Abstract. A 3-day period of strong, synoptic-scale stagnation, in which daytime boundary-layer winds were light and variable over the region, occurred in mid July of the 1995 Southern Oxidants Study centered on Nashville, Tennessee. Profiler winds showed light and variable flow throughout the mixed layer during the daytime, but at night in the layer between 100 and 2000 m AGL (which had been occupied by the daytime mixed layer) the winds accelerated to 5-10 m s^-1 as a result of nocturnal decoupling from surface friction, which produced inertial oscillations. In the present study, we investigate the effects of these wind changes on the buildup and transport of ozone (O₃). The primary measurement system used in this study was an airborne differential absorption lidar (DIAL) system that profiled O₃ in the boundary layer as the airplane flew along. Vertical cross sections showed that O₃ concentrations exceeding 120 ppb extended up to nearly 2 km AGL, but that the O₃ hardly moved at all horizontally, instead forming a dome of pollution over or near the city. The analysis concentrates on four meteorological processes that determine the 3-D spatial distribution of O₃ and the interaction between urban and rural pollution:

1. daytime buildup of O₃ over the urban area,
2. the extent of the drift of pollution cloud during the day as it formed, which controls peak O₃ concentrations,
3. nighttime transport by the accelerated winds above the surface, and
4. vertical mixing of pollution layers the next day.

1. Introduction

Over relatively simple topography, such as that found near Nashville, Tennessee, the overriding control on the transport and dispersion of atmospheric pollutants is the large-scale (usually synoptic) wind speed and direction. Occasionally, the larger-scale flow becomes weak, indicating that synoptic controls on flow and transport are weak (except for the large-scale subsidence accompanying these conditions). Two implications of such stagnation are (1) that pollution emissions are confined to a much smaller volume than when stronger flow is present, often leading to the highest pollutant concentrations in a season, and (2) that smaller-mesoscale influences, such as surface heating differences, can express themselves as locally generated flows or variations in mixing properties.

Because stagnation periods routinely produce the highest pollution concentrations, it is important to understand the key meteorological processes affecting the distribution and magnitude of the ozone buildup. We identify four processes that control the spread of the urban plume. The first three represent a sequence occurring through different stages of the diurnal heating cycle, and the fourth, advection, represents a daytime control that proved to be significant even under very light wind conditions. The processes are as follows: (1) daytime buildup of ozone over the city. It is important to determine how confined in the vertical and horizontal the pollution cloud is. (2) Acceleration of the flow at night above the nocturnal inversion. This flow carries the urban pollution cloud away from the city and redistributes it over the nearby countryside. Occupying the layer that had been the daytime mixed layer, the resulting nocturnal flows often show considerable shear with height. Thus they may draw apart the ozone cloud that formed over the city during the day by advecting urban pollutants in one direction at one level and in different directions at higher levels. (3) Vertical mixing the next day. The layers that were formed over rural areas by nighttime advection of urban ozone become mixed as a result of surface heating and thus become part of the rural ozone background. Where ozone was brought in at low levels, surface ozone concentrations drop as lower O₃ air mixes down. Where ozone was brought in as elevated layers, surface concentrations increase by fumigation. (4) Daytime advection of ozone away from the urban center. Even under light wind conditions during the day, when photochemical processes are active, horizontal drift of the pollution cloud can be an important control over peak concentrations of O₃ and other pollutants.

The issue of how far O₃ travels during a stagnation day after it forms is complicated by the fact that even under weak wind conditions, some flow, light and variable in direction, exists. An important question is whether such light flow, when integrated over many hours, has enough bias to transport an urban pollution cloud a significant distance, or to what extent the urban emission cloud remains localized. Fixed-point or fixed-profile wind measurements give a clue as to how much transport can be expected, but they do not convincingly resolve the problem of how far pollutants can drift in light winds, because of instrument-uncertainty, sampling, and representativeness issues. The evolving pollution plume must actually be tracked or mapped out as a function of time. During the Nashville SOS campaign airborne measurements were available to map the distribution of O₃, and an airborne differential-absorption lidar (DIAL) was especially useful in showing the extent of the O₃ plume and the magnitude of the concentrations over a broad area.

Given these four effects, we can envision a scenario on the fate of urban pollutants and their impact on the rural background as follows: The nocturnal dispersal of the urban O₃ column, which builds up during the day and is carried over the countryside at night, contributes to an increasing rural background. Urban emissions on subsequent days are then released into higher background concentrations over the city, accounting for the increases in urban O₃ concentrations through an episode. If true, the significance of this scenario is that it establishes a close interrelationship between urban and rural O₃, urban O₃ contributing to the background, and higher background concentrations being responsible for increases in urban O₃.

During the Nashville/Middle Tennessee campaign of the Southern Oxidants Study (SOS) in the summer of 1995, a major stagnation episode occurred from July 11 to 13. The period of high pollution actually began on July 7, when anticyclonic conditions moved into the Tennessee area [McNider et al., this issue] , and mixed-layer concentrations of carbon monoxide (CO) exceeded free atmospheric values above 3 km above sea level (asl) as measured by the NOAA P-3. A major difference between July 11 to 13 stagnation episode and the other days of the period, which ended on July 17, was that the daytime mixed-layer winds exhibited a discernable mean direction before and after the episode, whereas during the episode, daytime winds were light and variable throughout the levels occupied by the mixed layer. For example, on July 10 the mean flow in the mixed layer was northerly over north central Tennessee, and on July 14 the mixed-layer flow was westerly in the morning veering to northeasterly in the afternoon, but the flow was stronger than on the previous three days and had a discernible direction.

The focus of this study is on the horizontal and vertical distribution of the O₃ and meteorological processes producing that distribution. The photochemistry on July 11 and 12 is described in a campanion paper by Valente et al. [this issue].

2. Background: Instrumentation and Synoptic Meteorology.

The June-July 1995 campaign of the Southern Oxidants Study (SOS 1995) took place in north central Tennessee in the vicinity of Nashville (Plate 1). The elevation of Nashville at the airport is 183 m above sea level. In addition to an extensive array of surface-based chemistry and meteorological instrumentation [Meagher et al, this issue], five aircraft were deployed, as described by Hübler et al. [this issue], including NOAA's P-3 Orion, DOE's G-1 Gulfstream, NOAA's Twin Otter, and TVA's Bell 205 helicopter performing in situ measurements. A CASA 212 aircraft contained the ozone-profiling lidar flown by the Environmental Technology Laboratory (ETL) of the National Oceanic and Atmospheric Administration's Environmental Research Laboratories (NOAA/ERL).

Plate 1. Map of Nashville and middle Tennessee area showing locations of major power plants (pluses), cities (squares), and SOS sampling sites (triangles). Power plants include Cumberland (CL) and Gallatin (GA), and sampling locations include Dickson (DK), New Hendersonville (NH), Youth, Inc. (YI), and downtown Nashville. Color indicates land use patterns in 1 km by 1 km blocks based on AVHRR satellite data [Eidenschink, 1992]. Color bar is as follows: 1, crop/mixed farming; 2, evergreen needle-leaf trees; 3, inland water; 4, deciduous broadleaf trees; 5, mixed woodland; and 6, short grass.

2.1. Airborne Ozone DIAL System.

An important new capability available during the SOS 1995 campaign was the determination of the horizontal and vertical distributions of O₃ concentrations using the ETL O₃ differential-absorption lidar (DIAL) system. This instrument, flying in a CASA 212 aircraft, pointed downward to provide vertical cross sections of O₃ along the flight track. The system was developed by the Environmental Monitoring Systems Laboratory of the Environmental Protection Agency in Las Vegas, Nevada, and has previously been flown over Detroit and Houston [McElroy et al., 1993 a, b; Moosmüller et al., 1993 a, b, c).

The heart of the DIAL system is a krypton-fluoride excimer laser that emits pulses of ultraviolet (UV) radiation at 20 Hz. The return signals for the O₃ channels are averaged over 90-m range gates in the vertical and 8 s (160 shots) along the flight track. With a typical aircraft speed of 65 m s^-1 (~4 km min^-1) the 8-s time resolution translates to 520-m horizontal resolution. The system employs a separate wavelength not absorbed by O₃ to determine aerosol backscatter. The aerosol data, taken at 15-m vertical resolution, are used in the ozone-calculation algorithm to correct for aerosol attenuation.

The root-mean-square (rms) errors of ozone concentration caused by signal noise are typically about 4 ppbv in the upper part of the ozone profiles and about 10 - 15 ppbv in the lower part [Alvarez et al. 1999]. This corresponds to rms errors in aerosol backscatter of about 0.5 x 10^-6 and 1.0 x 10^-6 m^-1 sr^-1, respectively. The errors thus increase with range from the lidar, as a result of absorption of the lidar signals along the beam path.

The beam from the krypton fluoride laser is Raman shifted in hydrogen and deuterium to yield five transmitter wavelengths at 277, 292, 313, 319 and 360 nm. The first four are absorbed by ozone and thus are used in the differential absorption calculation, and the fifth is the wavelength not absorbed by O₃ used for the aerosol correction. In principle, any on-line/off-line wavelength pairing of the four shorter wavelengths (277 - 319 nm) could be used for the ozone retrieval because of their different absorption by ozone. The choice of on-line/off-line wavelength pair is usually dictated by the amount of ozone along the lidar beam path, the desired range resolution, and the sensitivity to aerosol effects.

The fairly large wavelength separation between on-line and off-line channels makes it necessary to correct for differential backscatter and extinction caused by air molecules and aerosol particles. The aerosol backscatter and extinction information needed for this correction is derived from the 360-nm channel following the method described by Fernald [1984]. The implementation of the aerosol correction of the DIAL ozone data generally follows the approach first used by Browell et al. [1985]. This requires making assumptions about the aerosol properties which if invalid may lead to errors in the corrected ozone concentrations due to overestimation or underestimation of the aerosol correction term. By comparing the results for several DIAL wavelength pairs of differing sensitivity to the aerosol correction, these remaining errors are minimized.

The ozone data shown here have been retrieved using the 277/292 nm wavelength pair, primarily because this wavelength combination is least susceptible to systematic errors in ozone concentration from erroneous aerosol corrections. Prior to processing, after-pulsing effects introduced by the photomultipliers used to amplify the backscattered light were removed from the lidar signals. After-pulsing caused a noticeable bias at the far-range end of the lidar signals (i.e., close to the ground), which if not corrected for would lead to systematic errors in ozone concentration. Lidar signals that were recorded during an aircraft turn have not been used for retrieval since the lidar beam path is slanted rather than vertical. Also, lidar signals contaminated by strong backscatter and attenuation by clouds have been excluded from the ozone and backscatter retrieval.

By using the 277/292 nm wavelength pair for the ozone retrieval, systematic errors in ozone concentration due to an erroneous aerosol correction are kept below a few ppbv in most cases.

Another source of error is the attenuation of the on-line wavelength (277 nm) near the ground in regions of high O₃ concentrations, such as in the urban plume over Nashville. The resulting low signal-to-noise ratios (SNRs) cause large errors in the calculated O₃ values. By increasing the range-gate size and averaging over longer time periods, these errors in high-O₃ regions can be reduced significantly. Despite these errors in high-O₃ regions, most of the figures showing lidar ozone were produced from the high-resolution (90 m by 520 m) ozone data in order to resolve small-scale features. However, the urban ozone profiles described in section 3.4 were calculated by first retrieving ozone profiles from the original lidar data averaged over 150 m in the vertical and 24 s in time (1560 m horizontally) and then averaging these profiles over the horizontal extent of the urban plume.

2.2. NOAA 915-MHz Wind Profiler

During SOS 1995, scientists from NOAA/ETL deployed a 915-MHz wind profiler at each of the following sites (all locations in Tennessee): Dickson, Dupont, Inc., New Hendersonville, and Youth, Inc. (see Plate 1). In addition, scientists from the Pacific Northwest Laboratories (PNL) operated a profiler at Land Between the Lakes (LBL), Kentucky, which shows up in the northwest corner of some of the flight-track figures. The profiler at Dupont, Inc. was operated in an experimental mode to obtain high temporal resolution profiles of vertical velocity only, and these data are not included in this study. The 915-MHz profiler was developed at the NOAA Aeronomy Laboratory in Boulder, Colorado [Ecklund et al., 1988]. Engineers designed this instrument to be ideally suited for field study (i.e., small, rugged, lightweight, and easily transportable).

Radar scattering is from refractive-index inhomogeneities associated with atmospheric turbulence, in the absence of precipitation or other particle scatterers. Winds are obtained from the mean Doppler shift, and turbulence information can be deduced from the intensity and width of the Doppler spectrum [White, 1997]. In a favorable (moist) environment, the maximum detectable range of the profiler is typically 2-3 km with 60-m vertical resolution and 3-5 km with 420-m vertical resolution, although higher detectable ranges can be achieved in deep convection. The Doppler velocity spectra are analyzed in real time to calculate winds using a simple consensus algorithm. For SOS 1995 we applied the ETL wind editor [Weber et al., 1993] to the archived spectral moments to remove outliers and provide a measure of quality of the analyzed winds. The editor calculates winds using individual radial velocity triplets (i.e., at the highest time resolution possible) to provide better estimates of winds in convection. It uses temporal and spatial continuity as quality controls in generating hourly averaged wind profiles.

The depth of the daytime convective boundary layer (CBL), or mixing depth, is a critical parameter for air pollution applications but is often difficult to parameterize accurately in numerical models. Wind profilers provide an estimate of the mixing depth using the intensity of the backscatter signal, which is proportional to the refractive index structure function parameter C_n² [e.g., Battan, 1973]. Both observations and theory show that the profile of C_n² in the CBL decreases in the surface layer, becomes nearly constant with height (depending on the influences of entrainment) in the mixed layer, and then increases sharply to a peak associated with the inversion capping the mixed layer. The profiler technique for deducing mixing depth relies on finding this peak in the measured profiles of SNR or C_n² (White, 1993).

The profiler technique of estimating mixing depth has several advantages over the standard approach (i.e., the radiosonde) and other experimental approaches. The profiler measures successive profiles that detail the evolution of the convective boundary layer. For some applications, details of the morning evolution may be more important than the maximum mixing depth that is achieved in the afternoon. The single estimate of mixing depth derived from the radiosonde may also not be representative of the average in space and time, especially if the radiosonde rises through a penetrating thermal, for example. Profilers equipped with RASS provide another means of estimating mixing depth. RASS coverage, however, is not sufficient to detect boundary layers deeper than about 1.5 km, which occur often in the southeastern U.S. during the summertime. A disadvantage of vertical profiling is the limited horizontal area covered by a single instrument. SOS 1995 addressed this problem by deploying a network of profilers and using the information provided by research aircraft to fill in the gaps.

To calculate horizontal wind trajectories, we used the hourly edited winds from the ETL and PNL wind profilers to produce an interpolated wind field averaged over vertical layers for each hour, and then we used these horizontal hourly wind fields in succession to advect the fluid parcels. We did not include the vertical winds w in the trajectory analysis because of known problems with w measured by the 915-Mhz profilers [White, 1997]. An updated inverse-distance-squared weighting was applied to the winds at each hourly location in the trajectories to account for the profiler spacing.

2.3. Other Surface Instrumentation

At each of the profiler sites, NOAA/ETL deployed a package of standard meteorological instruments to measure temperature, pressure, relative humidity, solar and net radiation, and wind speed and direction. At Dickson and New Hendersonville the instruments were mounted on 10-m towers. The temperature, pressure, humidity, and net radiation sensors were deployed ~2 m above ground, and the wind and solar radiation sensors were mounted at the top of the tower. At Youth, Inc. the instruments were installed on a 20-m scaffold. Raw data (1 min) were collected on data loggers and transmitted to NOAA/ETL via telephone lines.

The profilers and meteorological sensors deployed during SOS 1995 supplemented a wide array of chemistry measurements that are described in other papers in this issue. Of these, we will show only hourly averages of the surface ozone concentration measured by scientists from the Tennessee Valley Authority (TVA) at the Dickson, Downtown Nashville, and Youth, Inc. sites.

2.4. Synoptic Meteorological Conditions

A strong, extensive anticyclone dominated the large-scale meteorology of the southeastern United States during the mid-July period. At 50 kPa (Figure 1) the broad ridge was centered over the Great Plains on the morning of July 11, and it strengthened and moved eastward over Indiana by the morning of July 14, the day just after the period of strong stagnation. At the surface (Figure 2) this translated to a broad ridge over the Ohio River Valley on July 11 which strengthened, becoming the western branch of a Bermuda high by the end of the period. Horizontal pressure gradients, and consequently the daytime winds, were very weak during the episode. Figure 3 shows that the averaged winds during the July 11 to 13 period (bottom panel) were 2-3 m s^-1 weaker than the campaign-averaged winds (top panel) over at least the lowest 3 km of the atmosphere. Another consequence of the anticyclonic conditions was strong subsidence, which suppressed mixing depths during this period.

The eastward drift of the high-pressure systems through the episode produced subtle differences in the meteorology of the central Tennessee area through the period. These differences are highlighted in the descriptions of the individual days. The days were all weekdays (Tuesday through Thursday), so the automotive and manufacturing pollution source strengths should be similar on all three days. The major difference was an increase of surface temperature of 5-6 over the period, which undoubtedly produced a greater demand for air conditioning both in buildings and in vehicles, greater vehicle evaporative losses, and faster photochemical production rates.

3. Results

Profiler winds for the stagnation period (July 11-13) and the preceding day (July 10) are shown in Figure 4. Daytime atmospheric boundary layer (ABL) wind data for the day before the episode indicate persistent northerly flow, whereas for the three days of the episode, the indicated ABL flow was less than 2 m s^-1 (see Figure 3) and variable in direction according to the hourly averaged winds. On all four nights the winds at the levels that had been occupied by the daytime CBL (generally the layer from about 100 to 2000 m agl) accelerated, as the nocturnal inversion formed and decoupled the winds aloft from the effects of surface friction. These winds veered in time and with height through the night. McNider et al. [1988, 1993] and Gupta et al. [1997] argued that the inertial oscillation of winds in a nocturnal low-level jet are capable of advecting pollutants over a wide area, even when large-scale pressure gradients are relatively weak and daytime winds are light, as in the present case. The nighttime winds in Figure 4 thus confirm the acceleration of the flow at night, and data from the other profilers in the region (Youth, Inc. and Dickson) show a similar pattern.

	Figure 4a. Hourly wind profiles measured by the wind profiler at New Hendersonville for 10-11 July. The wind barb convention is 1 full barb = 5 m s-1. Data labeled as "suspect" by the wind editing algorithm are shown with small diamonds at the base of the wind flag. The hourly estimates of mixing depth deduced from profiles of radar backscatter intensity are denoted by the circles.
	Figure 4b. As in Figure 5a, except for July 12 to 13.

The surface O₃, temperature, and specific humidity for the downtown site, which was located on top of an office building (the Polk Building), and two sites outside the Nashville urban area are shown in Figure 5. Also shown are the mixed-layer heights for the three sites determined from the profiler backscatter. The O₃ data show the daytime buildup to >100 ppb downtown and the simultaneous temperature increases for all three days. The highest O₃ value of 140 ppb occurred in the sharp late-afternoon peak on July 12 at the downtown site.

Figure 5. Time series of mixing depth, surface temperature, and surface specific humidity for Dickson (DIK, solid), New Hendersonville (HEN, dotted-dashed), and Youth, Inc. (YOU, dashed) and surface ozone concentration for Dickson (solid), downtown Nashville (NAS, solid with crosses), and Youth, Inc. (dashed). Temperature and specific humidity were measured at 2 m above ground at Dickson and New Hendersonville and at 20 m above ground at Youth, Inc. The sampling heights for ozone were 10 m at Dickson, 110 m at downtown Nashville, and 20 m at Youth, Inc.

As a backround for the first day of the episode, we consider conditions on the previous day, July 10. NOAA P-3 flights to the north of Nashville indicated high concentrations of O₃ in northern Kentucky just south of the Ohio River. The high O₃ (100 ppb) was accompanied by high SO₂, aerosol, and NO_y, signifying a power-plant origin from the many plants along the Ohio River valley. The O₃ distribution and the northerly flow provide evidence that pollution from this large source advected into north central Tennessee during the day and overnight on July 10.

3.1. July 11, 1995

Horizontal maps of the ozone distribution from the airborne ozone lidar averaged for two vertical intervals (Plate 2) beginning at 1019 Central Daylight Time (CDT) show high O₃ at lower levels in several locations surrounding the urban area. This was especially true over Nashville itself and to the north of the city on the first leg, which passed generally from west to east ~35 km north of the city. The vertical cross section of this leg (Plate 3) shows that this high O₃ region extended up to 1000 m ASL but ended there.

Plate 2. Color-coded flight tracks for airborne O3 differential absorption lidar (DIAL) flights occurring from 1009 to 1235 Central Daylight Time (CDT) on July 11. Color coding represents O3 concentrations averaged over 800-m intervals in the vertical, for (a) 500-1200 m asl, (b) 1200-2000 m asl. Times of the flight legs are indicated at the end points in Plate 2a, and data begin at 1019 CDT at the northwesternmost vertex. Color coding for O3 concentrations is in ppbv. Atmospheric chemistry sampling site at Giles County is indicated as GC.

Plate 3. Vertical cross section of O3 concentration along northernmost leg in Plate 3 (the 1019- 1043 CDT leg). O3 concentrations are in ppbv. Each minute of flight time represents ~4 km along the flight track. The entire track was thus ~88 km long.

Later lidar cross sections (not shown) give a picture of the buildup and structure of the urban O₃ cloud reaching 1500 m asl (at 1215 CDT) and 1600 m asl (at 1315 CDT) over Nashville during the daytime. After noon, a well-developed column of O₃ in the urban plume was evident over and elongated to the southwest of the city (Plate 4). Although the mixed-layer winds were light and variable as indicated by the profilers (Figure 4), winds above the mixed layer were stronger and northeasterly. This flow seemed to influence the movement of the urban plume, as helicopter measurements reported by Valente et al. [this issue] show clearly that urban O₃ drifted more than 30 km to the southwest below 500 m agl through the afternoon, in agreement with the airborne DIAL data. Another consequence of the drift from northeast (NE) to southwest (SW) was the merger of the plume from the Gallatin power plant with the urban plume, as shown by Valente et al. [this issue] and St. John et al. [this issue].

Plate 4. Vertically averaged O3 concentrations (ppbv) from 800 to 1500 m asl for all flight legs (after 1156 CDT) on July 11. Final flight leg from the southwestern vertex toward Nashville was flown from 1259 to ~1821 CDT.

The light flow in the mixed layer means that no new O₃ would be advected in from the Ohio River power plants to the north. Daytime dry deposition at the surface would continue to efficiently remove O₃, as a result of the strong vertical mixing in the daytime boundary layer bringing O₃-rich air down to the surface. Because this was a period of stagnation, rural isoprene and other biogenic hydrocarbons were isolated from NO_x in the urban and power-plant plumes, and thus O₃ production was inefficient in these regions, as reflected in the large differences between urban and rural O₃ concentrations. The net result of these effects should be that the high [O₃] evident outside the urban plume at the beginning of this day would not be maintained and should tend to decline through the day.

Figure 6 shows trajectories from the overnight profiler winds, interpolated from the data at Dickson, LBL, New Hendersonville, and Youth, Inc., predicting that at low levels (400-800 m asl) the pollution cloud from Nashville should end up centered ~90 km to the west of the city by the next morning, considering that the centroid of the plume started out ~30 km or so SSW of the city at sunset.

Figure 6. Overnight forward trajectories calculated from the hourly horizontally interpolated winds, averaged over 400-m vertical intervals. Symbols indicate hourly positions beginning at 2000 CDT on July 11 and ending at 0800 CDT on July 12. Trajectory was calculated from an origin ~20 km southwest of Nashville at 36.0 N, 86.9 W.

3.2. July 12, 1995

On the second day of the episode the CASA 212 flew two missions with the O₃ DIAL system, an early-morning flight coordinated with two other aircraft (the Twin Otter and the G-1) and a noon/early-afternoon flight to document the urban ozone plume. The early-morning flight took off about sunrise, and repeated a triangular pattern to the southwest of Nashville. A region of relatively high O₃ appeared ~85 km to the west-southwest of the city just east of the Johnsonville power plant, mostly in a layer between 800 and 1200 m asl but also below this level (see Plate 5). This ozone, which reached peak values of >95 ppb, as indicated by the vertical profile at 0731 CDT (Figure 7), represents the urban O₃ that had formed at this level over Nashville the previous day and had advected to this location overnight, as predicted by the trajectory analysis and as indicated on Plate 5. This is not Johnsonville O₃, because the previous day's O₃ from the power plant would have been advected well to the west, and this time of day is too early for new O₃ from the plant to form.

	Plate 5. Flight-track O3 data averaged over 800-m vertical intervals for the first flight track (0708-0805 CDT) on July 12: (a) lower level from 400 to 1200 m asl, and (b) upper level from 1200 to 2000 asl. Trajectory in Plate 5a was calculated from mean winds in the 800 m layer from 400 to 1200 m asl.
	Figure 7. Profiles of O3 averaged horizontally over the region of high O3 in the western part of the flight track shown in Plate 5, for each of the four repetitions of the track during the morning flight on July 12.

In these early-morning data, high ozone appeared only at levels below ~1200 m. Data from flight legs flown later in the day, as the morning mixed layer grew past these levels, show greater [O₃] at higher levels between 1200 and 2000 m and reduced values at the lower levels, but the total O₃ in the vertical column remained nearly constant. This indicates the role of vertical mixing in bringing lower-O₃ air down from aloft and mixing the higher formerly urban O₃ upward. Figure 7 shows this vertical redistribution by midmorning. Data from this early flight thus confirm the nighttime transport and daytime vertical mixing processes. The final profile in Figure 7 shows evidence of the beginning of increases at the lowest levels, from the start of photochemical-production processes after 1000 CDT.

The second lidar flight began at noon and traced a butterfly pattern that produced an X centered on Nashville (Plate 6). The pattern was flown twice. Plate 7 shows the two northwest (NW) to southeast (SE) legs and Plate 8, the SW to NE legs. Plate 6, which shows the horizontal O₃ distribution for the average of the two patterns, reveals how highly localized the O₃ cloud is over Nashville. The vertical cross sections, especially Plate 7, strikingly show the extent to which the O₃ remained directly over Nashville at levels between 500 m and 2 km asl, essentially forming a dome of pollution over the city with values aloft exceeding 120 ppb. The dome was somewhat more elongated in the SW-NE direction (Plate 8). The small high-O₃ section to the SSE of Nashville seen in the horizontal plot (Plate 6) represents Murfreesboro, a smaller urban area that has a Saturn automobile plant. The vertical cross section (not shown) of this southern leg of the flight shows a localized dome of O₃ there that is a miniature version of the Nashville dome. Very little horizontal transport of urban pollutants is therefore indicated on this day except at lower levels, where a drift to the south is indicated by helicopter data taken during the day mostly at ~500 m asl [Valente et al., this issue].

Plate 6. Color-coded airborne O3 (ppbv) flight-track data averaged vertically from 800 to 1500 m asl for both passes over the pattern during the afternoon flight on July 12, i.e., averaged over all times from 1205 to 1600 CDT. Times are indicated at the vertices of the pattern.

Plate 7. Vertical cross sections of O3 concentration and aerosol backscatter for the NW-SE flight legs passing over Nashville for the afternoon flights on July 12. (a) O3 concentrations (ppbv) for the first pass (1220-1248 CDT, scale at bottom left), (b) aerosol backscatter (10-5 m-1 sr-1) for the first pass (scale at bottom right), and (c) O3 concentration for the second pass (1405-1433 CDT).

Plate 8. Ozone cross sections as in Plate 7 except for SW-NE flight legs. (a) O3 concentrations for first pass (1317- 1345 CDT), and (b) O3 concentrations for second pass (1501-1529 CDT).

Trajectories of overnight transport from the Nashville O₃ dome at different vertical levels (Fig. 8) indicate urban O₃ should be advected to the north by the next morning, with transport to the north and NW favored at levels below 1 km and transport to the NE at higher levels.

Figure 8. Overnight forward trajectories calculated from the hourly horizontally interpolated winds averaged over 400-m vertical intervals. Symbols indicate hourly positions beginning at 2000 CDT on July 12 and ending at 0800 CDT on July 13. Trajectory was calculated from an origin centered on Nashville, at 36.2 N, 86.8 W.

Another striking feature of the vertical cross sections, illustrated in Plate 7, is the variation in mixed-layer heights over the region. The ABL in most directions around Nashville is ~1700 m deep on the first pass and 1900 m or more on the second pass. However, to the northwest of Nashville the mixing height is strongly suppressed, reaching only 1100 m on the first pass and barely 1500 m on the second. Moreover, mixing heights in the O₃ dome over Nashville itself tended to be greater than over the surrounding areas, as found previously by Trainer et al. [1995] over Birmingham, Alabama. These relationships are clear from the O₃ cross sections, where high O₃ values mark the mixed layer, and they are also clear from the aerosol cross sections (Plate 7b). Mixing-height variability will be further discussed later in section 3.4. It is also interesting that the dropouts (white vertical bars where data are missing), which were due mostly to clouds just below the aircraft, were most plentiful over the city, indicating that cloud activity tended to be stronger over the urban area. This tendency was also evident during the flights, because it was occasionally necessary to change flight level over the urban area to avoid clouds.

Trajectories of overnight transport from the Nashville O₃ dome at different vertical levels (Figure 8) indicate that urban O₃ should be advected to the north by the next morning, with transport to the north and NW favored at levels below 1 km and transport to the NE at higher levels.

3.3. July 13, 1995

The earlier legs of the July 13 O₃ lidar flights are shown in Plate 9, and they confirm a layer of high O₃ to the ENE of the city at higher levels (1200-2000 m) but not at lower levels, in agreement with the trajectory predictions. Moreover, NOAA P-3 flights found considerable urban O₃ during climbout to the north and west in the late morning. The northernmost airborne lidar leg, which was first flown at 1100 CDT and clearly shows the elevated high-O₃ layer to the east of Gallatin (GA), was repeated at ~1330 CDT. The second pass also showed high O₃ in this region (Plate 10), but higher values could be seen lower in the boundary layer as well, as a result of convective mixing. The effects of vertical mixing on the later profile are evident in the vertical profiles of the horizontally averaged ozone concentrations in this area (Figure 9). The mixing produced especially dramatic increases in [O₃] at the level of the minimum in the earlier profile at ~800 m asl. Increases throughout the profile and especially the large increases evident near the surface indicate photochemical production.

	Plate 9. Vertically averaged O3 concentrations for the early flight legs (before 1236 CDT) on July 13. (a) Lower levels from 400 to 1200 m asl, and (b) upper levels from 1200 to 2000 m asl. Times (CDT) of the legs are indicated at the corners of the pattern in Plate 9a.
	Plate 10. As Plate 9 except for later flight legs (after 1236 CDT).
	Figure 9. Vertical profiles of O3 concentration averaged in the horizontal over 72 s (~5 km) of the flight track to the northeast of Nashville for the first pass at ~1055 CDT (solid line) and second pass at ~1323 CDT (dotted line) on July 13.

A south to north leg flown over Nashville between 1230 and 1300 CDT shows the extent of the urban O₃ plume in that direction (see Plate 10). The plume dimension was ~30 km, somewhat longer than on the previous day, even though the indicated mixed-layer flow was still light. Overall, the O₃ DIAL flights indicated that considerable O₃ could be found mostly in patchy regions around Nashville.

3.4. Intraregional Variability

Profiles from the airborne O₃ DIAL system averaged over the urban pollution area between 1100 and 1520 CDT on each of the three episode days (Figure 10a) reveal an increase in ozone in the upper part of the profiles above 1.3 km asl from July 11 to 12 and 13. The urban pollution area was determined subjectively from both the vertical cross sections and the horizontal O₃ maps and differed from day to day. It was identified as the deep regions of high O₃ over Nashville and the contiguous areas of high O₃ downwind. The marked increases in the profiles above 1300 m after July 11 were due in part to the strong suppression of the urban mixing height on that day, which was relaxed somewhat on the following days. Because of the deeper penetration of the O₃ plume on July 12 and 13, the total-column O₃ was higher in the urban area on those days. The highest O₃ concentrations at all levels occurred on July 12, the day with the least horizontal transport and ventilation. Surface O₃ measurements at the downtown Nashville site (see also Figure 5) showed a sharp episode maximum (140 ppb) on July 12, and surface measurements also showed that this day had the largest difference between the downtown and the other sites (e.g., see Figure 5), consistent with the lack of horizontal drift or exchange seen in the O₃ DIAL cross sections.

Figure 10. Horizontally averaged O3 profiles (ppbv) for July 11 to 13 midday flights (after 1100 CDT), (a) averaged over all those individual profiles identified as being within the urban plume on each day and (b) averaged over all the individual profiles outside the urban plume (i.e., all profiles except those used in calculating the profiles in Figure 10a), representing mean rural O3 profiles; note difference in horizontal scale from Figure 10a. The regions identified as "urban" were determined subjectively and differed from day to day.

The evolution of O₃ concentrations averaged over the rural areas (i.e., all the data outside the urban plume, Figure 10) showed differences among the three days of the episode below 1200 m asl. O₃ concentrations decreased from July 11 to July 12 and then increased again on July 13, consistent with the surface measurements at Youth, Inc., and Dickson (symbols in bottom part of figure). Average profiles from the airborne O₃ lidar calculated for different sectors around (and outside of) the city indicated that these trends were consistent in all directions, and the magnitude of the difference was ~15 ppb. In the upper part of the mixed layer between 1200 and 2000 m asl the decrease in O₃ from July 11 to 12 was small, decreasing with height from a ~10 ppb difference at 1200 m. The increase from July 12 to 13, however was more significant, exceeding 15 ppb through the layer and reaching ~20 ppb just above 1500 m asl.

In addition to the high concentrations of pollutants that accumulate, stagnant atmospheric conditions also allow mesoscale differences in heating to express themselves. The dramatic differences in mixed-layer inversion heights z_i on July 12 appeared in both the O₃ data (Plates 7a, 7c) and in the aerosol backscatter data from the airborne lidar (e.g., Plate 7b). They are also very evident in the mixed-layer height from the profilers, based on analysis of radar backscatter (as described in section 2.2). The Dickson profiler, located to the WNW of Nashville, consistently showed afternoon mixing heights 400-500 m or more lower than the other profilers (Figure 5), and the tendency existed on light wind days throughout the campaign for the mixing height at Dickson to be lower than at the other sites. It is similar in magnitude to the difference Trainer et al. [1995] found between the urban values over Birmingham, Alabama, and the regional values outside the urban plume. (This contrast across the study region is apart from the fact that mixing heights at all sites were lower during the stagnation episode than during the rest of the campaign, reflecting the strong synoptic-scale subsidence.) The variability in mixing height within the region most likely resulted from differences in land use. Regions to the west through north of Nashville have deciduous forest (this area also produces more isoprene), suggesting higher evapotranspiration (during extended dry periods), whereas the regions in the other directions are more farmland or mixed. Daytime 2.5-m temperatures at Dickson (Figure 5) were cooler than downtown by 1-2 C and were about the same as the temperatures at Youth, which were measured at 20 m above the ground. If the Youth temperatures were reduced to 2.5 m as measured at Dickson, they would be at least 1-2 C warmer in the superadiabatic surface layer. Thus the region around Dickson did have a cooler surface.

The reason for the z_i differences can be stated as follows. Stronger flow over a region with patchy surface characteristics moves air columns in the convective boundary layer (CBL) over many surface types and produces a more uniform mixing height in the horizontal. Weak flow, however, allows air columns to dwell over regions of one surface type and thus also allows surface heating differences to express themselves as intraregional variations in mixing height. Very weak, stagnant flow, as in the present case, amplifies these differences even more.

Another feature where small-scale heating differences could be expressed under light flow conditions is the urban heat island, with inflow at low levels and outflow aloft (in the upper ABL). The low-level inflow could be a mechanism for bringing rural air, with its biogenic hydrocarbon load, into the city to interact with urban emissions. We performed Lagrangian particle calculations for the weakest-flow day, July 12, using profiler data and the algorithm of McNider et al. [1988] to see whether particles released at low levels at the three wind profiler sites, which were outside the city, would move toward the city. Trajectories from New Hendersonville moved southward during the afternoon in the general direction of the city, and those from Youth, Inc., moved westward and a little south, again in the general direction of the city. Dickson was too far away to reflect urban heating influences. Noting that the low-level helicopter flights reported by Valente et al. [this issue] showed low-level drift to the south during the afternoon, it is not clear whether the indicated flow from the profilers at New Hendersonville and Youth are part of this drift, or whether they actually do demonstrate a heat island effect. Higher mixing depths over the city are by themselves evidence of warmer air there, and an indicator that kinematic effects (heat island circulation) should also be present. A profiler just outside the urban area to the south or southwest might have answered this question, but none was sited in this region. It is also important to consider that under such light flow conditions, the averaged profiler wind direction can be unrepresentative of that which is experienced by a pollutant plume or cloud. What is really needed to establish the existence of such an inflow is evidence in the form of a measurement of rural biogenic emissions, emission products, or other tracer of rural air mass origin entering the outskirts of the urban area. Unfortunately, these measurements were taken outside the city during the 1995 campaign. In conclusion, an urban heat island effect could be an important aspect of light-wind meteorology near a city with important consequences for the kind of chemistry occurring in the urban area, and evidence from this campaign hints that it may be present. We are continuing to investigate this effect.

4. Conclusions

A strong, regional high-pressure system, which built over the southeastern United States during mid-July 1995, produced a period of synoptic stagnation lasting three days over Nashville starting July 11. Strong daytime insolation during this episode drove active photochemistry in pollution emitted from urban and other sources, and urban O₃ concentrations were high (>100 ppb) from day to day through the episode. Remote sensing measurements from ETL's airborne O₃-profiling lidar showed that the urban pollution hardly moved from its urban source, remaining confined laterally to the atmosphere over or very close to Nashville.

We focused our discussion on four critical meteorological processes that affect the distribution of O₃ during light-wind periods.

Daytime buildup. Airborne lidar measurements gave a unique view of this effect, showing that high O₃ concentrations existed not only at the surface but also reached levels high above the city to the top of the daytime urban mixed layer, which extended more than 2 km above the surface. These high concentrations remained as an intact plume that stayed close to the urban area.

Nighttime transport. Wind profiler data clearly showed acceleration of the flow in the layer between ~100 and 2000 m agl from light and variable during the day to 5-10 m s^-1 at night. Airborne lidar and in situ measurements found urban O₃ over rural areas near places predicted by trajectory models. The lidar showed these occurrences in layers.

Daytime vertical mixing. Lidar cross sections and profiles showed these O₃ layers mixing out during the day and becoming part of the rural mixed layer the next day.

Daytime drift of O₃ pollution cloud. Helicopter and airborne lidar measurements showed that the dimensions of the O₃ plume varied from day to day even in the light winds of the stagnation episode. The highest surface O₃ concentrations occurred on the day with the least "net ventilation," suggesting that the lateral dimensions achieved by the urban-emission cloud were an important control on peak O₃ concentrations, even more than the mixing height.

We then posed a scenario on the degree of interaction between urban and rural pollution, in which day-to-day increases in rural O₃ provided an increasingly higher base into which urban emissions were released. This would result in higher urban O₃, which then, at least in part, fed the further increases in rural O₃. Data presented in this study provide evidence for the portions of the scenario relating to the transport of urban O₃ into rural areas and the incorporation of O₃ into the rural mixed layer, as just described in the previous paragraphs. An increase in urban O₃ aloft in the urban cloud was also noted. The feedback in which we see convincing increases of 15-20 ppb in the rural background is also supported by this data set in the mixed layer from July 12 to 13. The drop in background O₃ from July 11 to 12 occurred because the area was cut off from the strong power-plant source activity to the north on July 12, and in fact, rural regions were also isolated from emissions in the local area from urban and other sources. The increases noted in the background in all directions from Nashville from July 12 to 13 were real and reflected the emissions of Nashville and the other local sources in north central Tennessee being distributed over the countryside at night.

Other consequences of large-scale stagnation that we explored related to the expression of mesoscale heating differences. Both airborne lidar and profiler data revealed strong variations in the mixed-layer heights within the region around Nashville, as the mixing depth to the northwest was strongly suppressed compared with depths in the other quadrants around the city. The mixed layer over the Nashville urban area itself was deeper than that in the surrounding rural areas. Strong stagnation would be the best opportunity to find evidence for an urban heat island circulation, and we found some evidence for it. Perhaps its effects were too subtle to detect clearly with the present data set, and we suggested that a different placement of profilers or a slightly different strategy for chemistry sampling, including samplers for tracers of rural air in the urban areas, might resolve this issue in future experiments. We are continuing to analyze the present data set for more convincing signatures of this effect.

Finally, we emphasize the key role of nighttime transport in the cycle of pollution dispersion. It has often been tempting to conclude that transport over relatively flat terrain is unimportant at night because the nocturnal inversion isolates the flow above from the surface, and the flow near the surface is generally light. In addition to photochemical production, vertical transport shuts down, especially near the surface, over simple topography and perhaps this leaves the impression that transport is quiescent during this period. This study has shown that nighttime processes have a critical role in the distribution of urban pollutants, clearing out the air over the city and replacing it with background air, and distributing urban pollutants over rural areas where they do become part of the background, as hypothesized by McNider et al. [1988, 1993]. Presumably, a similar cycle occurs for pollutants from power plants and other large sources. These flows, which occur above the nocturnal inversion in the layer that had been occupied by the daytime mixed layer (generally 100 m to 2 km agl), are important only because high concentrations of urban pollutants were carried upward to these levels during the day. Stagnation episodes thus seem to be marked by much less horizontal transport during the day than found under more normal flow conditions but much more transport at night than previously appreciated.

Acknowledgments. Special thanks go to Lisa Olivier, who created the land use map of the Nashville region shown in Plate 1, and Jerome Fast for his careful review of the manuscript and many helpful suggestions. The authors acknowledge Las Vegas EPA Environmental Monitoring Systems Laboratory, including Dr. James McElroy and the lidar development team, who developed the airborne O₃ DIAL system. Without this system, most of the conclusions in this study would not have been possible. Funding for this research was provided through the NOAA Health of the Atmosphere and the EPA Southern Oxidants Study programs. Kenneth J. Olszyna of TVA kindly supplied the surface ozone data. We also appreciate the contributions of Wynn Eberhard and Chuck Frush in the field acquisition of the data and Karsten Baumann and David H. Levinson for helpful suggestions on the manuscript.

References

Alvarez, R.J., C.J. Senff, R.M. Hardesty, D.D. Parish, W.T. Luke, T.B. Watson, P.H. Daum, and N. Gillani, Comparisons of airborne lidar measurements of ozone with airborne in situ measurements during the 1995 Southern Oxidants Study, J. Geophys. Res., to be published, 1999.

Battan, L.J., Radar Observation of the Atmosphere., pp. 323, Univ. of Chicago Press, Chicago, Ill. 1973.

Browell, E.V., S. Ismail, and S.T. Shipley, Ultraviolet DIAL measurements of O₃ profiles in regions of spatially inhomogeneous aerosols, Appl. Opt., 24, 2827-2836, 1985.

Ecklund, W.L., D.A. Carter, and B.B. Balsley, A UHF wind profiler for the boundary layer: Brief description and initial results, J. Atmos. Oceanic Technol., 5, 432-441, 1988.

Eidenschink, J.C., The 1990 coterminous U.S. AVHRR data set, Photogramm Eng. Remote Sens., 58, 809-813, 1992.

Fernald, F.G., Analysis of atmospheric lidar observations: Some comments, Appl. Opt., 23, 652-653.

Gupta, S., R.T. McNider, M. Trainer, R.J. Zamora, K.R. Knupp, and M.P. Singh, Nocturnal wind structure and plume growth rates due to inertial oscillations, J. Appl. Meteorol., 36, 1050-1063, 1997.

Hübler, G., et al., An overview of the airborne activities during the SOS 1995 Nashville/Middle Tennessee Ozone Study, J. Geophys. Res. (this issue).

McElroy, J.L., H. Moosmuller, R.M. Jorgensen, C.M. Edmonds, R.J. Alvarez II, and D.H. Bundy, Airborne UV-DIAL measurements of ozone distributions in southeastern Michigan, in Proceedings of the 86th AWMA Annual Meeting, Air and Waste Manage. Assoc., 1993a

McElroy, J.L., C.M. Edmonds, D.H. Bundy, H. Moosmuller, R.M. Jorgensen, and R.J. Alvarez II, Airborne UV-DIAL urban plume measurements, In Tunable Diode Laser Spectroscopy, Lidar, and DIAL Techniques for Environmental and Industrial Measurements, edited by A. Fried, D.K. Killinger, H.I. Schiff, Proc. SPIE, 2112, 251-260, 1993b.

McNider, R.T., M.D. Moran, and R.A. Pielke, Influence of diurnal and inertial boundary- layer oscillations on long-range dispersion, Atmos. Environ. 22, 2445-2462.

McNider, R.T., M.P. Singh, and J.T. Lin, Diurnal wind-structure variations and dispersion of pollutants in the boundary layer. Atmos. Environment, 27A, 2199-2214, 1993.

McNider, R.T., W.B. Norris, A.J. Song, R.L. Clymer, S. Gupta, R.M. Banta, R.J. Zamora, A.B. White, and M. Trainer. Meteorological conditions during the 1995 SOS Nashville/Middle Tennessee Field Intensive, J. Geophys. Res, (this issue)

Meagher, J.F., E.B. Cowling, F.C. Fehsenfeld, and W.J. Parkhurst, Ozone Formation and Transport in the southeastern United States: An Overview of SOS Nashville/Middle Tennessee Ozone Study., J. Geophys. Res. (this issue).

Moosmüller, H., R.J. Alvarez II, R.M. Jorgensen, C.M. Edmonds, D.H. Bundy, and J.L. McElroy, An airborne UV-DIAL system for ozone measurements: Field use and verification, in Proceedings of the AWMA/SPIE International Symposium on Optical Sensing for Environmental Monitoring, SP-89, 413-422, 1993a.

Moosmüller, H., R.J. Alvarez II, R.M. Jorgensen, C.M. Edmonds, D.H. Bundy, D. Diebel, M.P. Bristow, and J.L. McElroy, An airborne lidar system for tropospheric ozone measurement, in Proceedings of the AWMA 86th Annual Meeting, Air and Waste Manage. Assoc., 1993b.

Moosmüller, H., R.J. Alvarez II, C.M. Edmonds, R.M. Jorgensen, D.H. Bundy, and J.L. McElroy, Airborne ozone measurements with the US EPA UV-DIAL, in Optical Remote Sensing of the Atmosphere Technical Digest, (Opt. Soc. of Am.), 5, 176-179, 1993.

National Oceanic and Atmospheric Administration (NOAA), Climate Analysis Center, Daily weather maps, Weekly series, July 10-16, 1995, Washington, D.C., NOAA, pp. 8, 1995.

Senff, C.J., R.M. Hardesy, R.J. Alvarez, and S.D. Mayor, Airborne lidar characterization of power plant plumes during the 1995 Southern Oxidants Study, J. Geophys. Res, accepted for publication, 1999.

St. John, J.C., W.L. Chameides, and R. Saylor, Role of anthropogenic NO_x and VOC as ozone precursors: A case study from the SOS Nashville/Middle Tennessesee Ozone Study, J. Geophys. Res. (this issue).

Trainer, M., B.A. Ridley, M.P. Buhr, G. Kok, J. Walega, G. Hübler, D.D. Parrish, and F.C. Fehsenfeld, Regional ozone and urban plumes in the southeastern United States: Birmingham, a case study. J. Geophys. Res., 100, 18,823-18,834, 1995.

Valente, R.J. et al., Ozone production during an urban air stagnation episode over Nashville, Tennessee, J. Geophys. Res. (this issue).

Weber, B.L., D.B. Wuertz, D.C. Welsh, and R. McPeek, Quality controls for profiler measurements of winds and RASS temperatures, J. Atmos. Oceanic Technol., 10, 452-464, 1993.

White, A.B., Mixing depth detection using 915-MHz radar reflectivity data (preprint volume), in Eighth Symposium on Observations and Instrumentation, Am. Meteorol. Soc., Boston, Mass., 248-250, 1993.

White, A.B., Radar remote sensing of scalar and velocity microturbulence in the convective boundary layer, NOAA Tech. Memo, ERL ETL-276, NOAA Environ. Res. Lab., Boulder, Co., 127 pp., 1997

R.J. Alvarez, R.M. Banta (corresponding author), and R.M. Hardesty, Environmental Technology Laboratory, NOAA/ERL, 325 Broadway, Boulder, CO 80303. (e-mail: rbanta@etl.noaa.gov).

F.C. Fehsenfeld, D. Parrish, and M. Trainer, Aeronomy Laboratory, NOAA/ERL, Boulder, CO

S.D. Mayor, National Center for Atmospheric Research, Boulder, CO 80307.

R.T. McNider, University of Alabama at Huntsville, Huntsville, AL 35899.

C.J. Senff, Cooperative Institute for Research in the Environmental Sciences, University of Colorado at Boulder, CO 80309.

R.J. Valente, Tennessee Valley Authority, Muscle Shoals, AL 35662.