Separating emissions and meteorological impacts on peak ozone concentrations in Southern California using generalized additive modeling

Abstract

Ozone levels have been declining in the Los Angeles, CA, USA area for the last four decades, but there was a recent uptick in the 4th highest daily maximum 8-h (MDA8) ozone concentrations from 2014 to 2018 despite continued reductions in the estimated precursor emissions. In this study, we assess the emissions and meteorological impacts on the 4th highest MDA8 ozone concentrations to better understand the factors affecting the observed MDA8 ozone using a two-step generalized additive model (GAM)/least squares approach applied to the South Coast Air Basin (SoCAB) for the 1990 to 2019 period. The GAM model includes emissions, meteorological factors, large-scale climate variables, date, and the interactions between meteorology and emissions. A least squares method was applied to the GAM output to better capture the 4th highest MDA8 ozone. The resulting two-step model had an R2 of 0.98 and a slope of 1 between the observed and predicted 4th highest MDA8 ozone. Emissions and the interactions between the maximum temperature and emissions explain most of the variation in the peak MDA8 ozone concentrations. Declining emissions have lowered the 4th highest MDA8 ozone concentration. Meteorology explains the higher than expected 4th-high, ozone levels observed in 2014–2018, indicating that meteorology was a stronger forcer than the continued reductions in emissions during that time period. The model was applied to estimate future ozone levels. Meteorology developed from climate modeling of the representative concentration pathway (RCP) scenarios, and two sets of emissions were used in the application. The modeling results indicated climate trends will push ozone levels slightly higher if no further emissions reductions are realized and that of two emissions trajectories modeled, the more stringent is required to reliably meet the federal ozone standard given annual meteorological variability.