INVESTIGATE WILDFIRE IMPACTS ON OZONE PRODUCTION BY VERTICAL OBSERVATIONS AND PHOTOCHEMICAL MODELING

In troposphere, ozone is a toxic secondary pollutant produced when its precursors react in sunlight. An important source of ozone precursors is biomass burning. Here we investigate the impacts of 2016 Southeast U.S. Wildfires on ozone production by integrating vertical resolved ozone profiles and photochemical modeling. The results show that wildfires contributed to ozone lamina at the top of boundary layer and enhanced surface ozone up to about 10ppbv in Southeast U.S.. Ozone lidar observed a lower ozone change with respect to a fast growth of aerosol plume, of which the reason is also investigated. Current results indicate an effective integration of vertical observations and modeling for us to understand the ozone production from fires in troposphere.


INTRODUCTION
Tropospheric ozone (O3) plays an important role in climate and affects air quality. It is transported from stratosphere or produced by photochemical reactions of its precursor gases that are emitted by multiple sources (e.g. vegetation, lightning, fossil fuel combustion, biomass burning (BB)) [1]. It is fairly well established that BB can release substantial O3 precursors (nitrogen oxides, carbon monoxide, and hydrocarbons) that react in the presence of sunlight [2], and then affect regional air quality or far downwind by long-range transport [3,4]. However, understanding the O3 formation are complicated by fire emissions, photochemistry, meteorological patterns, and the mixing with extra airmasses [5]. Here we report a case study of wildfires. The main goal is to investigate the wildfire impacts on ozone production. Specific objectives include: (1) Identify and quantify fire impacts on vertical ozone profiles; (2) Estimate the potential impacts on surface ozone when smoke interacts with boundary layer; (3) Explore the dynamical and chemical driver for the observed ozone variations in smoke plumes.

METHODOLOGY
Here we use both ECC ozonesonde [6] and ground-based ozone lidar [7] at Huntsville station (34.725N, 86.645W) to obtain vertical resolved ozone profiles. The ozonesonde has vertical resolution of 100-150m, with precision better than 5% and an accuracy better than 10% for O3 measurements. O3 DIfferential Absorption Lidar (DIAL) determines O3 profile by analyzing how rapidly the backscattered signals at the two UV wavelengths diminish with altitude [8]. The Rocket-city O3 Quality Evaluation in the Troposphere (RO3QET) lidar at Huntsville can retrieve O3 between 0.1-12km with 2min temporal resolution, with 150m vertical resolution and 10% precision in the lower troposphere [9]. As the retrieved O3 number density is affected by large amounts of aerosol, we use the aerosolcorrection O3 in this study.
Vertical aerosol from ceilometer and DIAL as well as MODIS reflectance imagery are combined to identify the smoke plumes. CT25K ceilometer, in the University of Alabama in Huntsville (UAH), is a low-power pulsed lidar at high spatiotemporal resolution [10]. The backscatter signal consists of both molecular and aerosol components. We use the total backscatter intensity curtain as an indicator of aerosol structure. The aerosol extinction profiles have been retrieved from RO3QET at 299nm with an iterative DIAL algorithm [11]. We have applied the newly retrieved DIAL 299nm aerosol extinction here.
We use a fully coupled meteorology-chemistry model, the Weather Research and Forecasting with Chemistry (WRF-Chem) model [12]version 3.9.1, in this study. Gas and aerosols are calculated with the MOZCART chemical mechanism [13]. Other selected parameterizations include Morrison's microphysics scheme, the Rapid Radiative Transfer Model longwave and Goddard shortwave radiation schemes, the Re Monin-Obukhov surface layer, the Noah Land Surface Model, the YSU PBL, the G3 Cumulus clouds, and the Madronich F-TUV photolysis. The initial and lateral conditions are provided by NCEP NAM 12km Analysis data for meteorology and MOZART-4 global model [14] outputs for chemistry. Fire emissions are calculated from the Fire Inventory from NCAR (FINN) [15] with an online plume rise model. Anthropogenic emissions from NEI-2011 and biogenic emissions calculated online by MEGAN module are also included. We performed the simulations at 16km and 4km horizontal resolution for two domains. We use the second domain output for analysis. The vertical configurations include 60 unequally spaced layers below 50hPa, with 12 layers below 2km altitude. The model performance has been evaluated by surface data, vertical observations, and aerosol optical depth (not shown here).

RESULTS
The suggests that wildfires continued to burn and emit a significant amount of smoke over southeast. Figure 1a shows the continuous aerosol structure below 4km above ground level (AGL) indicated by UAH ceilometer backscatter in Nov 12-14. Several aerosol layers travelling over Huntsville, mostly in residual layer during nighttime. While an elevated aerosol plume around 2km persisted for nearly 24 hours, from Nov 12 12UTC to Nov 13 12UTC. Wildfires are supposed to affect this aerosol layer, as MODIS reflectance product shows overhead smoke plumes over Huntsville.

Vertical observations analysis
The ozonesonde was also launched at UAH campus at 19UTC on Nov 12 (local time 1pm). Figure 1b shows the profiles of O3 VMR, RH, θ, as well as backscatter intensity from ceilometer at the sonde launch time. Upon the heating surface layer, θ is uniform across the convective boundary layer (CBL), indicating a high degree of vertical mixing. This corresponds to the uniform aerosol and O3 distribution across CBL. Above that, two inversion layers at 1.4-1.5km and 2.3-2.4km AGL, respectively, can be identified with clear knees in the θ profile. Above 1.5km is free troposphere (FT). The upper inversion layer is likely caused by wind shear. Between the two inversion layers, enhanced (+12ppbv) O3 lamina coexisted with that elevated aerosol plume. O3 VMR increased from 43.8ppbv (on average in CBL) up to a peak of 55.6ppbv at 1.8km AGL. This coexistence of fire-impacted aerosol plume and enhanced O3 indicates a fire-impacted ozone lamina in FT. To justify the hypothesis, we use sensitivity test by WRF-Chem to quantify the fire impacts later. In addition, this thick ozone lamina (nearly 1km thickness) was below a stratospheric intrusion layer, suggested by the extremely dry air (RH<5%) at 2.3-3.0km. As the fire-impacted O3 lamina is wet (RH>50%), we could not claim any contribution from the stratospheric O3 to this O3 lamina through one sounding profile.
The most severe particulate air pollution occurred in Nov 13-14, indicated by the highest backscatter signal in ceilometer (Figure 1a). We carried out 3h DIAL measurement to observe the vertical O3 and aerosol distribution under such a high aerosol loading condition, which covers the heavy aerosol subsidence process at nighttime. Figure 1c shows the time-height curtain of aerosol extinction coefficient retrieved at 299nm and O3 after aerosol correction. The aerosol extinction below 2km is above 0.25km -1 , indicating heavy aerosol loading below the capping inversion layer. After 2UTC, a plume with over 1.0km -1 extinction presents at 1km. It gradually subsides during transport in residual layer. At 4UTC, extinction across the residual layer column has exceeded 1.0km -1 , with a thicker layer between two finer layers. This specified fine structure is highly consistent with that depicted by ceilometer. There are two obvious O3 laminae illustrated by DIAL O3 curtain, with O3 VMR around 50-70ppbv. The 4km O3 lamina on Nov 13 nighttime was elevated above the heavy aerosol column, indicating it is dominated by other O3 source (e.g. STT) rather than fire emissions. Below that, another O3 lamina was aloft around 2km--same altitude as on Nov 12--corresponding to about 0.2-0.6km -1 aerosol extinction. However, we notice a lower O3 change with respect to the highest aerosol extinction over 1.0km -1 . It is unclear for the source of the ~2km O3 lamina and the O3 suppression within the heavy aerosol plume at night. To interpret those features in Nov 12-14, we use photochemical simulations to quantify the contribution of fire or extra sources to the observed ozone formation. Besides the local impacts, the impacts on a regional scale and surface ozone are also explored.  Figure 2 shows the modeled vertical sections of O3 and increased O3 from fire along the latitude of Huntsville station from west to east over Southeast U.S. in Nov 12-14 at 19UTC. The curtain on Nov 12 shows that the enhanced O3 lamina at 2km spread widely across 10 longitude (from 78W to 88W). This widely spreading layer is carried in light or calm wind field. As discussed before, this thick layer is aloft between two inversion layers: the bottom one is PBL entrainment zone and the upper one is likely formed by wind shear. The wind field simulation in Figure 2a and weather map (not shown here) confirms the formation of that wind shear, when a large-scale anticyclone circulation of winds around a high pressure blowing westward below 2km, while the jet stream above 3km (concurrent with stratospheric intrusion O3) blowing eastward. This structure is a key for capping the plume with enhanced O3 above PBL height. On Nov 13, the stratification feature only persisted over west of 82W, consistent with DIAL observation at Huntsville station (86.645W), when the wind shear structure vanished at east coast. On Nov 14, as the wind field below 5km is dominated by westerly wind, there is no obvious laminae structure presented over southeast. Figure 2b shows the contribution from fire to the ozone lamina from a vertical viewpoint. The result shows that up to 5-6ppbv O3 enhancement across 10 longitude is caused by fire emissions. The fire impacts on Nov 12 daytime across Huntsville latitude present a stratification feature, with obvious smoke-induced O3 lamina above PBL but little impacts in PBL. It also suggests that the wildfires take responsibility for half of the ozone enhancement over Huntsville on Nov 12. It is worth discussing the extra contribution from stratospheric intrusion or anthropogenic emissions sources in the ongoing studies. Figure 3 shows the fire smoke in MODIS and the modeled fire impacts on surface O3 in Nov 12-14 19UTC, respectively. On Nov 12, different from the elevated 2km O3 enhancement near Huntsville (Figure 2b), the surface O3 enhancement over 10ppbv is shown at southern AL and northern FL (Figure 3d), downwind of the fire locations (Figure 3a). The lower O3 production near fires also explains the low CBL ozone observed over Huntsville. In Nov 13-14, when high pressure circulation accumulated smoke around fire sources (Figure 3b, c), model results show more concentrated fire impacts on ozone production around the Southern Appalachians (Figure 3e, f). As O3 production from fires is lower on Nov 13 (Figure 3e), and the photochemical reaction is inactive at nighttime during DIAL measurement, a lack of ozone source likely caused the observed low ozone change during heavy aerosol loading at night. In addition, another sensitivity test with anthropogenic emissions on and off (not shown here) indicates NOx titration effect around urban and sub-urban area (Huntsville). Mixing of fire and anthropogenic pollution during the high pressure circulation needs to be studied.

CONCLUSIONS
Due to the multitude of tropospheric ozone sources and a lack of coincident measurements, coordinating observations and modeling are crucial for understanding O3 production from fires. The vertical profiles and surface stations are essential to inform model, regional photochemical model will be good complementary to stationary and surface observations by its spatial coverage. It is critical to identify and exclude other O3 sources for fire-impacts estimation in future studies.