As part of Flagship Project 3,Charlotte examines how green spaces can be analysed using machine e learning in order to positively influence people’s quality of life. The project specifically aims to develop a map with indicators for socially and ecologically valuable green spaces. To achieve this, Street View images, aerial photographs, and satellite images are combined using deep learning and analysed together with existing map data. In the interview, Charlotte explains what personally motivates her to work on this project.
Charlotte, can you tell us a little about your background?
Charlotte: I started my studies in mechanical engineering at EPFL in Lausanne. Along the way, the topic became too specific for me. I wanted something broader. I was fascinated by tools and methods that could be applied to model and eventually solve a wide range of real-world problems at a larger scale. That led me to the master's in computational science and engineering, also at EPFL. There, I specialised in computer vision* for earth monitoring. I did my thesis developing an algorithm to monitor coral reef environments based on underwater imagery. Just recently, I started my PhD at UZH, working on the 3rd flagship project, focusing on green spaces and their effect on human wellbeing. dritten Flagship-Projekt focusing on green spaces and their effect on human wellbeing.
Why did you choose to come and work on this project?
What first drew me to this project was the goal of improving the quality of green spaces for citizens. Everyone should be able to benefit from the positive impact of nature on wellbeing. I believe that integrating nature and green spaces into everyday life is very important. Personally, I integrate nature into my daily life through outdoor sports and simply spending time outside, so nature is essential to my wellbeing. Working on a project that helps bring it to more people resonates deeply with me. Beyond the goal itself, I am also interested in the project from a research perspective. On one side, there is the data-driven work of mapping and analysing green spaces. On the other side, we are trying to model something deeply subjective: How do people actually experience these spaces and what do they mean for their wellbeing? Developing AI models to extract and quantify something so subjective is what makes this technically challenging and exciting to me.
In the following video, Charlotte explains what distinguishes a good green space from a bad one — and how she goes about finding out.
In the video, Charlotte talks about the city, but she is actually referring to the whole of the canton of Zurich.
Your project has a strong machine learning focus, but the aim is to link green spaces to wellbeing. What do you see as big challenges here?
To make the link from green spaces to wellbeing, you have to focus on people, as they are the ones experiencing wellbeing or the lack of it. Collecting data on people is already difficult in general, but machine learning requires large quantities of data, which makes it even more difficult. Another challenge – and one of the most interesting aspects, in my opinion – is model interpretability. When you train a model on images, it learns to make the predictions you have trained it to do, however, we don't always know what it has actually learned. For example, it might predict an image as being a good green space not because of the quality of the green space, but because of an unrelated feature such as the weather or the lighting.
“Understanding what the model has truly learned is difficult because these models behave like black boxes. What we need to do is develop methods to open them up and extract that knowledge”
Charlotte Sertic
Machine learning is developing very quickly right now - what does this pace of change mean for you and your work?
For our research, the fast pace of development can actually be a positive thing. New models emerge that we can directly use to advance our own work, for example to help create or process datasets. But on the other hand, there is a constant need to stay up to date. Things can change very quickly, so you have to remain agile in the way you work and be ready to adapt.
What are you most excited about in your work for the next four years?
I am really excited to see this project come together with the other flagship projects and to collaborate and share knowledge with people from different backgrounds. In particular, I am looking forward to working with the Zurich University of the Arts, which brings a completely different way of looking at things. I am curious to see how their perspective connects with our scientific approach.
Charlotte Sertic did her bachelor’s degree in mechanical engineering. She then completed a master's in Computational Science and Engineering, specializing in Computer Vision for Earth monitoring. Since January 2026, she has been doing her PhD at Departement für mathematische Modellierung und Machine Learning (DM3L) im EcoVision LabHer dissertation is part of flagship project 3 des DIZH Public Data Labs. Dabei untersucht sie, wie Machine Learning helfen kann, Grünflächen zu analysieren und ihre Wirkung auf das Wohlbefinden zu verstehen.
* Computer vision is a subfield of AI research. The computer is trained to recognize meaning in images.(ZHAW))