Airborne lidar characterization of power plant plumes during the 1995 Southern Oxidants Study
Christoph J. Senff
R. Michael Hardesty and Raul J. Alvarez II
Shane D. Mayor
1. IntroductionDuring the summer of 1995, the National Oceanic and Atmospheric Administration's (NOAA's) nadir-viewing ozone lidar was operated on a Construcciones Aeronauticas SA (CASA) 212 aircraft for 6 weeks as part of the 1995 Southern Oxidants Study (SOS) Nashville/Middle Tennessee Intensive Ozone Study [Meagher et al., 1998]. The Nashville experiment, which included aircraft sampling as well as the operation of an intensive network of sophisticated air-quality monitoring stations, was aimed at comprehensively addressing the issue of production and accumulation of tropospheric ozone in and around a midsize southeastern city. A map of the region, showing the city as well as several large power plants, the emissions from which affect local ozone levels, is shown in Figure 1.
The CASA was one of six aircraft deployed during the 1995 SOS campaign [Hübler et al., 1998] over both regional and metropolitan areas to study the spatial extent of ozone pollution, the role of the interaction of precursors from different sources, and vertical and horizontal mixing processes. Four of the aircraft carried a variety of in situ sensors to measure ozone and its precursors, other relevant trace gases, and local meteorological parameters. The primary role of the CASA, with its ozone and aerosol profiling lidar, was to complement the in situ measurements with remotely measured profiles of ozone and aerosol backscatter extending from above the mixed layer to the surface. The lidar measurements provided a vertical context for the in situ measurements obtained by the other aircraft, which have little information on the representativeness of their measurements at distances away from the aircraft. Additionally, the vertical aerosol profiles observed by the CASA were used to deduce information on the state and evolution of the mixed layer.
During SOS, the aircraft activities were coordinated to address five experimental objectives [Hübler et al., 1998], described as follows: (1) urban plume studies, ozone formation in the Nashville urban plume; (2) power plant plume studies, evolution of emissions and ozone, plume interactions; (3) subregional characterization, interaction of power plant and urban plumes and the manifestation in local ozone accumulation; (4) regional studies, context of Nashville measurements within the regional setting; and (5) aircraft intercomparisons, simultaneous operation of aircraft to obtain intercomparisons of aircraft measurements and data sets.
Although the NOAA airborne ozone lidar was applied to all of the above objectives except regional studies (limited range of the CASA excluded this application), this paper describes use of the instrument in a study of the Cumberland power plant plume. Results of the lidar measurements were analyzed to investigate the change in plume size and shape, as well as ozone production in the Cumberland plume as it was advected downwind from the plant on 2 days characterized by distinctly different meteorology. In addition, we used the airborne lidar data to determine the boundary layer (BL) height and to relate the vertical extent of the Cumberland plume to the BL depth.
Airborne ozone and aerosol lidars with similar capabilities as the NOAA/Environmental Technology Laboratory (ETL) airborne ozone lidar have been used before to study tropospheric ozone and aerosol characteristics [e.g., Browell et al., 1983; Browell, 1995] as well as to measure BL depth and investigate its impact on air quality [e.g., Browell et al., 1985b]. Also, studies of power plant plumes with lidar have been reported in the literature [e.g., Johnson, 1969; Uthe and Wilson, 1979; Uthe et al., 1980]. However, we are not aware of any other airborne lidar investigations of ozone production rates in power plant plumes.
Other related papers describe use of the lidar results for studying the Nashville urban plume evolution during a stagnation event [Banta et al., 1998] and present intercomparisons of the airborne lidar ozone measurements with those from in situ sensors mounted on the other aircraft [Alvarez et al., this issue].
2. Application of Airborne Lidar for Power Plant Plume CharacterizationAs noted above, one of the primary goals of the 1995 SOS campaign was to investigate plumes emitted by power plants in northern Tennessee and assess the extent to which they contribute to the high ozone episodes often encountered in the Nashville area during summer. Coal-fired power plants in the Nashville area are significant sources of NOx, which in the presence of nonmethane hydrocarbons (NMHC) can lead to catalytic formation of ozone downwind of the power plant. However, close to the stack there is a temporary decrease in ozone levels due to "titration" by high levels of NO in the near field of the plume. Farther downstream, ozone levels above the local background indicate net ozone production due to the reaction of plume NOx with volatile organic compounds (VOC) that are entrained into the plume in the dilution process. The basic chemistry of this process is well understood [White et al., 1976; Trainer et al., 1995]. However, questions remain as to how much ozone is actually produced in the power plant plume, how the ozone production efficiency depends on the chemical composition of the plume, what the relative importance of power plants as sources of tropospheric ozone is compared to other sources, and what measures could be taken to effectively reduce ozone production in power plant plumes.
To address these questions, power plant plumes in the greater Nashville area were investigated in a number of cases during the 1995 SOS field study, both with airborne in situ sensors as well as with the NOAA airborne ozone and aerosol lidar. Ryerson et al.  report on an extensive power plant plume study on July 7 during which the NOAA-P3 aircraft intercepted plumes from four different power plants at multiple distances downwind of the stack. Ryerson et al. studied the evolution of SO2, NOx, and O3 concentrations in the plumes and derived ozone production efficiencies per unit NOx emitted. While aircraft in situ probing of the plumes with an array of chemical species sensors is indispensable for investigating the chemical processes in the plume, coverage of the plume area is usually limited, since the plume is typically only transected at a single height. This means that assumptions must be made about the size and shape of the plume when the in situ data are interpreted. With a remote sensor, on the other hand, such as NOAA's airborne ozone lidar, the two-dimensional structure of the plume can be mapped out during a single overpass. Typically flown at an altitude of around 3000 m above sea level (asl), the downward looking lidar gives a detailed picture of the plume for each overflight. Generally, the plume shape can be deduced either from the aerosol signature due to enhanced backscatter from particles in the plume or from the ozone depletion near the stack and the ozone enhancement farther downwind. At the same time, gradients in the aerosol return can be used to characterize height and structure of the BL. BL dynamics play an important role as they control to a large extent the lateral and vertical growth of the plume. Also, the data gathered when mapping power plant plumes with the ozone lidar can be used to calculate ozone production rates as a function of distance downwind. Of course, the ozone lidar cannot yield information on other important chemical species in the plume. Hence in situ and lidar measurements of power plant plumes complement each other well, yielding a more complete picture of the processes that occur as the plume travels downwind of the stack and interacts with its environment.
3. Lidar Characteristics and Methodology of Data AnalysisThe differential absorption lidar (DIAL) system used to profile ozone and aerosol from the CASA aircraft is described in detail by Alvarez et al. [this issue]. The lidar incorporates a KrF excimer laser that directs a laser beam into hydrogen and deuterium-filled Raman cells to produce simultaneous pulses at five wavelengths (277, 292, 313, 319, and 360 nm) in the ultraviolet region of the spectrum. These light pulses are then emitted into the atmosphere via a nadir-pointing mirror. The atmospheric returns at each wavelength are gathered by a 50 cm diameter telescope and separated by means of a spectrograph. The 277-319 nm channels are each digitized at 5 MHz (30 m range resolution), whereas the 360 nm channel is digitized at 10 MHz (15 m range resolution).
The aerosol backscatter and extinction information is deduced from the 360 nm channel following the approach of Fernald , while a pair of the four remaining channels is used to retrieve ozone concentration profiles, making use of their different light absorption by ozone. For most of the SOS data set, we chose the 277/292 nm wavelength pair for ozone retrieval, primarily since this wavelength combination is least susceptible to systematic errors in ozone concentration due to aerosol effects. The fairly large separation between the two DIAL wavelengths makes it necessary to correct for differential backscatter and extinction due to air molecules and aerosol particles. The method that we apply to correct for these effects generally follows the approach that was first used by Browell et al. [1985a]. Owing to a limited knowledge about the aerosol optical properties, some uncertainty in the aerosol-corrected DIAL ozone profiles remains. As described by Alvarez et al. [this issue], we applied significant effort to minimize these remaining aerosol correction errors, making use of the information at all the lidar wavelengths. In addition, we correct for biases in the data due to instrumental effects that were specific to the SOS 95 data set. Lidar signals that were recorded during an aircraft turn have not been used for retrieval because the lidar beam path is slanted rather than vertical. Also, lidar signals contaminated by strong backscatter and attenuation due to clouds have been excluded from the ozone and backscatter retrieval.
Although the lidar produces 20 pulses per second, significant averaging of the backscattered signals, both in range and over multiple pulses, is employed to reduce uncertainty in the ozone calculation. The number of pulses averaged can be varied as part of the data analysis. For most of the SOS data set, we averaged 160 laser pulses, resulting in a temporal measurement resolution of 8 s. At the nominal CASA airspeed of 65 m s-1, this gives a horizontal resolution of 520 m in the computed ozone and aerosol fields. Because both aerosol and ozone usually exhibit more vertical than horizontal structure, ozone concentrations and aerosol backscatter were computed with 90 and 15 m vertical resolution, respectively.
Near-range uncertainties in the overlap between the transmitter beam and the receiving telescope's field of view resulted in a minimum range of about 800-1000 m during SOS. Thus, although the plane typically flew at 3000 m, most of the reliable measurements begin at about 2000 m asl. Since the surface elevation in the study area varies from 200 to 350 m asl, and the ozone retrieval algorithm requires vertical averaging and differentiation of the lidar signals, the lowest height with continuous lidar ozone measurements is about 450 m asl (or 100 to 250 m above ground level).
Comparisons of the DIAL ozone measurements with those from the in situ instruments aboard other SOS aircraft produced an average difference of about 4 ppbv, with the largest offset between lidar and in situ sensors being less than 10 ppbv [Alvarez et al., this issue]. The rms errors in ozone concentration due to statistically uncorrelated fluctuations in the lidar measurements varied form about 3 to 14 ppbv increasing with range away from the lidar [Alvarez et al., this issue]. The statistical errors of the lidar ozone measurements are especially noticeable at lower altitudes since the lidar signals become weaker with increasing range due to absorption and scattering of the laser light. Of course, the rms errors in the DIAL ozone measurements can be reduced significantly by further averaging the lidar data horizontally and vertically. However, investigations of small-scale features such as the power plant plume study presented in this paper require a high spatial resolution in order to resolve the relevant process scales.
The lidar signal at 360 nm was also used to estimate the height of the BL using a method described by Senff et al. . Since the 360 nm signal is not absorbed by ozone, variations in signal intensity with range reflect to a large extent changes in aerosol backscatter along the beam path. More specifically, the derivative of the logarithm of the range-corrected signal at 360 nm, d/dR ln(P360×R2), is proportional to the aerosol backscatter gradient. This quantity is a sensitive indicator for the BL height, since the transition from the mixed layer to the free troposphere is often characterized by a significant drop in aerosol backscatter.
4. The Cumberland Power Plant Plume
Power plant plume studies represented an important component of the Nashville experiment. Because three large Tennessee Valley Authority (TVA) fossil fuel power plants are situated within 80 km of downtown Nashville, it was critical that the chemistry in the power plant plumes in both rural and urban regions, as well as the dynamics and evolution of the plumes and their relationship to mixed layer height and local meteorology, be well understood and observed. The work described in this paper focuses on the evolution of and ozone production in the plume emitted from the Cumberland power plant, located approximately 70 km NW of downtown Nashville. The Cumberland plant is the largest in terms of generating capacity (2600 MW) and NOx emissions (~3 kg s-1) within the Nashville study area. The plant produces ~0.5 kg s-1 of SO2; this figure was much higher prior to 1995, when 95% efficient scrubbers were installed at the plant. The stacks of the plant are 193 m high.
The Cumberland power plant plume was investigated with the airborne lidar on several days during the SOS 1995 campaign. Here we report on two cases, July 7 and 19, which were both characterized by a northwesterly synoptic-scale flow, but otherwise had distinctly different meteorological conditions. The implications of these different conditions on the plume characteristics will be discussed in the following two sections. We chose July 7 and 19 because in both cases the Cumberland plume was clearly defined and well separated from other power plant plumes or the Nashville urban plume. On other days, most notably July 18, the Cumberland plume had merged with other power plant plumes, making a quantitative analysis of the plume more difficult. These data sets offer the opportunity for a future investigation of the chemistry that occurs when plumes of different ages and composition merge.
To support our investigation of the Cumberland plume, we used data provided by radar wind profilers as well as meteorological surface data. During the SOS 1995 campaign, three radar wind profilers were deployed by NOAA to provide continuous profiling of winds and estimates of the BL height in the study area [White and Gottas, 1997]. The profiler closest to the Cumberland power plant was located at the Dickson surface station, about 30 km southeast of the plant (Figure 1). Therefore the profiler and surface data from the Dickson station were used to complement the lidar data in the Cumberland plume study.
5. July 7 Case
The synoptic weather situation on the morning of July 7 was characterized by high pressure to the west of Tennessee causing northwesterly winds over the SOS study area. Figures 2a and 2b show vertical profiles of wind speed and direction measured with the wind profiler at Dickson. Hourly averages ending at 1100 and 1200 LT are depicted covering the time period during which the CASA aircraft mapped out the Cumberland power plant plume. In addition, an estimate of the BL height derived from the lidar data (as described in section 3) is plotted in Figures 2a and 2b for the same two time intervals. The two wind speed profiles show values of about 5 m s-1 below 1000 m asl. Above 1000 m asl, wind speeds increase to 15-20 m s-1 at 2000 m asl. The wind direction is almost constant with height, decreasing from 340° near the surface to around 315° at 1000 m asl and then increasing again slightly. The BL height increases from ~1050 to 1500 m asl reflecting the late-morning rise of the BL top due to strong convective turbulence.
On July 7, the CASA aircraft equipped with the ozone lidar took off at 0944 LT and headed in a northwesterly direction toward the Cumberland power plant. To examine both the spatial and temporal evolution of the Cumberland power plant plume, the same flight track consisting of three to four transects across the plume was flown twice between 1000 and 1150 LT. Then the CASA headed back east to investigate the developing urban ozone plume over Nashville. Plate 1a shows the CASA flight track during the first mapping of the Cumberland power plant plume from 1000 to 1045 LT. The colored bands in Plate 1a represent vertical averages of ozone concentration measured with the ozone lidar at two different height ranges below the aircraft. The flight track is depicted as a thin black line inside the colored bands. The left panel of Plate 1a shows ozone values averaged over a height range of 500 to 1000 m asl, whereas in the right panel ozone values are averaged over 1000 to 1500 m asl. To facilitate plotting of these "maps" of ozone concentration the DIAL measurements taken every 520 m (8 s) along the flight path were interpolated onto a 1 × 1 km grid. Each grid point was assigned a weighted average of all the ozone measurements within a 4 × 4 km square with the grid point in the center. The weights were chosen as the inverse of the squared distance between the grid point and the location of the ozone measurements along the flight track. In the left panel of Plate 1a the times when the CASA aircraft flew a turn and started a new flight leg are plotted next to the flight track. As mentioned before, data taken during aircraft turns have been excluded from the ozone calculations. Because of the interpolation procedure used to create the ozone plots, these data gaps at the turns do not appear in Plate 1a.
During the first mapping, the Cumberland power plant plume was intercepted three times by the CASA aircraft. First on the N-S leg between 1002 and 1013 LT, then farther downwind on the SW-NE leg ending at 1025, and then again on the E-W leg between 1031 and 1043. In the left panel of Plate 1a at lower heights (500-1000 m asl), the plume can be clearly identified during the first two transects by its low-ozone signature which is due to "titration" of ozone at high NO concentrations close to the stack. Farther downwind the drop in ozone becomes less pronounced as the plume is diluted by entrainment of ambient air, and plume NOx reacts with precursor gases to form ozone. To verify that the areas of ozone depletion in the left panel of Plate 1a are in fact associated with the Cumberland power plant plume, a back trajectory analysis is performed starting roughly at the plume center during the third transect. The back trajectory was calculated from the hourly E-W and N-S wind components measured with the wind profiler at Dickson. The trajectory arrow closest to the starting point represents the path traveled by an air parcel during the time elapsed between the starting time of the trajectory analysis and the last full hour prior (in this case 40 min from 1000 until 1040). All the other trajectory arrows correspond to a 1-hour travel time. The back trajectory lines up almost exactly with the center of the plume on each transect and the location of the Cumberland power plant. This clearly demonstrates that the ozone depletion seen in the left panel of Plate 1a is linked to the Cumberland power plant plume. Since the plume is mostly confined to heights below 1000 m asl (see Plate 2a) the plume ozone signature cannot be identified in the right panel of Plate 1a. East of the third plume transect, the lidar detected a small region of above-background ozone. Most likely this patch of high ozone is associated with the plume, indicating net ozone production at the plume edge. However, judging by the lidar ozone data only, it cannot be ruled out that this region of above-background ozone may have another origin. Later on we will discuss in more detail whether or not these ozone enhancement areas are associated with the Cumberland plume and what implications this has for the ozone production calculations.
In Plate 1b, the CASA flight track and the DIAL ozone measurements along the flight path are shown in the same manner as in Plate 1a, but for the second plume mapping between 1100 and 1151 LT. The flight pattern was very similar to the first plume mapping except that after crossing the plume for the third time (during the E-W leg from 1128 to 1139), the CASA turned east crossing the Cumberland plume a fourth time farther downwind and then continued on toward Nashville. Again, the back trajectory shown in the left panel of Plate 1b confirms that the swath of below-background ozone levels is caused by the Cumberland plume. The trajectory and the plume signature at the first transect (1101 to 1112) do not quite line up. This can probably be explained by the fact that with 5 m s-1 winds (see Figure 2a) the trajectory analysis from hourly averaged wind data does not resolve details below about 20 km, which is larger than the distance between adjacent transects. During the second mapping of the Cumberland plume, the low-ozone signature is also visible above 1000 m asl (Plate 1b, right panel). This is in agreement with the increase in BL height to ~1500 m (Figures 2a and 2b), which allowed the plume to grow in the vertical to heights above 1000 m. Areas of enhanced ozone at the edges of the plume are clearly visible in the third and fourth plume crossing, both at lower and upper height levels. The appearance of these isolated patches of above-background ozone at the edges of the plume in two consecutive transects is a strong indication that these regions are associated with the plume.
To document the size and shape of the Cumberland plume in more detail, time-height cross sections of ozone concentration measured with the DIAL are shown in Plates 2a and 2b. The data are plotted from 500 to 1800 m asl for the same two time spans as in Plates 1a and 1b. The data were interpolated from their original 520 m horizontal × 90 m vertical resolution onto a 130 m × 9 m grid in order to avoid "blocky-looking" color plots. One minute of data in Plates 2a and 2b correspond to a horizontal distance of about 4 km as indicated by the scale in the upper right corner of the first color panel. Reddish colors represent ozone values of 70 ppbv and above, while dark purple and black indicate ozone concentrations of 15 ppbv or lower. Sometimes, at low altitudes, isolated red-colored patches occur close to green or blue patches, indicating large fluctuations in ozone concentration. As mentioned in section 3, these fluctuations are due to processing the lidar data at high resolution despite low signal-to-noise ratios near the ground. We chose to process the lidar data at this high vertical and horizontal resolution in order to attempt to resolve small features such as the Cumberland plume in detail. The Cumberland plume is again easily identified by the ozone depletion in the core of the plume. For easier reference, the plume transect numbers and the distances between the Cumberland plant and the plume crossings are plotted next to the respective plume signatures. During the first plume mapping, the plume was intercepted at 19, 34, and 41 km downwind of the plant (transects 1, 2, and 3 in Plate 2a), while during the second mapping the plume crossings occurred at 15, 29, 41, and 54 km (transects 1, 2, 3, and 4 in Plate 2b). For the plume transects closest to the Cumberland power plant, ozone values of less than 15 ppbv are observed in the core of the plume. This is a significant drop from the 45-55 ppbv background ozone levels found on either side of the plume. Farther downwind, ozone levels in the plume core increased to about 25-35 ppbv due to dilution by entrainment of ambient air and conversion of plume NOx to ozone. Above-background ozone levels of ~70 ppbv at the edge of the plume are visible in plume transect 3 in Plate 2a (near 1036) and in plume transects 3 and 4 in Plate 2b (near 1131 and 1147). The in situ measurements by Ryerson et al.  confirm both the magnitude of the ozone enhancement at the plume edges as well as its asymmetric pattern.
The shape of the plume for all seven transects in Plates 2a and 2b is fairly symmetric with a clearly defined center, except for transect 2 in Plate 2a where the plume appears to have two cores with low ozone values. The plume retains its regular shape out to downwind distances of more than 50 km. Most likely, this can be attributed to the relatively steady and swift northwesterly wind which carried the plume downwind at about 5 m s-1 (Figure 2a). As is evident from Figure 2b, there was practically no directional wind shear with height which could have torn the plume apart as it traveled downwind. Also, owing to the fairly strong synoptic-scale flow, the influence of local wind fields on plume shape and direction was minimal. In both cases, the Cumberland plume grew in size both laterally and vertically as it was advected downwind. The lateral growth is probably due to diffusion and horizontal turbulence, while the height growth is caused by vertical convective mixing in the growing daytime BL. The plume signature of transect 2 in Plate 2a, for example, shows low-ozone updrafts as well as entrainment of background air near the top of the plume, indicating considerable vertical mixing. During the first plume mapping, the Cumberland plume grew from about 1000 m to about 1200 m asl as it advected downwind. During the second pass, the plume height increased rapidly from about 1200 to 1800 m asl. To compare the vertical extent of the Cumberland plume with the depth of the BL, the BL height as measured with the lidar is overplotted as a black line in Plates 2a and 2b. The resolution of the BL height measurement is 8 s or 520 m horizontal. Over the course of the two plume mappings, the height of the BL increases from about 700 m to around 1800 m asl. The undulations at the top of the BL are due to strong updrafts pushing into the free troposphere aloft followed by entrainment of tropospheric air into the BL. The BL height measurement also shows drops and rises of a few hundred meters at scales of about 10 min (40 km horizontally). For example, the BL height drops significantly between 1127 and 1138. These slow variations in BL depth could be due to varying surface conditions, such as different land use, surface moisture etc., resulting in locally different BL heights, as found by Banta et al. . In all cases, the vertical extent of the Cumberland plume is almost exactly the same as the BL height. This demonstrates clearly that vertical, convective mixing processes spread the plume across the entire depth of the BL.
In the transition zone between near-field depletion of ozone in the Cumberland plume and enhancement of ozone farther downwind, ozone is not an ideal plume tracer. At transect 3 in Plate 2a and transect 4 in Plate 2b, the location of the plume boundaries becomes less obvious. In principle, the aerosol backscatter data derived from the lidar's 360 nm signal should also reveal the plume structure due to enhanced backscatter from the particles in the plume. However, in the July 7 case, the aerosol backscatter data (not shown) only show a faint plume signature for the very first plume transect at 1007. For the remaining transects, the plume aerosol signature becomes indistinguishable from the background, precluding the backscatter data from being used to identify the plume. In situ measurements of other trace gases such as NOx or SO2 would be helpful in those cases where the lidar data do not provide clear evidence of the plume width. However, such in situ measurements taken at the same time as the lidar measurements at various altitude levels in the BL are not available. Therefore we are left with the lidar ozone measurements to provide us with information on plume size and shape.
From the DIAL ozone data at the various plume transects, the ozone production rates in the plume can be derived as a function of plume travel time using an approach similar to the one employed by Ryerson et al. . By integrating the ozone concentration difference between plume and background across the plume for each transect, the effect of plume dilution is accounted for [White et al., 1976]. The difference in plume-integrated ozone concentrations at two adjacent transects is then a measure for ozone production or destruction in the plume between these transects. The cross-plume integral of ozone concentration differences is calculated as follows: for each transect, the time series of ozone concentration including a few minutes worth of data before and after the plume crossing are displayed for each measurement height. As an example, the ozone time series for transect 2 during the second plume mapping (at 1119 LT) at a height of 950 m asl is shown in Figure 3. Background regions (marked by vertical dotted bars) are first chosen far to the left and right of the plume location and then moved in toward the plume until the measured ozone concentrations deviate significantly from the background, that is, until a drop in ozone (at plume crossings close to the power plant) or an enhancement in ozone (at the wings of the plume farther downwind) is detected. The background ozone concentration is then determined by a straight-line fit across the plume. The differences between the interpolated background and the plume ozone measurements are summed for all data points of the time series that fall between the middle points of each background region (depicted by a thick black line in Figure 3), and then are multiplied by the vertical (90 m) and horizontal (520 m) resolution. This process is repeated for all heights that apparently are affected by the plume, and the results are added. Since the ozone DIAL data do not extend all the way to the surface, we assume that the ozone concentrations between the lowest height at about 500 m asl and the surface (typically 200-350 m asl) are the same as at the lowest measurement height. Since the BL appears to be well-mixed, this is a reasonable assumption to make. The height asl of the surface is determined from the location of the ground backscatter peak in the lidar's 360 nm signal. Furthermore, since the CASA aircraft did not intersect the plume at a 90° angle, making the plume width appear larger than it is, the above integrations were corrected accordingly. The angle is taken as the angle between the CASA flight direction and the line connecting the Cumberland power plant location with the center of the plume at each crossing.
Plume-integrated ozone and ozone production rates calculated with the above mentioned method may be subject to systematic errors for those transects where the horizontal extent cannot be determined unambiguously from the lidar ozone data. For example, at transect 3 in Plate 2a and transect 4 in Plate 2b plume-integrated ozone and ozone production rates will vary depending on how much of the enhanced-ozone areas is included in the calculation. To assess these errors due to ambiguity in plume width, the calculations of integrated ozone and ozone production were performed three times with slightly shifted ozone background regions, that is, for three different assumed plume widths. The range of these calculations is a measure of how critically plume-integrated ozone and the average ozone production rate in the plume depend on plume width. The placement of the background regions for these three calculations is shown in Plates 2a and 2b for each plume cross section at one particular height.
Figure 4 shows the integrated plume ozone difference with
respect to the background as a function of plume travel time for both
plume mappings on July 7. The plume travel time was estimated
using the back trajectory analysis. The plume-integrated ozone is
displayed in units of 1024 molecules per meter, indicating the
number of ozone molecules in a 1-m-thick slice perpendicular to the
direction in which the plume is moving. This unit was primarily
chosen to facilitate intercomparison with the results of Ryerson et al.
. The data shown in Figure 4 are the average of three
calculations with slightly different ozone background regions, that
is, different assumed plume widths. The error bars represent the
range of these three calculations. For the most part, plume-integrated ozone does not depend significantly on exactly where the
plume boundaries are placed, except for transect 4 of the second
plume mapping. However, the larger variation at this data point does
not affect the general trend of integrated ozone as a function of
plume travel time. The curves for both plume mappings show a
similar behavior: the integrated ozone difference drops from about
molecules m-1 at around 1 hour plume travel time down to
-3 to -3.5×1024
molecules m-1 at ~1.6 hours plume travel time, after
which both curves rise again at about the same rate as they were
falling before. This means that ozone in the plume is destroyed
during the first 2 hours as it travels downwind. After about 2 hours,
ozone is produced in the plume. The CASA did not fly far enough
downwind to observe net ozone production in the Cumberland
plume, resulting in positive integrated ozone differences. However,
by extrapolating the two curves in Figure 4 beyond 3 hours plume
travel time, we can estimate that the zero crossing occurs
somewhere between 3 and 3.4 hours of plume travel time. We can
perform a consistency check by extrapolating the two curves back
to 0 hour travel time. Immediately at the source point of the power
plant plume, the difference between ozone concentrations inside the
plume and the background should be zero. The extrapolated curves
cross the 0-hour line at
Our findings compare favorably with the results of Ryerson et al.
, who also investigated the Cumberland plume on July 7 with
in situ sensors mounted on the NOAA-P3 aircraft. They flew much
farther downwind of the plant, tracking the plume to just over 9
hours of plume travel time, thus documenting the region of plume-enhanced ozone much more extensively, but did not fly as close to
the source as the CASA. The magnitude of the integrated ozone
values are in good agreement in the region where the NOAA-P3 and
the CASA lidar measurements overlap. Ryerson et al. converted
their integrated plume ozone measurements to ozone fluxes by
multiplying with the average BL wind speed (5 m s-1, Figure 2a).
Multiplying the data in Figure 4 with a wind speed of 5 m s-1 would
yield fluxes of
In Figure 5, the plume ozone production rate is displayed in units of ppbv per hour. The production rates are derived from the data shown in Figure 4 by taking the difference in integrated ozone between adjacent plume transects and multiplying by the average plume area. For convenience, the production rates are converted from molecules m-3 h-1 to ppbv h-1 by estimating the air density in the plume from surface pressure and temperature measurements at Dickson. The plume area was calculated by integrating the 520 m × 90 m range cells of the lidar ozone measurements across the plume. As plume boundaries, we chose the "inner" limits of the background regions right and left of the plume (see Figure 3). Similar to Figure 4, the ozone productions rates in Figure 5 are the average of three calculations with different ozone background regions, with the error bars representing the range of these three retrievals. The ozone production rates are also not significantly affected by the placement of the plume boundaries. We observe negative ozone production rates of -5 to -8 ppbv h-1 that increase to zero at about 1.7 hours plume travel time (corresponding to the inflection point in Figure 4) and then become positive and rise to around 3 to 4 ppbv h-1. Although the difference in integrated ozone per unit time between transect 3 and 4 during the second plume mapping (dashed line in Figure 4) is twice as large as for the previous transect pair, the plume ozone production rate does not reflect this, since the plume area increases significantly at the same time.
6. July 19 Case
Surface weather maps for the morning of July 19 show an area of weak high pressure west and north of the SOS 1995 study area, resulting in generally northerly winds over Tennessee. Owing to weak pressure gradients across the central and southeastern portion of the United States (less than 5 hPa pressure difference from the Great Lakes south to Kentucky and Tennessee), surface winds were very light. In support of this, wind profiler measurements at Dickson show much lower wind speeds on July 19 than on July 7. In Figures 6a and 6b, hourly averaged wind speed and wind direction profiles are displayed for 1000-1100 and 1100-1200 LT, covering the time period during which the Cumberland plume was investigated with the lidar on July 19. The BL height estimates derived from the lidar data are plotted as horizontal lines for the same two time periods. In contrast to the July 7 case, the BL height on July 19 only increases slightly, from around 900 to 1150 m asl between 1000 and 1200 LT. Within the BL, wind speeds were less than 2 m s-1, and the wind direction was generally from the northwest except for the lowest 200 m of the second profile (1100-1200 LT), where south to southwesterly winds were observed. While wind speeds only gradually increased above the BL to around 5-9 m s-1 at 2500 m asl, the wind direction profiles show a strong wind shear above the BL top. The winds shift about 90°, from north to east in a layer about 500 m thick just above the BL top. Above 1600 m asl, the winds shift back to a northerly direction.
Plate 3 shows the CASA flight track with superimposed lidar ozone measurements from 1050 to 1150 LT. The data are displayed in the same way as for the July 7 case (see Plates 1a and 1b), except that the color scale extends to ozone concentrations of 100 ppbv rather than 75 ppbv. The CASA aircraft left Nashville at 0935 LT, again heading northwest toward the Cumberland plant. After flying a 100 km north-south leg twice for intercomparison purposes with other aircraft equipped with in situ ozone sensors [Alvarez et al., this issue], the CASA circled the Cumberland plant on the north side and then started mapping the Cumberland plume downwind of the stack. Between 1055 and 1117, the CASA crossed the plume three times very close to the stack. It intercepted the plume several more times farther downwind, flying in a zig-zag pattern. The Cumberland plume can again be readily identified by the ozone deficit inside the plume, both at lower and higher altitudes (left and right panels in Plate 3). In the transect starting at 1138 and ending at 1150, ozone data are missing in the expected location of the plume. This gap is due to cumulus clouds blocking the lidar beam, thereby preventing lidar ozone measurements in or below clouds. The slight drop in ozone concentration left and right of the gap, as well as a slight ozone enhancement at the plume edges, suggests the presence of the plume. However, owing to a lack of data in the plume core, this transect cannot be used to characterize the plume. Three additional plume crossings farther downwind were also contaminated by cloud interference.
The back trajectory starting at around 1130 at the center of the plume in plume transect 5 lines up fairly well with the low-ozone plume signature in the earlier transects. The bend in the plume path between transect 4 (1117 to 1126) and transect 5 (1126 to 1138) especially is reproduced very well by the trajectory. The fact that the trajectory passes about 10 km east of the Cumberland plant can be attributed to increased back trajectory uncertainty due to the light-wind conditions on July 19. In this case, the profiler wind measurements at Dickson may not be representative of the wind field near the Cumberland plant. At heights above 1000 m asl (right panel in Plate 3), the plume path is shifted slightly to the southwest in agreement with the winds coming from easterly directions above 1000 m (Figure 6b). The high ozone concentrations of around 90 ppbv observed adjacent to the Cumberland plume at altitudes below 1000 m are not due to ozone production in the plume. The region of high ozone is too large and too close to the plant for that to be the case. Most likely, this patch of high ozone had been advected overnight to the Cumberland plant vicinity from another pollution source.
Plate 4 shows a time-height cross section of the lidar ozone measurements covering the same time span as Plate 3 except that the last 15 min, which include the plume transect contaminated by clouds, were omitted. The data are displayed in the same manner as the July 7 data in Plates 2a and 2b. Again, the black line represents the BL height estimates derived from the lidar data. During the first 35 min of the time period displayed in Plate 4, the BL height stays at around 1000 m asl and then increases to about 1300 m asl (near 1125). The area of high background ozone concentrations close to the power plant shows up prominently in Plate 4 (1052 to 1115). The high ozone levels, which reach values of 100 ppbv and more, are clearly confined to the BL. Above the BL top, background ozone levels drop to around 65 ppbv. The Cumberland plume is again easily identified by its low-ozone signature. The plume was intercepted five times at 7, 11, 12, 21, and 29 km downwind. In all cases, the plume shape is very irregular. The upper part of the plume extends several hundred meters above the top of the BL (except for transect 5) and is shifted horizontally with respect to the bottom part of the plume. This shift increases with downwind distance and leads to a breakup into two plumes with separate cores that are only loosely connected (see transect 4 at 1122). This situation is very different from the July 7 case where the Cumberland plume had a very regular, symmetric shape and was confined to the BL at all times. Typically, the transport of trace gases or particles between the BL and the free troposphere is suppressed. A possible explanation for the Cumberland plume reaching far above the BL is that the plume penetrated a very low or not very well-defined BL top right after it was released from the power plant stack. The estimated travel time of the plume intercepted closest to the stack (transect 2) is about 1 hour, meaning that the plume was released from the stack at around 1000 LT. BL height estimates from the lidar data earlier in the flight (not shown) indicate that the BL height at that time in the Cumberland plant vicinity was about 500 m asl. With the stacks of the plant reaching to 330 m asl, it is conceivable that just after release part of the plume penetrated the stable layer, capping the BL. During their downwind travel, the portions of the plume above and below the BL top were subjected to different meteorological conditions such as different wind direction and turbulence characteristics, which explains the odd shape of the Cumberland plume on July 19. The strong directional wind shear just above the BL causes the portion of the plume above the BL top to drift in a southwesterly direction, while the plume part within the BL travels in a southeasterly direction. In transects that were flown in an east-west direction (transects 2 and 4), this drift of the upper part of the plume manifests as a shift to the right, that is, observed later in time, with respect to the bottom part of the plume. Correspondingly, plume crossings from west to east (transects 1, 3, and 5) show a drift to the left. Again, as the Cumberland plume travels downwind, it becomes considerably wider but only grows slightly in height from 1400 to 1600 m asl. The ozone levels in the plume core drop steadily, reaching their lowest values (less than 10 ppbv) at 21 km downwind (transect 4). Eight kilometers farther downwind (transect 5) the plume ozone levels have increased slightly, indicating that ozone production in the plume has started.
Plate 5 shows a time-height cross section of the aerosol backscatter coefficient in units of 10-5 m-1 sr-1 retrieved from the lidar's 360 nm channel. The time period covered in Plate 5 and the vertical extent of the data are the same as in Plate 4. Contrary to the July 7 case, enhanced backscatter from the Cumberland plume is clearly visible in plume transects 1 to 4. However, when comparing the aerosol and ozone plume signatures, it is evident that the aerosol signatures show smaller plume cross sections than the ozone data. Obviously, enhanced backscatter at 360 nm occurs in the core of the plume, but backscatter from the edges of the plume as determined from the plume ozone signature becomes indistinguishable from the background. The ozone measurements, on the other hand, still show a significant drop in ozone at the plume edges compared to background ozone levels. In transect 5, after the plume had been further diluted by entrainment of ambient air, the aerosol signature of the plume vanishes, similar to the July 7 case in which the plume could not be traced from the aerosol backscatter data in any of the plume crossings except the one closest to the stack. These observations indicate that only the plume cores fairly close to the source where particle concentrations are high leave a clear aerosol signature at the 360 nm lidar wavelength, whereas diluted plume air cannot be distinguished from background aerosol. The aerosol plume signature on July 19 is much stronger than on July 7 probably because the lower wind speeds on July 19 delay the dilution process, resulting in higher particle concentrations in the plume core. Another reason for higher backscatter from within the plume on July 19 could be deliquescence of the plume aerosol particles due to higher relative humidity. However, radiosonde measurements taken about 90 km east of the Cumberland plant indicate lower relative humidity on July 19 than on July 7.
Similar to the July 7 case, on July 19 plume ozone was integrated relative to the local background as a function of plume travel time, and an analysis of ozone production rates in the plume was performed, the results of which are displayed in Figures 7 and 8. Again, the data presented in Figures 7 and 8 are the average of three calculations with slightly different background regions; the error bars represent the range of those three retrievals. The placement of the background regions for these three calculations is shown in Plate 4 for each plume cross section at one particular height. Integrated ozone and ozone production rates are even less affected by the exact placement of the plume boundaries than on July 7. The measurements for plume transects 1 and 3 were averaged since these two plume crossings practically occurred at the same downwind distance; hence Figure 7 contains only four data points, although the plume was intercepted five times. For intercomparison purposes, the results of the second plume mapping on July 7 (dotted lines in Figures 7 and 8) are plotted together with the data for July 19 (solid lines). The plume-integrated ozone values on July 19 drop from about -2× 1024 molecules m-1 at ~1 hour plume travel time down to about -4×1024 molecules m-1 at 4 hours travel time and then increase again slightly. Extrapolation of the data back to 0 hour travel time yields an integrated ozone value of -0.3× 1024 molecules m-1, close to the expected value of zero. The slow drop in plume-integrated ozone again suggests incomplete dilution of the plume in the near field. The two curves for July 7 and 19 have a similar shape, and the measurements agree well close to the source. But plume-integrated ozone on July 7 increases sharply after 1.6 hours of travel time, while on July 19 the inflection point occurs much later at 4 hours. The ozone production rate in the Cumberland plume close to the stack (at about 1.3 hours of plume travel time) was about -8 ppbv h-1 in both cases (see Figure 8). On July 19, the ozone production rate increased gradually with time, crossing the zero line at 4 hours plume travel time and reaching a value of about 1.5 ppbv h-1 45 min later. In contrast, on July 7, ozone production rates rose sharply as the plume traveled downwind, reaching positive production rates of 3 ppbv h-1 already after 2 hours with a further slight increase after that.
The slower increase in ozone production rates on July 19 could at least partly be caused by an overestimation of the plume travel time on that day. Because of the low wind speeds on July 19, the Dickson profiler measurements may not be representative of the wind field along the plume path. More likely, the delayed ozone production on July 19 may have been caused by lower background levels of ozone precursor gases. Changes in plume chemistry due to different NOx emission rates of the Cumberland plant can be ruled out, since the NOx emissions were very similar on both days (8.2×1025 molecules s-1 on July 7 versus 7.3× 1025 molecules s-1 on July 19). Another possible explanation could be that less vigorous turbulent mixing on July 19 preserved pockets of high NO concentrations in the plume longer than on July 7. Eventual mixing of these high-NO regions with ambient air leads to ozone destruction much later than on July 7.
7. Summary and Conclusions
For two cases, July 7 and July 19, that were characterized by distinctly different meteorological conditions, we investigated the Cumberland power plant with the NOAA airborne ozone and aerosol lidar by flying multiple downwind transects across the plume. We documented the evolution of the plume's size and shape and derived ozone production rates in the plume. By comparing the vertical extent of the plume with BL height estimates derived from the lidar data, we could show that on July 7 due to strong vertical mixing the Cumberland plume was spread across the entire depth of the BL but did not reach into the free troposphere. On the other hand, on July 19, part of the plume had penetrated the BL top. Subject to different meteorological conditions, the two portions of the plume above and below the BL top evolved differently as they advected downwind. The plume shape on July 19 was very irregular, mostly due to a strong directional wind shear at the interface between mixed layer and free troposphere. This wind shear drew the plume apart, resulting in two smaller plumes that were only loosely connected. In contrast, on July 7, the plume was shaped symmetrically and had a clearly defined, single core. As the plume traveled downwind, it retained its shape to the farthest distance observed, over 50 km from its source.
By combining the ozone lidar measurements with a back trajectory analysis using wind profiler data, we could confirm that regions of low ozone were clearly linked to the Cumberland plume. At distances farther downwind of the power plant, we observed areas of above-background ozone close to the low-ozone plume signature, indicating net ozone production at the plume wings. Since in those cases the lidar ozone data did not yield clear evidence of the plume width, we could not establish exactly how far the plume wings extended and how much of the enhanced ozone regions were part of the plume. However, we demonstrated that this ambiguity in the horizontal extent of the plume did not significantly affect the determination of ozone production rates in the plume. Close to the plant where ozone is destroyed due to "titration" by elevated NO levels, we measured ozone destruction rates of -5 to -8 ppbv h-1. Farther downwind, where ozone is formed in the plume due to reaction of NOx with ozone precursor gases, we observed ozone production rates of about 4 ppbv h-1 on July 7 and 1.5 ppbv h-1 on July 19. These production rates, especially the higher rates on July 7, confirm that NOx-rich power plant plumes have the potential of raising local ozone levels significantly over the course of a day. On July 7, ozone production started after less than 2 hours of plume travel time, while on July 19 positive production rates were only observed after 4 hours. We speculated that this delayed onset of ozone production might be due to lower background concentrations of ozone precursor gases on July 19 on the grounds that the NOx emission rates of the Cumberland plant are similar for both days. Another possible explanation could be slower dilution of the plume on July 19 due to weaker turbulence, leaving pockets of high NO concentrations that destroy ambient ozone much later as the plume travels downwind. We also found that aerosol backscatter, observed with the lidar at 360 nm, was not a good tracer for plume dimensions in the case of the Cumberland plume. A clear aerosol signature was only detected in the undiluted core of the plume close to the plant.
The two plume case studies presented here clearly demonstrate the value of using airborne lidar for gathering information on important air quality issues such as the role of power plant plumes in regional ozone pollution. The July 19 case especially highlights how important it is to document the two-dimensional structure of the atmosphere. Because of the irregular plume shape on July 19, in situ probing of the plume would not have been adequate and would have probably produced misleading results. While lidar cannot replace in situ measurements of air pollutants, the unique information it yields on the state of the atmosphere can, in concert with in situ and other remote sensors, help address some of the scientific questions concerning ozone pollution in the lower troposphere.
Acknowledgments. Funding for this research was provided through the NOAA Health of the Atmosphere and the EPA Southern Oxidants Study Programs. The authors greatly appreciate comments and suggestions by Thomas Ryerson of NOAA/AL concerning power plant plume chemistry. The wind profiler data were provided by Allen White of NOAA/ETL, and L. Gautney (TVA) provided power plant emission rates. We also wish to thank Wynn Eberhard, Robert Banta, Charles Frush, Victor Kovalev, and Christopher Locker for their efforts during the field campaign. In addition, the authors acknowledge James McElroy and colleagues at the Las Vegas EPA Environmental Monitoring Systems Laboratory, who developed and tested the airborne ozone lidar before it was transferred to NOAA/ETL.
Alvarez, R. J., C. J. Senff, R. M. Hardesty, D. D. Parrish, 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., this issue.
Banta, R. M., et al., Daytime buildup and nighttime transport of urban ozone in the boundary layer during a stagnation episode, J. Geophys. Res., in press, 1998.
Browell, E. V., Airborne lidar measurements, Rev. Laser Eng., 23, 135-141, 1995.
Browell, E. V., A. F. Carter, S. T. Shipley, R. J. Allen, C. F. Butler, M. N. Mayo, J. H. Siviter Jr., and W. M. Hall, NASA multipurpose airborne DIAL system and measurements of ozone and aerosol profiles, Appl. Opt., 22, 522-534, 1983.
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, 1985a.
Browell, E. V., S. T. Shipley, C. F. Butler, and S. Ismail, Airborne lidar measurements of aerosols, mixed layer height, and ozone during 1980 PEPE/NEROS summer field experiment, NASA Ref. Publ., RP-1143, 1985b.
Fernald, F. G., Analysis of atmospheric lidar observations: Some comments, Appl. Opt., 23, 652-653, 1984.
Hübler, G., R. J. Alvarez II, P. H. Daum, W. Luke, D. Rider, and R. J. Valente, An overview of the airborne activities during the SOS 1995 Nashville/Middle Tennessee Ozone Study, J. Geophys. Res., in press, 1998.
Johnson, W. B., Jr., Lidar observations of the diffusion and rise of stack plumes, J. Appl. Meteorol., 8, 443-449, 1969.
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 the SOS Nashville/Middle Tennessee Ozone Study, J. Geophys. Res., in press, 1998.
Ryerson, T. B., et al., Emissions lifetimes and ozone formation in power plant plumes, J. Geophys. Res., in press, 1998.
Senff, C., J. Bösenberg, G. Peters, and T. Schaberl, Remote sensing of turbulent ozone fluxes and the ozone budget in the convective boundary layer with DIAL and Radar-RASS: A case study, Contrib. Atmos. Phys., 69(1), 161-176, 1996.
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.
Uthe, E. E., and W. E. Wilson, Lidar observations of the density and behavior
of the Labadie power plant plume, Atmos. Environ.,13,
Uthe, E. E., F. L. Ludwig, and F. Pooler Jr., Lidar observations of the diurnal behavior of the Cumberland power plant plume, J. Air Pollut. Control Assoc., 30(8), 889-893, 1980.
White, A. B., and D. J. Gottas, The use of Doppler wind profilers for air pollution studies, paper presented at COST-76 Profiler Workshop, European Commission, DG XII, COST, Engelberg, Switzerland, May 12-16, 1997.
White, W. H., J. A. Anderson, D. L. Blumenthal, R. B. Husar, N. V. Gillani, J. D. Husar, and W. E. J. Wilson, Formation and transport of secondary air pollutants: Ozone and aerosols in the St. Louis urban plume, Science, 194, 187-189, 1976.