Master’s thesis topics
Dr. Lukas Brunner
How to use this document
Here I present some topics for master’s thesis with me.
I try to keep this collection up-to-date but might not always manage to, so some topics might no longer be available. Typically, hidden slides are topics which have already been taken.
If you are interested in working with me for your thesis please also have a look at my homepage here and in particular at my supervision approach. Also please keep in mind that these are only general ideas for research questions, I expect you to contribute to developing your own, more specific research questions in the course of your thesis.
If you are interested or have any questions feel free to reach out to me: lukas.brunner@uni-hamburg.de
Lukas Brunner | 2
Representation of dry days in km-scale models
Top: (c) MPI-M/DKRZ | Bottom: Brunner
Background. The advent of global, km-scale models provides an exciting new source of climate information. These models provide a step change compared to the more traditional CMIP models in their resolution and their handling of convection.
Question. How well are dry days (i.e., days without/with hardly any precipitation) represented in the latest km-scale models compared to observations and CMIP models? How is this affected by the simplifications used in the model?
Tasks. Work with climate data from km-scale and CMIP models; develop a framework to analyse and compare precipitation in Python; investigate, visualise, and discuss results
Required technical skills (or willingness to learn): Python (xarray), Git, Linux command line
Literature: Brunner et al. (2025), Wille et al (accepted), Takasuka (preprint)
Lukas Brunner | 3
co-supervision with Benjamin Poschlod
Importance of km-scale model resolution for impacts
(c) MPI-M/DKRZ
Background. The advent of global, km-scale models provides an exciting new source of climate information. These models provide a step change compared to the more traditional CMIP models in their resolution and their handling of convection.
Question. What is the effect of km-scale resolution (compared to the about 100km resolution in earlier models) for climate impacts and risks? Which regions are particularly affected because they have, e.g., high population density and/or benefit strongly from high-resolution.
Tasks. Work with climate data from km-scale and CMIP models; develop a framework to analyse climate extremes and their changes Python; investigate, visualise, and discuss results
Required technical skills (or willingness to learn): Python (xarray), Git, Linux command line
Literature: Brunner et al. (2025)
Lukas Brunner | 4
co-supervision with Benjamin Poschlod
AMOC shutdown impacts on European climate from km-scale models
(c) MPI-M/DKRZ
Background. The advent of global, km-scale models provides an exciting new source of climate information. These models provide a step change compared to the more traditional CMIP models in their resolution and their handling of convection.
Questions. How does European (extreme) climate react to a shutdown of the Atlantic Meridional Overturning Circulation (AMOC)? What additional details on impacts can we learn from simulating AMOC shutdown at km-scale? How do the results differ from conventional CMIP models?
Tasks. Work with climate data from dedicated km-scale ICON runs; develop a framework to analyse data in Python; investigate, visualise, and discuss results
Required technical skills (or willingness to learn): Python (xarray), Git
Literature: Westen and Baatsen (2025), Brunner et al. (2025)
Lukas Brunner | 5
co-supervision with Keno Riechers (MPI-M)
Discriminate climate observations, predictions, and projections using (explainable) machine learning methods
Figure: adapted from Befort et al. 2022
Background. The application of machine learning methods for climate sciences allows for novel questions and types of analysis on the intersection between climate and data science.
Questions. (How long) can we distinguish climate predictions from the observations they were initialised with? For how long does the initialisation make them distinct from free running climate projections? What geographical areas provide most skill for a separation of observations, predictions, and projections?
Tasks. Work with climate data from CMIP models and observations; develop machine learning classifiers and explainable machine learning tools, building on earlier work; investigate, visualise, and discuss results
Required technical skills (or willingness to learn): Python (xarray, keras), Git
Lukas Brunner | 6
co-supervision with Leo Borchert
Daily temperature increments in models and observations
Figure: Brunner
Background. Many model evaluation methods focus on climatological means and their variability on different time-scales. Changes between time steps such as temperatures between one day and the next (daily increments) are less investigated.
Questions. How are daily temperature increments distributed around the globe and between models and observations? What role does model resolution play (potentially drawing on km-scale models)? Do the increments change with global warming?
Tasks. Work with climate data from CMIP models and observations; develop a framework to analyse and compare increments in Python; investigate, visualise, and discuss results
Required technical skills (or willingness to learn): Python (xarray), Git
Literature: Xu et al. (2020)
Lukas Brunner | 7
Genealogy from the first coupled models�to the latest km-scale models
Top: Edwards 2011, Bottom: Masson and Knutti 2011
Background. Modern climate models build on a long history of development, leading to complex interdependencies between them. Model output has been used to trace resulting similarities.
Questions. How does the next generation of km-scale models relate to older models? Can we trace model genealogy all the way from the first coupled models in the 1990 to today?
Tasks. Work with output from more than 150 climate models; apply standard clustering approaches and novel machine learning techniques to reveal dependencies; investigate, visualise, and discuss results
Required technical skills (or willingness to learn): Python (xarray, keras), Git, general knowledge of climate models
Literature: Masson and Knutti 2011, Merrifield et al. 2023
Lukas Brunner | 8
Earlier topics
Follow-ups might be possible…
Lukas Brunner | 9
Separating source of uncertainty of future climate extremes
Figure: Lehner et al. 2020
Background. Projections of future climate and in particular of extremes are inherently uncertainty. This uncertainty emerges from three main sources: societal and technological development (scenario), imperfect representation of the climate system in numerical models (model), internal climate variability (Int. variability).
Questions. What are the relative contributions of these three �main sources to the overall uncertainty? How does this depend �on the extreme index, region, and time horizon considered? �How does it compare to mean values?
Tasks. Work with climate data from CMIP models; implement a�partitioning of uncertainty into its sources based on existing �work; investigate, visualise, and discuss results
Required technical skills (or willingness to learn): Python, Git, �general knowledge of climate models
Literature: Hawkins and Sutton 2009, Lehner et al. 2020
Lukas Brunner | 10
Global mean temperature
Link between blocking in different locations�and temperature extremes
Figure: Brunner et al. 2018
Background. Blocking effects on temperature depend strongly on the location of the blocking and the season. In a recently published study a systematic bias in the definition of temperature extremes which might affect the link to blocking.
Questions. How is blocking at different locations connected to temperature extremes. (How) is the connection affected by a recently discovered bias in temperature extremes?
Tasks. Work with climate data from CMIP models and observations; develop a framework to analyse and link blocking and extremes in Python; investigate, visualise, and discuss results
Required technical skills (or willingness to learn): Python (xarray), Git
Literature: Brunner et al. 2024, Brunner et al. 2018
Lukas Brunner | 11
How will heat wave properties develop under climate change?
Background�Climate change impacts properties of heat waves such as intensity with severe implications for society. However, other properties such as duration are less investigated.
Research question�How do heat wave properties respond to global warming in different regions and based on different metrics?
Tasks�Build a framework to identify heat waves in CMIP6 data based on Python. Investigate and interpret changes and analyse possible drivers.
Lukas Brunner | 12