Ozone is a central species in atmospheric chemistry. It is a major air pollutant and greenhouse gas and the primary source of the hydroxyl radical (OH), which is the main tropospheric oxidant (Monks et al., 2015). It is produced within the troposphere by complicated chemical mechanisms involving hydrogen oxide radicals (HO x ≡ OH + peroxy), nitrogen oxide radicals (NO x ≡ NO + NO 2 ), volatile organic compounds (VOCs), and ozone itself. It is extensively observed from surface sites, aircraft, sondes, and satellites and is thus an important indicator of skill for chemical transport models (Hu et al., 2017). At the same time, comparisons with observations can be successful for the wrong reasons. Extensive intercomparisons of global models often show similar tropospheric ozone burdens but large differences in chemical source and sink magnitudes (Wu et al., 2007;Young et al., 2018), implying large differences in sensitivity to perturbations. This is a particular problem for chemistry-climate models that aim to quantify chemical feedbacks on climate change.
Here we compare two state-of-the-science atmospheric chemistry modules, GEOS-Chem and CAM-chem, within the Community Earth System Model (CESM2) (Danabasoglu et al., 2020). CAM-chem is the resident atmospheric chemistry module in CESM2 and as such has a large user base (Lamarque et al., 2012;Tilmes et al., 2015Tilmes et al., , 2016;;Emmons et al., 2020). GEOS-Chem is used by hundreds of research groups worldwide as an offline chemical transport model (CTM) driven by the GEOS archive of external meteorological data (Bey et al., 2001). Offline here is defined by contrast to online models that perform their own simulations of atmospheric dynamics (Brasseur and Jacob, 2017). GEOS-Chem is grid-independent and modularized so that the chemical module describing local operations in 1-D model columns (including emissions, chemistry, and deposition) is separated from the transport module (Long et al., 2015). This allows independent implementation of the GEOS-Chem chemical module in online models, where chemical transport is done as part of the simulation of atmospheric dynamics (Hu et al., 2018;Lin et al., 2020;Lu et al., 2020;Keller et al., 2021). The GEOS-Chem chemical module has been previously coupled to the WRF and GEOS meteorological models to investigate aerosolchemistry-climate feedbacks (Feng et al., 2021;Moch et al., 2022) and powers the GEOS global chemical forecasts (GEOS-CF) (Keller et al., 2021). The same GEOS-Chem scientific code base is used as in the offline CTM such that version updates developed for the CTM can be seamlessly passed on to the online applications. Fritz et al. (2022) implemented the GEOS-Chem chemical module in CESM2 as the first application of that module to an open-source Earth system model (ESM) for com-munity use. GEOS-Chem offers an alternative representation of atmospheric chemistry to CAM-chem within CESM2, contributing to the MUSICA (MUlti-Scale Infrastructure for Chemistry and Aerosols; Pfister et al., 2020) vision for CESM of allowing users to choose among a range of options for atmospheric chemistry. The GEOS-Chem emission component (HEMCO; Keller et al., 2014) has been previously implemented in MUSICA (Lin et al., 2021). Fritz et al. (2022) presented general comparisons between GEOS-Chem and CAM-chem in the CESM2 environment. They found good agreement between the two modules for stratospheric ozone but lower tropospheric ozone in GEOS-Chem due to tropospheric halogen chemistry not considered in CAMchem. They found several challenges in the implementation of the GEOS-Chem chemical module within CESM2. For example, CESM2 uses the MAM4 (Modal Aerosol Model version 4; Liu et al., 2016) modal aerosol microphysics to simulate aerosol-cloud interactions and aerosol-radiation interactions, while GEOS-Chem uses either bulk or sectional representations of aerosol microphysics. CESM2 does not couple convective transport with scavenging of water-soluble species, but this is a major process in the GEOS-Chem CTM to prevent unphysical buildup of water-soluble species in the upper troposphere (Balkanski et al., 1993;Liu et al., 2001). If convective transport and scavenging are applied sequentially, instead of being coupled, then water-soluble species can reach the upper troposphere in deep convective updrafts and disperse on the model grid scale to avoid scavenging. Indeed, Fritz et al. (2022) found large overestimates of uppertropospheric aerosol in GEOS-Chem within CESM2 as compared to the offline GEOS-Chem.
Our work builds on the Fritz et al. (2022) initial implementation of GEOS-Chem in CESM2 to address the previous challenges and to give a more thorough evaluation with observations and intercomparison with CAM-chem. We focus on tropospheric ozone and related oxidant chemistry from both a global perspective (ozonesonde and ATom-1 aircraft observations) and polluted conditions over East Asia (KORUS-AQ aircraft observations). KORUS-AQ, conducted in May-June 2016, is of particular interest because of the previously identified large differences between offline GEOS-Chem and CAM-chem in simulating the aircraft observations including 20-30 ppb differences in ozone (Park et al., 2021). We analyze the individual processes driving differences between GEOS-Chem and CAM-chem and use observations to arbitrate when possible. This process-based intercomparison of GEOS-Chem and CAM-chem leverages the unique capability of comparing these two major representations side by side in a common ESM environment where specific causes of model differences can be attributed to different representations of chemistry. As part of resolving differences in photolysis rates, we implement into CAM-chem the Fast-JX photolysis scheme used in GEOS-Chem (Bian and Prather, 2002), further contributing to the MUSICA vision of process-level modularization of atmospheric chemistry models.
We use a beta version of CESM 2.3 including the CAM6 Community Atmosphere Model (CAM tag version cam6_3_095), which has provided the basis for the integration of the GEOS-Chem module into the mainline CESM code. All simulations are for the year 2016 with an 18-month initialization period. The year was chosen for evaluation with the ATom (Wofsy et al., 2018) and KORUS-AQ (Crawford et al., 2021) aircraft campaigns. We use a global 0.9°× 1.25°g rid with 32 vertical layers up to 2 hPa. We use the "F" compsets in CESM, which use active atmosphere and land models with prescribed sea surface temperatures, sea ice, and greenhouse gases for current climate (CMIP6 SSP2-4.5 scenario). The model reproduces a given meteorological year by nudging winds and temperature (using the FCnudged configuration in CAM6) to a 3-hourly MERRA2 meteorological reanalysis produced by the NASA Global Modeling and Assimilation Office. This nudging is done with a 50 h relaxation time that allows CAM to generate its own physics, including the hydrological cycle and the effects of aerosols on clouds.
CAM-chem is the standard representation of tropospheric-stratospheric chemistry in CESM2, currently using the MOZART-TS1 (Model for OZone And Related chemical Tracers; Emmons et al., 2020) mechanism and the Modal Aerosol Model with four modes (MAM4; Liu et al., 2016) as the default. MOZART-TS1 includes 229 chemical species and 541 reactions. Photolysis is calculated using a lookup table based on the Tropospheric Ultraviolet and Visible (TUV) radiation model, which takes into account the impact of clouds but not aerosols (Kinnison et al., 2007). A sensitivity simulation developed for this project uses Fast-JX instead of the TUV lookup table for photolysis.
The CAM-chem version in our work uses HEMCO for emissions but is otherwise unmodified. HEMCO is the standard emission component of GEOS-Chem (Keller et al., 2014), now implemented in CESM as part of MUSICA (Lin et al., 2021). It allows the use of any emission inventories on any grid to be supplied to the model in netCDF format at runtime with options to add, supersede, and scale emissions.
Here we use the same emissions in GEOS-Chem and CAMchem processed through HEMCO. This includes global anthropogenic emissions from the CEDSv2 inventory (Community Emissions Data System;McDuffie et al., 2021) superseded by the KORUSv5 inventory (Woo et al., 2020) over East Asia. Fire emissions are from the GFED4.1s inventory (van der Werf et al., 2017;Randerson et al., 2018). HEMCO has extensions to use emission modules dependent on environmental variables, and this is applied to soil NO x emis-sions from Hudman et al. (2012) and ocean iodine emissions from Sherwen et al. (2016a, b). We otherwise use emissions computed from other modules in CESM to enforce consistency of the atmospheric chemistry simulation with other CESM components. This includes biogenic VOC emissions from MEGANv2.1 (Guenther et al., 2012) computed with the Community Land Model (CLM) and lightning NO x , dust, and sea salt emissions from CAM (Price et al., 1997;Mahowald et al., 2006a, b;Lamarque et al., 2012).
Unless explicitly written otherwise, GEOS-Chem in this work refers to the online implementation of the GEOS-Chem chemical module within the CESM2 model and not the offline CTM. We use GEOS-Chem version 14.1.1 (https://doi.org/10.5281/zenodo.7696632, The International GEOS-Chem User Community, 2023) with the addition of particulate nitrate (pNO 3 -) photolysis following Shah et al. (2023), which was subsequently implemented in version 14.2.0 (https://doi.org/10.5281/zenodo.8411433). The same GEOS-Chem chemical module and MERRA-2 meteorological fields are used in CESM2 and in the offline CTM simulations presented here. The GEOS-Chem chemical mechanism has 286 species and 914 reactions with a development history independent of MOZART-TS1. It features recent major updates to NO x heterogeneous and cloud chemistry (Holmes et al., 2019), isoprene chemistry (Bates and Jacob, 2019), aromatic chemistry (Bates et al., 2021), and Cl-Br-I tropospheric halogen chemistry (Wang et al., 2021). Photolysis is calculated using the Fast-JX model (Bian and Prather, 2002) with consistent aerosol and overhead column ozone information from the GEOS-Chem simulation (Eastham et al., 2014). No aerosol microphysics is included here so that aerosol concentrations are represented by the bulk masses of their chemical components (Park et al., 2004;Pai et al., 2020) but with four size bins for dust and two for sea salt aerosol (Alexander et al., 2005;Fairlie et al., 2010). Fritz et al. (2022) describe the original implementation of GEOS-Chem within CESM2. They developed an interface to pass input data to GEOS-Chem, run the GEOS-Chem chemical module, and export the updated chemical species concentrations. The interface converts between the bulk aerosols in GEOS-Chem and the modal aerosols in MAM4 for aerosolradiation and aerosol-cloud interactions. Coupling of the GEOS-Chem chemical module to CESM2 required the adaptation of several components for compatibility with CESM2 or consistency with CAM-chem. We summarize in Table 1 the important differences between the atmospheric chemistry representations in CAM-chem, GEOS-Chem within CESM2, and the offline GEOS-Chem CTM.
Here we make several improvements and corrections to the original implementation of GEOS-Chem within CESM2 by Fritz et al. (2022). We simulate nucleation in MAM4 by passing the gas-phase H 2 SO 4 production rate computed a With 50 h relaxation time, nudging U, V, and T. b GEOS-Chem bulk aerosols masses are mapped to MAM4 modes for aerosol-radiation and aerosol-cloud interaction effects within CESM2. See Fritz et al. (2022) for the species mapping between GEOS-Chem species to MAM4 aerosols. c Sectional aerosol microphysics are available in GEOS-Chem (Yu and Luo, 2009;Kodros and Pierce, 2017) but are not used here. d SOA denotes secondary organic aerosol; VBS denotes volatility basis set. GEOS-Chem here uses the complex SOA option from Pye et al. (2010). in GEOS-Chem from the SO 2 + OH reaction. We add an aerosol sink in the upper troposphere and lower stratosphere following Hodzic et al. (2015Hodzic et al. ( , 2016) ) to compensate for CESM2's omission of coupling convective transport and scavenging. We correct the sea surface temperatures passed to HEMCO, which results in inorganic iodine emissions being 1 % of the previously incorrectly calculated value. We also add numerous GEOS-Chem diagnostics for analyzing model output, including individual reaction rates and total production and loss rates for individual species.
Figure 1 shows the mean photolysis frequencies (J values) for NO 2 (J NO 2 ) and O 3 to O( 1 D) (J O1D ) simulated by the GEOS-Chem model (with Fast-JX) and the difference with CAM-chem (with TUV lookup table) in surface air in July. Photolysis rates in GEOS-Chem with Fast-JX are generally lower than in CAM-chem with TUV. Differences for J NO 2 are typically 0 %-10 % over oceans and 10 %-20 % over land, while differences for J O1D are typically 10 %-20 % over oceans and 20 %-40 % over land. There are some larger differences in polluted and open-fire regions, such as in East Asia and Siberia, and at high latitudes.
Fast-JX and TUV use the same spectroscopic data from the NASA JPL recommendations (Burkholder et al., 2020). Fast-JX includes aerosol extinction, but TUV does not, which explains the larger differences over polluted and openfire regions. Differences over the oceans are mainly due to clouds. While Fast-JX and TUV both represent effects of cloud extinction, treatment of cloud scattering between the two schemes is different. The effects of aerosol-cloud interactions on cloud properties through MAM4 cause GEOS-Chem and CAM-chem to have different cloud optical depths that can lead to further differences. Cloud effects are particularly large at high latitudes because of extensive cloud cover and low sun angles. Sensitivity simulations for clear sky (no cloud or aerosol extinction input to the photolysis schemes) show smaller differences between Fast-JX and TUV, generally less than 10 % for J NO 2 and less than 20 % for J O1D , while sensitivity simulations with Fast-JX implemented in CAM-chem show less than 5 % differences for J NO 2 and J O1D everywhere compared to Fast-JX implemented in GEOS-Chem.
Figure 2 shows photolysis frequencies from the KORUS-AQ and ATom-1 campaigns derived from actinic flux measurements (Hall et al., 2018;Crawford et al., 2021), compared to the photolysis frequencies computed by Fast-JX and TUV sampled along the aircraft flight tracks. J NO 2 values agree within 10 %, and there is no systematic bias relative to observations. Fast-JX values tend to be higher than TUV at high altitudes, and this can be attributed to cloud effects as discussed above. J O1D values also show good agreement for ATom-1, but observed values for KORUS-AQ are much lower than for ATom-1 in the same season, which is captured by Fast-JX but not by TUV (which is 30 % too high). We find that the overestimate of J O1D by TUV during KORUS-AQ is due in part to not accounting for aerosol extinction. Comparison of clear-sky J values shows that there is some additional unidentified factor causing TUV to be too high during KORUS-AQ, and this disappears when Fast-JX is implemented in CAM-chem. In what follows the CAM-chem simulation uses the TUV lookup table, but we will comment as appropriate on the effect of switching to Fast-JX.
Table 2 shows global tropospheric ozone and OH budgets from GEOS-Chem and CAM-chem compared to the literature. Ozone budgets from the two models and the multimodel mean in Young et al. (2018) are within 10 % of each other. The larger chemical production and shorter chemical lifetime in GEOS-Chem are mainly due to photolysis of particulate nitrate (Shah et al., 2023), without which chemical production in GEOS-Chem decreases by 10 % to 4902 Tg a -1 , and the tropospheric ozone burden decreases by 5 % to 332 Tg. The lower dry deposition in GEOS-Chem reflects lower ozone deposition to the ocean (Pound et al., 2020). GEOS-Chem and CAM-chem have the same global OH concentrations, on the high end of the range of values from the ACCMIP and CCMI model ensembles (Naik et al., 2013;Zhao et al., 2019). The lifetime of methylchloroform against loss to tropospheric OH is 5.4 and 5.3 years, respectively, in GEOS-Chem and CAM-chem, 15 % lower than 6.3 ± 0.4 years inferred from observations (Prather et al., 2012).
Figure 3 shows the spatial distribution of annual mean OH concentrations simulated by GEOS-Chem and the difference with CAM-chem. Despite having the same global mean OH concentrations, the two models have large regional differences. GEOS-Chem is up to 30 % lower than CAM-chem over the continents, particularly over polluted regions, due to lower J (O 1 D) and possibly higher OH reactivity. Over the Amazon and Congo basins where NO x is low, isoprene does not titrate OH in GEOS-Chem due to recent updates in isoprene oxidation chemistry incorporating H-shift isomerization of isoprene-hydroxy-peroxy radicals to recycle OH, which sustains OH under low-NO conditions (Bates and Jacob, 2019).
Figure 4 shows annual mean surface and 500 hPa ozone and NO x concentrations simulated by GEOS-Chem and differences with CAM-chem. Ozone differences are generally smaller than 5 ppb, indicating a remarkable degree of agreement. The largest surface differences are at southern midlatitudes due to slower ozone deposition to the ocean in GEOS-Chem. At 500 hPa, GEOS-Chem has lower ozone at high latitudes due to tropospheric halogen chemistry. This chemistry increases ozone destruction through catalytic ozone loss cycles driven by iodine and bromine and decreases ozone production by conversion of NO x to halogen nitrates. Tropospheric halogen chemistry is not represented in the default configuration of CAM-chem. N 2 O 5 uptake in clouds, included in GEOS-Chem (Holmes et al., 2019) but not in CAM-chem, also contributes to the lower GEOS-Chem ozone at high northern latitudes. Particulate nitrate photolysis in GEOS-Chem corrects for a missing NO x source in the remote troposphere (Shah et al., 2023) and accounts for the higher NO x and ozone than CAM-chem over the oceans.
Table 3 shows global budget terms in the troposphere from sensitivity simulations varying the most important differences between GEOS-Chem and CAM-chem. The largest controlling factors for tropospheric ozone differences between GEOS-Chem and CAM-chem are nitrate photolysis and tropospheric halogen chemistry, which increase and decrease the tropospheric ozone burden by 5 % and 4 %, respectively. The global tropospheric NO x burden is 4 % lower in GEOS-Chem than CAM-chem because of conversion to halogen nitrates and use of Fast-JX offsetting the effect of ni- trate photolysis. Using Fast-JX for photolysis in CAM-chem results in a 7 % decrease in tropospheric NO x , which we attribute to lower J NO 2 in surface air over continents (Fig. 1). Table 2 includes a residual term in the tropospheric ozone budget as a balance between the chemical production, chemical loss, deposition, and (negligible) accumulation terms. This residual term of 341-380 Tg a -1 is expected to represent stratosphere-troposphere exchange (STE), which is not explicitly diagnosed in CESM2, and falls within the range of literature values listed in Table 2. The residual changes slightly in the sensitivity simulations of Table 3 in a way that is consistent with the tropospheric ozone burden, as increasing tropospheric ozone decreases STE while increasing deposition.
Fritz et al. ( 2022) previously found tropospheric ozone in GEOS-Chem to be 30 % lower than CAM-chem in the extratropics because of halogen chemistry, but iodine emissions in that simulation were 100-fold too high because the interface to HEMCO erroneously passed 2 m temperature instead of sea surface temperature to the iodine emissions module (Sect. 2.2). With corrected iodine emissions, we find only a 4 % decrease of tropospheric ozone in GEOS-Chem due to tropospheric halogen chemistry. The magnitude of the effect of halogen chemistry on ozone is uncertain, ranging from 10 % to 19 % in previous implementations in offline GEOS-Chem (Sherwen et al., 2016b;Wang et al., 2021) and CAMchem (Saiz-Lopez et al., 2012), but all models report lowertropospheric ozone as a result. We find that halogen chemistry has a smaller effect on tropospheric ozone in GEOS-Chem within CESM2 than offline due to weaker wind speeds and lower sea surface temperatures in CESM2, resulting in weaker sea salt and gaseous iodine emissions.
Figure 5 compares annual mean ozone vertical profiles simulated by GEOS-Chem and CAM-chem to ozonesonde observations for 2016 from the World Ozone and Ultraviolet Radiation Data Centre (WOUDC), averaged across nine regions following Tilmes et al. (2012). Figure 6 compares April 2016 monthly mean surface ozone simulated by GEOS-Chem and CAM-chem to background surface ozone observations from 10 sites of the NOAA ESRL Global Monitoring Division (McClure-Begley et al., 2013) and 5 remote sites in China under the World Meteorological Organization Global Atmosphere Watch Programme. Both models match the observations well and are within 5-10 ppb of each other. GEOS-Chem has lower ozone at high northern latitudes (up to 10 ppb at the surface) because of halogen chemistry.
Figure 7 shows tropospheric profiles of OH, NO, and CO simulated by GEOS-Chem and CAM-chem over the oceans in comparison to observations from the ATom-1 campaign. Model profiles compared to aircraft observations are computed at model runtime by sampling the two closest time steps and four closest grid boxes to the time-varying flight track data and then interpolated to the aircraft time and location. The 0.9°× 1.25°resolution of the simulation is well adapted to the scales sampled by ATom. Both models generally agree with OH observations within uncertainty. Both models fit NO observations within a factor of 2 in the Northern Hemisphere but have large underestimates in the Southern Hemisphere. This underestimate is a known model issue in previous offline GEOS-Chem simulations (Travis et al., 2020) and is not correctable by nitrate photolysis because particulate nitrate concentrations in the Southern Hemisphere are low (Shah et al., 2023). Observations show a NO increase in the upper troposphere of the Southern Hemisphere that is captured by CAM-chem but not GEOS-Chem. Previous work has shown that offline GEOS-Chem simulations capture this increase of NO in the upper troposphere (Shah et al., 2023). A sensitivity GEOS-Chem simulation without tropospheric halogen chemistry, as shown in Fig. 6, also captures this increase. The difference is due to CESM2 not accounting for the scavenging in convective updrafts of soluble halogen gases such as HBr and HOBr. This increases the formation of stable halogen nitrates in the upper troposphere where thermolysis and hydrolysis are slow (Wang et al., 2021).
Both models underestimate CO in the Northern Hemisphere, which is a known issue attributed to excessive OH (Gaubert et al., 2020) or missing emissions of CO and its precursors (Park et al., 2021;Tang et al., 2023). CAM-chem has 10-20 ppb lower CO globally compared to GEOS-Chem that is likely driven by differences in OH. In the Southern Hemisphere the difference is driven by improvements in isoprene oxidation in GEOS-Chem by Bates and Jacob (2019), which recycles OH through H-shift isomerization of isoprene-hydroxy-peroxy radicals under low-NO conditions, seen in observations by Wells et al. (2020). This leads to faster in situ isoprene oxidation and a higher CO yield. This is not included in CAM-chem's default MOZART-TS1 mechanism used in this work but is included in the updated MOZART-TS2 mechanism (Schwantes et al., 2020(Schwantes et al., , 2022)).
We use comparison to observations from the KORUS-AQ campaign (1 May to 10 June 2016) over the Seoul Metropolitan Area (SMA; 37-37.6°N, 126.6-127.7°E) as illustrative of a polluted atmosphere. Figure 8 shows median concentration profiles of oxidants and related species. Observations are compared to GEOS-Chem and CAM-chem sampled along the flight tracks and to GEOS-Chem sensitivity simulations without particulate nitrate photolysis and without the nitrate correction applied in CESM2 for lack of scavenging in convective updrafts. Also shown in the figure are vertical profiles from an offline nested GEOS-Chem simulation at 0.25°× 0.3125°resolution reported by Yang et al. 2021) intercomparison, and this was attributed to its stratospheric ozone influx, but here GEOS-Chem uses the same dynamics and hence the same stratospheric influx. The success of CAM-chem in KORUS-AQ reflects instead its nonaccounting of tropospheric halogen chemistry as a sink of ozone, which in GEOS-Chem needs to be compensated by particulate nitrate photolysis. Both models are too high compared to observations above 4 km altitude, which is due at least in GEOS-Chem to excessive particulate nitrate resulting from inadequate convective scavenging. Particulate nitrate photolysis increases free-tropospheric NO x and ozone production, but this depends on the nitrate concentration. CAM-chem does not simulate nitrate. Because GEOS-Chem nitrate is not removed in convective updrafts in the CESM2 environment, our standard implementation within CESM2 corrects nitrate using the same photolytic sink that CAM-chem applies for SOA with a rate of 0.0004 × J NO 2 and no products (Hodzic et al., 2015(Hodzic et al., , 2016) ) to avoid buildup in the upper troposphere. But that correction is apparently insufficient because particulate nitrate is overestimated relative to observations above 4 km altitude, an overestimate not seen in the offline GEOS-Chem simulation of Yang et al. (2023) and which would be worse if we did not apply the correction (Fig. 8b and c). A solution would be to replace CESM convective transport with the GEOS-Chem offline convective transport and scavenging module using archived CESM convective mass fluxes, and this has been done before when coupling GEOS-Chem to the GEOS and Beijing Climate Center (BCC) ESMs, which had the same problem of not scavenging water-soluble species in convective updrafts (Yu et al., 2018;Lu et al., 2020). A more comprehensive solution would be to include scavenging of watersoluble species in the CESM2 convection scheme. This is implemented for MAM aerosols (Wang et al., 2013) but not for gas-phase species or aerosols only represented in GEOS-Chem, including nitrate.
The simulations of particulate nitrate and peroxyacetyl nitrate (PAN) within CESM show a sharp drop of concentrations with altitude above the surface, whereas the observations and the offline GEOS-Chem simulation of Yang et al. (2023) show a mixed-layer structure extending to 1-2 km altitude. This likely reflects a bias in the CESM2 boundary layer mixing scheme that would need to be investigated further. Boundary layer mixing in the offline GEOS-Chem model is a standard non-local scheme from Lin and McElroy (2010). The PAN simulations in GEOS-Chem and CAMchem otherwise agree closely, indicating similar production from VOC chemistry, and are lower than the offline GEOS-Chem simulation, which includes additional emissions of volatile chemical products (VCPs) as a source of acetaldehyde leading to PAN production (Yang et al., 2023).
GEOS-Chem has been implemented as an atmospheric chemistry module in the NCAR Community Earth System Model (CESM2) to serve as alternative to CAM-chem and contribute to the MUSICA vision of plug-and-play modularization of atmospheric chemistry within CESM (Pfister et al., 2020). Here we presented an intercomparison and evaluation with observations of tropospheric oxidant simulations with these two modules. The intercomparison covered the full year of 2016, allowing evaluation with the ATom-1 aircraft campaign over the remote Pacific and Atlantic and the KORUS-AQ aircraft campaign over the Seoul Metropolitan Area (SMA). Both GEOS-Chem and CAM-chem used the same emissions processed through HEMCO (Lin et al., 2021) and the same coupling to other CESM2 modules. GEOS-Chem uses the Fast-JX scheme for photolysis, while CAM-chem uses a lookup table based on TUV. Both schemes agree to within 10 % when compared to J NO 2 and J O1D photolysis frequencies observations in ATom-1, but observations in KORUS-AQ show that CAM-chem overestimates J O1D , while GEOS-Chem does not. One major difference is that TUV does not account for extinction by aerosols, while Fast-JX does. We implemented Fast-JX in CAM-chem and find that it resolves most of the photolysis differences with GEOS-Chem.
Global tropospheric ozone budget terms in GEOS-Chem and CAM-chem agree within 10 %, compared to a much wider spread in the literature and due in part to canceling effects. Differences between the two models are mostly driven by aerosol nitrate photolysis, N 2 O 5 uptake in clouds, and tropospheric halogen chemistry, all of which are included in GEOS-Chem but not in CAM-chem. Aerosol nitrate photolysis in GEOS-Chem produces NO x and enhances ozone production, compensating for losses from N 2 O 5 uptake in clouds and tropospheric halogen chemistry. Annual mean ozone concentrations agree within 5 ppb between GEOS-Chem and CAM-chem almost everywhere. Lower ozone deposition to the oceans in GEOS-Chem results in higher surface ozone at southern midlatitudes. Tropospheric halogen chemistry results in lower ozone at high northern latitudes. Tropospheric NO x in GEOS-Chem is higher than CAM-chem in the tropics due to nitrate photolysis and lower at high latitudes due to N 2 O 5 uptake by cloud and formation of halogen nitrates. The global mean tropospheric OH concentration is identical between the two models, but there are large differences over the continents driven by photolysis and by isoprene chemistry.
Both GEOS-Chem and CAM-chem show good agreement with annual mean ozonesonde observations and background surface ozone observations over the range of latitudes. Comparison to ATom-1 observations in July-August 2016 shows good agreement for OH concentrations in both the Northern Hemisphere and Southern Hemisphere (NH and SH) within the measurement accuracy and for NO x in the NH, but NO x in the SH is underestimated. GEOS-Chem shows a depletion of NO x in the SH upper troposphere that is due to formation of halogen nitrates and is not seen in the observations. However, the offline GEOS-Chem simulation does not show this problem. One issue in CESM2 is the lack of scavenging of water-soluble species, including halogen radical reservoirs in convective updrafts. Both GEOS-Chem and CAM-chem underestimate CO in the NH, but CAM-chem is consistently lower than GEOS-Chem due to higher OH in the NH and suppression of CO production from isoprene oxidation in the SH.
Comparison with KORUS-AQ aircraft observations allowed model evaluation for polluted conditions. Ozone concentrations in GEOS-Chem and CAM-chem are higher than observed above 4 km altitude, which in GEOS-Chem is due to excessive particulate nitrate photolyzing to produce excessive NO x . Lack of scavenging of water-soluble species in convective updrafts is a major shortcoming in CESM2 that hinders proper representation of nitrate in the upper troposphere. Simulation of peroxyacetyl nitrate (PAN) in KORUS-AQ shows good agreement between GEOS-Chem and CAMchem and with observations, indicating consistency in the VOC chemistry producing PAN. However, the decreases of PAN and particulate nitrate mixing ratios with altitude in the lower 2 km are much sharper than observed or simulated by the offline GEOS-Chem model, implying insufficient boundary layer mixing in CESM2.
Overall, we have shown that GEOS-Chem provides a high-quality simulation of tropospheric oxidant chemistry in CESM2 and can contribute modules for alternative representations of atmospheric chemistry to serve the MUSICA vision.
Code availability. A fork of an alpha version (cam6_3_095) of the Community Atmosphere Model (CAM) including GEOS-Chem is available at https://github.com/geoschem/CAM (Fritz et al., 2023) and is used in this work. CAM-chem using HEMCO for emissions is implemented in the mainline CAM code as of cam6_3_118 (https://github.com/ESCOMP/CAM/tree/cam6_3_ 118, The Community Atmosphere Model Developers, 2023a). GEOS-Chem within CESM2 is implemented in the mainline CAM code as of cam6_3_147 (https://github.com/escomp/cam/ tree/cam6_3_147, The Community Atmosphere Model Developers, 2023b). Author contributions. LKE, SDE, and DJJ conceived of the project, acquired funding, and supervised the work. HL, EWL, XF, SDE, and TMF performed software development. HL, LKE, LHY, XF, RD, SZ, YT, MMK, NKC, and DJJ analyzed the model results. SZ prepared the KORUSv5 emission inventory for input into HEMCO and CESM2. HL performed the visualization and preparation of the original draft. All authors contributed to review and editing of the manuscript. Competing interests. The contact author has declared that none of the authors has any competing interests. Disclaimer. Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors.
Atmos. Chem. Phys., 24, 8607-8624, 2024 https://doi.org/10.5194/acp-24-8607-2024
https://doi.org/10.5194/acp-24-8607-2024 Atmos. Chem. Phys., 24, 8607-8624, 2024