Dr. Willi Schimmel

Mathematical Modelling · Atmospheric Science · Scientific Programming · Data Science

Mathematics gives atmospheric complexity a shape we can model, test against observations, and refine in code.

Email willischimmel@gmx.de
LinkedIn Willi Schimmel
GitHub KarlJohnsonnn
ORCID 0000-0001-8428-6445
Base Leipzig, Germany

Operational MLCloudnet integration

Invited TalkEGU 2025

NumericsHPC & sparse solvers

Field WorkChile & Switzerland

CoordinationResearch projects

The Arc

From industrial mechanic to postdoctoral researcher: a non-linear path through applied mathematics, numerical weather-prediction development, custom machine-learning models, and atmospheric field data. Each stage added a practical layer to how I solve research and software problems, and over the years I have also kept improving how I plan, coordinate, and manage projects.

Industrial Mechanic → Abitur → Applied Mathematics Hands-on systems intuition carried from the factory floor into scientific computing and solver design.

Master Student → Fast Numerical Solvers Fortran solvers, sparse linear algebra, and HPC workflows for atmospheric multiphase chemistry.

PhD Candidate → Machine Learning Models Custom neural-network models that moved from thesis work into community-wide operational use.

Selected Highlights

Operational Impact Research code used beyond the thesis

VoodooNet moved from a PhD prototype into the operational Cloudnet processing chain maintained by FMI.

Scientific Visibility Invited EGU 2025 contribution

Session AS1.14: cloud-microphysics perturbation work combining remote sensing, in-situ observations, and COSMO-SPECS modelling. Abstract doi:10.5194/egusphere-egu25-12739.

Research Coordination Better structure for complex work

Field-campaign experience and proposal-writing training support how I plan, communicate, and finish collaborative research software.

Experience

Post-Doc Researcher

2023 – present

TROPOS Leipzig · Modeling of Atmospheric Processes

Cloud-seeding simulations with COSMO-SPECS in hectometer-scale large-eddy mode.
3D spectral-bin microphysics, cloud tracking, Lagrangian analysis, and radar forward operators.
ETH Zurich CLOUDLAB collaboration; validation with field observations in Switzerland and Chile.

PhD Researcher

2018 – 2022

Leipzig University · Remote Sensing & Arctic System / TROPOS

Built PyTorch + CUDA model for hydrometeor classification in Doppler cloud-radar data.
Multi-sensor retrieval methods and data-processing workflows for atmospheric targets.
Code integrated into the operational Cloudnet processing chain.

Research Assistant

2014 – 2018

TROPOS Leipzig · Modeling of Atmospheric Processes

Fortran solver for atmospheric chemistry and combustion kinetics.
Sparse linear algebra and numerical optimization for stiff ODE systems.
Graph-theoretical methods to reduce complex reaction networks.

Industrial Mechanic

2006 – 2010

Sitec Industrietechnologie GmbH · Chemnitz

Customized assembly systems, laser machine tools, and industrial prototypes.
Practical intuition that still shapes computational problem-solving.

Toolkit

Languages: Python, R, Fortran, Bash, CUDA, MATLAB, basic C/C++
Numerics: Sparse linear algebra, numerical optimization, fast numerical solvers, stiff ODE systems, HPC, parallel computing
ML and AI: PyTorch, GPU workflows, supervised learning, deployment
Platforms: GNU/Linux, macOS, Git, xarray, numpy, scipy, pandas, LaTeX
Research Coordination: Field campaigns, proposal writing, collaborative model validation, scientific communication

Selected Projects

2025 Cminor v1.0

Advanced graph-based reduction for large multiphase chemical systems, supporting faster simulations and a GMD publication.

2023 – present Cloud-seeding modeling (COSMO)

Coordinates LES ensembles, forward operators, and field validation to compare modeled and observed cloud responses.

2020 – 2022 VoodooNet

Built PyTorch/CUDA cloud-radar classifier that turned thesis research into operational atmospheric data processing.

2014 – 2018 Multiphase Chemistry Solver

Developed Fortran solver workflows for stiff chemistry mechanisms using sparse matrices, Rosenbrock methods, and HPC practice.

Talks & peer-reviewed publications

EGU General Assembly 2025 · Session AS1.14 · Invited / solicited talk Combined Remote-Sensing, In-Situ and Modelling of Cloud Microphysical Perturbations in Supercooled Stratus Clouds

Opened Atmospheric Sciences sessions with results from CLOUDLAB/PolarCAP work on aerosol-cloud interactions, field data, and COSMO-SPECS ensembles. Abstract doi:10.5194/egusphere-egu25-12739.

3rd Leibniz MMS Days 2018 · Leipzig · Conference talk Numerical simulation of large atmospheric multiphase mechanisms and detailed combustion kinetics

Presentation on the Fortran-based solver for large kinetic systems: Rosenbrock methods, sparse Jacobians, TROPOS and ChemKin mechanism formats, with examples from atmospheric multiphase and gas-phase combustion. Archived with doi:10.5446/35355.

Ohneiser, K., Seifert, P., Schimmel, W., et al. (2026). Ice-nucleating particle depletion in the wintertime boundary layer in the pre-Alpine region during stratus cloud conditions. Atmospheric Chemistry and Physics.
Ohneiser, K., Seifert, P., Schimmel, W., et al. (2025). Impact of seeder-feeder cloud interaction on precipitation formation. Atmospheric Chemistry and Physics.
Rug, L., Schimmel, W., Hoffmann, F., Knoth, O. (2025). Cminor v1.0 chemical mechanism integrator. Geoscientific Model Development.
Merker, J., Schimmel, W. (2023). On Sensitivity of Data Models w.r.t. Training Data. In Proceedings of the 2023 6th International Conference on Mathematics and Statistics, ACM.
Schimmel, W., Kalesse-Los, H., Maahn, M., et al. (2022). Identifying cloud droplets beyond lidar attenuation from vertically pointing cloud radar observations using artificial neural networks. Atmospheric Measurement Techniques.
Vogl, T., Maahn, M., Kneifel, S., Schimmel, W., Moisseev, D., Kalesse-Los, H. (2022). Using artificial neural networks to predict riming from Doppler cloud radar observations. Atmospheric Measurement Techniques.
Kalesse-Los, H., Schimmel, W., Luke, E., Seifert, P. (2022). Evaluating cloud liquid detection against Cloudnet using cloud radar Doppler spectra in a pre-trained artificial neural network. Atmospheric Measurement Techniques.
Hamadneh, T., Merker, J., Schimmel, W., Schuldt, G. (2022). Simplicial Bernstein form and positivity certificates for solutions obtained in a stationary digital twin by Bernstein Bubnov-Galerkin method. In 2022 5th International Conference on Mathematics and Statistics, ACM.