Paul Griffiths

Surface ozone is an air pollutant that contributes to hundreds of thousands of premature deaths annually. Accurate short-term ozone forecasts may allow improved policy actions to reduce the risk to human health. However, forecasting surface ozone is a difficult problem as its concentrations are controlled by a number of physical and chemical processes that act on varying timescales. We implement a state-of-the-art transformer-based model, the temporal fusion transformer, trained on observational data from three European countries. In four-day forecasts of daily maximum 8-hour ozone (DMA8), our novel approach is highly skillful (MAE = 4.9 ppb, coefficient of determination ) and generalizes well to data from 13 other European countries unseen during training (MAE = 5.0 ppb, ). The model outperforms other machine learning models on our data (ridge regression, random forests, and long short-term memory networks) and compares favorably to the performance of other published deep learning architectures tested on different data. Furthermore, we illustrate that the model pays attention to physical variables known to control ozone concentrations and that the attention mechanism allows the model to use the most relevant days of past ozone concentrations to make accurate forecasts on test data. The skillful performance of the model, particularly in generalizing to unseen European countries, suggests that machine learning methods may provide a computationally cheap approach for accurate air quality forecasting across Europe.

Machine learning (ML) is transforming atmospheric chemistry, offering powerful tools to address challenges in tropospheric ozone research, a critical area for climate resilience and public health. As in adjacent fields, ML approaches complement existing research by learning patterns from ever-increasing volumes of atmospheric and environmental data relevant to ozone. We highlight the rapid progress made in the field since Phase 1 of the Tropospheric Ozone Assessment Report (TOAR), focussing particularly on the most active areas of research, namely short-term ozone forecasting, emulation of atmospheric chemistry and the use of remote sensing for ozone estimation. This review provides a comprehensive synthesis of recent advancements, highlights critical challenges, and proposes actionable pathways to develop ML in ozone research. Further advances hinge on addressing domain-specific issues such as the dependence of ozone concentrations on several poorly observed precursor species, as well as making progress on generic ML challenges such as the definition of suitable benchmarks and developing robust, explainable models. Reaping the full potential of ML for ozone research and operational applications will require close collaborations across atmospheric chemistry, ML and computational science and vigilant pursuit of the rapid developments in adjacent fields.

Global surface warming has accelerated since around 2010, relative to the preceding half century1–3. This has coincided with East Asian efforts to reduce air pollution through restricted atmospheric aerosol and precursor emissions4,5. A direct link between the two has, however, not yet been established. Here we show, using a large set of simulations from eight Earth System Models, how a time-evolving 75% reduction in East Asian sulfate emissions partially unmasks greenhouse gas-driven warming and influences the spatial pattern of surface temperature change. We find a rapidly evolving global, annual mean warming of 0.07 ± 0.05 °C, sufficient to be a main driver of the uptick in global warming rate since 2010. We also find North-Pacific warming and a top-of-atmosphere radiative imbalance that are qualitatively consistent with recent observations. East Asian aerosol cleanup is thus likely a key contributor to recent global warming acceleration and to Pacific warming trends.

The Aerosol Chemistry Model Intercomparison Project (AerChemMIP) was endorsed by the Coupled Model Intercomparison Project 6 (CMIP6) and was designed to quantify the climate and air quality impacts of aerosols and chemically reactive gases. AerChemMIP provided the first consistent calculation of effective radiative forcing (ERF) for a wide range of forcing agents, which was a vital contribution to the Sixth Assessment Report (AR6) of the Intergovernmental Panel on Climate Change. It supported the quantification of composition–climate feedback parameters and the climate response to short-lived climate forcers (SLCFs), as well as enabled the future impacts of air pollution mitigation to be identified, and the study of interactions between climate and air quality in a transient simulations. Here we review AerChemMIP in detail and assess the project against its stated objectives, its contribution to the CMIP6 project, and the wider scientific efforts designed to understand the role of aerosols and chemistry in the Earth system. We assess the successes of the project and the remaining challenges and gaps. We conclude with some recommendations that we hope will provide input to planning for future MIPs in this area. In particular, we highlight the necessity of sufficient ensemble size for the attribution of regional climate responses and the need for coordination across projects to ensure key science questions are addressed. Summary data for CMIP6 and AerChemMIP models such as model components, model configurations, and emergent quantities are included.