Improved Atmospheric Composition Forecasting In GEOS-CF Version 2

Improved Atmospheric Composition Forecasting In GEOS-CF Version 2

Authors: Viral Shah, Pam Wales, and Carl Malings

Editors: Bennett Erdman

Published May 12, 2026

The GMAO recently released a major update of its Composition Forecast system (GEOS-CF, version 2), substantially improving its global atmospheric composition forecasting capabilities, and providing users and with better predictions of key atmospheric constituents. Here we give an overview of the capabilities and performance of the new version.

The GEOS Composition Forecasting (GEOS-CF; Keller et al., 2021; Knowland et al., 2022) system provides daily estimates and 5-day forecasts of global atmospheric composition at 25-km horizontal resolution, extending from Earth's surface to approximately 80 km in altitude. The system embeds the GEOS-Chem model within GMAO’s GEOS model, a capability built on the long-standing collaboration between GMAO and the GEOS-Chem community.

GEOS-CF data support a variety of research and applications, including NASA satellite missions, aircraft campaigns and instrument teams, local and regional air quality forecasters, and environmental health practitioners (Malings et al., 2026). Consistent with NASA's open science principles, the system uses open-source software and provides open access to data products through multiple channels, including direct file access, interactive visualization, an Application Programming Interface (API), and the Google Earth Engine; see GEOS-CF Data Access for details. GEOS-CF also contributes to NASA’s Earth Science to Action mission to address societal challenges by supporting air quality monitoring and management in many areas of the world.


System Updates

GEOS-CF v2 includes improved atmospheric physics and chemistry, more recent emissions data, direct assimilation of satellite ozone data, and the addition of carbon dioxide (CO2) and methane (CH4) forecasts. These changes are summarized in Table 1 and described in the updated GEOS-CF File Specifications. The File Specifications also lists changes in the file naming convections and archived fields between the two versions. The CO2 and CH4 output are under evaluation prior to public release.


GEOS-CF v1GEOS-CF v2
GEOS AGCM VersionIcarus-1_0v10.23.0 (Jason-23_0)
Meteorological replayGEOS FP-ITGEOS IT
Composition data assimilationNo assimilation (stratospheric ozone nudged to GEOS-FP)Assimilation of ozone satellite data (MLS, OMI)
GEOS-Chem version12.0.114.0.0
Anthropogenic emissionsHTAP (base year 2010)CEDS (base year 2019)
Greenhouse gas forecasts-CO2 and CH4 (from GOCART, to be released soon)

Evaluation of Tropospheric Composition

The updates introduced in GEOS-CF v2 have led to substantial improvements in the simulated atmospheric composition. We highlight key results, focusing on tropospheric ozone and PM2.5, two species of prime importance for tropospheric composition and air quality.


Figure 1: Tropospheric ozone distribution from ozonesondes and GEOS-CF. The left panels show annual mean ozone columns between the surface and 300 hPa from the observations and GEOS-CF v2. The values are shown in Dobson Units (DU; 1 DU ≡ 2.69×1016 molecules cm-2). The right panels show the ozone profiles for the northern extratropics and from the observations and from GEOS-CF v2, GEOS-CF v1, and a GEOS-CF v2 run without ozone data assimilation. The ozonesonde observations are from the SHADOZ, WOUDC, and NOAA networks.


Figure 1 compares the tropospheric ozone distribution from GEOS-CF with balloon-borne ozonesonde observations for 2023. GEOS-CF v2 matches the observed ozone columns, integrated between the surface and 300 hPa, with little bias and high spatial correlation. The figure also shows ozone profiles over the northern extratropics, where GEOS-CF v1 underestimated the ozone mixing ratios in the free troposphere. The bias with respect to the observations is reduced in GEOS-CF v2. The improvement in free tropospheric ozone is due to assimilation of satellite ozone data, since a GEOS-CF v2 simulation without ozone data assimilation still underestimates observed mixing ratios. The GEOS-CF ozone profiles provide prior information in retrieving tropospheric ozone profiles from TEMPO data; thus, these improvements are expected to contribute to improved accuracy of TEMPO retrievals.

Figure 2: 95th percentile maximum daily 8-hour average (MDA8) ozone over the US in 2023. The maps show the observed values from the EPA Air Quality Service (AQS) monitors, and the values from GEOS-CF v2. The right panel shows the probability distribution of the annual 95th percentile MDA8 ozone across sites as observed and as simulated by GEOS-CF v2 and v1.

Figure 2 compares the distribution of surface ozone over the US from GEOS-CF and EPA air quality observations. At each site, it shows the annual 95th percentile of the maximum daily 8-hour average (MDA8), a metric closely tied to ozone air quality impacts and policies. Peak MDA8 ozone values typically occur in summer afternoons, during sunny, stagnant conditions that favor photochemical production of ozone. GEOS-CF v2 reproduces the observed spatial patterns, with higher ozone values on the east coast, the Midwest, and California, but the values are uniformly biased high by about 10 ppbv, likely reflecting uncertainties in boundary layer mixing and dry deposition in the model. Still, GEOS-CF v2 shows a marked improvement in ozone compared to GEOS-CF v1, as seen in the probability distribution of MDA8 ozone values. In GEOS-CF v1, MDA8 ozone values exceeded 80 ppbv at 25% of the sites, but such values are hardly present in the observations and GEOS-CF v2. The improvement in GEOS-CF v2 surface ozone is largely due to the update to anthropogenic emissions of nitrogen oxides (NOx) and volatile organic compounds (VOC).


Figure 3: Daily mean PM2.5 concentrations over the US in 2023. The maps show the observed values from the EPA Air Quality Service (AQS) monitors, and those from GEOS-CF v2. The right panel shows the monthly variation in average PM2.5 for all sites from the observations, GEOS-CF v2, and GEOS-CF v1.

GEOS-CF v2 also shows significant skill in simulating surface PM2.5 concentrations. Figure 3 shows the distribution and the seasonal variation of observed and GEOS-CF PM2.5 concentrations. GEOS-CF v2 shows similar spatial patterns of PM2.5 as observed, with a mean bias of 15%. The bias is larger in the winter than in summer, likely because of a wintertime overestimate in nitrate and primary organic aerosols. However, the bias is reduced in GEOS-CF v2 compared to GEOS-CF v1, where the previous version overestimated PM2.5 concentrations by a factor of 2. Several features in the new version contribute to this improvement, including updated emission inventories; improved chemistry in GEOS-Chem, particularly in the formation of sulfate and nitrate aerosols; and changes to convective transport and scavenging in GEOS. 

In summary, GEOS-CF v2 improves global atmospheric composition forecasting through updates to model physics, chemistry, emissions, and data assimilation, leading to more accurate simulations of tropospheric ozone and PM2.5. These improvements will benefit scientific research, satellite retrievals, mission planning, and air quality focused applications using data from GEOS-CF.

Video: Comparison of hourly tropospheric NO2 columns over North America on December 31, 2025, from GEOS-CF versions 1 (top) and 2 (bottom). 


References
Keller, C. A., Knowland, K. E., Duncan, B. N., Liu, J., Anderson, D. C., Das, S., Lucchesi, R. A., Lundgren, E. W., Nicely, J. M., Nielsen, E., Ott, L. E., Saunders, E., Strode, S. A., Wales, P. A., Jacob, D. J., and Pawson, S.: Description of the NASA GEOS Composition Forecast Modeling System GEOS‐CF v1.0, J. Adv. Model. Earth Syst. https://doi.org/10.1029/2020MS002413, 2021.

Knowland, K. E., Keller, C. A., Wales, P. A., Wargan, K., Coy, L., Johnson, M. S., et al. NASA GEOS Composition Forecast Modeling System GEOS‐CF v1.0: Stratospheric composition. J. Adv. Model. Earth Syst. https://doi.org/10.1029/2021MS002852, 2022.

Malings, C. A., Shah, V, Wales, P. A., Knowland, K. E., Keller, C. A., Wayman, C., Ardizzone, J., Ott, L.E., Pawson, S., Jenness, A., Rau, A., Mao, J., Dong, Z., Johnson, M.S., Roots, M., Mandarino, F., Lazrak, N., Cardenas, B, and Mwaniki, G.: A Survey of Applications of the NASA GEOS-CF Global Atmospheric Composition Forecasts: case studies for NASA open data and Earth science to action. Bull. Am. Meteorol. Soc. https://doi.org/10.1175/BAMS-D-25-0131.1, 2026.