2013 Uintah Basin Winter Ozone & Air Quality Study

FINAL REPORT 2013 UINTAH BASIN WINTER OZONE & AIR QUALITY STUDY

SETH LYMAN, MARC MANSFIELD, AND HOWARD SHORTHILL UTAH STATE UNIVERSITY, COMMERCIALIZATION AND REGIONAL DEVELOPMENT

BA TAH SIN IN U

DOCUMENT NUMBER: CRD/13.10297B 3 Y D UBOS2 U REVISION: ORIGINAL RELEASE 0 13 S T O Z ON E DATE: OCTOBER 24, 2013

Table of Contents

SPATIAL, SEASONAL, AND INTER-ANNUAL ASPECTS OF WINTERTIME OZONE Introduction ...... 1 Methods ...... 1 Ozone Measurements ...... 1 Ozone Precursor Measurements ...... 2 Meteorological Measurements ...... 3 Spatial Data Analysis ...... 3 Results and Discussion ...... 4 Ozone Concentrations and Distribution ...... 4 Distribution of Nonmethane Hydrocarbons ...... 12 Distribution of Oxides of Nitrogen ...... 19 Influence of Transport Patterns on Ozone During an Inversion Episode ...... 25 Seasonal Changes in Ozone and Precursor Chemistry ...... 27 Interannual Variability in Wintertime Ozone ...... 31 Summary ...... 34 Acknowledgements ...... 34 USE OF THE CALMET DIAGNOSTIC MODEL TO SIMULATE WINTER INVERSIONS ...... 35 Introduction ...... 35 Methods ...... 35 Results and Discussion ...... 37 CALPUFF Model Animations ...... 37 Wind Speed and Direction ...... 38 Temperature ...... 42 Acknowledgements ...... 44 References ...... 45

WINTERTIME EMISSIONS OF HYDROCARBONS FROM PRODUCED WATER EVAPORATION FACILITIES Introduction ...... 47 Methods ...... 48 Pond Facilities Descriptions ...... 48 Dynamic Flux Chamber Measurements ...... 49 Meteorological Measurements ...... 50 Water Samples ...... 51 Results and Discussion ...... 51 Water Chemistry ...... 51 Emissions from Different Pond Surface Types ...... 52

i Factors that Influenced Emission Rates ...... 54 Comparison to Basin-wide Emissions ...... 55 Uncertainties and Next Steps ...... 56 Summary ...... 56 Acknowledgements ...... 56 References ...... 57

SURVEY OF BASINS AND VALLEYS IN THE WESTERN USA FOR THE CAPACITY TO PRODUCE WINTER OZONE Executive Summary ...... 59 Introduction ...... 64 Survey of Thermal Inversions ...... 66 Survey of Snow Cover ...... 71 Survey of Ozone-Favorable Meteorology ...... 71 Survey of Oil, Gas, and Coal Bed Methane Production ...... 76 Survey of Population ...... 78 Survey of Wind Speed ...... 79 Survey of Total Ultraviolet Radiation ...... 80 Ozone Monitoring in the Upper Snake River Plain ...... 82 Average Winter Ozone Concentrations in Seven Basins ...... 86 Discussion of the Observed Ozone Concentrations in Seven Basins ...... 88 Predictions for the Ozone Concentration in the Remaining Five Basins ...... 89 Case-by-Case Examination of Ozone Seasons in Each of the Seven Basins ...... 89 A. Uintah Basin ...... 90 B. Upper Green River Basin ...... 95 C. Price River Valley ...... 100 D. Wind River Basin ...... 102 E. Upper Snake River Plain ...... 105 F. Cache Valley ...... 107 G. Salt Lake Valley ...... 109 Yellowstone Caldera ...... 116 Blow-over from the Wasatch Front into the Uintah Basin ...... 117 Unanswered Questions ...... 118 Conclusions ...... 119 Acknowledgements ...... 119 References ...... 120 Appendix A ...... 122 Appendix B ...... 126

ii ANALYSES OF HEALTH IMPACTS AND RISKS DUE TO UINTAH BASIN AIR QUALITY Abstract ...... 141 Study 1. Winter Ozone and Asthma in the Uintah Basin ...... 142 Introduction ...... 142 Methods ...... 144 Seasonal Asthma Trends in the Uintah Basin ...... 144 Additional Evidence for Lack of Correlation between Ozone and Asthma ...... 146 Conclusions ...... 150 Acknowledgements ...... 150 Study 2. Health Risk Assessments of Uintah Basin Air Quality ...... 151 Results from Sublette County, Wyoming, and Motivation of our Study ...... 151 Preliminary Screening ...... 151 Additional Screening ...... 153 Estimates of the Annual Concentrations of Benzene and Additional VOC ...... 155 Conclusions ...... 156 Acknowledgements ...... 156 References ...... 157 Appendix: Seasonal Trends in Asthma and Correlations with the School Year ...... 160

iii SPATIAL, SEASONAL, AND INTER-ANNUAL ASPECTS OF WINTERTIME OZONE

Seth Lyman, Marc Mansfield, Howard Shorthill, Randy Anderson, Chad Mangum, Jordan Evans, and Tate Shorthill

Introduction

Ozone concentrations have been measured continuously in the Uintah Basin since fall 2009. During winter 2009-10, monitoring stations at Ouray and Red Wash observed ozone concentrations that exceeded the Environmental Protection Agency (EPA) standard. Following the discovery of this new phenomenon, stakeholder concern led to the establishment of several additional air quality monitoring stations to support subsequent studies. Seventeen stations operated around the Uintah Basin during winter 2010-11, and 30 operated during winter 2011-12. Martin et al. (2011) and Lyman and Shorthill (2013) highlight the results of these studies. This section contains an analysis of ozone, precursor, and meteorology data from 20 monitoring sites that operated around the Basin during winter 2012-13, and an analysis of variability across the four years of available ozone data.

Methods

Ozone Measurements

Ten of the air quality monitoring stations in the Uintah Basin during winter 2012-13 were operated by Utah State University (USU), and ten were operated by other organizations. Table 1 contains a list of all monitoring stations, including locations, elevations, and responsible operators. Data and methods used for stations operated by other organizations were obtained from EPA’s AQS database (https://ofmext.epa.gov/AQDMRS/aqdmrs.html). We utilized 2B Technology Model 205 or 202 ozone monitors at most of the stations, and we operated an Ecotech Model 9810 ozone analyzer at the Horsepool site. We performed calibration checks at all USU stations at least every other week using NIST-traceable ozone standards. Calibration checks passed if monitors reported in the range of ±5 ppb when exposed to 0 ppb ozone, and if monitors were within ±7% deviation from expected values when exposed to higher concentrations of ozone. We only included data bracketed by successful calibration checks in the final dataset.

1 Table 1. Air quality monitoring stations that operated winter 2012-13.

Data Elevation Site Name Operator Availability Latitude Longitude (m) VOC NOx Dinosaur NM NPS AQS dbase 40.4371 -109.3047 1463 N/A N/A Duchesne USU USU 40.1615 -110.4011 1682 N/A N/A

Fruitland UDAQ AQS dbase 40.2087 -110.8403 2021 canister active NO, NO2

Seven Sisters USU USU 39.9813 -109.3454 1618 canister passive NO2 Gusher USU USU 40.2935 -109.6575 1557 N/A N/A

Horsepool USU USU 40.1437 -109.4672 1569 active active NO, NO2 , NOy Lapoint USU USU 40.4040 -109.8157 1674 N/A N/A Little Mtn. USFS AQS dbase 40.5368 -109.7001 2624 N/A N/A Mountain Home USU USU 40.4319 -110.3821 2234 N/A N/A

Myton EPA/STI AQS dbase 40.1948 -110.0622 1550 N/A active NO, NO2

Ouray EPA/Golder AQS dbase 40.0548 -109.6880 1464 N/A active NO, NO2

Rabbit Mtn. Enefit/Tetra. AQS dbase 39.8687 -109.0973 1879 N/A active NO, NO2

Rangely NPS/BLM AQS dbase 40.0869 -108.7616 1648 N/A active NO, NO2

Red Wash EPA/Golder AQS dbase 40.1972 -109.3525 1689 N/A active NO, NO2

Roosevelt UDAQ/USU AQS dbase 40.2942 -110.0090 1587 active active NO, NO2 , NOy Sand Wash USU USU 39.8390 -109.9150 1416 N/A N/A Seep Ridge USU USU 39.7539 -109.5460 1975 N/A N/A

Vernal UDAQ AQS dbase 40.4531 -109.5097 1606 canister active NO, NO2

Wells Draw USU USU 40.0670 -110.1510 1768 canister Passive NO2

Whiterocks EPA/STI AQS dbase 40.4694 -109.9304 1841 N/A active NO, NO2

Ozone Precursor Measurements

We measured NO, true NO2 (via a photolytic converter), and NOy at Roosevelt and Horsepool with AQD/Teledyne-API and Ecotech systems, respectively, and calibrated the systems weekly with NO standards, monthly with NO2 standards via gas phase titration, and at the beginning and end of the campaign with nitric acid and n-butyl nitrate permeation tubes (for NOy). A number of sites operated by other organizations measured NO and NO2 via a molybdenum converter-based system (measurement contains some NOy), and we obtained these data from EPA’s AQS database. We performed supplemental NO2 measurements at some sites from February 1 through February 8 using Radiello NO2 passive samplers (Table 1). We deployed Radiello samplers for one week and analyzed them according to Radiello protocols on a Dionex ion chromatograph. Radiello samples were blank corrected and also corrected to concentrations of NO2 measured with automated instruments at co-located sites.

We measured 57 ozone-forming nonmethane hydrocarbons (NMHC) in 30-minute and hourly samples at Horsepool and Roosevelt, respectively, during January and February. NMHC were analyzed by sample

2 concentration on activated carbon traps, followed by desorption into automated gas chromatography- flame ionization detection systems. We calibrated these systems every other week with certified gas standards. We also performed supplemental NMHC measurements at some sites from February 1 through February 8 using evacuated stainless steel canisters. Canisters were filled from 7 to 9 AM on 1, 3, 6, and 8 February. Some of these samples were analyzed with the Roosevelt automated gas chromatograph, but most were analyzed by a commercial laboratory using gas chromatography and flame ionization detection. Canister samples utilized automated sampling timers and critical orifice- based flow controllers. All wetted parts were either stainless steel or stainless steel coated with deactivated fused silica. All canister sampling components were cleaned between each use by repeatedly flushing with hot, humidified nitrogen. EPA PAMS compounds (EPA, 2003) were measured by the automated systems at Horsepool and Roosevelt, and PAMS compounds and methanol were measured with canister systems. Methanol was analyzed via gas chromatography and mass spectrometry by a commercial laboratory.

Meteorological Measurements

We operated solar radiation sensors, including incoming and outgoing short wave, long wave, UV-A, and UV-B at Horsepool, and incoming and outgoing shortwave radiation at Roosevelt. We operated a comprehensive, research grade meteorological instrument suite at Horsepool and lower cost meteorological measurements (Davis VantagePro) at some other sites. In addition, we downloaded meteorological data collected by others from the EPA AQS database and from mesowest.utah.edu.

Spatial Data Analysis

Data from all sites were visualized and interpolated using ArcGIS software. Relationships among measured parameters and a variety of spatial variables were investigated using ArcGIS and SPSS software. The following variables were considered in terms of proximity to study sites (5, 10, 15, 20, 25, or 50 km radius): • Number of producing wells (oil, gas, both), • Amount of oil and gas production (February 2013), • Number of drill rigs,

• Number of compressor stations and gas plants (with estimated NOx and VOC emissions), • Number and surface acres of produced water ponds, • Human population, • Elevation, and • Difference between elevation of site and of surrounding terrain.

These parameters were analyzed in correlation and multiple regression analyses to determine the best predictors of observed ozone and precursor concentrations.

3 Results and Discussion

Ozone Concentrations and Distribution

Measurements show that, relative to other years, winter 2012-13 had the highest number of ozone exceedance days, highest maximum concentrations of ozone, and largest affected area of elevated ozone. Five research sites experienced more than 30 days of exceedances of EPA’s National Ambient Air Quality Standard (NAAQS)1. The only sites in the Basin that had fourth highest daily maximum ozone concentrations below 75 ppb were those more than 2000 m above sea level (Tables 1 and 2).

In contrast, air quality monitoring stations outside of the Uintah Basin but within the intermountain West, including those within Wyoming’s Upper Green River Basin, did not experience elevated ozone (Table 2), despite the existence of strong, persistent temperature inversions in many basins during Winter 2012-13, including the Idaho Falls area, Salt Lake Valley, and Cache Valley (Hall et al., 2013). Salt Lake and Cache valleys experienced many exceedances of EPA’s NAAQS for particulate matter smaller than 2.5 microns (PM2.5; data not shown), but neither valley had elevated ozone. The Upper Green River Basin did not have persistent snow cover during winter 2012-13, which is likely the reason elevated ozone was not observed there.

The Uintah Basin experienced seven distinct inversion episodes during which ozone at multiple monitoring stations exceeded the NAAQS (Figure 1). Of these, a 13-day episode that began 15 January and continued through 27 January was the longest. Episodes in January and early February tended to be longer, whereas episodes in late February and early March were shorter and exhibited more rapid increases in ozone concentrations. The highest ozone concentrations of the study period were observed in early March, while the greatest number of sites exceeded the NAAQS in late January. Adequate snow cover and a series of prolonged inversion conditions (Figure 2) allowed for the observed ozone production. As soon as temperatures warmed enough to melt the snow from the Basin, which occurred around 10 March, significant ozone production ceased.

1 The NAAQS for ozone is 75 ppb, calculated over three calendar years as the average of the fourth highest 8-hr average daily maximum ozone concentration, though any measured concentration above 75 ppb commonly is referred to as “exceedance.”

4 Table 2. Statistics for 8-hour average ozone concentrations at sites around the Uintah Basin and at select sites in Utah and the intermountain West. Sites within 4th Highest Days of Uintah Basin Mean Median Max Min Daily Max Exceedance Dinosaur NM 45.8 41.5 126.0 2.0 113.4 26 Duchesne 34.7 34.1 112.4 2.6 87.1 12 Fruitland 39.8 41.6 64.5 5.8 60.5 0 Gusher 58.9 53.6 129.7 13.7 117.6 38 Horsepool 57.8 49.7 139.0 17.0 131.5 41 Lapoint 65.0 59.9 115.6 30.8 109.4 26 Little Mountain 49.4 49.7 72.3 29.3 66.9 0 Mountain Home 46.8 46.6 74.7 14.9 71.0 0 Myton 52.6 48.5 109.4 15.6 97.7 17 Ouray 47.7 39.9 141.6 6.1 132.4 39 Rabbit Mtn. 44.6 43.0 107.3 18.1 82.5 8 Rangely 38.2 35.8 106.1 6.1 91.0 11 Red Wash 50.1 42.5 124.0 11.4 114.0 36 Roosevelt 44.9 40.1 110.8 9.8 104.0 29 Sand Wash 55.0 49.5 127.5 11.1 122.0 33 Seep Ridge 48.7 47.3 95.1 14.9 80.8 8 Seven Sisters 47.4 41.1 152.0 4.0 137.7 31 Vernal 37.1 33.4 114.9 5.4 102.1 22 Wells Draw 49.5 45.8 131.5 15.1 108.0 26 Whiterocks 55.9 54.8 95.7 31.8 86.8 7 Sites outside 4th Highest Days of Uintah Basin Mean Median Max Min Daily Max Exceedance Boulder (WY) 31.8 32.4 49.8 5.8 47.6 0 Idaho Falls 35.1 35.7 51.2 9.2 49.0 0 Logan 17.5 16.8 44.0 2.0 43.3 0 Meeker 39.7 40.6 59.1 20.8 55.9 0 Price 34.9 35.3 57.8 8.4 51.1 0 Rifle 22.0 21.5 51.6 2.0 46.6 0 Salt Lake City 15.9 14.0 47.8 2.0 46.0 0

Figure 1. Time series of 8-hour average ozone concentrations at all monitoring sites in the Uintah Basin, winter 2012-13. EPA NAAQS of 75 ppb is shown as a red dashed line.

Figure 2. Time series of average snow depth from five stations in the Uintah Basin, pseudo-lapse rate for the Basin, 8-hour average ozone at Ouray, and average total daytime UV-A and UV-B radiation (average during daytime hours of the sum of upwelling and downwelling UV-A and UV-B) at Horsepool, winter 2012-13. The pseudo-lapse rate was derived from the change in temperature with elevation at surface meteorological stations in the Basin. The dashed black line indicates a lapse rate of zero and an ozone concentration of 75 ppb. A more negative lapse rate indicates a stronger inversion.

6 Highest ozone concentrations were observed consistently in the area of the Basin that lies south of Vernal containing a high density of natural gas wells (Figures 3, 4, 5), though ozone distribution exhibited unique characteristics during different inversion episodes. Figure 3 shows that for the entire study period, high elevation sites and sites on the margins of the Basin had the lowest ozone concentrations, while lower elevation sites and sites in the area south of Vernal had the highest.

During the 13-day inversion episode that culminated on 26 January, elevated ozone concentrations extended far beyond the Ouray-Horsepool-Seven Sisters area south of Vernal, and eight-hour average ozone for that day exceeded 100 ppb at ten sites, including sites on the edge of the Basin like Rangely and Rabbit Mountain (Figure 4). In contrast, during a shorter inversion episode in early March, only four sites exceeded 100 ppb, and many sites on the margins of the Basin experienced no exceedances of the NAAQS for ozone (Figure 5).

Figure 3. Fourth-highest daily maximum 8-hour average ozone concentrations for all sites in the Uintah Basin, winter 2012-13. The black line on the color scale indicates 75 ppb.

Figure 4. Maximum 8-hour average ozone concentrations for all sites in the Uintah Basin, 26 January.

Figure 5. Maximum 8-hour average ozone concentrations for all sites in the Uintah Basin, 1 March.

Ozone concentrations during inversion episodes were strongly correlated with elevation, were correlated with proximity to oil and gas operations (Figure 6), and were less consistently correlated with other spatial metrics (Table 3). The correlation of ozone with oil wells was much weaker than the correlation with gas wells. Gas wells tended to be in areas of lower elevation, while oil wells did not (R2

8 = 0.26, p = 0.02 for relationship between average elevation in a 25 km radius around study sites and the number of producing gas wells in the same radius), which could at least partly explain the better correlation of ozone with gas production. The average day-night difference in ozone concentration at a given site was significantly correlated with elevation and the number of people living nearby (Table 3).

Figure 6. Relationship between fourth highest daily maximum 8-hr average ozone concentration for winter 2012-13 and average elevation within 10 km of a monitoring station (in blue) and number of producing oil and gas wells within 10 km of a monitoring station (in red). Linear regression lines and Pearson R2 values are also shown.

Correlations of ozone with elevation and oil and gas production metrics were often weaker when maximum ozone on individual high ozone days, rather than ozone statistics for the whole winter season, was used (Table 3). For example, proximity to oil and gas production was not a significant predictor of ozone concentrations on 6 February, a day with high ozone, but it was a significant predictor on 26 January and 1 March, the days with the highest ozone values of the entire season. Correlations on 26 January and 1 March were weaker than those for the whole season. It is probable that meteorological conditions (especially air transport) on individual high ozone days obscured the impact of proximity to oil and gas production on ozone concentrations.

Correlations for spatial parameters were calculated within 5, 10, 15, 20, 25, and 50 km radii from sites, and the radii with strongest correlations are shown in Table 3. In general, correlations at different radii were similar to those shown, but correlations with oil and gas metrics tended to be strongest at larger radii, while correlations with population and average elevation were stronger at smaller radii.

We also explored whether the difference between the study site elevation and the average elevation of the surrounding terrain was a useful predictor of ozone concentrations. For example, the Sand Wash site was deep in the mouth of Desolation Canyon, and the average terrain in a 10 km radius around the site was 218 m higher than the elevation of the site itself. We hypothesized that sites surrounded by higher terrain would have stronger local inversion characteristics, perhaps leading to increased ozone production. No significant correlations were observed, however, between ozone concentrations and

9 the difference between site elevation and surrounding terrain elevation, indicating that sites in canyons or river bottoms were not more prone to ozone production than nearby areas.

Table 3. Pearson correlation coefficients (r) for relationships between ozone and spatial parameters. “N.S.” Indicates that the correlation was not significant (α = 0.05).

Site Average # Oil & # Oil # Gas Monthly Monthly

Elevation Elevation Gas Wells Wells Wells Oil Prod. Gas Prod. Population Radius from site -- 10 km 50 km 50 km 50 km 50 km 50 km 5 km Highest 8-hr -0.79 -0.89 0.77 N.S. 0.66 N.S. 0.65 N.S. ozone 4th highest 8-hr -0.76 -0.90 0.75 N.S. 0.64 N.S. 0.63 N.S. ozone # Exceedance -0.75 -0.90 0.74 N.S. 0.62 N.S. 0.61 N.S. days Day-night -0.61 -0.54 N.S. 0.53 N.S. 0.52 N.S. 0.56 difference ozone 26 Jan max. 8-hr -0.84 -0.93 0.67 N.S. 0.59 N.S. 0.57 N.S. ozone 6 Feb max. 8-hr N.S. -0.50 N.S. N.S. N.S. N.S. N.S. N.S. ozone 1 Mar max. 8-hr -0.61 -0.76 0.74 N.S. 0.65 N.S. 0.65 N.S. ozone

Multiple linear regression was employed to better understand the relationship between ozone and spatial parameters (as in Lyman and Gustin, 2009). In multiple regression analysis, a linear regression equation that incorporates more than one independent variable is used to predict the dependent variable, and the predictive value of independent variables can be individually assessed. Many of the independent variables considered in this study are correlated with each other, and the multiple regression method is able to show the value of each variable in predicting ozone concentration without the obscuring effects of this inter-correlation.

All parameters listed in the Methods section were utilized in the linear regression analysis. An iterative process of including and excluding parameters was employed to determine whether each possible independent variable added additional predictive power to the multiple regression model. Ultimately, elevation was found to be the best predictor of ozone concentration. The exact elevation of study sites was less useful, however, than the average elevation of the terrain surrounding the study sites, and average elevation in a 10 km radius of sites was the best predictor (Figure 6, Table 3).

While average elevation in a 10 km radius was able to explain 80% of the variability in the fourth highest 8-hour average ozone concentration among study sites (i.e., R2= 0.80), adding the number of producing oil and gas wells within 10 km of study sites with average elevation in a multiple regression model added 10% more predictive power (i.e., R2 = 0.90). In other words, with only information about elevation and

10 proximity to oil and gas activity, ozone concentrations during winter 2012-13 can be predicted at a given location in the Basin with 90% accuracy.

The spatial distribution of compressor stations, gas plants, and produced water impoundments was strongly correlated to that of oil and gas wells, and including these facilities in the analysis did not improve the quality of the regression. Oil production and gas production were worse predictors of ozone concentrations than the number of oil and gas wells. Using the number of gas wells within 10 km, rather than the number of oil and gas wells together, was as useful as the number of oil and gas wells at 2 2 predicting ozone (R = 0.90), but using the number of oil wells within 10 km was not (R = 0.81; proximity to oil wells variable was not significant, with p = 0.45). This finding could indicate that emissions associated with gas production are more important in ozone production than emissions associated with oil production, either because natural gas-related emissions are more reactive, or because they are more abundant on a per-well basis.

While correlations of ozone with proximity to oil and gas production were strongest when larger radii were used, in the multiple regression analysis the number of wells within 10 km was the strongest predictor of ozone concentrations. While average elevation within 10 km and number of producing wells within 10 km explained 90% of the variability in fourth highest ozone, substituting the number of producing wells within 50 km allowed for explanation of only 81% of variability, and the number of wells was not a significant independent variable (p = 0.42).

Located in Colorado on the eastern edge of the Uintah Basin, the Rangely site, which is regulatory, has three years of ozone data with an average greater than 75 ppb. Some have speculated that high ozone in Rangely is due to transport of ozone and precursors from Utah (Webb, 2011). While the multi-year Uintah Basin Ozone Study (UBOS) certainly has focused on Utah (e.g., maps of oil and gas production included in the 2012 UBOS final report only showed wells in Utah), a dense field of 668 producing oil and gas wells exists within 15 km of the Rangely monitoring station (Figure 7). We hypothesize that ozone experienced in Rangely is due to a combination of local (i.e., within Colorado) and regional (i.e., within the Uintah Basin) precursor sources.

Figure 7. Producing oil and gas wells in and around the Uintah Basin. Wells were considered “producing” if they reported oil or gas production during February 2013.

Distribution of Nonmethane Hydrocarbons

Concentrations of NMHC2 increased during inversion episodes, though the distribution of NMHC remained mostly the same. The highest NMHC concentrations were consistently observed at the Seven Sisters site, which is located in an area of intense natural gas production south of Vernal (Figures 8 and 9). At the height of an inversion episode on 6 February, the distributions of alkanes and aromatic compounds were similar, with highest concentrations at the Seven Sisters and Horsepool sites, followed by Roosevelt, and then by Vernal and Wells Draw (Figure 9). Fruitland, a site remote from oil and gas activity on the eastern edge of the Uintah Basin, consistently had the lowest NMHC concentrations. This spatial distribution of NMHC was similar to observations during non-inversion conditions during winter 2011-12 (Lyman et al., 2013).

2 The term “VOC” usually omits ethane since it is less reactive than most other volatile organics, while “NMHC” refers to all nonmethane organics, regardless of reactivity.

Figure 8. Sum of all alkanes measured on 1 February (the start of an inversion episode) at six sites around the Uintah Basin.

Figure 9. Sum of all alkanes measured on 6 February (day of highest ozone during the inversion episode noted in previous figure) at six sites around the Uintah Basin.

The ratio of alkanes to aromatics was higher at Wells Draw, a site in an area of intense oil production, than at Seven Sisters, a site in an area of mostly gas production, echoing observations made during non- inversion conditions in winter 2011-12 (Lyman et al., 2013). Alkene concentrations were also highest at Wells Draw (Figure 11). The source of higher alkenes in this oil producing region is uncertain, though fuel combustion is known to be an important source of alkenes (Doskey et al., 1992), and pump jack

13 engines in this area may be a significant alkene source. Methanol distribution was similar to alkane distribution (Figure 12).

Figure 10. Sum of all aromatics measured on 6 February (day of highest ozone during an inversion episode) at six sites around the Uintah Basin.

Figure 11. Sum of all alkenes measured on 6 February (day of highest ozone during an inversion episode) at six sites around the Uintah Basin.

Figure 12. Methanol measured on 6 February (day of highest ozone during an inversion episode) at five sites around the Uintah Basin.

Different NMHC have different reactivities (i.e., different ability to produce ozone). The ability of VOC to produce ozone can be measured via Maximum Incremental Reactivity (MIR; Carter, 2009). MIR uses box model simulations parameterized to maximize the sensitivity of ozone production to changes in NMHC concentrations. An MIR for a compound is the unit change in ozone production that occurs with a corresponding change in that compound. Since NMHC speciation is different at different sites around the Uintah Basin, MIR can provide a more useful means than NMHC abundance alone to determine the potential of NMHC to form ozone at various study sites. Ideally, MIR scales should be developed using meteorological and other conditions specific to the study area, but this is not yet feasible for the Uintah Basin because of uncertainties in wintertime ozone chemistry. In this work, we use the MIR scale from Carter (2009) as an approximation for MIR in the Uintah Basin during winter inversion conditions.

In Table 4, NMHC concentrations and corresponding MIR are shown for 6 February, the highest ozone day during an inversion episode that occurred during the first eight days of February. MIR are shown in units of µg of ozone produced per cubic meter of air. These values were derived by multiplying the concentration of individual compounds (units of µg per m3 of air) by the MIR for that compound (units of µg of ozone produced per µg of the organic compound under consideration) and indicate the maximum amount of ozone that could be produced from that amount of NMHC. Two µg O3 m-3 are approximately equivalent to 1 ppb of ozone.

15 Table 4. NMHC concentrations and MIR for six sites in the Uintah Basin, 6 February. Wells Seven Horsepool Vernal Draw Fruitland Sisters Roosevelt Concentrations Alkanes ppbC 3847.9 1428.4 2829.7 168.5 7679.9 3548.5 Alkenes ppbC 14.4 22.7 34.6 <2.0 21.8 20.2 Aromatics ppbC 105.0 54.8 36.0 7.2 251.9 107.3 Alkanes ppbv 2283.9 847.4 1699.4 100.7 4519.6 2117.1 Alkenes ppbv 9.9 14.3 19.5 <1.0 16.9 12.4 Aromatics ppbv 55.4 28.2 18.4 3.9 134.1 56.5 Methanol ppbv 23.4 <5 16.2 <5 57.5 -- Ethane ppbv 202.4 65.6 145.9 8.1 370.9 153.7 Maximum Incremental Reactivities

All NMHC µg O3 m-3 1917.3 827.6 1406.6 86.7 4076.6 1839.6

Alkanes µg O3 m-3 1649.8 633.4 1206.0 74.9 3405.5 1505.4

Alkenes µg O3 m-3 51.5 85.7 119.2 <10 105.2 104.7

Aromatics µg O3 m-3 203.1 109.3 69.3 11.8 541.9 230.3

Methanol µg O3 m-3 15.2 <4 10.5 <4 37.4 --

Ethane µg O3 m-3 202.4 65.6 145.9 8.1 370.9 153.7 Alkanes % of total 86.05 76.54 85.74 86.42 83.54 81.83 Alkenes % of total 2.69 10.36 8.48 <10 2.58 5.69 Aromatics % of total 10.59 13.21 4.93 13.58 13.29 12.52 Methanol % of total 0.79 <0.5 0.75 <0.5 0.92 -- Ethane % of total 10.56 7.93 10.37 9.31 9.10 8.36

At every site, including Fruitland, NMHC was comprised mostly of alkanes, and alkanes made up the majority of total MIR. In other words, the majority of ozone at the sites can be expected to have been produced from reactions involving alkanes. The percentage of total MIR attributable to aromatics and alkenes varied somewhat among sites. Aromatics contributed more than 10% of MIR at all sites except Wells Draw, and Vernal, Roosevelt, and Wells Draw were the only sites with more than 3% of MIR due to alkenes. Methanol contributed less than 1% of total MIR in all cases, while ethane, often considered too unreactive to regulate (EPA, 2003), made up 7-11% of total MIR at the study sites.

Because such a large percentage of NMHC in the Uintah Basin is alkanes (95-97% as ppbC at the six sites in this study), these compounds make up the vast majority of total MIR, even though they are low on the MIR scale relative to alkenes and most aromatics. Thus, while emissions controls that focus on aromatics and alkenes because of their high reactivity may produce more benefit at a lower cost, focus on highly reactive VOC alone is not likely to be adequate to control ozone pollution in the Basin.

16 Though only six distributed NMHC monitoring stations existed during winter 2012-13, statistically significant correlations were observed with a number of spatial and other variables. Figure 13 shows that the same predictors of ozone concentrations (area average elevation and proximity to oil and gas activity) were also strong predictors of NMHC. Alkanes, aromatics, and methanol were all strongly correlated with each other, with total NMHC, and with ozone (Table 5; Figure 14). Alkenes, on the other hand, exhibited few significant correlations, owing to their dramatically different distribution relative to other NMHC categories (Figure 11). Alkenes were correlated with proximity to oil wells, indicating that processes associated with oil production may be larger sources of alkenes than processes associated with natural gas production.

Figure 13. Relationship between 1-8 February average total NMHC concentration and average elevation within 25 km of monitoring stations (in blue) and number of producing oil and gas wells within 15 km of monitoring stations (in red). Linear regression lines and Pearson R2 values are also shown.

17 Table 5. Pearson correlations (r) for average NMHC concentrations for 1-8 February at the six sites shown in Table 4. “N.S.” Indicates the correlation was not significant (α = 0.10).

NMHC Alkanes Alkenes Aromatics Methanol (ppbC) (ppbC) (ppbC) (ppbC) (ppbv) Alkanes (ppbC) 0.99 -- N.S. 0.80 0.96 Alkenes (ppbC) N.S. N.S. -- N.S. N.S. Aromatics (ppbC) 0.82 0.80 N.S. -- 0.81 Methanol (ppbv) 0.95 0.96 N.S. 0.81 -- Highest 8-hr ozone 0.92 0.90 N.S. 0.81 0.81 4th highest 8-hr ozone 0.92 0.91 N.S. 0.87 0.86 26 Jan max. 8-hr ozone 0.81 N.S. N.S. 0.91 N.S. 6 Feb max. 8-hr ozone 0.84 0.86 N.S. N.S. N.S. 1 Mar max. 8-hr ozone 0.93 0.91 N.S. 0.93 0.95 # Wells in 15 km 0.94 0.91 N.S. 0.73 0.98 # Oil wells in 15 km N.S. N.S. 0.73 N.S. N.S. # Gas wells in 15 km 0.84 0.81 N.S. 0.85 0.96 Population in 15 km N.S. N.S. N.S. N.S. N.S. Avg. elevation in 25 km -0.83 -0.82 N.S. -0.75 -0.93

Figure 14. Relationship between 1-8 February average total NMHC concentration and fourth highest 8- hour ozone concentration at study sites. A linear regression line and Pearson R2 value is also shown.

18 Distribution of Oxides of Nitrogen

Figures 15 through 17 show the distribution of NOx during three inversion episodes. NOx tended to be highest in populated areas and areas with more natural gas production, especially the Seven Sisters site, which is in the area of maximum well density in the Basin and is near several large compressor stations and gas plants.

In general, NOx concentrations were highest during January inversion episodes and decreased as the winter progressed. The lapse rate in the Uintah Basin was most negative (indicating that inversions were stronger) in January and became less negative as the winter progressed (Figure 2). The stronger inversions in early winter likely were more effective at trapping ozone precursors, while stronger solar radiation later in winter likely inhibited inversion formation and allowed for more dilution of emitted precursors.

Figure 15. Average NOx at ten sites during an inversion episode, 20-26 January.

Figure 16. Average NOx at twelve sites during an inversion episode, 30 January through 5 February. NOx concentrations shown for Wells Draw and Seven Sisters are derived from passive Radiello samplers.

Figure 17. Average NOx at ten sites during an inversion episode, 1-3 March.

NOx concentrations at most sites were collected with instruments that utilize molybdenum oxide converters to transform NO2 to NO. Because these converters also transform some NOy (i.e., the sum of all reactive nitrogen compounds, including NO and NO2, HNO3, HONO, particulate nitrogen, and other compounds) to NO, the NO2 and NOx values obtained from these instruments were biased (high). Since

20 the NOx analyzer at Horsepool utilized a photolytic NO2 converter that does not transform NOy, NO2 concentrations at that site were not biased.

Since different sites may not have the same ratio of NOx to NOy, correcting NOx values collected with molybdenum converters to true NOx, or vice versa, is problematic. Fortunately, molybdenum converter NOx measurements were collected at Horsepool by the University of Colorado (CU research group of D. Helmig; instrument provided by Utah Department of Environmental Quality) from 1 through 18 February, and we were able to conduct a comparison of the two methods. The molybdenum converter- based instrument pulled air from 2 m above ground, while the USU instrument pulled from 4 m above ground, and the two inlets were about 50 m apart. Figures 18 and 19 show that NO concentrations measured by the two systems were comparable, but NOx concentrations were very different and were relatively weakly correlated. Figure 20 shows that the molybdenum converter-based NOx measurement was better correlated with NOy than with true NOx.

NOy concentrations at Horsepool, and presumably at sites throughout the Uintah Basin, were much higher than true NOx during inversion episodes, and the NOx concentrations presented in Figures 15 through 17 should be interpreted as NOx + some portion of NOy. For Figures 15 through 17, NOx at Horsepool was calculated based on the relationship between measured NOy and molybdenum converter NOx. In Figure 16, NOx for the Wells Draw and Seven Sisters sites was calculated based on the relationship between true NO2 measurements collected by Radiello passive samplers and molybdenum converter NOx measurements at Fruitland, Vernal, and Horsepool.

Figure 18. Comparison of NO measured by a photolytic NOx analyzer (USU) to a molybdenum converter- 2 based NOx analyzer (CU). The linear regression curve, R value, and slope of the relationship are also shown.

Figure 19. Comparison of NOx measured by a photolytic NOx analyzer (USU) with a molybdenum 2 converter-based NOx analyzer (CU). The linear regression curve, R value, and slope of the relationship are also shown.

Figure 20. Comparison of NOy (USU) with NOx measured by a molybdenum converter-based NOx analyzer (CU). The linear regression curve, R2 value, and slope of the relationship are also shown.

NOx concentrations at Vernal and Roosevelt were among the highest observed at any study site and exhibited the largest diurnal changes, with highest NOx in the morning hours (Figure 21), likely due to diurnal traffic patterns in these cities. Red Wash, located only 250 m from a well-traveled highway, also

22 showed significant diurnal variability, but the peak NOx at Red Wash was at midday. NOx also peaked at midday at the Ouray and, to some extent, Horsepool sites, similar to observations made during winter 2011-12 (Lyman et al., 2013). A counter at the Horsepool site during winter 2011-12 observed highest traffic at midday.

Figure 21. Diurnal change in NOx concentrations at 10 sites around the Uintah Basin during the 20-26 January inversion episode.

Table 6 and Figure 22 show that the spatial distribution of NOx was, in some instances, correlated with the spatial distribution of ozone. This relationship was not significant for the distribution of NOx over the entire study period, but was significant for NOx distribution during some inversion episodes. NOx distribution was strongly correlated with the difference between day and night ozone concentrations, probably because ozone can be destroyed by NO at night, and sites with high NOx also tend to be sites with higher nighttime NO. Correlations between NOx concentrations and proximity of study sites to oil and gas operations were either not significant or weak, but NOx was correlated with population and elevation.

NOx concentrations during the 31 January through 5 February period were not significantly correlated with total NMHC measured during the same period (p = 0.391), but they were correlated with aromatics 2 (R = 0.61, p = 0.07). When 1-8 February average NMHC and 31 January through 5 February NOx were used as independent variables to predict 6 February ozone in a multiple linear regression model, the NOx variable was not significant (p = 0.77). The NOx variable also was not a significant variable in prediction of the fourth highest 8-hour average ozone in a multiple regression with NMHC, which indicates that the correlation of NOx distribution with ozone concentrations in Table 6 may not be indicative of a strong causal relationship between NOx and ozone, but could instead be due to a correlation between NOx and other variables such as elevation. Also, the tendency of NOx to destroy ozone under some conditions may have confounded this relationship.

Table 6. Pearson correlations (r) of average NOx during different periods with ozone and spatial metrics. “N.S.” Indicates that the correlation was not significant (α = 0.10).

Avg. NOx Avg. NOx Avg. NOx Avg NOx 20-26 Jan. 1-3 Mar. 31 Jan.-5 Feb. 15 Jan.-10 Mar. Highest 8 -hr ozone N.S. 0.63 0.62 N.S. 4th highest 8-hr ozone N.S. 0.73 0.65 N.S. # Exceedance days N.S. 0.77 0.57 0.56 Day-night difference ozone 0.88 0.78 0.81 0.86 26 Jan max. 8-hr ozone N.S. N.S. 0.68 N.S. 6 Feb max. 8-hr ozone N.S. 0.67 0.50 N.S. 1 Mar max. 8-hr ozone N.S. N.S. 0.59 N.S. # wells in 5 km N.S. N.S. 0.51 N.S. Population in 5 km 0.83 0.64 0.57 0.82 Avg. elevation in 5 km -0.65 -0.79 -0.64 -0.64

Figure 22. Relationship between 31 January through 5 February molybdenum converter-based NOx concentration (contains some NOy) and fourth highest 8-hour ozone concentration at 12 study sites. A linear regression line and Pearson R2 value are also shown.

24 Influence of Transport Patterns on Ozone During an Inversion Episode

UBOS study participants and others operated dozens of meteorological stations in and around the Uintah Basin during winter 2012-13. Meteorology, particularly wind speed and direction, was an important determinant of ozone production in the Basin. Figures 23, 24, and 25 show ozone concentrations and wind vectors prior to and during an inversion episode that occurred from 31 January through 8 February.

Figure 23 shows wind conditions and the low ozone concentrations of 28 January, just two days after one of the season’s highest ozone days in the Basin. A storm front arrived on 27 and 28 January and was associated with relatively high winds in many parts of the Basin on 28 January, as shown. Winds at high elevation sites during this period were from the south. Wind speeds within the Basin were lower on subsequent days, allowing another inversion to form and ozone concentrations to rebuild. By the afternoon of 6 February (Figure 24), ozone concentrations exceeded 100 ppb at many sites. Wind at high elevations continued to be from the south throughout this period, and wind at lower elevation sites in the Basin was light and variable (Figure 24). On 8 February, however, wind at high elevation sites changed directions and blew from the north (Figure 25). Following this synoptic scale change, wind at low elevation sites within the Basin continued to be light and variable; nevertheless, a partial mix-out of ozone from the Basin occurred, and ozone concentrations dropped to less than 100 ppb at all sites.

Figure 23. Daily maximum ozone concentrations and wind vectors from surface sites on 28 January during a stormy period between inversion episodes. Arrows indicate wind direction. The black line on the ozone color scale indicates 75 ppb.

Figure 24. Daily maximum ozone concentrations and wind vectors from surface sites during an inversion episode, 6 February.

Figure 25. Daily maximum ozone concentrations and wind vectors from surface sites, 8 February.

Seasonal Changes in Ozone and Precursor Chemistry

While inversions became weaker (less negative lapse rate) as the winter ozone season proceeded from January through March, maximum ozone concentrations increased and took fewer inversion days to reach maxima (Figure 2). Figures 26, 27, and 28 show average ozone and precursor concentrations for each hour of the day during three inversion episodes that occurred in January, February, and March. Concentrations of ozone precursors decreased with each successive episode, but the amount of ozone produced each day (i.e., the difference between morning and afternoon ozone concentrations) increased. Average total NMHC (TNMHC) was 37% lower during 1-3 March than during 20-26 January, and NO2 was 63% lower, but daily ozone production was 64% higher during 1-3 March than during the earlier period. The average daytime total UV radiation (average of the sum of incoming and outgoing UV-A and UV-B for daylight hours) was 51% higher during 1-3 March, and daily maximum temperature was 10.5 °C higher, probably accounting for the observed increase in daily ozone production.

Figure 26. Diurnal average concentrations of ozone and precursors at Horsepool, 20-26 January.

Figure 27. Diurnal average concentrations of ozone and precursors at Horsepool, 4-6 February.

Figure 28. Diurnal average concentrations of ozone and precursors at Horsepool, 1-3 March.

Ozone production efficiency is the ratio of production of ozone and NO2 (also called “odd oxygen” or Ox) to the removal of NOx; more simply, it is the number of ozone molecules produced for each molecule of precursor consumed (Lin et al., 1988; Sillman, 1999). The linear regression slope of the relationship between ozone and NOz (NOz is the sum of NOx reaction products, or NOy – NOx) and between Ox and NOz have often been used to approximate ozone production efficiency, since the compounds that comprise NOz are end products of NOx photochemistry (e.g., Trainer et al., 1995). A higher ozone to NOz slope indicates that more ozone is produced per molecule of NOx consumed. Deposition or other loss of NOz from the atmosphere is not accounted for in this metric, however, and a higher slope could also mean that NOz is being removed more efficiently.

Figure 29 shows the slope of the relationship between ozone and NOz at Horsepool on days with maximum ozone greater than 90 ppb, calculated as described by Chou et al. (2009). This figure provides evidence that ozone production efficiency increases from early to late winter in the Uintah Basin, probably because temperature and available solar energy increases. Since some of the increase in the ozone to NOz slope over time could be due to increased NOz deposition, the values in Figure 29 should be used with caution. The trend of increasing efficiency in ozone production from early to late winter, however, is likely to be robust.

Figure 29. Slope of the linear regression relationship between Ox and NOz at Horsepool, January-March. Only days with maximum ozone greater than 90 ppb are shown.

Photochemical indicators are metrics based on ambient measurements that can indicate whether ozone production in a region can be more effectively reduced by VOC controls (VOC sensitive) or NOx controls (NOx sensitive) (Sillman, 1999). A number of photochemical indicators have been used, but the ratio of ozone to NOy has been most commonly used and has been verified in a number of studies (Stein et al., 2005). The rationale for use of this indicator ratio, as explained by Sillman (1999), is that the major sinks for radicals in the ozone photochemical process are peroxides and NOz, especially HNO3. Radical production exceeds the rate of NOx emissions in NOx sensitive regimes, while the opposite is true in VOC sensitive regimes (see also Kleinman, 1994). In cases with more radicals than NOx (NOx sensitive), less NOx is available to react with radicals to produce HNO3 and other NOz, so more peroxides are produced relative to HNO3. In the opposite case, more HNO3 and other NOz can be expected. Thus, the ratio of peroxides to NOz is a relative indicator of NOx versus VOC sensitivity. Ozone production is typically proportional to peroxide production, and since direct peroxide measurements are rarely available, the ratio of ozone/NOy or ozone/NOz is often used as a proxy.

Unfortunately, cutoff values or ranges that indicate NOx versus VOC sensitivity in the ozone/NOy ratio or other indicators can be different for different regions (Stein et al., 2005), and caution must be exercised in using cutoff values from other studies, especially when emissions or chemistry may be unique, as with wintertime ozone in the Uintah Basin. Determination of appropriate cutoff values to distinguish whether NOx or VOC controls would be more effective is best achieved through utilization of a photochemical model, and indicator ratios can be useful for verifying model-based determinations of NOx or VOC sensitivity.

Since a verified photochemical model is not yet available for the Uintah Basin, the value of the ratio of ozone/NOy that would indicate NOx versus VOC sensitivity cannot be reliably determined. However, the transition between these two states is gradual, and higher ozone/NOy ratios always indicate movement towards NOx sensitivity (Sillman, 1999). Figure 30 shows that the ozone/NOy ratio steadily increased from early to late winter in the Uintah Basin. The ozone/NOz ratio, another common indicator, showed

30 a similar trend (not shown). Together, these indicators provide evidence that more radicals are produced during late-season ozone events in the Uintah Basin (which also would tend to increase ozone production efficiency, as discussed above) and that early-season events are more likely to be VOC sensitive than late-season events.

In previous studies, the transition region between NOx and VOC sensitive regimes has been characterized by ozone/NOy ratios between 5.6 and 15 (Stein et al., 2005). The ozone/NOy ratio at Horsepool is within this range by mid-February. One possible scenario is that VOC controls in the Uintah Basin will be more effective at reducing ozone than NOx controls early in the year, but that NOx controls will have increasing effectiveness as the winter season proceeds. Seasonal transitions from NOx sensitivity in warm seasons to VOC sensitivity in cool seasons have been observed by others, and are likely due to decreased UV radiation and water vapor concentration, both factors that limit radical production (e.g., Jacob et al., 1995).

Figure 30. Ratio of ozone to NOy at Horsepool, January-March. Only days with maximum ozone greater than 90 ppb are shown.

Interannual Variability in Wintertime Ozone

Figure 31 shows a time series of ozone concentrations at several sites in the Uintah Basin from July 2009 through 15 March 2013. The Ouray and Red Wash air quality monitoring stations began operation in July 2009. During winter 2009-10, both sites experienced multiple exceedances of the NAAQS for ozone (75 ppb). Subsequently, regulatory monitors in Roosevelt, Vernal, and Rangely were added. As Figure 31 shows, NAAQS exceedances have been observed during three of the four years in the Uintah Basin for which continuous ozone monitoring data is available. The Utah Department of Environmental Quality also measured ozone in Vernal during 2006 and 2007. Those data are not publicly available and are not included here, but no wintertime NAAQS exceedances were measured during that period.

Figure 31. Time series of daily maximum 8-hour average ozone concentration at five sites in the Uintah Basin, July 2009-March 2013. The red dashed line shows 75 ppb, the EPA NAAQS for ozone.

Table 7 summarizes ozone statistics from the five sites shown in Figure 31 for each of the years that data are available. Data are organized by calendar year rather than by winter season, since summertime exceedances have also occurred (albeit rarely), and since a nonattainment designation, if made by EPA, will be based on calendar years. The Rangely site has three calendar years of regulatory ozone data, and the three-year average of the fourth highest daily maximum 8-hour average ozone concentration (i.e. “ozone design value”) for the site is 78.0 ppb. The Ouray and Red Wash sites are not regulatory, but their current ozone design values are 106.5 and 92.9 ppb, respectively. Ozone data has been collected in Vernal and Roosevelt since January 2011, but 2011 data were collected by USU and are not regulatory. These sites, as a result, only have two years of regulatory data. The two-year averages of the fourth highest regulatory value for Vernal and Roosevelt are 83.4 and 85.5, respectively. In order to maintain a design value less than 75 ppb, the fourth highest 8-hour average ozone concentration for 2014 will have to be no higher than 58 ppb for Vernal and 54 ppb for Roosevelt, which is unlikely.

32 Table 7. Ozone summary statistics for five sites in the Uintah Basin over five calendar years. The Vernal, Roosevelt, and Rangely sites are regulatory; Ouray and Red Wash are not. All values shown were calculated from daily maximum 8-hour average concentrations. 4th Highest # of Mean Median Max Min Daily Max Exceedance Days 2009 Ouray 47.2 47.9 101.5 23.4 67.4 1 Red Wash 44.8 43.7 72.3 27.3 67.6 0 Vernal ------Roosevelt ------Rangely ------2010 Ouray 56.7 54.5 123.6 20.3 117.3 40 Red Wash 54.4 53.6 105.4 17.0 98.9 30 Vernal ------Roosevelt ------Rangely 42.3 42.2 67.2 11.1 58.8 0 2011 (Vernal and Roosevelt sites were operated by USU and were not regulatory) Ouray 54.0 52.8 138.6 18.1 119.6 24 Red Wash 51.6 51.8 130.2 21.3 98.3 21 Vernal 55.5 55.6 95.1 33.1 84.9 7 Roosevelt 56.2 54.7 116.3 29.3 103.6 19 Rangely 48.6 50.0 88.6 21.9 73.4 3 2012 Ouray 49.3 50.5 76.5 18.8 67.6 1 Red Wash 47.5 48.8 69.5 21.5 66.4 0 Vernal 45.7 46.8 68.9 14.5 64.8 0 Roosevelt 50.3 51.6 70.9 14.6 67.0 0 Rangely 46.7 47.4 71.9 15.9 69.6 0 2013 (through 15 March) Ouray 81.8 79.8 141.6 38.4 132.4 39 Red Wash 74.8 73.6 124.0 38.8 114.0 36 Vernal 67.2 63.4 114.9 37.5 102.1 22 Roosevelt 68.0 64.2 110.8 37.0 104.0 29 Rangely 56.4 52.4 106.1 31.7 91.0 11

33 Summary

The spatial distribution of ozone and NMHC concentrations observed during winter 2012-13 largely can be explained by elevation and proximity to oil and gas production. This finding shows that ozone production in the Basin depends on intensity of inversions (elevation dependent) and intensity of emissions. The spatial distribution of ozone was found to be more closely correlated with NMHC than with NOx , perhaps indicating that NMHC sources are more important than NOx sources for ozone production.

Ozone production efficiency increased from early to late winter, allowing ozone to be produced more rapidly during later season ozone events, even while precursor concentrations were lower. During early winter, lower temperatures and reduced sunlight increase the likelihood that the Uintah Basin is VOC sensitive (i.e., VOC controls would be most effective at mitigating ozone pollution). It is possible that the Basin transitions from VOC to NOx sensitivity as the winter proceeds.

Though ozone concentrations above 75 ppb have occurred during three of the past four winters in the Uintah Basin, only the Rangely monitoring station has the requisite three years of regulatory data to permit a nonattainment area designation. By 2014, the Vernal and Roosevelt stations are likely to meet this benchmark.

Acknowledgements

We are grateful to the Uintah Impact Mitigation Special Service District; the Utah Science, Technology, and Research Initiative; the Department of Energy; and Mr. Marc Bingham for financial support of this work. We thank the Bureau of Land Management for providing equipment for remote ozone monitoring and to Detlev Helmig and Chelsea Stevens of the University of Colorado for providing access to molybdenum converter-based NOx data they collected at the Horsepool site. We are also grateful to the Utah Department of Environmental Quality for providing space for our monitors at Roosevelt.

USE OF THE CALMET DIAGNOSTIC MODEL TO SIMULATE WINTER INVERSIONS

Michael Christiansen and Trevor O’Neil Department of Chemistry, Utah State University, Vernal, UT

Introduction

The major precursors to ozone formation—NOx and VOC—are emitted by a variety of sources. To determine the contributions to ozone production of these different sources in the Uintah Basin, meteorological models that accurately reproduce transport conditions during winter inversions are needed. A realistic meteorological model of the region will enable us to observe how pollutant transport from sources within and outside of the Basin affects ozone and precursor concentrations in the Basin.

Methods

We incorporated data from all available meteorological stations, both at the surface and aloft, from an intensive UBOS study period (28 January to 8 February) for input to the most current version (6.42) of CALMET, an EPA-approved modeling program. This program was recently employed in a similar study of ozone precursor transport in the Upper Green River Basin of southwestern Wyoming (Rairigh, 2010). We utilized a horizontal grid resolution of 1.5 km2, and a 300 × 300 km modeling domain that extends 9 to 75 km beyond each edge of the Uintah Basin. Meteorological data were extracted from stations located either inside or up to 80 km beyond this area in all directions. Table 8 lists the 68 stations used. A 3D rendering of the modeling domain is shown in Figure 32. Upper air data were extracted from sites at Salt Lake City, UT; Grand Junction, CO; Denver, CO; and Riverton, WY. Vertical layers used for the model are shown in Table 9, and range from 10 to 3,500 meters above ground level.

Table 8. Stations from which data were used in the CALMET model. Station Station Name Station Station Name 1 OURAY 35 ROOSEVELT 2 MEEKER AIRPORT 36 CHEPITA 3 VERNAL AIRPORT 37 FRUITLAND 4 HEBER AIRPORT 38 PRICE 5 BRYSON CANYON 39 PORTABLE RWIS 6 BADGER WASH 40 DRY RIDGE 7 CART CREEK 41 ROOSEVELT 8 HANNA NEAR DUCHESNE NW 24 42 LITTLE RED FOX 9 BLACK TAIL 43 MOUNTAIN HOME 10 NUTTERS RANCH 44 MOON LAKE NEAR ALTONA 11 KINGS PONIT 45 INDIAN CANYON SUMMIT 12 HORSE RIDGE 46 US-40 @STARVATION 13 UPPER SAND WASH 47 YELLWOSTONE DRAINAGE 14 YAMPA PLATEAU 48 WEST FORK BLACK SMITH 15 DIAMOND RIM 49 VERNAL 16 FIVE MILE 50 SPLIT MOUNTAIN 17 BEAR RIVER 51 UPPER P.R. CANYON 18 DRAGON ROAD 52 CURRANT CREEK PEAK 19 GREEN RIVER 53 †RANGELEY COLORADO 20 LADORE 54 NORWAY 21 HUNTER CREEK 55 DEER VALLEY 22 RATTLESNAKE BENCH 56 WILD HORSE 23 DINOSAUR NM SUCCESS 57 RIFLE 24 CRAIG MOFATT AIRPORT 58 WINTER RIDGE 25 BRUIN POINT 59 MYTON 26 HELPER @ US 6 60 RAYS VALLEY 27 CALICO 61 *SAND WASH 28 SOLDIER SUMMIT 62 *PARIETTE DRAW 29 PRICE AIRPORT 63 *SEVEN SISTERS 30 FLATTOP MOUNTAIN 64 *SEEP RIDGE 31 PINTO 65 *WELLS DRAW 32 HEBER US-40 66 *MOUNTAIN HOME 33 SALT LAKE CITY 67 *GUSHER 34 PRICE 68 *HORSEPOOL *USU Sites †Site omitted because reported data were incomplete

Figure 32. 3D elevation map of the CALMET modeling domain. Approximate locations of Vernal, Price, and Grand Junction are shown.

Table 9. Vertical layers utilized in the CALMET model. Layer Top Layer Top Layer # (m above ground level) Layer # (m above ground level) 1 10 8 500 2 20 9 750 3 60 10 1000 4 100 11 2000 5 150 12 3000 6 200 13 3500 7 350

Results and Discussion

CALPUFF Model Animations

CALMET can be coupled with the 3D-modeling software CALPUFF, which generates animated plume dispersion models. In concert with this work, we created a video of a twelve-hour plume dispersion model that uses five hypothetical source points. These are publicly available online and illustrate the value of the CALMET modeling system (https://www.youtube.com/watch?v=5lTZeAfBSR4&feature=c4- overview&list=UUpUkAZfpeUBMmA_zAAERyZw).

37 Wind Speed and Direction

Ground-level (10 m) wind field vectors are shown below in Figures 33 through 36. Each of these figures was taken from 6 February 2013, the day of highest ozone concentration during an inversion period. On this day, the CALMET-predicted wind speeds within the Uintah Basin remained low (less than 3 m/s). At 3:00 local time (Figure 33), wind within the Basin flowed primarily from the south. By 9:00 (Figure 34), wind in the Basin was low and variable, but tended to blow mostly from east to west. This pattern held, for the most part, at 15:00 (Figure 35), but by 21:00 (Figure 36), winds were from the south on the east side of the Basin, and from the west on the west side of the Basin.

Figure 33. Ground-level (10 m) wind vectors at 3:00, 6 February.

Figure 34. Ground-level (10 m) wind vectors at 9:00, 6 February.

Figure 35. Ground-level (10 m) wind vectors at 15:00, 6 February.

Figure 36. Ground-level (10 m) wind vectors at 21:00, 6 February.

We extracted ground-level (10 m) wind directions and speeds, as well as temperature, from the model at the locations of select surface meteorological stations and compared them with measured data. Table 10 shows results from two stations, Dragon Road and Sand Wash. Modeled and measured wind speed, wind direction, and temperature were well correlated at Dragon Road, but modeled and measured wind speed and direction were poorly correlated at Sand Wash. The Sand Wash site is located within Desolation Canyon, and at the 1.5 km horizontal resolution used, CALMET was not able to fully resolve terrain or wind patterns within the canyon. Also, measured wind speed was low, often zero, in Sand Wash, while modeled wind speed was never zero, further confounding the comparison.

41 Table 10. Slope, R2 value, and percent deviation for the relationship between measured and modeled wind speed, wind direction, and temperature at the Dragon Road and Sand Wash sites. Percent deviation rows show values ± 95% confidence intervals. Dragon Road Sand Wash Wind Speed Slope 1.01 -0.03 R2 1.00 0.01 Percent Deviation 0.8 ± 0.3% 20 ± 84% Wind Direction Slope 0.68 0.19 R2 0.59 0.02 Percent Deviation 7.8 ± 10.2% 9.4 ± 16.8% Temperature Slope 0.93 1.04 R2 0.85 0.92 Percent Deviation 0.0 ± 0.1% 0.3 ± 0.1%

Temperature

A surface-level temperature contour map for 10:00 local time on 7 February 2013 is shown in Figure 37. This figure shows an inversion, with areas of lowest elevation having temperatures less than 0 °C, and areas of highest elevation having temperatures up to about 10 °C. Site S49 shown in Figure 33 has a temperature of more than 20 °C, which is extremely unlikely for early February. We expect the temperature at this site is inaccurate, and it will be removed from future analyses.

To determine whether changes in temperature with height were being accurately simulated by the model, we extracted temperature from the different vertical layers of CALMET and compared them to temperature measured with a moored balloon on 6 February at Pariette Draw. Figure 38 shows the average measured and modeled vertical temperature profile for the period from 12:00 to 15:00. CALMET does simulate the presence of a temperature inversion, but the inversion occurs higher in the model than in reality, and the model predicts a decrease in temperature with height under the inversion that does not exist in the measurements, leading to a divergence of about 6 °C between measurements and the model at 275 m above ground.

Figure 37. Surface temperature contour map for 10:00, 7 February.

Figure 38. Measured and modeled vertical temperature profile for 12:00 to 15:00, 6 February at Pariette Draw.

Acknowledgements

We gratefully acknowledge funding for this project from the Uintah Impact Mitigation Special Service District.

44 References

Carter, W.P.L. (2009), Development of the SAPRC-07 Chemical Mechanism and Updated Ozone Reactivity Scales, College of Engineering, University of California, Riverside, California.

Chou, C.C., C.Y. Tsai, C.J. Shiu, S.C. Liu, and T. Zhu (2009), Measurement of NOy during campaign of air quality research in Beijing 2006 (CAREBeijing-2006): Implications for the ozone production efficiency of NOx , J. Geophys. Res., 114, D00G01, doi: 10.1029/2008JD010446.

Doskey, P.V., J.A. Porter, and P.A. Scheff (1992), Source fingerprints for volatile non-methane hydrocarbons, J. Air Waste Manage. Assoc., 42, 1437-1445.

EPA (2003), Interim Guidance on Control of Volatile Organic Compounds in Ozone State Implementation Plans, OAR-2003-0032, Environmental Protection Agency, Research Triangle Park, North Carolina.

Hall, C., R. Anderson, M. Mansfield, S. Lyman, and H. Shorthill (2013), A Survey of Basins and Valleys in the Western U.S.A. for the Capacity to Produce Winter Ozone, Office of Commercialization and Regional Development, Utah State University, Logan, Utah.

Jacob, D.J., L.W. Horowitz, J.W. Munger, B.G. Heikes, R.R. Dickerson, R.S. Artz, W.C. Keene (1995), Seasonal transition from NOx - to hydrocarbon-limited conditions for ozone production over the eastern United States in September, J. Geophys. Res., 100, 9315-9324.

Kleinman, L.I. (1994), Low and high-NOx tropospheric photochemistry, J. Geophys. Res., 99, 16831- 16838.

Lin, X, M. Trainer, and S.C. Liu (1988), On the nonlinearity of tropospheric ozone, J. Geophys. Res., 93, 15879-15888.

Lyman, S., and H. Shorthill (Eds.) (2013), Final Report: 2012 Uintah Basin Winter Ozone and Air Quality Study, CRD13-320.32, Office of Commercialization and Regional Development, Utah State University, Logan, Utah. Lyman, S.N, and M.S. Gustin, (2009), Determinants of atmospheric mercury concentrations in Reno, Nevada, U.S.A., Sci. Tot. Environ., 408, 431-438.

Martin, R.S., K. Moore, M. Mansfield, S. Hill, K. Harper, and H. Shorthill (2011), Final Report: Uinta Basin Winter Ozone and Air Quality Study, EDL/11-039, Energy Dynamics Laboratory, Utah State University Research Foundation, Logan, Utah.

Rairigh, K. (2010), Draft Upper Green River Winter Ozone Study: CALMET Database Development – Phase II, Wyoming Department of Environmental Quality, Cheyenne, Wyoming.

Sillman, S. (1999), The relation between ozone, NOx , and hydrocarbons in urban and polluted rural environments, Atmos. Environ., 33, 1821-1845.

45 Stein, A.F., E. Mantilla, M.M. Millan (2005), Using measured and modeled indicators to assess ozone-NOx -VOC sensitivity in a western Mediterranean coastal environment, Atmos. Environ., 39, 7167- 7180.

Trainer, M., B.A. Ridley, M.P. Buhr, G. Kok, J. Walega, G. Hubler, D.D. Parrish (1995), Regional ozone and urban plumes in the southeastern United States: Birmingham, a case study, J. Geophys. Res., 100, 18823-18834.

Webb, D., (2011), ‘Big-city’ ozone goes rural, The Daily Sentinel: Grand Junction, Colorado , March 24

46 WINTERTIME EMISSIONS OF HYDROCARBONS FROM PRODUCED WATER EVAPORATION FACILITIES

Seth Lyman, Marc Mansfield, Howard Shorthill, Randy Anderson, Jordan Evans, Chad Mangum, and Tate Shorthill

Introduction

Ozone concentrations exceeding EPA standards have been observed in the Uintah Basin during winter inversion conditions. Local sources that emit ozone precursors, including volatile organic compounds (VOC) and nitrogen oxides (NOX) are largely responsible for the high ozone levels (Lyman and Shorthill, 2013), yet ozone precursor emissions from many of these sources, including upstream oil and gas production equipment and facilities, are inadequately characterized. The development of effective mitigation strategies for ozone pollution cannot proceed without a better understanding of these sources.

One source of VOC is the water that is brought to the surface during oil and gas production (Clark and Veil, 2009). In the Uintah Basin, this produced water is commonly disposed of via open-air evaporation, a technique involving the storage of produced water in shallow ponds where surface evaporation can take place, either passively or by active aeration (Clark and Veil, 2009). In 2012, more than five million barrels of produced water were evaporated from the 427 surface acres of produced water ponds in the Basin (http://oilgas.ogm.utah.gov/ Data_Center/DataCenter.cfm). Locations, ownership, acreage, and status of each produced water pond in the Uintah Basin were compiled and verified by our group from data obtained from the Utah Division of Oil, Gas, and Mining; site visits; and aerial photos.

Since produced water ponds may be responsible for a significant percentage of ozone-forming emissions in the Uintah Basin, knowledge of the types and amounts of VOC emitted from these ponds is vital to the development of effective mitigation practices. Until this study, however, no measurements of produced water emissions had ever been conducted in the Basin. In 2009, the U.S. Environmental Protection Agency (EPA) released a report of measurements of VOC emissions from two different evaporation facilities in Western Colorado (Thoma, 2009), and while an important first step, the study only covered two small facilities for a few days during the summer, at best a difficult extrapolation to the winter conditions of the Uintah Basin.

To better understand emissions from produced water ponds during winter inversions in the Uintah Basin, a dynamic flux chamber was used to measure methane, methanol, and speciated nonmethane hydrocarbon (NMHC) emissions at three separate pond facilities during February and early March 2013. As a rule, VOC signifies only those hydrocarbons that are more reactive than ethane, but since ethane is known to be a major component of NMHC in local ambient air and likely contributes significantly to ozone production (Lyman and Shorthill, 2013), NMHC was an important measurement focus. Though much of the total pond surface area at the evaporation facilities was frozen during the study period, areas that regularly received water deliveries remained free of ice. Emissions were measured from both frozen and open pond surfaces, and total emissions were estimated from these and all facilities in the Basin during winter conditions.

Figure 1. Locations of produced water pond facilities in the Uintah Basin. Each pond is shown as a blue square.

Methods

Pond Facilities Descriptions

To protect the interests of produced water disposal facility operators, non-disclosure agreements were established as a precondition to access of pond facilities. The locations and operators of the sampled facilities, therefore, will not be given in this report. Three facilities in different parts of the Uintah Basin were sampled during this study, and their combined acreage comprised 17% of the total pond surface area in the Basin.

Water was trucked from well-site storage tanks to the disposal facilities and emptied into a skim tank where gravity separated the water from the residual oil (see Figure 2 for a diagram of a typical facility). During winter, well-site storage tanks were usually heated, so the produced water was warm when introduced into the skim tanks. The open skim tanks used by some facilities presented a pure oil surface directly to the atmosphere, and inclusion of this source in future studies would be valuable. As it separated from the oil, water drained from the bottom of the skim tank into an evaporation pond. During warm seasons, this water would then be distributed to other ponds at the facility where it could be sprayed into the air or cascaded down a lined dike to facilitate evaporation. During winter, transfer to other ponds was curtailed, and active evaporation methods did not occur.

48 During the winter 2012-13 study, all ponds that regularly received water from skim tanks remained at least partially unfrozen, while most of the ponds that did not remained frozen on the surface and covered with snow. Some ponds, however, were so saline that despite subfreezing air temperatures the water remained open. The salinity of produced water depends on the formation from which it was derived (Clark and Viel, 2009) and, when first extracted, typically is equivalent to that of seawater. Salinity increases, however, as evaporation takes place at the disposal facilities, with some ponds becoming hypersaline.

Some produced water facilities had a small pond (usually netted) downstream from the skim tank to catch additional residual oil, followed by a larger pond that received water from the small netted pond. Other facilities did not have this smaller intermediate pond. All facilities had oil storage tanks into which the skimmed oil was pumped.

Figure 2. Illustration of a typical Uintah Basin produced water facility sampled by our group during winter 2012-13.

Dynamic Flux Chamber Measurements

Emissions of methane, non-methane hydrocarbons, and methanol were sampled using a modified version of the EPA Emission Isolation Flux Chamber (Eklund, 1992) that currently is widely used for air toxics emissions. Dynamic flux chambers measure emissions (or deposition) as the difference in concentrations inside and outside the chamber, multiplied by the flow rate and divided by the surface area covered by the chamber. The chamber system utilized in this study connected to a laser-based methane and carbon dioxide analyzer (LGR Ultraportable Greenhouse Gas Analyzer) and stainless steel evacuated air sampling canisters, which were analyzed for C2-C12 hydrocarbons and methanol. C2-C12

49 hydrocarbons were analyzed by GC-FID according to EPA PAMS analysis protocols (EPA, 1998), and methanol was analyzed according to EPA TO-15 protocols (EPA, 1999).

The laser-based analyzer switched every two minutes between sampling air inside and outside the chamber, while a pair of evacuated canisters pulled air from inside and outside the chamber over a period of 30-45 minutes. A system of flow meters and pumps regulated air sampling, and the flow through the chamber was kept at 5 L min-1. Concentrations, flows, and other information were logged with a Campbell Scientific CR1000 datalogger at five-second intervals. Figure 3 shows a photo of the chamber on a produced water pond.

Emissions were measured from a subset of ponds at each facility and from snow-covered ground near produced water ponds. Emissions were also measured from a clean, 1/8” thick Teflon surface to assess system contamination. Teflon surface emissions were measured periodically throughout the study to verify that the chamber system did not become contaminated over time. The chamber, tubing and wires, and foam floatation device were washed periodically with soap and water to remove oil and grease.

Figure 3. USU dynamic flux chamber on a produced water pond.

Meteorological Measurements

Detailed meteorological data were collected during all emissions measurement periods. These included solar radiation; wind speed, direction, and turbulence; ambient temperature, pressure, and humidity; water temperature inside and outside the chamber; and air pressure and temperature inside the chamber.

50 Water Samples

At each of the study’s produced water facilities, water samples were collected from most ponds and analyzed for salinity, pH, methane, methanol, and a suite of other organic compounds. Samples were analyzed for methane by method RSK-175 (Kampbell and Vandegrift, 1998), for methanol by EPA Method 8015B (EPA, 1996a), and for other organics by EPA Method 8260B (EPA, 1996b). When ice was present, water was collected from under the ice.

Results and Discussion

Water Chemistry

For simplicity, pond surfaces encountered during the study were classified as (1) frozen, (2) hypersaline, or (3) recently offloaded. These three types comprised all the pond surface area at sampled facilities. Frozen and hypersaline surfaces comprised about 90% of total pond surface area at the facilities sampled. Concentrations of organic compounds in hypersaline surfaces were below limits of detection, with the exception of methane, methanol, and octane (Table 1). These surfaces were several times more saline than other types. The recently offloaded surface type tended to have the highest concentrations of most organic compounds, and water under frozen surfaces had either slightly lower or similar concentrations. Concentrations of methanol in some samples were more than two orders of magnitude higher than concentrations of other organic compounds and comprised as much as 0.5% of produced water.

51 Table 1. Average concentrations of select organic compounds in, and chemical properties of, three produced water surface types encountered during the winter 2012-13 sampling season. N.D. means not detected.

Frozen Hypersaline Recently Offloaded Methane mg L-1 0.4 0.02 0.7 Methanol mg L-1 2690 6.3 2415 Isopropyl alcohol mg L-1 6.3 N.D. 10.6 n-Butyl alcohol mg L-1 5.3 N.D. 16.3 tert-Butylbenzene mg L-1 0.0 N.D. 0.2 Acetone mg L-1 1.5 N.D. 2.2 Benzene mg L-1 4.7 N.D. 12.6 Toluene mg L-1 7.3 N.D. 15.9 Xylenes, Total mg L-1 4.4 N.D. 7.5 Naphthalene mg L-1 0.1 N.D. 0.2 Other aromatics mg L-1 1.2 N.D. 2.3 n-Hexane mg L-1 0.3 N.D. 0.6 Cyclohexane mg L-1 0.4 N.D. 1.1 n-Octane mg L-1 0.7 0.1 2.3 pH -- 8.8 8.6 8.4 Conductivity mS cm-1 20 174 29 Salinity % NaCl sat. 34 203 56

Emissions from Different Pond Surface Types

Emissions from the Teflon surface were low, showing that the flux chamber system had consistent, low blanks (Figure 4). Emissions from hypersaline surfaces, frozen surfaces, and snow-covered ground were higher and more variable than emissions from the Teflon surface, but were not significantly different from one another (p = 0.55 to 0.90). Emissions from the recently offloaded surface type were more than two orders of magnitude higher than emissions from the other surfaces.

The recently offloaded surface type had higher emissions of aromatics and lower emissions of methanol as a percentage of total emissions than other surface types (Figure 5). Even though the recently offloaded surface type made up only 10% of the pond surface area at sampled produced water facilities, it accounted for 99.2% of all emissions from the facilities studied.

Figure 4. Emissions from different surface types at produced water facilities. The top of the colored column is the average total emissions from four types of compounds, and the contribution of each type is delineated with a different color. The black line on top of the colored column is the 90% confidence interval for the average total emissions.

Figure 5. Emissions of organic compounds from different surface types as a percentage of total emissions.

53 Factors that Influenced Emission Rates

In many cases, organic compound concentrations in water were strongly correlated with emission rates of that compound. For example, the R2 value for the correlation between toluene concentration in water and toluene emission flux was 0.91 (Figure 6). No significant correlation was observable for some other compounds, however. The average R2 value for 14 compounds for which water concentrations and emission rates are both available was 0.48.

Figure 6. Toluene concentration in unfrozen produced water versus toluene emission rate.

Meteorology (most particularly, temperature) also influenced emission rates. Figure 7 shows a dramatic increase in emissions of organic compounds from a produced water pond as the layer of ice that had formed overnight melted. Emissions from ponds with ice cover were much lower than from open ponds with similar organic compound concentrations in water. Thus, the contribution of produced water pond emissions to ozone production in the Uintah Basin is likely to depend strongly on ice cover (which, in turn, depends on temperature).

Figure 7. Change in emission rate of alkanes, aromatics, methanol, and methane throughout the day as temperature warms and ice covering a produced water pond melts.

54 Comparison to Basin-wide Emissions

To apply emissions data from this study to all produced water facilities in the Uintah Basin during the winter months, the following assumptions were made: (1) pond surface types follow a similar distribution, (2) produced water has a similar composition, and (3) observed emission rates are seasonally representative.

Table 2 shows estimated emissions from different surface types for the entire Uintah Basin, using the assumptions above. In Table 3, the emission rates from Table 2 are compared with total NMHC emissions from the 2012 UBOS emissions inventory (Mansfield et al., 2013). Table 3 shows that emissions from produced water ponds are a small portion of total inventoried wintertime NMHC emissions, but since more than 50% of emissions from produced water are aromatic compounds (many highly reactive in terms of ozone production), this source may still contribute significantly to wintertime ozone.

Table 3 provides an estimate of wintertime methanol emissions from all produced water facilities in the Uintah Basin, but no estimate exists of methanol emissions for all sources in the Basin, so the significance of this source is unknown. A comprehensive emissions inventory for wintertime emissions of methanol and formaldehyde is needed (emissions of formaldehyde appear to be associated with methanol emissions in many cases; Roberts et al., 2013).

Table 2. Estimated emission rate of organic compounds from all produced water facilities in the Uintah Basin, categorized by pond surface type. TNMHC is total nonmethane hydrocarbons and is the sum of alkanes, alkenes, and aromatics.

kg/h Frozen Hypersaline Recently Offloaded Alkanes 0.61 0.21 98.7 Alkenes 0.002 0.001 0.000 Aromatics 0.17 0.11 155.3 TNMHC 0.78 0.32 254.0 Methanol 0.77 1.10 117.0

Table 3. Wintertime produced water emissions in the Uintah Basin compared to total emissions from all sources (Mansfield et al., 2013). TNMHC is total nonmethane hydrocarbons and is the sum of alkanes, alkenes, and aromatics.

TNMHC Methanol kg/h from produced water 255 119 tons/month from produced water 196 91 Total tons/month for entire Basin 9200 -- % of total emissions from ponds 2.1% --

Uncertainties and Next Steps

During this single-season study, only a subset of total produced water facilities in the Uintah Basin was sampled. More study clearly is needed to better quantify emission rates and relationships between emissions, meteorology, and water chemistry. This study did not seek to quantify emissions from produced water storage tanks at well sites, from produced water loading and offloading areas, or from covered and uncovered skim tanks and oil storage tanks at produced water facilities. These additional produced water-related emission sources could substantially increase estimates of wintertime hydrocarbon emissions from produced water.

Summary

Produced water NMHC emissions are a small but meaningful percentage of total wintertime emissions in the Uintah Basin. Emissions from produced water are dependent on meteorology and water composition. More work is needed to refine current estimates of produced water emissions.

Acknowledgements

We gratefully acknowledge support from the Utah State and Institutional Trust Lands Administration (SITLA), the Uintah Impact Mitigation Special Service District (UIMSSD), the Utah Science, Technology, and Research Initiative (USTAR), and the United States Department of Energy (DOE).

56 References

Clark, C.E., and J.A. Veil (2009), Produced Water Volumes and Management Practices in the United States, ANL/EVS/R-09/1, Argonne National Laboratory, Argonne, Illinois. Eklund, B. (1992), Practical guidance for flux chamber measurements of fugitive volatile organic emission rates, J. Air Waste Mgmt. Assoc., 42, 1583-1591. EPA (1996b), Method 8015B: Nonhalogenated Organics Using GC/FID, United States Environmental Protection Agency. EPA (1996b), Method 8260B: Volatile Organic Compounds by Gas Chromatography/Mass Spectrometry (GC/MS), United States Environmental Protection Agency. EPA (1998), Technical Assistance Document for Sampling and Analysis of Ozone Precursors, EPA/600-R- 98/161, United States Environmental Protection Agency, Research Triangle Park, North Carolina. EPA (1999), Compendium of Methods for the Determination of Toxic Organic Compounds in Ambient Air, EPA/625/R-96/010b, United States Environmental Protection Agency, Cincinnati, Ohio. Kampbell, D.H., and S.A. Vandegrift (1998), Analysis of dissolved methane, ethane, and ethylene in ground water by a standard gas chromatographic technique, J. Chromatographic Sci, 36, 253- 256. Lyman, S., Mansfield, M., Shorthill, H., Anderson, R., Mangum, C., Evans, J., Shorthill, T., Horel, J., Crosman, E., Neeman, E., Christiansen, M., and O’Neil, T (2013), Distributed measurements of air quality and meteorology, in Final Report: 2013 Uintah Basin Winter Ozone and Air Quality Study, Environ, Novato, California. Lyman, S., and H. Shorthill (Eds.) (2013), Final Report: 2012 Uintah Basin Winter Ozone and Air Quality Study, CRD13-320.32, Office of Commercialization and Regional Development, Utah State University, Logan, Utah. Mansfield, M., D. Moss, C. Hall, E. Smith, and H. Shorthill (2013), Emissions inventory report, in Final Report: 2012 Uintah Basin Winter Ozone and Air Quality Study, CRD13-320.32, edited by S. Lyman and H. Shorthill, Office of Commercialization and Regional Development, Utah State University, Logan, Utah. Roberts, J., E.J. Williams, S.B. Brown, L. Lee, R. Cohen, S. Murphy, J. Gilman, J. de Gouw, C. Warneke, C. Young, P. Edwards, R. McLaren, J. Kercher, J. Thornton, C. Tsai, J. Stutz, R. Zamora (2013), Intensive measurements at the Horsepool site, in Final Report: 2012 Uintah Basin Winter Ozone and Air Quality Study, CRD13-320.32, edited by S. Lyman and H. Shorthill, Office of Commercialization and Regional Development, Utah State University, Logan, Utah. Thoma, E. (2009), Measurement of Emissions from Produced Water Ponds: Upstream Oil and Gas Study #1: Final Report, EPA/600/R-09/132, Office of Research and Development, U.S. Environmental Protection Agency, Cincinnati, Ohio

57 SURVEY OF BASINS AND VALLEYS IN THE WESTERN USA FOR THE CAPACITY TO PRODUCE WINTER OZONE

Courtney Hall, Randy Anderson, Marc Mansfield, Seth Lyman, and Howard Shorthill

Executive Summary

Extensive winter air quality monitoring in the Upper Green River Basin (UGRB) of Wyoming and the Uintah Basin (UB) of Utah indicates that simultaneous thermal inversions and snow cover are necessary for the formation of high concentrations of winter ozone. In this report, we will utilize the phrase “ozone-favorable meteorology,” to indicate this condition. We have developed a technique, described below, to estimate the number of days per year that any given basin or valley has ozone-favorable meteorology.

Ozone formation also requires the presence of ozone precursors, namely volatile organic compounds (VOC) and the oxides of nitrogen, NO and NO2, referred to collectively as NOx. There are three source categories of ozone precursors that concern us: the oil and gas extraction industries, internal combustion engines, and agricultural emissions. Different source categories produce different amounts and different mixes of ozone precursors, with different ozone reactivities, so the source category found in one basin may be more ozone-productive than another. Although NOx is required for ozone formation, it also acts to “titrate out,” or scavenge ozone. Under certain conditions, especially at higher NOx concentrations, it plays a dual role, acting not only to generate, but also to destroy ozone.

Prior to the accidental discovery of high winter ozone concentrations in the UGRB, wintertime monitoring of ozone was rare, and still is so today. Therefore, the wintertime ozone reactivities of many precursor source categories have not been confirmed by direct measurement. One of our goals has been to examine available meteorological data to identify basins and valleys with meteorology favorable to the formation of winter ozone, but with varying precursor amounts and source categories, in order to suggest basins that might be good candidates for ozone monitoring.

This led us to sample ozone in Idaho Falls, Idaho, in the USRP, during winter 2013. It presented itself as a basin with inversions and snow cover, but without any oil or gas production. Over a seven-week period from mid-January to mid-March, there was a continuous snow pack, and there were four multi- day inversion events. The 8-hour average ozone, however, never exceeded about 51 ppb.

We have surveyed 13 basins and valleys in the western USA for the occurrence of ozone-favorable meteorology. The 13 basins are listed in Table 1. For locations of each basin, see Figure 1. Each of the 13 will be represented in this report with an acronym, which is also listed.

Table 1. The 13 basins and valleys included in this study. Asterisks indicate the basins for which winter ozone measurements exist. Basin Acronym Major Population Centers Big Horn Basin BHB Cody, Greybull, & Worland, WY Cache Valley* CV Logan, UT Flathead Valley FV Kalispell, MT Lower Snake River Plain LSRP Burley & Twin Falls, ID Price River Valley* PRV Price, UT Salt Lake Valley* SaltLV Salt Lake City, UT San Luis Valley SanLV Alamosa, CO Uintah Basin* UB Vernal & Roosevelt, UT Upper Green River Basin* UGRB Big Piney & Pinedale, WY Upper Snake River Plain* USRP Idaho Fall & Rexburg, ID Utah Valley UV Provo, UT Wind River Basin* WRB Lander & Riverton, WY Yellowstone Caldera YC Lake Yellowstone, WY

Figure 1. Locations of the 13 basins and valleys considered in this study.

Two years ago, not too long before we initiated this study, winter ozone measurements had been taken, as far as we are aware, in only three basins or valleys: the SaltLV, the UGRB, and the UB. Since then, monitoring has also occurred in CV, the PRV, the USRP, and the WRB. In other words, we are aware of only three basins with four or more years of winter ozone monitoring, and of four other basins that have seen one or two seasons of monitoring. We have assigned each of these seven basins to one of four ranks. The ranking system groups basins together that have similar average ozone concentrations throughout the winter. Assigning two basins to the same rank might suggest that they have similar ozone systems, but too little is known at present to be sure of this. The four ranks are as follows. (February ozone averages are being used only as convenient markers for the ranks; as we see below, basins in the same rank actually track each other throughout all the winter months.)

• Rank 1: Uintah Basin o Mean February ozone of about 75 ppb o Many exceedences of the National Ambient Air Quality Standard (NAAQS).

• Rank 2: Upper Green River Basin o Mean February ozone of about 50 ppb o Exceedences of the NAAQS are less frequent but not uncommon in some winters. These exceedences have occurred with enough frequency that Sublette County is now formally in non-attainment for ozone.

• Rank 3: Wind River Basin, Price River Valley, and Upper Snake River Plain o Mean February ozone around 40 ppb. o No wintertime exceedences of the NAAQS have been recorded for these basins, though each has only been monitored for one or two winters.

• Rank 4: Cache Valley and Salt Lake Valley o Mean February ozone around 30 ppb. o Though these valleys have serious air quality problems, winter ozone is not one of them. No wintertime NAAQS ozone exceedences have been recorded.

Ozone averages in the UB during the winter of 2012, when there was no snow cover, are much lower than average. Based on this one season, we can state that the UB drops to Rank 3 when snow is not present.

We have also collected data on the production of oil, natural gas, and coal bed methane in all 13 basins, and will use these data as proxy indicators of the emission of ozone precursors by the oil and gas industries. Of the seven with existing winter ozone measurements, three (CV, SaltLV, USRP) have no oil or gas extraction, two (PR and WRB) have low to moderate extraction rates of less than 4 megatonnes/year (Mt/y), while the UB and the UGRB are high, with about 10 and 21 Mt/y, respectively.

Population statistics (2010 census) have also been assembled for each basin. We use the population of a basin or valley as a proxy indicator of the emission of ozone precursors by vehicular traffic. The most urbanized basins are SaltLV and UV, with populations of about one million and one-half million, respectively. Basins in the 100 to 250 thousand range are the USRP and LSRP, FV, and CV. All remaining basins (UB, SanLV, BHB, WRB, PRV, YC, and UGRB) have populations between about 10 and 50 thousand.

We have developed a computational technique, BASLARC (BASin Lapse Rate Calculator), which uses surface air temperature readings from throughout a basin to estimate the daily lapse rate of the basin, a quantity that indicates the presence and strength of thermal inversions. We use these calculations to estimate the average number of days per year that a basin has ozone-favorable meteorology. This average gives a meteorology score for each of the basins, and will be referred to as the BASLARC score for the basin. All seven of the basins that have been monitored for winter ozone have BASLARC scores of between 20 and 52 days per year, i.e., we estimate that they have between 20 and 52 days per year with simultaneous inversions and snow cover. Of the other six basins, BHB, SanLV, UV, and YC also have BASLARC scores of 20 or more, while FV and LSRP each have 11 or fewer.

It follows that meteorology alone is not a good predictor of the average winter ozone concentration. For example, of the seven basins with winter ozone measurements, two, the UB and UGRB, have high winter ozone (Rank 1 or 2) three have winter ozone at Rank 3 (PRV, USRP, and WRB), and two are low at Rank 4 (SaltLV and CV). This is in spite of the fact that all seven have “ozone-favorable” meteorology.

The magnitude of oil and gas production in a basin is a better predictor of winter ozone than meteorology. Only the two basins with production rates above 10 Mt/y, UB and UGRB, also have winter ozone in Ranks 1 or 2. The remaining five basins, with production rates below 4 Mt/y, all have winter ozone in Ranks 3 or 4.

Two valleys, Salt Lake and Cache, have other winter pollution problems associated with inversions; indeed, they are formally in non-attainment for aerosols in winter. Furthermore, they are known to have high ozone precursor concentrations during inversions. Nevertheless, as already mentioned, their winter ozone concentrations are well below background. We hope to develop a better understanding of the reasons for this through modeling. Our present hypothesis is based on the difference in their precursor emission sources: UB is high for oil/gas emissions, SaltLV for vehicular traffic, while CV is moderately high for vehicles but also has significant agricultural emissions. Since they have different source categories, we can surmise that the ozone reactivities of their precursor mixes are different. We hypothesize that the mix typical of the UB and UGRB is conducive to ozone formation in winter, while the mix typical of the SaltLV or CV is not. In fact, the ozone levels in SaltLV and CV are probably below background because of components that are scavenging ozone.

The data now indicate that: (1) Only basins with more than 10 Mt/y of oil and gas production and BASLARC scores of 30 or more days/y generate enough ozone to trigger exceedences of the NAAQS ozone standard. (2) Basins with 4 Mt/y or less of oil and gas production do not trigger NAAQS exceedences, in spite of high BASLARC scores. (3) Basins with other pollution problems but without oil and gas production have low winter ozone, presumably because ozone is being scavenged. These three rules are based on only a handful of basins, many of which have had only one or two seasons of ozone sampling. Therefore, more extensive data may one day force revisions of these three rules.

One result remains unexplained. The UGRB has higher petroleum extraction than the UB, with 21 and 10 Mt/y, respectively. It also has a higher BASLARC score, 52 vs. 33 days/year. Based on these two criteria, we might expect the UGRB to have higher ozone than the UB, but it does not, with the UB and the UGRB being assigned to ranks 1 and 2, respectively. In the report, we tender three hypotheses for this juxtaposition, and point out that modeling can probably help us select between them.

In one manner of speaking, winter ozone measurements also exist for the YC. The National Park Service conducts continuous ozone monitoring between 1 April and 31 October, and spring comes late to

Yellowstone, with the snowpack usually lasting until May. Furthermore, the average ozone concentration in April and May is rather high. The average for April is about 55 ppb, quite similar to the UGRB in February, although NAAQS exceedences are very rare. However, ozone concentrations in May do not appear to be correlated with the presence of snow cover. The ozone drops to lower values in the summer. Its ozone system seems to be fundamentally different from that of the UB. Additional speculative discussion about Yellowstone is given in the body of the report.

To the best of our knowledge, winter ozone has not been measured in the other five basins that we surveyed. Very tentatively, based on similarities between them and the seven basins with ozone measurements, we predict that UV will be rank 4, BHB and SanLV will be rank 3, and that FV and the LSRP will be either rank 3 or 4. See below for details.

One school of thought maintains that blow-over from urban regions, particularly the SaltLV and UV, contributes significantly to winter ozone in the UB. As we point out in this report, this hypothesis cannot at present be entirely ruled out, but we do give two reasons why we believe blow-over does not contribute significantly to the winter ozone problem in the UB. A better understanding of airflow between the two regions will be provided by the models that we are currently building.

In summary, according to the evidence available now, the presence of inversions with snow cover does not by itself produce high winter ozone. SaltLV, CV, PRV, WRB, and the USRP all provide evidence to the contrary. The presence of ozone precursors, without specifying the precursor source category, also does not guarantee high winter ozone. Furthermore, low to moderate oil or gas production, as is typical of the WRB and the PRV, has not led to high winter ozone. Of the seven basins and valleys for which winter ozone measurements have occurred, only two have produced enough ozone to threaten or to cause ozone non-attainment, namely UB and UGRB. These two basins also have the highest oil and gas production of all the basins surveyed. The hypothesis that is most consistent with all these facts is that ozone precursors emitted by an intensive petroleum extraction industry have the correct mix so that high winter ozone will occur when snow cover and thermal inversions are also occurring. If petroleum extraction does not occur or occurs at lower levels, then basins, including highly urbanized ones, do not appear capable of producing the correct precursor mix to generate high winter ozone.

Introduction

The Uintah Basin (UB) of eastern Utah and the Upper Green River Basin (UGRB) of western Wyoming are known to produce winter ozone at concentrations high enough to exceed United States Environmental Protection Agency (EPA) standards. Winter ozone is of concern because of possible adverse health effects and the economic burden associated with its control and abatement. In this report, we compare these two basins with 11 others throughout the western USA, focusing on the factors in each basin that are known or suspected to produce winter ozone. All 13 basins are listed in Table 1, and mapped in Figure 1. Each basin has been assigned an acronym, given in Table 1, by which it will be referred in this report. Our goal has been to find basins that have ozone-favorable meteorology, but that have varying amounts and blends of ozone precursors. Tables 2 and 3 give further data on each basin.

Experience indicates that thermal inversions combined with snow cover are necessary for the formation of winter ozone in both the UB and UGRB. In this report, we use the phrase “ozone-favorable meteorology” to indicate the condition of simultaneous thermal inversions and snow cover. We have developed an approach, the “BASin LApse Rate Calculator,” or BASLARC, based on the distribution of surface temperatures throughout each basin, to determine the intensity of thermal inversions on a daily basis. We have also collected data on snow cover in each of the basins.

Ozone precursor compounds, in the form of volatile organic compounds (VOC) and the oxides of nitrogen, NO and NO2, collectively known as NOx, must also be present in the air to produce ozone. The oil and gas extraction industries produce both VOC and NOx, as do internal combustion engines. To compare the magnitude of oil and gas production among different basins, we report here the annual (2012) production figures for oil, natural gas, and coal bed methane in each basin, and we employ 2010 population numbers to indirectly gauge the importance of emissions by vehicles.

We made an effort to include both rural and urban basins in this survey, and basins both with and without oil and gas extraction. Although we made a thorough attempt to identify a number of different candidate basins, we make no claim to have found or surveyed all basins with inversion/snow-cover meteorology or with oil/gas production. For example, we have found mention in the literature of winter inversions in the Yampa Valley (Steamboat Springs, CO) [Billings, et al. 2006], in Bear Lake Basin (Randolph, UT) [Palacios, et al. 2007] and in different regions of the Colorado Plateau [Whiteman, et al 1999]. A more thorough survey may be undertaken in the future.

Prior to the discovery of winter ozone in the UGRB, it was not common to monitor ozone during the winter. Therefore, winter ozone data are available for only about half of the basins, and for many of these, winter ozone sampling exists for only one or two seasons, as recorded in Table 3. The USRP has meteorological conditions similar to both the UB and the UGRB, namely an estimate of 33 days per year with simultaneous inversions and snow. When the preliminary results of our snow-inversion survey for the USRP became available, we recognized an opportunity to measure ozone in a basin with ozone- favorable meteorology but without petroleum production. We therefore obtained permission to monitor ozone at the Idaho National Laboratory (INL) in Idaho Falls, Idaho for a seven-week period from January to March 2013. The BASLARC technique, when applied to the USRP, indicates that there were 37 days, out of 54 days total during the measurement period, with simultaneous snow cover and thermal inversions. However, the highest daily ozone concentration (8-hour running mean) was 51 ppb, well below the National Ambient Air Quality Standard (NAAQS) of 75 ppb, and small in comparison to winter ozone measurements typical of the UB and UGRB. A complete report of the Idaho Falls ozone data and measurements is given below.

Of course, both fossil fuel production and population are reported on a county-by-county basis. For the purposes of estimating basin-wide contributions, we have accumulated contributions from individual counties as outlined in Appendix A. Because county boundaries are not always aligned with the edges of basins or valleys, this approach is sometimes approximate and arbitrary.

Table 2. The 13 basins and valleys included in this study. See map (Figure 1) for the location of each. Asterisks indicate the basins for which winter ozone data exist Population ~ altitude‡ of Basin 2010 census basin floor (m) Latitude BHB 48,000 1100 44.3 CV* 125,000 1300 41.8 FV 120,000 800 47.8 LSRP 163,000 1100 42.6 PRV* 32,000 1600 39.5 SaltLV* 1,064,000 1200 40.7 SanLV 45,000 2200 37.6 UB* 51,000 1400 40.4 UGRB* 10,000 1900 42.8 USRP* 216,000 1300 43.7 UV 517,000 1300 40.3 WRB* 40,000 1400 43.0 YC 21,000 2000 44.6 ‡ = Above sea level.

Table 3. Some characteristics of the basins surveyed. Annual Production in Megatonnes

BASLARC # Winter Mean [O3], ppb, Natural Coal Bed Basin Score‡ Seasons§ Feb. 1 – Feb. 29 Oil Gas Methane TOTAL BHB 20 1.28 0.27 1.55 CV 33 1 34.5 0 FV 11 0 LSRP 6 0 PRV 20 1 46.2 0.01 1.95 1.96 SaltLV 28 8 31.1 0 SanLV 27 0 UB 33 4 74.9 3.08 7.09 10.17 UGRB 52 9 53.0 0.92 20.55 21.47 USRP 33 1 41.8 0 UV 25 0 WRB 25 2 40.2 0.61 3.08 0.06 3.75 YC 38 0 ‡ = Average days/year with simultaneous snow cover and thermal inversions. § = Number of winter seasons that ozone has been monitored in corresponding basin.

Survey of Thermal Inversions

We use the so-called lapse rate, Γ, to quantify the intensity of an inversion. It is determined by the initial slope of the temperature-altitude profile, as shown in Figure 2. By definition, the lapse rate is negative when the atmosphere is inverted, and positive when it is not. For example, a lapse rate of +10 K/km means that the temperature decreases by 10 K (or equivalently by 10°C) as we rise through one kilometer, while a lapse rate of –10 K/km indicates a 10 K increase for each kilometer of altitude. The proper method of determining the temperature-altitude profile is through airborne measurements collected with moored balloons, free-floating balloons, or aircraft. Such measurements, however, are not generally available on a daily basis in most of the basins considered here. Therefore, to estimate the lapse rate on any given day, we have developed a technique we call BASLARC (BASin LApse Rate Calculator). It is based on the distribution of temperatures at various surface sites throughout the basin.

Typical Winter Summer Inversion A Altitude Inversion Layer

Temperature

Figure 2. Schematic diagrams of the altitude-temperature profiles in the atmosphere. In the typical summertime pattern, the air gets colder as we rise through the atmosphere. During a winter inversion, the air grows warmer as we climb, until the top of the inversion layer, after which it cools. Point A represents a point above the inversion layer. The lapse rate is determined by the slope of the profile near the surface, i.e., the slope of the dashed line.

Altitude-Temperature profiles, Wind River Basin

2012/07/05 [16.1]

1990/12/23 [-43.6] Altitude, m 1991/03/30 [-28.8]

1990/07/09 [23.7]

1400-40 1500-20 1600 17000 1800 1900 20 40 Temperature, °C

Figure 3. The BASLARC construction for four different days in the WRB. The lapse rate in K/km is indicated in brackets. (In the interest of full disclosure: These four days were “cherry-picked” to show large-magnitude lapse rates and strong altitude-temperature correlations. The daily traces often show more scatter.)

The BASLARC analysis begins with the daily maximum temperature measured at a number of different stations at different altitudes throughout the basin. These temperature data were downloaded from Utah State University’s Utah Climate Center (UCC), its URL is given in the References section. Appendix B gives lists of the stations employed for each basin. The temperature-altitude data can be displayed on a scatter plot, as in Figure 3, which shows the BASLARC construction for four days in the WRB. The daily lapse rate is taken as the negative of the slope of the best-fit line, using altitude and temperature, respectively, as independent and dependent variables.

We chose to base BASLARC on the daily maximum temperature at each site because it is the most common daily temperature datum found in the historical record. Another option would be to employ the average daily temperature, but it is not recorded as frequently. Since the daily maximum temperature and the daily maximum ozone concentration both usually occur in the afternoon hours, we also believed that using the daily maximum temperature would make BASLARC more sensitive to persistent inversion events, i.e., events for which the thermal inversion persists throughout the day, and that it would be a better indicator of the meteorological state of the atmosphere at the time of day that ozone formation is most important.

On the other hand, using the daily maximum temperature probably has its disadvantages. For example, if the passage of a warm or cold front produces a generalized heating or cooling trend, the daily maximum temperature might occur just before or just after midnight, or at some other time of day. The resulting BASLARC lapse rate might be misleading, assigning an inversion to a day in which none actually occurs. Furthermore, there is no guarantee that daily maxima at different sites throughout the basin occur at the same time of day. Finally, assigning a single lapse rate value to an entire day is not a good representation of the meteorological state of the basin on days that the weather is changing rapidly.

In order to have adequate sampling throughout a basin, we have programmed BASLARC to skip over any day for which there are less than five stations reporting a temperature. Because of data gaps, this means that there may be days for which BASLARC does not determine a lapse rate. Furthermore, with this constraint, BASLARC is inoperable in basins so small that five different stations cannot be found. In this report, we will refer to this as the “small-basin problem.” As we see below, the YC is borderline for the small-basin problem, with many days for which the lapse rate goes undetermined. This probably means that its BASLARC score of 38 is subject to error. Furthermore, the small-basin problem has made it impossible to study several interesting valleys, including Bear Lake Basin (Randolph, Rich County, Utah), Yampa Valley (Steamboat Springs, Routt County, Colorado), and the king of cold pools, Peter Sinks (Cache County, Utah). All are reported to have intense wintertime inversions [Palacios, et al 2007; Billings, et al 2006; Clements, et al 2003], but we were unable to find enough temperature stations to compute their BASLARC scores.

Nevertheless, by using as many stations as possible, we are usually able to assemble long, uninterrupted time series of the daily lapse rate in the basins considered in this study. Table 4 shows the approximate total length of the time series assembled for each of the basins, not counting small-basin gaps in the time series. Most of these time series terminate in 2012 or 2013.

The North American Mesoscale model (NAM) is a weather forecasting system provided by NOAA’s National Operational Model Archive & Distribution System. For our purposes, we use the “analysis- only” archived runs with the acronym NAM-ANL, its URL is given below in the References section. It is possible to extract from NAM-ANL, an independent determination of the occurrence of cold pooling in any particular region, at least at low (12-km) resolution. This determination is based on a quantity known as the potential temperature. We performed spot checks of the potential temperature in four of the basins (UB, USRP, PRV, and WRB) on four separate days during the winter of 2013. At least on these four days and in these four basins, BASLARC and NAM-ANL are in complete agreement in terms of predicting the presence or absence of thermal inversions.

Table 4. Lengths of both the lapse rate and the snow depth time series assembled for each basin. lapse rate snow depth lapse rate snow depth Basin (years) (years) Basin (years) (years) BHB 22 23 LSRP 22 22 CV 38 109 UGRB 21 22 FV 8 23 UB 63 64 USRP 23 24 UV 40 113 PRV 12 97 WRB 23 65 SanLV 22 22 YC 10 22 SaltLV 59 100

An explanation for the broad disparity in the lengths of the time seriesMean listed in Table 4 Biweekly Lapse is in order. Rate At the beginning of this study, we had arbitrarily decided to start each time series at about 1990, assuming that 20 or more years would be adequate to sample the climate trends of a basin. This explains why series of around 22 years commonly appear in Table 4. Eventually we learned that it was no more difficult to download all available years from the UCC, and so over time our procedure evolved to the point that we would download all available data. In such cases, time series may go back for a century or more. On the other hand, downloads are labor intensive – one file must be downloaded for 10 15 each line in each table of Appendix B – so we usually felt no compulsion to go back a second time to get older data from stations for which we already had 20 years of data. Time series shorter than 20 years also appear in Table 4. These are usually the result of long gaps caused by the small-basin problem when stations go off line. The snow series are usually longer than the lapse rate series for any given basin because 5 we do not require five stations for calculating snow depth averages.

Figure 4 shows the average lapse rate in each basin, displaying biweekly averages. For almost all the basins, the curves show definite wintertime minima, indicating that inversions are more 0 common then. Curiously, SanLV shows two minima, meaning that BASLARC is recognizing many summer inversions there as well. The depth of each wintertime minimum is an indicator of the severity of inversions Mean lapse rate, K/km in each basin. These depths range from the WRB at about –9 K/km to CV at about +1 K/km. FV is atypical in that well-defined winter minima are not seen.

Mean Biweekly Lapse Rate JASONDJFMAMJ -10 -5 CV YC FV SanLV USRP

10 15 SaltLV PRV UB LSRP UV UGRB BHB 5 WRB 0 Mean lapse rate, K/km

JASONDJFMAMJ

-10 -5 CV Figure 4. Mean biweekly lapse rate for each of the YC 13 basins. FV SanLV USRP SaltLV PRV 69 UB LSRP UV UGRB BHB WRB Mean Biweekly Lapse Rate 10 15 5 0 Mean lapse rate, K/km

Daily inversion probability JASONDJFMAMJ -10 -5 JASONDJFMAMJ CV YC FV SanLV USRP SaltLV PRV UB LSRP UV UGRB BHB WRB Probability

0.0 0.2 0.4 0.6 0.8 1.0 CV Figure 5. Probability, based on the BASLARC calculation, that a given day of YC the year has a temperature inversion. FV SanLV USRP Figure 5 shows the probability that any given calendaSaltLV r day will have an inversion. These are best-fit bell- shaped curves, which are almost always a good fit to thePRV data, especially with those basins for which the UB time series listed in Table 4 is long. In most cases, the inversion probability peaks in January, and is LSRP highest for the BHB. BASLARC estimates that about 85% of the daysUV there in early to mid-January have inversions. On the other extreme, BASLARC estimUGRB ates that the YC has about 25% odds of an inversion BHB during December. The summertime inversions in WRB SanLV are also obvious in Figure 5.

To avoid the small-basin problem, and to minimize the effect of temperature fluctuations or errors at any one site, we normally try to include as many temperature stations as possible within each basin. However, some care must be taken in the selection of the set of stations that represents the basin, the general rule being that the stations used must not extend, either horizontally or vertically, outside the typical cold pool that forms during an inversion. We have found, for example, that we should not use stations too far up in the mountains above the basin. (In effect, we want to stay in the linear part of the blue curve in Figure 2. If we go too far up the mountainside, we find ourselves in the vicinity of point A on that curve, and we misrepresent the lapse rate.) We often found it necessary to select an optimal set of stations by trial and error, inserting and removing stations at the outer boundaries of the basin until we found a set that maximized the summer-winter difference in the curves shown in Figure 4.

An interesting case in point is provided by the two basins that we call the Lower and the Upper Snake River Plains. We began our study of southern Idaho by taking a large set of stations from throughout

70 the Snake River Plain, from Saint Anthony on the east to Twin Falls and Hollister on the west, and treating them as representative of a single basin. This set of stations gave no strong winter inversions. This was a perplexing finding, since the region has a reputation for inversions. We eventually discovered that by taking two separate groupings of stations to form what we refer to here as the USRP and LSRP, then each grouping taken individually had wintertime inversions. We interpret this to mean that in winter, the two regions tend to form separate cold pools, and that one region can be inverted when the other is not.

Survey of Snow Cover

Most of the stations providing temperature data also record a daily snow depth. On any given day, we assemble all of the snow depths reported from throughout a basin and use these values to estimate the basin-wide snow depth for the day, interpreting the entry “trace” as 10 mm (≈ 0.4 in) . There are occasional obvious errors in the snow record, for example, one might find a station reporting a zero depth when all other stations in the basin report ample snow cover, or non-zero entries extending throughout an entire summer. Therefore, after assembling all the daily snow depths from throughout any one basin, we discard all those that are more than two standard deviations away from the mean, and then recalculate the mean and report it as the basin-wide average snow depth for the day. Figure 6 shows the snow depth through the course of the year, averaged biweekly. Average snow depths vary from a maximum of about 1100 mm (≈ 40 in) for the YC, to about 20 mm (≈ 0.8 in) for the LSRP. Average snow depths typically peak in January. Figure 7 shows the probability that any given calendar day will have snow cover (defined operationally as a day with average snow depth > 50 mm or about 2 in). Once again these probabilities are displayed with a best-fit bell-shaped curve. They extend from a maximum of around 100% for the YC, the USRP, and the UGRB, to a minimum of about 15% for the LSRP. These daily probabilities typically peak in January.

Survey of Ozone-Favorable Meteorology

By combining the time series data on inversions and snow cover, we are able to determine whether any given day has ozone-favorable meteorology. The bell-shaped curves in Figure 8 represent the probability that any given calendar day simultaneously has an inversion (BASLARC lapse rate < 0) and snow cover (mean snow depth > 50 mm). The total area under any of one of the curves is called the BASLARC score of the basin, and it represents the average number of days per year that the basin has ozone-favorable meteorology. These numbers are reported in Table 3, and as a bar chart in Figure 9. The five basins with meteorological conditions most favorable to winter ozone are the UGRB, with about 52 days per year, YC with 38, and the UB, CV, and USRP, each with about 33 days per year. A number of other basins, the SaltLV, the SanLV, the WRB, and UV, follow close behind, with around 25 to 30 days per year each, on average. BHB and PRV are each at about 20 days per year, while FV and the LSRP each display 11 or less. FV and LSRP score low primarily because they are at low altitude and therefore have low probabilities for snow cover.

Biweekly Mean Snow Depth

JASONDJFMAMJ Mean snow depth, mm

The YC provides an interesting case. BASLARC predicts that it has a relatively low frequency of inversions, topping out at about 25% in December (Figure 5). However, it has a very long snow season (Figure 6b), so its final BASLARC score is relatively large.

Biweekly Mean Snow Depth 0 50 100 150 200 250 300 350

CV JASONDJFMAMJ FV SanLV USRP SaltLV PRV UB LSRP UV UGRB BHB WRB Mean snow depth, mm

0 50 100 150 200 250 300 350 Figure 6a. Mean snow depth throughout the year is shown for all basins except the Yellowstone CV Caldera. Yellowstone gets its own panel because its curve is at least 3x higher than any of the others.FV SanLV USRP SaltLV PRV UB LSRP UV UGRB BHB WRB

Biweekly Mean Snow Depth, Yellowstone Caldera Mean snow depth, mm

JASONDJFMAMJ 0 100 200 300 400 500 600 700 800 900 1000 1100

Figure 6b. Mean snow depth in the Yellowstone Caldera.

Mean Biweekly Lapse Rate 10 15 5 0 Mean lapse rate, K/km

Daily snow cover probability JASONDJFMAMJ -10 -5 JASONDJFMAMJ CV YC FV SanLV USRP SaltLV PRV UB LSRP UV UGRB BHB WRB Probability

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 CV Figure 7. Probability that any given day of the year has snow cover, defined operationally as a snow YC depth greater than 50 mm. FV SanLV USRP SaltLV PRV UB LSRP UV UGRB BHB WRB

Mean Biweekly Lapse Rate 10 15 5 0 Mean lapse rate, K/km

Daily probability of inversions with snow cover JASONDJFMAMJ -10 -5 CV JASONDJFMAMJ YC FV SanLV USRP SaltLV PRV UB LSRP UV UGRB BHB WRB Probability

0.0 0.1 0.2 0.3 0.4 0.5 0.6 CV YC Figure 8. Probability of simultaneous inversion and snow cover on any given day. Area under each FV curve is the BASLARC score, or the average number of days per year with ozoneSanLV -favorable meteorology. USRP SaltLV PRV UB LSRP UV UGRB BHB WRB

Snow/Inversion days per year (average) Average snow/inversion days LSRP BHB CV FV PRV SltLV SanLV UB UGRB USRP UV WRB YC

0 5 10 15 20 25 30 35 40 45 50 55 Figure 9. BASLARC score, or average number of days per year with both inversions and snow cover, for each basin.

Survey of Oil, Gas, and Coal Bed Methane Production

Table 3 and Figure 10 give, for each basin, the total tonnage of oil (including “condensate”), natural gas, and coal bed methane produced during 2012, in millions of metric tons, or megatonnes (Mt). Of the 13 basins, the UGRB and the UB have the most intensive oil and gas extraction, producing respectively 21 and 10 megatonnes/year (Mt/y). Production figures for the WRB, the PRV, and the BHB are lower, being respectively about 4, 2, and 2 Mt/y. No oil or gas is produced in any of the other basins. These data were assembled by consulting websites of the respective state agencies responsible for collecting production data, see below for the URL’s, and see Appendix A for the calculations. Figure 10 gives a bar chart of the oil, natural gas, and coal bed methane production of each basin in 2012.

The non-methane content of coal bed methane is lower than that of natural gas, which means that tonne for tonne, it is less ozone-reactive. Therefore, Figure 10, which puts coal bed methane and natural gas on an equal footing, may overestimate the significance of this emission source for the PRV.

Oil, Natural Gas, Coal Bed Methane Production (2012)

UGRB

UB 10 15 20 25 5 Petroleum production (megatonnes) WRB PRV BHB 0

CBM natural gas oil

Figure 10. Oil, natural gas, and coal bed methane production numbers in 2012 (megatonnes/year). None of the other basins had any production.

Survey of Population

Population numbers for each basin (2010 census), rounded to the nearest thousand, are shown in Table 2 and Figure 11. See Appendix A for additional information. Population numbers range from around 10 thousand for the UGRB to over one million for the SaltLV.

Basin Population

6 10

5 10 Population

4 10 BHB CV FV LSRP PRV SaltLV SanLV UB UGRB USRP UV WRB YC 3 10

Figure 11. Population of each basin (2010 census). Note the logarithmic scale.

Survey of Wind Speed

The Desert Research Institute’s Western Regional Climate Center (see below for its URL) tabulates monthly average wind speed at 19 airports throughout the 13 basins under study, based on hourly data from 1996 to 2006. The February averages are displayed in Table 5 and Figure 12. Whenever data are available from more than one airport in any one basin, the individual results have been averaged to give a single basin-wide average.

Table 5. Mean Wind Speed. MWS Basin Basin Airport (m/s) Average BHB Cody 3.4 2.8 Greybull 2.6 Worland 2.3 CV Logan 2.3 2.3 FV Kalispell 1.7 1.7 LSRP Burley 4.2 4.5 Twin Falls 4.9 PRV Price 2.6 2.6 SaltLV Salt Lake City 3.4 3.4 SanLV Alamosa 3.0 3.0 UB Vernal 1.7 1.7 UGRB Big Piney 2.2 2.2 Pinedale 2.2 USRP Idaho Falls 3.6 3.2 Rexburg 2.8 UV Provo 2.7 2.7 WRB Lander 2.3 2.9 Riverton 3.4 YC Lake Yellowstone 0.8 0.8

Average February wind speed Average wind speed (m/s) BHB CV FV LSRP PRV SaltLV SanLV UB UGRB USRP UV WRB YC

012345 Figure 12. Mean wind speed during February for the 13 basins.

Survey of Total Ultraviolet Radiation

The energy source for ozone production is ultraviolet radiation from the sun. Because snow cover reflects rather than absorbs ultraviolet radiation, the total radiation is much larger in the presence of snow. It is also influenced by the solar angle, being larger when the sun is higher in the sky. Therefore, it increases as the winter progresses, and as we move south. There is about a 10° difference in latitude between the southernmost (SanLV) and the northernmost (FV) basins surveyed. Calculations of total ultraviolet radiation throughout the winter season are presented for each of the 13 basins in Figure 13 and Table 6. Displayed is the midday maximum in UV radiation, assuming clear skies, measured in W/m2, including contributions from both the UV-A and UV-B bands, i.e., from all wavelengths between 290 and 400 nm. The curves in Figure 13 and quantities in Table 6 were calculated as explained in Finlayson-Pitts and Pitts [2000]. The no-snow case corresponds to their “best estimate surface albedo,” tabulated in their Table 3.7, p. 66, and the snow case, corresponding to an albedo of 80%, is based on their Table 3.11, p. 71. These tables give the actinic flux, or intensity of radiation as a function of the wavelength and of the solar zenith angle, which was integrated numerically over all wavelengths between 290 and 400 nm. Page 58 of Finlayson-Pitts and Pitts explains how to calculate the solar zenith angle from the date, the time of day, and the latitude and longitude. Note that on 28 February in the presence of snow cover, the total UV radiation in SanLV is about 31% higher than that in FV, and that the UB flux is about 6% higher than that of the UGRB.

Mean Biweekly Lapse Rate

Table 6. Midday UV radiation at each basin, 28 February, W/m2. Basin Flux with snow Flux without snow SanLV 178 72

PRV 171 70 10 15 UV 168 69 UB 168 69 SaltLV 167 69 CV 162 67 5 LSRP 159 66 UGRB 158 66 WRB 157 66 0 USRP 154 65 BHB 151 64 Mean lapse rate, K/km YC 150 63 FV 136 59

Actinic Flux -- midday maximum JASONDJFMAMJ -10 -5 CV January February March YC FV SanLV USRP SaltLV PRV UB snow LSRP UV UGRB BHB WRB

Actinic flux (W/m2) no snow 50 100 150 200 250

0 CV Figure 13. Midday maximum in ultraviolet radiation for each of the basins as a function of the time of YC FV year, in both the snow and the no-snow cases.SanLV USRP SaltLV PRV UB 81 LSRP UV UGRB BHB WRB Ozone Monitoring in the Upper Snake River Plain

When our preliminary work indicated that the USRP had ozone-favorable meteorology, we saw the opportunity to sample ozone in a region without oil or gas development. We installed a monitor on the campus of Idaho National Laboratory (INL) in Idaho Falls, Idaho, at latitude 43.52172°, longitude -112.0529°, and altitude 1445 m. The monitor operated 54 days, 15 January to 10 March. Time series of the 1-hour average ozone concentration appear in Figures 14a-g. According to the BASLARC analysis for the USRP, there was continuous snow cover for the entire period, varying from a maximum of 192 mm on 23 February to a minimum of 65 mm on 10 March. The BASLARC analysis also found 37 inversion days, or 69% of the total time period. Two intense multi-day inversions occurred, one running 17-24 January, the other 13-15 February. Other, less strong multi-day inversions ran from 31 January to 4 February, and from 1-2 March. (Section 13 below contains time series plots of the lapse rate.) Nevertheless, the largest daily 8-hour ozone maximum was 51 ppb, occurring on 28 February and 5 March. As mentioned above, NOAA’s North American Mesoscale model weather forecasting model (NAM-ANL) confirmed that inversions were occurring during these same four time periods in the USRP.

On many days, ozone concentrations follow the expected diurnal pattern. The most likely explanation for departures from the typical diurnal pattern are atmospheric disturbances such as passing storms or fronts, or perhaps ozone titration due to plumes of ozone scrubbing compounds passing over the monitoring site. On a number of occasions, the ozone concentration is seen to drop below 15 ppb. Two ozone minima, 25-26 January and 5-7 February, may be the result of atmospheric disturbances that broke up the inversions of 17-24 January and of 31 January to 4 February. On the other hand, the 14 February minimum corresponds to an inversion day (inversion confirmed by BASLARC and NAM-ANL).

The Idaho Falls ozone data (USRP) will be made available upon request to the authors.

1 hr ozone at Idaho Falls

1/15 1/16 1/17 1/18 1/19 1/20 1/21 1/22 1/23 1/24 1/25 1/26 1/27 1/28 [O3] (ppb) 0 10 20 30 40 50 60 0 48 96 144 192 240 288 336

time (hrs) Figure 14A-G. Ozone time series in the USRP, Idaho Falls, Idaho, winter 2013. The zero-hour corresponds to the midnight preceding 15 January. Each vertical line indicates midnight. Dates on subsequent panels overlap to provide context. This panel shows 15-28 January.

1 hr ozone at Idaho Falls

1/22 1/23 1/24 1/25 1/26 1/27 1/28 1/29 1/30 1/31 2/01 2/02 2/03 2/04 [O3] (ppb) 0 10 20 30 40 50 60 168 216 264 312 360 408 456 504

time (hrs) Figure 14B. 22 January through 4 February.

1 hr ozone at Idaho Falls

1/29 1/30 1/31 2/01 2/02 2/03 2/04 2/05 2/06 2/07 2/08 2/09 2/10 2/11 [O3] (ppb) 0 10 20 30 40 50 60 336 384 432 480 528 576 624 672

time (hrs) Figure 14C. 29 January through 11 February.

1 hr ozone at Idaho Falls

2/05 2/06 2/07 2/08 2/09 2/10 2/11 2/12 2/13 2/14 2/15 2/16 2/17 2/18 [O3] (ppb) 0 10 20 30 40 50 60 504 552 600 648 696 744 792 840

time (hrs) Figure 14D. 5-18 February.

1 hr ozone at Idaho Falls

2/12 2/13 2/14 2/15 2/16 2/17 2/18 2/19 2/20 2/21 2/22 2/23 2/24 2/25 [O3] (ppb) 0 10 20 30 40 50 60 672 720 768 816 864 912 960 1008

time (hrs) Figure 14E. 12-25 February.

1 hr ozone at Idaho Falls

2/19 2/20 2/21 2/22 2/23 2/24 2/25 2/26 2/27 2/28 3/01 3/02 3/03 3/04 [O3] (ppb) 0 10 20 30 40 50 60 840 888 936 984 1032 1080 1128 1176 time (hrs)

Figure 14F. 19 February through 4 March.

1 hr ozone at Idaho Falls

2/26 2/27 2/28 3/01 3/02 3/03 3/04 3/05 3/06 3/07 3/08 3/09 3/10 3/11 [O3] (ppb) 0 10 20 30 40 50 60 1008 1056 1104 1152 1200 1248 1296 1344

time (hrs) Figure 14G. 26 February through 11 March.

Average Winter Ozone Concentrations in Seven Basins

As already mentioned, winter ozone monitoring has occurred in seven of the basins. Table 7 summarizes these measurements.

Table 7. Basins with winter ozone measurements. Mean Feb. Fraction of Feb. days Basin Monitor Location Operating Agency [O3]8hr, ppb with [O3]8hr > 75 ppb UB Ouray, UT EPA 74.9 0.460 UGRB Boulder, WY EPA 53.0 0.067 PRV Price, UT UDAQ 46.2 0.000 USRP Idaho Falls, ID USU/CRD 34.5 0.000 WRB Pavillion, WY WDEQ 40.2 0.000 CV Logan, UT UDAQ 34.5 0.000 SaltLV Salt Lake City, UT UDAQ 31.1 0.000

EPA, Utah Division of Air Quality (UDAQ), and Wyoming Department of Environmental Quality (WDEQ) ozone data are available at websites listed in the References section below. The Idaho Falls ozone data (USRP) will be provided upon request.

Figure 15 shows biweekly averages of the 8-hr daily ozone maximum, taken over all available data for the months of January through March for these seven basins. We assign each basin to one of four ranks, where any one rank includes all the basins with similar traces:

UB > UGRB > (WRB ≈ PRV ≈ USRP) > (CV ≈ SaltLV) Rank 1 Rank 2 Rank 3 Rank 4

The UB, with the highest ozone concentration, is assigned to Rank 1. The UGRB is assigned to Rank 2. Since the traces for PRV, USRP, and WRB, all lie close together, we assign them all to Rank 3. The traces for the CV and SaltLV are also similar, so they are both assigned to Rank 4. The last two columns of Table 7 indicate that Rank 1 has mean February ozone concentrations of about 75 ppb, Rank 2 of about 55 ppb, Rank 3 of about 40 ppb, and Rank 4 of about 30 ppb. Rank 1 has frequent exceedences of the 75 ppb NAAQS, Rank 2 has a few, and Ranks 3 and 4 have none. Pigeonholing basins into separate ranks does not necessarily suggest that all basins in one rank have similar ozone systems, only that their average concentrations are similar.

The 2012 season saw no snow in the UB, and the mean ozone data for that basin and season appear in Figure 15 with the label “UB – no snow.” That particular trace indicates that the Rank 3 basins have ozone concentrations comparable to those of the UB without snow.

Mean ozone concentrations

RANK 1

RANK 2

RANK 3 mean [O3], ppb

RANK 4 15 25 35 45 55 65 75 0123 month

UGRB WRB USRP PRV CV SaltLV UB UB -- no snow

Figure 15. Average biweekly ozone concentrations for the indicated basins. Basins with similar traces are assigned to the same rank, under the assumption that they have similar ozone-producing systems.

Discussion of the Observed Ozone Concentrations in Seven Basins

With only seven basins reporting any measurements of winter ozone, and with many of those only reporting for one or two years, it is difficult to draw definitive conclusions, but all seven basins are consistent with the following general rules:

A. Only the basins with high petroleum extraction rates, i.e., above 10 Mt/y, and with ozone- favorable meteorology, i.e., a BASLARC score of 30 or more days/y, have high winter ozone concentrations (Rank 1 or 2). Two basins, the Uintah and the Upper Green, satisfy this rule; with 10 Mt/y and 21 Mt/y, respectively, of petroleum extraction, and with BASLARC scores of 33 and 52. The UGRB has already gone into non-attainment for ozone, and it appears likely that the UB will also.

B. Basins with low petroleum extraction rates, less than 4 Mt/y, and that are not highly urbanized, lie in Rank 3, in spite of ozone-favorable meteorology. Three basins, the PRV, USRP, and the WRB, satisfy this rule, although they have BASLARC scores of 20 or more.

C. If ozone precursor emissions are dominated by internal combustion engines, as in highly urbanized valleys or basins, and if oil or gas extraction are absent, then winter ozone during inversions are below background, Rank 4, presumably because such emissions produce an abundance of compounds that actively scavenge ozone, even with ozone-favorable meteorology. The SaltLV is an example of this behavior. Because of similarities between ozone concentrations, the CV has also been assigned to Rank 4. CV is not as urbanized as SaltLV, but it does have significant urban and agricultural emissions [State of Idaho, Department of Environmental Quality, 2012].

The existence of an intense oil or gas industry is a stronger predictor of high ozone than meteorology for the basins studied. Nevertheless, one caveat bears repeating: CV, PRV, USRP, and WRB have only had winter ozone monitoring for one or two seasons at most. All our experience to date with the UB and UGRB indicates that one single season is not representative of multi-year climate patterns. As more winter ozone data accumulate, the above rules may need to be revised.

The steady climb of each trace in Figure 15 through the months of January and February has been attributed to two separate effects: the steadily increasing solar angle (see Figure 13), and the steadily increasing day length. Four of the traces in Figure 15 extend into the latter half of March. Those representing the UGRB and UB drop in the last half of March, which we attribute to melting of the snow pack. However, the other two traces, representing the WRB and the SaltLV, continue to climb into the last half of March, even though Figure 6 indicates that snow cover is usually gone by then. The WRB has sparse snow cover on average (see Figure 7), and it was especially sparse for the 2011 and 2012 season, which are the only seasons included in Figure 15. The fact that the Wind River trace continues to climb may therefore be due to limited sampling. However, the SaltLV trace includes data from eight separate winters, so the effect is probably real. SaltLV does not always have a persistent snow pack; there is often complete melting between storms (see below). Nevertheless, we expect it to have less snow on average in the latter half of March, so the continuing increase through March seems inconsistent with our expectation that winter ozone should correlate with the presence of a snow pack. This may instead by a manifestation of the “spring ozone maximum,” which refers to the well-documented fact that in many regions of the Northern Hemisphere, including many that rarely see snow (e.g., Hawaii and Greece), and for reasons that are not completely understood, ozone concentrations display a yearly

88 cycle that peaks in the spring [Monks, 2000]. We have also seen a lack of correlation between ozone and snow cover in the YC, and will comment on these results in Section 14 below.

Curiously, based on BASLARC scores (52 vs. 33 day/yr) and on oil/gas production (21 vs. 10 Mt/yr), we would predict that the UGRB should have higher ozone concentration than the UB. This result is not understood, but we can give three hypotheses:

1. It may be that inversions in the UGRB are more “leaky.” Drainage out of the UB is only through narrow Desolation Canyon, while the UGRB opens broadly on its southeastern flank. We also know that average February wind speeds in the UGRB are higher than in the UB, and, as we see below, a complete picture of the meteorology of the UGRB needs to include wind speed.

2. Perhaps both basins are VOC-limited. Then, a higher NOx emission in the UGRB could reduce its mean ozone concentration relative to UB’s.

3. Perhaps UB concentrations are higher because of blow-over of ozone precursors from the urbanized Wasatch Front to the west, i.e., from SaltLV and UV. No similar urban sources exist for the UGRB. (Nevertheless, see below for arguments against blow-over from the Wasatch Front.) At present, we tend to favor the first of these hypotheses, but in fact, any or all three could be contributory. More modeling and measurements would be required to select from among the three.

Predictions for the Ozone Concentration in the Remaining Five Basins

As previously mentioned, five basins have not had winter ozone measurements. Based on the three rules stated above, we can make predictions about their ozone concentrations. Three of the basins, BHB, SanLV, and UV, have ozone-favorable meteorology, with BASLARC scores of 20 or more days/year, but they have low to no petroleum production (1.2 Mt/y for BHB, zero for SanLV and UV). UV is urban and is expected to follow the pattern of SaltLV, so we predict it will be in Rank 4. BHB and SanLV are expected to follow the pattern expected of Rank 3. This leaves two basins, the FV and the LSRP, that have inversions (but not much snow because of their low altitude) and that have higher populations than many of the basins considered. Hence, they might follow the pattern of SaltLV and CV. Therefore, our prediction for both is that they will be in Rank 3 or 4.

Case-by-Case Examination of Ozone Seasons in Each of the Seven Basins

In this section, we present time series plots for many of the ozone seasons in the seven basins with existent ozone sampling. These are useful in demonstrating similarities and differences between the ozone systems in each of the seven basins. In most cases, daily values of the following four variables are plotted: ozone concentration (the daily maximum in the 8-hour running average), lapse rate, snow depth (both as determined by BASLARC), and the estimated intensity of ultraviolet radiation at its midday maximum. The daily ultraviolet intensity was calculated as outlined in Section 8, using the actinic flux tables of Finlayson-Pitts and Pitts [2000], switching from the no-snow to the snow case as the

89 average snow depth switches from below to above 50 mm, and vice versa. These radiation intensity curves are only meant to be suggestive. For example, they do not account for variations in cloud cover, nor do we wish to imply that the intensity switches discontinuously as the snow depth passes through 50 mm. The radiation intensity curves demonstrate the steady trend associated with the seasonal change in solar angle, and almost all of the time series plots displayed below show a correlation between ozone concentration and radiation intensity.

A. Uintah Basin

Time series plots for the UB appear in Figure 16a-d, for each of four seasons 2010-2013. These plots indicate many events, labeled A through P, where a negative lapse rate coincides with high ozone. Most of these events are multi-day, persistent inversions. We also see several late-winter events, labeled E, F, J, K, and P, when high ozone occurs even though the lapse rate is not extremely low. Presumably, this means that the UV intensity is so high that strong ozone events can occur during weaker inversions.

These plots demonstrate a strong correlation between snow cover, lapse rate, and ozone concentration. The snow pack generally disappears during the first half of March, and so also do the high ozone and the inversions. One of the reasons for displaying these plots for the UB is to indicate how different its ozone system seems to be from all the others. None of the other basins demonstrates the strong snow cover/lapse rate/ozone correlations that we see here.

Winter 2012 was unique in that there was no snowpack, and as a result there were very few inversion days and the ozone remained low.

In all four winters, either with or without snow, the ozone levels correlate well with UV intensity. Although it fluctuates, the ozone concentration is seen to trend with the UV intensity.

Snow depth, mm 400 350 300 250 200 150 100 50 0

8-hr ozone, ppb 150 120 90 60 30 0 03/12 F F 02/26 E E D D 02/12 C C Lapse rate Ozone UV radiation Snow depth 01/29 B B 01/15 A A 01/01

5 0 -5 20 15 10 -10 -15 -20 -25 Lapse rate, K/km rate, Lapse Uintah_2010.ps 2010/01/01 2010/03/15

80 40

200 160 120 UV radiation, W/m radiation, UV 2

Figure 16a. Ozone/lapse rate/snow depth time series for the UB, winter 2010. The daily BASLARC lapse rate is also shown. Events A-F are periods of high ozone, and are almost always correlated with minima in the lapse rate.

Snow depth, mm 400 350 300 250 200 150 100 50 0

8-hr ozone, ppb 150 120 90 60 30 0 03/12 K K 02/26 J J I I 02/12 H H Lapse rate Ozone UV radiation Snow depth 01/29 01/15 G G 01/01

5 0 -5 20 15 10 -10 -15 -20 -25 Lapse rate, K/km rate, Lapse Uintah_2011.ps 2011/01/01 2011/03/15

80 40

200 160 120 UV radiation, W/m radiation, UV 2

Figure 16b. Ozone/lapse rate/snow depth time series for the UB, winter 2011. The daily BASLARC lapse rate is also shown. Events G-K are periods of high ozone, and are almost always correlated with minima in the lapse rate.

Snow depth, mm 400 350 300 250 200 150 100 50 0

8-hr ozone, ppb 150 120 90 60 30 0 03/11 02/26 02/12 Lapse rate Ozone UV radiation Snow depth 01/29 01/15 01/01

5 0 -5 20 15 10 -10 -15 -20 -25 Lapse rate, K/km rate, Lapse Uintah_2012.ps 2012/01/01 2012/03/15

80 40

200 160 120 UV radiation, W/m radiation, UV 2

Figure 16c. Ozone/lapse rate/snow depth time series for the UB, winter 2012. The daily BASLARC lapse rate is also shown. No periods of high ozone were observed in 2012.

Snow depth, mm 400 350 300 250 200 150 100 50 0

8-hr ozone, ppb 150 120 90 60 30 0 03/12 Q Q P P 02/26 O O 02/12 N N Lapse rate Ozone UV radiation Snow depth 01/29 M M 01/15 L L 01/01

5 0 -5 20 15 10 -10 -15 -20 -25 Lapse rate, K/km rate, Lapse Uintah_2013.ps 2013/01/01 2013/03/15

80 40

200 160 120 UV radiation, W/m radiation, UV 2

Figure 16d. Ozone/lapse rate/snow depth time series for the UB, winter 2013. The daily BASLARC lapse rate is also shown. Events L-Q are periods of high ozone, and are almost always correlated with minima in the lapse rate.

B. Upper Green River Basin

Figure 17a,b shows time series for two heavy ozone seasons in the UGRB, namely years 2008 and 2011. The correlation between lapse rate and ozone here is not as strong as that observed in the UB. Some high ozone events (C, E, and G) do coincide with a low lapse rate, but others (A, B, D, F, and H) do not.

Perhaps the BASLARC predictions in the UGRB are suffering from the “small-basin” problem described above: As seen in Appendix B, we are using data from ten different stations, but gaps in the BASLARC time series indicate that not all ten are continuously operating. Several wild fluctuations in the BASLARC lapse rate are observed in the data (events I, J, K, and L) that might also imply that the BASLARC calculation is not functioning properly for the UGRB. As mentioned, we were able to confirm the BASLARC predictions of inversions in several basins against those of NAM-ANL, but time constraints have not yet permitted us to do the same for the UGRB. Such comparisons will be done for events I through L and at other times. If they indicate anything significant, we will post an addendum to this report.

Part of the issue seems to be is that inversions in the UGRB are more dynamic than in the UB. Figure 18a,b shows time series plots for the same two seasons (2008 and 2011) in which we have displayed average wind speed for the basin rather than the lapse rate. The UGRB ozone apparently is more strongly correlated with wind speed (events M through W) than with the BASLARC lapse rate. Our regression analysis (Mansfield and Hall, 2013) of winter ozone in both basins also saw stronger correlations with wind speed in the UGRB. Above, we speculated that the UGRB inversions are “leaky;” if the air mass is more dynamic, sloshing about in the basin, then perhaps it confuses the BASLARC lapse rate calculation.

Snow depth, mm 750 625 500 375 250 125 0

8-hr ozone, ppb 125 100 75 50 25 0 D C C B A I Lapse rate Ozone UV radiation Snow depth 12/15 12/29 01/12 01/26 02/09 02/23 03/08

30 20 10 -10 -20 -30 Lapse rate, K/km rate, Lapse UpperGreen_2008.ps 2007/12/15 2008/03/15

80 40

200 160 120 UV radiation, W/m radiation, UV 2

Figure 17a. Ozone/lapse rate/snow depth time series for the UGRB, winter 2008. Also shown is daily BASLARC lapse rate. Events A-D are high ozone periods that are often, but not always, correlated with minima in the lapse rate. Event I is a wide fluctuation in the lapse rate and may be spurious.

Snow depth, mm 750 625 500 375 250 125 0

8-hr ozone, ppb 125 100 75 50 25 0 H G G F E E Lapse rate Ozone UV radiation Snow depth L K J 12/15 12/29 01/12 01/26 02/09 02/23 03/09

30 20 10 -10 -20 -30 Lapse rate, K/km rate, Lapse UpperGreen_2011.ps 2010/12/15 2011/03/15

80 40

200 160 120 UV radiation, W/m radiation, UV 2

Figure 17b. Ozone/lapse rate/snow depth time series for the UGRB, winter 2011. Also shown is daily BASLARC lapse rate. Events E-H are high ozone periods that are often, but not always, correlated with minima in the lapse rate. Events J-L are wide fluctuations in the lapse rate and may be spurious.

Snow depth, mm 750 625 500 375 250 125 0

8-hr ozone, ppb 125 100 75 50 25 0 R R Q Q P P O O N N M M Wind speed Ozone UV radiation Snow depth 12/15 12/29 01/12 01/26 02/09 02/23 03/08

8 6 4 2 0 10 Wind speed, m/s speed, Wind WUpperGreen_2008.ps 2007/12/15 2008/03/15

80 40

200 160 120 UV radiation, W/m radiation, UV 2

Figure 18a. Ozone/wind speed/snow depth time series for the UGRB, winter 2008. Also shown is the daily average wind speed. Ozone maxima (events M-R) are almost always correlated with wind speed minima.

Snow depth, mm 750 625 500 375 250 125 0

8-hr ozone, ppb 125 100 75 50 25 0 W W V V U U T T S S Wind speed Ozone UV radiation Snow depth 12/15 12/29 01/12 01/26 02/09 02/23 03/09

8 6 4 2 0 10 Wind speed, m/s speed, Wind WUpperGreen_2011.ps 2010/12/15 2011/03/15

80 40

200 160 120 UV radiation, W/m radiation, UV 2

Figure 18b. Ozone/wind speed/snow depth time series for the UGRB, winter 2011. Also shown is the daily average wind speed. Ozone maxima (events S through W) are almost always correlated with wind speed minima.

C. Price River Valley

Figure 19 shows the ozone/lapse rate time series for the winter of 2013, the only season with winter ozone measurements. A correlation between lapse rate and snow depth is obvious. The ozone concentration is lower on average during the more intense inversions, with the lowest ozone, below 30 ppb, occurring during an intense inversion. This could be explained if the inversions are acting to trap pollutants that scavenge ozone. However, this ozone minimum occurs in January, so it may instead be due to low solar elevation.

100

Snow depth, mm 175 140 105 70 35 0

8-hr ozone, ppb 60 50 40 30 20 10 0 Lapse rate Ozone UV radiation Snow depth 11/01 11/15 11/29 12/13 12/27 01/10 01/24 02/07 02/21 03/07

5 -5 35 25 15 -15 -25 -35 Lapse rate, K/km rate, Lapse Price_2013.ps 2012/11/01 2013/03/15

90 60 30

150 120 UV radiation, W/m radiation, UV 2

Figure 19. Ozone/lapse rate/snow depth time series for the PRV, winter 2013.

101

D. Wind River Basin

The WRB is seen to have intense inversions, with lapse rates as low as about –40 K/km. As seen in Figure 20a,b, inversions are stabilized by the snowpack. All of the intense inversions, events A through H, occur with a snowpack present. In spite of these inversions and snow cover, however, ozone is not high. The highest 8-hr maximum is at about 55 ppb.

102

Snow depth, mm 240 200 160 120 80 40 0

8-hr ozone, ppb 60 50 40 30 20 10 0 03/10 E 02/24 D C Lapse rate Ozone UV radiation Snow depth B 02/10 A 01/27

5 -5 45 35 25 15 -15 -25 -35 -45 Lapse rate, K/km rate, Lapse WindRiver_2011.ps 2011/01/27 2011/03/15

90 60 30

180 150 120 UV radiation, W/m radiation, UV 2

Figure 20A. Ozone/lapse rate/snow depth time series for the WRB, winter 2011.

103

Snow depth, mm 240 200 160 120 80 40 0

8-hr ozone, ppb 60 50 40 30 20 10 0 Lapse rate Ozone UV radiation Snow depth H G F 12/15 12/29 01/12 01/26 02/09 02/23 03/08

5 -5 45 35 25 15 -15 -25 -35 -45 Lapse rate, K/km rate, Lapse WindRiver_2012.ps 2011/12/15 2012/03/15

90 60 30

180 150 120 UV radiation, W/m radiation, UV 2

Figure 20B. Ozone/lapse rate/snow depth time series for the WRB, winter 2012.

104

E. Upper Snake River Plain

Figure 21 displays the time series for winter 2013 in the USRP. Several multi-day inversion events are seen (labeled A through D in Figure 21). As mentioned, NOAA’s NAM-ANL model has confirmed that inversions are indeed occurring on these days. The ozone system here seems to be similar to that of the WRB: Inversions are occurring, but they are not correlated with ozone concentration. During this seven-week period, the 8-hr ozone concentration never gets above about 51 ppb.

105

Snow depth, mm 200 175 150 125 100 75 50 25 0

8-hr ozone, ppb 60 50 40 30 20 10 0 D 02/26 C 02/12 Lapse rate Ozone UV radiation Snow depth B 01/29 A 01/15

5 0 -5 -10 -15 -20 -25 Lapse rate, K/km rate, Lapse IFalls_2013.ps 2013/01/15 2013/03/11

80 40

200 160 120 UV radiation, W/m radiation, UV 2

Figure 21. Ozone/lapse rate/snow depth time series for the USRP, winter 2013. Events A through D are multiday inversions, but they are not accompanied by ozone maxima.

106

F. Cache Valley

CV had continuous ozone monitoring from 02 November 2012 to 15 March 2013. Its time series appears in Figure 22. Inversions are stabilized by the snow pack, but there do not seem to be any correlations between ozone concentration and lapse rate. Over the entire monitoring season, the 8-hr ozone is less than 45 ppb, and on about 16% of the sampling days, it is less than 20 ppb. Such low ozone numbers indicate the presence of an abundance of ozone scavengers in the air.

107

Snow depth, mm 450 400 350 300 250 200 150 100 50 0

8-hr ozone, ppb 45 30 15 0 Lapse rate Ozone UV radiation Snow depth 11/02 11/16 11/30 12/14 12/28 01/11 01/25 02/08 02/22 03/08

50 40 30 20 10 -10 -20 -30 -40 Lapse rate, K/km rate, Lapse Cache_2013.ps 2012/11/02 2013/03/15

90 60 30

180 150 120 UV radiation, W/m radiation, UV 2

Figure 22. Ozone/lapse rate/snow depth time series for CV, winter 2013.

108

G. Salt Lake Valley

Figure 23a-f displays six winter seasons, each one extending from December through March. The snowpack is not always persistent throughout each winter; we often see complete melting between storms. The ozone concentration rarely gets above 50 ppb, and then only in March. On 12% of the days between December and March, the ozone is less than 10 ppb, and as in the CV, such low concentrations must be the result of ozone scavenging. Contrary to the behavior in the UB, there are events in which a lapse rate minimum corresponds with an ozone minimum. This not only implies ozone scavenging, but it implies entrapment of scavengers by the inversions.

109

Snow depth, mm 350 300 250 200 150 100 50 0

8-hr ozone, ppb 60 50 40 30 20 10 0 Lapse rate Ozone UV radiation Snow depth 12/01 12/15 12/29 01/12 01/26 02/09 02/23 03/09 03/23

5 0 -5 20 15 10 -10 -15 -20 -25 Lapse rate, K/km rate, Lapse SaltLake_2006.ps 2005/12/01 2006/03/31

250 200 150 100 UV radiation, W/m radiation, UV 2

Figure 23A. Ozone/lapse rate/snow depth time series for SaltLV, winter 2006.

110

Snow depth, mm 350 300 250 200 150 100 50 0

8-hr ozone, ppb 60 50 40 30 20 10 0 Lapse rate Ozone UV radiation Snow depth 12/01 12/15 12/29 01/12 01/26 02/09 02/23 03/09 03/23

5 0 -5 20 15 10 -10 -15 -20 -25 Lapse rate, K/km rate, Lapse SaltLake_2007.ps 2006/12/01 2007/03/31

250 200 150 100 UV radiation, W/m radiation, UV 2

Figure 23B. Ozone/lapse rate/snow depth time series for SaltLV, winter 2007.

111

Snow depth, mm 350 300 250 200 150 100 50 0

8-hr ozone, ppb 60 50 40 30 20 10 0 Lapse rate Ozone UV radiation Snow depth 12/01 12/15 12/29 01/12 01/26 02/09 02/23 03/08 03/22

5 0 -5 20 15 10 -10 -15 -20 -25 Lapse rate, K/km rate, Lapse SaltLake_2008.ps 2007/12/01 2008/03/31

250 200 150 100 UV radiation, W/m radiation, UV 2

Figure 23C. Ozone/lapse rate/snow depth time series for SaltLV, winter 2008.

112

Snow depth, mm 350 300 250 200 150 100 50 0

8-hr ozone, ppb 60 50 40 30 20 10 0 Lapse rate Ozone UV radiation Snow depth 12/01 12/15 12/29 01/12 01/26 02/09 02/23 03/09 03/23

5 0 -5 20 15 10 -10 -15 -20 -25 Lapse rate, K/km rate, Lapse SaltLake_2009.ps 2008/12/01 2009/03/31

250 200 150 100 UV radiation, W/m radiation, UV 2

Figure 23D. Ozone/lapse rate/snow depth time series for SaltLV, winter 2009.

113

Snow depth, mm 350 300 250 200 150 100 50 0

8-hr ozone, ppb 60 50 40 30 20 10 0 Lapse rate Ozone UV radiation Snow depth 12/01 12/15 12/29 01/12 01/26 02/09 02/23 03/09 03/23

5 0 -5 20 15 10 -10 -15 -20 -25 Lapse rate, K/km rate, Lapse SaltLake_2010.ps 2009/12/01 2010/03/31

250 200 150 100 UV radiation, W/m radiation, UV 2

Figure 23E. Ozone/lapse rate/snow depth time series for SaltLV, winter 2010.

114

Snow depth, mm 350 300 250 200 150 100 50 0

8-hr ozone, ppb 60 50 40 30 20 10 0 Lapse rate Ozone UV radiation Snow depth 12/01 12/15 12/29 01/12 01/26 02/09 02/23 03/09 03/23

5 0 -5 20 15 10 -10 -15 -20 -25 Lapse rate, K/km rate, Lapse SaltLake_2011.ps 2010/12/01 2011/03/31

250 200 150 100 UV radiation, W/m radiation, UV 2

Figure 23F. Ozone/lapse rate/snow depth time series for SaltLV, winter 2011.

115

Yellowstone Caldera

In one manner of speaking, winter ozone measurements also exist for the YC. The National Park Service conducts continuous ozone monitoring between 1 April and 31 October, and spring comes late to Yellowstone, with the snowpack often lasting until May. The ozone concentrations are available at the EPA-AQS (see below for the URL). Furthermore, the average ozone concentration in April and May is rather high, but does not appear to be correlated with the presence of snow cover. For example, the average for April is about 55 ppb – very similar to the UGRB in February – and it drops from there to lower values in summer, see Figure 24. (However, NAAQS exceedences are very rare.) Based on our experience with the UB and UGRB, one might speculate that these high ozone numbers in Yellowstone are the result of elevated UV radiation brought about by reflection by the snow cover. However, as Figure 25 demonstrates, we do not see any correlation between snow cover and ozone concentration during the first two weeks of May. (In contrast, the UB shows significant correlation between snow cover and ozone concentration during the first two weeks of March.)

BASLARC calculations on Yellowstone have a serious small-basin problem. As can be seen in Appendix B, there are only six available stations. Whenever for any reason two or more of those do not report a temperature, no lapse rate is recorded. Although ozone monitoring starts on 1 April, and the snow pack usually persists past 1 May, lapse rate coverage is usually spotty in April, so it is difficult to determine, during April, if there are any correlations between lapse rate and ozone concentration. Because of the spotty nature of the lapse rate time series, its BASLARC score of 38 day/yr should probably be regarded as a poor approximation.

Apparently the ozone system in Yellowstone is fundamentally different from that of the UB. Perhaps this behavior is also a manifestation of the “spring ozone maximum” discussed above [Monks, 2000]

Mean ozone concentrations, Yellowstone Caldera mean [O3], ppb

Apr May Jun Jul Aug Sep Oct

30 35 40 45 50 55 60 Figure 24. Average ozone concentrations at Lake Yellowstone, Wyoming.

116

Ozone distribution (8-hr max) May 1 to May 15 Yellowstone Caldera

NO SNOW SNOW COVER COVER

80 to 84.9

75 to 79.9

70 to 74.9

65 to 69.9

60 to 64.9

55 to 59.9

50 to 54.9

45 to 49.9

40 to 44.9

35 to 39.9

Figure 25. Yellowstone ozone in May is not correlated with snow cover.

Blow-over from the Wasatch Front into the Uintah Basin

One school of thought maintains that blow-over from urban regions, particularly the SaltLV and UV, contributes significantly to winter ozone in the UB. Some transport of pollutants into the UB from the Wasatch Front probably occurs; however, there are two significant arguments against it being a major contributor to the UB’s ozone problem. First, it is not believed that air blowing over from SaltLV is effective at penetrating the cold pool of air found in the UB during thermal inversions. Second, the evidence seen so far indicates that the mix of precursors in the SaltLV is not particularly conducive to high ozone. As already mentioned, the SaltLV air seems to be good at scavenging ozone. We are currently constructing WRF (Weather Research and Forecasting) models of the airflow between the two

117 regions, and hope that these models will provide additional answers. (WRF is the state-of-the-art system for weather forecasting.) Nevertheless, any scenario in which blow-over contributes significantly to UB ozone would also have to explain why the UB has high winter ozone while the SaltLV does not. If anything, Wasatch Front air entering the UB should act to scrub out ozone.

Unanswered Questions

In this section, we note questions about the ozone systems in these various basins that remain unanswered.

1. If the steady climb exhibited by the traces appearing in Figure 15 is due to increasing solar elevation as the season progresses, and if snow cover always magnifies the available solar radiation, then why do only two of the four traces dip at the end of March? As mentioned above, the WRB behavior might be attributable to a small sample, but not that for the SaltLV. We see a similar inexplicable lack of correlation with snow cover in Yellowstone (Figure 25). We have suggested that these basins might be manifesting the “spring ozone maximum” [Monk 2000], but this question is still open.

2. Why is UB winter ozone higher than that of the UGRB, even though two indicators, the BASLARC score and the annual oil and gas production, seem to favor the UGRB? Above we have alluded to differences in drainage and flow patterns in the two basins, but this question is still unresolved.

3. Experience in the UB and the UGRB have conditioned us to expect ozone maxima during inversions; however, in some basins we see ozone minima during inversions. (USRP on 14 February 2013, SaltLV on a number of occasions.) A tentative explanation is that the inversion is trapping compounds that act to scrub ozone, but this is not established for certain.

4. Both UB and UGRB experience good and bad seasons for ozone. Most weather patterns last for a few days, but to have a bad ozone season, you need a persistent pattern that lasts for weeks. The UB ozone system has an “on/off switch”: A significant snowfall in December produces a snow pack that stabilizes cold pooling and keeps the atmosphere cold, thus ensuring that the snow pack endures until March, and as long as there is snow on the ground, we are at risk for high ozone. In other words, a one- or two-day event in December initiates a negative feedback loop that sets up a multi-week pattern. The UGRB, however, has a snow pack practically every year (for the six winter seasons from 2006 to 2011, the minimum January-February snow depth was about 125 mm; more than twice that is very common), but it has high ozone only during some winters. What is the switch in the UGRB that turns on a season of high ozone? In the UB, abundant snow cover seems enough to ensure high ozone, but this is not the case for the UGRB.

118

Conclusions

The process of comparing winter ozone phenomena in several basins helps us better understand the ozone system in any one. For example, BASLARC, a tool that we developed for characterizing inversions, indicates that high ozone occurs in the UB only during inversions, but BASLARC lapse rates alone are not good indicators of UGRB ozone. (The lapse rates must be combined with the wind speed.) We believe this indicates that UGRB inversions are more dynamic than those in the UB.

In some basins, such as SaltLV and CV, low winter ozone indicates ozone scavenging, but such scavenging is not occurring in the UB or UGRB systems. This is most likely attributable to the differences in precursor source emissions, which in the case of the UB and UGRB, is the oil and gas industry. The most obvious conclusion is that oil and gas emissions are ozone-productive, while the emissions in SaltLV and CV are not.

A survey of the seven basins for which winter ozone measurements are available indicates that intense oil and gas emissions are the strongest predictor of high winter ozone, and that intense emissions from other sources, vehicular or agricultural, are strong predictors of low winter ozone. Of course, it would be preferable to have data from additional basins and over a longer time span, but this conclusion is consistent with all the data presently available.

Acknowledgements

We would like to thank the staff at the Center for Advanced Energy Studies for facilitating the placement of an ozone monitor on their campus in Idaho Falls during the winter of 2013. We would also like to thank Cara Keslar, Wyoming Department of Environmental Quality, for consultations on ozone measurements in Pavillion, Wyoming (WRB). Financial support for this work was provided by the Uintah Impact Mitigation Special Service District and the Utah Science, Technology and Research Initiative.

119

References

A listing of books, periodicals, and websites consulted during the course of this study.

A. Websites providing meteorological data.

The Utah Climate Center website of Utah State University was the source of all the meteorological data used for the BASLARC analysis: http://climate.usurf.usu.edu/.

The Western Regional Climate Center of the Desert Research Institute was the source of wind speed data presented in Section 7: http://www.wrcc.dri.edu/htmlfiles/westwind.final.html.

B. Websites providing ozone data.

The Environmental Protection Agency (EPA) and Utah Division of Air Quality (UDAQ) data are available at EPA’s Air Quality System (AQS) website: www.epa.gov/ttn/airs/airsaqs/.

The Wyoming Department of Environmental Quality (WDEQ), Division of Air Quality ozone data for Pavillion, WY (BHB) are available at www.wyvisnet.com/data.aspx.

C. Websites providing oil, gas, and coal bed methane data.

Each individual state agency maintains a separate website, from which production data can be accessed both by county and by time period.

Utah Division of Oil, Gas, and Mining: http://oilgas.ogm.utah.gov/Data_Center.

Wyoming Oil and Gas Conservation Commission: http://wogcc.state.wy.us/.

Montana Board of Oil & Gas Conservation: http://www.bogc.dnrc.mt.gov/webapps/dataminer/.

Colorado Oil and Gas Conservation Commission: http://cogcc.state.co.us/cogis/ProductionSearch.asp.

D. Potential temperature maps for identifying thermal inversions.

NAM-ANL is an acronym for the North American Mesoscale model (analysis-only application), and is part of a model archiving system maintained by the National Operational Model Archive and Distribution System of the National Oceanic and Atmospheric Administration (NOAA). It is possible to download maps showing the location of cold pooling as predicted by the Weather Forecasting and Research (WRF) modeling system. These maps have been used to validate our BASLARC predictions about inversions in a few selected cases. The URL is http://nomads.ncdc.noaa.gov/NAM/analysis_only.

E. U.S. Census population data.

A convenient source for the population of a county as determined by the 2010 census is that county’s Wikipedia article: en.wikipedia.org/wiki.

120

F. Intensity of solar ultraviolet radiation.

Detailed algorithms for calculating the solar zenith angle as a function of latitude, longitude, time of day, and day of the year, and tables for the actinic flux as a function of solar zenith angle and wavelength: B.J. Finlayson-Pitts, J.N. Pitts, Jr., Chemistry of the Upper and Lower Atmosphere: Theory, Experiments, and Applications, Academic Press, New York, 2000.

G. Literature references to thermal inversions.

B.J. Billings, V. Grubisic, R.D. Borys, “Maintenance of a Mountain Valley Cold Pool: A Numerical Study, Monthly Weather Review, 134, 2266-2278 (2006).

P. Palacios, C. Luecke, and J. Robinson, "Climatology in the Bear Lake basin, Utah," Natural Resources and Environmental Issues: Vol. 14, Article 12 (2007). Available at: http://digitalcommons.usu.edu/ nrei/vol14/iss1/12.

C.D. Whiteman, X. Bian, S. Zhong, “Wintertime Evolution of the Temperature Inversion in the Colorado Plateau,” Journal of Applied Meteorology, 38, 1103-1117 (1999).

C.B. Clements, C.D. Whiteman, J.D. Horel, “Cold-Air-Pool Structure and Evolution in a Mountain Basin: Peter Sinks, Utah,” Journal of Applied Meteorology, 42, 752-768 (2003).

H. Other references cited.

A description of emissions sources in Cache Valley: State of Idaho, Department of Environmental Quality, Cache Valley Idaho PM2.5 Nonattainment Area State Implementation Plan, December 2012, www.deq.idaho.gov.

Another application of the BASLARC tool, and evidence of the differences between the Uintah and the Upper Green ozone systems: M.L. Mansfield and C.F. Hall, “Statistical Analysis of Winter Ozone Events,” accepted for publication in Air Quality, Atmosphere, and Health (2013).

A discussion of the spring ozone maximum: P.S. Monks, “A review of the observations and origins of the spring ozone maximum,” Atmospheric Environment, 34, 3545-3561 (2000).

121

APPENDIX A Population statistics; oil, natural gas, and coal bed methane production statistics. Production data in Bbl or Mcf appear in green; Mtonne in blue.

Oil§ Natural Gas§ Coal bed Methane§ Population‡ Bbls Mtonne Mcf Mtonne Mcf Mtonne BIG HORN BASIN Big Horn Co., WY 11,668 1,688,844 2,006,278 0 0.228 0.03846 0 Park Co., WY 28,205 7,068,314 10,190,982 0 0.955 0.1954 0 Washakie Co., WY 8,533 687,817 1,913,330 0 0.093 0.03668 0 TOTAL 48,406 1.28 Mt 0.27 Mt 0 Mt CACHE VALLEY Cache Co., UT 112,656 0 0 0 Franklin Co., ID 12,786 0 0 0 TOTAL 125,442 0 Mt 0 Mt 0 Mt FLATHEAD VALLEY Flathead Co., MT 90,928 0 0 0 Lake Co., MT 28,746 0 0 0 TOTAL 119,674 0 Mt 0 Mt 0 Mt LOWER SNAKE RIVER PLAIN Twin Falls Co., ID 77,230 0 0 0 Cassia Co., ID 22,952 0 0 0 Gooding Co., ID 15,464 0 0 0 Jerome Co., ID 22,374 0 0 0 Lincoln Co., ID 5,208 0 0 0 Minidoka Co., ID 20,069 0 0 0 TOTAL 163,297 0 Mt 0 Mt 0 Mt PRICE RIVER VALLEY Carbon Co., UT 21,403 80,859 0 90,970,761 0.01092 0 1.744 Emery Co., UT 10,860 2,440 0 10,904,667 0.00033 0 0.2090 TOTAL 32,263 0.01 Mt 0 Mt 1.95 Mt SALT LAKE VALLEY Salt Lake Co., UT 1,063,842 0 0 0 TOTAL 1,063,842 0 Mt 0 Mt 0 Mt ‡ = From 2010 Census. § = Annual production.

122

Oil§ Natural Gas§ Coal bed Methane§ Population‡ Bbls Mtonne Mcf Mtonne Mcf Mtonne SAN LUIS VALLEY Saguache Co., CO 6,108 0 0 0 Rio Grande Co., CO 11,982 0 0 0 Alamosa Co., CO 15,445 0 0 0 Conejos Co., CO 8,256 0 0 0 Costilla Co., CO 3,524 0 0 0 TOTAL 45,315 0 Mt 0 Mt 0 Mt UINTAH BASIN Duchesne Co., UT 18,607 14,415,407 41,674,583 0 1.948 0.7989 0 Uintah Co., UT 32,588 8,347,534 328,025,651 0 1.128 6.288 0 TOTAL 51,195 3.08 Mt 7.09 Mt 0 Mt UPPER GREEN RIVER BASIN Sublette Co., WY 10,247 6,796,014 1,071,917,224 0 0.9181 20.55 0 TOTAL 10,247 0.92 Mt 20.55 Mt 0 Mt UPPER SNAKE RIVER PLAIN Madison Co., ID 37,536 0 0 0 Jefferson Co., ID 26,140 0 0 0 Butte Co., ID 2,891 0 0 0 Bingham Co., ID 45,607 0 0 0 Bonneville Co., ID 104,234 0 0 0 TOTAL 216,408 0 Mt 0 Mt 0 Mt UTAH VALLEY Utah Co., UT 516,564 0 0 0 TOTAL 516,564 0 Mt 0 Mt 0 Mt WIND RIVER BASIN Fremont Co., WY 40,123 4,544,778 160,874,306 2,930,826 0.6140 3.084 0.0562 TOTAL 40,123 0.61 Mt 3.08 Mt 0.06 Mt YELLOWSTONE CALDERA Teton Co., WY 21,294 0 0 0 TOTAL 21,294 0 Mt 0 Mt 0 Mt ‡ = From 2010 Census. § = Annual production.

123

Notes:

(1) Census data is available at en.wikipedia.org/wiki.

(2) Oil and gas production data.

Utah Division of Oil, Gas, and Mining: http://oilgas.ogm.utah.gov/Data_Center (Select “Production.” Select “Prod by County.” Request “County Equals ”, then click SUBMIT.)

Wyoming Oil and Gas Conservation Commission: http://wogcc.state.wy.us/ (Select “County.” Then select “Table County Production.” Then select county name. For total oil and gas, unselect “CBM only.” For CBM only, select “CBM only.” Cut and paste results into a spreadsheet program.)

Montana Board of Oil & Gas Conservation: http://www.bogc.dnrc.mt.gov/webapps/dataminer/ Select “Production” then “Annual Production by County”

To date, Idaho has no oil or gas production.

Colorado Oil and Gas Conservation Commission: http://cogcc.state.co.us/cogis/ProductionSearch.asp

(3) The State of Utah database does not differentiate between natural gas and coal bed methane (CBM), reporting both as “gas.” However, it is safe to assume that Uintah and Duchesne counties produce only natural gas, and Carbon and Emery produce only CBM.

(4) Annual petroleum production is for the last complete calendar year, i.e., 2012.

(5) Conversion factor for oil. Oil production is measured by volume (barrels). Because different crudes have different densities, statistics based on mass are only approximate. Here we use the assumption that the density is 850 kg/m3. Also 1 barrel = 158.987 L, 1 Mbl = 1000 barrels.

1 bbl =

1 bbl 158.987 L m3 850 kg tonne Mt bbl 1000 L m3 1000 kg 106 tonne

= 1.351 × 10–7 Mt

(6) Conversion factor for natural gas and coal bed methane. Standard temperature and pressure for Mcf is 60 °F and 1 atm; conversion between volume and moles is assumed to follow ideal gas law:

124

n P 1 atm mol mol = = = 0.04220 = 42.20 V RT # # L atm & & L m3 %8.20574 ×10−2 % ( 288.75 K ( $ $ K mol ' '

Calculating mass in CH4 equivalents: 16.04 g/mol.

1 Mcf =

1 Mcf 1000 ft3 (0.3048)3 m3 42.20 mol 16.04 g Mt Mcf ft3 m3 mol 1012 g

= 1.917 × 10–8 Mt

125

APPENDIX B. METEOROLOGICAL STATIONS PROVIDING DATA

The following tables and maps list the meteorological stations consulted in each basin or valley. Maximum daily temperature and daily snow depth measurements at each station were accessed through Utah State University’s Utah Climate Center database. (See the References section for the URL.) Temperature-altitude data were used to estimate the daily lapse rate as described in the main body of the report.

Many stations are named for the nearest city or town. For example “MAESER 9NW” indicates a station found approximately 9 mi. northwest of the town of Maeser, Utah.

UINTAH BASIN STATIONS Station Name Lat Long Elev (m) 01 ALTAMONT 40.356 –110.288 1942 02 DINOSAUR QUARRY AREA 40.438 –109.304 1464 03 DUCHESNE 40.168 –110.395 1682 04 FT DUCHESNE 40.284 –109.861 1540 05 JENSEN 40.364 –109.345 1449 06 MAESER 9NW 40.560 –109.664 1963 07 MYTON 40.194 –110.062 1548 08 NAPLES 0.8 N 40.441 –109.491 1583 09 NEOLA 40.418 –110.051 1814 10 OURAY 4 NE 40.134 –109.642 1422 11 ROOSEVELT 0.2 WSW 40.295 –110.004 1575 12 ROOSEVELT 1.1 SE 40.286 –109.983 1532 13 ROOSEVELT RADIO 40.288 –109.959 1528 14 VERNAL 1.5 WSW 40.446 –109.563 1662 15 VERNAL 2SW 40.427 –109.553 1667 16 VERNAL 3.7 WNW 40.480 –109.596 1726 17 VERNAL MUNICIPAL AIRPORT 40.442 –109.514 1603 18 VERNAL 40.430 –109.510 1608 19 DINOSAUR NATIONAL MONUMENT 40.244 –108.972 1804

Landmarks and towns appearing on the Uintah Basin map:

D Duchesne FD Fort Duchesne DC Dinosaur, CO A Altamont O Ouray B Bonanza M Myton J Jensen R Roosevelt V Vernal

126

Uintah.map.ps Sites for temperature study

0 km 50 km 100 km 150 km

Duchesne Uintah 16 17 02 V 09 14 08

L Moffat 15 18 A J R 01 11 05 04 19 Wasatch 12 13 FD DC 07 03 M D O 10

O3 B Rio Blanco

Carbon

Garfield

Uintah Basin

127

UPPER GREEN RIVER BASIN STATIONS Station Lat Long Elev (m) 1 BIG PINEY MARBLETON AP 42.584 –110.107 2124 2 BONDURANT 43.228 –110.436 1992 3 BOULDER REARING STN 42.716 –109.690 2115 4 CORA 42.959 –109.996 2268 5 DANIEL FISH HATCHERY 42.928 –110.127 2236 6 PINEDALE 13.8 NW 42.982 –110.087 2275 7 PINEDALE 13 SE 42.752 –109.670 2149 8 PINEDALE 42.875 –109.859 2200 9 BIG PINEY (AMOS) 42.560 –110.100 2124 10 BIG PINEY 42.536 –110.114 2079

Landmarks and towns on the Upper Green map

LB La Barge YP Yukon Peak GP Gannett Peak DM Dike Mountain WB War Bonnet Peak WR Wind River Peak L Lander D Daniel BS Big Sandy Dam Fn Farson Br Boulder P Pinedale Bt Bondurant BP Big Piney Fe Fontenelle K Kemmerer

128

UpperGreen3.map.ps Sites for temperature study

0 km 50 km 100 km 150 km

Teton

YP 02 Fremont GP Bt

06 05 04 08 DM D P Br 07 WB L O3 Lincoln Sublette 03 O3 WR 01 BP 09 10

BS LB Sweetwater Fn

Upper Green River Basin

129

WIND RIVER BASIN STATIONS Station Lat Long Elev (m) 1 BOYSEN DAM 43.405 –108.163 1479 2 BURRIS 43.366 –109.276 1865 3 DIVERSION DAM 43.228 –108.949 1699 4 LANDER 11 SSE 42.675 –108.669 1760 5 LANDER 1.7 N 42.857 –108.724 1613 6 LANDER 1N 42.845 –108.741 1621 7 LANDER 2.0 NNE 42.857 –108.719 1609 8 LANDER 5.7 SW 42.763 –108.793 1856 9 LANDER 6.0 SE 42.762 –108.654 1665 10 LANDER 8.7 SSE 42.710 –108.679 1773 11 LANDER 0.9 W 42.829 –108.745 1644 12 LANDER HUNT FLD AP 42.815 –108.726 1704 13 PAVILLION 43.246 –108.694 1658 14 RIVERTON 0.5 NW 43.029 –108.387 1518 15 RIVERTON 43.066 –108.477 1697 16 SHOSHONI 43.237 –108.110 1474 17 LANDER HUNT FIELD 42.810 –108.730 1693 18 RIVERTON RAOB 43.060 –108.480 1703 19 LANDER AP 42.818 –108.728 1694 20 RIVERTON 43.031 –108.374 1510

Landmarks and towns appearing on the Wind River map.

DM Dike Mountain WB War Bonnet Peak WR Wind River Peak L Lander P Pinedale R Riverton S Shoshoni BD Boysen Dam SP South Pass City

Two ozone stations are shown in blue: One near Pavillion, one near South Pass City. Ocean Lake is outlined in blue.

130

WindRiver.map.ps Sites for temperature study

0 km 50 km 100 km

Hot Springs

BD 02 01

O3 13 16 03 P S

15 R 18 14 20

Fremont 05 07 DM 06 L 11 1912 17 WB 09 Sublette 08 10 WR 04

Wind River Basin

131

SALT LAKE VALLEY STATIONS Station Name Lat Long Elev (m) 1 DRAPER 40.533 -111.833 1412 2 SALT LAKE CITY E BENCH 40.736 -111.817 1478 3 GARFIELD 40.724 -112.198 1310 4 CITY CREEK WTP 40.840 -111.831 1625 5 GRANGER 40.700 -111.967 1298 6 HERRIMAN 40.517 -112.033 1525 7 HOLLADAY 0.5 ESE 40.666 -111.812 1396 8 HOLLADAY 0.7 WNW 40.674 -111.832 1358 9 COTTONWOOD WEIR 40.624 -111.787 1512 10 KEARNS 40.667 -112.000 1360 11 LOWER MILL CREEK 40.700 -111.783 1513 12 MAGNA 1.2 W 40.703 -112.109 1334 13 MAGNA ASARCO FARM 40.700 -112.100 1315 14 MIDVALE 40.600 -111.917 1324 15 MTN DELL DAM 40.750 -111.722 1652 16 MOUNT OLYMPUS 0.5 SW 40.680 -111.795 1535 17 MURRAY 40.650 -111.900 1314 18 DRAPER-POINT OF MTN 40.488 -111.900 1372 19 RED BUTTE 1 40.767 -111.833 1498 20 REVERE 40.583 -112.083 1544 21 RIVERTON 40.517 -111.983 1421 22 SALTAIR SALT PLT 40.767 -112.117 1283 23 SALTAIR 40.783 -112.167 1284 24 SANDY 1.6 NE 40.599 -111.855 1409 25 SANDY 2.0 ESE 40.557 -111.821 1479 26 SANDY 2.0 NE 40.605 -111.852 1413 27 SANDY 2.1 WNW 40.585 -111.888 1363 28 SALT LAKE CITY 2.1 NE 40.789 -111.873 1551 29 SALT LAKE CITY 3.3 E 40.775 -111.866 1392 30 SALT LAKE CITY 3.6 SE 40.734 -111.833 1399 31 SALT LAKE CITY 4.5 SE 40.722 -111.817 1443 32 SALT LAKE CITY 4.9 SE 40.721 -111.825 1405 33 SALT LAKE CITY INTL AP 40.778 -111.969 1288 34 SALT LAKE CITY 40.767 -111.883 1312 35 SALT LAKE TRIAD CTR 40.771 -111.896 1305 36 SOUTH SALT LAKE 2.4 NE 40.735 -111.865 1317 37 TAYLORSVILLE 0.3 NNW 40.674 -111.932 1302 38 TAYLORSVILLE 2.4 SW 40.647 -111.964 1366 39 TERMINAL 40.750 -112.000 1290 40 BOUNTIFUL-VAL VERDA 40.855 -111.890 1384 41 WEST VALLEY CITY 2.5 ESE 40.680 -111.963 1316 42 WEST JORDAN 0.7 SW 40.593 -112.005 1439 43 WEST JORDAN 1.6 NNW 40.619 -112.011 1442 44 A S R RSCH LAB 40.700 -111.917 1296

132

Towns and landmarks on the Salt Lake Valley map.

Cv Centerville Bf Bountiful WC Woods Cross NS North Salt Lake EP Ensign Peak SL Salt Lake City SP Summit Park Ma Magna WV West Valley City Ta Taylorsville Mu Murray To Tooele BC Bingham Canyon He Herriman WJ West Jordan Sa Sandy Dr Draper Al Alta

SaltLake3.map.ps Sites for temperature study

0 km 10 km 20 km 30 km 40 km 50 km

WC 40 NS 04

EP 23 33 28 34SL 29 19 22 35 15 39 3002 SP 36 32 31 03 Ma 05 44 12 13 WV 11 41 37 0816 10 Mu Ta 07 17 38 09 43 WJ 26 14 2724 Al 42 20 Sa BC 25 01 To 06 Dr He 21 18

Salt Lake Valley

133

BIG HORN BASIN STATIONS Station Name Lat Long Elev (m) 1 BASIN 44.379 –108.031 1170 2 BLACK MTN 43.656 –107.737 1718 3 BUFFALO BILL DAM 44.508 –109.168 1577 4 CLARK 3 NE 44.937 –109.137 1247 5 CODY 12 SE 44.414 –108.901 1600 6 CODY 2.4 WSW 44.510 –109.099 1573 7 CODY 3.8 44.488 –108.992 1570 8 CODY 5.0 ESE 44.498 –108.956 1571 9 CODY 7.2 NNE 44.608 –108.977 1481 10 CODY 7.3 SW 44.453 –109.169 1652 11 CODY 44.515 –109.045 1547 12 DEAVER 44.892 –108.593 1251 13 GREYBULL S BIG HORN 44.517 –108.082 1199 14 GREYBULL 44.491 –108.041 1177 15 LOVELL 2.4 SSW 44.806 –108.413 1204 16 LOVELL 44.838 –108.404 1169 17 MEETEETSE 0.2 NNE 44.158 –108.870 1756 18 POWELL 1.5 SW 44.744 –108.784 1352 19 POWELL 3.9 ENE 44.783 –108.688 1307 20 POWELL 5.1 S 44.684 –108.740 1328 21 POWELL FLD STN 44.776 –108.759 1332 22 RAIRDEN 2WSW 44.179 –107.938 1225 23 SHELL 1NE 44.548 –107.763 1306 24 SUNSHINE 3 NE 44.076 –108.955 1901 25 TEN SLEEP 26.6 S 43.653 –107.373 1525 26 TEN SLEEP 4.8 NNW 44.102 –107.473 1381 27 TENSLEEP 4NE 44.065 –107.382 1468 28 THERMOPOLIS 1.2 SW 43.636 –108.230 1390 29 THERMOPOLIS 25WNW 43.723 –108.696 1679 30 THERMOPOLIS 9NE 43.754 –108.144 1305 31 THERMOPOLIS 0.9 SW 43.638 –108.225 1344 32 THERMOPOLIS 43.648 –108.204 1315 33 WAPITI 1NE 44.474 –109.426 1719 34 WORLAND 0.5 W 44.017 –107.965 1235 35 WORLAND 7.8 SW 43.920 –108.059 1261 36 WORLAND 43.966 –107.951 1272 37 CODY MUNI (AWOS) 44.510 –109.010 1553 38 SOUTH BIG HORN CO 44.510 –108.080 1201 39 WORLAND MUNICIPAL 43.960 –107.950 1294 40 HEART MTN 44.705 –108.956 1460 41 TENSLEEP 16SSE 43.811 –107.365 1426 42 WORLAND 44.011 –107.969 1237

134

CACHE VALLEY Station Name Lat Long Elev (m) 1 AMALGA 41.833 –111.900 1373 2 BENSON 1.0 SSW 41.787 –111.928 1350 3 CACHE JUNCTION 41.820 –111.980 1351 4 CLARKSTON 41.917 –112.050 1533 5 CLG WARD UTAH ST UNIV 41.667 –111.900 1369 6 CORNISH SUGAR FACTORY 41.983 –111.950 1366 7 CUTLER DAM UP&L 41.833 –112.056 1308 8 HILLS 0.9 SW 41.660 –111.890 1372 9 HYDE_PARK_0.7 41.805 –111.823 1376 10 HYRUM 41.639 –111.865 1418 11 LEWISTON 41.967 –111.833 1136 12 LOGAN 0.5 E 41.741 –111.825 1394 13 LOGAN 0.5 ENE 41.744 –111.826 1391 14 LOGAN 0.7 SW 41.732 –111.845 1376 15 LOGAN 0.9 SE 41.731 –111.822 1387 16 LOGAN 2.5 ENE 41.757 –111.793 1471 17 LOGAN-CACHE AIRPORT 41.780 –111.850 1358 18 LOGAN CACHE AP 41.787 –111.853 1358 19 LOGAN SUGAR FACTORY 41.700 –111.867 1366 20 LOGAN 41.783 –111.850 1358 21 LOGAN USU EXP STN 41.767 –111.817 1406 22 LOGAN UTAH ST UNIV 41.746 –111.803 1460 23 MENDON 0.3 NE 41.714 –111.975 1517 24 MILLVILLE 0.3 SE 41.677 –111.815 1418 25 MILLVILLE 41.683 –111.850 1402 26 NORTH LOGAN 0.7 41.772 –111.801 1440 27 NORTH LOGAN 0.8 ESE 41.769 –111.799 1448 28 PRESTON 0.8 SE 42.089 –111.864 1436 29 PRESTON 42.093 –111.868 1463 30 PROVIDENCE 0.8 WNW 41.711 –111.826 1380 31 RICHMOND 0.5 WNW 41.922 –111.819 1395 32 RICHMOND 1.6 NE 41.938 –111.790 1502 33 RICHMOND 41.906 –111.810 1427 34 SMITHFIELD 41.833 –111.917 1383 35 TRENTON 41.915 –111.913 1358 36 WESTON 42.036 –111.962 1452 37 LOGAN 5 SW EXP FARM 41.666 –111.891 1368 38 PROVIDENCE 0.5 NW 41.710 –111.820 1394

135

FLATHEAD VALLEY STATIONS Station Name Lat Long Elev (m) 1 ARLEE 16.7 WNW 47.295 –114.393 896 2 BIGFORK 13 S 47.875 –114.033 887 3 BIGFORK 2.6 SSE 48.023 –114.069 950 4 BIGFORK 3.2 ENE 48.074 –114.012 930 5 CHARLO 4.9 S 47.371 –114.162 828 6 COLUMBIA FALLS 7.7 S 48.2620 –114.156 932 7 CRESTON 48.189 –114.134 896 8 HOT SPRINGS 47.600 –114.683 884 9 KALISPELL 10.1 SW 48.091 –114.464 992 10 MARION 1.1 SE 48.096 –114.643 1236 11 POLSON .5 E 47.692 –114.151 889 12 POLSON KERR DAM 47.678 –114.242 832 13 POLSON 47.675 –114.101 917 14 ROLLINS 0.5 WSW 47.903 –114.208 895 15 RONAN 4.1 NE 47.573 –114.044 1040 16 ROUND BUTTE 1 NNW 47.538 –114.276 945 17 SAINT IGNATIUS 47.315 –114.098 884

LOWER SNAKE RIVER PLAIN Station Name Lat Long Elev (m) 1 AMERICAN FALLS 8.2 W 42.791 –113.015 1394 2 BURLEY MUNI AP 42.542 –113.766 1262 3 HAZELTON 42.597 –114.138 1238 4 HOLLISTER 42.353 –114.574 1379 5 JEROME CO AP 42.727 –114.456 1234 6 JEROME 42.733 –114.519 1140 7 MINIDOKA DAM 42.677 –113.500 1269 8 OAKLEY 42.234 –113.898 1390 9 PAUL 1ENE 42.628 –113.762 1265 10 SHOSHONE 1 WNW 42.938 –114.417 1204 11 TWIN FALLS 4.9 WSW 42.543 –114.552 1175 12 TWIN FALLS 6 E 42.546 –114.346 1207 13 TWIN FALLS-KMVT 42.581 –114.457 1119 14 TWIN FALLS SUN VLY RGNL AP 42.482 –114.487 1265 15 BURLEY MUNICIPAL AR 42.530 –113.760 1267 16 JEROME 42.730 –114.450 1233 17 TWIN FALLS/JOSLIN 42.480 –114.480 1265 18 BURLEY 2S 42.519 –113.803 1376 19 MALTA AVIATION 42.302 –113.335 1384

136

UPPER SNAKE RIVER PLAIN Station Name Lat Long Elev (m) 1 ARCO 17 SW 43.462 –113.556 1815 2 ARCO 43.636 –113.299 1623 3 BLACKFOOT FIRE DEPT 43.197 –112.353 1383 4 DUBOIS EXP STN 44.244 –112.201 1661 5 GROUSE 43.719 –113.547 1829 6 HAMER 4 NW 43.966 –112.264 1460 7 HOWE 43.783 –113.003 1469 8 IDAHO FALLS 16 SE 43.346 –111.785 1776 9 IDAHO FALLS 46 W 43.532 –112.942 1505 10 IDAHO FALLS FANNING FLD 43.516 –112.067 1441 11 IDAHO FALLS - KIFI 43.513 –112.013 1445 12 MONTEVIEW 44.037 –112.516 1475 13 REXBURG BYU IDAHO 43.808 –111.789 1525 14 SAINT ANTHONY 43.952 –111.679 1497 15 ST ANTHONY 1 WNW 43.969 –111.713 1509 16 SUGAR CITY 43.887 –111.737 1501 17 REXBURG 43.830 –111.800 1481 18 CRATERS OF THE MOON 43.465 –113.558 1797 19 IDAHO FALLS FAA AP 43.517 –112.067 1441 20 MACKAY LOST RIVER RS 43.918 –113.632 1797 21 ARCO IDAHO 43.623 –113.387 1640

PRICE RIVER VALLEY Station Name Lat Long Elev (m) 1 CASTLE DALE 39.208 –111.012 1713 2 CASTLEDALE HUNTER UP&L 39.174 –111.029 1725 3 FERRON 39.088 –111.132 1805 4 FLATTOP MOUNTAIN - CLEVELAND 39.338 –110.600 1865 5 HELPER-CARBON PLT 39.727 –110.866 1859 6 HIAWATHA 39.483 –111.017 2219 HUNTINGTON FARM UP&L 39.367 –111.067 1900 7 PRICE KPUC 39.617 –110.750 1797 8 PRICE 39.530 –110.810 1721 9 PRICE 3 E 39.608 –110.751 1778 10 PRICE BLM 39.599 –110.819 1690 11 PRICE CARBON CO AP 39.609 –110.755 1805 12 PRICE/CARBON COUNTY 39.545 –110.750 1785 13 PRICE/CARBON(RAMOS) 39.610 –110.750 1796 14 PRICE GAME FARM 39.617 –110.833 1702 15 SUNNYSIDE CITY 39.553 –110.388 1993 16 SUNNYSIDE CITY CTR 39.552 –110.385 1990 17 VICTOR 39.400 –110.683 1600 18 WELLINGTON 3 E 39.545 –110.686 1646 19 WELLINGTON 39.533 –110.717 1651

137

SAN LUIS VALLEY Station Name Lat Long Elev (m) 1 ALAMOSA 1.2 NE 37.480 –105.855 2301 2 ALAMOSA 1.6 NW 37.488 –105.890 2302 3 ALAMOSA 2.1 SSW 37.440 –105.885 2298 4 ALAMOSA 2S 37.442 –105.861 2296 5 ALAMOSA 4.1 ENE 37.500 –105.804 2299 6 ALAMOSA 6.9 NW 37.536 –105.965 2310 7 CENTER 1.2 NE 37.764 –106.093 2324 8 CENTER 4 SSW 37.707 –106.144 2339 9 CRESTONE 2 SE 37.981 –105.690 2440 10 DEL NORTE 2.5 ENE 37.691 –106.308 2391 11 DEL NORTE 2.6 W 37.674 –106.400 2423 12 DEL NORTE 2E 37.674 –106.325 2397 13 DEL NORTE 3.3 E 37.679 –106.292 2385 14 FORT GARLAND 5.8 ESE 37.390 –105.336 2547 15 GREAT SAND DUNES 7.0 SSW 37.652 –105.579 2405 16 GREAT SAND DUNES NM 37.733 –105.512 2494 17 LA GARITA 5.6 WSW 37.792 –106.341 2515 18 MANASSA 37.174 –105.939 2344 19 MONTE VISTA 2.7 SSE 37.546 –106.122 2333 20 MONTE VISTA 2W 37.581 –106.187 2345 21 MONTE VISTA 3.6 NE 37.613 –106.094 2330 22 MONTE VISTA 4.1 S 37.522 –106.161 2347 23 MONTE VISTA 7.0 SE 37.521 –106.039 2317 24 MONTE VISTA 0.7 W 37.580 –106.157 2340 25 SAGUACHE 2 WNW 38.099 –106.171 2384 26 SAN LUIS 8.8 SW 37.085 –105.496 2477 27 VILLA GROVE 8.2 W 38.236 –106.100 2801 28 WAVERLY 1W 37.430 –106.032 2317 29 ALAMOSA SAN LUIS VA 37.430 –105.860 2296 30 BLANCA 4 NW 37.479 –105.572 2390 31 SAGUACHE 38.086 –106.144 2347

138

UTAH VALLEY Station Name Lat Long Elev (m) 1 ALPINE 40.464 –111.771 1545 2 PROVO BYU 40.246 –111.651 1393 3 FAIRFIELD 40.270 –112.094 1487 4 PLEASANT GROVE 40.368 –111.734 1436 5 PLEASANT GROVE UTAH 40.431 –111.750 1585 6 PLEASANT GROVE 1.0 WNW 40.378 –111.749 1407 7 SANTAQUIN CHLORINATOR 39.958 –111.779 1573 8 SPANISH FORK PWR HOUSE 40.080 –111.604 1439 9 LEHI 2.0 NNE 40.431 –111.844 1463 10 OREM TRTMT PLT 40.277 –111.737 1375 11 OREM 2.6 S 40.261 –111.689 1440 12 PROVO MUNI (AWOS) 40.210 –111.710 1369 13 SPANISH FORK 1.3 SSE 40.098 –111.628 1441 14 SPANISH FORK 1.0 W 40.114 –111.658 1396 15 UTAH LAKE LEHI 40.360 –111.897 1373

YELLOWSTONE CALDERA Station Name Lat Long Elev (m) 1 LAKE YELLOWSTONE 44.562 –110.399 2399 2 OLD FAITHFUL 44.457 –110.833 2243 3 SNAKE RIVER 44.133 –110.666 2098 4 YELLOWSTONE LAKE 44.544 –110.421 2388 5 YELLOWSTONE NP E ENT 44.489 –110.004 2119 6 YELLOWSTONE LAKE RA 44.550 –110.410 2368

139

ANALYSES OF HEALTH IMPACTS AND RISKS DUE TO UINTAH BASIN AIR QUALITY

Courtney Hall, Colten Dofelmire, Chad Mangum, Marc Mansfield, Seth Lyman, Howard Shorthill

Abstract

This report documents two different studies on health impacts and risks due to air quality in the Uintah Basin.

The first study is a statistical analysis exploring correlations between ozone concentrations and reported asthma cases in the Uintah Basin. Ozone is known to exacerbate asthma symptoms. We compared asthma cases at Ashley Regional Medical Center in Vernal, UT, to ozone concentrations at Ouray, UT, where high ozone concentrations have been measured in the winter. We could not find any definitive relationship or correlation between ozone concentration and asthma cases. Our inability to elucidate these relationships may be at least partially due to the small dataset available.

The second study applies EPA-recommended analyses to assess possible health risks due to toxic air contaminants (TAC) in the Uintah Basin. Such risks are categorized as acute (short term) or chronic (long term). The only acute risk identified in this study is that of ozone. Results for chronic risks are less conclusive because they rely on year-round measurements. (Since our primary concern has always been winter air quality, year-round measurements are not available.) We have determined that benzene, a compound known to pose both cancer and non-cancer chronic hazards, is likely to present low risk, but could present a marginal risk if it turns out that its average summer concentration is high. If the summer benzene concentration proves to be more than about 50% higher than the winter concentration, then we begin to cross over a threshold at which the EPA believes, “some sort of remediation is desirable.” Summertime measurements of benzene concentrations in the Basin, therefore, are called for.

141 Study 1. Winter Ozone and Asthma in the Uintah Basin

Introduction

Elevated surface ozone concentrations have been observed in the Uintah Basin since 2009. While ozone in the stratosphere protects us from the sun, “breathing air containing ozone can reduce lung function and increase respiratory symptoms, thereby aggravating asthma or other respiratory conditions” [Environmental Protection Agency, 2013]. Our objective here is to determine if high winter ozone concentrations are correlated with asthma-related hospital visits at Ashley Regional Medical Center.

A Utah Department of Health document, Asthma in Utah: Burden Report 2012, [Beck and Baxter, 2012] reports that Uintah and Duchesne counties have higher asthma emergency department visits and higher asthma hospitalization rates than the state average. This has led to a misconception that the two counties have higher asthma rates than the state average, when in fact the same report finds that Uintah and Duchesne counties are “similar to [the] state” average for the “prevalence of current asthma [… in] Utah adults.” Closer inspection of the report indicates the most likely explanation for this discrepancy. Those counties with higher than average emergency department visits or higher than average hospitalization rates turn out to be predominantly rural, while the opposite is true of the more urban regions of the state. In other words, the Uintah Basin does not seem to be atypical in terms of total asthma cases, but it does seem to follow the well-known pattern that individuals in rural areas are more likely to seek medical treatment at a hospital emergency room than at a clinic. The Department of Health Burden Report does not specifically address ozone concentrations in the Basin, but it also does not find an asthma rate in the Basin significantly different from the state average.

High ozone concentrations are known to increase airway responsiveness to allergens in asthmatics [Kehrl et al., 1999]. Several studies have linked high ozone concentrations with increased asthma cases in hospitals [Silverman and Ito, 2010; Lin et al., 2008; Mar and Koenig, 2009; Babin et al., 2008; Strickland et al., 2010]. These studies were performed in areas where ozone levels are highest in the spring and summer.

Asthma is either caused or exacerbated by many factors, including genetics, diet, obesity, viral infections, allergen exposures, endotoxins, and atmospheric pollutants, including ozone. In fact, even exposure to non-immunological stimuli (strong smells, cold air, fog, smoke, exercise, dust, etc.) can exacerbate asthma symptoms in susceptible individuals [Sousa et al., 2013; Lenney, 2009; Corrigan, 2012]. Because there are so many potential triggers, isolating the effect of any one stimulus, such as ozone, is difficult except with large data sets. In essence, the ozone signal is swamped by the signals from all other stimuli. The various studies cited above were drawn either from large population distributions or long time series. As we show in this report, detecting ozone-asthma connections in small population centers such as the Uintah Basin is probably not possible, even when ozone concentrations exceed EPA standards.

A study connecting winter ozone concentrations and clinic visits for adverse respiratory-related effects was recently performed in Sublette County, Wyoming [Pride et al., 2013], which is the only other region known to have local wintertime ozone production leading to exceedances of EPA air quality standards. The study covered a 4-year period from 1/1/2008 to 12/31/2011, and found a 3% increase in clinic visits for every 10 ppb increase in ground-level ozone on the previous day. (The 95% confidence limits are 0%

142 and 7%.) The study was designed to compare the ozone exposure on some given day with that of the same days of the week within the same month. (In other words, the third Wednesday in February 2009 is compared with all other Wednesdays in February.) This allows the study to control for weekly and for seasonal trends. Cases of adverse respiratory effects were identified from electronic billing records generated by the only two clinics in the area, and were based on diagnostic codes corresponding to diagnoses of acute bronchitis, asthma, chronic obstructive pulmonary disease (COPD), pneumonia, upper respiratory tract infections, and “other respiratory.” The complete study included about 13,000 clinic visits, distributed as follows: upper respiratory infections, 73.3%; COPD, 15.4%; asthma, 6.2%; pneumonia, 2.4%; acute bronchitis and “other,” 1.4% each. Wyoming Department of Environmental Quality makes public announcements of ozone forecasts; these “ozone notification days” permit citizens to limit exposure and to voluntarily control emissions. Due to concerns that ozone forecasts would by themselves produce additional clinic visits, the authors performed follow-up analyses by eliminating notification days from the study. These follow-up calculations produced essentially the same result: a 3% uptick in next-day visits for every 10 ppb uptick in ozone concentration. Nevertheless, there are two obvious caveats to the study: (1) The study counts visits, not organic illnesses, and (2) the various diagnostic classifications include a broad spectrum of disease severity. It would be incorrect to say that the study finds a 3% increase in severe diseases like COPD or asthma.

Although there is much evidence that ozone affects respiratory outcomes in general, the relationship between ozone concentrations and health outcomes in the Uintah Basin is not well understood. For instance, cold weather may decrease exposure to winter ozone. Ozone levels are usually much lower indoors because ozone reacts quickly with surfaces such as furniture [World Health Organization, 2006; Geyh et al., 2000]. The effect of ambient ozone concentrations is weaker with a higher usage of air conditioning compared to open window ventilation [Geyh et al., 2000; Liu et al., 1997]. A heated building would probably provide similar protection. Outdoor ozone levels, therefore, may not accurately represent population exposure when cold weather forces individuals inside.

Additionally, heat seems to worsen the effect of high ozone concentrations. The effects of ozone on respiratory admissions increase during warmer weather and ozone adds to mortality risk during heat waves [World Health Organization, 2008; Filleul et al., 2006]. A study in the Atlanta, Georgia, metropolitan area found the impact of ozone on pediatric asthma or wheeze in the warm season to be about twice that of the cold season [Strickland et al., 2010]. A deeper understanding of the effect of winter ozone on the population of the Uintah Basin is needed.

As reported below, we examined seasonal asthma trends in Vernal and found a statistically significant asthma maximum in September. Pediatric asthma cases occur during a 3-week period between about August 25 and September 15 at three times the annual rate. Local physicians are familiar with the maximum and attribute it to exposure to pollen [Anderson, 2013]. It has been suggested that similar maxima in other regions are attributable to upper respiratory viral infections [Johnston et al., 1996; Carlsen et al., 1984; Johnston et al., 2005; Grech et al., 2002]. (See Appendix for additional discussion of seasonal asthma maxima.) In any case, ozone concentrations are low during this same period, so that ozone can be ruled out as the cause of the September maximum. Furthermore, linear regression calculations failed to find any correlation between ozone and asthma in the Uintah Basin, with R2 correlation coefficients of 0.02 or lower, see below.

143 Methods

Ashley Regional Medical Center in Vernal provided de-identified data for 424 asthma cases during a 3.5- year (1277-day) period between 1/1/2009 and 6/30/2012. The 424 cases fall in the following four categories: (1) patients presenting at the emergency room and then discharged, (2) patients presenting at the emergency room and then admitted to hospital, (3) direct admission to hospital by a pediatrician, (4) direct admission to hospital by a primary-care physician. The data do not include presentations to clinics that did not result in hospital admission. All age groups are represented, with 153 children’s cases (ages 0 to 17) and 271 adult cases (ages 18 and above).

Measured ozone concentrations were available for Ouray, UT, starting 7/29/2009. Ouray is about 30 miles south of Vernal, but ozone data for Vernal is not available for the entire time period. Ozone concentrations at Ouray during ozone episodes are almost always higher than at Vernal, but both sites lie within the same airshed, are typically covered by the same cold air pool during wintertime inversions, and have concentrations that usually are highly correlated [Mansfield and Hall, 2013]. For this study, we use the daily maximum in the 8-hour running average concentration.

Data on exposure to cigarette smoke and pets were incomplete, so these effects were not analyzed.

We examined the data for seasonal trends in asthma and also performed several linear regression analyses. As explained below, the seasonal asthma trend is uncorrelated with the seasonal ozone trend, and the linear regressions show no correlation between asthma cases and ozone.

Seasonal Asthma Trends in the Uintah Basin

Ashley Regional Medical Center registered 153 pediatric asthma cases (ages 0 – 17) and 271 adult cases (ages 18+) over the 1277-day period of the study, for rates of 153/1277 = 0.12 cases/day for children, 271/1277 = 0.21 cases/day for adults, and an overall rate of 0.33 cases/day. Figure 1 shows the daily case rate for each calendar month. We note that for the month of September, the case rate climbs to almost 0.5 cases/day, and that adults and children both contribute to this September maximum. A less prominent maximum also occurs in March, but it appears to be due only to adults. A more detailed analysis (see Appendix) sheds additional light on the September maximum. As we show there, the pediatric case rate for one 3-week period increases to 0.33 cases/day, almost three times higher than the year-round average. Interestingly, the beginning of the 3-week period coincides almost exactly with the opening of school. Whether or not this is coincidental is not known, but correlations between ozone cases and the school calendar have been reported in other regions (see Appendix).

The Appendix shows the odds that this 3-week maximum has occurred entirely by chance are about 23 in one million, indicating that its occurrence is almost certainly non-random, i.e., that it has some underlying cause. As already mentioned, possibilities are allergens or respiratory infections, or both. The March maximum may also have an underlying cause, but it may also be a random fluctuation.

Figure 1 also shows monthly average ozone concentrations. These are largest in January, February, and March, and do not seem to be well correlated with the occurrence of asthma. The lack of correlation is seen in the scatter plot of Figure 2, which plots average monthly case rate against average monthly ozone concentration. With a correlation coefficient of only 0.014, we have to conclude that ozone concentration does not have a noticeable effect on asthma cases in the basin.

144

Asthma cases by month, Ashley Regional Medical Center Monthly average ozone concentration, Ouray, Utah 75

Ozone Adults + Children 60 [O3], 8 hr, ppb 45

Adults 30 cases per day 15

Children 0 0.0 0.1 0.2 0.3 0.4 0.5 1 2 3 4 5 6 7 8 9 10 11 12 Month

Figure 1. Monthly average asthma cases per day, in red, green, or blue for children, adults, and the entire population, respectively. Monthly average ozone concentrations are also shown. Horizontal lines are overall averages. A statistically significant asthma maximum occurs in September.

145 Scatter plot, asthma cases vs. ozone concentration

Sep R2 = 0.014

Mar

Aug

May Dec Oct Montly average cases/day Feb

Jul Apr Jun Nov Jan 0.25 0.30 0.35 0.40 0.45 0.50 40 45 50 55 60 65 70 75 Monthly average [O3]

Figure 2. Scatter plot of monthly average case rate vs. monthly average ozone concentration. Also shown is the correlation line. The square of the correlation coefficient is 0.014.

Additional Evidence for Lack of Correlation between Ozone and Asthma

The distribution of ozone concentrations does not appear to depend on whether an asthma case has occurred (Figure 3). Ozone concentrations on days with an asthma case have a mean of 51 ppb and a standard deviation of 14.14 ppb. The mean for all of the concentrations for the period is 52 ppb and the standard deviation is 16 ppb, slightly higher than on days with asthma cases. An independent samples t- test for these means returned a p-value of 0.057, which indicates that the difference between the two means is only marginally significant. In other words, ozone concentrations on days with asthma cases do not appear to be different than ozone concentrations on days without asthma cases.

Winter ozone concentrations vary by year. For example, the 8-hour ozone concentration in Ouray rose above 75 ppb 42 times in 2010, but never in 2012. We compared the number of asthma cases by year,

146 focusing on only the first quarter of the year, when ozone concentrations are historically highest (Figure 4). A one-sample chi-square test confirmed that the variation in the number of asthma cases by year is insignificant (p-value of 0.601). One-sample chi-square tests also confirmed that asthma cases were not unequally distributed among the months of the year (p-value of 0.103) and among quarters of the year (p-value of 0.100). These tests suggest that asthma cases did not increase during the ozone season.

Figure 3. The bar chart on the left shows the distribution of ozone concentrations on days with no asthma cases, that on the right for days with asthma cases. Statistical tests verify that the two distributions are equivalent.

147

Figure 4. The number of asthma cases during the first quarter of each year is statistically the same in both good and bad ozone years.

148 Figure 5 shows a scatter plot giving the number of cases on any one day as a function of the ozone concentration on that day. In this case, the linear regression correlation coefficient is extremely small (0.003), indicating lack of correlation.

Scatter plot, asthma cases vs. ozone concentration

R2 = 0.003 Asthma cases per day 0123

0 25 50 75 100 125 150

[O3] (8 hr), ppb

Figure 5. Asthma cases per day vs. ozone concentration. The linear regression correlation line is also shown. The squared correlation coefficient is only 0.003.

149 Conclusions

We were unable to elucidate a relationship between ambient ozone concentrations and asthma-related hospital visits or admissions. Scientific studies have clearly documented that this relationship exists, but other factors, such as other triggers for asthma symptoms and less ozone exposure because of cold winter temperatures, have obscured the effects of ozone on asthma in these data. Scientific studies capable of detecting the effects of ozone on asthma apparently require large data sets in order to see the relatively small ozone signal. Ozone does not appear to be among the dominant factors causing asthma symptoms in the Uintah Basin.

The data do demonstrate a statistically significant asthma maximum in the Vernal area in September that is not related to ozone concentration, and which coincides, perhaps coincidentally, with the opening of the school year.

Future studies should include more types of respiratory cases, such as COPD, in addition to asthma. A study that examines the relationship between high ozone concentrations and lung volume in children or adults could also provide more information.

Acknowledgements

We are grateful for the generous financial support of the Uintah Impact Mitigation Special Service District and the Utah Science, Technology, and Research Initiative. We are also grateful to Arnita Perry and Debbie Spafford at Ashley Regional Medical Center for providing us with the asthma data.

150 Study 2. Health Risk Assessments of Uintah Basin Air Quality

Results from Sublette County, Wyoming, and Motivation of our Study

In response to public concerns over the health impacts of ozone and toxic air contaminants (TAC) in the Upper Green River Basin of Sublette County, Wyoming, the Sublette County Commissioners, the Wyoming Department of Health, and the Wyoming Department of Environmental Quality contracted with Sierra Research, Inc., to conduct a health risk assessment for the county [Walther, 2011]. From February 2009 to March 2010, Sierra monitored the ambient concentrations of 51 TAC and ozone. In regards to the health impacts of ozone, they concluded, “Because the highest measured concentration was less than the [NAAQS] standard [of 75 ppb], it can be concluded that ozone levels […] were low enough to avoid any direct health impacts (i.e., NAAQS are set to protect public health, including the most sensitive subpopulations of infants, the elderly and the ill, with a large margin of safety)” [Walther, 2011]. However, as luck would have it, the Wyoming study coincided with a low ozone season, and as is well known, Sublette County has experienced higher ozone concentrations in other winters. As for the other 51 TAC, the report concluded, “The estimated health impacts of the 51 TAC monitored in the study are not high enough to suggest a need for a more refined health risk assessment of the TAC in the ambient air in and near Sublette County” [Walther, 2011].

As part of the Uintah Basin air quality study, we have accumulated seasonal concentration data on a number of volatile organic compounds (VOC). The list of compounds we chose to measure was based on ozone reactivity, not on direct toxicity to humans. Nevertheless, our list contains known carcinogens, and so we thought it would be informative to submit our concentration data to an analysis similar to that performed in Wyoming. This is not intended to be an in-depth analysis of the toxicity of contaminants in the Uintah Basin airshed, but rather a preliminary assessment to determine if further study is warranted.

The concentration data that we use for this study were taken at two different sites in the Uintah Basin. The Horsepool site is in Wonsits Valley, central Uintah County, in a region of intense natural gas production, with data collection occurring between December 2012 and March 2013. The Roosevelt site is in the city of Roosevelt, eastern Duchesne County, near Constitution Park, with data collection from mid-December 2012 to the end of February 2013.

Preliminary Screening

Health effects of inhaled compounds are classified in three different ways: chronic cancer, chronic non- cancer, and acute. Chronic risks are those associated with long-term exposure and whose health effects are considered to be cumulative over a lifetime. Acute risks are those associated with a short-term (less than an hour or a day) exposure [EPA-904-B-06-001, 2010]. Following EPA recommendations [EPA-904- B-06-001, 2010], a preliminary screen of all of the detected compounds is performed by comparing the largest daily average concentration encountered in the dataset with a screening concentration. Each compound can be screened in this way for chronic cancer risk, chronic non-cancer risk, and acute risk. Appropriate screening factors have never been determined for all possible compounds, and we can only include compounds with known toxicities. Furthermore, this first analysis is only intended to be a

151 preliminary screening. Compounds that fail to pass are not necessarily posing a threat; rather the procedure serves to identify compounds that might require further attention.

We first discuss chronic non-cancer effects. For each compound, a screening value is obtained, and the compound is assumed to fail the screening if the maximum daily average concentration exceeds the screening value. The chronic non-cancer screening value is derived from its reference concentration (RfC), which is defined at the EPA/IRIS (Integrated Risk Information System) website as “an estimate (with uncertainty spanning perhaps an order of magnitude) of a continuous inhalation exposure to the human population (including sensitive subgroups) that is likely to be without an appreciable risk of deleterious effects during a lifetime” [EPA/IRIS]. Chronic non-cancer screening values are taken to be 10% of the RfC both by the Wyoming study [Walther 2011] and in recommendations from EPA, [EPA- 904-B-06-001, 2010] for essentially two reasons: first, in the interest of being more conservative, and second, to correct for synergistic hazards that might occur when individuals are exposed to mixtures rather than pure compounds. The screening values we use come from one or the other of those two documents.

Table 1 compares the maximum daily average concentration at Horsepool and Roosevelt of compounds for which a chronic non-cancer screening value is available. Compounds that fail the screening (i.e., those for which the maximum daily average exceeds the screening value) are highlighted in red.

Table 1. Preliminary non-cancer chronic screening of compounds in the air of the Uintah Basin. All concentrations are in μg/m3. Red ink indicates compounds that do not pass the screening. (N-hexane was not measured at Horsepool.) Maximum Daily Average Concentration Non-Cancer Chronic Compound Horsepool Roosevelt Screening Value (10% Of Rfc) benzene 19.2 27.0 3 cyclohexane 31.2 43.3 600 ethylbenzene 4.59 3.04 100 isopropylbenzene (cumene) 22.4 39.7 40 m,p-xylenes 3.07 13.3 10 o-xylene 3.13 2.73 10 n-hexane ----- 131 70 Styrene 20.3 41.8 100 Toluene 26.6 32.1 40

We next present the results of the chronic cancer screening. Again the procedure calls for us to compare the maximum daily average concentration against a screening value. Screening values for chronic cancer risks were selected either from the Wyoming or the EPA document, and protocols for their definition are also given in the EPA document [Walther, 2011; EPA-904-B-06-001, 2010]. Only two of our compounds have tabulated cancer screening values. The results are shown in Table 2, and all compounds that do not pass this initial screening are again highlighted in red.

152 Table 2. Preliminary cancer screening of compounds. All concentrations are in μg/m3. Red ink indicates compounds that do not pass the screening. Maximum Daily Average Concentration Compound Horsepool Roosevelt Cancer Screening Value benzene 19.2 27.0 0.13 ethylbenzene 4.59 3.04 0.4

Finally, we discuss the results of the acute screening study. Sources that we consulted for acute screening values were the Wyoming and EPA documents already mentioned [Walther, 2011; EPA-904-B- 06-001, 2010], the California Office of Environmental Health Hazard Assessment, [California OEHHA, 2013] and the EPA Acute Exposure Guidelines [EPA AEGL, 2013]. In the case of screening for acute risks, many more screening values are available. However, the procedure is essentially the same, compounds do not pass the screening if the maximum daily average concentration is greater than the screening value. Only one compound, ozone, failed the acute screening test.

In summary, five compounds, benzene, ethylbenzene, m-xylene, o-xylene, and n-hexane, are deemed to have failed the initial screening for chronic concerns. Only one compound, ozone, failed the initial screening for acute hazards. We reiterate that at this level of screening these compounds do not necessarily constitute a health risk. The purpose of the preliminary screening is to identify compounds for greater scrutiny. For example, in the case of the non-cancer chronic test, a compound fails if the highest daily average in the dataset is larger than 10% of a concentration which is “likely to be without deleterious effects” even when inhaled over a lifetime.

Additional Screening

Following EPA guidance [EPA-904-B-06-001, 2010], additional screening calculations are called for to characterize both the excess cancer risk (defined as the amount of risk from some specified source over and above the risk presented by all other sources) and the non-cancer chronic risk from all the compounds that failed to pass the preliminary screening.

Assessment of the chronic risk requires year-round averages of pollutant concentrations, but this presents a challenge for us, since measurements were only made in winter. We worked an estimation approach that gives the annual average concentrations of the compounds of interest (next section) listed in the following table. However, the estimation procedure may be inaccurate. We will use these estimates in the following calculations, and then point out afterwards the effects of any errors.

Table 3. Estimates of the average annual concentration of five compounds in the Uintah Basin (counting m- and p-xylene as separate compounds). Estimated Annual Compound Concentration µg/m3 benzene 3.5 ethylbenzene 0.43 m,p-xylenes 2.0 n-hexane 17

153

Assessment of the chronic cancer risk is obtained by dividing the annual average concentration by the screening value for those compounds for which a cancer screening value is available:

3.5 BENZENE: = 27 0.13

0.43 ETHYLBENZENE: =1.1 0.4

The accumulated cancer risk is the sum of the above ≈ 28, which is, interestingly, the same value obtained for Sublette County [Walther, 2011, p. 39]. Obviously, the cancer risk is dominated by benzene. Because of the way in which the chronic screening values are determined, this represents a lifetime excess cancer risk of 28 individuals in a population of one million. Quoting EPA:

The level of total cancer risk that is of concern is a matter of personal, community, and regulatory judgment. In general, the USEPA considers excess cancer risks that are below about 1 chance in 1,000,000 (1×10-6 or 1E-06) to be so small as to be negligible, and risks above 1E-04 [100 in 1,000,000] to be sufficiently large that some sort of remediation is desirable. Excess cancer risks that range between 1E-06 and 1E-04 are generally considered to be acceptable … although this is evaluated on a case-by-case basis and EPA may determine that risks lower than 1E-04 [100 in 1,000,000] are not sufficiently protective and warrant remedial action. [EPA Region 8, 2013]

The 100/1,000,000 threshold is reached for benzene at an annual average concentration of 13 µg/m3, that for ethylbenzene at a concentration of 40 µg/m3.

To assess the non-cancer chronic risk, each annual average concentration is divided by its chronic RfC value (defined above) to obtain its chronic health hazard index:

3.5 BENZENE: = 0.12 30

0.43 ETHYLBENZENE: = 0.00043 1000

2.0 m,p-XYLENES: = 0.020 100

17 n-HEXANE: = 0.024 700

154 The total chronic health hazard index is determined by summing all of the above:

0.12 + 0.00043 + 0.020 + 0.024 ≈ 0.16.

This index is also dominated by benzene. A total chronic health hazard index below 1.0 is considered to be less than significant [Walther, 2011].

In summary, our estimate of the cancer risk due to the VOC compounds measured in the air of the Uintah Basin is 28/1,000,000. Our estimate of the non-cancer chronic health hazard is 0.16. The acceptability thresholds established by EPA for these risks are 100/1,000,000 and 1.0, respectively. Benzene is the dominant contributor to both of these risks. The only compound identified to pose an acute risk to health in the Uintah Basin is ozone.

Estimates of the Annual Concentrations of Benzene and Additional VOC

Our measurements of VOC concentrations in the Uintah Basin only occurred in wintertime, while the computations of health risks given above require annual averages. Because of the obvious differences in meteorology, we expect summer and winter VOC concentrations to be considerably different.

Wintertime concentrations of VOC are correlated with the presence of thermal inversions because inversions indicate a stable, dense layer of air adjacent to the ground surface. The strength of a thermal inversion is quantified by the so-called lapse rate, which measures the rate of temperature rise or fall with increasing altitude. We have already estimated the lapse rate in the Uintah Basin on any given day for the last 60 years from the distribution of surface temperatures measured on that day. By comparing lapse rates and benzene concentrations on those days for which the pollutants have been measured, we find that concentrations are high when the atmosphere is inverted, and we were able to derive a mathematical formula to estimate the pollutant concentration on any given day from the lapse rate data. We then applied that formula over all 60 years of lapse rate data, permitting us to estimate pollutant concentration on any given day, summer or winter. These values were then averaged to obtain a year-round average for benzene of 3.5 µg/m3.

In emissions scenarios such as these, we find that concentrations of any two compounds obey proportionalities, e.g., the [butane]/[benzene] ratio is relatively constant from day to day. These proportionalities were obtained from the wintertime measurements, and applied to the annual averages, to obtain all of the other average concentrations given in Table 3.

The average benzene concentration in mid-January in Roosevelt is 8.5 µg/m3. The mid-July value obtained by this approximation technique is much lower at 2.0 µg/m3, and the annual average, as cited above, is 3.5 µg/m3. The technique obviously predicts that summer concentrations are lower, but it essentially equates the concentration on a non-inverted day in January to the concentration in July, which is, in fact, an untenable approximation. One could argue that concentrations in summer should be lower because the atmosphere is usually better mixed. On the other hand, one could argue that emissions of gases evaporating from sources such as tank batteries and evaporation ponds are higher in summer because of the elevated temperature. In the Wyoming study, the maximum observed concentrations almost always occurred in the summer [Walther, 2011, see for example Tables E-1 to E- 12]. Such maxima may be the result of spikes when an emission plume from some source passes over

155 the monitor, so they do not necessarily imply that the average summer concentration is greater than that of the winter. But then, there are also emission plumes blowing around in the winter. Based on these results, we believe it is prudent to recognize that we do not have good estimates of the year- round average concentrations of benzene or other VOC, and that it is difficult to know for certain if the summer concentrations are higher or lower than the winter ones.

If the annual average concentration of benzene turns out to be the same as the January value of 8.5 µg/m3, then the cancer risk calculated above rises to nearly 70/1,000,000, while the non-cancer chronic health hazard rises to nearly 0.4. If the summer average turns out to be about 50% higher than the winter average, then we start to enter the domain in which the EPA believes, “Some sort of remediation is desirable.” We believe, therefore, that monitoring of VOC in summer as well as in winter is called for, so that we are better able to assess the risks associated with benzene inhalation.

Conclusions

We have assessed both acute (short term) and chronic (long term) health risks. The only acute risk identified is that of ozone. The fact that ozone makes the list is no surprise, but it is reassuring that no other acute hazard was found. Our results for chronic health risks are inconclusive because a full assessment requires annual measurements, while for now only winter measurements have occurred. The only possible chronic hazard appears to be benzene (a carcinogen), but it should be of concern only if its typical summertime concentration is somewhat larger than that of winter. We suggest that summertime measurements of benzene be performed to clarify this question.

Acknowledgements

We are grateful for financial support for this project from the Uintah Impact Mitigation Special Service District and the Utah Science Technology and Research Initiative.

156 References

Anderson, D.G., MD; Ashley Regional Medical Center, Vernal, Utah; private communication, 2013.

Babin, Steven, et al. "Medicaid patient asthma-related acute care visits and their associations with ozone and particulates in Washington, DC, from 1994-2005." International Journal of Environmental Health Research, 2008: 209-221.

Beck, C., and K. Baxter, "Asthma in Utah: Burden Report 2012," Utah Department of Health, Asthma Program, health.utah.gov/asthma.

California Office of Environmental Health Hazard Assessment (OEHHA), http://www.oehha.ca.gov/ air/allrels.html.

Carlsen, K.H., I. Orstavik, J. Leegaard, H. Høeg, “Respiratory virus infections and aeroallergens in acute bronchial asthma,” Archives of Disease in Childhood, 59, 310-315 (1984).

Corrigan, C., “Mechanisms of Asthma,” Medicine, 40, 223-227 (2012).

Crighton, E.J., M.M. Mamdani, R.E.G. Upshur, “A population based time series analysis of asthma hospitalisations in Ontario, Canada: 1988 to 2000.” BMC Health Services Research, 1:7 (2001).

Environmental Protection Agency. EPA IRIS (Integrated Risk Information System) http://www.epa.gov/iris/.

Environmental Protection Agency, Region 8. HH: Risk Characterization. http://www2.epa.gov/ region8/hh-risk-characterization.

Environmental Protection Agency. A Preliminary Risk-Based Screening Approach for Air Toxics Monitoring Data Sets, Version 2, October 2010. U.S. Environmental Protection Agency, EPA- 904-B-06-001. www.epa.gov/region4/air/airtoxic/Screening_111610_KMEL.pdf.

Environmental Protection Agency. EPA AEGLs (Acute Exposure Guideline Levels) http://www.epa.gov/ opptintr/aegl/.

Environmental Protection Agency. Ground level ozone standards designations. Feb 1, 2013. http://www.epa.gov/ozonedesignations/ (accessed June 1, 2013).

Filleul, Laurent, et al. "The Relation Between Temperature, Ozone, and Mortality in Nine French Cities During the Heat Wave of 2003." Environmental Health Perspectives, 2006: 1344-1347.

Fleming, D.M., K.W. Cross, R. Sunderland, A.M. Ross, “Comparison of the seasonal patterns of asthma identified in general practitioner episodes, hospital admissions, and deaths,” Thorax, 55, 662- 665 (2000).

157 Geyh, Alison S., Jianping Xue, Haluk Ozkaynak, and John D. Spengler. "The Harvard Southern California Chronic Ozone Exposure Study: Assessing Ozone Exposure of Grade-School-Age Children in Two Southern California Communities." Envirnomental Health Perspectives, 2000: 265-270.

Grech, V., M. Balzan, R.P. Asciak, A. Buhagiar, “Seasonal Variations in Hospital Admissions for Asthma in Malta,” Journal of Asthma, 39, 263-268 (2002).

Han, Y-Y, Y-L Lee, YL Guo, “Indoor environmental risk factors and seasonal variation of childhood asthma,” Pediatric Allergy and Immunology, 20, 748-756 (2009).

Harju, T., T. Keistinen, T. Tuuponen, S.L. Kivela, “Seasonal variation in childhood asthma hospitalisations in Finland, 1972-1992,” European Journal of Pediatrics, 156, 436-439 (1997).

Johnston, N.W., S.L. Johnston, J.M. Duncan, J.M. Greene, T. Kebadze, P.K. Keith, M. Roy, S. Wasserman, M.R. Sears, “The September epidemic of asthma exacerbations in children: A search for etiology,” Journal of Allergy Clin Immunol., 115, 133-138 (2005).

Johnston, S.L., P.K. Pattenmore, G. Sanderson, S. Smith, M.J. Campbell, L.K. Josephs, A. Cunningham, B.S. Robinson, S.H. Myint, M.E. Ward, D.A. Tyrrell, S.T. Holgate, “The relationship between upper respiratory infections and hospital admissions for asthma: a time-trend analysis,” American Journal of Respiratory and Critical Care Medicine, 154, 654-60 (1996).

Kehrl, Howard R., David B. Peden, Beth Ball, Larry J. Folinsbee, and Donald Horstman. "Increased specific airway reactivity of persons with mild allergic asthma after 7.6 hours of exposure to 0.16 ppm ozone." Allergy and Clinical Immunology, 1999: 1198-1204.

Khot, A., R. Burn, N. Evans, C. Lenney, W. Lenney, “Seasonal variation and time trends in childhood asthma in England and Wales 1975-81,” British Medical Journal, 289, 235-237 (1984).

Kimbell-Dunn, M., N. Pearce, R. Beasley, “Seasonal variation in asthma hospitalizations and death rates in New Zealand,” Respirology, 5, 241-246 (2000).

Lenney, W., “The aetiology of childhood asthma,” Pediatrics and Child Health, 19, 257-260 (2009).

Lin, Shao, Xiu Liu, Linh H Le, and Syni-An Hwang. "Chronic Exposure to Ambient Ozone and Asthma Hosptial Admissions among Children." Environmental Health Perspectives, 2008: 1725-1730.

Liu, Lee-Jane Sally, Ralph Delfino, and Petros Koutrakis. "Ozone Exposure Assessment in a Southern California Community." Environmental Health Perspectives, 1997: 58-65.

Mansfield, M.L. and C.F. Hall, “Statistical Analysis of Winter Ozone Events,” accepted for publication in Air Quality, Atmosphere, and Health (2013).

Mar, Therese F, and Jane Q Koenig. "Relationship between visits to emergency departments for asthma and ozone exposure in greater Seattle, Washington." Annals of Allergy, asthma, & immunology, 2009: 474-479.

158 Monteil, M.A., S. Juman, R. Hassanally, K.P. Williams, L. Pierre, M. Rahaman, H. Singh, A. Trinidade, “Descriptive epidemiology of asthma in Trinidad, West Indies,” Journal of Asthma, 37, 677-785 (2000).

Pendergraft, T.B., R.H. Stanford, R. Beasley, D.A. Stempel, T. McLaughlin, “Seasonal Variation in Asthma- Related Hospital and Intensive Care Unit Admissions,” Journal of Asthma, 42, 265-271 (2005).

Pride, Kerry, J. Peel, B. Robinson, A. Busacker, Grandpre F. Yip, and T. Murphy. Associations of Short- Term Exposure to Ozone and Respiratory Outpatient Clinic Visits - Sublette County, Wyoming, 2008-2011. Public Health Report, Cheyenne: State of Wyoming Department of Health Public Health Division, 2013.

Silverman, Robert A, and Kazuhiko Ito. "Age-related association of fine particles and ozone with severe acute asthma in New York City." Allergy and Clinical Immunology, 2010: 367-373.

Sousa, S.I.V., M.C.M. Alvim-Ferraz, and F.G. Martins. "Health effects of ozone focusing on childhood asthma: What is now known - a review from an epidemiological point of view." Chemosphere, 2013: 2051-2058.

Strickland, M.J., L.A. Darrow, M. Klein, W.D. Flanders, J.A. Sarnat, L.A. Waller, S.E. Sarnat, J.A. Mulholland, P.E. Tolbert, “Short-term associations between Ambient Air Polllutants and Pediatric Asthma Emergency Department Visits,” American Journal of Respiratory and Critical Care Medicine, 182, 307-316 (2010).

Walther, E. Screening Health Risk Assessment: Sublette County, Wyoming, SR2011-01-03, January 2011, Sierra Research, Inc., Sacramento, CA. http://www.sublettewyo.com/DocumentCenter/ Home/View/438.

Weiss, K.B. “Seasonal Trends in US Asthma Hospitalizations and Mortality,” Journal of the American Medical Society, 263, 2323-2328 (1990).

World Health Organization. Air Quality Guidelines Global Update 2005. Report, Copenhagen: World Health Organization, 2006.

World Health Organization. Health risks of ozone from long-range transboundary air pollution. Copenhagen: World Health Organization, 2008.

159 APPENDIX: Seasonal Trends in Asthma and Correlations with the School Year

Figure 1 indicates that asthma cases in the Uintah Basin follow seasonal trends. Such trends are of interest because they help indicate the most active asthma triggers. Since ozone is not high in September, it is probably not a dominant asthma trigger in the Uintah Basin.

Similar seasonal trends in asthma cases appear throughout the world. Table 4 summarizes the results of 14 studies from around the world. The seasonal trends do not seem to be identical from one region to the next: some have narrow maxima, others broad. While the month in which the maximum occurs also varies, almost invariably the studies describe a maximum occurring sometime in autumn or winter. Of particular interest is the study by Johnston, et al., [2005] examining pediatric asthma cases in the province of Ontario, Canada. The maximum they describe is narrow, spanning only three weeks, and sharp (see their Figure 1). So sharp, in fact, that they use the phrase “September epidemic.”

It is also interesting that some of the studies mention that seasonal asthma trends correlate with the local scholastic calendar [Johnston et al., 1996; Grech et al., 2002; Johnston et al., 2005]. It is not difficult to see how seasonal trends could arise from asthma triggers such as pollen or spores whose release follows an annual pattern. More difficult to explain are these correlations with the school year. Viral infections of the upper respiratory tract have been proposed, under the assumption that the classroom brings children together and facilitates the spread of viruses [Johnston et al., 1996; Carlsen et al., 1984; Johnston et al., 2005; Grech et al., 2002]. Viral infections can also explain how adults are able to participate in any school-caused maximum: The children bring the virus home and infect the rest of the household. Other explanations are the lower use of control medications, and that allergens are found at higher levels in the school environment than at home [Johnston et al., 2005]. Of course, correlation is not causation, and at least in some cases it could just be that some other trigger happens to coincide with the opening of school.

In any case, a school-year correlation appears in the Vernal asthma data. Figure 6 shows the number of cumulative pediatric asthma cases relative to the first day of school. In other words, day 0 on the horizontal axis is the opening day of school, while the vertical axis displays the total number of cases seen since the first day of school, counting backwards on the days prior to day 0. (Over the three fall seasons covered by this study, day 0 in the Uintah School District has been the fourth Wednesday in August.) The average slope of the curve indicates the case rate. For 90 days prior to the first day of school, and for about 70 days between day +21 and day +90, the case rate is in the vicinity of 0.11 cases/day, essentially the same as the overall rate. However, between day 0 and day +20, i.e., for the first three weeks of school, the case rate is about 3 times higher, around 0.33 cases/day.

As mentioned, this could be a random occurrence. If it were random, it would mean that 21 cases happened to fall in one particular set of 63 days (a 3-week period during each of three different years) out of a total of 153 cases in 1277 days. The probability that 21 or more cases would fall in any one set of 63 days when the overall rate is 153 cases in 1277 days, assuming that all cases occur randomly is

153 " % 153! n 153−n −6 ∑ $ ' q (1− q) ≅ 22.87×10 n=21 # n!(153− n)!& where q = 63/1277. In other words, the odds that this 3-week maximum has occurred by chance are about 23 in one million.

160 Table 4. Summary of 14 studies that report seasonal asthma trends. Author(s) & Region Description of Seasonal Trends Ages Studied Johnston, et al. (1996) “Admissions for asthma are more frequent during periods of All Southampton, England school attendance.” Pendergraft, et al. Peak from October through February. ≥ 5 USA (all regions) Strickland, et al. Admissions November through April are 21% higher than May 5-17 Atlanta, GA (metro) through October. Han, et al. “a peak […] in winter and a nadir […] in summer.” 6-15 South Taiwan Johnston, et al. (2005) “Asthma exacerbation […] occur[s] globally after school 5-15 Ontario, Canada returns.” Strong spike in pediatric hospital treatments during September is termed “September epidemic.” Crighton, et al. “A small peak in hospitalisations in May, a trough in July, and All Ontario, Canada a large peak between September and November.” Harju, et al. Primary peak in October, secondary in May. < 15 Finland Weiss Peak from September to November. 5-34 USA Fleming, et al “Particularly high rates of admission occurred in September.” 0-14 England and Wales Grech, et al Peak in January and trough in August. Pediatric* Malta and adult Kimbell-Dunn, et al Peak in autumn (April-June), sharp decrease in summer 5-44 New Zealand (January). Monteil, et al. “Highest in the last quarter of the year; lowest in July and All Trinidad August.” Carlsen, et al. “More attacks [occur…] during spring and autumn.” 2-18 Norway Khot, et al. Highest in September, also high in June, July, and October. 4-14 England and Wales * “In school-aged children, the end of school in June was associated with a sharp (91%) drop in admissions, and restarting school in October was associated with an even sharper (165%) rise” [Grech et al, 2002].

161 Asthma cases, cumulative count, relative to school opening 15 30 45 Cumulative cases -30 -15 0 -90 -75 -60 -45 -30 -15 0 15 30 45 60 75 90 Relative day

Figure 6. Over a period of three weeks starting from the opening day of school, pediatric asthma cases in the Vernal area occur at about three times the usual rate. We show that the effect is not due to chance.

162