A Geographically Weighted Gaussian Process Regression (GW-GPR) emulator of anthropogenic PM2.5 from the GEOS-Chem High Performance (GCHP) 13.0.0 global chemical transport model

Abstract

Chemical transport modelling has been an essential tool to study the impacts of socio-economic changes and policies on air quality and associated social costs due to human health impacts. However, high computational and human resource demands limit the use of state-of-the-art chemical transport models outside of the atmospheric science community. We address this limitation by training Geographically Weighted Gaussian Process Regressors (GW-GPR) on the outputs of a series of perturbation experiments from the GEOS-Chem High Performance global chemical transport model (GCHP 13.0.0). The Gaussian Process Regressor relates changes in annual mean surface anthropogenic PM2.5 in each GCHP model grid cell to changes in short-lived air pollutant emissions and atmospheric CH4 and CO2 concentrations. In comparison to existing linearized and regionalized approaches, our method can account for sub-regional changes in air pollutant emission patterns and incorporates the non-linear response of secondary air pollutants to precursor and greenhouse gas emissions. We evaluate our emulator by predicting the global distribution of PM2.5 in 2050 (relative to 2014) under 4 sets of climate and air pollution control policy scenarios. The emulator reproduces grid cell scale changes in anthropogenic PM2.5 (R2=0.94–0.99 over the 4 scenarios tested), and associated global changes in premature mortalities at 95 % confidence level, while requiring <10 s of CPU time (vs. ∼3000 CPU hours for GCHP) for each scenario. We demonstrate the utility of the emulator by projecting global trends of population-weighted PM2.5 from the AerChemMIP ensemble, where the emulator prediction falls within the ensemble range. To our knowledge, the GW-GPR emulator is the first global-scale emulator operating at grid cell scale with explicit consideration of non-linearities in atmospheric chemistry, climate change, and provides predictive uncertainties. The accuracy, speed and simplicity of the emulator also show the capability of machine learning algorithms in emulating global atmospheric chemistry models, and in making atmospheric chemistry modelling accessible for global climate/air pollution scenario analysis and integrated assessment.