Meteorological factors affect the concentrations and distributions of pollutants during episodes of degraded air quality. Over the last 10 years, the upper Green River basin (UGRB) of Sublette County, Wyoming, has experienced numerous wintertime ozone episodes stimulated by emissions from oil and natural gas development operations, resulting in the region being determined to be in marginal nonattainment of the National Ambient Air Quality Standards. Examination of surface wind field patterns in the UGRB using observations from a network of surface monitoring stations for 2011 and 2012, with an emphasis on ozone-episode days, confirms that increased ozone concentrations are most frequently measured on days on which winds are light and variable. Dispersion and dilution of ozone and its precursor pollutants on these days is therefore inefficient, and so these episodes invariably occur within and close by the gas fields. On days that instead experience afternoon southeasterly winds, episodes can often be observed at locations on the northwestern perimeter of the basin remote from pollutant source regions. Simulations using the Weather Research and Forecasting Model, conducted for the case study of 15 February 2011, identify these southeasterlies as barrier winds caused by southwesterly flow at 700 hPa impinging on the Wind River Mountains that flank the UGRB to the northeast. Characterization of the barrier wind and the overall airflow patterns facilitates more accurate future forecasting of the time-dependent geographical distribution of increased concentrations of ozone and other pollutants in the region.
The upper Green River basin (UGRB) in Sublette County, Wyoming, is home to one of the fastest-growing natural gas developments in the United States (Carter and Seinfeld 2012; Field et al. 2015a). Since natural gas exploration projects commenced in the basin some 15–20 years ago, adding to the existing oil-field operations, development has accelerated with project authorizations through Records of Decision being awarded by the Bureau of Land Management for more than 8500 new wells in the Pinedale Anticline Project Area (PAPA), the Jonah Field II Project Area, and the Jonah Infill Drilling Project Area (BLM 1998, 2006, 2008). The relatively flat (average north–south gradient: 1:460) UGRB is an almost treeless, sparsely covered grassland and sagebrush landscape close to Wyoming’s western border with Idaho, at an altitude of approximately 2150 m (see Fig. 1). Mountain ranges enclose the basin on three sides, leaving the south side open to lower-elevation plains and the Interstate-80 highway corridor. It is bordered to the west by the Wyoming Range, which reaches an altitude of 3468 m, to the north by the Gros Ventre Range, ascending to 3570 m, and to the east and northeast by the Wind River Range, rising to 4209 m.
Discovery of high ozone (O3) mixing ratios in the UGRB in 2005 prompted the Wyoming State Department of Environmental Quality (WDEQ) Air Quality Division to initiate coordinated air-quality monitoring and modeling studies for the region. During the first observation period in the winter of 2005, surface O3 mixing ratios were observed to exceed the then-current 8-h National Ambient Air Quality Standard (NAAQS) of 84 ppbv at both the Boulder, Wyoming (BLDR), and Jonah, Wyoming, monitoring sites, where maximum values of 89 and 98 ppbv, respectively, were measured (WDEQ 2011a).
In 2008, the NAAQS for O3 was reduced from 84 ppbv to a value of 75 ppbv for the fourth-highest daily maximum 8-h average O3 mixing ratio, averaged over 3 yr. Subsequent monitoring throughout the county has confirmed that O3 “exceedances” have occurred on numerous occasions, with the result that the U.S. Environmental Protection Agency (EPA) designated Sublette County and adjacent parts of Lincoln and Sweetwater Counties as an area of “marginal nonattainment” for O3 pollution (EPA 2015). Observations of episodes of high O3 during winter in rural Wyoming and more recently in the Uintah basin of Utah (Martin et al. 2011; Edwards et al. 2013; Karion et al. 2013; Oltmans et al. 2014; Ahmadov et al. 2015) are of interest, first because of concerns about their associated potential detrimental health effects and second because of their apparently anomalous wintertime occurrence. In contrast, episodes elsewhere invariably take place in summer, when small solar zenith angles for much of the day lead to high ultraviolet radiative flux intensities that promote high photochemical O3 production rates (Fehsenfeld et al. 1983; Liu et al. 1987; Logan 1985, 1989; Nunnermacker et al. 1998; Vukovich et al. 1977). Summertime episodes of high O3 such as those found in the Los Angeles basin often occur when prevailing anticyclonic systems encourage the formation of subsidence inversions and stagnant wind conditions, which limit pollutant mixing and dispersion (Godish 2003).
Similar trapping of pollutant emissions, but by radiation inversions in wintertime, has been recognized for some time (Yu and Pielke 1986; Allwine et al. 1992), but its potential consequence—the photochemical formation of secondary pollutant species—was not widely realized until high O3 mixing ratios were observed in the UGRB and subsequently explained (Schnell et al. 2009). A combination of appropriate clear-sky meteorological conditions is required to promote the occurrence of local O3 episodes (Schnell et al. 2009; Stoeckenius and Ma 2010), including sufficient snow cover of depth ≥15 cm (Oltmans et al. 2014), which increases the albedo and hence the actinic flux, reduces O3 deposition losses at the surface (Bocquet et al. 2007; Edwards et al. 2014), and facilitates the simultaneous formation of a strong nocturnal radiation inversion and an underlying shallow boundary layer “cold pool” that persists throughout the following day with minimal depth modification as a result of snow-cover inhibiting surface heating. Also needed are high concentrations of volatile organic compounds (VOC) and oxides of nitrogen (NOx), the pollutant precursors needed for O3 formation, which in the UGRB are emitted primarily by operations at local oil and gas fields (Field et al. 2015a).
High concentrations of O3 precursors invariably occur when the cold pool is particularly stable and shallow and the surface wind speeds are low. Indeed, O3 episodes in the UGRB are only observed when these conditions are met (Oltmans et al. 2014). When the surface boundary layer deepens and/or wind speeds increase, not only do O3 concentrations decrease by dilution and dispersion as expected, but so do those of precursor species and their chemical derivatives, leading to a concomitant decrease in the rate of O3 formation. Ozone concentrations therefore quickly drop below the episode-level threshold. The relationship among boundary layer depth, wind speed, and observed O3 concentration is consequently more complex than for systems that contain unreactive species of interest, including relatively involatile particulates and trace gases (e.g., Pal et al. 2014). Modeling wintertime cold-pool systems that contain reactive pollutant species has therefore proved to be problematic and continues to pose a significant challenge because of the broad range of temporal and spatial scales of the influencing meteorological processes and the often complex chemistry (Baker et al. 2011). This situation is unfortunate because it is recognized that meteorological factors—in particular, those affecting surface wind fields and the formation, persistence, and stability of cold pools—are thought to play a more important role in determining year-to-year differences in the frequency and magnitude of episodes of high O3 than do those resulting from interannual minor variations in O3-precursor emission rates (Oltmans et al. 2014).
The evolution, persistence, and dispersion of trapped shallow cold pools, formed by radiative cooling and downslope flow at night, have been the subject of numerous investigations (e.g., Banta and Cotton 1981; Billings et al. 2006; Lareau et al. 2013; Neemann et al. 2015; Reeves and Stensrud 2009; Vrhovec and Hrabar 1996; Whiteman 1982; Whiteman et al. 2001; Zhong et al. 2001). Below the layer’s capping inversion, winds in the cold surface layer have long been recognized to often be relatively weak and disorganized (Egan and Schiermeier 1986; Whiteman et al. 1999) with parcel trajectories frequently reflecting the influence of terrain channeling on gravity-driven drainage flow. Indeed, interactions with the terrain are believed to affect local surface wind directions significantly, especially under stagnant conditions, and this process can lead to enhanced broadening of advecting pollutant plumes (Allwine et al. 1992; Hanna 1990). In addition, the trapping inversion is thought to essentially isolate and decouple surface winds from the larger-scale flow aloft (Vrhovec 1991).
The overall picture that emerges is that, in both valleys and wider relatively flat basins such as the UGRB, the combination of stagnant conditions and vertical stability in the surface layer ensures that emitted pollutants accumulate with minimal dispersion until such time as the cold surface layer dissipates as a result of the destruction of the capping inversion and/or the onset of larger-scale winds (Allwine et al. 1992; Lareau et al. 2013; Savov et al. 2002). As a consequence, higher concentrations of secondary pollutants such as O3 are usually found close to precursor source locations. On occasion, however, organized surface wind patterns become established below the capping inversion and transport reactant and product species downwind to distant locations. Surface-flow characterization is then needed not only to understand the observed spatial pollutant distributions and their relationship to emission sources (e.g., Field et al. 2015a,b) but also to facilitate the development of models used by regulators to formulate wintertime episode mitigation and control strategies. In the UGRB, meteorological data have until the last decade been too sparse to generate a comprehensive understanding of pollutant transport. The advent and subsequent expansion of monitoring sites since 2005 have now largely removed these limitations, making it possible to examine in detail the effect of meteorological factors on specific pollutant distributions.
In this study we focus primarily on a scenario in which pollutants are advected to remote locations that were not anticipated to be affected by ongoing natural gas development and processing operations. These impacts occur during periods of southeasterly winds, which studies with the Weather Research and Forecasting (WRF) Model that are presented here now show are due to the previously unrecognized development of a barrier wind along the Wind River Mountains. Other observed surface wind regimes and their effect on O3 distributions are also briefly described, to contrast with and provide a context for that associated with formation of southeasterly barrier winds. More detailed discussions of O3 and precursor VOC distributions in the UGRB can be found elsewhere (Field et al. 2015a,b).
Barrier winds are often found in the lower atmosphere adjacent to elongated mountain ranges whenever stable air is directed toward a topographic ridge and Froude numbers are less than about 1.0 (Pierrehumbert and Wyman 1985). Air impinging on the barrier slows as it is forced to ascend the mountain slopes, resulting in convergence. The consequent damming of stable air causes a surface pressure increase along the side of the mountain chain, creating a horizontal pressure gradient force (PGF) below the height of the barrier. This leads to a lower-level flow that is initially directed away from the mountains, before being transformed by Coriolis deflection into a quasigeostrophic wind blowing parallel to the mountain barrier if appropriate conditions persist for more than a few hours. Schwerdtfeger (1975) conducted the seminal work on barrier winds using observations in Antarctica. Others (Richwein 1980; Parish 1982) have since noted that barrier winds occur with some frequency in midlatitudes. They have subsequently been documented extensively (e.g., Bell and Bosart 1988; Colle and Mass 1995; McCauley and Sturman 1999).
Because the terrain forces such motion, barrier winds occur below the crest of the mountains and, depending on the magnitude of the PGF, can dominate flow in an area as broad as 100 km or more adjacent to the barrier (Schwerdtfeger 1975; Parish 1982). Thus, if formed, barrier winds along the windward (southwestern) flank of the Wind River Mountains would be southeasterlies that could extend across much of the width of the UGRB because at its widest the basin is less than 100 km wide and is even narrower to the north. The magnitude of barrier wind speeds depends on the depth of the airflow, the stability of the air, and the height of the barrier, with maximum wind speeds typically occurring just below midheight of the mountains (Parish 1982).
2. UGRB ozone and wind observations in the winter of 2011
Observations of both pollutant concentrations and meteorological parameters in the UGRB have received a particular emphasis during the “winter O3 formation season” (January–March), the period during which O3 episodes primarily occur (Oltmans et al. 2014). This study focuses on data from 2011, the year in which the highest recorded 8-h average O3 mixing ratio (126 ppbv) was measured, and in particular on the example case of 15 February 2011. Between 1 January and 31 March 2011, 8-h average O3 mixing ratios exceeded the NAAQS of 75 ppbv on 14 days; 1-h average O3 mixing ratios above 75 ppbv were measured on 29 days. The majority of the 2011 exceedances were observed at the WDEQ Air Quality Division BLDR site, although about one-quarter of 1- and 8-h exceedances occurred on the northwestern side of the basin, at the remote sites Wyoming Range (WYRA) and/or Daniel (DANI). The locations of these and other monitoring sites are shown in Fig. 1. Not surprising is that mixing ratios of O3 and its precursors display considerable day-to-day variability across the basin, reflecting temporal changes in the surface wind field and other factors. Thus on some days, and in particular when winds are light and variable, higher O3 mixing ratios, should they arise, are predominantly centered close to pollutant emission sources in the natural gas fields, whereas on other days, diverse distribution patterns are found, depending on the extent to which advection processes influence local species concentrations (Field et al. 2015b). Day-to-day and hourly variations of 1-h average O3 mixing ratios at BLDR, WYRA, DANI, Pinedale (PINE), and Jule Spring (JUSP) are illustrated by the 48-h time series plots shown in Fig. 2. The plots in the figure show that O3 mixing ratios exceeded 75 ppbv on both days, but at different locations. On 14 February 2011, high O3 mixing ratios were measured close to the PAPA and Jonah gas fields at BLDR and JUSP, respectively, whereas on the following day, they were instead observed only at the remote sites WYRA and DANI.
Data are available from 16 surface monitoring sites throughout the basin, composed of seven Federal Equivalent Method (FEM) sites, one Clean Air Status and Trend Network (CASTNET) site, the local airfields near Pinedale and Big Piney, remote automated weather stations (RAWS), and Snowpack Telemetry (SNOTEL) meteorological monitoring stations, nine of which only reported standard meteorological measurements (temperature, pressure, relative humidity, wind speed, wind direction, solar radiation, and precipitation; WDEQ 2015). In addition, a subset of meteorological and trace-gas measurements were carried out on a 75-m-high tower located just east of the Jonah field (WDEQ 2011b). Primary components of the 2011 data used for O3 distribution and wind field pattern analyses were hourly values of O3 mixing ratio, surface wind speed, wind direction, and temperature. Ozone-precursor VOC distributions could not be determined because continuous measurements of methane and VOCs were only carried out at three surface sites in 2011 and 2012, none of which are in the northwestern UGRB. Ten additional instrumented mesonet monitoring stations (Table 1; sites 23–32) were established throughout the basin during this study in 2012, a year with only three episodes of high O3, to obtain basic meteorological state data at a higher spatial resolution. A comprehensive list of surface monitoring sites for 2011 and 2012 is given in Table 1. Analysis of the 2-yr dataset allows surface wind patterns in the UGRB to be identified in the broader context of the prevailing synoptic conditions.
For the January–March “O3 season,” observed wind patterns in both 2011 and 2012 can be roughly classified into five categories, of which three in particular are associated with periods of high O3. Table 2 summarizes the number of days on which each wind pattern was observed. In 2011, 65 days experienced light and variable winds in the shallow cold surface layer during the morning hours. On 44 of these days, the wind field was modified later in the day because of erosion of the capping inversion as it succumbed to the influence of relatively weak northwesterly or westerly flow aloft. This wind pattern is designated pattern I and is associated with seven O3-episode days. On the remaining 21 days, 13 of which experienced high O3 mixing ratios, the inversion, and consequently the underlying light and variable winds, persisted throughout the daytime hours (pattern II). A third wind pattern (pattern III) was characterized by southeasterly winds flowing parallel to the Wind River Mountains that develop in late morning and early afternoon and spread southward throughout the day. During the winter of 2011, this pattern was observed on 12 days, all in February and March. Ozone episodes occurred on five of these days. On six other 2011 days (two with an O3 episode), morning southeasterly winds were observed that were later replaced by southwesterly winds (pattern IV). A fifth wind pattern (pattern V), which only occurred on three days in late March of 2012, was characterized by daytime strong (5–15 m s−1) southerly winds that developed from overnight southeasterlies at sunrise. Last, there were seven days in 2011, with two O3 episodes, and 16 days in 2012 that each exhibited a unique, different wind pattern that did not conform to those in patterns I–V. They could therefore not be categorized.
In summary, Table 2 shows that 20 of the 29 episodes in 2011 occurred on days on which either pattern I or II was prevalent. Highest O3 concentrations for these 20 episodes were found close to precursor pollutant sources, as were those for the two episodes on days on which the wind pattern could not be categorized. In contrast, for the five episodes on pattern-III days and the two episodes on pattern-IV days, increased O3 values were most apparent at remote sites in the northwest of the basin. As Table 2 shows, there were only three episodes in 2012.
Two mechanisms potentially contribute to enhanced O3 mixing ratios: first, when O3 forms at the site of interest from precursor species that have been emitted either locally or elsewhere and transported to that location and, second, when remotely produced O3 is advected to the site. It is therefore not unexpected that in the UGRB both the BLDR and University of Wyoming (UW) monitoring sites in particular experience locally produced O3, because they are situated close to areas of heavy gas-field development (Fig. 1b) (Field et al. 2015b). Indeed, observations show that O3 maxima at both sites are not uniformly associated with winds from any particular direction, except perhaps for a slight preponderance of southwesterly winds. Figure 3 shows wind roses for the BLDR and WYRA sites during January–March of 2011. For the entire period, resultant winds that are based on hourly observations at BLDR were 2.6 m s−1 from 316°. For periods when 1-h O3 concentrations exceeded 75 ppbv, 89% of which occurred between 13 February and 15 March, resultant winds were only 0.4 m s−1. In general, high O3 mixing ratios were most often associated with surface wind speeds of less than 4 m s−1, supporting the contention that low surface wind speeds are conducive to pollutant accumulation and periods of increased O3 (Yu and Pielke 1986; WDEQ 2011a).
Resultant winds at WYRA, located on the northwestern side of the basin in an area with very little industrial development, were 1.6 m s−1 from 291° for the January–March 2011 period, whereas during episodes when 1-h O3 concentrations exceeded 75 ppbv, they were 2.1 m s−1 from 142°. This resultant wind speed is slightly higher than that at any other site when high O3 is present. The WYRA wind roses in Fig. 3 illustrate these wind patterns and demonstrate that at least 70% of the O3 episodes at this site occur on days for which winds are from the southeast with speeds of 1–5 m s−1, implying that the transport time of air parcels to WYRA from the northernmost parts of the PAPA gas field, some 40 km to the southeast, is likely to be just 3–5 h and occasionally perhaps as little as 2.5 h. Analogous wind roses for DANI (not shown) display similar but slightly less uniform wind behavior patterns for both the entire 3-month period and O3-episode days.
The O3 mixing ratio time series plots in Fig. 2 are commensurate with these observations. Thus on 14 February 2011, when an O3 episode occurred at BLDR and JUSP (and UW) but not at WYRA, DANI, or other sites, pattern-II (persistent light and variable) winds were measured throughout the day. At BLDR, the 24-h surface wind speed average was 2.2 ± 0.9 m s−1 (1σ standard deviation), with a maximum hourly average value of 4.6 m s−1 [at 0200 mountain standard time (MST)]. Wind directions were variable, ranging predominantly from south through northwest to northeast. Winds (1.9 ± 0.2 m s−1) between 1500 and 1900 MST (2200 and 0200 UTC), when the O3 mixing ratio exceeded the episode threshold and reached its maximum (105 ppbv), conformed to the overall 24-h directional pattern, varying from south to north-northeast via west. At WYRA, the corresponding values for the 24-h-average winds and maximum wind speeds were 2.0 ± 0.8 m s−1 and 3.7 m s−1 (at 1100 MST; 1800 UTC), respectively, and during the BLDR O3 episode wind speeds averaged 2.1 ± 0.8 m s−1, with a direction that varied over a 155° range from east-northeast to southwest.
On 15 February 2011, when pattern-III winds prevailed, high O3 mixing ratios (>75 ppbv) were measured at WYRA from about 1300 until 1900 MST (0200 UTC), an hour after sunset, by which time O3 formation had long since ceased. At DANI, the episode occurred about an hour earlier. High episode-level O3 values were not measured elsewhere on this day. At WYRA, the 24-h surface wind speed average was 2.4 ± 1.1 m s−1. Hourly average wind speeds during the night increased from a low of 1.1 m s−1 (at 0200 MST; 0900 UTC) to a maximum of 4.3 m s−1 (at 0800 and 0900 MST; 1500 and 1600 UTC). During the day, they decreased to 3.1 m s−1 by 1300 MST (2000 UTC) and subsequently reached a low of 0.5 m s−1 by 1800 MST (0100 UTC). Speeds again increased somewhat after sunset, to 2.0 m s−1 by 2000 MST (0300 UTC), returning to 1.1 m s−1 by midnight. During both the night and daylight hours up to 1600 MST (2300 UTC), surface winds were essentially southeasterly, increasingly so up to midafternoon, with less variability (range: 113°–170°; average: 143° ± 16°). Winds subsequently became more southerly throughout the evening hours and were ultimately westerly by midnight. These wind characteristics are typical of those for O3-episode days at WYRA, as already noted.
Surface winds at BLDR on 15 February 2011 were generally north-northwesterly (342° ± 40°) at 2.2 ± 0.7 m s−1 until 1000 MST (1700 UTC). Within the next hour they rapidly transitioned to southeasterlies which strengthened considerably over the next two hours, achieving a wind speed of 5.6 m s−1 by 1300 MST (2000 UTC). Thereafter, up to 1800 MST (0100 UTC), during the period of the O3 episode at WYRA, wind speeds declined to 3.6 m s−1 as the winds steadily adopted a more east-southeasterly direction. By 2000 MST (0300 UTC) and thereafter, winds were easterly, trending east-northeasterly, with speeds fluctuating between 1.8 and 4.0 m s−1. It is clear that the late-morning and afternoon persistent southeasterlies facilitated the transport of pollutant species from source regions near BLDR toward WYRA. Pollutant advection pathways are obviously best appreciated by considering the overall wind field in the region of interest, not just wind characteristics at source and receptor sites. To this end, Fig. 4 shows hourly wind vector time series plots at all of the monitoring sites operating on 15 February 2011. Stations are listed in a roughly clockwise progression, starting in the northeast at the Pinedale CASTNET station (PND165), Half Moon Lake (HAFW4), and so on (see Fig. 1), followed by those on the east side of the basin, down to JUSP, and then back north along the western side of the basin [Big Piney Airport (KBPI), DANI, WYRA, and U.S. 191 between Pinedale and Bondurant (KRIM)]. The seven stations listed after KRIM are all located either outside the basin airshed [Shute Creek (KSHC), Moxa Arch (MOXA), and South Pass (SOPA)] or on its perimeter [Snider basin (SNIW4), Mount Coffin (MCOBT), Big Sandy Opening (BGSW4), and South Pass (KSOU)]. Potentially useful measurements at intermediate sites between BLDR and WYRA, such as at Mesa (MESA) and Trapper’s Point (TRAPP; Table 1), were unfortunately not obtained in 2011. The plots show that in the morning, after sunrise, most of the stations along the east side of the valley reported light and variable winds. By local noon (1900 UTC), however, the winds at all six of these stations, from HAFW4 south to UW, had become southeasterly and by 1500 (2200 UTC) were uniformly established all along the western side of the Wind River Mountains, extending to WYRA and KRIM. This directional transition occurred rapidly, within an hour, at most of these sites, starting earlier at the more northerly sites HAFW4, PND165, and PINE and later at those farther south. Also apparent, at the higher-elevation perimeter sites KSOU and SOPA, southeast of the basin, is a strong, persistent southwesterly flow, commensurate with the regional North American Mesoscale Forecast System (NAM) 700-hPa reanalysis wind field for this day, shown in Fig. 5. [Surface wind observations at 0300, 1100, 1300, and 1500 MST (1200, 1800, 2000, and 2200 UTC) are also depicted on the geographic maps shown in Fig. 7, which is described in more detail later in the paper.]
The time-varying winds on 15 February 2011 provide an archetypical example of wind pattern III. It is noteworthy that measured episode-level O3 mixing ratios at WYRA and DANI and the period of sustained southeasterlies occur simultaneously, strongly supporting the contention that O3 episodes at these remote sites result from transport of O3 and/or its precursors from sources within the basin that are to the southeast (WDEQ 2011c). The timing of O3 maxima at WYRA and DANI and periods of southeasterly winds show a similar correspondence on other episode days.
In addition, we note that the O3 mixing ratio plots in Fig. 2 show that at night, when the boundary layer capping inversion tends to become more firmly established, O3 mixing ratios are reduced to background levels of 30–50 ppbv at the five depicted sites. Similar values are measured at all other monitoring sites in the basin, suggesting that there is minimal carryover of excess O3 to the subsequent day. This behavior is in contrast to that during quiescent wintertime periods with prolonged stable inversions in the Uintah basin, when carryover of O3 from day to day is considered likely (Oltmans et al. 2014). In the UGRB, O3 concentrations are thought to decline during the evening and night as a result of ventilation, partial titration by nitric oxide (NO), and loss to the snow-covered surface. Day-to-day variations in O3 in the URGB are similarly not believed to be significantly influenced by long-range transport of O3 and/or its precursors from upwind urban areas such as Salt Lake City, Utah, ~290 km to the southwest, because of the presence of wind barriers such as the Wyoming Range in the regional topography.
3. Characterization of southeasterly winds in the UGRB: A WRF case study
The characteristics, nature, and underlying driving forces of the observed southeasterly winds that blow concurrently with at least 70% of the O3 episodes at the remote WYRA and DANI sites on the northwestern perimeter of the UGRB were investigated by carrying out finescale numerical simulations using the WRF Model. It has been previously assumed that the prevailing southeasterly flow that occasionally develops in the basin is caused by strong well-developed upslope flow. The reinforcing augmentative influence of an approaching cold front, at least for an event on 10 March 2011, has also been suggested (WDEQ 2011a). Upslope flow, however, invariably requires surface heating by solar radiation and would be weakened or even negated by surface cooling caused by the extensive snow cover reported for O3-episode days to have been greater on average in the winter of 2011 than in any of the preceding four years, especially in the northernmost areas of the basin (WDEQ 2011a,b). If thermally driven upslope flow were to have caused the southeasterlies, it would be expected that they would have occurred on many more days and in particular on those when quiescent conditions prevailed and solar flux intensities were sufficient to promote O3 production. As noted above in section 2, there were 13 such episode days, none of which developed afternoon southeasterlies. We therefore propose that the southeasterly flow regime is instead caused by blocking of stable air against the windward (western) side of the Wind River Mountains, resulting in the formation of a barrier wind (Schwerdtfeger 1975).
As discussed below, synoptic conditions over western Wyoming on 15 February 2011 show little evidence of any prefrontal activity or other potentially complicating factors that could obscure or otherwise obfuscate the identification of the basic processes at work that generate afternoon southeasterly flow in the UGRB. For this reason, this day was chosen to provide a case study for wind field simulation. Numerical modeling studies were conducted using WRF, version 3.4.1. A model domain was used that consisted of four superimposed increasingly smaller domains with 27-, 9-, 3-, and 1-km grid resolutions, centered over the middle of the UGRB. The innermost domain was composed of 181 × 181 grid points. Vertical resolution consisted of 70 sigma levels, with increasing resolution toward the surface. Key parameterizations used for the run are the following: Lin microphysics scheme, Goddard scheme for longwave radiation physics, Dudia shortwave radiation scheme, MM5 surface layer similarity with the unified Noah land surface model, and the Yonsei University boundary layer physics scheme. Here we report results of a 24-h simulation commencing at 0000 UTC 15 February 2011 (1700 MST 14 February 2011) with the analysis grids from the National Centers for Environmental Prediction (NCEP) 12-km NAM used to initialize the run. The choice of the 0000 UTC initiation time allowed a model spinup time of 12 h, which was more than sufficient to allow adjustments of mass and wind to the finescale terrain.
Figure 5 illustrates the large-scale conditions at 1800 UTC (1100 MST) for the 15 February case, just prior to the onset of coherent southeasterly flow along the western flank of the Wind River Mountains. A surface high pressure system (Fig. 5a) was centered just south of the basin in western Colorado. At 700 hPa (Fig. 5b), a pronounced ridge of high pressure is present over the entire western United States. Winds are from the southwest across Wyoming at about 10–15 m s−1. Since the UGRB is situated at an elevation of ~2150 m, the 700-hPa surface is about 750 m above the surface and is a relevant indicator of flow in the free atmosphere. Because the Wind River Mountains extend above 3500 m, with 12 distinct peaks rising to over 4000 m, the 700-hPa flow is well below crest level. The southwesterly flows shown in Fig. 5b are directed nearly orthogonal to the elevated terrain. Upper-level conditions (Figs. 5c,d) also show ridging, with the axis passing near the UGRB. The NCEP 12-km NAM pressure, temperature, and wind field data depicted in Fig. 5 are consistent with ambient fair-weather conditions with no organized cyclone development or well-developed cold front. Interpretation of the southeast wind regime in the eastern UGRB cannot therefore be plausibly attributed to prefrontal activity.
A classic signature of barrier winds, aside from the mountain-parallel low-level winds, is a surface pressure surplus on the windward side of the mountain range. Results from the WRF simulation for the innermost domain show that blocking of the stable air at low levels is apparent. Streamlines of the wind at about 180 m above ground level, together with the sea level perturbation pressure from WRF, are shown in Fig. 6. Perturbation pressure is computed by subtracting the mean sea level surface pressure across the domain from the individual gridpoint values of surface pressure. Notwithstanding acknowledged difficulties in estimating sea level pressure over complex mountainous terrain, a pattern emerges that is consistent with the wind observations illustrated in Fig. 4. As shown below, such surface pressure patterns are also supported by isobaric height measurements. At 1200 UTC (0500 MST; Fig. 6a), sea level pressure in the UGRB is disorganized and no appreciable gradient exists across the valley floor. Streamlines of the 180-m wind field are also disorganized, with no coherent wind direction pattern apparent in the UGRB. By 1800 UTC (1100 MST; Fig. 6b), surface sea level pressure perturbations show an elongated pressure surplus along the Wind River Mountains that is coincident with an organization of the flow streamlines in a barrier-parallel direction. Surface pressure perturbation patterns become more apparent during the afternoon (Figs. 6c,d) with blocking by the Wind River Mountains, resulting in a 1–2-hPa gradient across the basin with highest pressures along the barrier. In response to the organized horizontal pressure field, the flow expands throughout the afternoon hours and extends nearly across the entire 100-km-wide basin by 2200 UTC (1500 MST).
Simulated high perturbation pressure values adjacent to the Wind River Mountains conclusively demonstrate that the barrier-parallel flow cannot be the result of thermally induced effects. If anabatic circulations were present, lower pressure would become established against the elevated terrain. The presence of a pressure surplus along the mountain barrier concurrent with the development of mountain-parallel flow unequivocally demonstrates that barrier wind formation processes are at work.
Comparison of WRF results with available data is largely limited to the less-than-numerous surface-station observations. Upper-level data over the UGRB for almost all of February of 2011 are limited to those obtained by the WDEQ at four levels (3, 25, 50, and 73 m AGL) of the “tall tower” located (42.42°, −109.56°) just east of the Jonah field, north of JUSP. WDEQ minisodar and tethered balloon profiles, all near BLDR, were obtained, but in the winter of 2011 were limited to seven days only, all between 28 February and 12 March (WDEQ 2011b). They suggest that the inversion heights on measurement days were generally no higher than 50–150 m (Rappenglück et al. 2014; Oltmans et al. 2014). The closest National Weather Service rawinsonde ascents are conducted at Riverton, Wyoming, situated on the east side of the Wind River Mountains, and at Salt Lake City, west of the Wasatch Mountain Range, 290 km to the southwest. Figure 7 depicts the 10-m wind from the WRF simulations and the coincident surface wind observations at sites within the basin. Wind observations are typically made at 10 m above ground level at WDEQ sites and at 3 m at other sites. At 1200 UTC (0500 MST; Fig. 7a), observations and WRF results suggest a disorganized wind pattern with weak wind speeds in the central and northern UGRB. By 1800 UTC (1100 MST; Fig. 7b), surface wind measurements and WRF 10-m winds indicate that southeast flow is becoming established along the eastern part of the basin. WRF-simulated winds and actual wind observations show that the southeast wind regime becomes enhanced during the afternoon (Figs. 7c,d). WRF wind directions and wind speeds are in good agreement with those observed both at the surface sites and at the three upper levels at the tall tower.
Classification of the southeast flow as a barrier wind can be conclusively demonstrated by cross sections of the pressure field normal to the Wind River Mountains. The location of the cross section is indicated in Fig. 7a. The dynamics of the flow adjacent to the mountains is revealed by a detailed examination of the isobaric height field derived by the WRF simulation. In particular, analysis of perturbations in the isobaric height field illustrates the vertical structure of the horizontal pressure gradient force. Height perturbations are computed by subtracting the mean isobaric height from the isobaric height at individual grid points. Such perturbations of isobaric height are similar to D values (Bellamy 1945), which are departures of isobaric heights from the respective mean isobaric heights of a U.S. Standard Atmosphere. Thus variations of height along an isobaric surface are proportional to horizontal variations of pressure. Use of isobaric perturbation heights therefore allows inspection of the horizontal pressure gradient force at all levels when examining cross sections.
Figure 8 illustrates the perturbations of the isobaric height field. For barrier winds, highest perturbation heights should occur adjacent to the mountain barrier and decrease in an upstream direction (in this case to the southwest) from the ridge. Figure 8a shows that no significant stable air damming is present at 1200 UTC (0500 MST). This is consistent with the surface pressure field illustrated in Fig. 6a and with the disorganized wind field at that time.
Blocking of the southwesterly 700-hPa flow by the Wind River Mountains is apparent by 1800 UTC (1100 MST; Fig. 8b) with an isobaric height gradient of about 6 m across the valley floor over the approximately 90-km cross-sectional horizontal distance. These results support the surface pressure analyses shown in Fig. 6, for which each figure panel shows a pressure surplus against the Wind River Mountains, supporting the assertion that the mountain-parallel observed and simulated winds are indeed barrier winds. Isobaric height gradients increase slightly during afternoon hours (Figs. 8c,d) such that by 2200 UTC (1500 MST) a height difference of ~8 m can be seen over the cross section. This corresponds to a horizontal pressure gradient of ~0.8 hPa over 90 km, equivalent to a geostrophic wind of ~8.7 m s−1, sufficient to support the mountain-parallel wind regime.
Further evidence of the barrier nature of the flow regime is also seen in the WRF-simulated wind directions and wind speeds along the same cross section (Fig. 9). At 2000 UTC (1300 MST) the southeasterly wind regime is restricted to a surface layer that is about 500 m deep. Wind speeds decrease to the west, consistent with cold-air damming against the Wind River Mountains, which results in the most pronounced barrier wind characteristics being adjacent to the elevated terrain. The sloping isogons and isotachs are clearly tied to the mountains and are indicative of a barrier wind. Observed southeast flow that developed on 15 February 2011 is clearly the result of blocking of the stable air by the terrain.
It is noteworthy that development of the southeasterly winds associated with O3 episodes at DANI and WYRA always occurs just prior to the initial observations of increasing O3 levels at those sites and, furthermore, that they are subsequently sustained throughout the episode periods. Interesting is that, although O3 levels simultaneously increase somewhat at BLDR, PINE, and nearby adjacent sites, they are always well below the 75-ppbv episode threshold level, suggesting that the southeasterlies create an effective channel through which pollutants are advected directly from their sources in the oil and gas fields to areas in the northwest of the basin, bypassing communities to the northeast, a conclusion commensurate with the computed streamlines shown in Fig. 6 and the wind vectors shown in Fig. 7.
Photochemical O3 production and buildup in the shallow stable surface layer over the snow-covered UGRB in winter follows a typical daytime pattern, with highest concentrations usually being observed during the afternoon. The quiescent conditions that generally prevail ensure that the highest O3 mixing ratios are mainly found in the vicinity of precursor sources. Measurements in the winter of 2011 confirm that about one-quarter of high-O3 events are experienced late in the day, however, at remote UGRB monitoring sites north and west of precursor source areas. Observational meteorological data show that afternoon southeasterly winds arise on at least 70% of these episode days, suggesting that pollutants are advected directly to the receptor sites from the source areas. It has been proposed that such transport is associated with prefrontal circulations augmented by thermally driven upslope flow. A case-study analysis of synoptic flow patterns and consideration of the likelihood of radiative surface heating provide little support for this assertion. We contend instead that transport processes are the direct result of barrier winds that develop as stable air is blocked by the Wind River Mountains along the northeastern and eastern side of the basin. To be specific, when 700-hPa winds with a southwesterly component impinge normally on the mountains, flows at lower levels are blocked by the orography. Low-level mass and wind fields consequently adjust (Schwerdtfeger 1975), resulting in below-crest southeasterly flow parallel to the mountains. This barrier wind regime is not unusual in the UGRB, where it becomes established by late morning or early afternoon.
An exemplary case of barrier wind development occurred on 15 February 2011. Simulations of the atmospheric conditions over the UGRB on this day were carried out using the WRF Model with horizontal resolution down to 1 km and vertical resolution decreasing with height, using 70 sigma levels. NAM analysis grids were used to initialize the model run. The simulation shows no organized surface airflow pattern early in the day, but following blocking of the stable air at low levels and the consequent development of an elongated pressure surplus along the mountain chain that leads to a pressure gradient across the basin, the genesis in late morning and early afternoon of a coherent southeasterly flow parallel to the mountains. The computed temporal evolution of the direction and speed of these winds agrees well with that observed. Conclusive evidence that the southeast flow constitutes a barrier wind is demonstrated by calculated perturbations of the isobaric height field, particularly adjacent to the barrier, that appear concomitantly with the developing horizontal pressure gradient force.
In the 2011 winter O3 season, afternoon southeasterlies developed only during February and March, on ~20% of days, of which 42% experienced high O3 mixing ratios at WYRA and/or DANI. These sites were also impacted on two additional days, when southeasterlies were displaced by southwesterlies later in the day. We conclude that while barrier wind southeasterlies may not constitute the predominant wind pattern, they occur not infrequently, and should therefore not be overlooked when considering pollutant advection, especially as they appear to transport those pollutants to communities in the northwestern parts of the basin while simultaneously redirecting them away from northeastern locations such as the principal regional city, Pinedale. Other WYRA episode days with pattern-III winds were observed that resulted in the same pollutant dispersion pattern. However, there may be other yet to be discovered episodes when conditions are very similar but with sufficient minor differences to trigger the development of different O3 distributions, thereby possibly exposing Pinedale and other communities to high O3 concentrations. Identifying and understanding the likelihood of these alternative possibilities would be greatly facilitated by the development of more advanced prognostic meteorological models, which would not only enable the prediction of location-specific pollutant episodes at a higher resolution but also indirectly lead to the design of more effective air-pollution mitigation and control strategies. To this end, improvements in the boundary layer and land-surface schemes currently incorporated in the models would be particularly beneficial, as suggested previously (Baker et al. 2011), and would go some way toward achieving these objectives.
This work was supported in part by the Air Quality Division of the Wyoming Department of Environmental Quality. The authors thank Jeff Soltis and technical support staff of the UW Department of Atmospheric Science for help with the UW observational wind field study carried out in 2012.