Building a ML model for improving climate change prediction

Air quality models are used to predict air pollution abundances and provide input into health impact assessment modelling, in order to understand the health burden associated with population exposure to air pollutants. However, these models are limited by their large computational cost. This study used a machine learning model to accurately represent a complex air quality model to predict long–term air quality and the associated health impacts in China from changes in emissions.

The emulator was built using 55 separate high resolution simulations, using ~1 year of wall clock time on the ARC3 HPC system. The emulator allows thousands of simulations to be run in a small fraction of the time of the complex model, and was used to show the benefits of reducing many different pollution sources across different regions of China.

Sectoral contributions to ambient fine particulate matter (PM2.5, annual−mean) concentrations and exposure in China for 2012, 2020, and 2020 minus 2012.

References

Conibear, L., Reddington, C. L., Silver, B. J., Chen, Y., Knote, C., Arnold, S. R., & Spracklen, D. V. (2021). Statistical emulation of winter ambient fine particulate matter concentrations from emission changes in China. GeoHealth, 5, e2021GH000391. https://doi.org/10.1029/2021GH000391

Conibear, L., Reddington, C. L., Silver, B. J., Chen, Y., Arnold, S. R., & Spracklen, D. V. (2022). Emission sector impacts on air quality and public health in China from 2010 to 2020. GeoHealth, 6, e2021GH000567. https://doi.org/10.1029/2021GH000567