BSc und MSc Theses

In the Viererfeld we operate a LI-COR Carbon Node, an instrument that measures the exchange of water vapour and carbon dioxide between the land surface and the atmosphere. An earlier side by side comparison with a reference instrument showed that the Carbon Node reports water vapour fluxes roughly twice as high; which of the two instruments is at fault remains unresolved. At the current site we now have radiation measurements, soil temperatures at three depths and soil moisture measurements available, which allows the surface energy balance to be determined independently of the flux instrument and used as a benchmark. This balance states: net radiation minus soil heat flux equals sensible heat flux plus latent heat flux. The latent heat flux is the energy equivalent of the water vapour flux, that is, precisely the quantity under suspicion, and the energy it claims cannot exceed what radiation and soil heat flux make available. The thesis covers the processing and quality control of the flux, radiation and soil data, the calculation of the soil heat flux from the soil temperature profile and soil moisture, the analysis of energy balance closure at half hourly and daily resolution. What we offer is work with genuine field data from an ongoing measurement campaign, an insight into micrometeorology and water balance research, and a clearly defined result with direct consequences for our measurement operations.

Level: BSc / MSc

Methods: Statistical analyses

Prerequisites: Interest in measurement technology and data analysis together with basic skills in R or Python; familiarity with the eddy covariance method is not required and will be taught during the thesis

Supervisor: Dr. Jörg Franke (joerg.franke@unibe.ch)

Advisor: Patrick Kallabis (patrick.kallabis@unibe.ch)

Eddy covariance stations measure the surface energy balance precisely, but they are expensive, fixed to one place and represent a single footprint. Teuling et al. (2024) showed that a smartphone with a thermal sensor, its light sensor and a few external weather sensors can estimate evapotranspiration surprisingly well, with mean RMSE values of 0.10 mm h⁻¹ against a lysimeter and 0.05 mm h⁻¹ against eddy covariance.

Our new low cost system “PocketFlux” takes this idea further: instead of evapotranspiration alone, the goal is the full surface energy balance, i.e. short- and longwave radiation, sensible and latent heat flux, measured with a rugged smartphone with integrated thermal camera and a Bluetooth weather meter.

A working Android app already exists, and first field data were collected in September 2026. The app runs two processing modes side by side: one reproduces the method of Teuling et al. (2024), the other our own PocketFlux implementation, so that any gain can be quantified directly. In the thesis, the regression coefficients are calibrated against a SwissFluxNet station. The instrument is then operated at the Viererfeld in Bern next to our LI-COR CarbonNode, which provides turbulent fluxes, all four radiation components and soil moisture at 10, 20 and 50 cm as an independent reference. This shows how well the approach transfers to a new site and which terms need site-specific calibration.

Level: BSc / MSc

Methods: Low cost systems, Caliration, Fieldwork

Prerequisites: Basic knowledge in R or Python

Supervisor: Dr. Jörg Franke (joerg.franke@unibe.ch)

Advisor: Patrick Kallabis (patrick.kallabis@unibe.ch)

The Vierer- and Mittelfeld, situated at the northeastern edge of the Länggasse, shall be transformed into housing complex for ca. 3000 people over the next 5-10 years. As a showcase for urban densification and socio-ecological sustainability, it offers unique potentials for transdisciplinary research on sustainable urban development. Taking this opportunity, different research groups of the GIUB and OCCR have, together with various stakeholders (e.g., Stadtgrün, Tiefbauaumt, Quartierverein, Gymnasium Neufeld, PHBern, BFH), recently launched the “UrbanLab Viererfeld”, which also got funded by “Engaged UniBE” initiative. In this context, multiple possibilities for BSc and MSc thesis topics arise – using different methodologies and often in collaboration with other research units and stakeholders:

• Collection and analysis of environmental data serving as a basis for monitoring and modelling (e.g., measuring air pollution, noise, air/surface/soil temperatures, subsurface water flows ...)
•  Analyses of social perceptions, framings, and appraisals of the Viererfeld and its surroundings across different segments of the Bernese population; including analyses of care infrastructures and urban planning processes (e.g., discourse analysis, surveys, interviews, participatory approaches)
• Evaluation of Potentials and challenges of citizen science in climate-sensitive urban development (e.g., with surrounding neighbourhood, upper secondary school classes from Gymnasium Neufeld, ...)

Level: BSc / MSc

Methods: Measurements, statistical analyses, GIS-based mapping, interviews, surveys, etc.

Prerequisites: Statistics (basics), GIS (basics), interest in inter- and transdisciplinary approaches

Supervisors: Dr. Moritz Gubler (moritz.gubler@unibe.ch); Prof. Dr. Stefan Brönnimann (stefan.broennimann@unibe.ch); PD Dr. Jeannine Wintzer (jeannine.wintzer@unibe.ch); Dr. Jörg Franke (joerg.franke@unibe.ch); Dr. Deniz Ay (deniz.ay@unibe.ch)

Advisors: Prof. Dr. Edouard Davin (Wyss Academy for Nature), (edouard.davin@unibe.ch); Dr. Apolline Saucy (ISPM), (apolline.saucy@unibe.ch); Dr. Sophie Meyer (ISPM), (apolline.saucy@unibe.ch); Patrick Kallabis, (patrick.kallabis@unibe.ch)

Many Swiss towns want to know where heat builds up, but a dense measurement network is costly to install and maintain. Within the myurbanclimate project we develop a planning tool that uses Swiss-wide geodata to place sensors where they add the most information. The aim is to capture the urban temperature field with as few sensors as possible while losing as little information as possible. So far, the tool has mainly been developed with data from Bern. Olten is therefore a real test case: can the tool plan a good network in a city it has never seen?

Together with the city of Olten, the student set up a real network, checks the mounting sites on public ground (typically street lamps), installs the low-cost sensors and runs the network over the summer. The measured temperatures are then compared with the temperature field and the uncertainty the tool predicted before any sensor was installed. In a second step, the data are used to describe Olten’s urban climate: UHI in space and time, heat days and tropical nights, and the neighbourhoods with the highest heat exposure. The results go back to the city as a basis for heat adaptation

Level: MSc

Methods: Set-up of a monitoring network, fieldwork, evaluatuion and urban climate analysis.

Prerequisites: Basic knowledge in R or Python.

Supervisors: Prof. Dr. Stefan Brönnimann, (stefan.broennimann@unibe.ch); Dr. Moritz Gubler, (moritz.gubler@unibe.ch)

Advisor: Patrick Kallabis, (patrick.kallabis@unibe.ch)

The o-meters (Bernometer, Thunometer, Bielometer) show current and forecast air temperature on a 50 m grid for three Swiss cities. A random forest model combines weather forecasts (MeteoSwiss for Bern, Open-Meteo for Thun and Biel) with geodata such as terrain, land use and infrastructure. Every 15 minutes, the modelled field is corrected with live data from the low-cost city networks by interpolating the residuals between measured and modelled temperature. This set-up works, but it leaves several open questions that can each become a thesis:

(1) Model intercomparison: do gradient boosting, neural networks or Gaussian processes beat the current random forest when trained on the same predictors and tested on stations left out of training?

(2) Humidity and heat index: the networks also measure relative humidity, which is not used yet. Maps of humidity and heat index would describe heat stress better than temperature alone.

(3) Residual field: the current correction interpolates residuals by distance only. Modelling them explicitly with geodata shows where and why the model fails.

(4) Systematic errors: comparing forecasts from different sources and lead times with the later measurements reveals systematic biases, e.g. by time of day, weather situation or location, and gives the uncertainty of a local forecast.

5) Optionally, every thesis can include work on our new O-meter app.

Level: BSc / MSc

Methods: Statistical modelling, machine learning, uncertainty analysis.

Prerequisites: Knowledge in R / Python or willing to dive in. Willing to dive into modelling and machine learning.

Supervisor: Prof. Dr. Stefan Brönnimann, (stefan.broennimann@unibe.ch)

Advisor: Patrick Kallabis, (patrick.kallabis@unibe.ch)

The shortest walking route is not necessarily the one with the most shadow. On hot days, this can become an issue especially for vulnerable people. Our new feature CoolPath finds routes that avoid sun at the expected time of passage. A working prototype already runs in the O-meter app for Bern (not published yet): shade is computed from swissSURFACE3D and swissALTI3D at 2 m resolution every 30 minutes, arcades are treated as fully shaded, and each footpath segment gets a perceived length in which sunny metres count more than shaded ones. The approach builds on CoolWalks (Wolf, Vierø & Szell, 2025). A first example from Bern: in early September at noon, the route from the GIUB to Schauplatzgasse has 29 % fewer sunny metres for only 2 % detour or even 66 % fewer sunny meters for 5 % detour.

The thesis develops the prototype into a scientifically grounded tool. First, the shade field is combined with the o-meter temperature field or heat index, so that routes also avoid warm areas and not only direct sun. Second, the weight of shade is so far set by hand and correspond not to specific literature. Third, the workflow is transferred to Thun and Biel, where the app currently falls back to standard routing

Level: MSc

Methods: Geodata processing, shade-aware routing, analysis of real route choices

Prerequisites: Interest in urban climate, heat exposure and geodata. Basic knowledge in Python or willingness to dive in. Kotlin is optional (vibe coding is ok, if you understand what you are doing). Ready to do fieldwork in cities

Supervisor: Prof. Dr. Stefan Brönnimann (stefan.broennimann@unibe.ch)

Advisor: Patrick Kallabis (patrick.kallabis@unibe.ch)

The Viererfeld is a large open area in Bern next to the city and the planned site of a new urban quarter. During a 24-hour measurement campaign, thermal drone images of the site were taken at different times of the day and night, while measurement towers recorded the atmospheric conditions on the ground. Together, these data show how grassland, paths, trees and neighbouring buildings heat up during the day and cool down at night.

The thesis first processes the thermal images into comparable surface temperature maps over the full day and evaluates the tower time series. Surface and air temperatures are then linked to a 3D model of the Viererfeld, which gives slope, exposure, shading and sky view for every point. This makes it possible to explain the observed patterns, for example why some surfaces cool faster at night or where cold air collects.

Level: BSc / MSc

Methods: Thermal image analysis, evaluation of tower measurements, combination with a 3D model

Prequisites: Interest in urban climate and data analysis. Basic knowledge in R or Python

Supervisor: Dr. Moritz Gubler (moritz.gubler@unibe.ch)

Advisor: Patrick Kallabis (patrick.kallabis@unibe.ch)

This thesis will investigate the possible mechanisms driving the controversially discussed post-volcanic winter warming at high latitudes through analysis of a large ensemble of climate model simulations spanning multiple major eruptions. By utilizing ensemble modeling approaches, we aim to quantify the relative contributions of different physical processes—including stratospheric aerosol radiative forcing, Arctic Oscillation modulation, polar vortex dynamics, and ocean-atmosphere teleconnections—to the development of a warming signal. The large ensemble framework will enable robust statistical analysis of the likelihood for a warming, its magnitude and persistence while accounting for internal climate variability, allowing us to isolate the volcanic signal from natural fluctuations. Through systematic analysis across different eruption magnitudes and seasonal timing, this research will advance our understanding of volcano-climate interactions and improve predictive capabilities for post-eruption climate anomalies in polar regions.

Level: MSc

Methods: Statistical analysis of a model simulations ensemble

Prerequisites: Basic knowledge and interest in paleo-climatology and atmospheric dynamics. Basics in data analysis with R or Python. Willingness to work with large data sets

Supervisors: Dr. Jörg Franke (joerg.franke@unibe.ch)

Advisor: Dr. Ralf Hand (Ralf.Hand@dwd.de)

HIST-DAILY is a dataset of daily and sub-daily European weather observations from before 1900. The first version brings together station series in a standardized exchange format and is described in a forthcoming data descriptor paper. This thesis will contribute to a second version of the dataset by preprocessing and formatting additional station series into the standard format and applying quality control procedures. To assess the added value of the new data, the series will be validated against existing records and reanalysis products to check how much they improve spatial and temporal coverage. Depending on progress, there is the possibility of contributing to a scientific publication describing the extended dataset.

Level: BSc / MSc

Methods: Data preprocessing, formatting into a standardized station data format, and quality control of historical meteorological series

Prerequisites: Basic Python or R skills. Interest in historical climatology and data rescue. Attention to detail

Supervisor: Prof. Dr. Stefan Brönnimann (stefan.broennimann@unibe.ch)

Advisor: Carlota Corbella (carlota.corbellaalcantara@unibe.ch)