About the project

Cities and property owners need fast, reliable facts about their sites to plan for climate adaptation, biodiversity and water management. Krontech is a research project where a practical tool is being built that turns aerial images and GIS data into clear, property‑level numbers—like how much area is green vs. paved, canopy cover, and where upgrades (e.g., green roofs) would help the most.

We already combine orthophotos, laser‑scanned height models (LiDAR), NDVI and terrain (e.g., slope). We are also training AI segmentation models to make the maps sharper. A key challenge is what lies under tree canopies—grass, bare soil, paths? Getting this right matters for runoff, cooling and habitat.

What you will do

• Train and compare segmentation models for urban land cover at high resolution.
• Fuse multiple data sources (orthophotos, LiDAR/DSM‑DTM, NDVI, slope) and test what each adds.
• Use multi‑year imagery to handle seasons and improve class separation.
• Tackle the under‑the‑trees problem using canopy height, crown shape and context (park vs. street).
• Turn maps into parcel‑level metrics that support reporting (e.g., CSRD) and simple targets like the “3‑30‑300” rule of thumb.
• (Nice‑to‑have) Build a small web demo to visualize parcels, masks and KPIs.

Possible thesis angles (pick one or mix two)

A. Under‑canopy inference – Predict the ground cover beneath tree crowns using fused data and simple context rules.

B. Multi‑source & multi‑season fusion – Show how extra inputs (height, NDVI, seasons) improve over single‑date RGB.

C. From maps to metrics – Turn segmentation into clear property KPIs and a first take on 3‑30‑300, with basic uncertainty.

Data you will use

• High‑resolution orthophotos (several years for the same sites)
• LiDAR‑derived DSM/DTM and canopy metrics
• NDVI and terrain derivatives (slope/aspect)
• Parcel boundaries and other GIS layers (roads, buildings, parking)
• Optional: soil/landform and carbon‑storage layers where available
  (Access and licensing follow Swedish public data policies and partner agreements.)

What you bring

There are no strict requirements. It helps if you have:

• Python skills and interest in ML/computer vision or remote sensing
• Some GIS experience (QGIS/GeoPandas/GDAL/rasterio)
• Curiosity about cities, climate adaptation and biodiversity
• Bonus: PyTorch/TensorFlow, web mapping (Leaflet/Mapbox), basic SQL

Supervision & collaboration

• Industrial supervisor: IVL (e.g., Jens Wilhelmsson/Agnes Olsson)
• Academic supervisor/examiner: arranged with your university
• Real‑estate partners provide pilot data and feedback

Host Organization

IVL Swedish Environmental Research Institute (IVL), together with participating real‑estate partners.

Location: Stockholm/Malmö/Göteborg

Credits: 30 ECTS

Group size: 1–2 students

Start date: Flexible (earliest January 2026)