We give here a narrative description of the current standard version of the GEOS-Chem model, with two purposes:

We strongly encourage you to be generous in citations—this not only recognizes the developer's work but also increases the traceability of your paper. Offering co-authorship to developers is encouraged for new developments flagged in this narrative if they are important for your work. It may also be appropriate to offer co-authorship for older model developments if they were new when you started your work. See the New GEOS-Chem Developments page for more specific information on the developer(s) to be credited, and contact the Model Scientist or appropriate Working Group chair if you need guidance.

The narrative below is reviewed and updated by the GEOS-Chem Steering Committee at every new X.Y model version release.

Citing GEOS-Chem

GEOS-Chem should be referenced by its version number X.Y.Z and corresponding DOI. See the history of model versions and their DOIs. The website http://www.geos-chem.org is also a useful reference. We strongly encourage you to cite GEOS-Chem journal publications, both for your general use of GEOS-Chem and for your specific applications. Consult the narrative below for referencing specific components. For questions on citations please contact the relevant Working Group Chair or Model Scientist.

Model name

The name "GEOS-Chem" was coined in 2001 and is first referred to in Bey et al. [2001]. It is not an acronym - there is nothing to spell out. GEOS stands for Goddard Earth Observing System and Chem stands for Chemistry but calling it the "Goddard Earth Observing System - Chemistry" model would be inappropriate because the GEOS Earth System Model can use other chemical modules besides GEOS-Chem, and GEOS-Chem can use other meteorological drivers besides GEOS.

If an abbreviated name for GEOS-Chem needs to be used, such as in a Figure or other context where space is limited, then 'GC' is acceptable and in fact frequently used for informal communication within the GEOS-Chem community. No other abbreviation is acceptable. In particular, 'GEOS' should not be used because of confusion with the GEOS Earth System Model.

Original/historical references

Bey et al. [2001] is the first reference to GEOS-Chem that includes a detailed model description. It is suitable and widely used as an original reference for the model. It only describes a model for gas-phase tropospheric oxidant chemistry. Other original references are:


GEOS-Chem is a grid-independent model. It operates on 1-D columns with default or user-specified horizontal gridpoints, vertical gridpoints, and timesteps. The GEOS-Chem chemical module updates column concentrations for the effects of emissions, chemistry, aerosol microphysics, and deposition at each time step. This chemical module can be implemented in three different configurations:

Meteorological fields and grid resolution

GEOS-Chem in off-line mode (Classic or GCHP) is driven by assimilated meteorological data from the Goddard Earth Observation System (GEOS) of the NASA Global Modeling and Assimilation Office (GMAO). The two GEOS data archives used by GEOS-Chem are:

These archives have 3-hour temporal resolution for 3-D fields and 1-hour resolution for 2-D fields. GEOS-Chem simulations can be conducted at the native spatial resolution of the GEOS fields or at coarser resolutions. GEOS-Chem Classic simulations can also be conducted in nested mode (see Nesting below), while GCHP simulations can be conducted in stretched-grid mode [Bindle et al., 2021]. The default timesteps are optimized to balance accuracy and speed as described by Philip et al. [2016].

GEOS-Chem can also use off-line meteorological fields from the GISS GCM for future climates and paleoclimates. These implementations are referred to as the GCAP and ICECAP models. GCAP 2.0 is described by Murray et al. [2021] and is a new development in version 13.1.0

The GEOS-Chem chemical module can be used in on-line applications on any grid of the parent meteorological model:


The nested capability for GEOS-Chem was originally developed by Y. X. Wang et al. [2004]. It allows simulations at the native-grid horizontal resolution of the GEOS data over a user-selected regional domain with dynamic boundary conditions from a coarser global simulation. The nesting can either be 1-way, with no influence from the nested domain on the global domain, or 2-way where the two domains interact with each other. The 2-way nesting capability with multiple nests is described by Yan et al. [2014] and on this wiki page.

The current nested version of GEOS-Chem Classic uses GEOS-FP data with 0.25° x 0.3125° resolution or MERRA-2 data with 0.5° x 0.625° resolution within the nested domain. The capability to operate at 0.25° x 0.3125°resolution with full aerosol-oxidant chemistry was originally developed by Zhang et al. [2015] for East Asia and Kim et al. [2015] for North America. FlexGrid allows users to define any nested domain at runtime, with no pre-processing of meteorological or other data files, requiring only the generation of boundary condition files at the global model resolution (new development in version 12.4.0).

GCHP cannot use nested grids but can instead use the stretched-grid confguration of Bindle et al. [2021] to provide high resolution over regions of interest This is a new development in version 13.0.0.

Transport and deposition

GEOS-Chem Classic uses the TPCORE advection algorithm of Lin and Rood [1996] on the latitude-longitude grid of the archived GEOS meteorological data. GCHP uses the FV3 advection algorithm of Putnam and Lin [2007] on a cubed sphere grid after remapping the archived GEOS meteorological data on that grid. Convective transport in GEOS-Chem is computed from the convective mass fluxes in the meteorological archive as described by Wu et al. [2007]. Boundary layer mixing in GEOS-Chem uses either the non-local scheme implemented by Lin and McElroy [2010] or full mixing up to the GEOS-diagnosed mixing depth.

The wet deposition scheme in GEOS-Chem is described by Liu et al. [2001] for water-soluble aerosols and by Amos et al. [2012] for gases. Henry's law constants are from the compilation by Sander [2015] including for water-soluble organics [Safieddine and Heald, 2017]. Scavenging of aerosol by snow and cold/mixed precipitation is described by Q. Wang et al. [2011, 2014]. Faster scavenging as described by Luo et al. [2019] is an option in the model (new development in version 12.7.0)

Dry deposition is based on the resistance-in-series scheme of Wesely [1989] as implemented by Y. Wang et al. [1998a]. Aerosol deposition is from Zhang et al. [2001]. Aerosol deposition to snow/ice is described by Fisher et al. [2011]. Gravitational settling is from Fairlie et al. [2007] for dust and Alexander et al. [2005] for coarse sea salt. Sea-salt deposition is from Jaegle et al. [2011]. Cold-temperature HNO3 deposition is from Jaegle et al. [2018] (new development in version 12.6.0). There is an option for dependence of stomatal conductance on CO2 levels [Franks et al., 2013] and this is a new development in version 12.6.0. Ozone deposition to the ocean is from Pound et al. [2020] and is a new development in version 12.8.0.

See the mercury section for description of air-sea-land exchange of mercury.

Radiation and photolysis

GEOS-Chem can calculate the radiative forcing from changes in atmospheric composition using the optional RRTMG module. Implementation of RRTMG in GEOS-Chem is described in Heald et al. [2014].

Photolysis frequencies for stratospheric and tropospheric chemistry are calculated with the Fast-JX code of Bian and Prather [2002] as implemented in GEOS-Chem by Mao et al. [2010] for the troposphere and Eastham et al. [2014] for the stratosphere.

The effect of aerosols on photolysis rates is described by Latimer and Martin [2019] (new development in version 12.6.0). There is an option to add absorption of UV by brown carbon [Hammer et al., 2016]. There is an option to add aerosol nitrate photolysis following Kasibhatla et al. [2018] and this is a new development in version 12.6.0.


All GEOS-Chem emissions are configured at run-time using the HEMCO 3.0 facility described by Lin et al. [2021]; this is a new development in version 13.1.0. HEMCO allows users to mix and match inventories from the GEOS-Chem library or add their own, apply scaling factors, overlay and mask inventories, etc. without having to edit or compile the code. HEMCO also has extensions to compute emissions with meteorological dependencies and to process other input/output data in GEOS-Chem. Version 3.0 has a number of new features including greater modularity for adaptation to other models and an intermediate grid for more accurate regional masking of emissions.

Emissions of dust aerosol, lightning NOx, biogenic VOCs, soil NOx, and sea salt aerosol are dependent on the local meteorological conditions. These emissions are computed off-line at the native resolution of the GEOS meteorological data and then archived along with the GEOS data as input to GEOS-Chem. In that way, emissions in GEOS-Chem remain the same at any model resolution. Users can also choose to compute emissions on-line rather than using the off-line emission files. Off-line biogenic VOCs, soil NOx and sea salt aerosol emissions are described in Weng et al. [2020]. The default capability for off-line emissions is a new development in version 12.4.0.

Anthropogenic. Global anthropogenic emissions up to 2017 are from the CEDS inventory of McDuffie et al. [2020] and this is a new development in version 13.0.0. EDGAR v4.3.2 [Crippa et al., 2018] with trash emissions from Wiedinmyer et al. [2014] is available as an alternative option to CEDS (trash emissions are already included in CEDS). Ethane emissions from Tzompa-Sosa et al. [2016] and propane emissions from Xiao et al. [2008] overwrite the corrsponding CEDS and EDGAR v4.3.2 emissions in the default model. Diurnal and weekend/weekday vatiations of emissions are from van Donkelaar et al. [2008]. Diurnal variation of Chinese power plant emissions is from X. Liu et al. [2019] and this is a new development in version 13.1. Vertical allocation of emissions by sector follows Hemispheric CMAQ [US EPA, 2019] and this is a new development in version 13.1.0.

Future anthropogenic emissions following the RCP scenarios have been implemented into GEOS-Chem by Holmes et al. [2013].

Aircraft. Aircraft emissions are from the AEIC inventory [Stettler et al., 2011].

Ships. Global shipping emissions are from CEDS. Shipping emissions of NOx are processed by the PARANOX module of Vinken et al. [2012] to account for ozone and HNO3 production in the plume. The PARANOX module was updated by Holmes et al. [2014].

Open Fires. Emissions from open fires for individual years are from the GFED4.1s inventory with options to use instead the FINNv1.5 inventory [Wiedinmyer et al., 2011], the QFED inventory, or the GFAS inventory. BB4CMIP historical fire emissions for 1750-2014 are from van Merle et al. [2017] and this is a new development in version 12.6.0.

Lightning. Lightning NOx emissions are as described by Murray et al. [2012] to match OTD/LIS climatological observations of lightning flashes. The climatology has been updated to 2019 and this is a new development in version 12.9.0.

Biogenic VOCs. Biogenic VOC emissions in GEOS-Chem are from the MEGAN v2.1 inventory of Guenther et al. [2012] as implemented by Hu et al. [2015b]. Leaf area indices (LAIs) used in MEGAN v2.1 are from the Yuan et al. [2011] MODIS product for 2005-2016. Dependence on CO2 was added by Tai et al. [2013]. Acetaldehyde emissions are from Millet et al. (2010). Biogenic non-agricultural ammonia sources are from GEIA.

Soils. Biogenic soil NOx emissions are from Hudman et al. [2012].

Ocean. Marine emissions of DMS are from the Lana et al. [2011] dataset as implemented in GEOS-Chem by Breider et al. [2017]. Air-sea exchange of acetone assumes fixed ocean concentrations as described by Fischer et al. [2012]. Ocean acetaldehyde emissions are from Millet et al. (2010). Ammonia emissions from Arctic seabirds are from Croft et al. [2016]. Ocean ammonia emissions are from GEIA [Bouwman et al., 1997].

Volcanoes. Eruptive and non-eruptive volcanic SO2 emissions for individual years are from the AEROCOM data base. Update to 2019 is a new development in version 12.5.0.

Other. See the carbon gases section for GEOS-Chem references on emissions of CO2 and methane. See the aerosols section for GEOS-Chem references on primary aerosol emissions. See the mercury section for GEOS-Chem references on emissions of mercury. See the POPs section for GEOS-Chem references on emissions of persistent organic pollutants (POPs).


GEOS-Chem simulates detailed oxidant-aerosol chemistry in the troposphere and stratosphere. The chemical solver is KPP [Damian et al., 2002] as implemented in GEOS-Chem with the FlexChem interface.

Chemical kinetics

Chemical mechanism kinetics generally follow JPL/IUPAC recommendations but go beyond the recommendations for specific aspects of the mechanism including for:

See the radiation section for the calculation of photolysis frequencies. Methane is prescribed as a surface boundary condition from monthly mean maps of spatially-interpolated NOAA flask data, and subsequently allowed to advect and react [Murray et al., 2016]. Water is specified from the driving meteorological fields in the troposphere but is transported as a reactive tracer in the stratosphere.

Reactive uptake of NO2, NO3, and N2O5 by aerosols is as described by Holmes et al. [2019], with reactive uptake coefficients for N2O5 on sulfate-nitrate-ammonium-organic aerosol from McDuffie et al. [2018ab]. This is a new development in version 12.6.0. HO2 uptake is from Mao et al. [2013] with a reactive uptake coefficient of 0.2 for conversion to H2O. Acid uptake by dust particles from Fairlie et al. [2010] is provided as an option. Aerosol hygroscopicity for calculating surface areas is from Latimer and Martin [2019] and this is a new development in version 12.6.0. Cloudwater pH is calculated following Shah et al. [2020] and this is a new development in version 12.9.0.

Reactive uptake of nitrogen oxides by clouds accounts for entrainment in the subgrid cloudy fraction of gridboxes and this is a new development in version 12.6.0. The same treatment is also applied for halogen reactive uptake by clouds starting with version 12.9.0.

GEOS-Chem simulations can be configured to have full chemistry only in the troposphere (“troposphere-only simulation”) with simple linear representation of stratospheric chemistry following the Linoz algorithm of McLinden et al. [2000] for ozone and 2-D (latitude-altitude) monthly mean sources and loss rate constants for other gases [Murray et al., 2012].

Aerosol processes

Sulfate-nitrate-ammonium aerosol. The original SNA aerosol simulation in GEOS-Chem coupled to gas-phase chemistry was developed by Park et al. [2004]. SNA thermodynamics are computed with the ISORROPIA thermodynamic module [Fontoukis and Nenes, 2007], most recently updated to version 2.2.

Carbonaceous aerosol. Q. Wang et al. [2014] describes the current BC simulation in GEOS-Chem. Organic aerosol in the default model follows the simple, irreversible, direct yield scheme of Pai et al. [2020]. Complex SOA can be used as an option following the simplified Volatility Basis Set (VBS) scheme of Pye et al. [2010] and the aqueous-phase isoprene SOA scheme of Marais et al. [2016] coupled to the isoprene gas-phase chemistry mechanism.

Dust aerosol. The dust simulation in GEOS-Chem is described by Fairlie et al. [2007]. Dust size distributions are from Li Zhang et al. [2013]. Fine anthropogenic dust from combustion and industrial sources is from the AFCID inventory of Philip et al. [2017].

Sea salt. The sea salt aerosol simulation in GEOS-Chem is described by Jaegle et al. [2011].

Marine POA. There is an option to emit marine POA following Gantt et al. [2015].

Aerosol microphysics. Two alternate simulations of aerosol microphysics are implemented in GEOS-Chem: the TOMAS simulation [Kodros and Pierce, 2017] and the APM simulation [Yu and Luo, 2009]. TOMAS has new developments in version 12.3.0. APM has new developments in version 12.6.0

Aerosol optical depth. Aerosol optical depth affecting photolysis rates is calculated in GEOS-Chem using RH-dependent aerosol optical properties from Latimer and Martin [2019]. Dust optics are from Ridley et al. [2012]. These calculations can be performed at user-specified wavelengths from 230 nm to 56 um when using RRTMG (see the radiation section).

Aerosol-only simulation. In addition to the fully coupled gas-aerosol simulation described in the Tropospheric Chemistry section, there is an option to conduct aerosol-only simulations using fixed 3-D monthly oxidant concentrations (from a GEOS-Chem simulation of old vintage) and simple SOA. This is described by Leibensperger et al. [2012].

Carbon gases

CO2. The current form of the simulation is described by Nassar et al. [2010]. Anthropogenic emissions are from ODIAC2019 [Oda and Maksyutov, 2011; Oda et al., 2018] and this is a new development in version 13.0.0.

Methane. The current form of the simulation is described by Maasakkers et al. [2019]. Updated soil uptake from the MeMo model v1.0 [Murguia-Flores et al., 2018] is a new development in version 12.7.0. Updated emission from fuel exploitation is from Scarpelli et al. [2020a] and is a new development in version 13.0.0. Anthropogenic emissions from Mexico are from Scarpelli et al. [2020b] and this is a new development in version 13.1.

CO. Simulation of CO in GEOS-Chem can be conducted either as part of the standard full-chemistry simulation or as a separate tagged-tracer simulation that resolves CO sources from individual regions or processes, and uses archived OH fields from a full-chemistry simulation to compute the CO sink. The most recent version is described by Fisher et al. [2017].


The original GEOS-Chem coupled atmosphere-ocean simulation of mercury was described by Selin et al. [2007] for the atmosphere and by Strode et al. [2007] for the ocean. Extension to a coupled atmosphere-ocean-land model was described by Selin et al. [2008]. The current version of the atmospheric simulation is described by Horowitz et al. [2017], and the current version of the ocean simulation is described by Soerensen et al. [2010], with updated ocean rate coefficients from Song et al. [2015]. Treatment of Arctic sea ice and rivers is as described by Fisher et al. [2012, 2013]. Gas-aerosol partitioning of Hg(II) is from Amos et al. [2012].There is an option to couple GEOS-Chem with the terrestrial mercury module developed by Smith-Downey et al. [2010].

Anthropogenic emissions are from Y. Zhang et al. [2016]. Future SRES emission scenarios have been implemented by Corbitt et al. [2011]. Options are available to use anthropogenic emissions from Streets et al. [2019] or from EDGAR v4.2 [Muntean et al., 2018], and these are new developments in version 13.0.0.

Persistent Organic Pollutants (POPs)

The model includes a simulation of PAHs as described by Friedman et al. [2014].

Model diagnostics

The model offers detailed output diagnostics in NetCDF format including species concentrations, production and loss rates, family production and loss rates, emissions, deposition fluxes and velocities, budgets and fluxes, time series at fixed locations or along selected aircraft flight tracks and satellite orbits, etc. See the GEOS-Chem wiki diagnostics page for more information. The NOAA Obspack diagnostic is available for comparison of model output to compiled global suborbital observations of greenhouse gases.

Surface ozone and HNO3 concentrations can be diagnosed below the lowest model gridpoint to take into account aerodynamic resistance to deposition [Travis and Jacob, 2019] (new development in version 12.6.0).

PM2.5 mass concentration is diagnosed in the model as described here.

Model Adjoint

See the GEOS-Chem adjoint wiki page for description and references.