Lasers Over Hamburg: How the City Is Mapping Itself in 3D And Why That’s Great News for Data and AI Enthusiasts

If you live in Hamburg or nearby, you probably noticed it this week: the unusual aircraft noise during the day and sometimes even at night, and a small plane flying back and forth in very systematic lines across the sky.

Naturally, Hamburg, the best city in the world, explains exactly what is going on on its official website. The city is conducting a new airborne laser-scanning campaign to create an extremely detailed three-dimensional map of Hamburg. The goal is to capture the exact shape of the city, including every building, tree, street, and piece of terrain, by collecting millions of precise laser measurements from an aircraft flying high above Hamburg.

This campaign is part of a long-term effort to keep Hamburg’s geospatial data up to date, and it is actually the fourth campaign of this kind, following earlier citywide scans conducted in 2010, 2020, and 2022.

How the Laser Scanning Works

The technology behind this project is called airborne LiDAR (Light Detection and Ranging), and while the name sounds technical, the idea is surprisingly intuitive.

An aircraft flies across the city in carefully planned parallel lines while carrying a laser scanner mounted underneath the plane. This scanner sends hundreds of thousands of tiny laser pulses toward the ground every second. Each laser pulse travels to the surface and reflects back to the sensor, and by measuring how long the light takes to return, the system can calculate the exact distance between the aircraft and the surface below.

When this happens millions or even billions of times during a flight, the result is a massive collection of points in space, often called a point cloud, where each point represents a precise location on the Earth’s surface.

The data looks like a dense cloud of dots outlining buildings, trees, roads, and terrain — essentially forming a digital 3D version of Hamburg.

Lidar point cloud for a street (source)

Many laser pulses partially pass through tree canopies before hitting the ground, which means the system can often capture both the tops of trees and the ground underneath, allowing extremely accurate terrain models to be built even in forested areas.

Why Cities Do These Scans

Cities repeat these scanning campaigns regularly because the urban landscape is constantly changing. New buildings appear, infrastructure evolves, trees grow or disappear, and terrain changes through construction and environmental processes.

By updating this data regularly, Hamburg can maintain extremely accurate digital models of the city that help with a wide range of important tasks.

For example, the data can be used to:

• plan urban development and infrastructure

• simulate flooding and stormwater flow

• measure tree coverage and green areas

• evaluate rooftop solar potential

• maintain accurate digital city models (digital twin)

• improve environmental and climate planning

In other words, this dataset becomes a foundational layer of information for many modern smart city applications.

The data is public

What makes Hamburg particularly impressive in this space is not just that the city collects such high-quality data, but that it openly publishes much of it.

Many cities collect similar datasets but keep them restricted or difficult to access, which limits the broader innovation that could happen around them. Hamburg, however, provides large parts of its geospatial data through its open geodata portal, allowing researchers, developers, startups, and curious citizens to explore and use the information.

This approach turns public infrastructure data into something much more powerful: a shared resource that can enable new ideas and technologies far beyond the original purpose of the scan. If you are looking for interesting data for your next ML project, this is gold and open the door to many exciting projects.

Ideas for ML Projects

Here are a few ideas that could turn Hamburg’s open LiDAR data into fascinating machine learning experiments:

1. Tree Detection and Classification

Using point cloud neural networks, you could train a model to detect and segment individual trees across the city and estimate properties such as height, canopy volume, or species categories. This type of model is extremely valuable for urban forestry and climate research. (recent paper)

2. Building Roof Classification

A deep learning model could analyze roof shapes from the LiDAR data and classify them into categories such as flat roofs, pitched roofs, or complex structures.

Such models can be used to estimate solar energy potential across the entire city. (paper 1, paper 2)

3. Automatic 3D Building Reconstruction

Using neural networks designed for point cloud processing (such as PointNet or KPConv), you could attempt to reconstruct simplified building models directly from the LiDAR data.

This could contribute to automated digital twin generation. (recent paper)

4. Urban Change Detection

By comparing the current LiDAR campaign with previous scans from 2022 or 2020, a model could detect where the city has changed — identifying new construction, demolished structures, or changes in vegetation. (recent paper)

5. Road and Infrastructure Extraction

Deep learning models can also be trained to automatically identify roads, bridges, and other infrastructure within the point cloud.

This could support automated map updates or transportation planning tools. (paper 1, paper 2)

So the next time you hear a small aircraft passing above Hamburg in the middle of the night, it might not just be noise. It might be the sound of millions of laser measurements quietly capturing the city in three dimensions, creating a dataset that researchers, engineers, and curious minds will explore for years to come.