Multiscale Modeling of Background Ozone: Research Needs to Inform and Improve Air Quality Management

Introduction

For many decades, ground-level ozone has been recognized as a pollutant that causes adverse human health effects and damages crops and ecosystems. Scientific understanding about the ambient concentrations at which such damages occur has evolved, triggering several revisions to the National Ambient Air Quality Standard (NAAQS) for ozone that gradually reduced critical thresholds. Grid-based photochemical air quality modeling systems (AQM) such as the Community Multiscale Air Quality (CMAQ)¹ model and the Comprehensive Air Quality Model with Extensions (CAMx)² have long been used to develop air quality management strategies aimed at reducing emissions of precursor species (nitrogen oxides and volatile organic compounds) for areas exceeding the ozone NAAQS. As the NAAQS became more stringent, processes considered in AQM were expanded from representing photochemistry causing elevated ozone levels in highly polluted urban airsheds such as the Los Angeles basin to also considering regional aspects such as the multi-state transport of ozone and its precursors and the effects of long-range international transport.

Viewed from an air quality management perspective, the magnitude of ozone formed from locally controllable anthropogenic precursor emissions relative to background ozone (BG O3) levels (i.e., ozone formed from non-local anthropogenic precursor emission sources and local and non-local natural sources, regardless of where the ozone formation occurs) has generally decreased over time. This increases emphasis on representing the processes controlling BG O3 and characterizing its temporal and spatial fluctuations, especially for ozone exceedance events. The latter is particularly important because although the relative magnitude of typical average BG O3 concentrations with respect to the NAAQS has increased, BG O3 contributions to specific exceedance events have not increased in some cases^3,4.

Increased focus on BG O3 processes is evident in the broader research community⁵ and in EPA activities. For example, the EPA has summarized the state of BG O3 in policy assessments^6–9, participated in broader community research syntheses^4,10, and released guidance documents for BG O3 related events^11–12. Improving the representation of BG O3 in AQM improves their fidelity, which increases our confidence in the models’ ability to differentiate BG O3 from controllable sources. Improved differentiation promotes effective controls on anthropogenic sources. In this article, we provide our perspective on improving the characterization of BG O3 in AQM and discuss examples of using such information in air quality management.

Linking hemispheric-to-global and regional modeling in planning applications

AQM for current air quality planning are typically regional scale rather than global models. Therefore, species concentrations are prescribed along the lateral boundaries of the region that is modeled (“lateral boundary conditions” or LBC) and sometimes at the top of the model to reflect high ozone in the lower stratosphere. Historically, source attribution has differentiated BG O3 sources within regional scale models but did not consider source contributions to LBC. Increasingly, source attribution is also performed in multi-scale applications using zero-out techniques^3,7,13,14, AQM equipped with enhanced source tracking or source attribution capabilities^15,16, or by combinations of techniques¹⁷. Figure 1 illustrates how source-apportionment modeling performed on the hemispheric to global scale, e.g. with the hemispheric CMAQ (H-CMAQ)¹⁸ model, can provide additional information on the processes and source contributions to the “boundary condition” contribution estimated by a regional-scale CMAQ source-apportionment application.

Figure 1.

Example ozone source apportionment analysis performed on both a U.S. regional CMAQ and a Northern Hemisphere H-CMAQ modeling domain. Note that the H-CMAQ source contributions are shown for a longer time period to illustrate their temporal fluctuations.

Hemispheric-to-global modeling research needs

Natural sources, international anthropogenic sources, and ozone of stratospheric origin are all significant contributors to concentrations along the boundaries of the regional-scale modeling domain (Figure 1). Future research should target improving the representation of these sources in hemispheric models like H-CMAQ or H-CAMx or global models that are often used to generate LBC for regional AQM¹⁹. Improvements to these models should include the following:

• Better constraints for stratospheric ozone and stratosphere-troposphere air mass exchange. The hemispheric and global models differ in their characterization of stratospheric ozone, employing a variety of techniques, including parameterizations based on historical ozonesonde observations, using climatological satellite-derived concentrations fields, and simulating fully explicit stratospheric chemistry. Research is needed to systematically compare the strengths and shortcomings of the different approaches as they relate to air quality management applications. Stratosphere-troposphere air mass exchange occurs over a range of space and time scales, from individual short-lived stratospheric deep intrusion events that can be associated with strong fronts and may mix stratospheric air to the surface to slower mixing processes that control the large-scale ozone budget in the mid-to upper troposphere. Capturing the continuum of meteorological processes for the AQM requires representing dynamics in the tropopause with adequate vertical resolution and structure.

• Improved representation of natural emissions. Natural sources contributing to the tropospheric ozone budget include biogenic emissions of volatile organic compounds (VOC), emissions of nitrogen oxide (NO) from lightning, emissions of NO from soils, emissions of halogens from marine sources, and emissions of VOC and nitrogen oxides (NO_x) from fires. Some sources, like soil NO, have both natural (via nitrogen fixation) and anthropogenic (via fertilization) contributions. Thus, it is important to understand the spatiotemporal distribution of each source, the space and time scales of influence, and how their relative contributions evolve over time. Quantifying the magnitude and temporal and spatial variations of these emission sources is uncertain due to incomplete knowledge about some of the underlying processes, as well as limitations in accurately representing these processes which often involve the interactions between the atmosphere, biosphere, and hydrosphere contributing to variability in these emissions in an AQM framework. Development work should be aimed at improved representations of these emissions within the AQM system while constraining emission estimates with remote sensing information where appropriate.

• Improved characterization of international emissions. Detailed sector-specific emission inventories and approaches for processing such inventories into spatially, temporally, and chemically resolved fields for AQM have long been available over the U.S., Canada, and parts of Europe, but the same level of detail is generally unavailable elsewhere. International collaborations such as the Global Emissions Initiative (GEIA) or the Community Emissions Data System (CEDS) provide opportunities to address such data gaps in the future and are particularly important to improve the characterization of emissions from countries such as Mexico and China that have been shown to impact modeled BG O3 over the U.S.⁹ Ocean-going ships burning fossil fuels also contribute a significant fraction of the global NO_x and SO₂ emissions. Changing trade patterns, growing shipping traffic, and changing fuel types and utilization will continue to alter the composition and amounts of shipping emissions embedded within air masses as they undergo long-range transport and impact downwind continents. Emerging satellite-observed vessel activity data sets may help to provide more accurate estimates of the evolving contribution to BG O3.

• Enhanced characterization of BG O3 through data assimilation. The development needs listed above are aimed at improving the ability of the AQM to simulate tropospheric ozone through better process representations and input data. A parallel approach to improve the representation of tropospheric ozone that influences BG O3 in regional-scale AQM applications could leverage data assimilation in hemispheric or global scale models. Several assimilation studies have used satellite ozone and nitrogen dioxide to identify and update emissions estimates or add temporal trends^20–23. The Health and Air Quality Applied Science Team (HAQAST; haqast.org) included several assimilation projects designed to directly improve boundary conditions or to infer emission updates. Projects like these contribute data products that are useful to air quality managers and technology transfer between research institutions and the operational air quality agencies. Figure 2 provides a qualitative comparison of satellite products and a H-CMAQ model simulation. Differences between the satellite products and the model could potentially be leveraged with assimilation to update global emissions. Although data assimilation has the potential to improve the characterization of ozone inflow into regional domains for current and past conditions, it typically does not classify the sources contributing to this inflow. Further, data assimilation cannot project future changes in BG O3 that respond to changing emissions and climate.

Figure 2.

Example comparison of satellite-derived (left column) and modeled (center column) NO₂ (top row) and O₃ (bottom row) column densities vertically integrated throughout the troposphere for summer 2016. Relative differences between the satellite-derived and modeled values are shown in the right column. Note that in this example comparison, model values were limited to hours near the satellite overpass time, clear sky conditions, and did not have the satellite averaging kernel applied.

Regional-scale modeling needs

The previous section listed key development needs to improve the representation of BG O3 derived from hemispheric-to-global scale models and specified as boundary conditions for regional-scale AQM. In this section, we discuss several areas of research that would help ensure that regional-scale AQM adequately quantify the contributions of BG O3 to surface ozone.

• Representing vertical mixing between the free troposphere and boundary layer. Previous work^{19, 24, 25} showed that the sensitivity of surface O₃ concentrations towards LBC is generally most pronounced for those LBC that originate in the mid-troposphere. This implies that representing vertical mixing, including mixing by clouds, in AQM is key to capturing the effects of BG O3 on surface ozone. Accurate simulation of vertical mixing is especially important in the intermountain west, including ozone non-attainment areas in Colorado and Utah, where high elevation and complex terrain result in deep convective boundary layers and where BG O3 is a larger contributor compared to lower elevation states¹⁰. Liu et al.²⁶ performed a model intercomparison study and confirmed that differences in vertical grid structure and model representation of vertical mixing were the main reason for differences in boundary condition tracer concentrations at the surface.

• Improved representation of natural emissions. Similar to the hemispheric-to-global scale, an improved process-based representation of natural emissions on the regional scale would increase confidence in estimating BG O3 contributions to ground-level ozone. Over the U.S., emphasis should be placed on representing biogenic VOCs, lighting and soil NO, and fire emissions at finer temporal and spatial scales. Such improvements would improve quantifying the contributions to ozone from these sources in different areas (rural vs. urban) and on different days.

• Improved estimates of U.S. anthropogenic emissions. Accurate domestic emissions inventories for VOC and NOx are essential for AQM to simulate local ozone production, the relative effectiveness of VOC and NOx controls, and the relative importance of domestic versus background sources of ozone. VOC and NO_x emissions from oil and gas production activities are highly uncertain and may be underestimated by at least a factor of two^27–30. McDonald et al.³¹ found that consumer product VOC emissions are substantially underestimated causing AQM to underestimate local ozone production. Karamchandani et al.³²found that ozone trends over Los Angeles were more accurately modeled when the reported domestic VOC emissions were doubled. In contrast, Travis et al.³³ cite several studies suggesting that anthropogenic NOx emissions from several sectors were significantly overestimated and found that AQM performance for ozone in the southeastern U.S. improved when attempting to correct for this overestimation.

• Chemistry improvements. Photochemical mechanisms reflect the state of the science on chemical interactions, but there is considerable uncertainty in understanding the formation of organic nitrates in the oxidation of isoprene and terpene³⁴, in the nighttime production of organic nitrates³⁵, in radical budgets and nighttime radical reservoirs³⁶, and in aromatic oxidation chemistry^37,38. Given the continuing reductions in the mobile and industrial sources of NOx emissions and the transition of many urban areas from VOC sensitive to NOx sensitive photochemical regimes, the accurate simulation of VOC yields from the photochemical formation of organic and peroxy acyl nitrates sources becomes increasingly important. Research currently being conducted under EPA Science to Achieve Results (STAR) grants is expected to address some of these uncertainties and thereby advance the state of science.

• Diagnostic model performance evaluation. EPA guidance for modeling to demonstrate ozone model attainment demonstrations recommends using AQM in a relative sense³⁹. Accordingly, the model’s response to emission controls is quantified through relative response factors (RRF) that are multiplied by the observed baseline ozone design value to project whether the NAAQS will be attained. Using AQM in a relative sense is generally thought to be more reliable than using it in an absolute sense⁴⁰. However, even a relative use of model results can be problematic when AQM do not accurately simulate source contributions to ozone due to errors in the upstream meteorological model, errors in simulating regional transport and BG O3, and/or errors in modeled emissions inventories and local production of ozone. For example, AQM that overestimate regional BG O3 and underestimate local ozone production will provide inaccurate estimates of the relative ozone response to local emissions control measures. Research is needed in the use of diagnostic model evaluation techniques to evaluate the accuracy of modeled source contributions to ozone and to consider alternative ways to apply RRF at locations with high background ozone.

• Measurement needs to support regional model improvements. Advances in development of low-cost, compact, and accurate sensors should be leveraged to explore continuous aloft measurements of air pollution in the nocturnal residual layer by deploying such sensors on tall structures such as telecommunication towers. This would provide measurements relevant to quantifying the non-local contribution to surface-level air pollution⁴¹. Further, improved observational networks and additional field studies that collect continuous and co-located measurements (i.e., surface and vertical profiles) of speciated VOCs, NOx, and other key species relevant to O₃ would support more comprehensive model performance evaluations and aid in understanding the baseline of ozone and BG O3¹⁰.

Conclusion:

Accurately quantifying BG O3 is important for many air quality management applications. In this article, we have provided our perspective on research needs at both the regional and global scale to improve model-based estimates of BG O3. We hope this perspective can serve as road map as federal, state, and local agencies as well as academic partners develop research plans to address this challenge.