PySDM
pure-Python, open-source implementation of the SDM algorithm
PySDM is built around a Pythonic implementation of the Super-Droplet Method Monte-Carlo scheme for coagulation (Shima et al. 2009). The package features two number-crunching backends: multi-threaded CPU backend using Numba and a CUDA GPU backend using ThrustRTC and CURandRTC. PySDM ships with a set of examples shipped as Jupyter-notebooks, and covering simulation setups spanning box-model cases, adiabatic parcel simulations, single-column and two-dimensional kinematic frameworks (see animation below). Documentation of the package covers the API as well as tutorials exemplifying usage of PySDM from Python, Julia and Matlab. Development of PySDM is hosted on GitHub.
Release history and papers on PySDM
PySDM v1 constituted the MSc thesis of Piotr Bartman (Jagiellonian University, 2020) who had developed PySDM API, adaptive numerical solvers and package architecture. Features and examples of PySDM v1 are summarised in a Bartman et al. 2022 JOSS paper. Subsequent developments had been largely spearheaded by Emily de Jong and Clare Singer during their PhD studies at the Caltech CliMA team, and are summarised in the de Jong, Singer et al. 2023 JOSS paper. The original collisional breakup scheme (maintaining the key constant-state-vector size trait of SDM) had been defined and validated in de Jong et al. 2023 GMD paper (with section 3 based on MSc thesis of Oleksii Bulenok (Jagiellonian University, 2023)). The original adaptive time-stepping scheme for SDM introduced in PySDM v1 has been the topic of further research by Emma Ware which has been summarised in Ware et al. 2025 arXiv preprint.
Publications using PySDM
- Bieli et al. 2022 (JAMES): "An Efficient Bayesian Approach to Learning Droplet Collision Kernels: Proof of Concept Using “Cloudy,” a New n-Moment Bulk Microphysics Scheme"
- de Jong e al. 2022 (JAMES): "Spanning the Gap From Bulk to Bin: A Novel Spectral Microphysics Method"
- de Jong et al. 2023 (GMD): "Breakups are complicated: an efficient representation of collisional breakup in the superdroplet method"
- Singer et al. 2023 (AMS Annual Meeting): "Surface-Active Organics: Results from a Particle-Based Microphysics Model Calibrated with Laboratory Experiments"
- D'Aquino et al. 2024 (SoftwareX): "PyPartMC: A Pythonic interface to a particle-resolved, Monte Carlo aerosol simulation framework"
- Azimi et al. 2024 (JAMES): "Training Warm-Rain Bulk Microphysics Schemes Using Super-Droplet Simulations"
- Arabas et al. 2025 (JAMES): "Immersion Freezing in Particle-Based Aerosol-Cloud Microphysics: A Probabilistic Perspective on Singular and Time-Dependent Models"
- Lüttmer et al. 2025 (EGU General Assembly): "Cirrus formation in particle-based aerosol-cloud microphysics"
- Ware et al. 2025 (arXiv): "Adaptive time-stepping for the Super-Droplet Method Monte Carlo collision-coalescence scheme"
🎓 Student project opportunities
- Distributed-memory parallelism for PySDM using Numba-MPI
- Brownian coagulation in PySDM and development of aerosol-related examples
- Super-particle attribute initialisation using copulae
- Immersion freezing and coagulation: effective freezing-temperature spectra under singular and time-dependent models
- Venusian cloud particle-based microphysics (e.g., using the models introduced in Ando et al. 2025)
- Logic for serialising and deserialising simulation state
Contributors from our team
Piotr Bartman-Szwarc
Oleksii Bulenok
Emma Ware
Aleksandra Strząbała
Sanket Bhiogade
Konrad Bodzioch
Agnieszka Żaba
Agnieszka Makulska
