The Goddard Earth Observing System Composition Forecast (GEOS-CF) is a global atmospheric constituent forecasting system created and operated by the Global Modeling and Assimilation Office at the National Aeronautics and Space Administration (NASA) Goddard Space Flight Center. In alignment with the NASA Earth Science to Action Strategy, GEOS-CF forecasts support a variety of research and practical applications, including NASA satellite missions, aircraft campaigns and instrument teams, local and regional air quality forecasters, and public and occupational health experts investigating air pollutant exposure. In alignment with the NASA Open-Source Science Initiative, GEOS-CF supports several data access methods, thereby serving a range of users with varying technical capabilities. This paper surveys several case studies of applications making use of GEOS-CF, investigating how the data were accessed and examining the perceived benefits and limitations of GEOS-CF in each case. The global coverage and data availability of multiconstituent information at a relatively low data latency was broadly perceived as the strength of GEOS-CF, while cited weaknesses included the relatively coarse spatial resolution for local-scale applications, lack of outputs of interest to specific applications, and uncertain data quality for regions and outputs with limited validation observations. Our goal is to present these examples and lessons learned to the broader community both as illustrations of NASA's Earth Science to Action Strategy and Open-Source Science Initiative and to inform future developments of GEOS-CF and similar systems, with the aim of improving the free and open provision of actionable Earth science information for societal benefit.

Tropospheric Emissions: Monitoring of Pollution (TEMPO) is observing air quality andatmospheric composition over North America from a geostationary orbit since its operationsstarted in August 2023. TEMPO observes the continent every 40 to 60 minutes at a spatialresolution on the order of ~ 2 x 4.5 km 2 . Together with the Geostationary EnvironmentMonitoring Spectrometer (GEMS, launch 2020) monitoring Asia and the Sentinel-4/UVN(launch 2025) monitoring Europe, TEMPO is part of the current global constellation ofgeostationary sensors devoted to the observation of air quality. Like GEMS and Sentinel-4/UVN,TEMPO uses backscattered ultraviolet and visible solar radiation to retrieve atmosphericamounts of key trace gases and aerosols associated with air quality and atmospheric chemistry.Among the species retrieved from TEMPO observations of nitrogen dioxide and formaldehydeare important to understand emissions and atmospheric chemistry, including the formation anddestruction of tropospheric ozone.After multiple version updates over the first two years of the mission, the TEMPO Level 2 NO 2and HCHO products have undergone significant enhancements to improve the performance andaccuracy of the slant column retrievals, air mass factor calculations and post-processingcorrections including destriping for NO 2 and background for HCHO. We illustrate theperformance of both retrievals (version 3 & 4), evaluating their fitting uncertainty and showingcomparisons with independent correlative measurements and other satellite products showcasingsmall noise levels and remarkable accuracy with well quantified biases. We continue byillustrating the capacity of TEMPO products focusing on different case studies showingTEMPO’s high temporal and spatial resolution. We finalize discussing aspects of the retrievalsubject to improvement and our plans to address them.

High-resolution PM2.5 forecasts are increasingly produced with machine-learning models, yet practical guidance on how often these models should be retrained and validated remains limited. This study quantifies the impact of retraining frequency and bias correction on air-quality prediction skill across multiple cities with contrasting emission sources and meteorological regimes, using The Goddard Earth Observing System composition forecast (GEOS-CF) fused with in-situ observations. Site-specific models are trained on at least two years of hourly data, with bias correction and alternative update schedules (6-18-month baselines and 6-12-month retraining cycles) evaluated using RMSE, R2, and SHAP-based feature importance. Bias-corrected models consistently improves GEOS-CF forecasts by more than 107% in R2 and reduces RMSE by over 75%, with annual retraining providing the largest gains (13% increase in R2, 12% reduction in RMSE) relative to more frequent updates. SHAP analysis shows that the dominant predictors and their relative importance vary by city, with combinations of boundary-layer height, aerosol optical depth, humidity, wind, and nitrogen oxides driving PM2.5 levels, demonstrating that a single global pre-trained model is inadequate and that locally tuned models are required. Together, these results define minimum data requirements, preferred retraining intervals, and the need for site-specific bias-corrected models, offering concrete design rules for operational PM2.5 forecasting systems.

During the 2024 ozone (O 3 ) season, the first complete year with Tropospheric Emissions: Monitoring of Pollution (TEMPO) observations over the northeastern United States, TEMPO provided hourly daytime retrievals of formaldehyde (HCHO), a proxy for volatile organic compound reactivity, and nitrogen dioxide (NO 2 ), an O 3 precursor. Matched TEMPO Version 03 (V03) and profile-shape-adjusted Tropospheric Monitoring Instrument (TROPOMI) Version 02.06.00 early-afternoon (~13:30 LT) retrievals show similar spatial patterns and changes across April, the temperature-defined early-season period (1 May-15 June), and midsummer (16 June-5 August). For 2024 O 3 season daily regional means, TROPOMI relative to TEMPO shows normalized relative difference (NRD) of -2% for tropospheric NO 2 (r=0.60) and -44% for total HCHO (r=0.85) vertical column density. Both instruments capture seasonal shifts in O 3 photochemistry from April to midsummer: period-median early-afternoon HCHO increases by 110% in TEMPO and 103% in TROPOMI as biogenic emissions increase, while NO 2 declines by 15% and 19%, respectively, as photochemistry intensifies. TEMPO reveals a steeper daytime HCHO column build-up from early season to midsummer (+0.79 to +1.01×10 15 molec cm -2 h -1 , 08:30-14:30 LT), alongside faster morning surface O 3 build-up (+2.7 to +4.0 ppb h -1 ) and an earlier daily O 3 peak (15:00 to 12:00 LT). On high-O 3 days, TEMPO tropospheric NO 2 is enhanced in the morning relative to period-median conditions, while total HCHO is elevated throughout daytime. Early-season events are characterized by a suppressed nocturnal boundary layer that favors overnight precursor accumulation followed by morning NO 2 column enhancements, whereas midsummer events show larger absolute HCHO column anomalies and persistently warmer conditions.

Aerosols are a key climate forcer and harmful to human health at the surface. Accurately modeling aerosol optical properties, mass loading and their relationship is important for constraining aerosol-climate forcing and characterizing particulate matter pollution exposure. We investigate the drivers of uncertainties in the NASA Goddard Earth Observing System Chemistry Climate Model (GEOSCCM) in simulating aerosols by focusing on the link between aerosol optical properties and mass. We compare a GEOSCCM hindcast with long-term coincident observations including satellite AOD measurements, speciated PM2.5 datasets from observations-model data fusion, and ground-based measurements of aerosol mass and optical properties. We analyze regional trends and seasonal variations of AOD and PM2.5, and surface aerosol properties, including relative humidity's role in hygroscopic enhancement. This work also presents the first extensive assessment of GEOSCCM's aerosol component with observational data. Our findings show that biases in PM2.5 components and relative humidity significantly impact simulated aerosol scattering at the surface, while scattering efficiency assumptions align with observations. This indicates that errors in simulated scattering relate more to simulated aerosol speciated mass and relative humidity than optical properties and size distribution assumptions in GEOSCCM. Our work highlights the importance of relative humidity biases on aerosol scattering enhancement for climate models where meteorology is not prescribed. Findings suggest improvements in GEOSCCM aerosols mass and optical properties could be achieved through updating emission inventories, especially over biomass burning regions, reducing nitrate biases, and improving relative humidity simulation.

Abstract Many of today's most significant and urgent scientific questions, including those about the impact of human activity on air quality and radiative forcing, require the synthesis of observations of atmospheric constituents and scientific theory. The past three decades have each seen a step‐change in the coverage, accuracy, and precision of constituent observations, and hence the questions they can address, especially from the growing constellation of Earth‐observing satellites. This paper is the first comprehensive description of the design and implementation of NASA's Constituent Data Assimilation System (CoDAS), a software package for assimilating constituent observations into the Goddard Earth Observing System (GEOS), an integrated collection of Earth system models. CoDAS is a generalized, tracer‐agnostic system that extends previous capabilities to assimilate multiple species from multiple sensors configurable at run time. Data assimilation provides a means of monitoring constituent changes, intercomparing heterogeneous types of observations, and reconciling data and models. Assimilation produces an optimal synthesis of the ingested data and model, taking advantage of the fact that the statistics of model‐data differences are simpler than those of the observations themselves. The use of a model enables the propagation of information from all previous observations across space and time, which can improve estimates even where and when data are unavailable.

Stratospheric water vapor (SWV) is a greenhouse gas that has an important, yet uncertain, impact on the Earth's climate through its radiative effect and feedback. As the climate changes, it is thus critical to monitor and understand changes in SWV. NASA's Microwave Limb Sounder (MLS) aboard the Aura satellite has observed SWV since 2004 but is reaching end of life. The Stratospheric Aerosol and Gas Experiment (SAGE) missions observe SWV as well, with the SAGE III instrument operating on the International Space Station (ISS) since 2017. We use the constituent data assimilation capabilities of NASA's Goddard Earth Observing System (GEOS) to demonstrate that the up to 30 SAGE III/ISS profiles each day provide a useful constraint over the observed midlatitudes and tropics. We conclude that by assimilating SAGE III/ISS SWV into GEOS we can largely continue to monitor SWV after Aura MLS.

Reanalysis datasets are widely used to understand atmospheric processes; however, different reanalyses may give very different results for the same diagnostics. The Atmospheric Processes And their Role in Climate (APARC; formerly SPARC) Reanalysis Intercomparison Project, or S(soon to be A)-RIP (https://s-rip.github.io/), is a coordinated activity to compare key diagnostics among atmospheric reanalyses, identify differences among reanalyses and their underlying causes, provide guidance on appropriate usage of reanalyses in scientific studies, and contribute to future improvements in the reanalysis products via collaborations with reanalysis centers and data users. S-RIP Phase 1 (completed in early 2022) focused primarily on the upper troposphere and above and processes linking these regions to the troposphere and surface. We are broadening our efforts in Phase 2 (S-RIP2), with new directions including studies of the tropospheric circulation, extreme weather events, and their links to the stratosphere, along with evaluation of chemical reanalyses, both those with a stratosphere / upper troposphere focus and those that focus on air quality applications. This presentation will provide a summary of Phase 1 results and discussion of future directions for S-RIP2, emphasizing applications to composition and chemistry studies and capacity building for Early Career Scientists.

On June 15, 2022, a Virginia Department of Environmental Quality (DEQ) monitoring site in a rural area of southeast Virginia recorded its first ozone exceedance since 2016. On this day, the daily maximum 8-h average surface ozone concentration reached 75 ppb with a 1-h maximum ozone concentration of 82 ppb measured at 17:00 UTC. In this analysis, we utilize various observational data and models to determine the atmospheric conditions responsible for this ozone exceedance event. To do this, we (1) evaluate the accuracy of the Goddard Earth Observing System Composition Forecasting (GEOS-CF) and Weather Research and Forecasting with Chemistry (WRF-Chem) model forecasts for these conditions and (2) use them to investigate the mechanisms responsible for the high surface ozone at the Virginia DEQ monitoring site. Comparisons of model forecasted ozone with measured ozone by lidars from the Tropospheric Ozone Lidar Network (TOLNet) at NASA Langley and NASA Goddard reveal that both models forecasted ozone reasonably well near the surface. Observational data show an upper-level ridge and surface anticyclone over the eastern United States on June 14 and 15. Back trajectory calculations using the Hybrid Single-Particle Lagrangian Integrated Trajectory model (HYSPLIT) and model data suggest that a low-level plume of polluted air from the New York City/New Jersey region was transported along the east coast, arriving in southeast Virginia around 10:00 UTC on June 15. This plume quickly mixed down to the surface throughout the morning, elevating the surface ozone concentration as well as the concentrations of several precursor species.

Abstract This Algorithm Theoretical Basis Document (ATBD) describes the retrieval algorithm and sensitivities of the Version 3 cloud product derived from the spectra collected by the Tropospheric Emissions: Monitoring of POllution (TEMPO) instrument. The cloud product is primarily produced for supporting the retrievals of TEMPO trace gases that are important for understanding atmospheric chemistry and monitoring air pollution. The TEMPO cloud algorithm is adapted from NASA's Ozone Monitoring Instrument (OMI) oxygen collision complex (O2‐O2) cloud algorithm. The retrieval generates effective cloud fraction (ECF) from the normalized radiance at 466 nm and generates cloud optical centroid pressure (OCP) using the O2‐O2 column amount derived from the spectral absorption feature near 477 nm. The slant column of O2‐O2 is retrieved using Smithsonian Astrophysical Observatory's spectral fitting code with optimized retrieval parameters. ECF and OCP are used by TEMPO trace gas retrievals to calculate Air Mass Factors which convert slant columns to vertical columns. The sensitivities of the cloud retrieval to various input parameters are investigated.

Surface ozone (O 3 ) mixing ratios exceeding the National Ambient Air Quality Standard were measured at rural monitors along the Colorado Front Range on 17 April 2020 during the COVID‐19 lockdown. This unusual episode followed back‐to‐back upslope snowstorms and coincided with the presence of a deep stratospheric intrusion, but ground‐based lidar and ozonesonde measurements show that little, if any, of the O 3 ‐rich lower stratospheric air reached the surface. Instead, the statically stable lower stratospheric air suppressed the growth of the daytime boundary layer and trapped nitrogen oxides (NO x = NO + NO 2 ) and volatile organic compounds (VOCs) emitted by motor vehicles and oil and natural gas (O&NG) operations near the ground where the clear skies and extensive snow cover triggered a short‐lived photochemical episode similar to those observed in the O&NG producing basins of northeastern Utah and southwestern Wyoming. In this study, we use a combination of lidar, ozonesonde, and surface measurements, together with the WRF‐Chem and Goddard Earth Observing System composition forecast models, to describe the stratospheric intrusion and characterize the boundary layer structure, HYSPLIT back trajectories to show the low‐level transport of O 3 and its precursors to the exceedance sites, and surface measurements of NO x and VOCs together with a 0‐D box model to investigate the roles of urban and O&NG emissions and the COVID‐19 quarantine in the O 3 production. The box model showed the O 3 production to be NO x saturated, such that the NO x reductions associated with COVID‐19 exacerbated the event rather than mitigating it.

As poor air quality continues to pose threats to humanity, better modeling of atmospheric composition is critical both for environmental and climate monitoring. The NASA Goddard Earth Observing System Composition Forecast (GEOS-CF) provides daily global air quality forecasts of atmospheric composition. Due to the full chemistry mechanisms and transport of hundreds of chemical tracers, GEOS-CF forecasts cannot extend beyond 5 days, and probabilistic estimates of tracer concentrations are computationally infeasible to calculate. Probabilistic estimates are critical in quantifying the forecast uncertainty and are achieved by generating an ensemble of models with slightly different initializations. We describe a 2-yr study to build a deep learning ensemble emulator. In this study, we describe why models such as NVIDIA's FourCastNet are inadequate for learning atmospheric composition and can experience instabilities during training. We share competitive forecast skill for carbon monoxide, nitric oxide, nitrogen dioxide, and ozone, using a novel air quality conditional generative adversarial network (AQcGAN) built for emulating ensembles. Using this model, we are able to forecast ten days into the future generating both ensemble mean and spread in seconds for a given ensemble, offering a low-cost option for extending the forecast window in GEOS-CF and for quantifying uncertainty. Results show that when compared to the actual mean and spread of the ensembles, the emulator forecasts offer competitive skill in terms of low RMSE scores. We share results from an operational data assimilation study where the AQcGAN has been used as the ensemble. We describe future plans for integration with data assimilation methods and GEOS-CF.

Air quality (AQ) is a major and growing concern for public health around the world. Economic development, population growth, and climate change are all expected to exacerbate already poor AQ in many regions. Furthermore, AQ is often only sparsely monitored with reference-grade in-situ instruments. NASA resources and products have the potential to help in addressing this AQ data gap. The GEOS-CF (Goddard Earth Observing System-Composition Forecasting) global atmospheric composition modeling system is run each day at a global scale to provide recent estimates and five-day forecasts at hourly temporal resolution of atmospheric constituents relevant to AQ. NASA satellite missions (along with those of other space agencies) provide remotely sensed estimates of atmospheric composition relevant to AQ. This paper gives a brief overview of these capabilities, and outlines the efforts underway to combine model forecasts, satellite retrievals, and surface-based measurements to provide more comprehensive and accurate estimates and forecasts of local AQ which will be broadly applicable and accessible globally.

Tropospheric ozone is an air pollutant and a greenhouse gas whose anthropogenic production is limited principally by the supply of nitrogen oxides (NOx) from combustion. Tropospheric ozone in the northern hemisphere has been rising despite the flattening of NOx emissions in recent decades. Here we propose that this sustained increase could result from the photolysis of nitrate particles (pNO3-) to regenerate NOx. Including pNO3- photolysis in the GEOS-Chem atmospheric chemistry model improves the consistency with ozone observations. Our simulations show that pNO3- concentrations have increased since the 1960s because of rising ammonia and falling SO2 emissions, augmenting the increase in ozone in the northern extratropics by about 50% to better match the observed ozone trend. pNO3- will likely continue to increase through 2050, which would drive a continued increase in ozone even as NOx emissions decrease. More work is needed to better understand the mechanism and rates of pNO3- photolysis.

Stratospheric water vapor (SWV) is a greenhouse gas that has a significant, yet uncertain, impact on the Earth’s climate through its radiative effect and feedback. As the climate changes, it is thus critical to monitor and understand changes in SWV. NASA’s Microwave Limb Sounding (MLS) aboard the Aura satellite has observed SWV since 2004 but will soon reach end of life. The Stratospheric Aerosol and Gas Experiment (SAGE) missions observe SWV as well, with the SAGE III instrument operating on the International Space Station (ISS) since 2017. We use the constituent data assimilation capabilities of NASA’s Goddard Earth Observing System (GEOS) to demonstrate that the up to 30 SAGE III/ISS profiles each day provide a useful constraint on SWV over the observed midlatitudes and tropics. We conclude that assimilating SAGE III/ISS SWV into GEOS can continue the SWV climate data record of Aura MLS.

Abstract Integrating air quality information from models, satellites, and in situ monitors allows for both better estimation of air quality and better quantification of uncertainties in this estimation. Uncertainty quantification is important to appropriately convey confidence in these estimates and forecasts to users who will base decisions on these. Uncertainty quantification also allows tracing the value of information provided by different data sources. This can identify gaps in the monitoring network where additional data could further reduce uncertainties. This paper presents a framework for data fusion with uncertainty quantification, applicable to multiple air‐quality‐relevant pollutants. Testing of this framework in the context of nitrogen dioxide forecasting at sub‐city scales shows promising results, with confidence intervals typically encompassing the expected number of actual measurements during cross‐validation. The framework is now being implemented into an online tool to support local air quality management decision‐making. Future work will also include the incorporation of low‐cost air sensor data and the quantification of uncertainty at hyper‐local scales.

Satellite-derived spatiotemporal patterns of nitrogen oxide (NOx) emissions can improve accuracy of emission inventories to better support air quality and climate research and policy studies. In this study, we develop a new method by coupling the chemical transport Model-Independent SATellite-derived Emission estimation Algorithm for Mixed-sources (MISATEAM) with a divergence method to map high-resolution NOx emissions across US cities using TROPOspheric Monitoring Instrument (TROPOMI) tropospheric nitrogen dioxide (NO2) retrievals. The accuracy of the coupled method is validated through application to synthetic NO2 observations from the NASA-Unified Weather Research and Forecasting (NU-WRF) model, with a horizontal spatial resolution of 4 km × 4 km for 33 large and mid-size US cities. Validation reveals excellent agreement between inferred and NU-WRF-provided emission magnitudes (R= 0.99, normalized mean bias, NMB = −0.01) and a consistent spatial pattern when comparing emissions for individual grid cells (R=0.88±0.06). We then develop a TROPOMI-based database reporting annual emissions for 39 US cities at a horizontal spatial resolution of 0.05° × 0.05° from 2018 to 2021. This database demonstrates a strong correlation (R= 0.90) with the National Emission Inventory (NEI) but reveals some bias (NMB = −0.24). There are noticeable differences in the spatial patterns of emissions in some cities. Our analysis suggests that uncertainties in TROPOMI-based emissions and potential misallocation of emissions and/or missing sources in bottom-up emission inventories both contribute to these differences.

The effect of sea breeze circulation on stratification of the vertical formaldehyde (HCHO) concentration vertical profiles is explored using a regional atmospheric chemical transport model (CTM) for three synoptically stagnant days focused on the east coast of the U.S in June 2018. During this event, a significant thermal contrast between the Atlantic Ocean and the terrestrial regions (12-17 degrees C), observed by moderate resolution imaging spectroradiometer (MODIS) and well-captured by the WRF-CMAQ model (15-18 degrees C), is conducive to monsoon-like flow, perpendicular to the shorelines, carrying clean marine air masses over the land within a few hundreds of meters above the surface. In contrast, the westerly continental polluted air masses prevail in higher altitudes. These two conflicting flows result in atypical vertical shapes of HCHO concentrations increasing with altitude. This decoupling pattern is so pronounced that we observe total column HCHO negatively correlate with surface concentrations. Comparisons of an accredited global model, GEOS-CF, to surface wind measurements and MODIS skin temperature indicate its poor representation of the sea breeze timing and strength, resulting in GEOS-CF HCHO vertical shapes being drastically different from the WRF-CMAQ. Based on radiative transfer calculations, the differences in the vertical distribution of HCHO between the WRF-CMAQ and that of GEOS-CF in the first 3 km are sufficient to induce a 20-30% error in air mass factors (thus total vertical HCHO column abundances). Through an experiment involving converting HCHO total columns to surface mixing ratios, we demonstrate that GEOS-CF allocates noticeably more HCHO molecules (40-150%) to the surface layer due to the misrepresentation of the vertical shape of HCHO during the sea breeze event. It is known that a significant fraction of the human population lives in coastal areas prone to detrimental effects caused by air pollution, and elevated pollutant concentrations usually occur in synoptically stagnant atmospheric conditions when local circulation patterns come into play; accordingly, our experiments emphasize the importance of the effect a priori profiles can have on satellite-derived applications under such conditions. To ensure that the quantitative representation of satellite-based trace gas retrievals on a daily basis is trustworthy and useable for air quality applications, atmospheric models providing a priori profiles for satellite retrievals should be welltuned to reproduce complex local circulation such as sea-land breezes.

During polar spring, periods of elevated tropospheric bromine drive near complete removal of surface ozone. These events impact the tropospheric oxidative capacity and are an area of active research with multiple approaches for representing the underlying processes in global models. We present a method for parameterizing emissions of molecular bromine (Br 2 ) over the Arctic using satellite retrievals of bromine monoxide (BrO) from the Ozone Monitoring Instrument (OMI). OMI retrieves column BrO with daily near global coverage, and we use the GEOS‐Chem chemical mechanism, run online within the Goddard Earth Observing System Earth System Model to identify hotspots of BrO likely associated with polar processes. To account for uncertainties in modeling background BrO, hotspots are only identified where the difference between OMI and modeled columns exceeds a statistical threshold. The resulting hotspot columns are a lower‐limit for the portion of OMI BrO attributable to bromine explosion events. While these hotspots are correlated with BrO measured in the lower troposphere over the Arctic Ocean, a case study of missing detections of near‐surface BrO is identified. Daily flux of Br 2 is estimated from hotspot columns of BrO using internal model parameters. When the emissions are applied, BrO hotspots are modeled with a 5% low bias. The sensitivity of the resulting ozone simulations to the treatment of background uncertainties in the BrO column is demonstrated. While periods of isolated, large (>50%) decreases in surface ozone are modeled, this technique does not simulate the low ozone observed at coastal stations and consistently underestimates ozone loss during March.