Air Quality in the Northern Colorado Front Range Metro Area: The Front Range Air Pollution & Photochemistry Experiment (FRAPPÉ), 2019 update

Air Quality in the Northern Colorado Front Range Metro Area: The Front Range Air Pollution and Photochemistry Éxperiment (FRAPPÉ), by Frank Flocke, Gabriele Pfister, James H. Crawford, Kenneth E. Pickering, Gordon Pierce, Daniel Bon, Patrick Reddy 22 December 2019 https://doi.org/10.1029/2019JD031197

Earlier (2016) North Front Range Oil and Gas Air Pollutant Emission and Dispersion Study https://www.colorado.gov/airquality/tech_doc_repository.aspx?action=open&file=CSU_NFR_Report_Final_20160908.pdf

We describe the resources used, the deployment strategy, and the outcomes of the Front Range Air Pollution and Photochemistry Éxperiment (FRAPPÉ) experiment, which took place in the summer of 2014 in the Front Range of Colorado. We provide a history of air quality of the region and the outcomes of previously conducted experiments, describe the atmospheric conditions encountered during the campaign, and summarize the scientific findings that the campaign produced, together with the Deriving Information on Surface Conditions from Column and Vertically Resolved Observations Relevant to Air Quality (DISCOVER‐AQ) intensive, simultaneously carried out by the National Aeronautics and Space Administration. The goal of FRAPPÉ was to measure emission tracers and photochemical tracers from the ground and by aircraft to be able to quantify the contributions of various emission sectors to the photochemical production of ozone in the Colorado Front Range. We found major contributions from the fossil fuel extraction sector as well as the transportation sector, with minor contributions from agriculture, energy generation, and industry. The meteorological conditions were also found to be critical in creating situations conducive to high ozone in the area.

1 Introduction

Approximately 80% of Colorado’s population of about 5.6 million lives in the Colorado Northern Front Range Metropolitan Area (NFRMA), encompassing the cities of Denver, Boulder, Longmont, Greeley, and Fort Collins. The NFRMA currently is in nonattainment of the 8‐hr ozone National Ambient Air Quality Standard (NAAQS). While summertime ozone levels in metropolitan regions located in the eastern United States have sharply declined over the last two decades (Cooper et al., 2014; Simon et al., 2015), summertime ozone in the NRFMA and similarly in other regions of the western United States has increased or leveled off (Strode et al., 2015). There is uncertainty whether this is due to increased precursor emissions, upward trends in background ozone, a shift in chemical regimes, a combination thereof, or other factors.

The study of air quality in the NFRMA presents some complex and unique challenges. First, flow patterns are highly complex due to the unique meteorological situation, driven in particular by mountain‐valley circulation effects, as well as the high elevation and varied terrain. This complex meteorology, coupled with the mix of diverse pollution sources (e.g., urban emissions, strong point sources such as large power plants and industrial complexes and airports, and area sources such as agricultural emissions, emissions from oil and gas development, biogenic emissions, and wildfires), in the NFRMA present challenges with respect to characterizing, modeling, and forecasting the transport and photochemical processes contributing to local air quality.

The Front Range Air Pollution and Photochemical Experiment (FRAPPÉ) was designed to identify and quantify the main drivers of summertime ozone in the NFRMA. The FRAPPÉ effort involved about 50 scientists from National Center for Atmospheric Research (NCAR) and the university community and included instrumentation and deployment of the National Science Foundation/NCAR (NSF/NCAR) C‐130 aircraft as well as ground site measurements. The FRAPPÉ field experiment was carried out jointly with the fourth deployment of the National Aeronautics and Space Administration (NASA) Deriving Information on Surface Conditions from COlumn and VERtically Resolved Observations Relevant to Air Quality (DISCOVER‐AQ) between 15 July and 20 August 2014 with most DISCOVER‐AQ platforms concluding their measurements on 10 August. DISCOVER‐AQ involved about 100 scientists, numerous ground site and remote sensing measurements, and the instrumentation and deployment of three aircraft. Additional contributions were provided by the State of Colorado Department of Public Health and Environment (CDPHE). NASA DISCOVER‐AQ’s primary objective was to understand sources, transport, and chemical transformations of air pollutants, particularly those that lead to ground‐level ozone and particulate matter formation and how they relate to column observations from satellites (Crawford & Pickering, 2014). Chemical and meteorological observations were conducted from four different aircraft and multiple mobile vans, ozone sondes, lidars, tethered balloons, and numerous operational and additional surface sites. The measurements were supplemented by comprehensive modeling activities. The flight patterns of all aircraft and the full data set including data from surface sites and mobile platforms are publicly available at http://www‐air.larc.nasa.gov/missions/discover‐aq/discover‐aq.html [DISCOVER‐AQ Science Team, 2016].

FRAPPÉ and DISCOVER‐AQ were designed to address the following questions:

1. What are the factors controlling NFRMA surface ozone, and are current emission controls sufficient to reduce ozone levels below standards?
2. What are the relative contributions of local mountain‐valley recirculation patterns and long‐range transport to buildup of photochemical oxidants and particulates during smog episodes in the NFRMA in the summer?
3. What are the relative contributions of the diverse local sources of pollution to air quality degradation and photochemical oxidant formation in the NFRMA?
4. To what degree does pollution from both NFRMA sources and long‐range transport contribute to photochemical smog/ozone pollution, visibility degradation, and nitrate deposition in Rocky Mountain National Park and other designated wilderness areas to the west of NFRMA?
5. What is the impact of ozone precursor emissions from oil and gas exploration and production on the photochemical regime and the ozone production efficiency?
6. How is pollution from NFRMA affecting background mixing ratios of oxidants, specifically ozone and NO_x in surrounding rural areas? Are commonly used air quality models able to adequately represent the urban‐rural interface?
7. Is the NFRMA like many other urban areas in the United States shifting to a more NO_x‐limited chemical regime, and what are the implications for future air quality predictions and policy decisions?
8. How well do remotely sensed observations of pollutants such as NO₂ and aerosols relate to surface measurements in the highly variable terrain and land cover of the NFRMA?

To date, most of these questions have been addressed at least in parts by peer‐reviewed studies or reports. This paper intends to provide an overview of the FRAPPÉ and DISCOVER‐AQ Colorado campaign and the results that have been published thus far specifically as they relate to the overall objective of FRAPPÉ, which was to characterize the drivers of summertime ozone pollution in the NFRMA. A list of to‐date published studies can be found at https://www2.acom.ucar.edu/frappe/publications.

2 Front Range Meteorology and Emissions

The NFRMA is located between the Rocky Mountains to the west and the High Plains to the east. The greater Denver metropolitan area, one of the top 20 mega‐urban regions in the United States, is the largest urban area in the region with over 2 million inhabitants, followed by Colorado Springs, and Ft. Collins with about 250,000 inhabitants each, and Boulder, a city of about 100,000 people. Smaller communities embedded between the larger cities fill in the remaining space on the west side of the NFRMA, and in recent years, considerable urban expansion has taken place towards the eastern plains. The area between the mountains to the west and Interstate Highway 25 to the east has experienced steady growth since the mid‐1990s. The metropolitan area is also one of the nation’s fastest‐growing urban centers with Denver’s population expected to increase by nearly 50% by 2030 (e.g., see http://www.metrodenver.org/do‐business/demographics/population/).

The region also experienced large increases in oil and gas drilling activities with the number of oil and gas wells having nearly doubled in Weld County, to the northeast of Denver, between January 2008 and July 2015 to over 27,000 active well sites [Colorado Oil and Gas Conservation Commission (COGCC), 2016]. In addition, the area is also home to abundant livestock. There are roughly 2.5 million head of cattle in Colorado, of these about 1 million cattle in feedlots with the majority of these located NE of the NFRMA. Weld County in particular is the top agricultural producing county in Colorado (https://www.nass.usda.gov/Statistics_by_State/Colorado/Publications/Annual_Statistical_Bulletin/index.php). This results in a mix of different emission sources: urban, industrial, oil and gas exploration, and agriculture.

The NFRMA is located at an elevation of roughly 1,500–1,800 m, on the east slope of the Central Rocky Mountains. To the west of the NFRMA, the terrain becomes mountainous, mostly wooded, with scattered smaller communities up to elevations of about 3,000 m, and then transitions into the mostly uninhabited alpine region along the Continental Divide, reaching up to 4,300‐m altitude. Several major river canyons extend from the high terrain down into the NFRMA.

In the summer months, particularly during weak synoptic conditions, the local meteorology is mainly controlled by thermally driven, terrain‐induced, diurnal flow patterns (Johnson & Toth, 1982; Toth & Johnson, 1985, and references therein). This has unique consequences for the transport, mixing, and photochemical processing of local emissions (Haagenson, 1979; Greenland, 1980; Doran, 1996; Baumann et al., 1997; Olson et al., 1997). Briefly, during the night, radiative cooling of the land surface causes downslope drainage flows from the Continental Divide in the West, far into the Eastern Plains. The drainage flows are most pronounced in the canyons and river valleys (Doran, 1991; Johnson & Toth, 1982). During the morning hours, solar heating of the ground and mountain slopes causes upslope flow to develop, starting a few hours after sunrise in the foothills and slowly extending out into the eastern plains about 3 hr later. On days with strong upslope, the flow can reach the top of the Continental Divide by mid‐afternoon. In the late afternoon, flow reversal starts at the top of the Continental Divide and a convergence line forms and moves down and east during the evening hours. Finally, downslope winds take over the entire region again around midnight (Johnson & Toth, 1982, and references therein).

This meteorology can cause pooling of nighttime emissions (urban, agricultural, and oil and gas development) in the lower elevations, particularly the Platte River Valley. After sunrise, these emissions become subject to photochemical processing and transport to the west (back into the urban area) after onset of the upslope flows. Fresh emissions from the NFRMA, mixed with these partially processed air masses, are then transported into the mountains during the day. Once lofted above the Continental Divide, air can be entrained into the free troposphere, which is typically dominated by large‐scale westerly flow. This way, at night, processed air masses from the previous day can sometimes be transported back into the NFRMA with the downslope winds developing in the evening. In addition, especially during weak synoptic conditions, air masses entrained into the westerly flow over the Continental Divide can be mixed down from the free troposphere into the expanding boundary layer during the next morning. Both mechanisms act as an enhancement to photochemical processing of current emissions, causing the concentrations of ozone and other photooxidants to build up over the course of several days (Reddy & Pfister, 2016). In addition, upslope transport of photochemically active air masses impact ecologically sensitive and/or recreationally important areas west of the NFRMA, such as Rocky Mountain National Park (e.g., https://www.colorado.gov/pacific/cdphe/rocky‐mountain‐national‐park‐initiative) and numerous nearby Wilderness areas. These impacts include reduced visibility, enhanced deposition of nitrogen (Benedict et al., 2018; Darrouzet‐Nardi et al., 2012), and elevated ozone concentrations (Benedict et al., 2019; Brodin et al., 2010; Brodin et al., 2011). In contrast, stronger frontal passages can induce outflow of NFRMA pollution to the east, impacting the agricultural areas located downwind in the central Great Plains. Such events, which in the summer occur about every 5–10 days over the Central Rockies (e.g., Shafer & Steenburgh, 2008; Wiesmueller, 1982), are the major driver for cleaning out accumulated pollution from the NFRMA. Afternoon thunderstorms, which are common in summer but vary in frequency dependent on the synoptic weather situation, can dilute pollution by injecting upper tropospheric air into the PBL and limit the photochemical ozone production and accumulation of pollution because of the buildup of clouds and stronger winds. However, they generally have a much more local‐scale impact than the frontal passages mentioned above.

Reddy and Pfister (2016) showed that July daily 8‐hr average maximum (MDA8) ozone in Colorado is strongly correlated with 500‐hPa heights and strongly anticorrelated with 700‐hPa zonal winds. Compared with other years from 1995 through 2014, July–August of 2014 was in the 32 percentile for 500‐hPa heights and the 63 percentile for zonal winds. Consequently, compared with July–August from 1995 through 2014, the meteorology of the study period was on average less conducive to high‐ozone concentrations, less favorable for the kind of upper level high‐pressure systems associated with terrain‐induced circulations, and more conducive to midlevel westerlies that can lead to outflow of NFRMA ozone to the eastern plains. The authors were unable to find a long‐term climatology of terrain‐induced flows within the NFRMA, but the statistics cited above suggest that thermally driven upslope deep into the mountains and related flow regimes might have occurred less often than normal during the study period. This is in line with average summer ozone MDA8 values in 2014 being around 7 ppb lower than in previous years (see https://www.colorado.gov/airquality/tech_doc_repository.aspx?action=open&file=2014AnnualDataReport.pdf), and the average monthly temperatures in the Boulder Denver Forecast area in July and August were about 0.5 and 1°C lower than average, respectively (see https://www.esrl.noaa.gov/psd/boulder/Boulder.mm.html).

The NFRMA has been subject to several past field experiments, which focused on air quality in the Front Range urban area, a subset of which we briefly review below. Early studies were launched primarily as a result of Denver’s infamous “brown cloud,” which put Denver into the media spotlight as one of the places with the nation’s unhealthiest air quality during the 1960s and 1970s. Riehl and Crow (1962) sampled particulate matter in Denver and found average values close to 300 μg/m³ at some locations in the winter of 1961. The 1973 Denver Air Pollution Study (Russell, 1977) and the Denver Haze Study conducted in the winter of 1978–1979 (Groblicki et al., 1981; Wolff et al., 1981) also focused on episodes of winter pollution in Denver. In addition to large aerosol loadings, Denver would exceed the Air Quality standard for CO as often as 200 days per year. In 1987–1988 the Metro Denver Brown Cloud study provided objective apportionment to the observed brown cloud pollution over Denver (Sloane et al., 1991). Aerosols were mostly attributed to local ammonia emissions from livestock combining with SO₂, mostly from energy generation units (EGUs) and nitrate, mostly from traffic emissions, as well as from automobile emissions, and from road sanding and residential wood burning in the winter. During 1996–1997, measurements of aerosol composition and inorganic aerosol precursors were carried out in winter and summer months at several urban and rural sites during the Northern Front Range Air Quality Study (NFRAQS), summarized in Watson et al. (1998). The more recent studies showed that progress in reduction of automobile emissions, reduction of SO₂ from EGUs, the reduction of road sanding, and the use of oxygenated fuels all helped greatly in reducing the severity of Denver air pollution episodes, generally with respect to particulates (Lyons, 1990; Neff, 1997; Reddy et al., 1995; Rowe et al., 1990).

In summer 2012, the Summer Ozone Near Natural gas Emissions (SONNE) field campaign was conducted at the Boulder Atmospheric Observatory (BAO) Tower in Erie, CO, collecting continuous in situ measurements of NO_x and a suite of volatile organic carbons (VOCs; Gilman et al., 2010). These data have been used by McDuffie et al. (2016) together with data from the FRAPPÉ campaign to highlight the importance of oil and natural gas (O&NG) emissions relative to biogenic emissions on ozone production in the NFRMA. In late winter 2011, the Nitrogen, Aerosol Composition, and Halogens on a Tall Tower (NACHTT) field experiment took place during 3 weeks at the BAO tower with the objective to characterize air pollutants and oxidant formation during the winter season (Brown et al., 2013). It included near‐surface measurements of aerosols, soluble trace gases, and volatile organic compounds and also vertical profiles of some tracers. They found that nitrate was the dominant component of aerosol mass at BAO, but there were also significant contributions from both organics and sulfate. Large mixing ratios of light alkanes were attributable to emissions from oil and gas activities and were estimated to be responsible for up to 55% of the OH reactivity with VOCs, indicating that these VOC emissions are likely to be a large contributing factor to local ozone formation during summertime (Gilman et al., 2013; Swarthout et al., 2013).

Neither SONNE nor NACHTT, however, were designed to take a comprehensive look at all the different emission sources in the NFRMA, their chemical processing, and mixing as well as transport of pollutants into and out of the NFRMA. This was the objective of FRAPPÉ and DISCOVER‐AQ, with a specific focus on surface ozone formation processes.

3 Measurement Resources and Sampling Strategy

A large suite of airborne and ground‐based measurements of photochemical tracers and meteorological parameters were performed during the FRAPPÉ campaign. To the extent possible, ground resources were distributed between the sites and both aircraft and ground measurements were closely coordinated to maximize scientific return and minimize duplication.

3.1 Airborne Measurements

Four aircraft participated in the FRAPPÉ and DISCOVER‐AQ intensives. The airborne component of the FRAPPÉ experiment used the NSF/NCAR C‐130 aircraft equipped with a comprehensive photochemistry payload. Instruments are listed in Table 1. Two aircraft were operated and supported by NASA for DISCOVER‐AQ (NASA P‐3 and NASA B‐200). The NASA P‐3 carried a payload focused on basic in situ photochemical tracer and aerosol measurements for comparison with remote sensing data, as well as other supporting measurements (see Table 2 for details). The NASA B‐200 carried a High Spectral Resolution light detection and ranging (LIDAR) system measuring cloud and aerosol extinction profiles, as well as the Geo‐CAPE Airborne Simulator (GCAS), measuring slant column amounts of ozone, NO₂, and formaldehyde. NASA also supported a third aircraft (the Falcon), which carried the Geostationary Trace gas and Aerosol Sensor Optimization (GeoTASO) instrument. GeoTASO was a testbed instrument for the TEMPO and GEMS satellite missions, and measured slant columns of a number of photochemical and emission tracers including ozone, NO₂, formaldehyde, and SO₂.Table 1. Instruments and Measured Species on the NSF/NCAR C‐130 Aircraft

Measurement	Instrument	PI
Photolysis rates	HARP	Sam Hall, ACD, NCAR
O3, NO, NO2	Chemiluminescence	Andy Weinheimer, ACD, NCAR
Organic nitrates	TDLIF	Ron Cohen, UC‐Berkeley
PANs	PAN CIMS (PANs)	Frank Flocke, ACD, NCAR
SO2,HNO3	GTCIMS	Greg Huey, Georgia Tech
CO	Aerolaser AL5002	Teresa Campos, ACD, NCAR
CO2, CH4, H2O	Picarro	Teresa Campos, ACD, NCAR
CH2O, C2H6	CAMS	Alan Fried, CU
VOC	TOGA	Eric Apel, ACD, NCAR
PTRMS	Lisa Kaser, ACD, NCAR
AWAS	Don Blake, UC‐Irvine
CH4 isotopes	AWAS	A. Townsend‐Small, U.Cincinnati
OH, H2SO4, crigee intermed., HO2, RO2	HOx CIMS	Lee Mauldin, Chris Cantrell, CU
H2O2, ROOH	PCIMS	Brian Heikes, URI
Dan O’Sullivan, USNA
NH3	TILDAS	John Nowak, NOAA
Aerosol composition	AMS, CAPS	Roya Bahreini, UC‐Riverside
Aerosol composition	PILS	Delphine Farmer, CSU
Jim Smith, ACD, NCAR
Particle size distribution (8‐500 nm)	SMPS	Jim Smith, ACD, NCAR
Particle distributions	CN, PCASP	Mike Reeves, RAF, NCAR
Forward‐facing video	Digital camera	Stuart Beaton, Janine Aquino, RAF, NCAR
Aircraft state parameters, T, P, WD, and WS		Chris Webster, RAF, NCAR

• Abbreviations: AMS: aerosol mass spectrometer; AWAS: Advanced Whole Air Sampler; CAMS: Compact Atmospheric Multispecies Spectrometer; CAPS: CAvity Phase‐shift Spectrometer; CIMS: Chemical Ionization Mass Sectrometer; CN: condensation nuclei counter; GTCIMS: Georgia Tech CIMS; HARP: Hiaper Airborne Radiation Package; NCAR: National Center for Atmospheric Research; NSF: National Science Foundation; PCASP: Passive Cavity Aerosol Spectrometer Probe; PCIMS: peroxide CIMS; PILS: particle into liquid sampler; PTRMS: proton transfer mass spectrometer; SMPS: scanning mobility particle sizer; TILDAS: tunable infrared laser differential absorption spectrometer; TDLIF: thermal dissociation laser‐induced fluorescence; TOGA: trace organic gas analyzer.

Table 2. Instruments and Measured Species on the NASA P‐3 Aircraft

Measurement	Instrument	PI
Aerosol number, size distribution, composition, and physical properties	LARGE	Bruce Anderson, NASA
O3, NO, NO2, NOy	Chemiluminescence	Andy Weinheimer, ACD, NCAR
NO2, sum of peroxy nitrates, sum of alkyl nitrates, HNO3	TDLIF	Ron Cohen, UC‐Berkeley
CO, CH4, N2O	DACOM	Glenn Diskin, NASA
H2O	DLH	Glenn Diskin, NASA
CH2O	CAMS	Alan Fried, CU
VOC	PTR‐ToF‐MS	Armin Wisthaler, U. Innsbruck
Ethane	Aerodyne TILDAS	Tara Yakowitch, Aerodyne Res.
CO2	AVOCET	Melissa Yang, NASA
Aircraft state parameters, T, P, WD, WS		John Barrick, NASA

• Abbreviations: AVOCET: atmospheric vertical observations of CO₂ in the Earth’s troposphere; DACOM: differential absorption CO measurement; CAMS: Compact Atmospheric Multispecies Spectrometer; DLH: diode laser hygrometer; LARGE: Langley Aerosol Research Group Experiment; NASA: National Aeronautics and Space Administration; NCAR: National Center for Atmospheric Research; TILDAS: tunable infrared laser differential absorption spectrometer; TDLIF: thermal dissociation laser‐induced fluorescence; ToF: time of flight.

In addition to the GeoTaso instrument the campaigns also served as a testbed for the advancement of in situ aircraft instrumentation. This included a fast, infrared absorption spectrometer for the measurement of ethane with a 1‐s time resolution (Richter et al., 2015), a novel approach of measuring background water soluble aerosol composition (Fukami et al., 2016), and a Chemical Ionization Mass Sectrometer (CIMS) technique for the fast measurement of atmospheric peroxides (Treadaway et al., 2018).

The sampling strategies of the NASA aircraft were different from those of the NSF/NCAR C‐130 in that they were optimized for examining relationships between remotely sensed columns, in situ profiles, and surface observations. The NASA P‐3 thus focused on relatively unbiased, repeated vertical sampling of the planetary boundary layer above the Northern Front Range urban area during cloud‐free conditions, with sounding spirals located over the six ground super‐sites, which were located in Fort Collins, Greeley, downtown Denver, Littleton, Golden, and the BAO Tower in Erie. The spiral locations/ground super‐sites were chosen for optimal coverage of the expected emissions and transport patterns while providing adequate infrastructure for the ground based instrumentation (see also section 3.2). The B‐200 and Falcon aircraft mostly flew along the same tracks as the NASA P‐3, but at higher altitudes, to optimize the vertical coverage of their remote sensing payloads. Flight tracks and selected ground site locations are shown in Figure 1.

The NSF/NCAR C‐130 flights covered three major objectives: local, terrain‐driven (upslope) air mass chemistry and transport, local emissions, and larger scale transport, including inflow of polluted air into the NFRMA, as well as outflow into the eastern plains and injection of polluted air into the free troposphere. Emission flights typically involved lawnmower patterns over the areas of interest at altitudes of about 300‐m above the surface. Upslope flights were designed to probe the entire boundary layer and free troposphere above by flying stacked N‐S transects typically at three different altitudes and then shifting the longitude of the N‐S flight legs westward in a quasi‐Lagrangian fashion as the day progressed. NSF/NCAR C‐130 flight tracks are shown in Figure 2.

3.2 Ground Resources

In situ observations of chemical and meteorological parameters were collected at a number of surface sites across the NFRMA including numerous operational CDPHE ozone monitoring sites. Some of the CDPHE sites were enhanced by additional instrumentation, and a number of surface sites were added during the campaign. Comprehensive chemical and meteorological measurements were made at BAO in Erie and Platteville and ozone and VOC canister sampling was performed at several sites along a transect from the NFRMA to the mountains. Ozone measurements are also available from a number of high‐altitude sites: Two sites operated by the National Park Service are located in Rocky Mountain National Park (Benedict et al., 2019), with one of these (Longs Peak) also collecting wind and VOC measurements; three sites further South at Niwot Ridge are operated by the National Oceanic and Atmospheric Administration (NOAA); and two sites in the Foothills to the West of Denver were added by CDPHE. Ozone sondes were launched from Platteville, Fort Collins, and Golden, and ozone measurements were also collected from tethered balloon at varying sites across the NFRMA. Surface ozone measurements and selected NMHC canister samples were taken by the Institute of Arctic and Alpine Research (INSTAAR) group along an east‐west gradient spanning Boulder County from the eastern plains (1,660‐m elevation) and Niwot Ridge, a high‐altitude site near the Indian Peaks Wilderness, at 3,530‐m elevation (Brodin et al., 2011). NASA Pandora spectrometers were deployed at 14 sites, providing total column measurements of O₃ and NO₂. Figure 3 shows the tracks of the mobile ground resources, VOC canister sampling locations, and other ground measurement sites active during FRAPPÉ. Table 3 provides an overview of the measurements available from selected ground sites. A full list of measurements is available on the FRAPPÉ web site (https://www2.acom.ucar.edu/sites/default/files/frappe/FRAPPE‐DISCOVER‐AQAllMeasurements.pdf).

Table 3. Measurements Available From Selected Ground Sites During FRAPPÉ

Name	Latitude	Longitude	Elev.(m asl)	Measurements
BAO Towera,b,c	40.050	−105.004	1,580	O3, NOxy, CO, CO2, CH4, SO2, VOCd, N2O, NH3, PANs, OH reactivity, aerosol param., LIDAR, wind profiler, N2O5, org. acids, RONO2, met, ceilometer
Chatfield Parka,b,c	39.535	−105.070	1,676	O3, NO2, met, VOC, CH4
Denver‐LaCasaa,b,c	39.779	−105.005	1,601	O3, NO2, NOy, SO2, VOC, PM
Fort Collins‐Westa,b,c	40.593	−105.141	1,572	O3, NO2, met, VOC, CH4, TOLnet O3, MPL
NREL‐Goldena,b,c	39.744	−105.178	1,833	O3, NOxy, CO, CO2, SO2, CH4, VOC, tether sonde, flux tower, nephelometer, LIDAR, wind profiler, ceilometer, TOLnet O3
Plattevillea,b,c	40.183	−104.726	1,523	O3, NOxy, CO, CO2, SO2, CH4, VOCd, wind profiler, radiation
Aurora East/DUc	39.639	−104.569	1,802	O3, met
CAMP (Denver)	39.751	−104.988	1,594	O3, NO2, NOy, SO2, CO, VOC, PM
I‐25 Denverb	39.732	−105.015	1,587	O3, NO2
Niwot Ridge C1b,c	40.032	−105.533	2,886	O3, met
Rocky Flats ‐ N b,c	39.913	−105.189	1,803	O3, NO2, met
Squaw Mtn b,c	39.680	−105.493	3,492	O3
Table Mountain b,c	40.125	−105.237	1,687	O3
Welch	39.639	−105.140	1,743	O3, met
Weld Co. Tower b,c	40.386	−104.737	1,484	O3, NO2, met
Aspen Park‐RTD	39.542	−105.298	2,477	O3, met
Fort Collins‐CSU	40.577	−105.079	1,525	O3, met
South Boulder Creek	39.957	−105.238	1,671	O3
Welby	39.838	−104.950	1,556	O3, NOxy, met, SO2, CO
Fritz Peak	39.908	−105.492	2,681	O3
Squaw Mtn b,c	39.680	−105.493	3,492	O3
Mines Peak	39.794	−105.764	3,797	O3
Daniels Park	39.482	−104.931	1,978	O3
Golden Gate Fire	39.777	−105.368	2,452	O3
North Fork Fire St.	39.362	−105.245	2,323	O3
Jackson Res. SP	40.388	−104.095	1,361	O3
NCAR Mesa Lab	39.978	−105.275	1,862	O3
Briggsdale	40.650	−104.335	1,481	O3
Pawnee Buttes	40.808	−103.989	1,600	O3
INSTAAR Boulder	40.013	−105.253	1,610	O3, VOC canisters
Dawson School	40.064	−105.110	1.562	O3, VOC canisters
Betasso	40.017	−105.358	1,981	O3, VOC canisters
Sugar Loaf	40.019	−105.405	2,396	O3, VOC canisters
Coughlin Meadows	40.004	−105.479	2,530	O3, VOC canisters
Niwot Ridge C1 b,c	40.032	−105.533	2,880	O3
Niwot Ridge Soddie	40.048	−105.571	3,359	O3, VOC canisters
Niwot Ridge Saddle	40.054	−105.589	3,527	O3
Rocky Mountain NPd	40.278	−105.545	2,743	O3, NOxy, SO2, met, VOC, NH3, PANs, HNO3, H2O, aerosol composition
Rocky Mountain NP High Alt.	40.388	−105.688	3,545	O3, met
Foothills Lab	40.038	−105.243	1,615	FTIR integrated column

• Note. The first six sites are the spiral locations for the NASA P3‐B aircraft. Notes: (a) P3 spiral site, (b) NASA Pandora instrument location, (c) NASA AeroNET instrument location, and (d) in situ VOC; others are VOC canister samples during spiral overflights.
• Abbreviations: asl: above sea level; LIDAR: light detection and ranging; met: basic meteorological parameters; NASA: National Aeronautics and Space Administration; PANs: peroxy acyl nitrates; PM: particulate measurements; RONO2: alkyl nitrates; VOC: volatile organic compounds.

In addition to fixed surface sites, six mobile vans were operated during the campaign with the collected list of measurements listed in Table 4.Table 4. Mobile Unit Measurements Deployed During FRAPPÉ

Measurement	Platform	PI
O3, NO2, NO, CO, CO2, CH4, N2O, NH3	NOAA CSD van	Tom Ryerson, NOAA
O3, CO, CO2, CH4, VOCs (flask), CH4 isotopes	NOAA GMD van	Gabrielle Petron, NOAA
column NO2, O3, CO, VOCs	VIS/IR spectrometer trailer	Rainer Volkamer, CU
O3, NO2, NO, SO2, CO, CO2, CH4, VOCs (PTRMS), fast C2H6, N2O, NH3, H2O, AMS	Aerodyne van	Cody Floerchinger, Scott Herndon, Aerodyne
CO2, CH4, CH4 isotopes	INSTAARa van	Bruce Vaughn, CU
CO, CO2, CH4, N2O, NH3, H2O	Princeton van	Mark Zondlo, Princeton U.

• Abbreviations: AMS: aerosol mass spectrometer; FRAPPÉ: Front Range Air Pollution and Photochemistry Éxperiment; NOAA: National Oceanic and Atmospheric Administration; VOC: volatile organic carbon; VIS/IR: visible/infrared.
• ^a Institute for Artic and Alpine Research, University of Colorado Boulder.

Similar to the aircraft payload, a number of instrument innovations were tested and implemented during the ground component of FRAPPÉ. Kille et al. (2017) employed a novel approach to measuring fluxes of NO₂, ethane, and ammonia from individual area and point sources such as animal feed lots and oil and gas processing facilities using a solar occultation measurement technique mounted on a mobile platform. This technique relies on a mobile solar tracking device described in Baidar et al. (2016). Duvall et al. (2016) describe an effort to assess the usefulness of low‐cost sensors for ozone and NO₂ for scientific studies. FRAPPÉ measurements were also used to improve the accuracy and applicability of TOLNet (Tropospheric Ozone LIDAR Network, https://www‐air.larc.nasa.gov/missions/TOLNet/) measurements for large‐scale air quality studies (Wang et al., 2017).

The FRAPPÉ and DISCOVER‐AQ data sets are publicly available on NASA’s Airborne Science Data web repository (https://www‐air.larc.nasa.gov/missions/discover‐aq/discover‐aq.html). Quality controlled data from each instrument as well as merges on several different time bases are available for download. Also available are video files recorded by the aircraft as well as other ancillary data. The one‐dimensional data are archived in the commonly used ICARTT format (https://earthdata.nasa.gov/esdis/eso/standards‐and‐references/icartt‐file‐format). Also available for download are selected model results as indicated in Table 5 below.Table 5. Nonoperational Forecast Products Available From the NCAR/EOL FRAPPÉ Field Catalog

Modeling System	Information	Contact
NOAA/CIMSS RAQMS*	Global full chemistry with assimilation of AOD, total and stratospheric ozone (1°×1°)	Brad Pierce, NOAA/NESDIS
NASA GEOS‐5	Global with assimilation of total ozone. aerosols and CO tracers (0.3125°×0.25°)	Ken Pickering, University of MarylandArlindo da Silva, NASA/GSFC
NOAA/ARL NAQFC‐β	Regional full chemistry (4km); based on WRF/CMAQ	Ken Pickering, University of MarylandPius Lee, NOAA/ARL
MOZART‐4	Global with full chemistry (2°x2°) and chemical tracers (0.5°x0.5°)	Louisa Emmons, NCAR
NOAA/ESRL RAPChem	Regional full chemistry with assimilation of meteorology (13 km); based on WRF‐Chem	Steven Peckham, NOAA ESRL (now U.S. Army Corps of Engineers)
NCAR FLEXPART*	Forward/backward trajectories for selected sources and locations	Rajesh Kumar, NCARChristoph Knote, NCAR (now University of Munich)
NCAR/RAL WRF STEP	Regional with assimilation of meteorology and synthetic tracers (3 km)	Gabriele Pfister, NCAR
NCAR/MMM WRF*	Regional with Synthetic tracers (3 km)	Gabriele Pfister, NCAR

• Note. Raw model data available from the NASA Data Archive are indicated by an asterisk.
• Abbreviations: AOD: aerosol optical depth; ARL: Air Resources Laboratory; CMAQ: Community Multiscale Air Quality Modeling System; FLEXPART: FLEXible PARTicle dispersion model; GEOS‐5: Goddard Earth Observing System Model, Version 5; MOZART‐4: Model for OZone And Related Chemical Tracers version 4; NAQFC‐β: National Air Quality Forecasting Capability β version; NASA: National Aeronautics and Space Administration; NCAR: National Center for Atmospheric Research; NESDIS: National Environmental Satellite, Data, and Information Service; RAP‐Chem: Rapid Refresh Model with Chemistry; RAQMS: Real‐time Air Quality Modeling System; WRF: Weather Research and Forecasting model; WRF ARW STEP: WRF run under NCAR’s ShortTerm Explicit Prediction (STEP) program

3.3 Modeling Resources

For flight planning a combination of operational forecast products were used including weather forecast products from National Centers for Environmental Prediction (NCEP) Global Forecasting System (GFS) and NCEP North American Mesoscale Forecast System (NAM) together with products created specifically for the campaign (Table 5). A range of different forecast products is needed to optimally cover all the relevant aspects of flight planning and deployment of the ground‐based resources (e.g., mobile labs and tethered balloons) as well as to consider uncertainties associated with different forecast products. The products employed during FRAPPE and DISCOVER‐AQ encompass global and regional chemical transport models with full chemistry and inert tracers as well as simulations with the Lagrangian transport and dispersion model FLEXPART. Graphical output from these products is available from the NCAR/EOL FRAPPÉ Field Catalog (http://catalog.eol.ucar.edu/frappe/). The forecasts provide information on expected weather patterns including cloudiness, temperature and winds, air pollution levels, stratospheric intrusions, wildfire plumes, and other factors. Many of the forecast products have been evaluated and have been and remain an essential piece of information in post‐campaign analysis (e.g., Battye et al., 2016; Pfister et al., 2017; Sullivan et al., 2015; Sullivan et al., 2016; Vu et al., 2016). Additional model simulations have also been conducted after the campaign in support of the analysis of observations (e.g., Abdi‐Oskouei et al., 2018; Bahreini et al., 2018; Kaser et al., 2016; Kille et al., 2017; Pfister & Flocke, 2017).

4 Results

4.1 Meteorology and Transport

Accurate information about meteorology and transport is crucial for understanding air quality. The measurements made during FRAPPÉ and DISCOVER‐AQ provide a wealth of information to characterize transport patterns and evaluate chemical transport models.

A careful evaluation of modeled winds has been conducted by Pfister et al. (2017) and Pfister and Flocke (2017). Both studies used the WRF model to simulate the entire campaign period, but the simulations were conducted by different teams and for different purposes; the product used in Pfister et al. (2017) was designed to forecast pollution transport during the campaign compared to the study by Pfister and Flocke (2017), which objective was to conduct post‐campaign analysis for emission evaluation and ozone source contribution analysis (see the supporting information for a summary of this study). As a result, there are significant differences between these simulations in regard to the spatial and vertical resolutions, meteorological driving fields, physics options, nudging settings, and other configurations. As is a general problem for limited resolution models, both of these studies conclude that the complex topography poses large challenges to the model, but comparison to wind observations from surface sites, ozone sonde launches and aircraft shows that the WRF model overall simulates quite well the general flow patterns, which are dominated by mountain‐valley flows. The simulations predict a stronger upslope component (i.e., higher wind speeds) than what was measured at the surface sites. A likely reason for this is that the model does not represent the subgrid local topographic features and thus better represents the regional picture. The model employed a 4‐km spatial grid resolution, but ideally grid spacing of 1 km or less might be needed to resolve the narrow canyons stretching from the Front Range into the mountains. Also, surface sites might often be impacted by nearby buildings, canyons, or other topographical features and not necessarily provide representative observations for comparison to a 3‐ to 4‐km model grid size.

Pfister et al. (2017) analyze the typical summertime transport pattern in the Front Range using the tracer capability in WRF with chemically inert tracers representative of emissions in Colorado from O&NG activities, area (e.g., industrial) and mobile emissions, and agricultural emissions. They show that during nighttime, the urban emission tracer reveals a pooling of emissions from the Denver Metro Area to the north‐east toward the region of strong oil and gas sources. Flow reversal in the morning transports aged urban emissions together with fresh oil and gas emissions toward the Metro area. As is very common in the summer (Toth & Johnson, 1985), on most days during the study, upslope flows developed around mid‐morning starting at near‐Foothills sites and then spread East and West. These have the potential of carrying NFRMA pollution to the remote mountain areas. Six NSF/NCAR C‐130 flights were dedicated to following the upslope flows. During most of these flights, at least a partial separation between air masses originating further north (oil and gas related emissions) and further south (Denver area urban and industrial emissions) was observed; some mixing can occur, especially if the upslope winds have a northeasterly component.

An underestimate in the development of clouds and related thunderstorm activity might also be contributing to a stronger and more developed upslope component in the WRF model. During FRAPPÉ, thunderstorm activity frequently interrupted flow patterns. The comparison of WRF modeled downwelling shortwave radiation to measurements at three surface sites in the NFRMA showed an average of ~20% higher daily integrated shortwave radiation compared to measurements (Pfister & Flocke, 2017). A similar bias in solar radiation was also found by forecast products that were used during the campaign for flight planning (Pfister et al., 2017). This demonstrates a clear underprediction in cloudiness and suggests also an underprediction of storms, which frequently occurred in the afternoon during the campaign.

The campaign data also allow evaluation of the planetary boundary layer (PBL) heights (PBLH), which were retrieved from micropulse lidar data at Fort Collins West, Platteville, and NREL Golden (Hegarty et al., 2018; Pfister et al., 2017; Pfister & Flocke, 2017). This is not a true direct comparison but gives an indication of the model’s ability to reproduce the typical PBLH and its day‐to‐day variability. Despite the different definitions of PBLH, the model simulations conducted by Pfister and Flocke (2017) and Pfister et al. (2017) overall represent the day‐to‐day variability fairly well but have varying difficulties on individual days. For example, both simulations overestimate the growth and height of the PBL on 22 July 2014. This was a day with high and widespread ozone concentrations, which was caused in part by a slow growing PBL compared to more typical days. The temporal and vertical evolution of ozone concentrations on this day is reflected in Figure 4, which shows ozone measurement from all P‐3 spirals over the Fort Collins and Golden ground sites with profile measurements for 22 July highlighted.

The studies referenced in the paragraph above also characterized the vertical variability of trace gases within the PBL and their findings challenge the general assumption of a well‐mixed daytime boundary layer. This is corroborated by Tevlin et al. (2017), who analyzed ammonia measurements between the surface and 280 m above ground level from a moveable carriage at the BAO tower and found large changes in NH₃ from 0 to 280 m in height, and by Kaser et al. (2016) from analysis of aircraft profiles of ozone. Similar boundary layer gradients have also been observed during the 2011 DISCOVER‐AQ deployment in the Washington‐Baltimore urban corridor (Zhang et al., 2016). Due to the nonlinearities in many atmospheric reactions, this causes uncertainties in simulated ozone production rates because models generally do not resolve these inhomogeneities.

The large impact of transport and meteorology on air pollutants is demonstrated by Abdi‐Oskouei et al. (2018), who used ethane as the indicator of O&NG emissions and explored the sensitivity of ethane concentrations to different physical parameterizations and simulation setups in WRF‐Chem. From a range of sensitivity analysis, they find a large variability in modeled ambient ethane mixing ratios and also in the comparison with observed ethane mixing ratios for different physics parameterizations in WRF‐Chem, which highlights the important role of model configurations (i.e., meteorology and dynamics) on model performance and ultimately the assessment of emissions.

Upslope flows were prevalent on all high‐ozone days during the campaign with 22 July and 12 August being the days when the surface monitors in Rocky Mountain National Park experienced their highest hourly ozone mixing ratios (82 and 80 ppb at Trail Ridge Road and 76 and 78 ppb at Longs Peak, respectively). Sullivan et al. (2016) characterize the polluted air mass and assess transport throughout the 22 July 2014 upslope event using measurements from O₃ and wind profilers, O₃‐sondes, aircraft, and surface‐monitoring sites. Observations indicate that under relatively weak synoptic conditions thermally driven upslope flow was established throughout the Colorado Front Range during the pollution episode. As the thermally driven flow persisted throughout the day, O₃ concentrations increased and affected high‐elevation Rocky Mountain sites. Their analysis of field observations, coupled with modeling analyses, demonstrates a westerly return flow of polluted air aloft, indicating that the mountain‐plain solenoid circulation (Reddy & Pfister, 2016) was established and exacerbated surface ozone pollution within the Front Range. The 12th of August 2014 was another day with strong upslope flows and high‐ozone concentrations transporting pollution from the Front Range to the mountains, which was successfully followed with the NSF/NCAR C‐130 (Pfister et al., 2017). Figure 5 shows a vertical cross section of ozone concentrations measured from the aircraft. This graph demonstrates the upslope flow of ozone from the Front Range to the continental Divide. The measurements also demonstrate the occurrence of the hypothesized spill‐over of Front Range pollution into the valleys to the West of the Divide.

Another meteorological feature that can lead to high‐ozone concentrations includes the Denver Cyclone (Szoke, 1991; Szoke et al., 2006; Wilczak & Christian, 1990; Wilczak & Glendening, 1988). The Denver Cyclone is a mesoscale cyclonic gyre and can form when a low‐level, southeasterly flowing air mass crosses the Palmer Divide, which is to the south of Denver. The gyre spins up to the north of the Palmer Divide. Depending on the location of the center of the Cyclone, different types of emissions are entrained into the cyclone, leading to accumulation of precursors and efficient ozone production. During the campaign, Denver Cyclone conditions were experienced on 3 days: July 27 and July 28 and 4 August. Vu et al. (2016) investigate the impacts of the Denver Cyclone on regional air quality in the greater Denver area and showed that the meteorological patterns associated with the Denver Cyclone increased aerosol mass loadings in the Denver metropolitan area mainly by transporting aerosols and/or aerosol precursors from the northern regions, leading to impaired visibility and air quality deterioration. Figure 6 demonstrates how this effect results in transport of emissions from the oil and gas sector into the Denver urban area, here depicted by coloring the flight track by the mixing ratio of ethane, a specific oil and natural gas emission tracer. Ozone measured from the C‐130 aircraft over the north Denver urban area exceeded 90 ppbv on this day.

Table 6 contains a list of all days where the MDA8 was equal to or exceeded 70 ppbv (75 ppbv) at one or more of the CHPHE operational ground stations, which occurred on 10 (4) different days during the campaign. Most of these days were characterized by weak synoptic flow conditions, either caused by high pressure over the region or stationary fronts situated over north‐eastern Colorado. A nighttime cold pool of polluted air often forms in the Plate River valley, which is then entrained back into the urban area through the developing upslope flow. On 28 July and 4 August, a Denver cyclone developed (strongest in July, as described above). Afternoon thunderstorms often prevented selected stations from exceeding the MDA8 by reducing available sunlight and locally injecting free tropospheric air into the boundary layer. This occurred widespread on 10 and 14 July. Some buildup of regional ozone was observed from 2 to 4 August, but midday thunderstorms prevented the northern stations from reaching high MDA8 values on that day.Table 6. MDA8 (ppm) at the Operational Front Range Network Sites for Days When at Least One Station Had Values Equal to or Larger Than 0.07 ppm (Bold Italic) or Equal to or Larger Than 0.075 ppm (Red Bold Italic)

Location	10 July	14 July	17 July	19 July	22 July	28 July	2 August	3 August	4 August	13 August
Welby	0.062	0.058	0.065	0.062	0.067	0.067	0.063	0.066	0.065	0.061
Aurora East	0.057	0.056	0.054	0.064	0.058	0.058	0.054	0.056	0.077	0.053
S Boulder Creek	0.054	0.056	0.060	0.065	0.075	0.066	0.064	0.067	0.048	0.063
CAMP	0.051	0.050	0.054	0.060	0.059	0.058	0.058	0.064	0.052	0.053
La Casa	0.058	0.055	0.056	0.064	0.065	0.064	0.063	0.066	0.062	0.063
Chatfield	0.064	0.068	0.074	0.060	0.075	0.065	0.057	0.071	0.072	0.067
Welch	0.056	0.059	0.064	0.061	0.070	0.060	0.056	0.069	0.063	0.059
Rocky Flats N	0.063	0.062	0.070	0.072	0.082	0.073	0.068	0.074	0.060	0.071
NREL Golden	0.060	0.065	0.069	0.059	0.079	0.073	0.071	0.078	0.066	0.070
Aspen Park	0.056	0.070	0.064	0.055	0.069	0.062	0.055	0.064	0.057	0.055
FTC West	0.074	0.043	0.065	0.063	0.076	0.070	0.069	0.073	0.056	0.079
FTC CSU	0.069	0.048	0.073	0.074	0.072	0.067	0.067	0.072	0.051	0.072
Weld Tower	0.065	0.047	0.055	0.065	0.060	0.063	0.064	0.069	0.055	0.059

• Boldface is to highlight the values that exceed the NAAQS.

4.2 Emissions

Most of the study focus has been on anthropogenic emissions as wildfires played a negligible role during the campaign. To date, the most comprehensive evaluation of current emission inventories of gas phase species for the Colorado Front Range has been conducted by Pfister and Flocke (2017) using the WRF/Community Multiscale Air Quality Modeling System (CMAQ; see the supporting information for details on methodology and results). Through careful comparison of modeled concentrations with NCAR/NSF C‐130 aircraft data accounting for model differences in winds, they concluded that in the base inventory, all relevant emission species including NO_x and primary VOCs were significantly underestimated. At the time of the study the National Emissions Inventory (NEI) Environmental Protection Agency (EPA) inventory was not yet available and the base emissions were based on a combination of 2017 projected emissions and actual 2014 activity data for O&NG sources and electric generation units (EGU). The study found that all major emission sectors except EGU emissions needed adjustments to bring modeled and observed emission tracer mixing ratios into agreement.

While mobile emissions for the Denver area appeared in good agreement with the measurements, both on‐road and off‐road (construction) emissions for the urban areas outside of Denver (such as Boulder, Longmont, Loveland, Greeley, and Fort Collins) had to be doubled to agree with the measured values. Ethyne emissions from mobile sources had to be doubled in all areas to achieve better agreement with the model. In the case of O&NG extraction emissions, the emission factors across the entire region had to be increased by a factor of 2 for both NO_x and VOC, except for ethane, the measured mixing ratios of which agreed well with the model without needing adjustments. Using the ground‐based VOC canister samples, Pfister and Flocke (2017) identified a number of point sources from O&NG activities, which, at the time of measurement, emitted very large amounts of reactive and toxic VOC (e.g., up to 120 ppbv of benzene were measured immediately downwind of several facilities). Some of these sources were not part of the emission inventory, but from the sporadic observations made during FRAPPÉ, it is not possible to tell what fraction these sources contribute to the derived adjustments to the VOC emissions in the oil and gas sector.

A large underestimate in bottom‐up O&NG inventories has also been found in previous studies. For example, Pétron et al. (2014) implemented a top‐down approach using aircraft data for calculation of NMHC emissions over the Denver‐Julesburg Basin and concluded that the state inventory for total O&NG‐related NMHC emissions is at least a factor of 2 too low, while in the case of benzene emissions the state inventory was approximately seven times too low. VOC emissions from O&NG constitute the largest source of VOCs in the NFRMA by mass, which is in line with studies by Gilman et al. (2013) and Eisele et al. (2009).

Table S2 in the supporting information compares the adjusted (posteriori) NO_x and VOC emissions from Pfister and Flocke (2017) to 2011, 2014, and projected 2017 EPA NEI emissions for each county as well as the entire Front Range. The posteriori estimate is lower in NO_x and VOC than EPA 2011 but higher compared to EPA 2014 (10% for NO_x and 30% for VOC). The table demonstrates that the dominant VOC emission contribution to Front Range VOCs come from Weld County (about 60% of the total), which is almost entirely attributed to O&NG operations. The NO_x emissions in Weld County are also substantial, contributing about 30% of the total Front Range NO_x emissions. The Denver area emissions are about 2 times higher than those of Weld County but also contain much larger traffic volumes and some very large single sources such as the Comanche power plant and Denver International Airport. Note that the substantial emissions from Jefferson, Adams, and Arapahoe counties are essentially attributable to the Denver Metro area. The largest emissions are from the far western regions of Adams and Arapahoe counties, namely, Commerce City (the major industrial center of Denver, also including the Comanche Power Plant, which is responsible for almost half of the NO_x emissions) and the densely populated southern and southeastern Denver suburbs (including the Santa Fe Dr. industrial area, the Denver Tech Center, Centennial Airport), respectively. The eastern end of Jefferson County contains about one third of the western Denver suburban area. Weld County’s NO_x sources are more distributed and are a mix of O&NG operations (contributing about half), vehicles, and industrial emissions (contributing about one quarter each).

NO₂ emissions have been studied by Wild et al. (2017), who developed an accurate method for determination of primary NO₂ emissions. The NO_x fraction emitted as nitrogen dioxide (NO₂) is usually presumed to be small but can affect ozone production and distribution, and this fraction is generally not reported in emission inventories. Their new approach was demonstrated by measurement of on‐road vehicle emission plumes from a mobile laboratory during the campaign. The mean value of NO₂/NO_x of 0.053 per plume on roadways in northern Colorado derived by this study is not weighted by the total NO_x emissions from sampled vehicles thus cannot be directly compared to aggregate numbers from roadside monitors. Nevertheless, the estimated value falls below most of the aggregate emission estimates (0.05–0.23) from Europe but is higher than the 0.01–0.03 range found in a 2008 study in California (Bishop et al., 2010).

Ammonia measurements from the campaigns were used by Battye et al. (2016) to evaluate the EPA NEI 2005 emissions implemented in the National Air Quality Forecast Capability (NAQFC) CMAQ model. They find that the model underestimated NH₃ in north‐eastern Colorado by a factor of 2.7. Similar results were derived from both ground‐level monitors and satellite retrievals. The underestimation of NH₃ vapor was not accompanied by a comparable underestimation of particulate NH₄⁺. Seasonal patterns measured at an Ammonia Monitoring Network (AMoN) site in the NE NFRMA suggest that the underestimation of NH₃ is not limited to summer. Even though the Battye study used NH₃ emission estimates from a 2005 inventory, more recent inventories also appear to strongly underestimate this source. The change in estimated NH₃ emissions from the 2005 NEI to the 2011 NEI was only an increase of 10% over the Colorado Northern Front Range (U.S. Environmental Protection Agency [USEPA], 2009, 2015), and based on emission trend information from EPA (https://www.epa.gov/air‐emissions‐inventories/air‐pollutant‐emissions‐trends‐data), the Colorado‐wide NH₃ emissions between 2005 and 2014 have actually decreased by about 25%.

Ammonia emissions, together with methane and N₂O emissions, were also studied by Eilerman et al. (2016) using repeated diurnal and seasonal measurements from a mobile van of two dairy farms, a beef cattle feedlot (Concentrated Animal Feeding Operation, CAFO), and a sheep feedlot. A consistent diurnal pattern in the NH₃ to CH₄ enhancement ratio has been observed, with midday enhancement ratios approximately 4 times greater than nighttime values. This diurnal pattern is similar, with slight variations in magnitude, at different CAFOs and across seasons. The average NH₃ to CH₄ enhancement ratio is in agreement with statewide inventory averages and previous literature. Evaluation of the enhancement ratios is important as it can be used as a source signature to distinguish feedlot emissions from other NH₃ and CH₄ sources, such as fertilizer application and fossil fuel development.

Vertical column densities (VCDs) of NH₃ as well as NO₂ and C₂H₆ were retrieved by a mobile Solar Occultation Flux instrument, which had its first deployment during FRAPPÉ. Kille et al. (2017) demonstrate that the retrieved VCDs are complementary to in situ observations and report significant variability in the VCDs on scales smaller than 6 km with 50 % of the VCD variability at distances shorter than 2 km. This means that satellites currently might quantify less than 10 % of the observed VCD variability since the variability happens on scales smaller than current ground pixel sizes of satellites. In agreement with other studies, Kille et al. (2017) find that emission fluxes for NH₃ during the summer daytime are generally underestimated in the NEI 2011 emission inventory by a factor of 2–10 for the per head NH₃ emissions. They further find an underestimate in emissions of NO_x from microbiological activity in CAFO soils. The latter might account for ∼10 % of the total NO_x emission in Weld County, CO, and can double the NO_x source in some rural agricultural areas.

The analysis of a unique set of isotopic measurements of individual sources and of regional air by Townsend‐Small et al. (2016) looks into the contributions of methane from agricultural and from oil and gas sources. They show that δ²H‐CH₄, but not δ¹³C, signatures are consistent in air sampled downwind of landfills, cattle feedlots, and oil and gas wells (Townsend‐Small et al., 2016) and estimate, when applying these source signatures to air in ground and aircraft samples, that at least 50% of CH₄ emitted in the region is biogenic. A larger contribution from O&NG was found in a recent study by Kille et al. (2019), who analyzed column measurements of methane, ethane, and ammonia in the Denver‐Julesburg Basin during March 2015. Using tracer relationships and analysis of wind direction, they attributed 63 ± 17% of the measured excess methane amount to O&NG, 25 ± 10% to agriculture, and 12 ± 12% to other sources.

Finally, PM_2.5 sources were studied by Valerino et al. (2017) using detailed aerosol chemical composition measurements carried out at the Golden ground site. Organic matter made the dominant contribution to PM_2.5 concentrations, comprising an average of ~75% of PM_2.5 throughout the study period. The majority of the organic matter has been estimated to be related to secondary aerosol formation. Concentrations of aerosol nitrate, ammonium, and elemental carbon increased significantly during the daytime when the winds were from the northeast, indicating a strong local source for these compounds. The long‐range transport of wildfire emissions also impacted the Colorado Front Range for 1–2 days during, and this event was characterized by elevated concentrations of organic aerosol and of potassium. Bahreini et al. (2018) measured and simulated the distribution of organic aerosol (OA) during FRAPPÉ and found larger enhancements of oxygenated OA in air masses with large contributions of O&NG emissions, compared to air masses dominated by urban emissions, but modeling the observed OA distributions proved difficult.

4.3 Ozone

A comprehensive evaluation of WRF/CMAQ modeled ozone concentrations has been conducted by Pfister and Flocke (2017) through comparison to surface data, ozone sonde launches, and aircraft measurements. The CMAQ simulations employed the CB6r3 chemical mechanism (Emery et al., 2015) and represent the overall characteristics of the measurements with a major part of the discrepancies explained by the underestimate in clouds (as discussed earlier). This leads to a high bias in ozone photochemical production in the model, which is in addition to inadequate representation of the variability in free‐tropospheric ozone as well as other uncertainties of the modeled meteorology as discussed in section 4.1.

Using the optimized emission inventory (Table S2, see also section 4.2), Pfister and Flocke (2017) performed zero‐out emission scenarios to derive an estimate of the ozone source contributions. Details about the methodology used by Pfister and Flocke (2017) and results are provided in section 3 of the supporting information. By selectively turning off one or more emission sectors and comparing the ozone distribution in each run with the control run (all emissions turned on), they find an average enhancement in O₃ due to local anthropogenic emissions in the NFRMA of around 15–20 ppbv (see Figure S1 in the supporting information). Their estimates vary between 20 and 30 ppbv on several high‐ozone days with largest values typically centered around the Denver metro area; O₃ enhancement maxima reach up to 40 ppbv (as seen on 28 July 2014; Figure S2). This study concludes that locally produced ozone is the major driver of ozone pollution in the NFRMA. Furthermore, mobile sources and oil‐ and gas‐related emissions were estimated to be the largest contributors to local ozone production in the NFRMA, with both sectors contributing, on average, 30–40% each to total NFRMA ozone production on high‐ozone days. Typically, oil and gas emissions show a stronger influence in the northern part of the NFRMA and the northern foothills, while mobile emissions dominate farther south and in the southern foothills.

The recent study by Bien and Helmig (2018) supports these findings and concludes that the trends in calculated surface ozone in the NFRMA are primarily determined by local/regional emissions and production, rather than by large‐scale ozone transport into the area. They analyzed 15 years of surface ozone observations in Colorado, and compared ozone trends at urban stations with trends at remote and high‐elevation sites, finding significant trends only at stations inside the Front Range urban corridor.

Ozone source attribution techniques face specific challenges, in particular because of the nonlinearity in chemistry. A number of different modeling approaches have evolved over time (e.g., Butler et al., 2018; Collet et al., 2017; Emmons et al., 2012), each with their own limitations. The zero‐out modeling sensitivity technique applied by Pfister and Flocke (2017) determines the contribution as the absolute change in ozone assuming one or more entire source sector is turned off, whereas other FRAPPÉ studies involved statistical analysis of chemical and meteorological surface observations, sometimes in combination with airmass trajectory analysis and chemical box modeling. Most of these studies focused on estimating the enhancement in ozone production due to emissions from O&NG exploration and extraction activities based on data collected during the FRAPPÉ time frame as well as other summer time measurements made in recent years.

The large contributions of O&NG sources to local production of ozone found by Pfister and Flocke (2017) are corroborated by several studies. Cheadle et al. (2017) analyzed surface observations of O₃ and several other trace gases (including methane, ethane, CO, and nitrous oxide) from stationary and mobile platforms. This study is limited to few individual days; however, these days represent a typical summertime meteorological daytime upslope situation for the NFRMA, which also occurred frequently during FRAPPÉ. Cheadle et al. (2017) estimate that O&NG sources can contribute up to 30 ppbv to O₃ growth in the Greeley area NE of Denver. The paper by Oltmans et al. (2019) analyzed ozone measured during the FRAPPÉ campaign at surface sites, at the BAO tower (10, 100, and 300 m above ground level), from tethered ozone sondes operated on several days at Ft. Collins West as well as balloon sondes released at the Ft. Collins West and Platteville surface sites. A back‐trajectory analysis for the ozone measured at BAO shows that on high‐ozone days, transport is generally from the North‐East sector, also underscoring the strong contribution of O&NG emissions for local ozone production at BAO.

Similarly, from analysis of ozone and concurrent wind measurements, Evans and Helmig (2017) state that between 2009 and 2012, the two surface sites South Boulder and BAO show a preponderance of elevated daytime ozone levels associated with east‐to‐west airflow from regions with O&NG operations located east and northeast, and a relatively minor contribution of transport from the Denver Metropolitan area in the southeast. A significant influence of O&NG emissions from NE Colorado on regional ozone production is also supported by two other studies. Zaragoza et al. (2017) used concurrent measurements of PAN and PPN made during FRAPPÉ at the BAO location to show that days with high ozone and PAN production coincide with the largest PPN/PAN ratios of 0.2–0.25, exceeding ratios that have been observed in areas dominated by anthropogenic emissions. Because photochemical processing of air masses containing large amounts of alkanes, such as emissions from O&NG extraction activities, produce larger PPN/PAN ratios, Zaragoza et al. (2017) conclude that O&NG emissions play an important role in the production of ozone observed at BAO. Benedict et al. (2019) used measurements of gas and particle phase species made during FRAPPÉ at two high‐altitude sites on the east side of Rocky Mountain National Park and analyzed six high‐ozone events. Linear correlations between alkyl nitrates and ozone observed during two high‐ozone events show a dominant influence of VOCs from the oil and gas sector, and ~20 ppbv of additional ozone observed at the high‐altitude site were attributed to production from alkanes attributed to O&NG emissions. Three other high‐ozone events were associated with a mixture of VOCs from urban and oil and gas sources, and one high‐ozone event was driven by a stratospheric intrusion.

The above findings are somewhat in contrast to the study by McDuffie et al. (2016), who used observation derived VOC OH reactivity (OHR) and O₃ production efficiency (OPE) together with an observationally constrained box model to quantify the influence of O&NG emissions on O₃ production at BAO. The measurements were made in the summers of 2012 and 2014, with OHR measurements only available in 2012. The study suggests that O₃ production at BAO is NO_x sensitive and that O&NG VOC emissions contribute, on average, 17.4% (2.9 ppbv) to maximum photochemical O₃, scaling nonlinearly with changes in O&NG VOCs.

The disagreement of McDuffie et al. (2016) with other studies likely is related to the fact that their work only considers measurements made at a single site in the NFRMA with variable influence of transported O&NG emissions. Also, their results are highly reliant on the set of model experiments. There is, however, the potential for significant ozone production downwind of BAO given that, on average, McDuffie et al. calculate that oil‐ and gas‐related VOCs comprised 84% of the carbon and 56% of the OH reactivity.

The variability of results from the different studies described above demonstrates the spatial limitations of ground‐based measurements and the need for multiple approaches to assess regional photochemistry in a complex emission environment. Aircraft measurements allow a better representation of the conditions across multiple scales given the high spatial and temporal variability in ozone regimes and OPE across the NFRMA and the nearby mountains but are typically limited in the time period that can be covered.

Pfister and Flocke (2017) also evaluated other emission sectors and find that the contribution of industrial emissions to NFRMA ozone are smaller compared to the mobile or oil and gas sector on average, but industrial emissions can dominate locally, as is the case in the Commerce City area near downtown Denver. NO_x emissions from the Cherokee power plant as well as from Denver International Airport cause ozone titration locally due to large NOx emissions in a confined area (there are other large NOx sources in the area, such as the Suncor refinery, but these comprise less than 15% of the NO_x emissions from the two combined sources above). Further downwind the additional NOx from these point sources can result in net ozone production in the western part of the NFRMA on high‐ozone days, albeit over a limited area. Reduction of NO_x from these sources will result in higher ozone locally and slightly lower ozone in areas further downwind. The conclusions were derived from the WRF/CMAQ zero‐out scenarios performed by Flocke and Pfister et al. (2017) and are corroborated by simulations with two different chemical box models (the NCAR/University of Munich BOXMOX and the NASA Langley Steady State Model) both constrained by NSF/NCAR C‐130 observations.

Chemical box models were employed in other studies to assess the ozone production in the NFRMA during FRAPPÉ. Baier et al. (2017) used direct measurements of the net ozone production (P(O₃)) from the Measurement of Ozone Production Sensor (MOPS) made at the Golden surface site to evaluate a photochemical box model that was constrained by concurrent measurements of other chemical species. The results show that median diel modeled P(O₃) is underestimated relative to the MOPS by roughly a factor of 2 around midmorning when actinic flux increases and morning rush hour emissions of NOx and VOCs decrease. This result agrees with previous studies where a MOPS instrument was used to measure P(O₃) (Cazorla et al., 2012; Baier et al., 2015). The uncertainties in both the measurements and the model are substantial however, and the model‐data P(O₃) mismatch could be attributed to unknown processes in the chamber or from unknown peroxy radicals, which are not explicitly treated in the two chemical mechanisms tested (RACM2 and MCMv331). If the MOPS accurately depicted atmospheric P(O₃), then these results would imply that P(O₃) in Golden was NO_x‐sensitive for more of the day than what is calculated by regional‐scale models, widening the NO_x‐sensitive P(O₃) regime period from just the afternoon further into the morning. This could result in overestimation of modeled ozone production in this area. Further study would be needed to examine this issue in detail.

The large set of vertical profile measurements from FRAPPÉ and DISCOVER‐AQ together with chemical box modeling was explored by Schroeder et al. (2017) to determine the validity of using satellite column CH₂O/NO₂ ratios as an indicator of near‐surface O₃ production sensitivity. They used a steady‐state time‐dependent photochemical box model for each 10‐s averaged aircraft observation to constrain the model. In case of the FRAPPÉ observations, ratios of species measured on the C‐130 were used to estimate species, which were not measured on the NASA P3 aircraft. The model was constrained by measured CH₂O mixing ratios. The calculated radical budgets and instantaneous values for radical and ozone production and loss rates were then compared to the observed CH₂O/NO₂ ratio. For the NFRMA they found a range of column CH₂O/NO₂ ratios between 1.1 and 3.0 that could not be reliably classified as either radical limited or NO_x limited and instead are labeled as transition/ambiguous. This range of uncertainty is larger than the range of 1.0–2.0 reported in Duncan et al. (2010). The increase in uncertainty is attributed to nonhomogeneous vertical distributions of CH₂O and NO₂ in the lower troposphere, and the resulting fact that near‐surface conditions may be decoupled from conditions at the top of the PBL and the lower free troposphere. Nearly half of the column CH₂O/NO₂ ratios measured in the NFRMA fall in the range 0–2 (i.e., radical limited or in the transition/ambiguous regime) compared to DISCOVER‐AQ campaigns in Maryland and Houston where only 12% and 25% of the column CH₂O/NO₂ ratios fall in the 0–2 range, respectively.

Ozone concentrations at the surface are impacted not only by local chemical production but also by entrainment from above the PBL. Kaser et al. (2016) focus on the diurnal cycle of entrainment by applying a zeroth‐order jump model to over 200 NASA P‐3 aircraft profiles conducted over the six key surface sites. By analyzing the ozone variability above and below the atmospheric boundary layer height, the evolution of ozone in both regions throughout the day, and the boundary layer growth rate, they find that entrainment through PBL growth is most important in the early morning, enhancing surface ozone at a rate of ~5 ppbv/hr. The entrained ozone can originate from long‐range transport but might also be locally produced ozone from the previous day. The enhancement effect weakens around noon and changes sign to become a small dilution effect in the afternoon on the order of −1 ppbv/hr.

Sullivan et al. (2016) used a combination of high‐resolution O₃ measurements, chemical simulations, and back trajectories to analyze high‐ozone concentrations on 22 July at the Fort Collins‐West, NREL‐Golden, and Rocky Flats North surface sites, which exceeded the 2008 75 ppbv NAAQS on this day. They show that due to terrain‐induced recirculation (“mountain‐plain solenoid flow”) and mixing of ozone to the surface, the Fort Collins site received an additional increase in O₃, which pushed the site over the limit.

Stratosphere‐Troposphere exchange (STE) is another process that might contribute to elevated surface ozone concentrations. These events are more common in spring and early summer, but during the campaigns, one such event occurred from 4 to 8 August 2014 and has been studied by Sullivan et al. (2015). Analysis of observations from the Goddard Space Flight Center TROPospheric OZone (TROPOZ) Differential Absorption Lidar, the University of Wisconsin High Spectral Resolution Lidar, and multiple ozone sondes as well as Real‐time Air Quality Modeling System simulations estimate additional contributions from stratospheric air to upper free tropospheric ozone mixing ratios over Fort Collins of about 10–30 ppbv throughout July and August 2014. However, this study did not quantify the impact of STE on surface ozone.

5 Summary and Outlook

The NSF/NCAR and State of Colorado FRAPPÉ study and the fourth deployment of the NASA DISCOVER‐AQ experiment were conducted jointly in summer 2014 and provide an extremely rich observational data set to study summertime ozone pollution in the Northern Colorado Front Range Metropolitan Area (NFRMA) as well as to examine other environmental issues such as nitrogen deposition and greenhouse gas emissions from agricultural and O&NG activities in the region. This data set together with a range of modeling tools ranging from process models to 3‐D chemical transport models has provided the framework for characterizing transport patterns in the regions, evaluating and improving emission inventories, investigating chemical processes, and assessing source contributions just to name a few. The campaigns also served as testbeds for the development of new measurement technologies.

The main objective of FRAPPÉ was to understand the drivers of summertime ozone pollution. High‐ozone days in the NFRMA are typically characterized by high pressure, summertime meteorological situations that favor stable conditions and little external forcing. These conditions favor clear skies and the development of light upslope or cyclonic local winds, which cause mixing of local ozone precursor emissions, efficient ozone production, and transport into the western suburbs and adjacent mountains. Numerous studies have identified emissions from oil and gas activities as one of the major contributors to ozone, albeit the quantification of their role varies with the methodology applied and the considered region and time period. This emission sector is also identified as carrying the largest uncertainties in currently available emission inventories, both with regard to magnitude and emission factors. Emissions from the transportation sector are estimated as being about equally important as O&NG on average over the NFRMA. The benefit of multiscale and multifaceted experimental and modeling studies to derive robust conclusions on source contributions has been demonstrated. A multitude of studies has also focused on the transport of pollution from the NFRMA into the pristine mountain areas to the west of NFRMA and demonstrated the large impact the NFRMA human footprint exerts on the air quality in these remote regions.

The comprehensive suite of measurements provides the basis for careful evaluation of different processes in 3‐D chemical transport models and accurate model representation of meteorology and transport has shown crucial for the model’s chemical performance. Considerations need to be given to the representativeness of observations to ensure that model output is evaluated fairly and correctly. The studies point toward the need for improvements in simulating meteorology including the representation of clouds and winds, specifically wind speed at high‐altitude terrain. Data assimilation of clouds for improved simulation of photolysis rates such as recently shown by Ryu et al. (2018) could provide benefits in this regard. There is also a need to revisit current mechanism chemistry, including possible missing peroxy radicals at high NO or NO_x levels or the treatment of nitrate chemistry (Cassanelli et al., 2007). FRAPPÉ data have also been used in other studies such as on the nitrogen deposition in Rocky Mountain National Park (Benedict et al., 2018) or health impacts of oil‐ and gas‐related emissions on local residents (McKenzie et al., 2018).

A large number of science questions have been addressed and advanced with the to‐date published studies, but some data sets and topics remain insufficiently exploited and should be utilized in future studies. Among these are the pollution inflow into and outflow out of the NFRMA, the assessment of emissions from individual processes using VOC canister samples collected nearby selected emission sources, characterization and evaluation of powerplant emission from analysis of dedicated aircraft passes, the spatial and temporal variability in NO_x and VOC sensitivities across the region, or the application of inverse modeling techniques to better constrain emission inventories for different sectors.

Better constrained emission inventories and chemical processes greatly aid the process of developing effective ozone reduction strategies. Emissions in the NFRMA are constantly evolving suggesting the potential for year‐to‐year changes in photochemical O₃ sensitivities and emissions of VOC and NO_x. Repeated detailed aircraft measurements and intense ground‐based data like what was done during FRAPPÉ would greatly help to detect these trends, now that a baseline has been established with the 2014 campaigns.

Acknowledgments

The data from the field campaigns are publicly available at the FRAPPÉ/DISCOVER‐AQ data archive (http://www‐air.larc.nasa.gov/missions/discover‐aq/discover‐aq.html) including model output along the aircraft flight tracks. The authors would like to thank the State of Colorado/Colorado Department of Public Health and Environment and the National Science Foundation (NSF) for funding of FRAPPÉ and NASA for funding of DISCOVER‐AQ. We acknowledge the modeling contributions by Sojin Lee (Univ. of Houston). We acknowledge Scott Landes (CDPHE), Mary Barth (ACOM/NCAR), Geoff Tyndall (ACOM/NCAR), and John Orlando (ACOM/NCAR) as well as the three anonymous reviewers for valuable input and comments to the manuscript. The National Center for Atmospheric Research is sponsored by the National Science Foundation.