Software

Interactive PM2.5 forecasting for Europe at 1 km resolution

A web-based studio built on top of the CRAN-PM Vision Transformer model. CRAN-PM Studio lets you explore daily PM2.5 forecasts across Europe at 1 km resolution, browse interactive maps and time series, and inspect station-level diagnostics against the European Environment Agency (EEA) network. Powered by cross-resolution attention that fuses 25 km meteorological context with 1 km local PM2.5, plus elevation-aware self-attention and wind-guided cross-attention for physically consistent predictions in complex terrain.

MaintainersAmmar Kheder, Zhi-Song Liu, Michael Boy

Atmospherically Relevant Chemistry and Aerosol box model

Zero-dimensional process model for gas-phase chemistry coupled with aerosol formation and growth. Used for in-depth process studies and for generating training data for machine-learning emulators.

MaintainersPetri Clusius, Carlton Xavier, Michael Boy

Column model for biosphere–atmosphere interactions

One-dimensional chemistry-transport model coupling boundary-layer meteorology, gas-phase chemistry, and aerosol dynamics. Co-developed and maintained at AMC-Lahti for studies on biogenic VOCs, new-particle formation, and air-quality processes.

MaintainersMichael Boy, Petri Clusius, Carlton Xavier, Benjamin Foreback

Topography-aware pollutant-flow learning for high-resolution air quality

Neural model that explicitly incorporates terrain topography to predict pollutant transport at high spatial resolution. Used for fine-scale air-quality forecasting in regions with complex orography.

MaintainersAmmar Kheder, Helmi Toropainen, Wenqing Peng, Zhi-Song Liu, Michael Boy

Lagrangian air-mass trajectories coupled with column chemistry

Coupling of the FLEXPART particle-dispersion model with the SOSAA column model, used to track the chemical evolution of air masses arriving at a site (e.g., Beijing severe-haze events, Arctic transport).

MaintainersBenjamin Foreback, Petri Clusius

AI surrogate for stiff chemistry kinetics

Neural-network emulator that learns to integrate the stiff systems of ODEs describing atmospheric chemistry, enabling fast and differentiable predictions for air-quality forecasting and inverse problems.

MaintainersZhi-Song Liu, Petri Clusius, Michael Boy