A critical evaluation of present-day and future surface ozone as simulated by global chemistry-climate models in the Atmospheric Chemistry & Climate Model Intercomparison Project (ACCMIP)

Abstract:

In evaluating a future scenario for air quality, one can identify four major causal factors: (1) global emissions that alter atmospheric composition and thence baseline levels of surface ozone (O₃); (2) global changes in climate that also alter these baselines (e.g., temperature, water vapor); (3) climate-driven changes in the meteorological regimes of polluted regions that lead to air quality extreme (AQX) episodes; and (4) changes in the efficacy of local emissions to produce pollution within a governance region. While these factors are all part of a coupled system, a model that combines all would be difficult to verify. Thus an assessment approach would be to evaluate each factor separately using observations and an ensemble of models. In this study, we focus on factor (3), evaluating the ability of the models in the Atmospheric Chemistry & Climate Model Intercomparison Project (ACCMIP) to reproduce the observed present-day climatology (e.g. diurnal/seasonal cycles, AQX episode size) of surface O₃ in North America (NA) and Europe (EU). We then characterize future changes within these domains as well as south Asia (SA) for two experiments of RCP8.5 climate, one with O₃ precursor emissions representative of the 2100s (RCP8.5) and one representative of the 2000s (Cl2100Em2000). We find that most models simulate the observed climatology well, albeit biased high over the range of each domain’s probability distribution (Fig. 1). For RCP8.5, the ensemble mean shows an increase of ~10% in the mean annual maximum daily 8-h average (MDA8) over all domains, with the largest changes in winter months. For Cl2100Em2000, NA shows a small increase (+1%) in annual mean MDA8 while EU and SA show small decreases (-2% and -3%, respectively). Also for RCP8.5, most models show decreases in the mean size (S) and mean duration (D) of AQX episodes in EU (S = -28%, D = -17%) and increases in SA (+54%, +15%). The ensemble mean shows decreases in D (-7%) and increases in S (+21%) in NA, but the sign of the change in S is split between the models. For Cl2100Em2000, we find similar but smaller changes in EU (-10%, -11%) and SA (+16%, +5%), and small decreases in NA (-3%, -8%). Thus we conclude future changes in AQX episode size to be driven more so by changes in precursor emissions rather than climate. This also applies to changes in duration for SA but is reversed for NA and EU.