Florian Gallusser, M.Sc. - Data Science Chair

Florian Gallusser

Chair of Data Science (Informatik X)
University of Würzburg
Campus Hubland Nord
Emil-Fischer-Straße 50
97074 Würzburg
Germany

Email: florian.gallusser[at]uni-wuerzburg.de

Phone: (+49 931) 31-88409

Office: Room 50.03.009 (Zentrum für Künstliche Intelligenz und Data Science (CAIDAS))

About Me

I joined the Data Science Chair at the University of Würzburg in December 2023 for my PhD studies after having received my Master's degree in Computer Science from the University of Bonn in the same year.

In the context of the project BigData@Geo 2.0 my work currently lies at the intersection of Deep Learning and Climate Modeling.

If you are looking for a thesis topic or a research assistant position in those fields, don't hesitate to contact me!

Projects

BigData@Geo 2.0

Publications

2025[ to top ]

Exploring Design Choices for Autoregressive Deep Learning Climate Models. Gallusser, Florian; Hentschel, Simon; Krause, Anna; Hotho, Andreas. In ICLR 2025 Workshop on Tackling Climate Change with Machine Learning. 2025.
- [ Abstract ]
- [ BibTeX ]
- [ URL ]
- [ Download ]
Deep Learning (DL) models have achieved state-of-the-art performance in medium-range weather prediction (MWP) but often fail to maintain physically consistent rollouts beyond 14 days. In contrast, a few atmospheric models demonstrate stability over decades, though the key design choices enabling this remain unclear. This study quantitatively compares the long-term stability of three prominent DL-MWP architectures — FourCastNet, SFNO, and ClimaX — trained on ERA5 reanalysis data at 5.625° resolution. We systematically assess the impact of autoregressive training steps, model capacity, and choice of prognostic variables, identifying configurations that enable stable 10-year rollouts while preserving the statistical properties of the reference dataset. Notably, rollouts with SFNO exhibit the greatest robustness to hyperparameter choices, yet all models can experience instability depending on the random seed and the set of prognostic variables.

@inproceedings{${label}, abstract = {Deep Learning (DL) models have achieved state-of-the-art performance in medium-range weather prediction (MWP) but often fail to maintain physically consistent rollouts beyond 14 days. In contrast, a few atmospheric models demonstrate stability over decades, though the key design choices enabling this remain unclear. This study quantitatively compares the long-term stability of three prominent DL-MWP architectures — FourCastNet, SFNO, and ClimaX — trained on ERA5 reanalysis data at 5.625° resolution. We systematically assess the impact of autoregressive training steps, model capacity, and choice of prognostic variables, identifying configurations that enable stable 10-year rollouts while preserving the statistical properties of the reference dataset. Notably, rollouts with SFNO exhibit the greatest robustness to hyperparameter choices, yet all models can experience instability depending on the random seed and the set of prognostic variables.}, author = {Gallusser, Florian and Hentschel, Simon and Krause, Anna and Hotho, Andreas}, booktitle = {ICLR 2025 Workshop on Tackling Climate Change with Machine Learning}, keywords = {auhtor:hentschel}, title = {Exploring Design Choices for Autoregressive Deep Learning Climate Models}, year = 2025 }

2024[ to top ]

Global Vegetation Modeling with Pre-Trained Weather Transformers. Janetzky, Pascal; Gallusser, Florian; Hentschel, Simon; Hotho, Andreas; Krause, Anna. 2024.
- [ Abstract ]
- [ BibTeX ]
- [ URL ]
Accurate vegetation models can produce further insights into the complex inter- action between vegetation activity and ecosystem processes. Previous research has established that long-term trends and short-term variability of temperature and precipitation affect vegetation activity. Motivated by the recent success of Transformer-based Deep Learning models for medium-range weather forecast- ing, we adapt the publicly available pre-trained FourCastNet to model vegetation activity while accounting for the short-term dynamics of climate variability. We investigate how the learned global representation of the atmosphere’s state can be transferred to model the normalized difference vegetation index (NDVI). Our model globally estimates vegetation activity at a resolution of 0.25◦ while relying only on meteorological data. We demonstrate that leveraging pre-trained weather models improves the NDVI estimates compared to learning an NDVI model from scratch. Additionally, we compare our results to other recent data-driven NDVI modeling approaches from machine learning and ecology literature. We further provide experimental evidence on how much data and training time is necessary to turn FourCastNet into an effective vegetation model. Code and models will be made available upon publication.

@misc{janetzky2024global, abstract = {Accurate vegetation models can produce further insights into the complex inter- action between vegetation activity and ecosystem processes. Previous research has established that long-term trends and short-term variability of temperature and precipitation affect vegetation activity. Motivated by the recent success of Transformer-based Deep Learning models for medium-range weather forecast- ing, we adapt the publicly available pre-trained FourCastNet to model vegetation activity while accounting for the short-term dynamics of climate variability. We investigate how the learned global representation of the atmosphere’s state can be transferred to model the normalized difference vegetation index (NDVI). Our model globally estimates vegetation activity at a resolution of 0.25◦ while relying only on meteorological data. We demonstrate that leveraging pre-trained weather models improves the NDVI estimates compared to learning an NDVI model from scratch. Additionally, we compare our results to other recent data-driven NDVI modeling approaches from machine learning and ecology literature. We further provide experimental evidence on how much data and training time is necessary to turn FourCastNet into an effective vegetation model. Code and models will be made available upon publication.}, author = {Janetzky, Pascal and Gallusser, Florian and Hentschel, Simon and Hotho, Andreas and Krause, Anna}, keywords = {author:KRAUSE}, title = {Global Vegetation Modeling with Pre-Trained Weather Transformers}, year = 2024 }

2023[ to top ]

Multi-Task Transformer Visualization to build Trust for Clinical Outcome Prediction. Antweiler, Dario; Gallusser, Florian; Fuchs, Georg. In 2023 Workshop on Visual Analytics in Healthcare (VAHC), pp. 21–26. 2023.
- [ Abstract ]
- [ BibTeX ]
Clinical decision support systems based on machine learning are a rising application in healthcare. Early detection of deteriorating conditions provide the opportunity for medical intervention in hospital patients. Recent approaches increasingly rely on Large Language Models such as BERT, because patient data is often in the form of structured temporal data. These models are notoriously hard to interpret and therefore to trust, while precisely trust is an essential principle for technology in healthcare. We develop a visual analytics system to inspect, compare, and explain pre-trained transformer models for a given clinical outcome prediction task. The work is developed on the basis of a large hospital patient dataset and prediction tasks for acute kidney injury and heart failure. Discussion with healthcare professionals confirms that our system can lead to a faster decision process and improved modeling results.

@inproceedings{10356804, abstract = {Clinical decision support systems based on machine learning are a rising application in healthcare. Early detection of deteriorating conditions provide the opportunity for medical intervention in hospital patients. Recent approaches increasingly rely on Large Language Models such as BERT, because patient data is often in the form of structured temporal data. These models are notoriously hard to interpret and therefore to trust, while precisely trust is an essential principle for technology in healthcare. We develop a visual analytics system to inspect, compare, and explain pre-trained transformer models for a given clinical outcome prediction task. The work is developed on the basis of a large hospital patient dataset and prediction tasks for acute kidney injury and heart failure. Discussion with healthcare professionals confirms that our system can lead to a faster decision process and improved modeling results.}, author = {Antweiler, Dario and Gallusser, Florian and Fuchs, Georg}, booktitle = {2023 Workshop on Visual Analytics in Healthcare (VAHC)}, keywords = {clinical_outcome_prediction}, month = 10, pages = {21-26}, title = {Multi-Task Transformer Visualization to build Trust for Clinical Outcome Prediction}, year = 2023 }

Picture credits