Opportunities and challenges in using satellite data for U.S. air quality management

6-16-16 DRAFT ONLY do not cite or quote


Over the past half-century, long-term measurements of ambient air pollution from a network of ground-based monitors have been established to assess compliance with the U.S. National Ambient Air Quality Standards (NAAQS). These monitoring networks target the six NAAQS criteria pollutants and directly support regulatory decision-making on air quality such as attainment demonstrations. Pollutant concentrations measured by these monitoring networks are often examined alongside weather data to characterize the mechanisms controlling air pollution events, which in turn supports air quality forecasting, planning, and rule-making. Monitor data are also used to evaluate the skill and reliability of air quality models in regulatory and research frameworks. Ground-based monitors are essential to documenting the success of U.S. air quality programs, and they provide the only source of measurement data explicitly designed to capture human exposure to health-relevant pollutants. 

Still, ground-based measurements have limitations. With the exception of major cities, monitors are sparsely located across the United States. While most urban areas have at least one monitor for ozone (O3) and fine particulate matter (PM2.5), rural areas typically have no monitors, and there are no monitors over water bodies. Outside of the U.S., even large cities may have few or no air pollution monitors. For pollutants that regularly fall below the NAAQS, like carbon monoxide (CO), nitrogen dioxide (NO2), and sulfur dioxide (SO2), monitor placement is especially limited, even though these pollutants react in the atmosphere to create O3 and/or PM2.5, and serve as key markers for regulated emissions. 

Satellite data have been available to complement ground-based measurements for certain pollutants since the mid-1990s, but their use has been rare in the air quality management community. A number of barriers limit the adoption of satellite data outside of a fairly narrow group of atmospheric researchers, including: the fact that that satellites measure the “column” of air above the surface, rather than ground-level concentrations; that data from existing satellites are not available on an hourly basis akin to monitors; that satellite data are not provided at the resolution or gridding appropriate for utilization by regional air quality modelers; and that satellite data do not fit into the methods for compliance with the Clean Air Act set forth by the U.S. Environmental Protection Agency. 

In 2011, the NASA Applied Science Program established the NASA Air Quality Applied Sciences Team (AQAST), a five-year initiative to conduct research relevant to the integration of NASA data and tools – especially satellite data – with air quality management frameworks. AQAST and selected activities were featured in a 2014 issue of EM magazine, with the focus “Satellite Data for Air Quality Management.”

The establishment of “Tiger Teams” – collaborative projects that bring together AQAST members in support of a common goal – has been a major part of the AQAST structure. This website and related content have emerged from a two-year AQAST Tiger Team, focused on the application of satellite data to understand high-pollution events in the eastern U.S. To build a two-way dialogue and understand how satellite data could better address air quality issues, our team met monthly from spring 2014 until summer 2016. Through this time, our community of scientists and stakeholders has grown to include nine state agencies (Connecticut, Maine, Maryland, Missouri, New Hampshire, New York, Texas, Vermont, and Wisconsin), four EPA offices (OAQPS, OAR, Region 2, and Region 5), three multi-state organizations (MARAMA, NESCAUM, and Ozone Transport Commission), one consulting company (AER) and 10 AQAST-funded investigator teams (led by Greg Carmichael, Dan Cohan, Russ Dickerson, Bryan Duncan, Arlene Fiore, Daven Henze, Tracey Holloway, Edward Hyer, Daniel Jacob, and Gabriele Pfister). 

Our group launched with a fairly narrow mandate: How can satellite data support source-attribution of high O3 and PM2.5 events? Over the course of our discussions and work, however, the team realized that the needs of state air quality managers could be more effectively realized with a broader vision. As such, we sought to explore and synthesize the application of satellite data to a range of state-level air quality planning and management needs. 

These policy applications form the basis for our website structure, wherein our team members explore and synthesize the role of satellite data in the State Implementation Plan (SIP) process (Lead: Dickerson); how to evaluate air quality trends and patterns with satellite data (Lead: Duncan); steps to identify dust and smoke with satellite data (Lead: Hyer); evaluating models with satellite data (Lead: Holloway); the role of satellite data and models in source-apportionment methods (Lead: Henze); characterizing background ozone with satellite data and models (Lead: Fiore); and analyzing short-term air pollution episodes with satellite data (Lead: Pfister). 

To set the stage for each of these policy applications, we begin here with an overview of satellite data for policy applications, and answers to questions posed by our stakeholder partners. This overview and website aims to provide an explicit policy link between the many resources available for satellite data and analysis, including published papers (e.g., Duncan et al., 2014 (link); Streets et al., 2013 (link)) that offer valuable background and more depth on key issues. 

Where, when, and for which pollutants are satellite data available? 

Satellite data are available from two types of platforms: polar-orbiting and geostationary. Polar-orbiting satellites cover the whole earth, with coverage typically once a day or less. Geostationary satellites rotate with the earth, and “see” the region with which they rotate. This allows geostationary satellites to observe more frequently over a smaller area (e.g., North America). From the perspective of a local air manager, the frequency of coverage can be as high as sub-hourly for geostationary instruments as compared to typical once-per-day or less for polar-orbiting satellites. 

Current satellite products used to address air quality typically provide daily information from instruments aboard polar-orbiting satellites that fly in low-earth orbit. The time of day at which the instrument passes over a particular location depends on the orbit, the satellite, and the viewing characteristics of the instrument. The years that are available depends on when the satellite was launched, and when it ceased operational utility due to malfunction or retirement. An overview of satellite data products for air quality application is given in the table below. 

New instruments are being built, with expected launch dates in the 2016-2022 time frame. These will expand coverage to higher temporal coverage and even finer spatial resolution. For example, the TropOMI instrument planned for 2016 launch will be a higher-resolution polar-orbiting instrument that is similar to the current OMI instrument. Also in 2016, the GOES-R geostationary satellite will launch, with aerosol optical depth (AOD) available every 10 minutes over the U.S. In 2018 or later, the TEMPO satellite will launch, which will provide OMI-like data on an hourly basis over the U.S. 

Satellite capabilities are rapidly expanding. In the evolution of a satellite data product, early scientific research typically focuses on the methods used to create the satellite retrieval and its reliability. Advancing analysis methods can sometimes yield new data from older instruments, such that it is typical for new or improved data products to be introduced even without the launch of new technology. As data are recognized as useful by the scientific community, continuing research often applies the space-based data to policy-relevant applications. 

These “proof of concept” scientific studies can serve as the bridge between the research community and air quality managers. For example __, __, __ highlight the use of aerosol optical depth (AOD) for analysis of Particulate Matter (PM); and ___, ___, ___ present how satellite-derived NO2 and HCHO can support analysis of ozone production regime (determining how NOx or VOC reductions could impact ozone).

How should a new user access satellite data for air quality applications?

To facilitate the wider use of satellite data, a number of platforms have emerged to support data access. We have found that NASA WorldView and NASA Giovanni provide satellite data in a manner that is particularly relevant to air quality managers. NASA WorldView provides near-real-time data on a daily basis, well suited to event analysis, in a clean, intuitive interface. NASA Giovanni is a more flexible platform that allows averaging over time, the creation of a range of plot types and formats, and comparing among variables. However, with ease of use comes limitations in the functionality of each platform. For example, WorldView does not allow data averaging, and it is best suited to qualitative evaluation. Also, because WorldView is designed for near-real-time applications, it does not (yet?) include NO2. Although Giovanni has the capacity for most applications, it does not include HCHO, nor does it allow for differencing the same variable across time periods. Most of these limitations are correctable, and the future development of these platforms will benefit from a wider user base providing feedback and suggestions. These platforms, however, already allow for sub-setting data that can be downloaded for user-based analysis. Free software, Panoply, is provided by NASA GISS for easy plotting of data downloaded from these sites, or any others, in NetCDF format (a format commonly used for large datasets such as from satellite instruments or models).

More advanced users typically benefit from an investment in capacity-building to use datasets of interest in a flexible data-analysis software such as Matlab, GIS, or the free NCAR Command Language (NCL). By downloading satellite data and creating maps and plots with an analysis software, the user can directly compare satellite retrievals with other datasets (monitor values, model simulations, etc.). This approach also allows the user to create custom visualizations, and to access wider range of satellite data products. 

Other useful data portals include:

NASA Goddard, Total NO2 vertical column densities (VCD)

NASA Earth Data, Giovanni

NASA Goddard Earth Sciences Data and Information Services Center, Mirador

NASA Aura Validation Data Center (AVDC) Overview/Home

AVDC, Total NOVCD and Tropospheric NO2 CS30 VCD

NASA Earth Data, LANCE Rapid Response MODIS images

How have satellite data been used, or how could they be used, for air quality management applications?

Our discussions within our AQAST Tiger Team have highlighted a few data sources that bear the most immediate relevance for policy applications. Visible smoke and aerosol optical depth (AOD) support analysis of atmospheric particulates in general  including near-surface PM2.5 – although, it is important to recognize that column measurements are not necessarily directly translatable to surface air. Nitrogen dioxide is a powerful near-source tracer of fossil-fuel emissions providing information on pollution transport over local to regional scales (with a lifetime of hours to days), whereas CO is a tracer for fossil-fuel combustion, and the transport of associated emissions on regional to inter-hemispheric scales (with a lifetime of weeks to months) and provides information on pollution transport over local-to-regional small scales. Formaldehyde is a proxy for reactive VOCs, and may be used with NO2 to determine the likely response of ozone to emission reductions of NOx versus VOCs.

There are three main ways to use satellite data for policy applications: for qualitative applications, for quantitative applications, and for more advanced analysis.

For qualitative understanding, satellite images allow air quality managers to see and communicate spatial patterns, atmospheric transport, and trends in air pollution. Visualizations from satellite data depict transport of wildfire smoke and dust plumes across regions, continents, and even oceans. Maps of NO2 across the U.S. show clear patterns of cities and suburbs, even major interstate highways and railroads. Both NASA Worldview and Giovanni are well suited to these types of qualitative application.

Satellite data can also be used to quantify change and relative abundance. The measurement units from satellite instruments typically reflect column densities (e.g., molecules/cm2) or optical properties (e.g., the unitless AOD metric, which varies from 0 to 1). These units do not compare directly with atmospheric composition unit (e.g., molecules/cm3, ug/m3, or mixing ratio). However, the percentage change or the ratio of two related species provides a metric that may be directly compared, as shown for NO2 in the 20 largest U.S. cities (right-hand panel of the NO2 plots shown for each city). 

Beyond qualitative and simple quantitative calculations, satellite data support a wide range of advanced analysis, especially combining with complementary data sources. Satellite data are well suited to evaluate photochemical grid models, and to support the derivation of ground-level pollution estimates in unmonitored areas. Particulate matter estimates derived from AOD products retrieved from multiple satellite instruments are used in public health studies. These types of advanced analysis measures usually require that users download satellite data and customize analysis to suit applications.

How can satellite data be used to evaluate emissions inventories?

Because satellite data provide a “top down” view of the atmosphere with nearly continuous spatial coverage, they offer an unparalleled opportunity to evaluate emission inventories. Most emission inventories are developed using “bottom up” methods, based on energy use, land cover, and emission factors. Satellite data thus provides an external check that can support improved inventory development and applications. 

The most direct approach for emissions comparison with satellite data is to compare spatial and temporal patterns of short-lived, primary species, especially NO2. By comparing trends and maps of NO2, air quality managers can assess whether inventories are capturing the same features as the satellite instrument. This type of analysis can be done qualitatively – for example, comparing NO2 maps or trend-lines from NASA Giovanni with similar plots of NOx inventories over a particular region. It can also be performed quantitatively by comparing the results of a model simulation (where the emissions inventory was input) with satellite column data. Model-simulated NO2, for example, offers an “apples to apples” comparison with satellite column NO2, since the model accounts for the chemical and meteorological processes to which the emissions are subjected in the atmosphere. 

Still, there are barriers to emissions evaluations, discussed in Streets et al., 2013, and mentioned here with a focus on policy applications. 

Satellites measure what is in a column of air, subject to chemical and meteorological processing, whereas emissions are the direct release of chemicals prior to such processing. As such, the units are not comparable, and even patterns and trends may differ due to atmospheric processing. This limitation can be partially corrected by comparing satellite retrievals with the output of photochemical grid models. While a short-lived pollutant like NO2 is well suited for emissions evaluation, its short lifetime also introduces a high level of sensitivity to satellite overpass time, chemical processing, and boundary layer mixing. The highest resolution NO2 retrieval available is from the OMI instrument, which passes over the U.S. in the early afternoon. In many urban areas, surface NO2 is lowest in the afternoon due to boundary layer mixing and photochemistry, such that conclusions derived from satellite NO2 may be sensitive to these atmospheric (non-emissions) processes and/or the diurnal pattern of NO2 emissions associated with urban traffic patterns, power plant generation, and other factors.  

How reliable are satellite measurements of air pollutants?

Reliability is different than scientific uncertainty, or the error associated with each satellite data product. Uncertainty of the data product is a factor in its reliability, but the same data product may be reliable in one context and unreliable in a different context. For example, the spatial pattern of AOD is a reliable indicator of downwind transport of smoke over large scales, but it is not a reliable indicator of daily variability in urban PM. 

In general, satellite data are most reliable for qualitative trends and patterns, evaluated over longer time periods and wider spatial domains, for the actual species derived by the satellite retrieval. For example, maps showing NO2 trends over the U.S. show – with a high level of reliability  a decrease in NO2 over the past 10 years. It would be less reliable to quantify day-to-day change in NO2 over a particular location. Whereas NO2 is directly calculated from measured irradiances, AOD is a satellite-based proxy for particulates. Thus the derivation of particulate mass concentration from AOD is less reliable, because different scattering and absorbing characteristics of PM (e.g., sulfate versus black carbon, size distributions) affect AOD in different ways.

Regulatory applications of satellite data have been used primarily to supplement ground-based monitors. This can include supporting a weight-of-evidence exceptional event demonstration, evaluating photochemical grid models, or to characterizing air trends over a wide area. In this context, the reliability of satellite data can be directly assessed by comparison with the more limited ground-based data. There are ample precedents for the integration of uncertain data in decision-making frameworks, such as weather prediction, hazard response, and land use change. The widespread use of satellite data for air quality management may depend as much on developing relevant decision-making and communication protocols, as it will on developing improved satellite instrumentation and algorithms.

By understanding the needs of the air quality management community, researchers have been able to focus on the uncertainty characteristics of satellite data most relevant for policy applications. For example, ongoing work at Columbia University is assessing the reliability of the HCHO-to-NO2 ratio for ozone sensitivity in different environments. This type of work provides key information to support the appropriate use of satellite data in decision frameworks, and paves the way for technological and scientific improvements that can promote new applications.

How can I get started? 

NASA Worldview is the best starting point for users new to satellite data. First-time visitors to the site are offered an online tour, and the intuitive interface allows for easy navigation. Air quality issues begin with the visible imagery, as very high concentrations of smoke and dust appear gray and yellow, respectively. Worldview allows users to add information layers, including AOD and CO, as well as markers relevant to air quality such as fires and thermal anomalies and dust score. Worldview supports qualitative analysis, but limits data to daily values (no averaging function is included).

For quantitative analysis and a wider range of variables, NASA Giovanni is an easy next step up from WorldView. For many users, NASA Giovanni may meet their needs for trend analysis and map-making from satellite data.

It is likely that a new user may find that NASA Giovanni is too restrictive and/or is lacking key information. However, between Giovanni, Worldview, and other information sources (especially the overview provided by Duncan et al., 2014, and the table above), users can identify which data products will be useful, and download those data.

Users will select either “Level-3” (gridded) data or “Level-2” (granule) data. Level-3 is easier to use, as the data are already gridded in a format that can be plotted with analysis software. Level-2 granules are irregularly shaped polygons, and may be more challenging to plot. However, averaging to the grid (Level-3) may lower the information quality of the data for a particular time/place.

Beyond these online resources, the NASA Applied Remote Sensing Training (ARSET) program offers a range of free in-person and online courses to support training in satellite-data application.

As users gain familiarity with satellite data, the topics covered in the essays to follow on this website provide deeper discussions of specific policy applications. Whether through this website or other online, printed, and in-person resources, we appreciate your interest in this exciting initiative.  

Editors: Tracey Holloway and Arlene Fiore, co-leads of NASA AQAST Tiger Team on Eastern U.S. Episodes