Biomass burning is a major dynamic of the Earth system responsible for the emission of massive quantities of trace gases and aerosols to the atmosphere e.g.. To understand and quantify the effects of these biomass burning emissions on atmospheric composition, air quality, weather, and climate, many fire emission inventories have been developed at scales such as individual areas, countries or regions e.g., continents e.g., or the entire globe (e.g. FLAMBE (Naval Research Laboratory), GFED (G. van der Wref, VU University Amsterdam), FINN (National Center for Atmospheric Research, NCAR), GFAS (European Center for Medium-range Weather Forecast, ECMWF)) respectively.

The use of satellite Earth observation (EO) data is generally considered to be critical to providing the temporal coverage, spatial sampling frequency, and directly observable parameters necessary for creating these inventories, particularly so since landscape fires are highly variable emissions sources, and the exact amounts of material released by the combustion process is highly variable in both space and time .

When a landscape fire occurs, a rising plume created from the intense heat and convection produced by the energy released by burning vegetation interacts with the ambient atmosphere and transports the smoke emissions, affecting their longevity, chemical conversion, and fate . This makes the manner in which the fire emissions are injected into the atmosphere highly variable and sensitive to the smoke plume dynamics. To follow the terminology commonly used in the literature e.g., when not specified, the term “fire emission” refers to the gaseous and aerosols emissions only and not the heat fluxes (e.g. radiation) emitted by the fire.

Figure  shows EO satellite views of the evolution of the smoke plume generated by the “county fire”, which occurred in Ocala National Forest (Florida) in 2012. The fire was active from 5 to 13 April 2012 and burned across nearly 14 000 hectares (140 km2) of land. The apparent intensity and direction of travel of the smoke plume changes every day, and such variability is most likely related to both changes in the fire activity (for the former) and the local ambient atmospheric conditions (for the later). Together the fire and ambient atmospheric characteristics are the main drivers of the plume dynamics and therefore ultimately of the smoke emissions transport.

In addition to the use of in situ measurements e.g. and satellite Earth observation e.g., the wide-ranging controls on and impacts of landscape-scale fire emissions can be investigated using atmospheric chemistry transport models (CTMs) e.g.. Such models require information on the quantity and timing of the fire emissions, as well as their chemical makeup, and these generally come from the aforementioned emissions inventories. However, for a more complete representation of the source fires, many CTMs can also make use of information on the altitude at which the bulk of the emitted species is injected into the wider atmosphere, where they can fully interact with ambient atmospheric circulation. In a recent study on fire emission transport use the GEOS-Chem CTM with a horizontal resolution of 2∘ × 2.5∘ and 47σ levels forming a vertical stretched mesh with a resolution of 150–200 m near the planetary boundary layer (PBL). Since at these resolutions we cannot resolve the plume dynamics ≲100 m, parameterizations are therefore required to represent these “smoke plume injection heights” (InjH). The aim of this paper is to review the different approaches required for providing these parameterizations. The paper is structured as follows. First, Sect.  provides the background detail on fire plume observations and modelling in large-scale CTMs. The main physical mechanisms responsible for the fire plume dynamics are discussed in Sect. . The primary satellite EO data used currently to study plume injection height properties are detailed in Sect. . Then, the currently available injection height models and their implementations are discussed in Sect. . Finally, a summary and suggestions for further developments in this area are provided in Sect. .

Fire emissions are a particular case of emissions to the atmosphere, since they can be injected into the atmosphere far above the PBL and can thus potentially spread over a long distance according to local atmospheric circulation patterns. Only emissions from aircraft traffic and volcanoes , which are also coupled with intense dynamical mechanisms, offer a similar capability.

The question of the impact of fire emission injection in the atmosphere was first introduced by and was later extensively reported in EO data. For example, injections of gases and aerosols emitted from vegetation fires have been observed at various heights in troposphere and occasionally even the lower stratosphere . Smoke remnants from certain tropical fires have been observed at 15 km altitude , and plumes from individual Canadian stand-replacing forest fires can also reportedly approach such heights . For the largest events, observations from show that a single fire was able to induce a significant average surface temperature decrease at the hemispherical scale. The emissions from such large fire events are capable of spreading extremely rapidly, and show that the transport of emissions from an Australian fire in 2006 spread around the globe in only 12 days.

Using EO data from the Multi-angle Imaging SpectroRadiometer (MISR) instrument onboard the Terra satellite we estimate that 5 to 18 % of 664 plumes observed from boreal forest fires over Alaska and the Canadian Yukon in 2004 reached the free troposphere (FT). Using AI peak observation from TOMS, backward trajectories to identify location of the causal fire, and then GOES and/or American and Canadian fire report data bases for confirmation, identify a total of 17 plumes that reached an altitude of at least 10 km for the year 2002. Fires whose smoke columns reach these elevations are also likely to be those that emit large quantity of gases and aerosols; therefore, even though such large and intensely burning fires are relatively less common than smaller, less intense events, their impacts are likely to be much greater than the “average fire” . Other evidence shows that fires from agricultural or grassland vegetation type (usually less intense than those from boreal forest) can also generate plumes reaching the FT. show that half of the agricultural fires they observed over eastern Europe for the period 2006–2008 reach heights above the PBL. show that 26 % of the 27 Australian grassland fires they studied with various stereo-height retrieval algorithms rose above the PBL. In summary, the height to which biomass burning plumes rise, and the distance over which the emissions are therefore transported, is highly variable. Possibly even more variable than fire behaviour, since the same fire burning under different ambient atmospheric conditions will probably result in different plume behaviours. It is important to note however that certain atmospheric conditions are more favourable to fire occurrence than others, such as high pressure and/or low moisture (dry season) conditions .

Fully modelling the impacts of biomass burning emissions at large scales requires an understanding of plume dynamics, including their InjH. Some InjH inventories are already available, for example derived from satellite EO data of aerosols or CO. For example,

screened aerosol index (AI) measurements extracted from data collected by the Ozone Monitoring Instrument (OMI) and the Total Ozone Mapping Spectrometer (TOMS) to map high aerosol clouds (>5 km) related to wildfires over the period 1978–2009;

Although the above EO-based approaches are useful to understand and quantify the occurrence of wildfire plumes at different heights, they have limited sensitivity to the potential variability of InjH. Both inventories are therefore quite difficult to couple to fire emissions inventories and cannot be easily linked to particular fires and therefore to actual emission totals. Capturing the high variability of plume dynamics, estimating InjH, and implementing this within a CTM therefore remains a current topic of very active research , and the task of this paper is to review the different approaches currently available.

The injection height of a smoke plume is controlled by the plume dynamics, which are driven by both the energy released by the fire and the ambient atmospheric conditions (both stability and humidity) . In the time period between the emissions being first released by the combustion process (which happens at the flame scale of ∼ mm), and their later release into the wider atmosphere (which operates on a metre to kilometre scale), the smoke emissions are trapped in the plume (see Fig. ). Here the dynamics are dominated by

i.

the buoyancy flux induced by the convective heat flux (CHF) generated by the fire itself;

In some scenarios, the combination of these processes initially triggered by the heat released from the vegetation combustion is capable to producing deep convection in places where natural convection would not normally be possible; the so-called pyroconvection phenomena . show that in the case of large events like the Chisholm fire documented by, the energy budget of the plume is essentially driven by the latent heat released from the condensation of the entrained water vapour.

Depending on the quantity of water vapour condensed during the plumes development, three types of vegetation fire plume can be identified . i.

Dry smoke plumes containing water vapour rather than liquid droplets. These are typically created by smaller, weakly burning and low intensity fires and usually stay trapped in the PBL.

Since the initial trigger of plume rise is the heat released by the casual fire, InjH are strongly influenced by fire diurnal cycles . This leads to lower nocturnal InjH which are amplified by the combination of nighttime stable atmosphere and lower PBL . However some meteorological conditions can intensify fire activity over night, as for example the Santa Ana foehn wind , and keep them running. Few observations of nocturnal plumes triggered by those intense fires are available , and to our knowledge only tackle the issue of modelling nocturnal InjH. Their approach relies on a simulated diurnal cycle based on the high temporal resolution (∼15 min) fire radiative power (FRP) product of the geostationary orbiting satellite SEVIRI and the parameterization of (further discussed in Sect. ). Despite the low resolution of SEVIRI (>3 km), their empirical diurnal cycle captures the expected fire intensity increase at night, but no effects were found on InjH. Their resulting modelled InjH shows a strong diurnal pattern with low nocturnal InjH (e.g. maximum monthly mean nocturnal InjH lower than 2.5 km).

Of course, a full understanding of the complex coupled mechanisms inherent in fire plume dynamics is extremely challenging, and many points remain unclear: for example, the role of soot and aerosol in the heat transfer within the plume column and the effect of the number of initial cloud condensation nuclei on the triggering of pyroconvection .

Sensors and imagers onboard EO satellites can provide various information on wildfire plumes, including their trace gas ratios e.g., aerosol burden e.g., and their height, including on occasion the vertical distribution of material within them e.g.. provide a recent review of this topic. EO data also provide information on the characteristics of the causal fires themselves, including “active fire” (AF) products that detail the location, timing, and FRP of the landscape-scale fires occurring within the EO satellite pixels . FRP is a fire characteristic that has been shown to relate quite directly to the total heat produced by the combustion process and also to the rate of fuel consumption , trace gas , and aerosol e.g. emission. Such active fire products are usually derived from thermal wavelength Earth observations .

No satellite product is yet able to derive information on plume heights at a spatial and temporal resolution than matches those of sensors used for active fire detection and smoke emission estimation, such as e.g. the Moderate Resolution Imaging Spectroradiometer (MODIS), Meteosat SEVIRI, or the Geostationary Orbiting Environmental Satellite (GOES) . Therefore, determination of the injection heights at spatiotemporal scales and levels of completeness approximately matching these type of active fire observations is more likely to rely on InjH parameterizations.

A number of studies have determined the very serious implications that incorrect InjH estimates have on the ability of CTMs to represent emissions transport e.g.. Consequently it may also effect (i) “top-down” emission estimates based on the inversion of observed atmospheric concentrations of biomass burning species and (ii) radiative forcing studies . This section aims to review the different parameterizations that are currently available to tackle the issue of InjH. They are based either on empirical, deterministic, or statistical models.

Weakly burning landscape-scale fires appear to release their smoke mainly into the planetary boundary layer, but larger and/or more intensely burning wildfires produce smoke columns that can rise rapidly and semi-vertically above the source region, driven by the intense heat and convective energy released by the burning vegetation. These columns of hot smoke entrains cooler ambient air, developing into a rising plume within which the trace gases and aerosols are transported to potentially quite high altitudes, in the most extreme cases into the stratosphere. The characteristics of these rising plumes, and in particular the height that they reach before releasing the majority of the smoke burden, are now acknowledged as an important control on the atmospheric transport of emissions from certain of these larger fire events . However, results comparing model-based estimates of smoke plume rise parameter to actual plume height observations made from satellite EO instruments e.g. do not yet provide a strong quantitative agreement . Furthermore, the degree of improvement given by actually including plume rise parameterizations in atmospheric chemistry transport models can be difficult to interpret due to the complex interactions with other atmospheric processes .

Apart from simulations based on single fire events, where plume injection height is carefully prescribed or where highly detailed simulations are run at very high resolution , the impact of fire-induced up-draft on wildfire plume lofting appears, in general, to remain rather poorly understood and often weakly represented in current large-scale atmospheric modelling efforts. The impact of possibly coupled effects on ambient atmospheric processes, such as the convection induced by the nearby presence of a cold front, is also not well determined. At the scale of global CTMs, wildfire plume rise is generally represented by some form of parameterized model . The ideal parameterization should account for the main physical processes responsible for the plume dynamics, using inputs regarding the fire characteristics that are available from EO satellites in near real time and with concurrent measurements of fire activity and plume height from single fire events available to validate the resulting system (and reduce any impact from larger-scale transport effects that influence comparisons of downwind plume characteristics).

Despite a demonstrated diurnal bias of MISR-derived plume heights towards lower plumes , the current MISR data set for North America counts 22 fires with plume top higher that 4.5 km (see Figs.  and for an overview of the current MISR data over North America).

However, those high plumes might not be fully representative of standard fire behaviour as show that PyroCb plume maturity peaks around 18:00, and no fires are observed around that time with MISR (see local time observation distribution in Fig. d). Therefore, any PyroCb contained within the MISR-derived plume height data set are certainly few in number, which leads to questions regarding the full representativeness of a random selection of fire events selected from this sample . In their approach, apply several selection criteria when taking a subsample of fires extracted from the MISR plume height data set for use in evaluation of their parameterized plume rise model, which is an adaptation of the widely used model of . However, even with this carefully selected evaluation data set, the validation of this PRM model fails to show very convincing results. Nevertheless, in future, such validation (or optimization) of plume rise models should continue to pay attention to the quality of the evaluation data sets, including the following questions. i.

Are the fire activity (FRP) and the plume dynamics (plume top height) linked? A time delay is necessary for the plume to dynamically adjust to change in the forcing induced by the energy release by the fire. For example, during the simulation of PyroCb of the Chisholm fire by the ATHAM model, it takes 40 min for the plume to reach its stationary altitude with a constant forcing . As the smoke plumes observed by MISR are more likely in a relatively early stage of development due to the morning overpass of the Terra satellite (see Sect. 4.1.2), the effect of this time lag might be even more important than if fires were randomly observed at any time of their development.

Despite these difficulties, the range of relevant data provided on actively burning fires and their smoke plumes by EO satellites continues to grow e.g.. For example, GOES-R and Himawari-8 will provide capabilities similar to MODIS, with a temporal frequency potentially as high as 30 s, while Suomi NPP carrying VIIRS and TET-1/BIRDS will provide thermal bands with resolution up to 375 m. This will allow for detailed observations of pyroconvection during peak burning hours. These improving capabilities, together with continuing advances in the extent to which plume rise models can be parameterized and incorporated into large-scale atmospheric CTMs , can be expected to continue to advance the accuracy of smoke plume injection estimates and the resulting impact on long-range atmospheric transport of these globally important emissions.