.
The Urban Heat Pathfinder workflow does that. Sentinel-2 multispectral and Sentinel-3 thermal scenes come in raw. Cloud cover is masked using the SCL band. Thermal land-surface-temperature is computed per pixel, then aggregated zonally into the city's neighbourhood polygons. The output is a GeoJSON layer ready for a planning GIS.
Every transform is signed. The whole run sits inside a forensic.evidence-pack so the city can demonstrate this is the calculation we made, on these inputs, with this code.
Thirty nodes,
one .viow file.
From the actual UHP_Munich_RealData.viow shipped as a Vectorio reference demo. Seventeen unique TypeKeys, three forensic nodes, one orchestration coordinator.
ingestspatial.GdalIngest
→
reprojectspatial.GdalReproject
→
tilespatial.TileEnvelope
→
maskspatial.RasterQualityMask
→
indicesspatial.RasterBandMath
→
temporalspatial.TemporalComposite
→
zonalspatial.ZonalAggregation
→
fusespatial.LinearFusion
→
emitspatial.GeoJsonPublisher
Three forensic nodes sit alongside the data path — forensic.EvidenceCollector, forensic.EvidencePack, forensic.EvidenceVerifier — so the city's compliance officer can verify the run after the fact.
Four layers,
one pipeline.
The same UHP_Munich_RealData.viow renders four analytical layers a planner can toggle on the live map — composite Urban Heat Index, Land Surface Temperature, Surface Albedo, Vegetation (NDVI). Real screens from the Vectorio geospatial overlay.
Scene: Sentinel-2 L2A · Munich · 2025-08-15 · Pipeline: UHP_Munich_RealData.viow · Tile zoom 0–13.
What ESA hands you,
scene by scene.
Sentinel data ships at known shapes. The pipeline is sized to absorb them.
| Source | Per scene | Spectral bands | Cadence |
|---|---|---|---|
| Sentinel-2 MSI L2A | ~800 MB | 13 | every 5 days · 10 m resolution · ~110 × 110 km |
| Sentinel-3 SLSTR LST | ~400 MB | 9 | daily · 1 km resolution · land-surface temperature |
| SCL cloud-classification band | included | 1 | ships with L2A · used directly by RasterQualityMask |
| City neighbourhood polygons | ~3 MB | — | static · per city · fed to ZonalAggregation |
Why this matters.
- Free data, reproducible result. Anyone can re-run the exact same pipeline a year from now against the same Copernicus archive and get a byte-identical GeoJSON.
- Signed at every stage. The thermal anomaly map ships with a hash chain showing exactly which input scenes produced it. A regulator or a journalist can verify offline.
- Generalises. Swap Munich's polygons for Naples, Athens, Lisbon — same workflow, different city. Swap the thermal-index formula for a vegetation, flood, or methane index — same workflow, different mission.