Convective venting and surface ozone in Houston during TexAQS 2006

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115,, doi: /2009jd013301, 2010 Convective venting and surface ozone in Houston during TexAQS 2006 A. O. Langford, 1 S. C. Tucker, 1,2 C. J. Senff, 1,2 R. M. Banta, 1 W. A. Brewer, 1 R. J. Alvarez II, 1 R. M. Hardesty, 1 B. M. Lerner, 1,2 and E. J. Williams 1,2 Received 29 September 2009; revised 8 February 2010; accepted 5 April 2010; published 20 August [1] The influence of convective mixing on surface ozone in Houston during TexAQS 2006 is examined. We use airborne lidar measurements of ozone and ship based Doppler lidar measurements of winds, together with ship and ground based measurements of surface ozone to characterize horizontal and vertical mixing of ozone plumes from the Houston Ship Channel on two high ozone days. We show that a stable capping layer trapped the plume in the boundary layer on 31 August, while shallow convection associated with active fair weather cumulus clouds mixed the plume with free tropospheric air on 17 August. Deep convection associated with an isolated air mass thunderstorm further decreased surface ozone near Galveston Bay in the late afternoon. High ozone thus affected a smaller area for a shorter period on 17 August, despite similar background concentrations and local production. We generalize these findings by comparing Houston ozone concentrations to National Weather Service (Lake Charles, LA) radiosondes. We show that for 1 June to 15 September 2006, stable conditions with high background ozone occurred 18% of the days leading to mean daily 8 h concentrations of 73 ± 11 ppbv. Shallow and deep convection associated with moderate to strongly unstable conditions lowered the mean ozone to 50 ± 11 ppbv ( 29% of days), while weaker convection associated with marginally unstable conditions reduced the mean concentrations to 63 ± 13 ppbv ( 11%). We use these observations to derive simple relationships between surface ozone and convective indicators that may prove useful for parameterization of convective venting in air quality models. Citation: Langford, A. O., S. C. Tucker, C. J. Senff, R. M. Banta, W. A. Brewer, R. J. Alvarez II, R. M. Hardesty, B. M. Lerner, and E. J. Williams (2010), Convective venting and surface ozone in Houston during TexAQS 2006, J. Geophys. Res., 115,, doi: /2009jd Introduction [2] Houston, Texas, experiences some of the highest ambient ozone (O 3 ) levels in the United States [Kleinman et al., 2002] with frequent violations of the National Ambient Air Quality Standards (NAAQS) [Environmental Protection Agency (EPA), 2008] during the summer and early fall. Local photochemistry is the primary source for this elevated ozone, as high temperatures and abundant sunlight lead to rapid production of ozone from highly reactive volatile organic compounds (HRVOCs) and nitrogen oxides (NO x ) released by industrial sources along the Houston Ship Channel and Galveston Bay [Daum et al., 2004; Ryerson et al., 2003]. These emissions often create plumes of high ozone that are advected over Houston in the afternoon and early evening by the bay and sea breeze and merge with ozone from urban and 1 Chemical Sciences Division, Earth System Research Laboratory, NOAA, Boulder, Colorado, USA. 2 Cooperative Institute for Research in Environmental Sciences, University of Colorado, Boulder, Colorado, USA. Copyright 2010 by the American Geophysical Union /10/2009JD regional sources. Under certain meteorological conditions, these emissions lead to high ozone episodes that affect the greater Houston Galveston Brazoria (HGB) area. Previous studies have shown that particularly severe episodes occur when large scale northerly or northeasterly flow elevates the regional ozone background [Langford et al., 2009; Nielsen Gammon et al., 2005; Rappenglück et al., 2008] and interacts with the sea breeze to create stagnation events that promote production and accumulation of ozone [Banta et al., 2005; Darby, 2005]. [3] Several high ozone episodes occurred in the Houston area during the late summer and early fall of 2006 when the field intensive portion of the second Texas Air Quality Study (TexAQS 2006) was underway [Parrish et al., 2009]. The most serious took place 31 August to 1 September, when 8 h mean ozone mixing ratios in excess of 120 ppbv were measured on two consecutive days. This episode, which has been described elsewhere [Rappenglück et al., 2008], was in many ways similar to the case study by Banta et al. [2005] from the TexAQS 2000 field campaign. Rappenglück et al. [2008] concluded that high background ozone was a particularly important factor during this epi- 1of18

2 sode, as well as others that occurred later in September. They pointed out that all of these episodes closely followed the passage of cold fronts when the large scale flow was from the north or northeast, bringing air with relatively high ozone to the Houston area. Langford et al. [2009] reached similar conclusions but suggested that suppression of convective mixing by the stable air behind these cold fronts may also have been a key factor since this resulted in greater accumulation of locally produced ozone within the boundary layer. [4] It is well established that shallow and deep convection vent ozone and other pollutants from the mixed layer into the free troposphere [National Research Council, 1992; Cotton et al., 1995]. Deep convection can rapidly transport boundary layer air well into the middle and upper troposphere [Dickerson et al., 1987], not only decreasing concentrations of pollutants near the surface but also increasing the potential for long range transport [Thompson et al., 1994] and ozone production in the middle and upper troposphere [Pickering et al., 1992]. Active fair weather cumulus clouds formed by shallow convection can transport air from the boundary layer into the lower free troposphere and entrain free tropospheric air into the mixed layer [Stull, 1985]. Cloud venting associated with shallow convection has been confirmed by many observational studies including SF 6 tracer release experiments [Ching and Alkezweeny, 1986], in situ aircraft measurements [Greenhut, 1986], and lidar measurements [Ching et al., 1988]. Recent model studies have also examined this process. Angevine [2005] used a 1 D model with a mass flux diffusion scheme to reproduce observed decreases in CO when shallow cumulus clouds were present during case studies at the Great Plains Atmospheric Radiation Measurement (ARM) site and Nashville, TN. Vilà Guerau de Arellano et al. [2005] examined the same ARM case study with large eddy simulation models and found that active fair weather clouds decreased boundary layer ozone by up to 50% compared to clear sky conditions. [5] In this paper, we examine the influence of convective mixing on surface ozone in the Houston area during the summer of We begin with a case study contrasting the high ozone day of 31 August, where there was little or no convective mixing between the boundary layer and free troposphere, with another high ozone day, 17 August, where there was both shallow and deep convective activity. We use radiosonde data from the NOAA R/V Ronald H. Brown and the Lake Charles, LA National Weather Service (NWS) station to characterize the structure of the lower troposphere and assess the probability of convective mixing. Measurements from the downward looking ozone lidar (tunable optical profiler for aerosols and ozone, TOPAZ) aboard the NOAA Twin Otter are used to examine the horizontal and vertical structure of ozone plumes emanating from the ship channel and measurements from the scanning high resolution Doppler lidar (HRDL) aboard the R/V Brown are used to examine the structure of the horizontal and vertical winds. In situ measurements of ozone and related species from the R/V Brown and the network of Continuous Ambient Monitoring Stations (CAMS) maintained by the Texas Commission on Environmental Quality (TCEQ) are used to study the influence of convective venting on surface ozone both during the case study and the rest of the summer of Finally, we compare the daily maximum 8 h ozone concentrations measured in the Houston area during the summers of 2006 and 2000 (Tex- AQS I) with the convective potential inferred from the Lake Charles soundings and contrast the influence of convective venting to that of advective venting by strong southerly flow and the nocturnal low level jet [Tucker et al., 2010]. 2. Case Studies: High Ozone Episodes of 17 and 31 August [6] The high ozone episodes of 17 and 31 August embodied many of the characteristics identified in previous studies [Banta et al., 2005; Rappenglück et al., 2008]. Both episodes began with relatively high background ozone associated with large scale northeasterly or northerly flow and significant emissions of HRVOCs and NO x from sources along the ship channel. The surface winds were similar, changing from light and northerly or northeasterly in the morning to light and easterly by midafternoon and southeasterly by evening as the bay and sea breezes developed in turn. Both days had high peak temperatures, viz C (98 F) and 34.4 C (94 F) on the 17th and 31st, respectively, at Houston Intercontinental Airport. [7] Plume animations from the TCEQ show that urban and industrial emissions of HRVOCs and NO x from the Houston Ship Channel coincided with the highest measured surface ozone concentrations on both the 17 and 31 of August. Figures 1a and 1b show the locations of the approximately 40 regulatory or supplemental Continuous Ambient Monitoring Stations (CAMS) maintained by the TCEQ to monitor ozone and/or other chemical and meteorological parameters in the Houston area during the Tex- AQS 2006 study. These stations recorded ambient surface ozone concentrations every 5 min from which 1 and 8 h averages were calculated to derive the air quality index (AQI) [EPA, 2008]. The solid black symbols show the Houston stations, while the open symbols show stations in Brazoria (circles) and around Galveston Bay (squares). Summaries of these measurements are available on the TCEQ Web site ( monitoring/air/monops/sigevents06.html). [8] The colored contours in Figure 1a show the surface distribution inferred from the 5 min ozone measurements at 2100 UT (1500 CST) on 31 August. Figure 1b is similar, but for 2100 UT on 17 August. Both plots show plumes of high ozone advected westward from the ship channel by the bay breeze. The plumes appear qualitatively different, however, because the sea breeze front had propagated much further inland on 31 August and distorted the ozone plume. [9] The overall air quality in Houston was rated as Unhealthy on 17 August and as Very Unhealthy on 31 August. Eight hour mixing ratios in excess of 75 ppbv ( Unhealthy for Sensitive Groups or Orange ) were measured by 32% (13 of 41) of the reporting CAMS on the 17th and by 40% (15 of 37) of the reporting CAMS stations on 31 August. Two stations on the 17th and seven stations (19%) on the 31st exceeded 95 ppbv ( Unhealthy or Red ). Three stations (8%) reported mixing ratios in the Very Unhealthy or Purple range (>115 ppbv). These 2of18

3 [10] The number of sites reporting elevated ozone on both days was strongly influenced by the regional ozone background. Radar wind profiler based back trajectories [White et al., 2006] (not shown) indicate that air parcels arriving over Houston between 500 and 2500 m above mean sea level (ASL) on the afternoon of 17 August passed over southern Louisiana the previous day, while parcels arriving on 31 August passed over northern Louisiana, Arkansas, and Oklahoma during the previous 24 h. Trajectories at 1000 m ASL calculated using the NOAA ARL (Air Resources Laboratory) HYSPLIT model (available at passed near the Chicago area 48 h previously. This continental flow increased the 8 h mean regional ozone background determined from principal component analysis (PCA) of the CAMS measurements to 66 ppbv on 17 August and to 73 ppbv on 31 August [Langford et al., 2009], already comparable to the 2008 National Ambient Air Quality Standard (NAAQS) design value of 75 ppbv considered Unhealthy for Sensitive Groups. The difference between this background and the highest reported 8 h mean concentrations, a rough indicator of the contribution due to local ozone production, was 53 ppbv. A similar estimate for 17 August gives a local contribution of 47 ppbv, about 12% lower than the estimate for 31 August. The comparable estimates are consistent with the surface meteorology and the fact that both episodes occurred on a Thursday so that industrial activities and traffic patterns were likely to be similar. Figure 1. Map of Houston and surrounding areas showing locations of CAMS maintained by the TCEQ in the Houston area during the TexAQS 2006 study. The solid symbols show Houston stations, and the open symbols show stations in Brazoria to the south (circles) and around Galveston Bay (squares) to the east. The red symbols identify the CAMS stations used to calculate the Houston wide mean. The cross shows the location of the LaPorte Airport, and the open circle the University of Houston. The colored contours show the surface distribution inferred from the 5 min ozone measurements at (a) 2100 UT (1500 CST) on 31 August and (b) 2100 UT on 17 August. The solid (AB and A B ) and dotted (ab and a b ) heavy black lines show locations of the Twin Otter and R/V Brown, respectively. Wind barbs show the contemporary 5 min surface winds. color designations represent the EPA air quality index (AQI) and do not correspond to the 5 min color scale in Figure 1. Four monitoring stations downwind of the ship channel (Bayland Park (C53), Westhollow (C410), Tom Bass (C558), and Park Place (C416)) reported 1 h averages exceeding 124 ppbv on both days, and 1 h average mixing ratios of up to 158 and 147 ppbv were measured in this area on the 17th and 31st, respectively. The corresponding 8 h average ozone concentrations were 113 and 126 ppbv. The latter was the highest 8 h mean ozone reported by the Texaswide CAMS network during all of Meteorological Context [11] The upper level flow over the Gulf Coast was dominated by a persistent ridge over the southeastern United States throughout most of August This high pressure region is prominent as a large red area in Figure 2b, which plots the 700 hpa ( 3200 m ASL) geopotential surfaces and wind vectors from the NOAA National Centers for Environmental Prediction (NCEP) Reanalysis [Kalnay et al., 1996] for 0000 UT on 18 August (1800 central standard time, CST, on 17 August). The anticyclonic flow around the high produced east northeasterly winds over southeast Texas and Louisiana, including both Houston and Lake Charles, the nearest National Weather Service (NWS) radiosonde station 200 km to the east (the rectangular box encloses both Houston and Lake Charles). The regional synoptic pattern shifted during the last few days of August, and the Houston weather during much of September was dominated by a series of upper level troughs and surface cold fronts [Rappenglück et al., 2008]. A cold front associated with a particularly deep trough passed through the Houston area on the 29th. A cutoff low subsequently formed over the Midwestern United States, with a downward sloping tongue of subsiding upper tropospheric (and possibly lower stratospheric) air extending southward on the western flank [Langford et al., 1996]. This cutoff low lies over South Carolina in Figure 2a, which plots the 700 hpa geopotential surfaces and wind vectors at 0000 UT on 1 September (i.e., 1800 CST on 31 August). The cyclonic circulation around the low combined with the anticyclonic flow around a high pressure ridge over west Texas to produce strong northerly flow above east Texas and Louisiana. [12] The most significant difference between the meteorological conditions on 17 and 31 August is manifested in 3of18

4 front to the east of Galveston Bay evolved into an isolated air mass thunderstorm about 40 km in diameter. This westward moving cumulonimbus is directly over Galveston Bay in the 2339 UT GOES 12 image (Figure 4b). The shadows in the satellite image show the cloud top to be 8 12 km tall. Figure 4b also shows widespread thunderstorms to the north of Houston, as well as extensive cloud cover near Lake Charles, LA (LCH), 200 km to the east. In contrast, only a few scattered clouds are visible to the north of Houston at 2339 UT on 31 August (Figure 4a). [13] Radiosondes were launched from the NOAA R/V Ronald H. Brown in the Galveston area at 2300 UT on both 17 and 31 August. Figure 5 plots the profiles of (Figure 5a) virtual potential temperature and relative humidity and (Figure 5b) wind speed and direction on 31 August. The R/ V Brown was located 75 km south of LaPorte near Galveston Island (point a in Figure 1a) when the radiosonde was launched at 2300 UT. Figure 5a shows a well mixed Figure 2. Geopotential surface and wind vectors at 700 hpa ( 3200 m ASL) from the NOAA NCEP Reanalysis for (a) 0000 UT on 1 September (1800 CST on 31 August) and (b) 0000 UT on 18 August (1800 CST on 17 August). The black rectangle encloses the Gulf Coast from Houston to Lake Charles, LA. Figure 3, which shows the 1 km GOES 12 visible images of the Houston area at 2004 UT (1404 CST) on 31 August and 2009 UT (1409 CST) on 17 August. The skies were almost completely clear on 31 August (Figure 3a), whereas widely scattered fair weather cumulus (FWC) clouds indicative of shallow convection were present on 17 August (Figure 3b). The FWC density on 17 August is greatest in the low level convergence zones along the sea breeze front near the coast and above Houston (box) where the sea breeze urban heat island effect enhances convection [Orville et al., 2000; Steiger et al., 2002]. These clouds reduced the integrated solar visible radiation by 15% at both Bayland Park (C53) and at Deer Park (C35) near LaPorte (cross in Figures 1a and 1b). The associated reduction in spatially averaged NO 2 photolysis rates downwind of the ship channel is consistent with the differences in estimated total ozone production noted above. Figure 4 shows images taken nearly 4 h later at 2339 UT (1739 CST) on both days. Most FWC formation had subsided by 2339 UT on 17 August as solar heating diminished, but the cumulus along the sea breeze Figure 3. GOES 12 1 km visible images of the Gulf Coast including Houston and Lake Charles, LA (LCH) from (a) 2004 UT (1404 CST) on 31 August and (b) 2009 UT (1409 CST) on 17 August. The Houston area is enclosed in the white rectangle. 4of18

5 Figure 4. GOES 12 1 km visible images of the Gulf Coast including Houston and Lake Charles, LA (LCH) from (a) 2339 UT (1739 CST) on 31 August and (b) 2339 UT (1739 CST) on 17 August. shallow boundary layer over the Gulf of Mexico, with a deep inversion (gray shaded region) extending from 500 to 1800 m ASL above. The higher layer of dry stable air at 3400 m ASL is due to upper tropospheric (or possibly even lower stratospheric) air descending on the west side of the cutoff low in Figure 2a. The wind profile in Figure 5b shows SSE flow associated with the sea breeze below 500 m ASL and northerly winds with strong vertical shear ( 12 m s 1 km 1 ) throughout most of the inversion layer. Although this shear can generate wave activity and turbulent mixing, the inversion inhibited convective lifting above 500 m ASL and the skies remained clear since the lifting condensation level (LCL) lies near the top of the inversion at 1800 m ASL. The level of free convection (LFC) is near 3400 m ASL. The lifted index (LI), a measure of the stability of the lower and middle troposphere based on the temperature difference between a parcel lifted along the moist adiabat from the LCL to 500 hpa, and the surrounding air is Values of the LI between 3 C and 0 C indicate stable conditions, values between 0 C and 3 C indicate marginally unstable conditions, and values from 3 C to 6 C indicate moderately unstable conditions with potential for deep convection [e.g., DeRubertis, 2006, and references therein]. Daytime convective mixing is weaker above the cool Gulf waters than above land, and radiosondes launched at 1800 and 2100 UT from the University of Houston (open circle in Figure 1) [Rappenglück et al., 2008, Figure 5] showed the top of the mixed layer (and base of the shear zone) inland to be 1700 m ASL. The next University of Houston sounding at 0000 UT followed the arrival of the sea breeze from the Gulf, however, and was similar to the 2300 UT Galveston sounding with the top of the mixed layer at 750 m ASL. [14] The NOAA high resolution Doppler lidar (HRDL) on the R/V Brown also measured vertical and horizontal winds near Houston on both days. HRDL is a 2 mm coherent Doppler lidar that provides line of sight wind velocity estimates by measuring spectral shifts in backscattered radiation caused by atmospheric motion of aerosols [Grund et al., 2000]. The lidar was operated in a scanning mode with regular vertical stare periods to measure the vertical wind component for 5 8 min out of every 15 min. Figure 6a shows HRDL backscatter signal strength (an uncalibrated measure of aerosol surface area) and Figure 6b, the vertical velocity measurements from the afternoon of 31 August made while the ship was cruising 8 km off the coast of Galveston (points b to a in Figure 1a). The colored barbs in Figure 6a show horizontal wind profiles measured between the vertical stare periods. The abrupt decrease in backscatter signal near 2200 m ASL shows weak vertical mixing above this altitude. Figure 6b shows weak updrafts (yellow) extending from the surface to 500 m ASL in agreement with the thermal structure measured by the radiosonde (see Figure 5a). HRDL also shows weak downward motion (<1 m s 1 ) above the mixed layer, consistent with the return flow behind the sea breeze front [Atkinson, 1981]. The persistent layer of low aerosol backscatter within the northerly flow near 1500 m ASL underscores the weak vertical mixing. [15] The measurements from 17 August show a very different vertical structure. The radiosonde was launched over Galveston Bay but drifted southwestward over land with the prevailing northeasterly flow (see Figure 5d). The balloon passed through a 700 m deep inversion above the cooler waters of the bay consistent with marine boundary layer depths measured by Tucker et al. [2009] before reaching land and entering the terrestrial mixed layer (vertical dotted line in Figure 5c). The altitudes of the LCL and LFC are 1150 and 1760 m ASL, respectively, and the location of the LCL at the top of the mixed layer explains the widespread FWC. The relative humidity profile in Figure 5c shows that the radiosonde did not pass through any of the scattered clouds and the water vapor mixing ratio (not shown) was well mixed to above 3500 m ASL. The lifted index is 5.9 C; values below 6 C indicate very unstable conditions and the possibility of intense convection as evidenced by the thunderstorm seen in Figure 4b less than 40 min later. [16] The R/V Brown was near LaPorte in the north end of Galveston Bay between 2100 and 2130 UT on 17 August and en route to the Gulf of Mexico when the radiosonde was launched at 2300 UT (point a in Figure 1b). Figure 7a plots 5of18

6 Figure 5. Profiles of (a) virtual potential temperature and relative humidity and (b) wind speed and direction from a radiosonde launched from the R/V Brown near Galveston at 2300 UT (1700 CST) on 31 August. The gray shading highlights the low lying inversion and shear layer. (c) Virtual potential temperature and relative humidity and (d) wind speed and direction from a radiosonde launched from the R/V Brown in Galveston Bay at 2300 UT (1700 CST) on 17 August. the HRDL horizontal wind profiles (color coded wind barbs) measured between the vertical stare periods and the backscatter intensities measured during the vertical stare periods on 17 August. Figure 7b shows the corresponding vertical velocities. The winds rotated more to the north above the easterly bay breeze in the HRDL profile than was measured by the radiosonde launched 2 h later (see Figure 5d), but otherwise the measured winds are quite similar. The vertical velocities are mostly upward near the surface, with deep updrafts and downdrafts consistent with vigorous convective mixing (a rising thermal can be seen near 2124 UT). The backscatter signal is robust up to about 1200 m ASL then decreases gradually with increasing altitude up to about 2500 m ASL where it falls off rapidly. The gradual decrease in signal suggests a deep entrainment zone extending past the LFC to 2500 m ASL Mixing of High Ozone Plumes [17] The vertical structure of the ozone plumes advected westward from the ship channel was probed by the NOAA downward looking TOPAZ differential absorption lidar (DIAL) flown aboard the NOAA Twin Otter during the TexAQS 2006 field intensive [Alvarez et al., 2006]. TOPAZ measured the vertical distribution of ozone in the boundary layer and lower free troposphere with 90 m vertical range bins (smoothed by a 450 m window) and 600 m horizontal 6of18

7 Figure 6. HRDL measurements of (a) horizontal winds (color coded wind barbs) and vertical backscatter signal strength and (b) vertical velocities from the afternoon of 31 August while the ship was off the coast near Galveston. resolutions (10 s integration) to within 300 m of the surface. TOPAZ also provided uncalibrated aerosol backscatter profiles that can be used to derive mixed layer height [White et al., 1999; Nielsen Gammon et al., 2008]. Intercomparison with in situ measurements during TexAQS 2006 found generally good agreement but did indicate that the TOPAZ measurements have a negative bias at low altitudes when the lidar beams are strongly attenuated. This bias becomes significant (up to 15%) for measurements directly beneath high ozone plumes. [18] The Twin Otter flew missions in the Houston area on both 17 and 31 August. Ozone measurements were made from 1922 to 2321 UT on the 17th and from 1759 to 2238 UT on the 31st. Figure 8a shows a time height curtain plot of the ozone mixing ratios measured by TOPAZ on 31 August from an altitude of 3420 m ASL along the north to south flight leg represented by the solid black line (AB) in Figure 1a. The flight path crosses the ozone plume emanating from the ship channel slightly east of the surface maximum in Figure 1a and less than 30 minutes after the time represented by the surface contours. The width of the plume at the surface from Figure 1a (as defined by the 125 ppbv surface contours) is indicated by the horizontal bar at the bottom of the plot. The dash dotted lines show the LCL and LFC from the 2300 UT sounding plotted in Figure 5a. Figure 8b is similar to Figure 8a but shows measurements from 17 August. The two plumes also appear qualitatively different in vertical cross section Capped Boundary Layer [19] The stable capping inversion seen in the radiosonde and HRDL profiles on 31 August limits vertical mixing between the mixed layer and the free troposphere. The dotted lines in Figure 8a define the entrainment zone (EZ) determined from the 5% and 95% points in the transition from peak ozone to free tropospheric air [Cohn and Angevine, 2000; Nelson et al., 1989] along the solid vertical line near the center of the ozone plume. The bottom of the EZ defined from the ozone DIAL measurements is in good agreement with the LCL from the 2300 UT sounding. This boundary is also seen to roughly correspond to the 80 ppbv ozone contours to the north and south of the main plume. The EZ is 400 m thick or only 20% of the mean mixed layer height. This value is near the limit that would be expected when the boundary layer is shear dominated [Gryning and Batchvarova, 1994]. Nearly identical results are obtained when the higher resolution aerosol backscatter 7of18

8 Figure 7. HRDL measurements of (a) horizontal winds (color coded wind barbs) and vertical backscatter signal strength and (b) vertical velocities from the afternoon of 17 August while the ship was near LaPorte, TX. profiles are used. These observations are consistent with recent model simulations [Dacre et al., 2007], which also show the limited role of shallow convection as a mechanism for ventilating the boundary layer in a coastal environment when there is a capping layer. [20] The ozone vertical profile (averaged over 100 s or 6 km) from the TOPAZ measurements near the center of the plume (i.e., along the solid vertical line shown in Figure 8a) is represented by the solid black circles in Figure 9a. These profiles have been corrected for the bias noted above. The dotted line shows the uncorrected profiles, which are also used in Figure 8. The error bars represent the standard deviation of the averaged profiles, which includes both measurement uncertainties and the horizontal variability of ozone over the 6 km average. The measurement uncertainties were similar on both days; thus, the horizontal variability below 1200 m ASL appears to be much greater on 31 August than on 17 August. The vertical dashed black line extrapolates the mean ozone between 500 and 1000 m ASL to the surface. The blue and red profiles were measured 20 km to the north of the plume center (left dashed vertical line in Figure 8a) and 30 km to the south (right dashed vertical line in Figure 8a). The blue and red dashed lines extrapolate these profiles to the surface. The vertical dotted black line shows the regional ozone background [Langford et al., 2009]. [21] Figure 9a also plots 30 min averaged wind barbs centered at 2030 UT on 31 August from the 915 MHz wind profiler operated by the TCEQ at the LaPorte Airport (cross in Figure 1a). The wind profile is similar to that seen in the HRDL and radiosonde profiles described above with strong northerly flow above the LCL ( 1800 m ASL) and both directional and velocity shear beneath. The Gulf sea breeze front is evident as the shallow layer of southeasterly winds below 400 m ASL [Banta et al., 2005]. The solid circles in Figure 10a transect the plume through the ozone maximum just below the LCL. These points correspond to the solid horizontal line in Figure 8a. Figure 10a also shows transects above the EZ near 2500 m ASL in the free troposphere ( plus ) and near the top of the EZ (open circles). The former corresponds to the dashed line in Figure 8a and shows a nearly constant mixing ratio of 68 ± 5 ppbv that differs from the regional 8 h background value of 73 ppbv [Langford et al., 2009] by less than 10%. The transect near the top of the EZ shows three narrow spikes of high ozone separated by about 8 km. These spikes are attributed to 8of18

9 Figure 8. Time height curtain plot of the ozone mixing ratios measured by TOPAZ along the appropriate transect from Figure 1 on (a) 31 August and (b) 17 August. The meanings of the various lines are described in the text and in the captions for Figures 9 and 10. The bars along the bottom show the width of the 125 ppbv contours defining the surface plumes in Figure 1. gravity waves created. The vertical transport and mixing associated with this process will be described in more detail in a subsequent publication. [22] The curvature in the black profile plotted in Figure 9a suggests that the convective boundary layer extended only up to 1000 m ASL at the plume center. The blue and red profiles show similar mixed layer depths 20 km to the north and 30 km to the south, respectively. These values are much lower than the 1700 m ASL mixing height measured by the 2100 UT University of Houston radiosonde only 25 km to the northeast of the plume center [Rappenglück et al., 2008, Figure 5]. However, a TOPAZ ozone profile measured over the University of Houston at 2113 UT (green line) does appear to be well mixed up to 1700 m ASL. These differences reflect the spatial variability of the mixed layer depth around Houston and other coastal areas [Nielsen Gammon et al., 2008]. [23] Figures 8a and 9a show the bulk of the ozone plume on 31 August lay between the top of the convectively mixed layer and the upper stable layer. This can be explained by the strong low level convergence along the Gulf breeze front [Atkinson, 1981; Banta et al., 2005], which appears to have lifted the plume above the convectively mixed layer. The plume was then spread transversely by turbulent mixing within the shear layer and stretched to the south by the northwesterly winds. Thus, the plume is much broader aloft than near the surface and the 120 ppbv DIAL contours in Figure 8a nearly 45 km wide near the LCL Shallow Convection [24] The partial gaps in the TOPAZ measurements on 17 August (see Figure 8b) are caused by scattered FWC clouds that block the downward looking lidar (see Figure 3b). The tops of these active clouds protrude above the LFC, venting ozone, and other pollutants from the mixed layer into the lower free troposphere and entraining cleaner free tropospheric air into the mixed layer [Stull, 1985]. The dotted lines in Figure 8b define the entrainment zone (EZ) determined as before along the solid vertical line near the center of the plume. The vertical profile along this line (averaged over 100 s or 6 km) is represented by the solid black circles in Figure 9b. The bottom of the EZ defined in this manner is in good agreement with the LCL from the 2300 UT sounding. This boundary is also seen to roughly correspond to the 90 ppbv ozone contours (green) to the south of the main plume (i.e., to the right in Figure 8b). The EZ thickness is 850 m or nearly 50% of the mixed layer height showing vigorous mixing between the boundary layer and the free troposphere. [25] The solid circles in Figure 10b transect the plume just below the LCL through the ozone maximum. These points correspond to the solid horizontal line in Figure 8b. As before, Figure 10b also shows transects above the EZ near 2500 m ASL in the free troposphere ( plus ) and near the top of the EZ (open circles). The former corresponds to the dashed horizontal line in Figure 8b and shows a nearly constant mixing ratio of 62 ± 3 ppbv that also differs from the regional 8 h background value of 66 ppbv [Langford et al., 2009] by less than 10%. The transect near the top of the EZ shows higher ozone above the more vigorous thermals south of downtown Houston ( 29.6 N) and the sea breeze convergence zone ( N) (see Figure 3b). [26] Figure 9b also plots 30 min averaged wind barbs centered at 2130 UT from the 915 MHz wind profiler operated by the TCEQ at the LaPorte Airport (cross in Figure 1b). The profiler shows that the winds above the surface layer were mostly from the east northeast and less than 5 m s 1, similar to the radiosonde profile 2 h later. However, unlike the balloon sounding, the winds at LaPorte became almost northerly just below the LCL, before rotating back to the east northeast at higher altitudes. This profile is very similar to that measured by HRDL (see Figure 7a) at about the same time and at very nearly the same location. This localized northerly wind is attributed to the inflow created by the convective updrafts above Houston. [27] The strong vertical mixing and relatively weak winds on 17 August limit horizontal dispersion of the ozone plume. TCEQ animations show the surface plume in Figure 1b retained the appearance of a bulls eye with very little spreading as it moved westward. The highest mixing ratios found by TOPAZ, 200 ppbv in Figure 10b, were measured almost directly above the surface maximum in Figure 1b, and the dark orange ( ppbv) contours 9of18

10 Figure 9. (a). TOPAZ vertical ozone profiles (averaged over 100 s or 6 km) on 31 August along the three vertical lines in Figure 8a. The solid black circles correspond to the solid line through the center of the plume. The error bars show the standard deviations of the hundreds of averages and include both measurement uncertainties and horizontal variability. The blue and red solid circles correspond to the left (north) and right (south) dashed lines in Figure 8a. The dashed lines in all three profiles extrapolate the m ASL mean value to the surface. The green profile was measured over the University of Houston at 2113 UT; symbols and error bars are omitted for clarity. The barbs show the 30 min averaged winds (centered at 2030 UT) from the 915 MHz wind profiler operated by the TCEQ at the LaPorte Airport. (b) Same as Figure 9a but for 17 August (2130 UT wind barbs). in Figure 1b coincide with the red (>160 ppbv) contours in Figure 8b. The main body of the plume aloft, as defined by the (yellow) 120 ppbv contours in Figure 8b, is less than 30 km wide and changes little with altitude. Vertical profiles measured 25 km south of the plume center (right dashed vertical line in Figure 8b and solid red circles in Figure 9b) and 25 km north (left dashed vertical line in Figure 8b and solid blue squares in Figure 9b) show mixing ratios near the background of 66 ppbv at all altitudes above 800 m ASL. The Brazoria profile also shows a layer of elevated at the top of the entrainment zone that was likely mixed upward from the high ozone plume and advected to the southwest. [28] Despite the qualitative differences between the two plume structures plotted in Figures 8a and 8b, the total ozone obtained when the profiles and transects in Figures 9 and 10 are integrated differs by less than 5% between 17 and 31 August when the appropriate regional background is subtracted. The TOPAZ measurements thus also show that the total ozone within the two plumes was similar, in agreement with the other estimates of the relative production given above Deep Convection [29] The impending arrival of the developing thunderstorm over Galveston Bay on the afternoon of the 17th forced the Twin Otter to suspend ozone DIAL measurements at 2321 UT. The aircraft was located 50 km west of Galveston Bay at this time, so there is no direct information about the ozone distribution aloft in the vicinity of the thunderstorm. However, surface measurements of ozone and other chemical and aerosol species [Parrish et al., 2009, and references therein] were made by in situ instruments aboard the R/V Ronald H. Brown as the thunderstorm passed overhead. The Brown was traversing Galveston Bay en route to the Gulf of Mexico at this time (point a to point b in Figure 1b). The top two frames of Figure 11 again show (1) signal strength (db) and (2) ship motion corrected vertical velocities measured by HRDL between 1700 and 1825 CST on the evening of 17 August (2300 UT on 17 August to 0025 UT on 18 August). The time series ends 30 min before local sundown. The bottom two frames plot H 2 O, O 3, and CO mixing ratios and wind speed and direction measured over the same interval by the in situ instruments aboard the ship. The surface winds associated with the sea breeze were light (<1 m s 1 ) and from the southeast at 2300 UT, and the HRDL vertical profiles between 2303 and 2324 UT were similar to those measured 90 min earlier (Figure 7) with alternating regions of upward and downward motion associated with the weakening thermals. The surface wind rotated from the east to the northwest as surface air was entrained by the growing storm, and the HRDL vertical wind profiles at 2335 UT in Figure 11b show the organized updraft associated with this inflow. The surface winds then shifted abruptly to the southeast and increased to 8ms 1 as the gust front [Darby et al., 2002] associated with the following downdraft reached the ship at 2339 UT. Figure 4b shows the thunderstorm directly over Galveston Bay at this 10 of 18

11 parameter (a measure of the humidity dependence of light scattering) increased from 0.56 to The latter value is typical of clean background marine air and the low POM fraction shows that the aerosols sampled after 0000 UT were well aged, while those sampled before the storm were relatively fresh [Quinn et al., 2005]. Figure 10. (a) Transects corresponding to the solid and dashed horizontal lines in Figure 8a for 31 August. The horizontal dashed line shows the mean ozone mixing ratio measured above the plume in the lower free troposphere. (b) Same as Figure 10a but for 17 August and Figure 8b. The letters A and B show the ends of the Twin Otter flight legs. time. The gust front was followed by a brief period of rain at 2345 UT, and the winds shifted to the northwest and increased as the wake of descending free tropospheric air from the receding thunderstorm surrounded the ship. [30] The third panel from the top in Figure 11 shows that the measured H 2 O, O 3, and CO mixing ratios were relatively constant at the beginning of the time series (data contaminated by local ship plumes have been removed from the plots). All three abruptly decreased by 15% 20% with the arrival of the gust front. Ozone decreased from 81 to 66 ppbv (dashed lines), very near the values measured by TOPAZ above the boundary layer and the regional background estimated from the CAMS measurements. Carbon monoxide (CO) decreased from 150 to 125 ppbv. Massoli et al. [2009] noted that instrumentation aboard the R/V Brown also detected a qualitative change in aerosol properties at this time. The submicron particle organic mass (POM) fraction decreased from 0.54 to 0.34, and the g 2.3. Impact of Convective Venting on Surface Ozone [31] The TOPAZ measurements described above show that less ozone was transported from the surface into the free troposphere on 31 August, when shallow convection was weak, than on 17 August, when shallow convection was strong. The combination of weak vertical transport and strong vertical wind shear on the 31st meant that ozone from the plume affected a larger area for a longer period than on the 17th, as suggested by the relative areas of the surface ozone plumes in Figure 1. Figure 12a plots the 1 h mean ozone mixing ratios on both the 17th and 31st, averaged over all of the 30+ Houston CAMS (solid symbols in Figures 1a and 1b). Although the mean concentrations increase more slowly on the morning of 17 August than on 31 August, the peak concentrations are almost identical, and the integrated ozone on the 17th is 12% lower than that on the 31st in agreement with the other estimates of relative ozone production given above. Figure 12b is similar to Figure 12a but plots the mean mixing ratios measured at the four CAMS station in the vicinity of the ship channel (C35, C552, C572, and C1015) represented by the solid squares in Figures 1a and 1b. Figure 12c plots the mean of the four CAMS stations in West Houston (C410, C554, C559, and C562) represented by the solid triangles in Figures 1a and 1b. The integrated ozone near the ship channel was actually 10% higher on the 17th than on the 31st, while the integrated ozone in SW Brazoria and West Houston was 45% and 35% lower, respectively. Furthermore, the ratios of total NO x, methane, nonmethane hydrocarbons, and net solar radiation measured between midnight and sunset ( 1800 CST) on the 17th and 31st at Lynchburg Ferry (C1015) (see Figure 1b), one of the ship channel stations, were 1.25, 1.01, 1.23, and 1.00, respectively. This suggests that photochemical production of ozone under the clear skies near Galveston Bay may have been, if anything, slightly larger on the 17th than on the 31st. Although the mean photochemical production of ozone was probably slightly less on the 17th because of the scattered clouds [Vilà Guerau de Arellano et al., 2005], the larger relative decrease in ozone as the plume was advected westward on the 17th is more likely due to venting of the mixed layer by the active shallow convection that occurred throughout the day. Other loss processes (i.e., titration by locally emitted NO or surface deposition) would have decreased surface ozone more on the 31st than on the 17th due to the shallow convective boundary layer. Conversely, the more rapid rise in ozone on the morning of the 31st is consistent with the weak vertical mixing on that day. [32] Another consequence of the weak vertical transport and strong vertical wind shear in the northerly winds on 31 August is that much of the ozone and other pollutants produced on the 31st was advected out over the Gulf and recirculated inland to contribute to the poor air quality on 1 September [Rappenglück et al., 2008]. This southerly transport is seen not only in the TOPAZ measurements (see 11 of 18

12 Figure 11. (top) Signal strength (db) and ship motion corrected vertical velocities measured by HRDL between 1700 and 1825 CST on the evening of 17 August (2300 UT on 17 August to 0025 UT on 18 August). The time series ends 30 min before local sundown. (bottom) H 2 O, O 3, and CO mixing ratios and wind speed and direction measured over the same interval by instruments aboard the ship. Figure 8a) but also in the surface measurements from Danciger (C618) and Lake Jackson (C1016), the two southernmost CAMS in Brazoria (see Figure 1). The average of the hourly ozone mixing ratios from these two sites for both days is plotted in Figure 12d. The qualitative behavior is similar to that seen to the north in west Houston (Figure 12c), with 50% less integrated ozone at the surface in Brazoria on the 17th than on the 31st, consistent with the differences in the southernmost (red) TOPAZ ozone profiles from Figure 9. [33] Finally, Figure 12e plots the mean hourly ozone mixing ratios from Seabrook Friendship Park (C45) and 12 of 18

13 CST when the hourly mean mixing ratio on 17 August decreased to 50% of that measured in the ship channel. The concentrations slowly returned to previous levels over the next few hours. Local NO x measurements suggest that these changes are not caused by local surface chemistry. However, these changes are consistent with an influx of free tropospheric air in the thunderstorm downdrafts as seen aboard the R/V Brown or an influx of cleaner maritime air from the Gulf of Mexico in the thunderstorm inflow. Figure 12. Diurnal variation of the 1 h mean ozone mixing ratios on 17 and 31 August averaged over (a) all of the Houston CAMS (solid black symbols in Figures 1a and 1b), (b) four CAMS station in the vicinity of the ship channel (C35, C552, C572, and C1015) represented by solid squares in Figure 1, (c) four CAMS stations in West Houston (C410, C554, C559, and C562) represented by the solid triangles in Figure 1, (d) two CAMS stations in southwest Brazoria (C618 and C1016), and (e) two stations on the west side of Galveston Bay (C45 and C620). The error bars show the standard deviations of the means. Texas City (C620), located 10 and 25 km south of the ship channel, respectively, on the western shore of Galveston Bay. The diurnal variation of the mean mixing ratios at these sites is similar to that seen in the ship channel on both days, with concentrations 20% 25% lower than those in the ship channel at local noon and 10% lower than those in the ship channel between 1600 and 1900 CST on both days. A much more significant difference is seen between 1900 and Convective Potential and Surface Air Quality in Houston [34] The TOPAZ measurements from 17 August show the impact of shallow convective venting on plumes of ozone formed from HRVOC and NO x emissions in the Houston area. Much of the ozone from the ship channel plume was mixed into the free troposphere, and although high surface concentrations still occurred, they affected a limited area for a relatively short time period compared to 31 August. Localized deep convection further reduced surface concentrations of O 3, CO, and other gases and aerosols around Galveston Bay in the late afternoon. The measurements from 31 August show the consequences when the plume is introduced into a more stable mixed layer. The pollutants remain trapped at lower altitudes where they affect a wider area, and may be recirculated over the Gulf of Mexico and back by the diurnally varying land sea breeze to adversely affect air quality the following day. The broader impact of convective venting on surface air quality in Houston and the relative importance of shallow and deep convection is not necessarily obvious from measurements made on only 2 days. However, cloud venting should be an important process in the Houston area throughout the summer when the sea breeze urban heat island effect enhances both shallow convection and thunderstorm activity [Orville et al., 2000; Steiger et al., 2002] Comparison of Galveston and Lake Charles Soundings [35] Radiosondes were launched in the Houston and Galveston area only during the TexAQS 2006 study and thus are not generally available for a long term analysis of the relationship between convective potential and surface ozone. In order to assess the influence of convective venting on other days when a less complete suite of measurements is available, we compare the daily ozone measurements from the Houston CAMS network with the twice daily radiosondes from Lake Charles, LA. Lake Charles lies 200 km east of Houston, a similar distance from the Gulf of Mexico and near a large body of water. The synoptic upper air plots from Figure 2 and the GOES images from Figures 3 and 4 suggest that the mesoscale and large scale influences will often be similar. We can test this assumption for the 2 days of the case study by comparing the Galveston radiosondes from Figure 5 with the afternoon Lake Charles soundings. [36] Figure 13a plots the lowest 3500 m (i.e., hpa) of the virtual potential temperature and relative humidity profiles from the Lake Charles sounding for 0000 UT on 1 September (1800 CST on 31 August). Figure 13b plots the corresponding wind speed and direction profiles. These profiles are seen to be similar to those measured over Gal- 13 of 18

14 Figure 13. Lower tropospheric profiles of (a) virtual potential temperature and relative humidity and (b) wind speed and direction from the Lake Charles, LA, sounding for 0000 UT on 1 September (1800 CST on 31 August). (c and d) Same as Figures 13a and 13b but for 0000 UT on 18 August (1800 CST). veston Bay 1 h earlier (see Figures 5a and 5b). The layer of high static stability and strong wind shear (gray shading) is clearly seen but is much narrower and higher than in the Galveston sounding. The high static stability and strong shear together produce a bulk Richardson number of 0.29, only slightly larger than the value of 0.25 generally associated with wave activity and turbulent mixing [e.g., Holton, 1992]. The stable layer at 3400 m ASL is essentially the same as in the Galveston sounding. The LCL is also at a similar height (832 hpa or 1850 m ASL), but the LFC is higher (395 hpa or 7600 m ASL). The middle and upper troposphere is extremely dry and the total precipitable water in the sounding is only 30.4 mm. The lifted index is +3.0, similar to the +2.2 determined from the Galveston sounding. [37] The lifted index is usually correlated with the convective available potential energy (CAPE), also known as the buoyancy energy, calculated by integrating vertically the local buoyancy of a parcel from the level of free convection (LFC) to the equilibrium level (EL). CAPE is a measure of the maximum possible kinetic energy a statically unstable parcel can acquire [Holton, 1992], and the square root gives the maximum possible updraft vertical velocities if thunderstorms were to form. The CAPE estimated from the 31 August/1 September Lake Charles 14 of 18