Research programme
My research studies forecasting systems as a complete chain: from distributed observations and learned spatial structure to evaluation, model selection and deployment decisions.
Central premise
Environmental forecasting combines observations from sensors, satellites, reanalysis products and contextual data. Those sources are neither perfectly aligned nor permanently stable.
I develop and evaluate models that represent spatial relationships explicitly, learn temporal dynamics at the appropriate scales and expose the uncertainty behind the final model-selection decision.
Four connected areas
The areas below are not separate silos. Each contributes to a forecasting system that can be inspected from observation to action.
Evaluation and selection methods for forecasting systems that must remain dependable under missing data, sensor failures, spatial misalignment and changing deployment conditions.
Graph neural networks that capture relationships across distributed environmental sensors, cities and global observational grids.
The effect of missingness, observational uncertainty, aggregation and measurement inconsistencies on learned models and the decisions built on them.
AI methods for air quality, atmospheric aerosols, greenhouse-gas concentrations, urban noise and other environmental phenomena.
Research chain
Reliability can fail at any point, so each stage is treated as part of the research problem.
Sensor, satellite and contextual environmental variables.
Graphs, temporal windows and multi-scale dependencies.
Neural architectures evaluated across realistic datasets.
Missingness, misalignment, degradation and reference uncertainty.
Traceable model selection under explicit constraints.
Collaboration
I am interested in research involving large environmental observational datasets, sensor-network quality control, graph-based forecasting, robustness under distribution shift, meteorological extremes and reproducible environmental AI.
Discuss a collaboration