Air Quality Impacts from Oil and Natural Gas Development in Colorado

Page 3 22 Introduction Page 3 23 Resources Utilized and Published Work Page 5 24 Monitoring Networks Page 13 25 Nitrogen Oxides Page 13 26


52
The development of hydraulic fracturing techniques has made it profitable to extract petroleum hydro-53 carbons from geologic shale formations. The application of this technology has caused a surge in new oil 54 and natural gas (O&NG) drilling in shale basins across the United States, including in Colorado, where 55 most of the activity has been in the Denver Julesburg Basin (DJB). In 2017, there were over 53,000 ac-56 tive O&NG wells in Colorado. The O&NG development is concentrated in a number of lower elevation 57 basins, with ≈ 24,000 wells (January 2018) located in Weld County in the DJB, which in 2017 produced 58 ≈90% of the oil in Colorado [Swain, 2018]. From 2010-2018, annual natural gas production in Weld 59 County increased by a factor of 3.5, and annual oil production by a factor of ≈ 6.5 [Drilling-Edge, 2019]. 60 Some of the growth of the O&NG industry has occurred within the periphery of urban and residential 61 areas, raising concerns within communities. Proximity to O&NG operations has been associated with 62 human health effects, with atmospheric emissions being a primary pathway of exposure. Types and 63 causes of emissions are dependent on multiple variables and stages of the well development, and can 64 arise, for instance, from heavy equipment use at the site and vehicle traffic, power generation, drilling 65 operation, spillage and evaporation of fracking fluid, flowback of the extracted petroleum products, flar-66 ing, and fugitive or controlled hydrocarbon emissions during loading and transportation [Adgate et al., 67 2014]. Fracking fluid is a mixture of a multitude of synthetic chemicals, with the composition typically 68 kept proprietary by operators. Silica, added to the fracking fluid, dust, and soot/particles from diesel 69 engines contribute to particulate exposure. Gaseous emissions arise from fracking fluid additives, con-70 trolled venting, flaring, and leakage of equipment, storage tanks, and pipelines. Directly emitted gaseous 71 pollutants of concern for human health are hydrogen sulfide and petroleum constituents, including aro-72 matic and polycyclic aromatic hydrocarbons. For southwestern Colorado, including Mesa Verde National Park, an ozone enhancement of 9.6 ppb was 98 estimated for the ozone medium daytime 8-hour average ozone (MDA8) from O&NG influences 99 [Rodriguez et al., 2009].  Elevated concentrations of reactive nitrogen were associated with emissions from oil and gas operations, which are frequently co-located with agricultural production and livestock feeding areas in the region, and from urban areas.    Air quality monitoring and air sampling is conducted by the CDPHE, NOAA, the National Park Service, 174 Boulder County, and the City of Longmont. A map showing the distribution of monitoring sites and 175 measured species is shown in Figure 2. Ozone is monitored at the highest number of sites, followed by 176 PM2.5, and nitrogen oxides. There currently is only one location with continuous VOCs monitoring 177 (Boulder Reservoir); however, two more sites are anticipated to begin VOCs monitoring within the next 178 year The O&NG industry is the single most significant source of methane in Colorado [Pétron et al., 2012]. 206 Quantifying the methane flux from the O&NG industry has been challenging as the geographic area of 207 the O&NG activities overlaps with agricultural, beef, and dairy production areas, which all constitute sig-208 nificant methane sources. Methane to non-methane hydrocarbon (NMHC) relationships, in particular 209 those with ethane and propane, and the stable isotopic signature of methane, have been used to deci-210 pher the O&NG contribution to the total methane flux. Point source measurements near emission 211 sources, mobile lab ground surveying, and aircraft observations, in combination with inventory infor-212 mation, have been used to derive basin-wide O&NG methane flux estimates. Three NOAA studies, cov-213 ering the observations during three short time windows within the 2008-2015 period, are summarized in 214 Figure 2: (a) Air monitoring sites in Colorado that report ozone, nitrogen oxides, and PM2.5 data to the EPA Air Quality System archive (map downloaded from https://www.epa.gov/outdoor-air-quality-data/interactive-mapair-quality-monitors). Additional sites operated by NOAA and regional municipalities, and sites that monitor methane and VOCs were added to the map. (b) Distribution of oil and gas well sites (purple dots) for the same map area within the State of Colorado (map downloaded from https://cogcc.state.co.us/#/home).      variable, and estimated to enhance the Intermountain West regional summer MDA8 by 0.3 -1.5 ppb [Lu 372 et al., 2016]. Regional ozone production is further promoted by the dry and sunny climate. Combined, 373 these conditions make it more challenging for western States, including the NCFR, to control its ozone as 374 it leaves less room than in other regions for local ozone production to exceed the standard [ whereas transport from surrounding areas, including the DMA, brought in air with lower ozone levels 384 (Figure 4b). The geographical overlap of the source footprint with highest ozone with the area of high-385 est O&NG well density (Figure 4a), provided credence to the argument that O&NG industry emissions 386 played an important role in ozone production and high ozone occurrences. Daytime summer ozone pro-387 duction rates of 7 -8 ppb hr -1 have been seen in ambient diurnal ozone data [ widely across the region, with lower ratios present in the DMA, and higher ratios in the VOC-rich DJB. 392 These different air masses can mix during transport and recirculation, causing a wide range of spatial 393 and temporal differing conditions and ozone production regimes. 394 Two studies estimated the contribution of O&NG VOCs to the total reactivity with the OH radical using 395 VOC speciation and atmospheric concentrations at the BAO. This variable can serve as a metric for the 396 chemical reactivity of air and its potential for producing ozone. A NOAA study estimated that 55 +/-18% 397 of the reactivity was attributable to O&NG emissions [Gilman et al., 2013]. Swarthout et al. [2013], us-398 ing a similar approach but with independently collected data, determined a value of 57%. While OH re-399 activity does not directly translate to ozone production, based on these results, both groups predicted 400 that O&NG VOC emissions would enhance and play a significant role in the regional ozone budget. It 401 should be noted that these measurements were conducted in the late winter, when ozone production is 402 relatively moderate in the NCFR. Therefore, these findings represent, for example, lower influence from 403 biogenic VOCs. VOCs reactivity in a photochemical model. Their findings showed "that O&NG alkanes contribute over 406 80% to the observed carbon mixing ratio, roughly 50% to the regional VOC OH reactivity, and approxi-407 mately 20% to regional photochemical ozone production." Using observations from BAO for correlation 408 analyses and modeling of oxidation chemistry, Lindaas et al.
[2019] stipulated that O&NG emissions con-409 tribute to ozone production on high ozone days; however, that study fell short of providing a quantita-410 tive estimate. Another modeling study by NCAR scientists [Pfister et al., 2017b], building on FRAPPE data 411 for the wider NCFR area, concluded that on average, O&NG emissions contribute 30-40% to the local 412 ozone production on high ozone days. It needs to be emphasized that all of these studies derived esti-413 mates for the ozone produced regionally, not the total ozone, which is also determined by the back-414 ground that is transported into the region (see above). showed an association of high ozone days with transport from sectors with intense O&NG production 423 towards the northeast. The authors concluded that O&NG emissions were an important source of 424 ozone precursors and are crucial in producing peak ozone events in the NCFR. The ozone production 425 chemistry is primarily driven by VOCs of anthropogenic origin; biogenic emissions appear to have a mi-426 nor contribution to the NCFR ozone production chemistry [ convective uplifting that is pulling in air from the east. The flow reverses during the night, with cooler 495 air from higher elevations descending the mountains and forcing west to east air transport. 496 for the wider NCFR. The transition time between these two flow regimes changes with distance from 502 the mountain slopes, with locations further east experiencing an on average later onset of upslope flow 503 conditions. The diurnal flow regimes are most pronounced during the summer because of the larger 504 solar irradiance that is providing the thermal forcing. 505 The upslope flow paths are somewhat segregated, such that there is a separation of air masses that are 506 heavily influenced by O&NG emissions to the north of the DMA, whereas air masses south of North Den-507 ver are more strongly influenced by urban emissions [Pfister et al., 2017a]. Air enriched with emissions 508 from urban, traffic, O&NG, and other regional sources can get 'trapped' along the mountain slopes dur-509 ing late afternoon. This is reflected by highest ozone levels being observed at monitoring sites nearest 510 to the mountain slopes [Bien and Helmig, 2018 Although a classical view of high pollution episodes invokes a stagnant high pressure region, usually over 516 flat terrain, the meteorology of the NCFR leads to more complex circulation regimes. A common regime 517 23 518 519 in the winter occurs with downslope westerly warm winds from the Rocky Mountains flowing over 520 colder air drawn from the east toward a low pressure trough along the foothills or due to lee-cyclogene-521 sis located over southeast Colorado [Neff, 1997]. During the summer, the 'Denver Cyclone' is often ob-522 served. These conditions provide a similar opportunity for trapping pollutants near the surface [Wilczak 523 and Glendening, 1988;Wilczak and Christian, 1990;Szoke, 1991]: In this case, the Denver Cyclone occurs 524 nearer the surface with warmer air aloft from the south that originated over the Palmer Divide, a ridge 525 extending to the east and south of Denver. As the air from the east (underlying the warmer air aloft) 526 carries pollutants and precursors from the eastern plains, the air can stagnate as it encounters the topo-527 graphic barrier to the west. This circulation pattern can cause pollution to circulate and accumulate for 528 several days, leading to increases of secondary pollutants. Vu et al. [2016] demonstrate an up to an 529 80% increase in aerosol constituents during a cyclone episode during FRAPPE. 530 The frequency and prominence of high ozone occurrences is correlated with high pressure systems that 531 promote high temperatures, stagnant air circulation, and sunny weather, conditions that combined are 532 favorable for photochemical ozone production. Reddy and Pfister [2016] investigated this relationship 533 and proposed a method in which monthly 500-mbar pressure heights were used for correcting the year-534 to-year variability in the fourth highest 8-hour ozone average. Further, these conditions promote cyclic 535 terrain-driven circulations that reduce pollution transport away from sources. The authors recommend 536 correcting annual MDA8 data using monthly 500-mbar pressure heights for reducing weather influences 537 on ozone trends. 538

539
Emission inventories have been developed by state and national regulatory agencies in support of air 540 quality modeling and for directing policy development. These bottom-up inventories are based on emis-541 sions estimates of facility types and operations with regional/basin-wide scaling using best available fa-542 information. [Vaughn et al., 2018] demonstrated that this may be part of inventory uncertainties and 554 discrepancy between bottom-up and top-down emission estimates. Further, the lack of temporal infor-555 mation makes source apportionment and model performance evaluation more difficult. 556 In the most extensive evaluation of methane and VOC emissions representation in the Western Regional 557 Air WRAP Phase III inventory [WRAP, 2009] to date, [Pétron et al., 2012] concluded that "there are nota-558 ble inconsistencies between our results and state and national regulatory inventories". They further 559 stated "Our analysis suggests that the emissions of the species we measured are most likely underesti-560 mated in current inventories and that the uncertainties attached to these estimates can be as high as a 561 factor of two". Results also showed that methane sources from natural gas industries in Colorado were 562 most likely underestimated by at least a factor of two. Besides methane and total VOC, the study also 563 assessed benzene, and concluded that for this species State inventory estimates were too low by at least 564 a factor of five. Levi total VOC estimate by at least a factor of three. 582 Taken together, these available comparison studies highlight the deviations between the bottom-up and 583 top-town emissions estimates. Unfortunately, there is a scarcity of top-down estimates available for 584 this evaluation, and each of these have relatively large uncertainty windows themselves. Nonetheless, 585 these disagreements diminish the confidence in the State's bottom-up inventories, and air quality mod-586 eling that is building on these most likely under-predicted emissions. 587 Lastly, the methane flux estimates listed in Table 2, covering observations between 2008-2015, do not  616 show any changes in the total methane flux that are outside of the uncertainty windows of the individ-617 ual observations. Assuming that the VOC/methane ratio has remained constant, these methane flux de-618

Changes in O&NG Emissions and Atmospheric Concentrations
terminations do not suggest changes in basin-wide VOC emissions. Considering the large increase in 619 natural gas production during this time period, a reduction in the fraction of emitted methane (relative 620 to the produced quantity of natural gas) and VOCs appears probable [Peischl et al., 2018]. 621 Available VOC emissions estimates, differentiated by ethane, benzene, and total VOC, is provided in Ta   time. Certainly, none of these data show increases that scale with the DJB O&NG production increase 637 (e.g. 3.5-6.5 times for natural gas and oil, respectively, for 2010-2018). Therefore, it appears likely that 638 relative emissions rates have declined, likely due to the implementation of stricter emission controls. 639 However, the growth of the number of operations has probably counteracted those relative emissions 640 reductions, resulting in overall basin-wide stable total emissions. 641 These populations are at greatest risks for these exposures. Secondary products that are formed via 648 photochemical processing of emissions during transport are another concern. Here, the pollutants of 649

Oil and Natural Gas Emissions and Air Quality
importance are ozone and PM2.5. These species are transported across a wide spatial scale in the NCFR, 650 thereby affecting a much larger population. In excess of 3.5 million people live in the NCFR ozone NAA. 651 Approximately half of the NAA (mostly the northern part) is moderately to heavily influenced by O&NG 652 emissions. This part of the NCFR is where O&NG emissions have the greatest impact on ozone and ex-653 ceedances of the NAAQS. Atmospheric levels of particulates are relatively modest in the NCFR, with par-654 ticulate air quality thresholds being exceeded only occasionally, for instance during wildfire plume 655 transport events and wintertime inversion conditions. Nonetheless, health impacts from particulates 656 originating from O&NG sources are estimated to be similar as for ozone [Fann et al., 2018]. However, 657 ozone is currently the much more recognized regional pollutant. 658 Emissions of most primary air pollutants continue downwards trends in most of the United States. This 659 also applies to surface ozone; implementation of pollution control measures has resulted in declining 660 surface ozone across wide geographical scales in developed North American and European countries 661 [Fleming et al., 2018]. For instance, the compilation of ozone trends shown in Figure 6 provides a nice 662 testimony for decreases in ozone production across the U.S. These downward trends are particularly 663 remarkable in light of the population growth, increase in energy demand and production, and climate 664 change, which is driving higher ozone production rates from the increase of ozone precursors and faster 665 reaction rates in a warmer climate. Assessments in the magnitude of this effect vary by study. This 666 ozone 'climate penalty' potentially can be rather significant, with some estimates predicting an up to 3-6 667 ppb increase in surface ozone per degree of temperature increase [Rasmussen et al., 2012]. ing increasing ozone. Inclusion of 2016-2018 data in the ozone trend analysis indicates a steadily declin-705 ing regional Design Value in the last seven years (Supplemental Materials). 706 The DMA/NCFR ozone behavior deviates from that of most other regions in the U.S. This is most evident 707 in the summer daytime average ozone trends ( Figure 6). While this ozone metric has clearly (at many 708 sites with statistical significance) been heading downwards across the U.S., increasing values were de-709 termined for most sites in the DMA/NCFR. Persisting elevated ozone conditions were evident during 710 2018; ozone data collected by CDPHE in the NCRF were higher than in any of the previous five years, 711 with a season maximum of 89 ppb and 32 exceedance days of the 8-hour 70 ppb NAAQS at the Boulder 712 Reservoir site alone (see Figure 7 for   30 smaller than for other U.S. NAA, making meeting the standard more challenging. However, the local 737 ozone production is mostly within the control of the State. Meeting the standard is calling for a con-738 certed and aggressive effort in curbing regional ozone precursor emissions. and better define particulates pollution, and to regulate particulates and secondary aerosol pre-760 cursor emissions from the industry. 761 -VOCs data, mostly from occasional and campaign-type data, as well as the CDPHE monitoring at 762 Platteville, clearly show a strong contribution from O&NG operations on total VOCs and the 763 ozone-producing VOC reactivity in the region. VOC monitoring is crucial for assessing O&NG air 764 quality impacts. The current distribution of monitoring sites has a number of shortcomings for 765 evaluating and monitoring changes of O&NG emissions. VOC monitoring is needed near opera-766 tions to assess facility emissions and exposure risks of nearby residencies. This monitoring 767 needs to be expanded to activities such as flowback, liquid unloading, and wastewater separa-768 tion, which appear to be associated with high emissions and which have been mostly neglected 769 or been underrepresented in previous assessments. In order to capture the high variability of 770 these emission, this monitoring should be at high time resolution, ideally in real time. VOC mon-771 itoring needs to be tailored for characterizing emission trends, representative for a wide re-772 gional footprint. This can, for instance, be achieved by sampling at elevated sites or/and from 773 inlets high above the surface, and best during mid-day to afternoon hours, when chances to 774 sample mixed boundary layer air are highest. This monitoring would be most promising if it is 775 conducted continuously, and at highest possible accuracy. Continuous, concurrent, and coordi-776 nated monitoring at strategically selected sites upwind and downwind of the DJB would allow 777 assessing changes in basin-wide emissions. 778 31 -VOCs emitted from O&NG sources constitute the majority of the OH reactivity in the DJB north 779 of the DMA. These emissions contribute to a temporal and locally variable ozone production. 780 Summertime elevated ozone occurrences show a high correlation to transport from O&NG ex-781 traction regions and atmospheric O&NG influences. Due to the ozone production dynamics and 782 air circulation patterns, the daytime peak maximum ozone values are often observed along the 783 NCFR foothills, tens of kilometers downwind of the O&NG emissions source regions, and 784 thereby impacting communities outside of the production regions. These downwind air quality 785 impacts from O&NG industries should be a strong consideration in the design of monitoring net-786 works and decision-making on regulating existing and new O&NG development in the region. 787 -Several independent measurements near O&NG operations have shown spikes with highly ele-788 vated concentrations of BTEX compounds that exceed health risks thresholds for nearby resi-789 dents. Highest concentrations have been reported downwind of disposal facilities, rather than 790 from well pads. Available data are mostly from short episodic measurements. This clearly 791 demonstrates that characterization of BTEX emissions warrants more attention. This needs to 792 include markably during the past five years. A concerted effort building on these capabilities by regu-804 larly (e.g. monthly) light aircraft profiling, could, within a short time frame, yield significant im-805 provements of the basin-wide total emissions characterization. 806 -Assessments of ozone contribution from O&NG emissions will have high uncertainty, and will 807 under-predict the true ozone production as long as they rely on underestimated O&NG inven-808 tory emissions. Ozone impact studies need to be revisited with consideration on the most real-809 istic NOx and VOC emissions from the industry. 810 -Ozone pollution in the NCFR is well within the range where ecosystem impacts and production 811 yield losses in agriculture are predicted. Given the size of the agricultural industry, and from 812 available literature on ozone effects on crops, it is expected that the economic loss to the State's 813 farming industry from the O&NG-contributed ozone may be quite significant. A quantification 814 of the actual revenue loss is needed for evaluating these adverse economic impacts of O&NG 815 industry emissions. 816 -There has been a remarkable growth in the number of peer-review studies on air quality im-817 pacts from O&NG emissions. Consideration of the findings from these resources, and closer 818