Wildfire and the atmosphere: Modelling the chemical and dynamic interactions at the regional scale

Abstract

Forest fires release significant amounts of trace gases and aerosols into the atmosphere. Depending on meteorological conditions, fire emissions can efficiently reduce air quality and visibility, even far away from emission sources. In 2005, an arson forest fire burned nearly 700 ha near Lançon-de-Provence, southeast France. This paper explores the impact of this Mediterranean fire on the atmospheric dynamics and chemistry downwind of the burning region. The fire smoke plume was observed by the MODIS-AQUA instrument several kilometres downwind of the burning area out of the Mediterranean coast. Signatures of the fire plume on air pollutants were measured at surface stations in southeastern France by the air quality network AtmoPACA. Ground-based measurements revealed unusually high concentrations of aerosols and a well marked depletion of ozone concentrations on the day of the fire. The Lançon-de-Provence fire propagation was successfully simulated by the semi-physical fire spread model ForeFire. ForeFire provided the burnt area at high temporal and spatial resolutions. The burnt areas were scaled to compute the fire heat and water vapour fluxes in the three-dimensional meso-scale non-hydrostatic meteorological model MesoNH. The simulated fire plume kept confined in the boundary layer with high values of turbulent kinetic energy. The plume was advected several kilometres downwind of the ignition area by the Mistral winds in accordance with the MODIS and AtmoPACA observations. The vertical plume development was found to be more sensitive to the sensible heat flux than to the fire released moisture. The burnt area information is also used to compute emissions of a fire aerosol-like tracer and gaseous pollutants, using emission factors for Mediterranean vegetation. The coupled model simulated high concentrations of the fire aerosol-like tracer downwind of the burning zone at the right timing compared to ground-based measurements. A chemical reaction mechanism was coupled on-line to the MesoNH model to account for gaseous chemistry evolution in the fire plume. High levels of ozone precursors (NO_x, CO) were simulated in the smoke plume which led to the depletion of ozone levels above and downwind of the burning zone. This depletion of ozone was indeed observed at ground-based stations but with a higher impact than simulated. The difference may be explained by the simplified design of the model with no anthropogenic sources and no interaction of the smoke aerosols with the photolysis rates. Ozone production was modelled tens of kilometres downwind of the ignition zone out of the coast.

Introduction

The latest report of the IPCC (2007) highlights that climate change is very likely to impact fire risk in the Mediterranean Basin region. In fact, even if Mediterranean wildfires are mostly human-induced, the study of Moriondo et al. (2006) based on regional modelling indicates that fire frequency, fire severity and the length of the fire season would increase under future climatic conditions (based on the IPCC A2 and B2 scenarios). Furthermore, the analysis of Pausas (2004) for the eastern Iberian Peninsula confirms that a relationship exists between fire events and seasonal meteorological conditions (summer temperatures and mean rainfall), as also shown by annual data reported by the European Forest Fires Information System of the Joint Research Centre (EFFIS, 2008).

Within the context of increasing fire risk, it is necessary to investigate the dynamics and chemistry of forest fires, which are a threat not only to local ecosystems but also to public health. In the vicinity of the fire, biomass burning produces high concentrations of carbon dioxide (CO₂), carbon monoxide (CO), methane (CH₄), nitrogen oxides (NO_x), volatile organic compounds (VOCs) and particulate compounds (Lobert and Warnatz, 1993). CO₂ and CH₄ are the most important greenhouse gases responsible for the “enhanced greenhouse effect”. Moreover, CH₄ together with CO, NO_x and VOCs are chemically active gases hazardous to human health both directly and indirectly, since they are precursor gases of tropospheric ozone (O₃); in the troposphere, NO_x from combustion also allow the re-generation of the hydroxyl radical (OH) that, in turn, catalyzes the O₃ production. Lastly, biomass burning particulates can reduce visibility and air quality on a local scale and aerosols can affect the radiation budget of the Earth, impacting global and regional climate. Wildfire emissions can also be transported over considerable distance, spreading their effects from local to regional and occasionally global scales, depending on the efficiency of atmospheric transport (Takegawa et al., 2003; Miranda et al., 2008; Bytnerowicz et al., 2010).

The extent of the degradation of air quality due to forest fires has been quantified at different scales. Fire experimental fields in France (Barboni et al., 2010) and Portugal (Miranda et al., 2005) revealed concentrations of toxic air pollutants well above exposure limit values settled by the European Legislation established in the Council Directive 2008/50/EC. Miranda et al. (2005) analysed concentrations of particulate matter (PM), NO_x, CO and sulphur dioxides (SO₂) during an experimental field fire performed in 2002 at Gestosa. This experiment stresses the critical situation in terms of local air quality that can occur during a fire episode and affect the personnel involved in fire-fighting operations. The maximum hourly averaged values for aerosol particles with an aerodynamic diameter lesser than 2.5 μm (hereafter PM_2.5) and smaller than 10 μm (hereafter PM₁₀) were, respectively, 2350 μg m⁻³ and 1430 μg m⁻³. Gestosa PM concentrations are in the range of the hourly averaged data recorded in operational conditions during a wildfire in Greece: 3350 μg m⁻³ and 1300 μg m⁻³, respectively. Similar PM concentrations were observed near the Quinault fire (Trentmann et al., 2002). During Gestosa-2002, CO concentrations peaked at nearly 60 mg m⁻³, nitric oxide (NO) at 600 μg m⁻³ and nitrogen dioxide (NO₂) at 500 μg m⁻³. On a regional scale, Phuleria et al. (2005) measured pollutant gases and PM concentrations in the Los Angeles (LA) basin before, during, and after the October 2003 Southern California wildfires. They documented a strong degradation of urban LA air quality due to the fires. Downwind of the fires, the greatest impact was observed on coarse-PM concentrations which exceeded typical background concentrations by factors of three or four: PM₁₀ concentrations were near or above 200 μg m⁻³ during the fires. During the same event, CO was increased by nearly 12 ppmv and NO reached 100 ppbv. Interestingly, NO₂ levels remained essentially unchanged and O₃ concentrations decreased by about 25–50%. The authors proposed the reduction in photochemical activity due to the fire smoke blanketing the LA basin as a possible explanation for the NO₂ and O₃ fire response. O₃ depletion was also documented by aircraft measurements in young biomass-burning plumes, close to the fire, in South Africa (Hobbs et al., 2003) and Namibia (Jost et al., 2003). In addition, aircraft measurements investigated the change of the mixing ratio of many species in the fire plume moving away from the ignition point. Jost et al. (2003) measured 1703 ppbv of CO over the fire, rapidly decreasing by one-third at a distance of 4 km downwind of the fire. Yokelson et al. (2007) observed average PM₁₀ values for vertical profiles that ranged from 70–120 μg m⁻³ at 300–500 m to 30–60 μg m⁻³ near the top (∼3000 m) in central Brazil.

In Europe, episodes of trans-boundary fire tracer dispersion have already been observed and modelled. Saarikoski et al. (2007) and Sofiev et al. (2008) considered the influence of emissions from Russian and Baltic wildfires on air quality in northern Europe during spring and summer 2006. They used a Lagrangian dispersion model, SILAM, with fire emissions based on the Moderate Resolution Imaging Spectro-radiometer (MODIS) hot spots to simulate the observed increase of fire pollutants recorded at ground-base stations. A similar approach was used by Tressol et al. (2008) based on the FLEXPART model to assess the main origin of strong anomalies of O₃, CO and NO_x registered by MOZAIC aircraft above Frankfurt during the 2003 heat wave, when severe wildfire activity hit Portugal. Lagrangian models succeed in reproducing the main characteristics of fire plumes advection. The most important limitation of the current versions is the treatment of fire injection height which is generally kept constant. Lagrangian models are currently not designed to investigate the strong updrafts and convective fluxes associated with wildfires.

Hodzic et al. (2007) investigated the effects of forest fires on air quality in Europe during summer 2003 using the meso-scale chemistry transport model CHIMERE. CHIMERE has been improved to include the MODIS smoke emissions inventory and implements a new parametrisation to simulate the injection of smoke particles. The injection height is calculated as a function of atmospheric conditions and fire characteristics, retrieved from the MODIS inventory. The parametrisation allows for the simulation of the transport of smoke plume at the right altitude. This approach relies on the accuracy of satellite measurements.

Another approach is that of Turquety et al. (2009). The authors used the Infrared Atmospheric Sounding Interferometer (IASI) for the monitoring of CO during the summer 2007 Greek fires. Once a retrieval algorithm of CO vertical profiles is defined, CO mixing ratios are analyzed close to the fires and in the transported plume: this technique allows to study the dispersion of fire tracer and to roughly estimate the general level of injection of the fire plume.

The next level of complexity of fire–atmosphere coupling is the Eulerian high-resolution model ATHAM (Oberhuber et al., 1998). In order to investigate the connection between wildfire and atmosphere, in terms of both dynamics and chemistry, the active tracer atmospheric model ATHAM was forced utilizing wildfires parameters such as heat release and aerosol fluxes, obtained from ground-based observations. A simplified design was chosen with a static fire front and fire fluxes held constant throughout the simulation. Although these simplifications, the ATHAM model successfully simulated the transport of fire emissions, chemical processes leading to the formation of tropospheric O₃ in a young biomass-burning plume and radiative effects in a smoke plume, and pyro-convection (Trentmann et al., 2002, 2003, 2006; Luderer et al., 2006). However, the cited fire–atmosphere coupling does not consider a temporal and spatial evolution of fire characteristics and there is no feedback from the atmosphere to the fire.

The interaction between the atmosphere and the fire can be fully resolved using a fire spread model coupled with an atmospheric model. Fire spread models vary from empirical (Clark et al., 2004) to physics-based systems (Linn et al., 2002, among others). For a complete review of fire spread models the reader is referred to Sullivan (2007a,b). Semi-physical fire spread models are a good compromise being based on physical laws whose complexity is reduced by imposing realistic assumptions (Sullivan, 2007a; Filippi et al., 2009). Coupled fire–atmosphere model studies normally focus on small scale atmospheric processes, since at these resolutions these models are able to reproduce fire-induced effects on wind and turbulence that have been measured on field campaigns, as, for example, during the Fire-Flux experiment (Clements et al., 2007). At larger scales (meso-scale), an example of one-way coupling between the fire and the atmosphere is illustrated by the work of Miranda (2004). The author coupled the meteorological model MEMO to a semi-empirical fire progression model FARSITE and successfully reproduced the effects of the forest fires on the air quality in Lisbon during summer 2003. However, at resolutions much higher than the fire front resolution, coupled models still have limits and constraints that need to be further explored. The difference between resolution of the meso-scale atmospheric model and the coupled high-resolution fire spread model imposes the parametrisation of sub-grid fire processes. Mesoscale models incorporate various parametrisations to include sub-grid vertical transport, but strong vertical updrafts associated with intense heat sources, such as wildfires, are frequently ignored, or, their impact is diluted, at the resolution typical of large-scale models (Freitas et al., 2006). This deficiency implies that the fire injection height may be underestimated. The fire injection height is an important parameter necessary for the study of air quality during fire episodes. If pollutants are released in the Planet Boundary layer (PBL), removal processes are more efficient and can shorten pollutant residence time (Chatfield and Delany, 1990). On the contrary, when emitted into the free troposphere, characterized by faster winds, the pollutants can be transported considerably further and affect air quality from the local through the regional and global scales. Several studies have been carried out to investigate the height to which smoke plumes rise and the variability of this altitude due to atmospheric conditions and fire characteristics (Labonne and Chevallier, 2007; Kahn et al., 2008; Martin et al., 2010; Guan et al., 2010). Current methods to parametrise plume lifting are based on a one-dimensional entrainment plume rise model embedded in a host model (Freitas et al., 2007) or on a mixed eddy diffusivity – mass flux scheme for convective boundary layer plumes (Rio et al., 2010). So far, these approaches have been validated for African and Amazonian fires for which elevated injection heights have been observed.

The objective of this study is to explore fire impact on atmospheric dynamics and chemistry downwind of a burning area located in the Mediterranean region. The atmospheric meso-scale model MesoNH is coupled with the semi-physical fire spread model ForeFire to simulate the Lançon-de-Provence 2005 forest fire. A chemical reaction mechanism is coupled on-line to the MesoNH model to account for gaseous chemistry evolution in the fire plume. In Section 2, a brief description of the atmospheric and the fire spread model is given, precising model setup and initialization. Section 3 introduces the Lançon-de-Provence 2005 case study, describing the fire history and synoptic meteorological conditions before the wildfire burst out. In Section 4, the fire plume dynamics and chemical composition are compared with MODIS observations and ground-based measurements registered by the air quality survey network available in southeastern France (AtmoPACA). Finally, conclusions are summarized in Section 5.