Urban solar potential analysis sounds straightforward: take 3D building geometry, calculate roof surface orientation and area, factor in irradiance data, output a map. In practice, working with real-world CityGML data — especially Japan's PLATEAU dataset — surfaces every hard problem in spatial ETL at once.
This talk is about building that workflow honestly: where pipelines break, how you design transformations that survive contact with real data, and what it looks like when it finally works.

The Data Challenges
- Parsing and flattening nested CityGML building attributes across large LOD2 datasets
- Coordinate system handling — JGD2011/EPSG:6668 and the precision issues that follow
- Extracting and classifying roof surfaces for irradiance calculation
- Joining geometry with solar irradiance reference data at scale
- Managing intermediate and output volume for a whole city district

The Workflow
We'll walk through the pipeline stage by stage — not as a clean success story, but as a real engineering process. Where did we need a transformation type that didn't exist yet? Where did intermediate data look completely wrong until we found a subtle coordinate issue?
Each stage will be visible as a node graph, with data inspectable at every step — geometries, attributes, intermediate outputs — so the audience can follow the logic without reading code.
The Output — and What It Represents
The final stage produces solar potential scores per building surface, visualized directly within the workflow — geometries coloured by score, attributes inspectable in place.
But the more valuable output is the workflow itself. Every transformation step is documented, configurable, and reusable. Swap the city district, update the irradiance reference data, re-run. What started as a one-off analysis becomes a repeatable spatial data pattern worth sharing with the community.

What You'll Take Away

A practical map of the pain points in PLATEAU/CityGML-based analysis workflows, and approaches that work
A mental model for designing reproducible, visual ETL pipelines for complex geospatial data
A look at what browser-native, collaborative workflow tooling can do for this class of problem

Who Should Attend
Geospatial developers, GIS analysts, and data engineers working with 3D city models, CityGML, or large-scale urban datasets — and anyone interested in making spatial data pipelines more reproducible, inspectable, and shareable.