The Dust Emission Potential of Agricultural‐Like Fires—Theoretical Estimates From Two Conceptually Different Dust Emission Parameterizations

Agricultural fires affecting grass‐, crop‐ and shrublands represent a major, mainly anthropogenically driven disturbance of many ecosystems. In addition to emissions of carbonaceous aerosol, they were found to inject also mineral dust particles into the atmosphere. The fires can significantly modulate the near‐surface wind patterns so that conditions suitable for dust emission occur. However, the exact emission mechanism has not been investigated so far, but is inevitable for the understanding of its impacts on the Earth system. Here, we test two dust emission parameterizations representing saltation bombardment (SALT) and direct aerodynamic dust entrainment by (convective) turbulence (convective turbulent dust emission, CTDE) in the context of fire‐modulated wind patterns using large‐eddy simulation with an idealized setup to represent typical agricultural fire settings. Favorable aerodynamic preconditions for the initialization of both emission processes are found, however, with sometimes significant differences in dust emission flux depending on specific wind and fire properties. The strong fire‐induced modulations of the instantaneous momentum flux suggest that CTDE can be a very potent emission process in the fire vicinity. Nevertheless, fire impacts on the friction velocity can be significant too, so that dust emission through SALT is facilitated as well. Ultimately, the specific aerodynamic conditions within pyro‐convectively modulated wind patterns require the development of a parameterization that can describe these unique fire‐related dust emissions and their influencing factors properly. This will finally allow for considering fire‐induced dust emissions in aerosol‐atmosphere models and an investigation of its atmospheric impacts such as on the radiation budget.

2 of 20 climate change has impacted the global fire regime drastically within the most recent past. Changing climate conditions alter the fire risks in many regions around the globe due to shifts in precipitation and temperature patterns. This goes hand in hand with more frequent and intense heat waves eventually leading to temporary or persistent drought conditions (e.g., Bowman et al., 2017;Westerling, 2006). The dried-out vegetation is then more vulnerable to any kind of ignition, which can lead to an increased destructiveness of both natural and prescribed fires. This includes, for example, large parts of Eurasia during the hot and dry spring and summer seasons from 2018 to 2020, the Brazilian rain forests (mainly pushed by arson) in 2019, parts of Australia during the turn of the year 2019/20 (Sanderson & Fisher, 2020) as well as the western US (Higuera & Abatzoglou, 2020). Most of these fires were not just limited to one particular fire type, instead a variety of different landscapes was affected.
In general, the burning vegetation type strongly determines the intensity and destructiveness of a fire as it is usually a function of the available fuel load, that is, the (dry) biomass (e.g., Albini, 1993). While forests can provide a large fuel load and thus are linked to higher burning temperatures, the available fuel load is much more limited in case of grass-and croplands (e.g., Clements et al., 2007). Therefore, these fires are less intense and burn at lower temperatures. This behavior can be expressed by the sensible heat flux generated during the combustion process. Depending on the vegetation type and the environmental conditions, large variations ranging from several kW 2 m E  for weak grassland fires to a few thousand kW 2 m E  for intense crown fires are observed (Frankman et al., 2013;Lareau & Clements, 2017). Although the heat flux can vary substantially in space and time even within a single fire, a rough categorization can be made: Grassland fires are usually linked to sensible heat fluxes below 100 kW 2 m E  , while shrubland fires can reach intensities in an order of magnitude of some hundred kW 2 m E  (Clements et al., 2007;Frankman et al., 2013;Lareau & Clements, 2017). Stronger heat fluxes are usually linked to natural forest fires, where the major burning center is often elevated from the ground and thus impacts on the soil surface are much more limited (Clark et al., 1999). In contrast, the fire intensity of prescribed forest fires is usually lower and they can burn through the coppice as well and the soil surface may experience stronger impacts.
All kinds of wildland fires are well known as a major hazard due to their huge impacts on the atmosphere, biosphere and society. For instance, they destroy the vegetation cover, threaten the wildlife and the human population; the emitted combustion gases and aerosol particles can harm human health and such fires are found to impact the local weather conditions and even the large-scale climate (e.g., Bowman & Johnston, 2005;Forster et al., 2007;Kumar et al., 2011). While climate effects emerge primarily from the interaction of the fire emissions with the Earth's radiation budget and cloud microphysics, a fire represents also a huge disturbance of the local wind, temperature and stratification patterns. In particular, the fire updraft and the accompanied inflow motions can strongly modulate the wind regime within and around the burning fire area (Clements et al., 2008;Palmer, 1981;Peterson et al., 2015). Furthermore, fires are linked to enduring modifications of the soil surface as the soil characteristics are altered by the fire heat (Dukes et al., 2018;Ravi et al., 2012). A reduction of the soil moisture, modifications of the soil texture as well as changes of the grain size distribution can be observed, which, depending on the fire type, burning temperature, and predominant soil type, result in different impacts on the erodibility of a fire-affected soil surface (Dukes et al., 2018;Kavouras et al., 2012;Pérez-Cabello et al., 2006). This includes both the erosion due to water runoff as well as Aeolian erosion generated by aerodynamic forces. Here, the majority of investigations found an increasing number of fine particles due to breakdown of larger soil aggregates (Albalasmeh et al., 2013;Levin et al., 2012;McNabb & Swanson, 1990), although under some circumstances also the formation of coarse-mode particles can occur as a result of aggregation processes (Blank et al., 1996;Vermeire et al., 2005). As the combustion process also consumes the soil-protecting vegetation cover, the soil surface is at least partly exposed to the atmospheric forces. In concert, both effects lead to a reduction of the required aerodynamic lifting forces for the emission of soil-dust particles as shown by several investigations focusing on post-fire dust emissions (Dukes et al., 2018;Merino-Martín et al., 2014;Ravi et al., 2012;Whicker et al., 2006).
While increased post-fire dust emission fluxes were already the subject of numerous studies (e.g., Dukes et al., 2018;Ravi et al., 2012;Whicker et al., 2006), less attention is paid to the co-emission of soil-dust particles during the fire. However, the interplay of a largely fire-cleared soil surface with the fire's pyro-convective aerodynamic forces does likely enable the emission of soil-dust particles, too. Remote sensing 3 of 20 and in-situ observations have found significant fractions of mineral dust particles within smoke plumes emerging from wildfires in semi-arid regions (Kavouras et al., 2012;Nisantzi et al., 2014;Radke et al., 1991;Schlosser et al., 2017). Despite mixing processes, which may have entrained dust from origins outside the fire, evidence is growing that the strong changes in atmospheric flow patterns within and around a fire are linked directly to the mobilization of soil-dust particles and a subsequent uplift of those particles through the fire's updraft (Kavouras et al., 2012;Palmer, 1981). Particularly, the formation of a convergence close to the surface due to the ascent of heated air can cause a compensating motion that accelerates the horizontal winds and strengthens the near-surface turbulence drastically (Clements et al., 2008;Palmer, 1981). Depending on the fire type and the atmospheric preconditions, these fire-modulated wind forces can be sufficient to mobilize soil particles from the ground, leading to an enrichment of mineral dust particles within the outflow air masses of a fire (Maenhaut et al., 1996;Nisantzi et al., 2014;Radke et al., 1991;Reid et al., 1998).
Such fire-related dust emissions are not yet considered in the classic fire emission modeling, for which so far the main focus lies on the mostly carbonaceous gases and aerosol particles that form as a result of the biomass combustion. However, due to the large fire activity within semi-arid landscapes, fire-induced dust emissions might represent a noteworthy part of the emitted particle mass and thus add to the total atmospheric dust load, especially on a regional scale in the strongest fire-affected areas (Kavouras et al., 2012;Nisantzi et al., 2014;Schlosser et al., 2017). A neglect of this particular dust source could lead to a systematic underrepresentation of the global dust burden and thus to uncertainties concerning the aerosol-climate feedback due to their impacts on the radiation budget or cloud formation (e.g., Tegen & Schepanski, 2018). In addition, if mineral dust is mixed with primary combustion aerosol during the pyro-convectively driven emission process, the physical and chemical properties of the dust particles and eventually their impacts on weather, climate, and human health can be altered (Chalbot et al., 2013;Hand et al., 2010). Thus, a better understanding of the fire-driven dust emission processes is crucial to evaluate these impacts, especially with regard to the anthropogenic interferences that lead to an increase in fire risk in the context of climate change (Jolly et al., 2015;Westerling & Bryant, 2008).
The capability of the enhanced fire-induced winds to mobilize mineral dust particles from the ground was already investigated by Wagner et al. (2018). By defining different fire scenarios, this study has shown that fire can substantially alter the wind patterns in the surrounding area. For instance, the pyro-convective updraft that forms due to the ascent of heated air causes the development of a convergence at the surface, which again leads to an acceleration of the horizontal winds directed toward the convergence zone that is characterized by highly turbulent, convective motions. The main fire updraft and thus the convergence zone forms thereby slightly downwind of the burning fire area if an ambient wind forcing is present such that the enhanced horizontal winds affect particularly the upwind areas. In summary, the study of Wagner et al. (2018) has focused on the changes of the wind speed provoked by the fires and their general ability to exceed typical dust emission threshold velocities rather than on the underlying physical mechanisms leading to those emissions. In contrast, the present study deals with the different emission processes that can take place in the fire environment and investigates the fire-induced changes in the aerodynamic parameters that drive dust emission. We focus on fire setups representative of agricultural grass-, crop-and shrubland fires, where the burning takes place closely above the soil surface and thus fire-related emission of dust particles is more likely. To better understand the potential of pyro-convective aerodynamic forces to emit dust, we apply two conceptually different dust emission parameterizations. Comparison of the dust emissions provides important insights into fire-dust dynamics and serves as a basis to represent fire-related dust emissions on larger spatial scales in the future. The present paper is structured as follows: First, the relevant dust emission mechanisms and their parameterizations are introduced in Section 2. Section 3 focuses on the modeling strategy that is chosen to investigate the research objective outlined above. The corresponding results are presented in Section 4, before a discussion and conclusion closes the paper.

Dust Uplift Mechanisms
The wind-driven emission of mineral dust has been a research subject for a long time (Bagnold, 1941;Kok et al., 2012;Shao, 2008). Experimental data and theoretical understanding have revealed that soil particles with a diameter of around 70 E m can be most easily mobilized by the wind drag, as the combined 5 of 20 key drivers of dust emission via SALT or CTDE, respectively. The model's complexity was kept simple as the focus lies on the fire impacts on the near-surface wind patterns and the corresponding dust emission potential, without any feedback from the atmosphere on the fire evolution. Thus, we represented the fire by a constant and stationary sensible heat flux at the lower boundary of the model domain as it is typical for agricultural fires burning close to the surface. To ensure realistic ambient atmospheric conditions, the fire was ignited within an already turbulent, well-mixed boundary layer, whose evolution was initiated using a typical atmospheric profile gained from a mesoscale model simulation covering the Sahel zone by Tegen et al. (2013). The study's design is described in further detail in Wagner et al. (2018) where the interested reader can also find a more detailed explanation of the underlying concept of pyro-convectively driven dust emissions. The present investigation builds on that study, however, complemented by further simulations.
In the framework of this study, LES was set up with the All Scale Atmospheric Model (ASAM), a powerful numerical solver that has shown its general suitability over a wide range of atmospheric applications including high resolution small-scale process studies (Doyle et al., 2011;Hinneburg & Knoth, 2005;Jähn et al., 2016). ASAM solves the three-dimensional, fully compressible, non-hydrostatic Euler equations by means of a split-explicit Runge-Kutta time-integration scheme (Jähn et al., 2015;Knoth & Wensch, 2014). The model's architecture allows for a high spatio-temporal resolution at which turbulence can be resolved directly and only a subgrid part needs to be parameterized. The relevant grid spacing close to the surface was set to 10 m, valid for all three dimensions x, y, and z, with an integration time step of 0.2 s and an output time step of 10 s.
In order to capture a wide range of different fire types and atmospheric environments, multiple cases were simulated. Within these simulations either the fire intensity in terms of the fire's sensible heat flux, the actively burning fire size, or the mean geostrophic wind velocity was changed. The fire's sensible heat flux was varied between 50 and 270 kW 2 m E  to capture typical agricultural fire intensities reaching from weak grass-and cropland fires to more intense shrubland fires (e.g., Clements et al., 2007;Frankman et al., 2013;Lareau & Clements, 2017). The reference value of the sensible heat flux for variations of the ambient wind forcing and the burning size was set to 150 kW 2 m E  . As the fire intensity can vary widely even within a certain fire type depending on the fuel load, the fire characteristics, and the general environmental conditions, the applied values cannot be linked directly to a specific scenario, however, they lie within a typical range of values representative for the addressed field of application. Variations of the mean geostrophic wind velocity were set to range from quite calm to windier conditions (1-5 m 1 s E  ) representing characteristic scenarios for low intensity fires burning under largely controlled circumstances. The reference value of 3 m 1 s E  , used as background scenario for variations of fire strength and size, has been chosen comparable to wind speeds that are frequently reported by studies focusing on natural fire events or during fire experiments (e.g., Clements et al., 2007;Frankman et al., 2013). Furthermore, the impacts of a changing actively burning fire area on the dust emission potential were considered using size variations reaching from 2,400 to 20,400 2 m E . Here, 7,000 2 m E was set as the reference value.
To estimate the vertical dust emission fluxes, we used two fundamentally different parameterizations to gain insight into the potential of fires to emit dust: (a) the parameterization from Marticorena and Bergametti (1995) as refined by Tegen et al. (2002), which represents SALT; and (b) the parameterization from Klose et al. (2014), which represents direct aerodynamic entrainment due to (convective) turbulence (convective turbulent dust emission, CTDE). The chosen SALT parameterization allows for the representation of different soil types expressed by fractions of clay, silt, fine/medium sand, and coarse sand as well as considerations of soil moisture, roughness length and vegetation cover in order to reflect the impacts of surface properties on dust emission. A key part of the SALT parameterization of Marticorena and Bergametti (1995) is the drag partitioning scheme that accounts for roughness elements such as gravels or vegetation. These disturbing elements can consume a part of the momentum provided by the wind that is then not available for dust emission from the soil surface. To express the fraction of the friction velocity * E u that can act on the erodible surface, an effective friction velocity eff E f can be used: with 0s E z being the smooth roughness length of the soil and 0 E z the aerodynamic roughness length associated with the roughness elements (King et al., 2005;Marticorena & Bergametti, 1995;Marticorena et al., 1997). Different estimates have been established for 0s E z reaching from values defined by the largest grains of the soil ( 0 / 30 , e.g., Marticorena and Bergametti (1995)) to significantly larger values in the order of up to 0.1 cm as determined by field studies of bare soil surfaces (Lancaster & Baas, 1998;Wolfe & Nickling, 1996) and suggested by a comparision study of King et al. (2005). The value E X can be considered as a measure for the distance how long the flow behind roughness elements is impacted by them and was estimated as E X = 10 cm by Marticorena and Bergametti (1995). However, based on further measurements, MacKinnon et al. (2004) suggested E X = 12,255 cm to account for non-solid roughness elements such as vegetation and was therefore applied here. In the context of wildfires, this selection may allow for a representation of incompletely burned or unburned vegetation in the fire vicinity. Following the concept of SALT, dust emission (in terms of the vertical dust flux) depends on the friction velocity * E u at the ground. We assumed that the momentum flux is constant within the lowermost levels of the boundary layer and therefore * E u can be extrapolated from the 10 m level. The formulation given by Equation 2 using the logarithmic wind profile assumes neutral atmospheric conditions. In contrast, in the vicinity of fires the lower atmosphere is characterized by heated raising air and thus quite unstable conditions. To take these effects on * E u into account, we applied a stability correction following Benoit (1977) and calculated the friction velocity * E u iteratively based on the lower atmospheric wind and temperature conditions derived from the LES following Equation 6 of Benoit (1977). Furthermore, the parameterization of SALT assumes that saltation is in equilibrium with the atmospheric forcing (Barchyn et al., 2014;Neakrase et al., 2016;Owen, 1964). This premise is more justified for large-scale processes than for small-scale phenomena dominated by turbulence such as the pyro-convectively driven winds. This might lead to a systematic overestimation of the total dust emissions. In the high resolution LES we are using here, * E u represents a grid cell of 10 10 2  m during 10 s output frequency. We consider this spatio-temporal resolution small enough to represent the characteristics of the pyro-convective flow with a relatively small variability within each grid cell/time step, but large enough to approximately approach a saltation equilibrium within small-scale processes (Neakrase et al., 2016;Spiga et al., 2016). Soil moisture can affect the onset and strength of the emission fluxes by altering the threshold friction velocity. Different concepts exist to account for those impacts. For SALT, we have applied the parameterization based on Fécan et al. (1999) who showed that the threshold friction velocity increases significantly with an increasing soil moisture content as the adsorbtion of water affects the cohesive forces of the soil particles. However, the onset of this increase depends on the soil type and its clay content, and depending on the specific soil, small soil moisture contents do not affect dust emission at all (Fécan et al., 1999).
A parameterization of direct aerodynamic dust entrainment with a special focus on convective-turbulent motions was established by Klose et al. (2014) based on work published in Shao (2012, 2013). CTDE is described stochastically to take the large variations of the inter-particle cohesive forces and the chaotic nature of the aerodynamic forces within a turbulent atmosphere into account. Hence, probability density functions represent both the soil-dependent cohesive forces and the aerodynamic lifting force as a function of the instantaneous momentum flux a E  at 10 m height. The CTDE scheme of Klose et al. (2014) also allows for corrections of the dust emission flux for the presence of roughness elements such as vegetation and soil moisture. The soil moisture correction of CTDE considers the effect of soil moisture on the inter-particle cohesive forces, that is, the capillary forces, and the adsorptive film that covers the soil particles (Fécan et al., 1999;McKenna-Neuman & Nickling, 1989), which both ultimately affect the strength of the cohesive forces and thus reduce dust emission strength (Klose et al., 2014). The drag partitioning parameterization from Raupach et al. (1993) was used by Klose et al. (2014) to account for the effect of roughness on the momentum flux. For better comparability with the SALT implementation, we here apply the parameterization of Marticorena and Bergametti (1995) as refined by King et al. (2005) too and express the impacts of roughness elements on the instantaneous momentum flux a E  instead on the friction velocity as it is done 7 of 20 for SALT. The CTDE scheme was calibrated and evaluated against field measurements (Klose et al., 2014) and applied to estimate the dust emission in dust devils (Klose & Shao, 2016). Currently, it only represents the aerodynamic emission of dust particles. The entrainment of larger particles, for example, sand-sized, which could enter into non-equilibrium saltation or even into suspension in strong convective updrafts is not yet included in the scheme. The focus of the parameterization on convective turbulence makes it an excellent test bed to investigate pyro-convectively driven dust emissions through direct aerodynamic entrainment. Both dust emission schemes, the formulation of SALT from Tegen et al. (2002) and the CTDE approach from Klose et al. (2014), were configured with similar but also idealized soil-surface conditions. We have defined a main scenario that is used as a reference with conditions favorable for emission and tested the robustness of this approach by further sensitivity studies. The baseline assumptions for the reference case include a soil moisture of zero, as a consequence of the fire-related dehydration of the soil's top layer due to both the heat impact of the fire and the desiccative effect of the fire winds, and a fully erodible surface. The latter assumes that the fire largely consumed the soil-covering vegetation and left bare soil behind. While that is sufficiently adequate within the actively burning area, in particular for fragile growths such as grass or stubble, it is quite idealized for regions downstream of the fire where the vegetation provides the fuel to keep the fire going, or for shrublands where the hardwood growths are expected to withstand the flames at least partly (Levin et al., 2012) and thus limits the erodible surface. Therefore, we have conducted some sensitivity studies that investigate the impact of an increased surface roughness on the dust emission flux of both schemes. Additional investigations cover the impact of soil moisture on the emission strength for the case that the soil surface may not dry out completely by the fire impact and a residual soil moisture remains. We have used sandy loam as the underlying reference soil type, a very common soil within landscapes affected by agricultural fires such as within the Sahel region (e.g., Chalbot et al., 2013;Dukes et al., 2018;Kavouras et al., 2012). However, agricultural fires are not limited to just one soil type, instead they can also occur in regions where other soil types dominate. Hence, we have also tested other soil types that are present in frequently fire-affected regions. Their particle size distributions were taken from Klose et al. (2021) and are shown in Figure 1. For use with SALT, the particle size distributions have been converted into the four particle size populations used by Tegen et al. (2002) and are given by Table 1. Dust emission caused by SALT requires the so-called sandblasting efficiency E , a measure that transforms the horizontal saltation flux into a vertical dust flux. The sandblasting efficiency was found to be highly dependent on the clay fraction of the soil and can be calculated as follows: The resulting values of E  for the applied soil types are given in Table 1 as well.
In a nutshell, to investigate fire-related dust emissions, we applied two fundamentally different dust emission approaches, SALT and CTDE, that each rely on different aerodynamic preconditions to generate  10.1029/2020JD034355 9 of 20 convergence is located. Here, the main fire updraft has formed and the atmosphere is dominated by the highly convective-turbulent upward motions, which boost a E  . These turbulence-dominated upwind conditions are also transported downstream so that strongly enhanced a E  values last for further 100 m before decreasing again quite rapidly toward an average value. In summary, the wind field quantities the two dust emission parameterizations build on show a quite distinct behavior in the surrounding of the fire. While the friction velocity * E u increases moderately, and this in particular within the actively burning area, the instantaneous momentum flux a E  is highly sensitive to the fire-induced pyro-convective forces and can increase by more than one order of magnitude. In presence of an ambient wind forcing, this mainly affects the areas downstream of the burning, while the immediate fire area is less affected. Consequently, the application of the corresponding dust emission parameterizations leads to qualitatively and quantitatively different results. Figures 2e and 2f contrast the spatial distribution of the time-averaged dust emission fluxes through SALT (Figure 2e) and of CTDE (Figure 2f). In order to focus on the general characteristics of the emission patterns and taking into account that the underlying soil-surface conditions are highly idealized, the figures represent dust emission fluxes that are normalized by the maximum value of each scheme. Figures 2e and 2f reveal the coincidence of the SALT dust emission fluxes with the areas of the enhanced * E u and that of the CTDE fluxes with increased values of a E  (cf. Figure 2a). These strong dependencies, which are a direct result from the different parameterizations, determine the relative strength of dust emission with respect to the actively burning fire area. The strongest emissions through SALT occur near the outflow edge of the burning area and largely affect the burning area itself. The strongest CTDE fluxes are located further downstream and occur well outside of the active fire area. This spatial displacement of the peak emission fluxes is particularly visible in the cross-sections shown in Figure 2g. SALT dust emissions peak within the actively burning area, while the peak CTDE fluxes are shifted roughly 50 m downstream. Thus, SALT can be expected to be a more important dust emission process within the burning area as a result of the accelerated inflow winds, while CTDE is more relevant in its outflow edge as there the turbulent-convective updraft motions dominate the wind field. However, the enhanced turbulent motions also lead to slightly enhanced a E  values within the burning area, which results in small CTDE fluxes upwind and within the first tens of meters into the burning area where the wind field is more dominated by horizontal rather than vertical motions, that reach less than 5% of their maximum strength occurring at the convergence.
In addition to the previous graphs, Figure 2h shows the same situation now using absolute values instead of the normalized values. Due to the idealized character of the underlying soil-surface conditions and the lack of measurement data for evaluation, these values should not be seen already as solid estimates of dust emission fluxes occurring in real fire situations. They rather can provide first insights into which dust emission process might dominate fire-related dust emissions in general, in particular if soil-surface conditions are highly susceptible. The data reveal that the peak dust emissions related to SALT exceed those by CTDE slightly. It is important to note that these estimates do not account for limitations in the availability of either sand particles for saltation or loose dust particles for direct aerodynamic entrainment. So both SALT and CTDE estimates are likely smaller in reality. Although the strength of dust emission via SALT is quite sensitive already to small changes of * E u due to the polynomial proportionality, the highly convective-turbulent environment of the fire-induced wind patterns appears to strongly support CTDE, too. The strength and the extent of the increased values of * E u and a E  depend of course on the dimension and intensity of the fire as well as on the strength of the ambient geostrophic wind velocity. Therefore, the given values can only be seen as a qualitative picture of how these variables behave in the situation of a fire-modulated wind field.    Figure 3 WAGNER ET AL.

Impact of Changing Soil-Surface Conditions on Dust Emission Fluxes
The behavior of the two dust emission schemes was investigated so far only with idealized and homogeneous soil-surface properties such as a soil moisture of zero and a completely fire-cleared surface within and also around the burning fire area. To better understand how a more realistic setup can affect the fire-related dust emission fluxes, further simulations were conducted in which different aerodynamic roughness lengths were assigned for the fire area and its surrounding or where the soil was not completely dry. Furthermore, in addition to the sandy loam soil, we investigated the fire-related dust emission potential of other soil types as shown in Figure 1 or given by Table 1, respectively, which are also common in fire-prone regions. All additional simulations are compared to our idealized reference scenario that was discussed in detail in Section 4.1. Therefore, the fire-related dust emission fluxes within this section are always normalized to those of the reference simulation. Figure 4 shows the dust emission fluxes obtained for different soil types. It can be seen that normalized fire-related dust emission fluxes vary by almost two orders of magnitude across the different soil types. For individual soil types, SALT and CTDE differ in behavior, too. The changes in the SALT dust fluxes are largely caused by the different values of the sandblasting efficiency and result in comparably small dust emission fluxes for the sandy clay loam and loamy sand soil due to their small clay content, while loam and clay loam emit more dust compared to our reference of sandy loam soil. For most of the soil types, the fire-related CTDE fluxes behave similar to those of SALT, however, the sandy clay loam and sand soils stand out as their effects are opposite in the SALT and CTDE results. Our results demonstrate that the fire-related dust emissions are sensitive to soil texture in both the SALT and CTDE scheme and that sandy loam, our reference soil type, seems to be intermediate in emissivity for both mechanisms. If a soil with a large fraction of sand-sized particles is additionally characterized by a topsoil layer composed of fine dust particles (e.g., the applied sand soil), SALT can become an even more effective dust generation process after the fine topsoil material is exhausted by CTDE. The sometimes strong fire-related updrafts may be able to emit also particles larger than 20 m. This cannot be tested in our current setup, but it would increase the dust emission fluxes for soil types with a larger coarse dust and sand fraction. Furthermore, the strong surface heating caused by the fire can modify the soil's particle size distributions as larger particles may disintegrate or oppositely form due to aggregation processes (e.g., Levin et al., 2012;Mc-Nabb & Swanson, 1990;Vermeire et al., 2005). This would also affect the inter-particle binding forces of the dust particles and thus the erodibility of the soil but the implications of these complex interactions cannot be addressed in the framework of this study.
The dependency of the normalized fire-related dust emission fluxes within the fire area on the soil moisture content is given by Figure 5. Small soil moisture contents do not impact dust emission fluxes as the adsorption of small amounts of water does not significantly alter the cohesive forces of the soil particles and thus dust emission is not affected (e.g., Fécan et al., 1999). When the soil moisture increases further, the dust emission fluxes decrease. In the case of SALT, soil moisture is accounted for in terms of an increased threshold friction velocity for the initialization of SALT, which starts to be effective at values of 0.07 3 m E 3 m E  . The reduction of dust emission occurs quite rapidly and already at a soil moisture content of around 0.1 3 m E 3 m E  the dust emission fluxes caused by SALT become largely negligible. In contrast, the reduction of the CTDE fluxes, whose parameterization does not consider the threshold friction velocity but quantifies the effect of an increasing soil moisture content on the cohesive/capillary forces, occurs much more gradually and emissions  Hence, even relatively wet soil surfaces would not completely impede fire-related dust emissions based on the tested parameterizations and parameters, in particular CTDE. In general, the combination of the fire heat and the increased evaporation caused by the strong (and hot) fire-related winds leads to a quick drying of the uppermost soil layer (Gillette, 1999;Tegen et al., 2002) and can easily reduce the soil moisture content to values close to zero. The assumption of a largely dry soil therefore appears plausible.
The presence of roughness elements can have huge impacts on the strength of dust emission fluxes. In the context of agricultural fires, such roughness elements might be vegetation remnants such as hardwood growths that survive the fire impact within the burning area or that have simply not yet been affected by the fire. The effect of roughness elements on dust emission is twofold: First, they cover a part of the soil surface preventing dust emission. Second, they consume a part of the momentum provided by the wind that is then not available for the mobilization of the soil grains. While the cover fraction reduces the erodible area in its sim- are strongly modified by pyro-convection. Our analyses have shown that aerodynamic conditions favorable for the initialization of SALT and CTDE exist in nearly all of the investigated fire setups representative for typical agriculture-related burning conditions. Both dust emission parameterization schemes produced considerable amounts of mineral dust in the vicinity of the burning area. However, important differences exist between the resulting SALT and CTDE fluxes within and outside of the burning area, for example, if the atmospheric wind forcing or the fire characteristics such as intensity or the size of the burning area change. These differences are caused by the fire-modulated near-surface aerodynamic situation, which is characterized by two dominant wind regimes. Close to the convergence zone, usually 16 of 20 at the downward end of the fire area, the atmosphere is dominated by convective-turbulent upward motions, favoring CTDE. Upstream of the convergence zone, mostly within the burning area, the accelerated horizontal winds blowing toward the convergence area are characterized by increased friction velocities that ultimately promote SALT. Of course, both wind regimes superimpose to some degree and are strongly controlled by the ambient wind forcing. Therefore, in particular within the actively burning fire area, the aerodynamic conditions are favorable for dust emissions due to both processes, however, with different intensities depending on the exact location and the fire and environmental wind conditions.
The scheme-normalized dust emission fluxes within the immediate burning fire area were found to be strongly dependent on variations of basic fire and wind field parameters. This provides useful starting points for further investigations, as the impact of the ambient wind velocity, the size of the burning fire area and the fire intensity on dust emission can be approximated for both dust emission processes by simple regression types. The parameters needed for a future implementation of pyro-convectively driven dust emission into aerosol-atmosphere models can be retrieved either from satellite observations monitoring the wildfire activity or are available from the model meteorology. While increasing fire intensities and larger burning areas generally lead to stronger dust emission fluxes independent of the dust emission process, the opposing impact of changing ambient wind speeds may have noticeable implications for real fires. Within the burning area, the strength of SALT decreases if the ambient wind becomes less intense, while CTDE increases. This is due to the fire updraft, which is located above the fire in weak-wind conditions, and which leads only to comparably small increases of *,A fire E u (relevant for SALT), but to strong increases of ,A fire a E  (relevant for CTDE). Consequently, a parameterization representing SALT does not produce significantly enhanced dust emissions in such weak-wind cases. However, taking the current formulation of CTDE as a baseline implies that even weak agricultural management fires, which are preferably ignited during low-wind conditions, can be linked to an injection of a considerable amount of mineral dust into the atmosphere. If this can be proven correct by further investigations, the potential of such prescribed agricultural fires as source of mineral dust emission might be substantial and can have a considerable atmospheric relevance. For agricultural fires, CTDE can be a very important dust emission process due to the pyro-convective forces inducing strong modulations to the instantaneous momentum flux . Even though the friction velocity is less strongly affected, during favorable conditions (i.e., particularly strong burns or fires with a strong background wind forcing) the contribution of SALT generated by gusts can be significant and may even dominate over CTDE. Therefore, both processes are important if fire-induced dust emissions are described.
So far, our study's design is idealized and the dust emission schemes are not yet adjusted to natural fire conditions, first and foremost with respect to the soil-surface conditions. These simplifications include an unlimited particle supply from a completely dry sandy loam soil surface whose properties were not altered by fire impacts. Furthermore, we assumed a complete removal of the vegetation cover within the burning area as a result of the combustion process. This may be sufficiently fulfilled for crops and grasslands, but remains highly idealized for shrubs. Here, the unburned vegetation remnants would act as further roughness elements within the flow and consume a larger part of the momentum provided by the pyro-convective forces that is then not available for dust mobilization. Sensitivity studies have shown that increasing roughness and a larger soil moisture content reduce dust emission for both processes, however, under conditions typical for a burning fire, this reduction should not prevent dust emission at all. On the other hand, in particular during situations where the aerodynamic forcing is already weak, even a comparably small increase of the surface roughness (e.g., due to vegetation debris or larger ash particles) might suppress dust emission more effectively.
Although the strongest emission fluxes are found within or close to the actively burning area as here the heat release of the fire has the strongest impacts on the aerodynamic conditions, the fire-generated turbulence can be transported further downstream by the ambient wind into a larger surrounding of the burning activity. While within the actively burning area, soil-surface conditions more susceptible to wind erosion can be assumed, the situation is different for the surrounding areas. However, if patches of bare soil are present, which is typical in many semi-arid landscapes such as these dominated by heterogeneously distributed shrubs (Dukes et al., 2018), dust emission can be generated here as well. Such situations would again facilitate CTDE as the interaction of the fire-related wind flow with the vegetation can increase the turbulence locally (Klose et al., 2014) and CTDE may still be an effective emission process under such conditions.

of 20
However, this is highly case-dependent and cannot be generalized although a probabilistic approach seems feasible. Nevertheless, according to the generally more appropriate soil-surface conditions within the burning fire area, the main focus for further developments should lay here initially.
An explicit representation of pyro-convectively driven dust emissions in large-scale aerosol-atmosphere models is necessary to quantify the regional and global relevance of this emission pathway, in particular with respect to other emission processes. Therefore, a parameterization approach is required that accounts for both dust emission processes and prioritizes their contribution depending on the specific fire and environmental conditions. It should also include the impacts of an incomplete vegetation combustion depending on the particular fire type and account for fire-specific soil properties. The latter might include, if available, locally modified particle size distributions typical for a fire-altered soil type and possible effects of the fire on soil-surface cohesion, which would affect both dust emission processes. While post-fire soil properties are known, the exact soil conditions during a fire remain a source of uncertainty. All those limitations make it difficult to really quantify the importance of this specific emission process yet. However, using the results of the current study, it is already feasible to derive a first rough estimate of the contribution of fire-related dust emissions on a global scale. Based on the strength of the simulated dust emission fluxes within the fire area and taking their variability resulting from different wind velocities and fire properties into account, they can provide a starting point and be scaled up by using data of the global burned area. The mean dust emission fluxes within the fire area lie in the order of roughly 1-20 g 2 m E would mean that such fires might contribute 0.06%-6% to the global dust emissions. Such fire-related dust emissions appear particularly relevant as they would largely occur outside of typical dust emitting regions such as the global dust belt. However, the limitations of our simulations mentioned above lead to a large range of uncertainty. This includes several aspects such as the emission of larger and heavier particles in stronger fires, or dust emissions in the surrounding of the burning area, but also a severe reduction of the dust emission fluxes due to a remaining soil coverage or further fire impacts on the soil properties. To reduce uncertainty, it is important that in addition to a further optimization of the dust emission parameterization also more detailed measurements and field studies are conducted that address dust emission in the context of fires and their concomitants, which would help to prove/manifest or to revise the findings of this idealized model study. Eventually, such a dust emission parameterization tuned for fire conditions enables us to determine the amount of mineral dust that is emitted through pyro-convection. This appears especially relevant as many kinds of environmental fires are supposed to increase in number, frequency, and intensity as a response to climate change (Bowman et al., 2017;Jolly et al., 2015;Westerling & Bryant, 2008). If the related fraction of fire-induced dust emissions can be estimated properly, the climate impacts of airborne mineral dust in general and that of dust mixed with combustion aerosol as a consequence of the fire-emission can be evaluated more precisely.

Data Availability Statement
The data used for the analysis are available at Zenodo, see Wagner (2021).