Evaluation of MERRA-2-Based Ozone Profile Simulations: Recent Successes and Ongoing Challenges

Model simulations of ozone (O₃) are useful for filling in temporal and geographical observational gaps and providing interpretation for observed O₃ variability and trends. Extending a previous analysis that examined global ozonesonde profile variability and its link to meteorological conditions, we use global ozonesonde networks and satellite total column O₃ to evaluate nearly 40 years (1980 – 2016) of model O₃ profiles from two simulations using reanalysis meteorology. Both models are constrained by meteorology from the NASA Modern-Era Retrospective Analysis for Research Applications Version 2 (MERRA-2), and both run versions of the NASA Global Modeling Initiative (GMI) chemical mechanism. The first model (M2 GMI Replay) runs a previous version of the GMI mechanism, but uses the Replay technique to recalculate meteorological variables every 15 minutes in the model run. The second model (GMI CTM) uses the latest version of the GMI mechanism in a chemistry transport model that uses the 3-hour average MERRA-2 meteorology fields. Over 50,000 ozonesonde profiles from 38 stations around the globe provide a highly vertically resolved reference with which to compare the differing implementations of the models. In this study, we report four major findings: 1) Ozonesonde measurements are highly correlated (r > 0.9) with both models, particularly at and above the upper troposphere/lower stratosphere. The O₃ correlations and biases continuously improve in later years in the model runs. 2) In general, tropospheric O₃ in GMI CTM is 10 – 15 % lower than M2 GMI Replay. Biases near the surface average 5 – 15 ppbv (10 – 30 %) higher than the sondes in M2 GMI Replay, but are near zero in GMI CTM. 3) Both models struggle to reproduce observed mid-tropospheric O₃ enhancements in the tropics, especially in biomass burning affected portions of the Atlantic where biases are frequently -15 ppbv (-25 %) at 5 km. 4) The assimilation of data into MERRA-2 from a new generation of microwave satellite instruments starting in 1998 introduces a step change in total O₃ in both models. This change has a larger effect on GMI CTM, leading to an increase in mid-latitude total O₃ of >10 DU after 1998, and worsening the agreement with sonde and satellite total O₃ observations. Excessive transport of O₃ to the mid-latitude lower stratosphere, identified by differences in stratospheric tracers in both models, leads to the high total O₃ in GMI CTM. We demonstrate the benefits of the updated GMI mechanism for improving simulated lower tropospheric O₃ (GMI CTM), and the Replay technique (M2 GMI Replay) for dampening the effects of the data assimilation change on modeled lower stratospheric and total O₃. The figure shows latitude-height curtains of median O₃ mixing ratio differences between all ozonesonde profiles (~50,000 samples) and coincident M2 GMI Replay (a) and GMI CTM (b) O₃ profiles. Blue colors indicate that the modeled O₃ is biased low, and red colors indicate that the model is biased high. The site locations are listed at the top of the figure at their respective site latitudes.