Previous studies on net ozone production rates (PO3) and their sensitivities to precursors relied on limited in-situ data, often coarse and uncertain chemical transport models (CTMs), and ozone indicators like the formaldehyde-to-nitrogen dioxide ratio (FNR). However, FNR fails to fully capture PO3's complex relationships with pollution, light, and water vapor. To address this, we refine the satellite-based PO3 product from Souri et al. (2025) with key advancements: (i) a deep neural network to parametrize high-dimensional non-linear ozone chemistry without the need for empirical linearization of atmospheric conditions, (ii) incorporation of water vapor, (iii) improved error characterization, and (iv) the application of a finer CTM to dynamically convert column retrievals into near-surface mixing ratios. Our PO3 sensitivity maps surpass traditional FNR-based assessments by quantifying sensitivity magnitudes - factoring in photolysis rates and water vapor - with greater spatial information. Our new product provides daily near-clear sky PO3 and sensitivity maps using bias-corrected OMI (2005-2019, 0.25 degrees x 0.25 degrees) and TROPOMI (2018-2023, 0.1 degrees x 0.1 degrees), with values aligning within 10 %. High PO3 rates (> 8 ppbv h(-1)) appear in urban and biomass-burning regions under strong photochemical activity, including during a heatwave in the northeastern U.S. Photolysis rates are the dominant factor dictating the seasonality of PO3 magnitudes and sensitivities. The stability and long-term records of OMI retrievals (2005-2019) enable us to provide the first global maps of PO3 linear trends showing a surge of > 30 % over China, the Middle East, and India, while a reduction in the eastern U.S., southern Europe, and several regions in Africa.

Recent progress in constraining the atmosphere's primary oxidant, the hydroxyl radical (OH), with machine learning (ML) and satellite data raises the intriguing possibility of also constraining individual OH chemical production and loss terms. Here, we present a methodology to constrain primary OH production (i.e., OH production from the reaction of water vapor with O1D) from 60 degrees S-60 degrees N at 500m above ground level (magl) (POH_500) using a combination of ML, satellite observations, and meteorological data. The aim of this work is to establish methodological feasibility and to assess and quantify the uncertainties of that methodology. This methodology produces geophysically credible distributions of POH_500 across all seasons, with seasonal variability being driven primarily by changes in water vapor and ozone photolysis rates. Regions with quantifiable 1 sigma uncertainties of 25% or less comprise approximately 68%-73% of global POH_500, suggesting the product is of sufficient quality to inform the relationship between POH and trends and variability in OH. The incorporation of additional satellite retrievals into the machine learning model as well as increased spatial and temporal averaging could reduce errors in regions with higher uncertainties, such as those areas with frequent clouds or biomass burning. Ultimately, the results presented here can provide a blueprint to observationally constrain other production and loss terms within the OH budget.

The accurate representation of tropospheric hydroxyl radical (TOH) is crucial for reasonably modeling methane concentrations — a potent greenhouse gas. We use an improved parameterization of TOH using an interpretable and agile machine learning module named ECCOH (pronounced "echo") in NASA's GEOS global model to unravel the intricacies of TOH to its key inputs. However, the accuracy of this model is hampered by the accurate representation of its critical inputs. Fortunately, retrieving trace gases like nitrogen dioxide (NO2) and formaldehyde (HCHO) from space-borne sensors, like the Aura Ozone Monitoring Instrument (OMI), has seen remarkable progress. Consequently, we leverage these observations to assess how they can effectively alleviate some biases in TOH and can help better reproduce its long-term trends. In contrast to the earlier investigations, the refined representation of TOH archives a finer spatial resolution (1x1 degrees), and it is more up to date (2005-2019), allowing for elucidating the impact of recent emission regulations, such as those imposed in China, on TOH. OMI NO2 yields valuable insights over biomass-burning areas in Eastern Europe and central Africa, where our prior emission estimates possess significant biases, mitigating regional TOH biases up to 20%. Oceanic HCHO concentrations, serving as a proxy for TOH due to the predominant chemical pathway of VOC oxidation through OH, are only moderately altered by OMI HCHO, attributed to low signal-to-noise ratios and satisfactory representation of HCHO in the a priori simulations. Ultimately, we disentangle the convoluted map of TOH linear trends by isolating five pivotal inputs to the TOH parameterization, including stratospheric ozone, tropospheric ozone, water vapor, HCHO, and NO2. Our results demonstrate that these five parameters can collectively explain 65% of the variability in TOH trends alone. With the deployment of new satellites with enhanced sensor configurations and better temporal resolutions, our mission at NASA is to exploit those observations to improve the representation of many variables highly linked to TOH.

Questions about how regulations have shaped ozone pollution regionally cannot be answered by studying observed surface ozone concentrations alone, as we must precisely determine ozone production rates, which refer to the amount of ozone molecules photochemically produced in the atmosphere. Through an extensive suite of NASA's airborne campaigns such as DISCOVER-AQs, KORUS-AQ, INTEX-B, SENEX, and ATOMs, constraining a well-characterized chemical box model, we establish a simple but robust relationship between ozone production rates and various observable parameters; some of these factors, fortunately, are being measured or constrained by satellite observations, which has allowed us to create the first-ever maps of ozone production rates across the globe. We quantitatively and qualitatively assess this product's efficacy through independent airborne campaigns and by contrasting extreme events to a norm. We have a clear path forward to enhance this innovative product using agile machine learning algorithms for the years from 2005 to 2024, along with its well-characterized error budget.

Abstract. Ozone pollution is secondarily produced through a complex, non-linear chemical process. Our understanding of the spatiotemporal variations in photochemically produced ozone (i.e., PO3) is limited to sparse aircraft campaigns and chemical transport models, which often carry significant biases. Hence, we present a novel satellite-derived PO3 product informed by bias-corrected TROPOMI HCHO, NO2, surface albedo data, and various models. These data are integrated into a parameterization that relies on HCHO, NO2, HCHO/NO2, jNO2, and jO1D. Despite its simplicity, it can reproduce ~90 % of the variance in observationally constrained PO3 with minimal biases in moderately to highly polluted regions. We map PO3 across various regions in July 2019 at a 0.1°×0.1° spatial resolution, revealing accelerated values (>8 ppbv/hr) in numerous cities throughout Asia and the Middle East, resulting from the elevated ozone precursors and enhanced photochemistry. In Europe and the United States, such high levels are only detected over Benelux, Los Angeles, and New York City. PO3 maxima are seen in various seasons, attributed to changes in photolysis rates, non-linear ozone chemistry, and fluctuations in HCHO and NO2. Satellite errors result in moderate errors (40–60 %) of PO3 estimates over cities on a monthly average, while these errors exceed 100 % in clean areas and under low light conditions. Using the current algorithm, we have demonstrated that satellite data can provide valuable information for robust PO3 estimation. This capability expands future research through the application of data to address significant scientific questions about the locally-produced PO3 hotspots, seasonality, and long-term trends.

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.

The tropospheric hydroxyl (TOH) radical is a key player in regulating oxidation of various compounds in Earth's atmosphere. Despite its pivotal role, the spatiotemporal distributions of OH are poorly constrained. Past modeling studies suggest that the main drivers of OH, including NO2, tropospheric ozone (TO3), and H2O(v), have increased TOH globally. However, these findings often offer a global average and may not include more recent changes in diverse compounds emitted on various spatiotemporal scales. Here, we aim to deepen our understanding of global TOH trends for more recent years (2005–2019) at 1×1°. To achieve this, we use satellite observations of HCHO and NO2 to constrain simulated TOH using a technique based on a Bayesian data fusion method, alongside a machine learning module named the Efficient CH4-CO-OH (ECCOH) configuration, which is integrated into NASA's Goddard Earth Observing System (GEOS) global model. This innovative module helps efficiently predict the convoluted response of TOH to its drivers and proxies in a statistical way. Aura Ozone Monitoring Instrument (OMI) NO2 observations suggest that the simulation has high biases for biomass burning activities in Africa and eastern Europe, resulting in a regional overestimation of up to 20 % in TOH. OMI HCHO primarily impacts the oceans, where TOH linearly correlates with this proxy. Five key parameters, i.e., TO3, H2O(v), NO2, HCHO, and stratospheric ozone, can collectively explain 65 % of the variance in TOH trends. The overall trend of TOH influenced by NO2 remains positive, but it varies greatly because of the differences in the signs of anthropogenic emissions. Over the oceans, TOH trends are primarily positive in the Northern Hemisphere, resulting from the upward trends in HCHO, TO3, and H2O(v). Using the present framework, we can tap the power of satellites to quickly gain a deeper understanding of simulated TOH trends and biases.

AbstractDespite its importance for the global oxidative capacity, spatially resolved trends and variability of the hydroxyl radical (OH) are poorly constrained. We demonstrate the utility of a tropospheric column OH (TCOH) product, created from machine learning and satellite proxy data, in determining the spatial variability in trends of tropical OH over the oceans during September through November. While OH increases domain‐wide by 2.1%/decade from 2005–2019, we find significant spatial heterogeneity in regional trends, with decreases in some areas of 2.5%/decade. Our analysis of the trends in the proxy data indicate anthropogenic‐driven changes in emissions of OH drivers as well as increasing temperatures cause these trends. This OH product is potentially a significant advance in constraining OH spatial variability and serves as a useful complement to existing tools in understanding the atmospheric oxidative capacity. Comprehensive observations of TCOH are required to assess the fidelity of this method.

Behind the scenes of a remote sensing mission there are complex decision making and planning operations. Streamlining these operations, with a quantitative scientific value framework, aids efficient and optimized science data collection. While there have been previous efforts to quantify the science value for specific science scenarios, our work aims to develop a general framework which can be applied across different scenarios. We describe a pipeline of processes which combines model forecast and observation data, in computational forms, as dictated by the mission objectives set forth by subject matter experts. The framework is described with use cases involving the monitoring of nitrogen dioxide (NO2) concentrations over the Gulf of Mexico and methane concentrations over interior Alaska.

The hydroxyl radical (OH) fuels atmospheric chemical cycling as the main sink for methane and a driver of the formation and loss of many air pollutants, but direct OH observations are sparse. We develop and evaluate an observation-based proxy for short-term, spatial variations in OH (ProxyOH) in the remote marine troposphere using comprehensive measurements from the NASA Atmospheric Tomography (ATom) airborne campaign. ProxyOH is a reduced form of the OH steady-state equation representing the dominant OH production and loss pathways in the remote marine troposphere, according to box model simulations of OH constrained with ATom observations. ProxyOH comprises only eight variables that are generally observed by routine ground- or satellite-based instruments. ProxyOH scales linearly with in situ [OH] spatial variations along the ATom flight tracks (median r2 = 0.90, interquartile range = 0.80 to 0.94 across 2-km altitude by 20° latitudinal regions). We deconstruct spatial variations in ProxyOH as a first-order approximation of the sensitivity of OH variations to individual terms. Two terms modulate within-region ProxyOH variations-water vapor (H2O) and, to a lesser extent, nitric oxide (NO). This implies that a limited set of observations could offer an avenue for observation-based mapping of OH spatial variations over much of the remote marine troposphere. Both H2O and NO are expected to change with climate, while NO also varies strongly with human activities. We also illustrate the utility of ProxyOH as a process-based approach for evaluating intermodel differences in remote marine tropospheric OH.

Despite its importance in controlling the abundance of methane (CH4) and a myriad of other tropospheric species, the hydroxyl radical (OH) is poorly constrained due to its large spatial heterogeneity and the inability to measure tropospheric OH with satellites. Here, we present a methodology to infer tropospheric column OH (TCOH) in the tropics over the open oceans using a combination of a machine learning model, output from a simulation of the GEOS model, and satellite observations. Our overall goals are to assess the feasibility of our methodology, to identify potential limitations, and to suggest areas of improvement in the current observational network. The methodology reproduces the variability of TCOH from independent 3D model output and of observations from the Atmospheric Tomography mission (ATom). While the methodology also reproduces the magnitude of the 3D model validation set, the accuracy of the magnitude when applied to observations is uncertain because current observations are insufficient to fully evaluate the machine learning model. Despite large uncertainties in some of the satellite retrievals necessary to infer OH, particularly for NO2 and formaldehyde (HCHO), current satellite observations are of sufficient quality to apply the machine learning methodology, resulting in an error comparable to that of in situ OH observations. Finally, the methodology is not limited to a specific suite of satellite retrievals. Comparison of TCOH determined from two sets of retrievals does show, however, that systematic biases in NO2, resulting both from retrieval algorithm and instrumental differences, lead to relative biases in the calculated TCOH. Further evaluation of NO2 retrievals in the remote atmosphere is needed to determine their accuracy. With slight modifications, a similar methodology could likely be expanded to the extratropics and over land, with the benefits of increasing our understanding of the atmospheric oxidation capacity and, for instance, informing understanding of recent CH4 trends.

Transitioning to a sustainable energy system poses a massive challenge to communities, nations, and the global economy in the next decade and beyond. A growing portfolio of satellite data products is available to support this transition. Satellite data complement other information sources to provide a more complete picture of the global energy system, often with continuous spatial coverage over targeted areas or even the entire Earth. We find that satellite data are already being applied to a wide range of energy issues with varying information needs, from planning and operation of renewable energy projects, to tracking changing patterns in energy access and use, to monitoring environmental impacts and verifying the effectiveness of emissions reduction efforts. While satellite data could play a larger role throughout the policy and planning lifecycle, there are technical, social, and structural barriers to their increased use. We conclude with a discussion of opportunities for satellite data applications to energy and recommendations for research to maximize the value of satellite data for sustainable energy transitions.

We present a methodology that uses gradient-boosted regression trees (a machine learning technique) and a full-chemistry simulation (i.e., training dataset) from a chemistry–climate model (CCM) to efficiently generate a parameterization of tropospheric hydroxyl radical (OH) that is a function of chemical, dynamical, and solar irradiance variables. This surrogate model of OH is designed to be integrated into a CCM and allow for computationally efficient simulation of nonlinear feedbacks between OH and tropospheric constituents that have loss by reaction with OH as their primary sinks (e.g., carbon monoxide (CO), methane (CH4), volatile organic compounds (VOCs)). Such a model framework is advantageous for studies that require multi-decadal simulations of CH4 or multi-year sensitivity simulations to understand the causes of trends and variations of CO and CH4. To allow the user to easily target the training dataset towards a desired application, we are outlining a methodology to generate a parameterization of OH and not presenting an “off-the-shelf” version of a parameterization to be incorporated into a CCM. This provides for the relatively easy creation of a new parameterization in response to, for example, changes in research goals or the underlying CCM chemistry and/or dynamics schemes. We show that a sample parameterization of OH generated from a CCM simulation is able to reproduce OH concentrations with a normalized root-mean-square error of approximately 5 % and capture the global mean methane lifetime within approximately 1 %. Our calculated accuracy of the parameterization assumes inputs being within the bounds of the training dataset. Large excursions from these bounds will likely decrease the overall accuracy. However, we show that the sample parameterization predicts large deviations in OH for an El Niño event that was not part of the training dataset and that the spatial distribution and strength of these deviations are consistent with the event. This result gives confidence in the fidelity of a parameterization developed with our methodology to simulate the spatial and temporal responses of OH to perturbations from large variations in the chemical, dynamical, and solar irradiance drivers of OH. In addition, we discuss how two machine learning metrics, Gain feature importance and Shapley additive explanations values, indicate that the behavior of a parameterization of OH generally accords with our understanding of OH chemistry, even though there are no physics- or chemistry-based constraints on the parameterization.

A sample parameterization of OH for July created to be used with the ECCOH module of the GEOS Earth System Model. The methodology to generate the parameterization is described in detail in the GMD article "A Machine Learning Methodology for the Generation of a Parameterization of the Hydroxyl Radical". The parameterization is included as an example and should not be used for research purposes without consulting the publication and/or the authors. We have also included the dataset used to train the parameterization (OHParameterization_TrainingSet_1980_2019_M07.dat) as well as the training targets (OHParameterization_Targets_1980_2019_M07.dat). The training data are from the NASA MERRA2 GMI simulation (https://acd-ext.gsfc.nasa.gov/Projects/GEOSCCM/MERRA2GMI/). The scripts used to generate the training dataset and the parameterization can be found at https://zenodo.org/record/6046037.

Background and aim: With the recent advancement of Earth observation, computer simulations, and low-cost monitoring technologies, the capacity to obtain accurate geospatial data has increased tremendously, particularly for locations without expansive in-situ monitoring networks. Here we present an optimized machine learning model to estimate near real-time air pollutant concentrations in selected locations in Africa and Latin America using a combination of NASA Goddard Earth Observing System Composition Forecasts (GEOS-CF) and low-cost sensor data. Methods: We use a machine learning approach to estimate near real-time air pollutants concentration at selected locations across Africa and Latin America. Several meteorological and chemical parameters are retrieved from NASA's GEOS-CF to train a bias corrector model and predict corrected concentration estimates for these locations which are then validated against local monitoring data. We also conduct an explainability approach via SHAP Analysis to quantify the model contributing factors across these locations and track the model performance in extreme conditions. Results: The optimized machine learning model shows good agreement with ground air quality data, with R2 values from 0.61- 0.65 for locations in Mexico City (Mexico), Bogotá (Columbia) and Kigali (Rwanda). Via SHAP analysis, we demonstrate the need to conduct a measure of variance to determine model performance for different conditions and intervals, knowing that the model performance can be affected by various training conditions. Conclusion: Combining observations and optimized model simulations using machine learning techniques can significantly improve air quality forecasts in low- and middle-income countries, where the rapid pace of industrialization and communities are highly susceptible to air pollution health effects, and rarely have local air quality data and health risks alerting systems in place. These results are being used to assist air quality managers and environmental agencies in these locations to improve risk communication and reduce health burdens associated with outdoor air pollution.