Multidecadal increases in global tropospheric ozone derived from ozonesonde and surface site observations: can models reproduce ozone trends?

Abstract

Abstract. Despite decades of effort, the drivers of global long-term trends in tropospheric ozone are not well understood, impacting estimates of ozone radiative forcing and the global ozone budget. We analyze tropospheric ozone trends since 1980 using ozonesondes and remote surface measurements around the globe and investigate the ability of two atmospheric chemical transport models, GEOS-Chem and MERRA2-GMI, to reproduce these trends. Global tropospheric ozone trends measured at 25 ozonesonde sites from 1990–2017 (nine sites since 1980s) show increasing trends averaging 1.8 ± 1.3 ppb per decade across sites in the free troposphere (800–400 hPa). Relative trends in sondes are more pronounced closer to the surface (3.5 % per decade above 700 hPa, 4.3 % per decade below 700 hPa on average), suggesting the importance of surface emissions (anthropogenic, soil NOx, impacts on biogenic volatile organic compounds (VOCs) from land use changes, etc.) in observed changes. While most surface sites (148 of 238) in the United States and Europe exhibit decreases in high daytime ozone values due to regulatory efforts, 73 % of global sites outside these regions (24 of 33 sites) show increases from 1990–2014 that average 1.4 ± 0.9 ppb per decade. In all regions, increasing ozone trends both at the surface and aloft are at least partially attributable to increases in 5th percentile ozone, which average 1.8 ± 1.3 ppb per decade and reflect the global increase of baseline ozone in rural areas. Observed ozone percentile distributions at the surface have shifted notably across the globe: all regions show increases in low tails (i.e., below 25th percentile), North America and Europe show decreases in high tails (above 75th percentile), and the Southern Hemisphere and Japan show increases across the entire distribution. Three model simulations comprising different emissions inventories, chemical schemes, and resolutions, sampled at the same locations and times of observations, are not able to replicate long-term ozone trends either at the surface or free troposphere, often underestimating trends. We find that ∼75 % of the average ozone trend from 800–400 hPa across the 25 ozonesonde sites is captured by MERRA2-GMI, and <20 % is captured by GEOS-Chem. MERRA2-GMI performs better than GEOS-Chem in the northern midlatitude free troposphere, reproducing nearly half of increasing trends since 1990 and capturing stratosphere–troposphere exchange (STE) determined via a stratospheric ozone tracer. While all models tend to capture the direction of shifts in the ozone distribution and typically capture changes in high and low tails, they tend to underestimate the magnitude of the shift in medians. However, each model shows an 8 %–12 % (or 23–32 Tg) increase in total tropospheric ozone burden from 1980 to 2017. Sensitivity simulations using GEOS-Chem and the stratospheric ozone tracer in MERRA2-GMI suggest that in the northern midlatitudes and high latitudes, dynamics such as STE are most important for reproducing ozone trends in models in the middle and upper troposphere, while emissions are more important closer to the surface. Our model evaluation for the last 4 decades reveals that the recent version of the GEOS-Chem model underpredicts free tropospheric ozone across this long time period, particularly in winter and spring over midlatitudes to high latitudes. Such widespread model underestimation of tropospheric ozone highlights the need for better understanding of the processes that transport ozone and promote its production.