Figure 1. Study area map with FLEXPART-WRF model domain. Rectangles mark the WRF domains D1 which has a horizontal resolution of 12 km and D2 (4 km resolution). SEACIONS locations are marked by black circles (St. Louis, Missouri station (STL) is a gray circle). Wildfire emissions from FLAMBE, August-September 2013, with eleven Pyro convective and five high-altitude injection elevated smoke plume activity areas are marked, red crosses indicate pyroCb plumes which transport smoke into the upper troposphere and lower stratosphere. Blue crosses indicate high-altitude injection of smoke in the absence of large pyroCb which transport smoke into the mid-troposphere).
Figure 1 displays the seven ozonesonde sites across the U.S. for the South East American Consortium for Intensive Ozone Network Study (SEACIONS) — archived at http://croc.gsfc.nasa.gov/seacions/. Here, we focus on the sole midwestern U.S. site located at the James S. McDonnell Planetarium in Forest Park (90.27W, 38.63N, 181 m ASL), ∼5 km west of downtown St. Louis, Missouri. Balloon soundings carrying an ozonesonde and a radiosonde were made near daily between 10:00 to 14:00 (UTC - 5 h). The exact launch time depended on clearance from the Air Traffic Control of the local airport and weather conditions. The first flight was made on 8 August 2013 and the last on 22 September 2013. There were breaks scheduled between the launches in order to cover the time period of the larger SEAC4RS flight mission (Toon et al., 2016). A total of 28 ozonesondes were launched during this period from St. Louis. The balloons reached a maximum altitude of 22 to 36 km with an average ascent rate of about ∼5 ms-1. The instruments fell mostly within 70 km of the launch site.
Vertical curtain plots of ozone concentrations are shown in Figure 2, displayed in 500 m bins up to 15 km height. The figure shows structures in the tropospheric ozone profile which will be analyzed for impacts of stratospheric air masses and biomass burning sources using the modeling techniques demonstrated by Thompson et al. (2007, 2011b).
3 Numerical Models
3.1. Stratospheric-to-Tropospheric Transport ozone tracer model
To simulate the stratospheric air masses entering the troposphere during the SEACIONS field campaign, the NASA Goddard Earth Observing System model (GEOS-5, Rienecker et al., 2008) was used. The model analysis was provided on 72 levels from the surface to 0.01 hPa (∼75 km), every 6 hours. Initial conditions for the atmospheric component were taken from uncoupled experiments forced by the observed sea surface temperature. The model defines a stratospheric intrusion as air masses with a high ozone composition, potential vorticity greater than 1.5 potential vorticity units (PVU; 1 PVU = 10-6 m2 s-1 K kg-1), and a relative humidity (RH) of less than 20% (e.g. Holton et al., 1995; Waugh and Funatsu, 2003; Stohl, 2003). Here we categorize each STT by temporal duration, vertical extent, and meteorological cause and relate them to the contributions to the tropospheric ozone column. Previously, GEOS-5 has been used extensively for stratospheric intrusions and has been validated with ozonesonde observations (e.g. Stajner et al., 2008; Wargan et al., 2015). A review of data assimilation methodology applied to chemical constituents, including ozone, can be found in Lahoz et al. (2007).
3.2. Dispersion Model
3.2.1. FLEXPART-WRF Lagrangian Dispersion Model
The FLEXPART-WRF version 3.1 (Brioude et al., 2013) was used as a Lagrangian particle dispersion model using wind fields from the Weather Research and Forecasting (WRF version 3.4, Skamarock et al., 2008). Simulations were made for seven days to relate ozonesonde profiles to biomass burning events. FLEXPART-WRF was used to simulate carbon monoxide (CO) as a passive tracer for biomass burning plumes. The plume injection heights were adjusted to match the impacts with the vertical ozone profiles at a resolution of 100 m. The model does not include chemical loss or production in its trajectory calculations. The simulations can be used to estimate CO concentrations using multiple particles each representing a fixed amount of CO emitted and transported as a passive scalar. Therefore, trajectories aged 5 to 15 days can be considered as mixing ratios above background (Brioude et al., 2007). WRF was initialized using the North American Regional Reanalysis (NARR, Mesinger et al., 2006) wind fields, with hourly temporal resolution, two nested domains with a horizontal resolution of 27 km and 9 km, and 40 pressure levels. Figure 1 shows a map of the two nested WRF domains. Nine individual WRF simulations were performed to cover the entire time span of the SEACIONS mission. Each simulation lasted 162 h with 42 h of spin-up time and the remaining five days were used for the analysis (de Foy et al., 2014).
To determine the transport impacting each ozone measurement, a plume origin path was calculated. The plume path was determined based on several thousand forward trajectory particles released from a box surrounding the location and time of biomass burning present over the 44 day period. Plume origin path distributions were output in 1 day intervals on a 2.5˚ x 2.5˚ output grid at 500 m vertical resolution covering the U.S., with a 0.25˚ x 0.25˚ at 100 m vertical resolution nested grid over St. Louis. FLEXPART-WRF outputs the plume origin paths in units of s kg-1 m3, which represents the residence time of the plume per grid cell divided by the air density. The residence of each plume was calculated for the surface to upper troposphere (~0 to 15 km) layers of the atmosphere. This is known as a weighted plume residence time (Lal et al., 2014).
The dispersion of a plume origin path forward in time indicates the likely source regions contribution of the ozone precursors to the measured ozone but over the previous 7 days. For plumes passing through these layers, the residence time of the forward trajectories can interact with stratospheric intrusions allowing for the detection of mixed plumes. Hourly biomass burning emissions were provided by the Fire Locating and Modeling of Burning Emissions (FLAMBE) program (http://www.nrlmry.navy.mil/flambe/) (Reid et al., 2009). The emission factors are outlined by Ferek et al. (1998). FLAMBE provides quasi-operational, with fire location, instantaneous estimates of fire size, and smoke emission flux in kg m-2 generated in near real time for the Western Hemisphere (Reid et al., 2009). The inventory uses active fire detection from the GOES’s Wildfire Automated Biomass Burning Algorithm (WF-ABBA) and MODIS’s Active Fire Products to detect biomass burning activity in near-real-time (Reid et al., 2009). The fire emissions are calculated in grams of CO from the inventories’ burned area (m2) and smoke aerosol emissions (CO in gm-2) variables. The fire emissions are mapped onto a 0.25˚ x 0.25˚ grid as input to FLEXPART-WRF. With this technique, the quantity of CO emitted into each plume from several fire source regions across the U.S. was tabulated. The CO tracer has no chemical or depositional removal processes and was treated as a passive tracer. The NARR analysis contains information that can be used to derive stratospheric ozone values above and within the troposphere. We used these values to calculate the quantity of ozone transported from the stratosphere to the location where a plume was released. NARR provides more information on the stratospheric intrusions structure with temporal coverage 8-times daily and a 32 km spatial grid. Stratospheric intrusions in NARR were calculated using PVU distribution on isentropic surfaces, 320 K to 410 K (≈ 80 hPa). Specifically, with NARR we can closely match ozonesonde launch times with meteorological data (~18:00 UTC); this allows for better comparison with model and observed stratospheric intrusions.
3.2.2. FLEXPART-WRF high-altitude smoke injection
To account for the vertical extent (or plume injection height) of varying biomass burning emissions in the simulations, we employ the FLEXPART-WRF model. The BASELINE simulation represents biomass burning emissions as a uniform release within the well-mixed boundary layer up to ~3.5 km (e.g. Liousse et al., 1996; Colarco, 2004; Davison, 2004; Brioude et al., 2007). To represent plumes that penetrate well above the boundary layer, we use the U.S. Naval Research Laboratory (NRL) satellite-based pyroCb detection algorithm and inventory (Peterson et al., 2017a,b) — referred to hereafter as the PYRO simulation. For a confirmed pyroCb, the plume tops are set at the model determined Upper Troposphere and Lower Stratosphere (UTLS) boundary (~9 to 15 km). For non-pyroCb, plume tops are set to be near the model determined middle troposphere (~5 km). In both cases, the plume bottoms are set above the PBL, with the emissions uniformly distributed within the tropospheric model layers. Lastly, a method is presented that is a combination of the two methods mentioned above (BASELINE + PYRO) referred to hereafter as the combined method. The combined method was run as a single simulation and compared to each individual method. The combined simulation included all the emissions from the Baseline, with addition of the elevated emissions from the fires with pyro convective plumes identified in Figure 1, see Table 1 for more details.
Peterson et al. (2017a) examined a variety of wildfires and pyroconvective events in western North America during the summer (June to August) of 2013. At least 26 intense pyroCb events were inventoried, injecting smoke particles well into the UTLS. Several fires produced smaller pyroCbs and others injected smoke above the boundary layer in the absence of significant pyroCb activity (Figure 1). The NRL pyroCb detection algorithm is designed to detect large plumes that impact the UTLS, and in general, is not designed to capture lower tropospheric transport. The algorithm only detects anvil cloud tops; therefore, neither the exact quantitative plume height nor a plume bottom can be detected.
A variety of additional observations and methods exist to track the evolution of pyroCb smoke plumes in the UTLS after the initial injection (e.g., Fromm et al., 2010, 2019; Peterson et al., 2018). However, application of these methods is beyond the scope of this study. Evolution and transport of significant pyroCb plumes from the 2013 fire season are examined in greater detail by Peterson et al. (2015, 2017a,b) and Fromm et al. (2019).
4 Results and Discussions
4.1. The structure of tropospheric ozone columns