Quantifying Fire‐Driven Dust Emissions Using a Global Aerosol Model

Abstract

Vegetation fires have become increasingly recognized as a potential entrainment mechanism for mineral dust. However, the global importance of this emission pathway remains largely unknown. Based on previous LES investigations, we developed a parameterization that relates the dust emission potential of wildfires to observational data of the fire radiative power and further soil‐surface conditions. It was implemented into the aerosol‐climate model ICON‐HAM and simulations with and without the new emission pathway were conducted for the 10‐year period 2004–2013. Fire‐dust emissions can account for around 230 (190–255) Tg yr⁻¹, which represents around 18 (15–21) % of the total global dust emissions. These additional emissions originate largely from regions that are typically not known as significant sources of mineral dust. Locally, wildfires can enhance the presence of atmospheric dust particles and on the Southern hemisphere might even surpass other forms of dust emission. Highly dust active fire regions are identified in areas where burning grasslands create suitable emission conditions together with emissive soil types despite rather weak fires, for example, in Eastern Europe or the Central US. Fire‐dust emissions are subject to a strong seasonal cycle, mainly driven by the fire activity, following the hemispheric warm and dry seasons. Multi‐year comparisons with (dust) AOD observations revealed improvements due to the additional fire‐dust emissions, particularly in the most fire‐active regions on the Southern hemisphere. Nevertheless, further research and improvements of the parameterization are required to better classify the source areas and their variation with the changing climate and land use conditions.

Introduction

The capability of climate models to reproduce the global aerosol budget as realistically as possible is key for a correct estimate of the present, past, and future climate conditions. Aerosol particles affect virtually all atmospheric processes on Earth such as its radiation budget by scattering and absorbing solar and terrestrial radiation (e.g., Bellouin et al., 2020; J. Li et al., 2022; Tegen & Schepanski, 2018), the occurrence and lifetime of clouds (e.g., Z. Li, Rosenfeld, & Fan, 2017; Oreopoulos et al., 2020; Twomey, 1977) and their ability to form precipitation by acting as cloud condensation nuclei or ice nucleating particle (e.g., Andreae & Rosenfeld, 2008; Barry et al., 2021; Johnson et al., 2004; Mamouri et al., 2023; Roesch et al., 2021; Twohy et al., 2021). Additionally, aerosol particles can strongly affect land and marine ecosystems and their forms of life either due to fertilizing or toxic effects (e.g., Hamilton et al., 2022; W. Li, Xu, et al., 2017; Mahowald et al., 2011; Weis et al., 2022). The true scale of all of these impacts depends strongly on the aerosol particle considered. For example, mineral dust, a natural aerosol particle primarily emitted from bare soil surfaces such as deserts (e.g., Knippertz & Todd, 2012; Tegen & Schepanski, 2009), and black carbon (BC), released by various combustion processes of carbonaceous substances (e.g., Klimont et al., 2017; Wang et al., 2014), are two aerosol particles widely emitted over land surfaces with partly opposing impacts on the radiation balance, cloud microphysics, and ecosystems. While it is well established that wildfires and other forms of vegetation fires are a major source of atmospheric BC, it is less well known that such fires can also inject soil-dust particles into the atmosphere (e.g., Nisantzi et al., 2014; Palmer, 1981; Schlosser et al., 2017; Wagner et al., 2018).

The eligibility of vegetation fires to emit or resuspend soil-dust particles from the ground is well reported from observations and in situ measurements around the globe (e.g., Nisantzi et al., 2014; Palmer, 1981; Susott et al., 1991). Soil-dust particles were found within smoke plumes (Radke et al., 1991; Schlosser et al., 2017) and it has been shown that typical crustal-related elements have elevated concentrations within smoke plumes compared to fire-free background conditions (Gaudichet et al., 1995; Gill et al., 2024; Kavouras et al., 2012; Maenhaut et al., 1996; Perron et al., 2022). The driving mechanism are the fire-enhanced near-surface winds provoked by the pyro-convective forces resulting from the fire's heat release, and supported by an at least partial removal of the ground-covering vegetation due to the combustion process (Palmer, 1981; Susott et al., 1991). Although the majority of the observations of fire-related dust emissions suggests that the process is particularly frequent in semi-arid regions (e.g., Alves et al., 2010; Kavouras et al., 2012; Maudlin et al., 2015; Nisantzi et al., 2014), where landscapes are dominated by grasslands and shrublands, evidence is increasing that also other biomes, such as the Siberian tundra or tropical savannas, could be at least partly related to fire-dust emissions (Das et al., 2023; Mamouri et al., 2023; Pio et al., 2008). But also croplands that are in many regions often deliberately ignited as part of the usual farming routine were found to be a source of soil-dust particles (Chen et al., 2023). While direct measurements are still rare, increased post-fire dust emission activity (Dukes et al., 2018; Ravi et al., 2012; Yu & Ginoux, 2022) suggests that appropriate emission conditions exist also during the fire, if the fire is sufficiently strong to alter the near-surface winds and enable soil particle mobilization, which is the crucial precondition for dust emission in general.

Wildfires are a widespread global phenomenon affecting nearly all landscapes ranging from highly anthropogenic-influenced, cultivated landscapes to natural ecosystems such as tropical savannas or high-latitude tundras (e.g., Flannigan et al., 2009; Giglio et al., 2013; Tang et al., 2021). In the state-of-the-art aerosol-climate models, fires are only considered as a source of combustion products but not as a source of mineral dust. However, missing a component of the fire emissions could lead to a significant underestimation of their atmospheric impacts. Due to their opposing impacts on radiation, clouds, and ecosystems, it is important to assess not only the total aerosol load emitted by fires, in terms of the aerosol optical depth (AOD), but also the correct allocation to combustion and mineral dust particles becomes of high priority. This accounts all the more as fires, although always an integral part of Earth's history (Earl et al., 2015) and naturally sparked due to lightning strikes in dry regions, have become a strongly man-made disturbance of the Earth system (e.g., Achard et al., 2008; Nagy et al., 2018; Ward et al., 2018). Today, depending on the region, up to nearly 90% of all wildland fires are related to anthropogenic activity (Balch et al., 2017), provoked both intentionally for farming purposes, land modification or simply by arson, and unintentionally as a result of carelessness (Earl & Simmonds, 2018; Earl et al., 2015). On top of that, the escalating climate crisis acts as a reinforcing mechanism making fire occurrence both more likely and increase their typical strength and extent (e.g., Huang et al., 2015; Senande-Rivera et al., 2022) as global temperatures are expected to increase further during the upcoming decades (IPCC, 2023), making droughts more likely and more severe (Cook et al., 2018; Loukas et al., 2008). Therefore, the correct assessment of all fire emissions, including the role of co-emitted soil-dust particles, within aerosol-climate models becomes even more important.

The representation of the fire-driven dust emission path in numerical models requires the development of a dedicated parameterization that accounts for the specific aerodynamic and soil-surface conditions within and around a fire. To tackle the dust emission potential of fires, Wagner et al. (2018) have already proven by using a large eddy simulation (LES) model that the necessary aerodynamic preconditions for the mobilization of soil-dust particles exist within the wind fields around idealized fire setups mimicking typical medium-intense ground fires occurring in grass- and croplands. In a follow-up study, Wagner et al. (2021) coupled the fire-modulated winds with two different dust emission schemes, parameterizing dust entrainment based on saltation bombardment (Marticorena & Bergametti, 1995; Tegen et al., 2002) and convective-turbulent dust entrainment (Klose et al., 2014). This gave first quantitative insights into the dust emission behavior of such fires, including their dependence on fire strength and soil-surface conditions in the burning area. A key result was that the specific fire-related aerodynamic preconditions cannot be sufficiently described while focusing only on one dust emission process.

A first attempt to merge the ideal findings of Wagner et al. (2021) into an aerosol-chemistry model was conducted by Menut et al. (2022), who simulated fire-related dust emissions within Europe by a simplified approach using the regional model CHIMERE. To account for the fire-induced wind enhancement, an offset to the local background wind velocity at a fire location was applied. Despite regional improvements, Menut et al. (2022) could not identify a robust advancement of dust aerosol optical depth (DAOD) comparison by including fires as an additional dust source. Nevertheless, their study revealed that the focus on only classic wind-driven dust emissions often fails to reproduce a coherent picture of the atmospheric dust load. The consideration of the fire-driven dust emission pathway could be one missing process that helps to narrow the discrepancy between models and observations, which accompanies the dust model community already for a while (Cakmur et al., 2006; Zhao et al., 2022).

Although Hamilton et al. (2022) made an attempt to quantify the global dust emissions related to active and post fire activity, the share of fire-dust emissions on the total atmospheric dust load is still an open question. Furthermore, it is largely unknown where the fire-dust emission process is particularly dominant. This study aims to provide for the first time a global overview of fire-dust emissions by implementing the process into an aerosol-climate model. For this purpose, we have chosen the recently coupled model system of ICON-HAM (Salzmann et al., 2022) as the host model that is described in detail in Section 2 while an introduction of the different data sets is given in Section 3. Within the present study, we pursued an approach that builds directly upon the results of Wagner et al. (2021). The parameterization is presented in Section 4, followed by an overview of the conducted simulations in Section 5. The parameterization is analyzed for a 10-year period (2004–2013) and compared against observations of the atmospheric dust load as shown in Section 6, before the results of the global fire-dust emissions fluxes including seasonal and regional analyses are presented in Section 7. Limitations of the approach are critically discussed in Section 8, before a conclusion and short outlook on future research activities closes the paper.

The Aerosol-Climate Model ICON-HAM

ICON-HAM is a recently coupled model system consisting of the global weather and climate model ICON (ICOsahedral Nonhydrostatic atmosphere) and the aerosol module HAM (Hamburg Aerosol Module). It was first coupled using the version of ICON-A (Giorgetta et al., 2018) by Salzmann et al. (2022) and can be considered as the successor of ECHAM-HAM whose latest version ECHAM6.3–HAM2.3 was published by Tegen et al. (2019).

The Atmospheric Model ICON

The ICON model was developed by the German Weather Service (DWD, Deutscher Wetterdienst) together with the Max-Planck-Institute for Meteorology and already successfully introduced as the everyday German weather forecast model (Zängl et al., 2015). Furthermore, it was coupled to a variety of different sub-models to investigate specific components of the Earth system in more detail (e.g., Dipankar et al., 2015; Jungclaus et al., 2022; Rieger et al., 2015). For the propose of climate simulations, ICON-A was developed as the climate model version of ICON (Giorgetta et al., 2018). It uses the so-called Max-Planck-Institute physics package (Stevens et al., 2013) originating from the ECHAM6 general circulation model and was chosen as the new host model for HAM by Salzmann et al. (2022).

ICONs dynamic core includes local mass conservation and a mass consistent transport of tracers (Zängl et al., 2015). In contrast to many other models, the horizontal grid of ICON results from the projection of an icosahedron onto a sphere and thus differs from the ordinary latitudinal-longitudinal grid spacing. The resulting projection gets further divided multiple times, eventually forming triangular grid cells and thus avoiding singularities at the poles. The horizontal grid is now defined by a so-called RnBm grid. It combines the initial root division (R) in n sections with m bisection steps (B) whereby for realistic atmospheric simulations R is typically set to 2 and the resolution of the horizontal grid only defined by the number m of bisections B (Crueger et al., 2018; Giorgetta et al., 2018). An increasing number of bisections relates to a finer horizontal resolution. According to Salzmann et al. (2022), we use ICON-(HAM) in the R2B4 configuration, which results in 20,480 grid cells with a mean horizontal grid spacing of around 158 km. This resolution can still simulate baroclinic waves and the related eddy fluxes reasonably well (Giorgetta et al., 2018), while it keeps the computational costs at a manageable level for multi-year simulations. The vertical grid is built up on terrain-following hybrid sigma coordinates (Leuenberger et al., 2010). In total, 48 levels are used with a layer thickness of 40 m above ground that is steadily increased with height to the toplayer at 83 km (Giorgetta et al., 2018).

The Land Surface Model JSBACH

To account for land-atmosphere interactions, ICON-HAM is intrinsically coupled to the land-biosphere model JSBACH (Jena Scheme for Biosphere-Atmosphere interactions Coupled in Hamburg; Raddatz et al. (2007)). For example, JSBACH provides monthly resolved information of the leaf area index and the vegetation-free land areas derived from the US Geological Survey Global Land Cover Characteristics Database (Hagemann, 2002; Reick et al., 2013). Furthermore, JSBACH provides fractional information about land surface properties such as the presence of a vegetation coverage and the different biomes (land cover classes) within a grid cell, and provides their characteristic roughness length. The current version of ICON-A, that was coupled to HAM and is used in the present study, is only coupled to a simplified “lite” version of JSBACH4 that provides only static fields of certain properties of the land surface (Salzmann et al., 2022).

The Aerosol Modul HAM

The aerosol module HAM is an intrinsic part of the aerosol-chemistry model HAMMOZ, whereby MOZ comprises of the atmosphere-chemistry model MOZART (Model for OZone And Related chemical Tracers). Within the current study only the HAM component is used, that is, without the MOZ part. Instead, a simplified chemistry is used that relies on prescribed oxidant fields. ICON was coupled to the most recent version of HAM (v3.2), which was introduced and successfully evaluated in the framework of ECHAM6.3-HAM3.2 (Tegen et al., 2019). Originally, HAM was developed by Stier et al. (2005) and further improved by Zhang et al. (2012). It handles the emission, transport, and deposition of the most abundant atmospheric aerosol particles, namely sea salt, mineral dust, black and organic carbon as well as the related precursor gases from both natural and anthropogenic sources (Salzmann et al., 2022).

HAM uses prognostic variables to describe the emission, transport, alteration, and finally removal of the aerosol particles and its precursors. The interaction between aerosol, radiation, and clouds is parameterized using a radiation scheme, that accounts for the aerosol optical properties, and a two-moment cloud microphysic scheme (Salzmann et al. (2022) and references therein). In general, the aerosol microphysics within HAM can be described either by a modal approach represented by the M7 scheme (Vignati et al., 2004) or by the SALSA (Sectional Aerosol module for Large Scale Applications) scheme pursuing a bin approach (Kokkola et al., 2008). In the current study, we use the more common modal approach M7 that is characterized by the superposition of seven log-normal modes; four for hydrophilic aerosol particles including nucleation, Aitken, accumulation, and coarse mode, and three for hydrophobic aerosol ranging from Aitken to coarse mode, with fixed standard deviations for each mode (Vignati et al., 2004). All modes are characterized by an aerosol number concentration, their medium radius, and standard deviations (Tegen et al., 2019). The different aerosol species are internally mixed within each mode. Mineral dust can thereby be assigned to the accumulation mode for particles smaller than 0.5$$\,{\upmu }$$m and to the coarse mode for larger dust particles, both for the hydrophilic and hydrophobic category, while super-coarse particles are neglected due to their short atmospheric lifetime (Stier et al., 2005). The aerosol particles can be repartitionated between the modes, for example, due to growing processes, and can also change their hydrophobicity during the process (Salzmann et al., 2022; Stier et al., 2005). Within each mode, the mass concentration of each species is treated as a prognostic variable.

The emissions of the natural aerosol particles, such as sea salt and mineral dust, are mainly governed by the atmospheric conditions, namely the near-surface wind speed. Emission fluxes from biomass burning and all other anthropogenic sources are taken from ACCMIP (Atmospheric Chemistry and Climate Model Intercomparison Project; Lamarque et al. (2010)), which contains monthly mean emission fluxes with a horizontal resolution of 0.5$${}^{\circ}$$ between 1850 and 2000. For the more recent past from 2000 onwards and continuing into the future until 2100, one can choose from the different RCP (Representative Concentration Pathway) emission scenarios RCP2.6 to RCP8.5 describing possible trends of $${\text{CO}}_{2}$$ and other greenhouse gas emissions and thus impacting the strength of future global warming. Alternatively, emissions of species that originate from biomass burning, such as CO, OC, and $${\text{SO}}_{2}$$, can also be obtained from the Global Fire Assimilation System (GFAS), retrieved from satellite measurements of the fire radiative power (FRP); see details in Section 3.1. The emissions caused by biomass burning activity are injected into different vertical levels following the results of Val Martin et al. (2010). This means that three quarter of the emissions are injected into the planetary boundary layer (PBL), where they are equally distributed, while the remaining quarter is injected into the first (17%) and second (8%) level above the PBL.

Aerosol particles are finally removed from the atmosphere by gravitational sedimentation using the Stokes settling velocity as function of their median size (Seinfeld & Pandis, 2006; Stier et al., 2005), by dry deposition based on the resistance approach of Wesely (1989), and through wet deposition (Croft et al., 2009, 2010).

The ICON-HAM simulations are based on the Atmospheric Model Intercomparison Project (AMIP) experiment whose aim is a well balanced top-of-atmosphere energy budget to allow for realistic climate and Earth system modeling (Giorgetta et al., 2018). This includes the use of monthly fields of sea surface temperatures, sea ice conditions, ozone concentration, as well as annual concentrations of the most important atmospheric greenhouse gases, such as $${\text{CO}}_{2}$$ and $${\text{CH}}_{4}$$. Furthermore, spectral solar irradiance at the uppermost model level is provided. Other than that, the model is not nudged and runs freely. This means that the actual weather conditions and with that also the emissions of the wind-driven aerosol species deviate from the reality on a day-by-day comparison. However, longterm averages are expected to be sufficiently well captured due to the application of the boundary data.

Mineral Dust Emissions Within ICON-HAM

The emission of mineral dust within ICON-HAM is largely based on the emission scheme introduced by Tegen et al. (2002). Possible additional refinements concerning East Asian dust sources (Cheng et al., 2008) and a satellite-based approach using a preferential dust source map for the Saharan desert (Heinold et al., 2016; Schepanski, Tegen, Todd, et al., 2009) can be applied. This altogether represents the latest model version with regard to dust emission as it was used in ECHAM6.3-HAM3.2 as well (Tegen et al., 2019). In general, dust emission is driven by the interplay of the model's online calculated 10 m-wind velocity and the soil characteristics. The latter comprise soil humidity and snow coverage in potential dust source areas, which are themselves characterized by a constantly very low roughness length of 0.001 cm and do not change over time (Tegen et al., 2019). Thus, potential dust sources that could form (temporally) due to landcover modifications are not acknowledged. The actual dust emission process follows the physics-based formulation of saltation bombardment according to Marticorena and Bergametti (1995). It relies on the concept that the horizontal wind exerts a momentum flux to the surface that is characterized by the friction velocity (Kok et al., 2014; Lu & Shao, 1999; Marticorena & Bergametti, 1995). For every soil type exists a threshold friction velocity, at which the horizontal wind can overcome the cohesive forces of the soil-dust particles and saltation starts (Gillette & Walker, 1977; Kok, 2011; Marticorena & Bergametti, 1995). This horizontal saltation flux determines now the strength of the vertical dust flux, the actual emission flux, namely the injection of dust particles into the atmosphere, that depends on the soil type and is specified in ICON-HAM after Tegen et al. (2002) using the so-called sand-blasting efficiency.

The discrepancy between the rather coarse model resolution, that provides the 10 m wind velocity, and the subgrid processes governing dust emission represents a challenge for global dust modeling (Chappell et al., 2023; Pérez et al., 2011). To acknowledge this structural problem and in order to bring the model results in better accordance with observations, the dust emission fluxes of the model can be corrected, or, in other words, tuned. Therefore, a correction factor is introduced that can be applied to the threshold friction velocity if the model in its current resolution produces too much or too little dust (Albani et al., 2014; Chappell et al., 2023). This correction factor can be set globally to an uniform value or it can be differentiated for individual regions to acknowledge the different dust emission regimes there that are not always similarly affected by the model resolution. Within the latest version of ECHAM-HAM, Tegen et al. (2019) used regional tuning factors to bring the model's dust emission fluxes in better accordance to the model intercomparison study by Huneeus et al. (2011). In contrast, Salzmann et al. (2022) changed it back to a globally uniform correction factor of 0.9, which was close to the tuning factor of 0.86 that was used in a former version of ECHAM-HAM. However, during the coupling process they only aimed for a globally similar dust burden between the new ICON-HAM and its precursor and did not focus on the local emission strength.

A comparison between the ECHAM-HAM simulations of Tegen et al. (2019) and ICON-HAM after Salzmann et al. (2022) revealed a reduction of the global dust emission fluxes. Global dust emission fluxes of around 1,200 Tg yr⁻¹ in ECHAM-HAM, that are in good agreement with the median value of the multi-model study by Huneeus et al. (2011), decreased to values of just 600–900 Tg yr⁻¹ in ICON-HAM. This resulted in a decrease of the dust burden particularly over the so-called “dust belt” spanning from the Sahara toward East Asia, while an increase was identified off coast of West Africa. A similar pattern is present if the ICON-HAM results are compared to other aerosol models or dust retrievals (see validation in Section 6), pointing to a systematical mismatch of the ICON-HAM dust emissions in this particular version. However, it is crucial that the model maps the wind-driven dust emissions as accurately as possible (at least on a continental scale) before the additional fire-dust emissions are implemented as the different forms of mineral dust are not distinguishable after emission/injection anymore. Thus, we rejected the application of a globally uniform tuning factor and introduced regional factors again to compensate for the general underestimation of the total dust load as well as to overcome the regional differences, in particular over North Africa and the adjacent territories. Furthermore, the region around the Bodélé depression as one of the world's most active dust sources was isolated from the North African domain and treated differently. The related values of the applied regional tuning factors are given in Table 1. Further adjustments were introduced to the sand-blasting efficiency of the coarse-mode soil. With these adjustments we were able to increase the global dust emission fluxes by around 200–300 Tg yr⁻¹ as well as to reduce the regional mismatches. However, not all differences could be fully resolved, as some of those likely result from the unnudged ICON simulations or the model's general meteorology. A more detailed discussion and classification of the wind-driven dust emission fluxes retrieved with our applied settings is provided in the results section (Section 7.2).

Table 1 Tuning Factors Applied to the Threshold Friction Velocity for the Wind-Driven Dust Emissions of ICON-HAM

Region	Coordinates	Tuning factor
Northern Africa	20$${}^{\circ}$$W–35$${}^{\circ}$$E, 0–30$${}^{\circ}$$N	0.6
Middle East	35–60$${}^{\circ}$$E, 10–35$${}^{\circ}$$N	0.65
Asia	60–135$${}^{\circ}$$E, 10–80$${}^{\circ}$$N	0.8
Australia	110–155$${}^{\circ}$$E, 10–40$${}^{\circ}$$S	1.1
Bodélé	15–20$${}^{\circ}$$W, 15–20$${}^{\circ}$$N	0.85
all other regions	–	0.9

Applied Data Sets

GFAS

The Global Fire Assimilation System (GFAS), as a powerful data set for estimating biomass burning emissions on a global scale, comprises the emission fluxes of 40 gaseous and aerosol species as well as the underlying values of the FRP. GFAS is already partially used in the framework of ICON-HAM to map the combustion emissions linked to wildfires and other biomass burning activities. Its data are available since beginning of 2003 with a temporal resolution of 1 day and in the most recent version with a spatial resolution of 0.1$${}^{\circ}$$.

GFAS is based on measurements of the MODIS (Moderate Resolution Imaging Spectroradiometer) instruments onboard of the two polar orbiting satellites AQUA and TERRA and adjusted by SEVIRI (Spinning Enhanced Visible and Infra-Red Imager) data from the geostationary MSG (METEOSAT Second Generation) satellite where possible (Kaiser et al., 2012). They measure the thermal radiation emitted from fires within the 3.9 $${\upmu }$$m and 11 $${\upmu }$$m wavelength channels per pixel whereby a pixel covers roughly 1 km². These measurements are used to derive the MOD14 fire product (Giglio et al., 2006; Justice et al., 2002) that finally provides information on the FRP, which in turn can be linked to a dry matter combustion rate and eventually to fire emission rates (Kaiser et al., 2012; Wooster et al., 2005). The polar orbit of the two satellites allows for a global coverage but in turn limits the sampling frequency of possible fires to usually 3 or 4 observations per day (Kaiser et al., 2012). These observations are suitable enough to derive a daily average value of the FRP and thus a daily emission rate without assuming a specific behavior or a diurnal cycle of the fires. This simplification, together with the limited temporal coverage, can introduce an inaccuracy to the representation of fires. In particular small and weak fires might be missed, but Kaiser et al. (2012) found an overall satisfactory presentation of the diurnal variability. Furthermore, the GFAS retrieval is characterized by an improved detection threshold of small fires compared to other approaches, that are based on burnt area, and therefore should also be able to detect agricultural fires. Gaps in the observations (e.g., due to cloud coverage or thick smoke plumes) are corrected by an assimilation of earlier observations. Dubious observations of the FRP, that result from other heat sources, such as volcanos or industrial activities, are largely filtered out to ensure that the data set covers only biomass burning activity linked to fires (Kaiser et al., 2012).

MIDAS

The MIDAS (ModIs Dust AeroSol) data set of Gkikas et al. (2021) provides the dust aerosol optical depth (DAOD) at 550 nm and is to date available for the 15-year period 2003–2017 with a spatial resolution of 0.1$${}^{\circ}$$ and a temporal resolution of 1 day. It combines satellite observations of the AOD, retrieved from the MODIS instruments onboard of the AQUA satellite, with MERRA (Modern-Era Retrospective analysis for Research and Applications) reanalysis data of the DAOD-AOD-ratio to derive a quasi-observational data set of the DAOD on a global scale. The validation of the MIDAS DAOD has shown a good agreement with DAOD fractions derived from AERONET measurements with an overall small positive bias of only 2.7%. Thus, it provides a very suitable tool for model evaluations of the atmospheric dust load on a global scale where direct measurements of the DAOD are rare or simply not available, in particular for our simulation period 2004–2013.

AERONET

The AERONET (AErosol RObotic NETwork) program is a ground-based remote sensing network that uses sun-photometers to derive, among others, direct measurements of the AOD (Holben et al., 1998; Holben et al., 2001). AERONET provides continuous but cloud-screened long-term observations by several collaborators all over the world using a standardized measurement technique (Dubovik et al., 2000). The network was constantly expanded over the last 30 years and up to 600, at least at times operational, measuring stations guarantee a nearly global coverage (see also Figure 4). This makes the AERONET data to a very useful tool for the validation of modeled aerosol properties such as the AOD. Here, we use the quality-assured Level 2 AOD product from the current version 3.0 database.

Fire-Dust Parameterization

The aim of the current study is the investigation of dust emissions that are related to vegetation fires on a global scale using ICON-HAM. As a new emission process, that is now implemented for the first time into an aerosol-climate model, the parameterization for this specific purpose needs to be developed from scratch. This is done by an extension of the current dust emission module of HAM based on the previous studies of Wagner et al. (2018, 2021). In contrast to the “classic” purely wind-driven dust emission, fire-induced dust emission requires first and foremost additional input data about global fire occurrence and fire strength. This is gained from the GFAS data set that provides a measure of the fire strength by the FRP (see Section 3.1). First, the FRP fields were interpolated conservatively to the R2B4 resolution of the applied ICON grid, which is substantially coarser than the available GFAS resolution of 0.1$${}^{\circ}$$. The conservative interpolation assures that the total FRP within the model domain is conserved, meaning individual fires affect now a larger area (at least one grid cell) and the mean value of the FRP therein is accordingly reduced. Nevertheless, too small values of the FRP (smaller than 0.005 W m⁻²), as a possible result from averaging, are not considered for fire-dust emission. While GFAS provides a global coverage, its temporal resolution is limited to one data point per day. This inhibits the implementation of a diurnal cycle of fire activity and thus fires are considered to act as a potential source of dust at every time step as long as the FRP per grid cell is above the mentioned threshold (at least 24 hr).

Using offline dust modeling, Wagner et al. (2021) have already derived a dependency of the fire-dust emission fluxes on fire strength by applying two conceptually different dust emission approaches describing saltation bombardment and direct entrainment via turbulent-convective motions. It was shown that fire-dust emission depends not only on the fire strength, but that they are also sensitive to the ambient wind conditions and soil-surface properties such as soil moisture, surface roughness, and the soil type that is exposed to the fire-induced winds. While the impact of the soil moisture in the fire vicinity is negligible as the fire heats and dries out the top soil layer, the other factors can affect the strength of fire-dust emissions drastically and thus need to be considered in the development of the parameterization. It is well known that fires can have significant impacts on the particle size distribution (PSD) of a soil surface as larger particles can break apart and smaller particles can agglomerate (Blank et al., 1996; Chalbot et al., 2013; McNabb & Swanson, 1990). However, due to the large variability of fire-induced changes, that also depend on the soil type, and a lack of precise data about fire-affected PSDs, we decided to apply the strategy of Wagner et al. (2021) and calculated the fire-dust fluxes based on the soil types that are already implemented in ICON-HAM in the wind-driven dust emission scheme. As these soil types partly differ from those investigated in Wagner et al. (2021), we repeated the offline dust emission LES analysis for the five main soil types available in ICON-HAM namely coarse, coarse-medium, medium, medium-fine, and fine mode soil (Tegen et al., 2002). Based on these results, it was possible to derive an empirical relation of the theoretical fire-dust emission flux for both emission processes in response to the FRP for each soil type i and the different particle size classes as shown in Wagner et al. (2021). Thus, it exists a set of equations that relates a given FRP to a theoretical fire-dust emission flux, both for saltation $${F}_{\text{fdust,salt},i}$$ and turbulent-convective entrainment $${F}_{\text{fdust,conv},i}$$, in the form of 1 $${F}_{\text{fdust,salt},i}(D)={a}_{i}(D)\hspace*{.5em}FR{P}^{2}+{b}_{i}(D)\hspace*{.5em}FRP$$ 2 $\begin{align*}\begin{array}{lll}\hfill & \hfill & \hfill \\ \hfill {F}_{\text{fdust,conv},i}(D)& =\hfill & {c}_{i}(D)\hspace*{.5em}FRP\hfill \end{array} \end{align*}$ with the coefficients $${a}_{i}(D)$$, $${b}_{i}(D)$$, and $${c}_{i}(D)$$ that are given for all soil types i and particle size ranges D in Table A1 in Appendix A. These fire-dust emission fluxes can be further modified by the impact of the 10 m wind velocity $$\vert {v}_{10\mathrm{m}}\vert $$ at the fire location, which is taken from the ICON meteorology. The following relations hold: 3 $${F}_{\text{fdust,salt},i}(D)={d}_{i}(D)\hspace*{.5em}\vert {v}_{10\mathrm{m}}{\vert }^{{e}_{i}(D)}$$ 4 $\begin{align*}\begin{array}{lll}\hfill & \hfill & \hfill \\ \hfill {F}_{\text{fdust,conv},i}(D)& =\hfill & {f}_{i}(D)\hspace*{.5em}\mathrm{exp}\left(-{g}_{i}(D)\cdot \vert {v}_{10\mathrm{m}}\vert \right)\hfill \end{array}\end{align*}$ with the coefficients $${d}_{i}(D)$$, $${e}_{i}(D)$$, $${f}_{i}(D)$$, and $${g}_{i}(D)$$ as specified in Table A1. The dynamic impact of $$\vert {v}_{10\mathrm{m}}\vert $$ can substantially alter the effectiveness of the fire-driven dust emissions but was restricted to wind velocities between 0.5 and 6 m s⁻¹ to avoid extreme values that were not tested by the LES simulations of Wagner et al. (2018). As dust emission via saltation and turbulent entrainment are not fully independent from each other (Klamt et al., 2024; Klose et al., 2014; Stout & Zobeck, 1997), namely the other process is supposed to be at least partly accounted for within the respective scheme, a simple addition of both dust emission fluxes $${F}_{\text{fdust,salt}}$$ and $${F}_{\text{fdust,conv}}$$ would likely result in a substantial overestimation. However, due to their sometimes opposing behavior if applied to fire-affected wind fields (Wagner et al., 2021), the focus on just one of the emission processes likely under- or overestimates crucial emission scenarios depending on certain fire and wind conditions. Thus, a combination of both processes appears to be advised. Due to the lack of further evidence, we propose for simplicity an equal contribution of both dust emission processes to the total fire-dust emission flux $${F}_{\text{fdust}}(D)$$ in the following form: 5 $${F}_{\text{fdust},i}(D)=\frac{1}{2}\hspace*{.5em}{F}_{\text{fdust,salt},i}(D)+\frac{1}{2}\hspace*{.5em}{F}_{\text{fdust,conv},i}(D)$$

If further investigations figure out that a different allocation of both dust emission processes is more realistic, their individual contributions can be easily adjusted differently if desired. The resulting fire-dust emission fluxes F_fdust,i for all five soil types of ICON-HAM are shown as a function of the FRP in Figure 1. The additional impact of the ambient wind velocity following Equations 3 and 4 is indicated by the envelop of the solutions.

[IMAGE OMITTED. SEE PDF]

In addition to the soil type, the kind of burning vegetation can heavily impact the dust mobilization potential of fires too. Vegetation is well known for preventing dust emission by shielding the surface from the aerodynamic forces of the wind (e.g., Marticorena et al., 1997; Tegen et al., 2002; Wolfe & Nickling, 1996). If patches of bare soil are present within a sparsely vegetated landscape, the vegetation can act as roughness element and reduce the aerodynamic stress exerted to the surface and thus reduce or prevent dust emission (Raupach et al., 1993). This accounts all the more for fire-induced dust emission as here the vegetation is an integral ingredient for the fire. On the other hand, fires have also huge impacts on the vegetation itself as they can consume the ground-covering plants as fuel and open up new areas of bare soil (Pérez-Cabello et al., 2006; Ravi et al., 2012; Yu & Ginoux, 2022). While fragile growths, such as grass and crops, are likely to be effectively burned away, more fire-resistant hardwood growths, such as shrubs and trees, can remain as obstacles at the ground even during the fire. This affects the dust emission potential of fires twofold: it covers a part of the soil surface that is not available for emission and it increases the total roughness of the surface. The latter point is accounted by implementing an empirical roughness correction factor $${C}_{r}$$ based on the findings of Wagner et al. (2021) that is given by: 6 $${C}_{r}=0.0108\cdot {\overline{z}}_{0,\mathrm{v}\mathrm{e}\mathrm{g}}{[\mathrm{c}\mathrm{m}]}^{-0.686}$$ with $${\overline{z}}_{0,\mathrm{v}\mathrm{e}\mathrm{g}}$$ being the mean roughness length of the grid cell that results from the contributions of all fire-affected landcover types. The values of $${z}_{0,\mathrm{v}\mathrm{e}\mathrm{g}}$$ depend on the burning vegetation class and are taken from the JSBACH biome data set and are given in Table 2. For some landcover classes the vegetation density and thus their roughness length depends on the leaf area index (LAI), whereby two cases with a LAI below and above 1 are distinguished. The roughness correction following Equation 6 results in a strong reduction of the dust emission fluxes that can be described by a logarithmic decay. For example, if the surface roughness is increased to just 0.1 m as a typical mean value for a completely vegetated surface (Foken & Napo, 2008), the emission strength declines to less than 10% of its original value. The second effect of vegetation on dust emission, the incomplete burning process, is accounted by the introduction of an emissive area $${A}_{\mathrm{emi}}$$ that depends on the burning landcover class and is given in Table 2 as well. For fragile growths, such as grass and crops, the whole fire-affected area is considered to be prone to fire-dust emission, while for shrub-dominated landscapes this value is set to only 50% to account for unburned remnants that can remain in the fire area and prevent the emission there. Fires affecting ecosystems dominated by trees are even further limited and only 25% of the area can act as a possible dust source. Here, a substantial part of fires might only affect the crowns of trees and limit the dust mobilization potential at the surface and/or the surface itself is sheltered partly by a layer of dead or living biomass. Both processes together result in a (strong) reduction of the fire-dust emission flux obtained by Equation 5, so that the fire-dust emission flux $${F}_{\text{fdust,cor}}$$ corrected for the impacts of vegetation, is given by: 7 $${F}_{\text{fdust,cor},i}(D)={C}_{r}\cdot {\overline{A}}_{\text{emi}}\cdot {F}_{\text{fdust},i}(D)$$

Table 2 Vegetation-Related Parameters Applied to the Fire-Dust Emission Parameterization for the Different Landcover Types of ICON-HAM

Landcover type	Roughness length $${z}_{0,\mathrm{v}\mathrm{e}\mathrm{g}}$$ (m) for LAI $${\le} $$ 1	Roughness length $${z}_{0,\mathrm{v}\mathrm{e}\mathrm{g}}$$ (m) for LAI $${ >} $$ 1	Fraction of emissive area $${A}_{\text{emi}}$$
Tropical evergreen trees	2.0	2.0	0.25
Tropical deciduous trees	1.0	1.0	0.25
Extra-tropical evergreen trees	1.0	1.0	0.25
Extra-tropical evergreen trees	1.0	1.0	0.25
Raingreen shrubs	0.3	0.5	0.5
Deciduous shrubs	0.3	0.5	0.5
Grass	0.005	0.05	1
Pasture	0.005	0.05	1
Crops	0.005	0.05	1

For biomes where the roughness length is low and a large part of the surface considered emissive, the fire-dust emission fluxes are only slightly reduced, while fire-dust emission from forest-like biomes is now largely suppressed.

Within ICON-HAM, the soil type and the landcover type are available as fractions per grid cell. Both information are not independent from each other but only available per grid cell, which could introduce a source of uncertainty. While the fractional landcover type was already accounted in the vegetation correction (Equation 7), the final fire-dust emission flux per grid cell $${\mathcal{F}}_{\text{fdust}}(D)$$ needs to be weighed by the fraction of each soil type $${\omega }_{i}$$: 8 $${\mathcal{F}}_{\text{fdust}}(D)=\sum\limits _{i=\mathrm{soil}}{\omega }_{i}\cdot {F}_{\text{fdust,cor},i}(D)$$

The overall calculation of the fire-dust emission flux is conclusively illustrated in the flowchart in Figure 2. In a nutshell, the fire-dust emission flux $${\mathcal{F}}_{\text{fdust}}$$ for each particle size $$D$$ is a function of the fire strength in terms of the FRP, the 10 m-wind velocity $$\vert {v}_{10\mathrm{m}}\vert $$, the soil type, the roughness length of the burning vegetation $${z}_{0,\mathrm{v}\mathrm{e}\mathrm{g}}$$ as a function of the LAI, as well as the landcover class that determines the emissive fraction of the grid cell $${A}_{\text{emi}}$$, and can be written as follows: 9 $${\mathcal{F}}_{\text{fdust}}(D)=f\left(FRP,\vert {v}_{10\mathrm{m}}\vert ,\mathrm{soil},{\overset{-}{z}}_{{0}_{\text{veg}}}(LAI),{\overset{-}{A}}_{\text{emi}}\right)$$

[IMAGE OMITTED. SEE PDF]

The total dust mass $${\mathcal{M}}_{\text{fdust}}$$ that results from fire-emitted dust particles is eventually given by: 10 $${\mathcal{M}}_{\text{fdust}}=\frac{1}{6}\pi {\rho }_{p}\int \nolimits_{0}^{D={D}_{\text{max}}}{D}_{\text{fdust}}\hspace*{.5em}\mathrm{d}D$$ where $${\rho }_{p}$$ is the dust particle density similar to other dust particles and specified with 2.65 g cm⁻³. The mass distribution of the emitted dust particles between the accumulation and coarse mode and their partitioning into a soluble and insoluble part during the transport arises from the M7 scheme as an intrinsic part of the HAM module (see Section 2.3). In contrast to dust emissions caused by atmospheric motions alone, which are naturally emitted first into the lowermost model level, fire-dust emissions are subject to a different treatment as the strong pyro-convective forces can effectively lift even coarse-mode particles to higher altitudes (Wagner et al., 2018). To mimic this behavior, we applied the well-established fire-injection parameterization based on Val Martin et al. (2010), that is already part of ICON-HAM (see Section 2.3), to the fire-dust emission fluxes as well. Similarly to all other fire emissions, the fire-induced dust particles are distributed to 75% within the PBL and to 25% in the two model layers above. However, once injected into the atmosphere, the dust particles originating from fire activity are treated identically as the dust particles that are emitted by the purely wind-driven process and are thus not distinguishable for the model physics anymore. It yields now for every dust-containing mode x of HAM: 11 $${\mathcal{M}}_{\text{dust},x}={\mathcal{M}}_{\text{wdust},x}+{\mathcal{M}}_{\text{fdust},x}$$

Model Simulations

To test our approach and draw first conclusions regarding the regional and seasonal importance of fire-induced dust emissions, a sufficiently long simulation period is needed. The starting year was determined by the first availability of the GFAS FRP input data in 2003. As ICON-HAM requires at least a 3-month spin-up period before it is in a stable equilibrium (Salzmann et al., 2022), we start the analysis in 2004 and run the model for further 10 years until 2013 to get a robust data set. This ensures that the data are not too strongly affected by possible outliers, that is, a year with a particularly strong fire activity or some rare atmospheric/climate conditions. The 10-year period also assures meaningful comparisons against observations, since the model runs freely. Thus, the results need to be averaged over a sufficiently long time span to compensate for biases in the aerosol load introduced by the purely wind-driven aerosol emissions, such as (desert) dust, that are subject to the chaotically evolving wind patterns. The anthropogenic greenhouse gas emissions during the simulation period was described according to the RCP8.5 scenario (see Section 2.3).

In order to quantify the effect of the additional fire-dust emission, a “control” experiment was conducted, where the fire-dust parameterization was switched off, while all other parameters were kept constant, meaning that the atmospheric dust load is only governed by the wind-driven dust emissions. This simulation will serve as the baseline scenario whenever we investigate the impacts caused by the additional fire-dust emissions.

Validation

Although it is confirmed by several in situ and remote sensing observations (e.g., Kavouras et al., 2012; Nisantzi et al., 2014) that wildfires are related to the emission of soil-dust particles, detailed data of the emitted mass are largely missing. Thus, the validation of the fire-dust parameterization introduced here becomes quite challenging and is further complicated by the free-running, unnudged behavior of ICON-HAM in its current version. This prevents day-by-day comparisons at specific locations, since it cannot be expected that single dust events and the transport of other emissions including our fire-dust emissions are correctly reproduced in time and strength. Long-term averages of the (dust) AOD, either as yearly means or as mean over the whole simulation period 2004–2013, provide a way out as the weather-driven fluctuations are expected to cancel out each other over longer time periods.

Validation Against MIDAS

As the current study introduces an additional source of atmospheric dust, a validation against related measurements of the total atmospheric dust load is of highest interest. Unfortunately, it is rather difficult to retrieve the dust fraction from observations of the total AOD on a global basis, but MIDAS provides data of the DAOD as shown in Figure 3a as an average over the whole simulation period 2004–2013. The major dust sources of the so-called “dust belt” can be clearly identified with a mean DAOD of 0.3 and larger. Particularly dominant is the Bodélé depression with a mean DAOD of up to 0.8, but also other areas, mostly in the western parts of the Sahara, Sahel, and the adjacent Atlantic Ocean, are characterized by a high mean DAOD of up to 0.5.

[IMAGE OMITTED. SEE PDF]

When the modeled DAOD from both ICON-HAM simulations, the control run (Figure 3b) and the run with the fire-dust emissions (Figure 3c), is compared to MIDAS, the consistency of the rough regional structures is apparent although some regional structures got vanished as a result of the coarse model resolution. Enhanced values of the modeled DAOD are found for the control run (Figure 3b) over the Middle East, North Africa and the (sub)tropical Atlantic as part of the classic dust export route within the Saharan air layer (SAL) toward the Caribbean and South America. Two particular hotspots with a mean DAOD of up to 0.4 are found over and downwind of the Bodélé depression and the western Sahara in North Africa, while another one can be identified over the Rub-Al-Khall on the Arabian peninsula. Other parts of the Northern hemisphere experiences largely mean values below 0.1 mainly due to due transport processes rather than local source activity. This pattern does not change drastically by including the fire-dust emissions (Figure 3c) as the additional amount of the dust is relatively small, particularly in regions where the 'classic' dust sources are dominant, and distributed more widely over the globe. This is confirmed by the difference between both simulations (Figure 3d) where it becomes apparent that the mean DAOD increases nearly all over the globe slightly. Usually, this increase is in the order of 0.01–0.02, however, larger increases can be found over basically the whole African continent, parts of South America, Australia, and the region north of the Black and Caspian Sea. The strongest increase is located over the Eastern Sahel with the mean DAOD up to 0.08 higher, adding to the already high dust load in this particular area. In contrast, the additional dust mass within the Southern hemisphere, although less in absolute terms, means often a duplication of the mean DAOD compared to the scenario with wind-driven dust emissions only. Another region with an increased DAOD, that is not related to fire-dust emission, is found off coast of West Africa. Instead, the wind-driven dust emissions have increased in the Western Sahara due to atmospheric radiative feedbacks that are associated with the additional dust load induced by the fires over the Sahel and fostered by the unnudged model system as typical for North Africa (Heinold et al., 2008).

Compared to the MIDAS DAOD retrieval, the model runs without (Figure 3e) and with (Figure 3f) the new fire-dust parameterization show over large parts of the 'dust belt' an underestimation of the actual mean dust load, with a negative DAOD bias of around 0.1. The strongest underestimation occurs over the Taklamakan, whose emissions are unfortunately not well captured by the model. Over the western part of North Africa, the overall underestimation is interrupted by two areas where the modeled DAOD surpasses the observation, extending further into the tropical Atlantic. The eastern region comprises the Bodélé depression, while the other one is related to dust emissions in the Western Sahara, the two regions with the in total largest simulated DAOD (Figures 3b and 3c). This is most likely a result from the unnudged ICON model itself that has difficulties to correctly reproduce the wind patterns in the region, such as the African Easterly Jet, during the most active dust periods, so that the emitted dust is transported on a too southern route toward the Atlantic. This could result, despite the reasonably appropriate total emissions in the region (see also Table 4), to an overestimation of the dust load on the modeled transport routes and the related underestimation over the continental regions west of the Bodélé.

Apart from that, the original model version without the fire-dust emission shows a slight underestimation of the global mean dust load compared to the MIDAS data. This underestimation was eliminated by including the additional emission pathway and in fact results in a slight overestimation in many regions around the globe. The overall RMSE gets slightly reduced from 0.032 to 0.030, indicating an improvement after the implementation of the new dust source type in the rest of the globe, although the largest mismatches between the model and the observations could not be resolved as they are related to the purely wind-driven dust emissions and their atmospheric distribution over North Africa as explained above.

Validation Against AERONET Stations

In addition to the global DAOD analysis, AERONET stations are used for the validation of the fire-driven dust emissions. Measurements of the AOD at single stations allow for a more regional comparison and represent a largely independent data set of the atmospheric aerosol load. Figure 4a provides an overview of all AERONET stations used in the analysis, grouped together into (sub)continental regions that largely match with the continental borders, only with the omission of a few very remote stations. Due to the limitations arising from the unnudged ICON-HAM simulations, we can compare only yearly averages, where the impact of weather-driven variance is reduced and the data less noisy.

[IMAGE OMITTED. SEE PDF]

Firstly, Figure 5 presents a comparison of the modeled and measured AOD at 550 nm of all AERONET stations whose data availability allowed for the derivation of a yearly mean, both for the control run (Figure 5a) and the simulation with the additional fire-dust emissions (Figure 5b). Similarly to the MIDAS DAOD comparison, there is a general underestimation of the AOD, while the model overestimates the AOD only at a few stations noteworthy. This statement is true for both simulations, independent whether the fire-dust emissions are included. However, including this emission pathway results in a small improvement of all statistical values, for example, a reduction of the bias from −0.05 to −0.04 and the RSME from 0.14 to 0.13, while the pearson correlation factor increases from 0.67 to 0.71 and the standard deviation is reduced from 0.94 to 0.91. Although small, these changes are consistent with the findings arising from the MIDAS comparison and support the idea that an explicit consideration of fire-dust emissions leads to an improvement of the modeled AOD.

[IMAGE OMITTED. SEE PDF]

Although AERONET stations cover basically the whole globe, their distribution is unequal and dominated by the industrial regions in the Global North such as North America, Europe, and East Asia (cf. Figure 4a). These regions are usually not strongly affected by wildfires and thus might mask a stronger signal arising from regions where fire-dust emission is more distinct. Therefore, a separate analysis for the different (sub)continental regions is presented in Figure 6. Virtually no improvement of the modeled AOD was found for the AERONET stations in North America (Figure 6a) and Europe (Figure 6b), where the aerosol load is indeed largely dominated by other anthropogenic emissions which have not been altered in the model simulations. Nearly no improvement was found also for the Asian stations (Figure 6c), although this region is much more affected by wildfires. However, in particular within the central parts of the continent, where fire-dust emissions were found to be present in large quantities (see also Figure 10b), the density of stations is quite sparse, while many AERONET stations are located at the coastal regions of East Asia where again anthropogenic emissions dominate.

[IMAGE OMITTED. SEE PDF]

Compared to that, the total number of AERONET stations in Africa, South America, and Australia is much smaller (Figure 4a), but they are often much closer to the fire hotspots. Consequently, significant changes were found in all four of the (sub)continental regions (Figures 6d–6g) that point toward an improvement after the implementation of the fire-dust emissions. For example, the Northern African domain (Figure 6d), although largely dominated by desert dust emissions, many stations are located in and south of the Sahel, experiences a strong reduction of the RSME from 0.26 to 0.19 together with an increase of the correlation coefficient from 0.41 to 0.54. Similar improvements of the correlation coefficient are found in Southern Africa (Figure 6e), Australia (Figure 6g), and South America (Figure 6f) that often go hand in hand with a reduction of the bias, although a general underestimation of the AOD by the model still persists afterward. Nevertheless, the comparison of the modeled AOD with that retrieved from the AERONET stations of the Global South shows a significant improving trend after the implementation of the fire-dust emission pathway.

As the continental comparison is still rather coarse, we selected specific regions in Africa, South America, the Midwestern US, Australia, and Eastern Europe/Central Asia where the strongest fire-dust emissions occur (see also Figure 10b) and choose the AERONET stations therein for further analysis (Figure 4b). The regional extent of these regions is given in Table 3. When the yearly averages of all those stations are grouped together and compared to the modeled AOD (Figure 7), an overall improvement was found after the implementation of the fire-dust emissions as indicated by a reduction of the RSME and bias by each 0.02 with the strongest contribution coming from Northern Australia and the Amazons region. However, it also needs to be considered that the additional fire-dust emissions do not remain just in the fire-active regions but can also be transported further to the surrounding, so that improved correlations in more remote regions might still be attributed to the fire-dust emissions. Of course, it cannot be ruled out that also other fire-related emissions than soil-dust particles, for example, ash, could be partially responsible for the local improvement, but we suppose that the other combustion emissions are captured fairly well by the established emission inventory. Thus, the additional fire-dust emissions perform reasonably well and can be used for further analysis.

Table 3 Extent of the Regions With Particular Strong Fire-Dust Emissions as used for Figure 7

Region	Extent
Mid-western US	95–105$${}^{\circ}$$W, 30–52$${}^{\circ}$$N
Southern America	35–68$${}^{\circ}$$W, 5–28$${}^{\circ}$$S
Sahel	20$${}^{\circ}$$W–45$${}^{\circ}$$E, 5–15$${}^{\circ}$$N
Southern Africa	10–40$${}^{\circ}$$E, 5–30$${}^{\circ}$$S
Eastern Europe/Central Asia	30–80$${}^{\circ}$$E, 45–55$${}^{\circ}$$N
Northern Australia	120–150$${}^{\circ}$$E, 12–23$${}^{\circ}$$S
Eastern Australia	142–165$${}^{\circ}$$E, 20–40$${}^{\circ}$$S

[IMAGE OMITTED. SEE PDF]

Results

Global Fire Activity

Before putting the focus onto the fire-dust emissions more specifically, Figure 8 shows the global fire activity expressed as averaged FRP over the simulation period 2004–2013 and depicts the areas that can theoretically act as a source of fire-dust. The global hotspots of fire activity, with the on average most intense fires, lie within Africa around the Sahel and south of the equator in Southern Africa including Madagascar. The agricultural activity in these regions typically includes fires as a farming method to prepare the fields for the next cultivation cycle or to open up new areas for agricultural use (Sá et al., 2011; Wardell et al., 2004). Similarly, further areas with an intense fire activity can be found in the equatorial South America, particularly the Amazonas region, Southeast Asia and Northern Australia. Less frequently affected by strong fires are large parts of North America, Asia, Europe and the southern half of Australia.

[IMAGE OMITTED. SEE PDF]

However, the yearly mean provides only an incomplete picture of the regional fire hotspots. If focused on the seasonality of the fire activity (Figure 9), it becomes apparent that the fire seasons, and with it the related fire-dust emission potential, are often confined to only a few months per year. This can be either defined by the dry season within the tropics or by the summer season of the corresponding hemisphere, where the combination of above-average temperatures and drought conditions can increase the vulnerability to fires. During DJF (Figure 9a), the by far strongest fire activity is found within the Sahel in a quite confined band around 10$${}^{\circ}$$N, highly driven by man-made activities for fire clearing (Sá et al., 2011; Wardell et al., 2004), while other tropical regions around the globe are far less affected. This changes during MAM (Figure 9b), which can be considered as a transition period for the African fire activity with fading fire occurrence around the Sahel and the onset of fire activity in the Southern African territories. In addition, fires become more common in Latin America, Southeast Asia and the Siberian taiga. During JJA (Figure 9c) when the dry season peaks in the southern (sub)tropics, the fire activity in Southern Africa reaches its peak strength and fires in Southern America become more frequent and/or more intense as well. On the Northern hemisphere, regions in North America, such as Canada, Alaska, and California, experience their main fire season. To a smaller extent, some parts of Southern and Eastern Europe become more vulnerable to fires in the summer season too. In Asia, fires expand further toward the north and affect the Siberian tundra, although mean intensity is reduced. During the transition into SON (Figure 9d), the burning conditions in the Northern hemisphere become less optimal (boreal autumn) and fire occurrence is reduced again. In contrast, this season is associated with an increasing fire strength in the Southern hemisphere and South America, Australia, as well as Madagascar experience the most/strongest fires. In other parts of Southern Africa, fire activity remains high, but their mean strength is already decreasing, while the onset of the Sahelian fire season begins. In conclusion, most continents or continental subregions experience well defined fire seasons, but America and in particular Australia stand out as fires are common here throughout the year, although with varying intensity. Consequently, fire-dust emissions are expected to undergo similar regional and seasonal variations.

[IMAGE OMITTED. SEE PDF]

Wind-Driven Dust Emissions

Independent how strong fire-dust emissions are, the purely wind-driven dust emissions (Figure 10a) from deserts and other dry land areas dominate undoubtedly the global dust load and the fire-related dust emissions can only add to that. As expected, the emissions from the major deserts along the “dust belt” including the Sahara, the Arabian deserts, and Central/Eastern Asia dominate the globally emitted dust mass as confirmed by the emission sum in Table 4. The Sahara and the dust sources on the Arabian Peninsula alone are accounting for up to 800 Tg yr⁻¹ and thus roughly 75% of the global dust emissions. This is, although on the lower end, in good agreement with the multi-model comparison studies by Huneeus et al. (2011) and Wu et al. (2020) who found median dust emission fluxes from those source regions of roughly 900 and 1,400 Tg yr⁻¹, respectively. The Asian dust sources add up to around 200 Tg yr⁻¹ within our ICON-HAM simulations, which is well above the median value of Huneeus et al. (2011) but not even half of the median amount found by Wu et al. (2020), whose dust emission fluxes are in general much larger. Other well-known global deserts in North America, South America, the Namib and Kalahari in Southern Africa, and Australia are mapped reasonably well by ICON-HAM, too. However, their contribution to the global dust load is, with usually less than 10 Tg yr⁻¹, quite small and at the lower end of the comparisons by Huneeus et al. (2011) and Wu et al. (2020) but always within the range of variation. In total, the purely wind-driven dust emissions sum up to on average around 1,000 Tg yr⁻¹, with fluctuations between 900 and 1,100 Tg yr⁻¹, matching the multi-model mean by Huneeus et al. (2011) quite well, while they are by a factor of 2 smaller if compared to Wu et al. (2020). Nevertheless, it can be concluded that our model reproduce the global dust emissions reasonably well.

Table 4 Wind- and Fire-Driven Dust Emission Fluxes of ICON-HAM in Comparision to Other Studies

Region	ICON-HAM simulations (this study)	Multi-model comparison studies
Wind-driven dust emissions (Tg yr−1)	Fire-driven dust emissions (Tg yr−1)	Wind-driven dust emissions (Tg yr−1)
Huneeus et al. (2011)	Wu et al. (2020)
North America	3.6	16.7	2 (1.7–286)	5 (0.1–340)
South America	4.4	32.4	10 (0.2–186)	49 (3–365)
Europe	0.8	10.7	–	–
Northern Africa/Middle East	793.1	68.7	920 (272–3,419)	1,428 (508–3,800)
Southern Africa	3.4	37.7	12 (3–57)	20 (5–142)
Asia	200.8	38.1	137 (27–873)	502 (190–1,588)
Australia	9.3	27.8	31 (9–129)	56 (0.6–2,278)
Globe	1,015 (906–1,111)	232 (189–255)	1,112	2,060

[IMAGE OMITTED. SEE PDF]

Fire-Dust Emissions

Similarly to the wind-driven dust emissions, Figure 10b shows the 10-years-averaged dust emissions related to fire activity. If compared to Figure 10a, it becomes evident that fire-dust emissions are on average reduced in strength as their occurrence is limited to active fires and thus naturally less likely to occur than wind-driven dust emission events over the bare surfaces in the classic desert regions. However, fire-dust emissions occur more widely and affect also often regions that are not known as a significant source of mineral dust. For example, in the northern parts of Africa fire-dust emissions are largely limited to the savannas south of the Saharan desert, a global hotspot of fire activity as indicated by Figure 8, while north of it only along the Mediterranean coast additional emission occur. Together with the Saharan desert, lacking any vegetation that can burn and thus not a source of fire-dust emissions, nearly the whole Northern African land mass can be considered as a significant source region of soil-dust emissions. Similar statements can be made for Southern Africa and Australia, that are both characterized by deserts and other sparsely vegetated landscapes with regular wind erosion and further semi-arid regions that are often affected by fires. Here, however, wind- and fire-dust emissions are spatially not that strongly separated and many grid cells contain both areas of bare soil and suitable vegetated areas, that can be related to fire-dust emissions as it can be seen in Figure 10c, showing the regions with wind- and fire-driven dust emissions and their relative share.

If the focus shifts to other continents, potential dust sources expand also substantially, particularly in North and South America, where the Great Plains and savannas in Brazil can now also be attributed to mineral dust emission through fire activity. In Asia, fire-dust activity leads to an extension of possible dust source areas toward the north, affecting now Siberian tundras as well as the savannas in Southeast Asia. On these continents only a few locations remain without any significant dust emission at all, as there is neither dry and bare land allowing for wind erosion nor that fires are part of the local land-use practice or just naturally do not occur.

Figure 10c indicates all regions (grid cells) that are related to both fire- and wind-driven dust emissions and provides their relative share. Such regions comprise the Sahel, the dry southwestern parts of Africa, Argentina, the western half of the US, large parts of Australia as well as Central Asia, India, and the Northeast of China. Thereby, the fire-dust contribution dominates dust emission in the US, Southern Africa, Northern Australia, and parts of Asia, while Southern Australia, Argentina, and large parts of Asia are characterized by wind-driven dust emission and do not experience fires very frequently. Thus, the additional fire-dust emissions contribute only a comparably minor amount to the total dust emissions in that regions.

If the fire-related dust emission fluxes are expressed per FRP (Figure 11), it becomes apparent that the regions with the on average strongest FRP are not necessarily related to equally strong fire-dust emissions. In contrast, other regions, for example, the Great Plains, the areas north of the Black and Caspian Sea, some parts of the Ethiopian Highlands and the Sahelian savannas contribute significantly to fire-dust emissions, although the mean fire strength is relatively small. This behavior nicely illustrates that fire-dust emissions are not only a function of the FRP. In addition, further modulations arise from the burning vegetation type via the implemented roughness correction and the assumptions regarding the emissive area, which both together drastically reduce fire-dust emission for biomes with a high fraction of trees. This assumption largely excludes fires in rain forests, such as those in Indonesia, and fires in the boreal forests of Canada and parts of Siberia from becoming effective dust sources. In other regions, the vegetation reduces the dust emission strength despite quite intense fires. Another important player is the soil type at the fire location, which can modulate the dust emission flux, despite the same fire strength and similar vegetation cover, by more than one order of magnitude as already shown in Figure 1. Expressed in terms of the emitted dust mass, the most effective soil type is the coarse mode soil, which dominates the classic dust emission areas in the deserts of Northern and Southern Africa and Australia. The medium mode soil type is dominant in North America and large parts of Asia and contributes to the large fire-dust emissions there, despite the on average rather weak fires (cf. Figure 8). Thus, although the presence of a fire is an essential precondition for fire-dust emission, the soil-surface conditions modulate the emission strength significantly as it was already discussed in the methodology section (Sec. 4).

[IMAGE OMITTED. SEE PDF]

If the annual mean fire-dust emission fluxes are summed up for the different continental regions (Table 4), North Africa dominates again with additional emissions of up to 70 Tg yr⁻¹, largely driven by the anthropogenically caused, highly active fire regime in the savannas around the Sahel. However, this is only less than 10% of the total dust emissions of the Northern African domain. In contrast, the fire-induced dust emissions in other regions, such as Asia, Australia, Southern Africa, and South America, that contribute in absolute terms much less to the total fire-dust emissions, can exceed the purely wind-driven dust emissions by a multiple. Even Europe, hardly known as a dust source on a global scale, can contribute up to 10 Tg yr⁻¹ of fire-related dust emissions, mainly originating from the eastern parts of the domain with the fire-active regions around the Black Sea, for example, in Ukraine. Globally, the average fire-dust emissions within the investigated 10-years-period 2004–2013 sum up to 237 Tg yr⁻¹, with fluctuations between 189 and 255 Tg yr⁻¹. Therefore, fire-dust emission might be responsible for 18.8% (14.9%–20.9%) of the global dust load, that is found to be around 1,250 (1,124–1,359) Tg yr⁻¹, at least with respect to our rather coarse ICON-HAM simulations.

Seasonal Variations of Fire-Dust Emissions

Seasonal variations of the fire-dust emissions, that are given in Figure 12, follow closely the seasonal cycle of the fire activity as presented in Figure 9. Peak emission fluxes occur during DJF in Northern Africa around the Sahel and in South America, Southern Africa, and Australia during the related dry seasons in JJA and SON. However, if the seasonal maps of fire activity (Figure 9) and fire-dust emissions (Figure 12) are compared closely, some differences are apparent. The strongest fire activity in North America is found during JJA in the boreal forests of Canada and Alaska, but fire-dust emission fluxes are peaking in MAM in the Midwestern US. This area is furthermore a source region throughout the year despite rather weak fire activity, but influenced by the rather fragile vegetation affected by the fires and the highly emissive medium-mode soil type. While boreal forests burn quite intensively, the associated large surface roughness suppresses dust emission effectively. The fires occurring in the grasslands and steppes of the Midwestern US, in contrast, affect the soil surface more directly and the fire-driven winds can mobilize soil-dust particles easily. A region similarly prone to fire-driven dust emissions stretches from Eastern Europe, north of the Black and Caspian Sea, to Central Asia, where the strongest fire-dust emissions occur from MAM to SON. In contrast, the fires affecting parts of the Siberian taiga, particularly active during MAM and JJA, are again only barely linked to dust emissions.

[IMAGE OMITTED. SEE PDF]

In the Southern hemisphere, where fire-dust emissions are more clearly connected to the intensity of the fire season, the emissions dominate in the second half of the year, in particular during SON. Furthermore, these continental areas can act as a source of dust, released by vegetation fires, throughout many months of the year, and in some regions, such as Australia or South Africa, even for nearly the entire year. Despite the relative small emission fluxes, fire-dust emission can represent a significant aerosol source in the otherwise rather pristine Southern hemisphere and might have substantial impacts on the atmospheric conditions.

Interannual Variability of Fire-Dust Emissions

The “classic” dust emissions within ICON-HAM are purely driven by meteorological features, where the wind speed at the surface is the most important driver, and thus underlie random variations based on the chaotic weather regime within the unnudged model setup. In contrast, fire-dust emissions are mainly driven through the fire activity and only subsequently modified by an additional impact of the 10 m-wind velocity. Therefore, they are expected to follow the actual fire activity more closely and a comparison between the interannual variability of the fire intensity, in terms of the FRP and the fire-dust emissions becomes feasible.

Figure 13a shows the yearly averaged variations of the global mean FRP and the  related fire-dust emission fluxes between 2004 and 2013. The fire activity fluctuates usually only slightly around the mean value with only a few exceptions. 2007 marks a year with a clearly above-average fire activity, while 2009 and 2013 characterize the lower end within the simulation period. The global fire-dust emission fluxes show an even smaller variability and fluctuate often only slightly around the mean value of 232 Tg yr⁻¹. Just 2013, with emissions of only 190 Tg yr⁻¹, marks in accordance to the mean FRP again a negative outliner within the times series. While the 3 years with the lowest mean fire strength (2009, 2010, and 2013) are indeed also the years with the smallest total fire-dust emission fluxes, the opposite is not always true. 2007, characterized by the on average most intense fire activity, is only connected to an average fire-dust emission flux. Factors that sometimes prevent the direct link between fire activity and fire-dust emissions comprise, among others, that fires in individual years affect different biomes and soil types. Thus, the question whether and how strong the related fire-dust emissions are, depends strongly on those soil-surface properties. Nevertheless, the general trend always aligns and years with an above (below) average fire intensity are linked to an above (below) average fire-dust activity, which underlines the assumption that the FRP is at least globally the main driver of fire-dust emission.

[IMAGE OMITTED. SEE PDF]

If the same analysis is applied to the predefined (sub)continental regions (Figures 13b–13h), both the variation between individual years and the deviations between the mean FRP and the related fire-dust emission fluxes increase. Exemplary for North America (Figure 13c), 2013 as the year with the on average strongest fire activity is simultaneously the year with the lowest fire-dust emissions. Here, again, the likely cause are strong fires in the forest biomes of Canada and Alaska that are not very dust-active, while a year such as 2008 with a minimal fire strength can result in even slightly above average dust emission fluxes, if the highly emissive grasslands are burning. This changing fire behavior, that is in these regions largely governed by meteorological conditions that create “fire weather,” can result in quite huge fluctuations of the yearly fire-dust emissions in certain regions, such as North America where the emission fluxes vary between 10 and 25 Tg yr⁻¹. Similarly, the fire-dust emissions in Europe (Figure 13b) fluctuate between 5 and close to 20 Tg yr⁻¹, however, with a better accordance to above- and below-average values of the FRP. This points predominantly to fires that regularly affect the same or similar landscapes (here Eastern Europe), just with a large year-to-year variability in fire intensity.

The latter statement is also true for other regions such as South America (Figure 13d) with fire-dust emissions varying between 25 and 45 Tg yr⁻¹ or Australia (Figure 13h). However, the fire activity in Australia is strongly driven by weather and climate features, such as droughts linked to the El Niño Southern Oscillation (ENSO), and thus varies substantially (Mariani et al., 2016; Yu & Ginoux, 2021). For example, in 2010, outbound of a strong El Niño, the fire activity in Australia was minimal, while it increased more than fourfold in 2011 and 2012. Both years were characterized by strong La Niña conditions (Boening et al., 2012). Consequently, the fire-dust emissions vary considerably with minimal emissions of around 15 Tg yr⁻¹ in 2010 and more than 50 Tg yr⁻¹ 2 years later.

A much more uniform fire activity was observed within the 10-year period for the Northern African domain (Figure 13e), where the dominating fires in the sub-Saharan region are ignited on a routinely basis for farming and other land-clearing purposes. Therefore, the related fire-dust emissions vary only slightly from 65 to 75 Tg yr⁻¹, only with 2010 as an outlier with a well below average fire strength and thus fire-dust emissions of only 60 Tg yr⁻¹. A quite similar picture can be drawn for Southern Africa (Figure 13f), where the fire regime is also strongly controlled by human activity, and thus the related fire-dust emissions vary primarily between 30 and 45 Tg yr⁻¹. The situation in Asia (Figure 13g) is a bit more divers, as there are different fire regimes and fire-affected landscapes whose individual contributions fluctuate from year to year. This results in fire-dust emissions that vary by a factor of more than two, reaching from almost 20 to over 50 Tg yr⁻¹.

All in all, fire-dust emissions within certain regions can underlie a substantial interannual variability, particularly when the fire activity is related to varying weather and climate conditions such as the ENSO, while on a global scale years with an above- and below-average emission strength tend to compensate each other.

Vertical Distribution of the Atmospheric Dust

The increased atmospheric dust load provoked by the fire-dust emissions (cf. Figure 3d) might not only be the result of just more emissions, but the atmospheric life time of the dust particles can play a role too. Assumingly, the initial injection height of the dust particles determines their removal time from the atmosphere by deposition processes. Thus, the injection in higher atmospheric levels, partly even above the PBL, as implemented for the fire-dust emissions can result in an enhanced atmospheric concentration.

Figure 14 shows vertical profiles of the global dust mass concentration with and without the additional fire-dust emissions. As expected, the mean concentration increases as the new source is implemented, but the atmospheric dust mass does not increase uniformly. While the lowermost model layers do not show significant differences between the two simulations, the picture changes with altitude. In particular, the lower troposphere up to heights of 500 hPa experiences an increased dust load with the strongest difference occurring at around 750 hPa, and thus well above the level of the highest dust concentration without the fire-dust emissions (Figure 14a). This clearly empathizes the effect of the increased injection height of the fire-dust emissions into the whole PBL and the two model layers above. While the dust concentration in the mid troposphere is enhanced, the impact becomes less important for the upper troposphere, where in our setup even for the strongest fires no direct injection occurs and atmospheric motions alone are governing the vertical dust transport. If the dust aerosol is differentiated between the coarse (Figure 14b) and fine mode (Figure 14c), it become apparent that the increase is mainly driven by the coarse mode particles. This is consistent with the offline dust emission modeling based on the LES of Wagner et al. (2021), who found a significantly larger portion of coarse mode particles in the fire-driven dust emission fluxes compared to the classic, non-fire dust emissions. This suggests that vegetation fires, along with the associated dust emissions, might represent an atmospheric pathway for coarse-mode dust particles into the free troposphere, where they can undergo a long-range transport carrying them far away from their source regions (Ratcliffe et al., 2024; van der Does et al., 2018).

[IMAGE OMITTED. SEE PDF]

The vertical profiles of dust concentration and the additional impact of the fire-related emissions are not constant in the course of the year as the seasonal profiles in Figure 15 show. During DJF (Figure 15a), the global dust burden is more confined to the lower tropospheric layers, likely even to a large portion within the PBL. This is primarily related to the world's most dominant dust source, the Sahara, and the related dust transport pathway within the SAL, which is during DJF much more confined to the lower troposphere as the PBL is shallower (Schepanski, Tegen, & Macke, 2009). As the injection parameterization of the fire-dust emissions also depends on the PBL height, the fire-dust emissions, that also dominate during DJF over North Africa (Figure 12a), are injected to relatively low altitudes too.

[IMAGE OMITTED. SEE PDF]

During MAM (Figure 15b), the dust is distributed more widely within lower and mid tropospheric levels, and even present in considerable amounts in the upper troposphere up to 200 hPa. This change is again largely related to North Africa, where the PBL now deepens and the dust transport within the SAL occurs at higher altitudes (Schepanski, Tegen, & Macke, 2009). The already higher injection at the source regions supports also the transport into the upper troposphere. The total contribution of fire-dust emissions is relatively small during MAM as this is apparently the season with the weakest additional emissions. Furthermore, the share of North African fires, where the PBL is particularly deep, becomes smaller, while other notable fire-dust emissions occur in the extra-tropical regions with a much shallower PBL such as in the Midwestern US and in Central Asia. Thus, the additional fire-dust is more uniformly distributed within the troposphere.

During JJA (Figure 15c), the contribution of North Africa to the global (fire)-dust emissions declines further and the Southern hemisphere becomes more dominant. Thus, the vertical extent of the atmospheric dust concentration becomes smaller and peaks at altitudes of around 700–800 hPa, where also the maximum of the additional fire-dust is injected or transported to. While the purely wind-driven dust emissions decline further during SON, fire-dust emissions reach their peak season with numerous source areas around the global (cf. Figure 12d). Their relative contribution to the global dust load increases further, which is reflected in the vertical profile shown in Figure 15d too. The additional fire-dust aerosol is present nearly throughout the whole troposphere, with the strongest increase at approximately 700 hPa substantially above the maximum level of the non-fire dust emissions, once again highlighting the impact of the higher injection heights of the fire-dust emissions.

A closer examination of the vertical profiles for the different predefined continental regions (Figure 16) provides further insights into the regional importance of the additional fire-dust emissions. Nearly all of the continental regions on the Southern hemisphere experience a drastic increase of the mean atmospheric dust concentration, partially even by a factor of 2 or more as in Southern Africa (Figure 16f) or Australia (Figure 16h). Here, the fire-driven dust emissions are predominantly injected in and/or transported to elevated atmospheric levels with the maximum concentrations at a height of 700–800 hPa. For the North African domain (Figure 16e), the additional atmospheric dust load introduced by the fires is expectably small compared to the wind-driven dust emission but it is still notable in absolute terms and is present throughout all lower and mid-tropospheric levels with the peak difference at a height of 700 hPa. Interestingly, the additional fire-dust aerosol within the Asian domain (Figure 16g) is more confined to the lower troposphere down until the surface. This is supposingly a result of the majority of fire-dust emissions occurring north of 50$${}^{\circ}$$N, where the PBL height usually decreases, which in fact lowers the injection heights of fire-dust emissions there as well.

[IMAGE OMITTED. SEE PDF]

The small change of the atmospheric dust concentration within the European domain (Figure 16d), despite the large additional fire-dust emissions (compare Table 4), as well as the comparably large mean dust load there point to a dominant influence of Saharan dust being transported to Europe. In particular during MAM, the transport of Saharan dust is quite common over large parts of Europe (Marinou et al., 2017; Meloni et al., 2008). Thus, the impact of the local fire-dust emissions remains rather small. However, the advected Saharan is most pronounced in the atmospheric levels above the PBL almost up to the tropopause, which once again highlights the large contribution of advected dust. In contrast, the fire-dust emissions represent a local source and thus affect mostly the lower troposphere up to 700 hPa where they can have a stronger impact on local air quality. Similarly, the North American domain (Figure 16b) extends far onto the Atlantic Ocean and the Caribbean, a well known steady transport pathway of Saharan dust within the trade wind regime (Doherty et al., 2008; Prospero, 1996). In addition to that, the quite intense fire-dust emissions from the Sahelian fires are partly mixed within that transport pattern as well (Haywood et al., 2008; Paris et al., 2010). Thus, local (fire)-dust emissions are less pronounced when mapped by the averaged profiles within the large North American domain and only small changes can be found as a result of local fire activity.

The regional consideration showed that the additional fire-dust emissions can have quite distinct impacts on the local atmospheric dust load, reaching from rather small increases to more pronounced changes in the atmospheric levels where the majority of dust particles are present. This can strongly affect the atmospheric relevance of the dust particles with respect to radiative feedbacks or cloud microphysics.

Discussion and Limitations

The here proposed fire-dust parameterization was introduced and tested using the recently coupled global aerosol-climate model ICON-HAM (Salzmann et al., 2022). Therefore, certain aspects need to be taken into account when interpreting the results of our model simulations. While the global approach allows for an investigation of the possible importance of the additional fire-dust emissions in different regions over a sufficiently long time period, the rather coarse resolution of a global model introduces several limitations and uncertainties to the approach.

To bridge the results from the mainly LES-based studies of Wagner et al. (2018, 2021) to the much coarser ICON-HAM resolution, some adjustments and simplifications were unavoidable. This includes both how the dust emission process itself is treated as well as how many of the additional (external) factors, that can affect the fire-dust emission strength, are considered. Some fire and soil-surface properties are either not available on a global grid, at least not with a sufficiently high data quality, or whose numerical implementation into a global model, that runs over several years, would cause immense computational costs that are likely not compensated by the additional gain of information. For example, it was shown in Wagner et al. (2021) that dust emissions in the fire vicinity are not only sensitive to the fire strength but also to the fire size. The latter, however, is already very difficult to derive for an individual fire that usually moves through a landscape and therefore frequently changes its size and appearance. However, it becomes barely manageable for larger areas, such as represented by a grid cell, where multiple fires can burn at the same time. As fire size and fire strength are also often correlated, we propose that the additional value of a consideration of the fire size is quite small and have focused on fire strength alone.

Crucial for the deviation of fire-dust emissions are accurate data about the fire activity. As a measure of fire strength, we have used data of the FRP provided by GFAS that incorporates measurements from MODIS instruments onboard of polar-orbiting satellites. Thus, the temporal resolution of the global fire activity is limited to only one value per day and no information about a diurnal cycle is available. As in particular small fires, such as those ignited for agricultural purposes, have rather short burning durations, the limited overpasses of the satellites can miss some of these fires or, in contrast, if the fire was just burning during the satellite's overpass, overestimate the fire activity. Although the quality-checked GFAS retrieval accounts for it (Kaiser et al., 2012), some of the above mentioned problems cannot be fully excluded. Hence, also the related fire-driven dust emissions are subject to some uncertainties unavoidably related to the chosen approach. For example, savannah fires are typically extinguished during night (Roberts et al., 2009), so that a daily mean FRP can artificially enlarge the burning duration and thus the related fire-dust emissions. The interpolation of the fire data onto a coarser grid is inevitably related to a general loss of information. As result of the conservative interpolation of the GFAS data onto the coarser R2B4 grid of ICON, the maximum FRP strength per grid cell is reduced, which lowers the related fire-dust emission fluxes therein. However, as the fire-dust emission flux is a continuous function of the FRP (cf. Figure 1), even grid cells with a small average FRP are allowed to emit dust, which avoids the loss of the contribution of too small/weak fires due to the interpolation routine, although their emissions are now distributed over a much larger area, in fact over the size of one grid cell.

Furthermore, ICON-HAM displays some data, such as the soil type and the vegetation cover, as fractions per grid cell. Therefore, it is possible that parts of the soil and the vegetation affected by the fire do not always completely overlap with each other and/or with the actual fire, that was described by an averaged FRP over the entire grid cell. This can result in a misclassification of the actual fire-exposed soil and vegetation type, and an overestimation of the resulting fire-dust emissions is possible. On the other hand, the vegetation correction basically excludes fire-dust emissions from forest biomes, as we found no strong evidence that such landscapes are a significant source of soil-dust particles. However, this might also be related to a lack of measurements and investigations of such plumes, and some forest-dominated biomes can still provide suitable emission conditions for soil-dust particles. This would result in an underestimation of the global fire-dust emissions, in particular for fires in boreal forests, and with that of the dust load in high latitudes.

We assumed that the soil-dust particles, and in particular their PSD, were not changed due to the fire impact and that they retain their properties that govern the wind-driven dust emission too. However, the soil-surface structure is strongly affected by the heat of a fire, and thus the actual fire-dust emission fluxes from a fire-disturbed soil differ from those of an undisturbed soil. This can introduce some biases whereby their sign is unknown. Another important type of fire-related dust emissions are post-fire dust emissions that have not been considered in our study, but can play a substantial role for the atmospheric dust load in fire-prone regions. Post-fire dust emissions are again a purely wind-driven process, where dust particles are emitted from a fire-affected, vegetation-free surface as long as 1 year after a fire event, however, with the strongest emissions taking place within several weeks after the fire and thus in temporal dependence to the in-fire dust emissions (Yu & Ginoux, 2022). Thus, it is possible that a part of the missing additional dust load over the regions with a strong fire activity is related to post-fire emissions rather than arise from currently burning fires. As this process is missing in ICON-HAM and all other global aerosol models, it might be another factor for the still existing underestimation of the modeled AOD in these regions—even after the implementation of the fire-dust emissions—as the comparison with the AERONET stations has revealed. Therefore, it seems appropriate to consider post-fire dust emissions in further developments of the parameterization, for which the work of Yu and Ginoux (2022) represents an ideal foundation.

IOCN-HAM in its current version is unnudged and thus the atmospheric conditions evolve freely during the simulation period. In contrast to that, the model was fed with measured daily FRP data. This does not only mean that fire-dust emissions (and also all other types of emission) can be transported to different regions as they would have been in reality, but also that the current weather and wind conditions at the fire origin do not necessarily represent typical 'fire weather' conditions anymore. As an extreme case, it is even possible that rainy conditions are accompanying some fire events, which would accelerate the removal of the dust particles directly at the source and reduce the actual aerosol load. On the other hand, missing precipitation or other aerosol removal processes in the model subsequently to a fire event can artificially enlarge the atmospheric lifetime of the emissions. Overall, there are both enhancing and compensating effects on the fire-dust emissions that result from the non-nudged model version, so that we postulate that thanks to the chosen simulation period of 10 years the related biases should be reduced and it is not expected that they affect the general statements made within the current study.

The free-running behavior of the model also implies some limitations regarding the validation of our results. Nevertheless, comparisons of (multi-)year averages of the atmospheric dust load with observations of the (dust) AOD using the MIDAS data set (Gkikas et al., 2021) and against AOD measurements of AERONET stations (Holben et al., 2001) have revealed improvements after the fire-dust emissions were implemented, particularly for regions on the Southern hemisphere where fires are quite common. Hence, it appears increasingly likely that with the fire-dust emission process a significant dust emission pathway is missing in the current generation of atmosphere-climate models. A future consideration of the process could lead to a better characterization of the atmospheric dust load, particularly in fire-prone regions. This is all the more important as the soil-dust particles emitted during fire activity cannot be expected to be entirely present as pure mineral dust. Instead, they are often mixed internally with soot or other (in)organic combustion products, such as ash, which affects the chemical-physical properties of the dust and its atmospheric interaction with radiation and cloud microphysics (Hand et al., 2010; Phillips et al., 2022). However, within the current study an external mixture of mineral dust and BC was assumed, which has also implications on the total model AOD. While both aerosol classes simply sum up if an external mixing is considered, an internal mixture of the fire-dust emissions would reduce the total (dust) AOD. Furthermore, the emission process of combustion ash is likely very similar to that of soil-dust particles and could be partly included in the enhanced measured AOD over fire-active regions. Thus, the here presented values are likely an upper limit of the additional dust/aerosol load caused by vegetation fires. The impact of an internal mixing with other combustion aerosol can indeed be a fruitful linkage for future investigations of the atmospheric impacts of the fire-dust emissions.

Conclusion and Outlook

A newly developed parameterization that relates dust emissions to vegetation fires was implemented in ICON-HAM and analyzed for the 10-year period 2004–2013. Fire-dust emissions can contribute significantly to the atmospheric dust load with average emissions in the order of 230 (190–255) Tg yr⁻¹, which corresponds to up to 18 (15–21) % of the globally emitted dust mass with respect to our ICON-HAM simulations. In contrast to the classic, purely wind-driven dust emissions, that are largely confined to the 'dust belt' region, fire-dust emissions can occur nearly worldwide on basically all continents and represent a significant source of atmospheric dust. In regions, where classic dust reservoirs, such as deserts and other sparsely vegetated landscapes, are rare, fire-dust emissions can even dominate over the purely wind-driven dust emissions. Such regions are found within the semi-arid savannas of sub-Saharan Africa and parts of South America, where fires are not only a very common component of local land management and agricultural practices during the dry season, but are also used for large-scale deforestation. These regions are in good agreement with the hotspots of enhanced post-fire dust activity as found by Yu and Ginoux (2022), and confirm that the fire impacts on the vegetation cover create suitable conditions for dust emission during and after a fire.

While the global sum of fire-dust emissions is rather constant over time, interannual variability on a regional perspective is more pronounced and dust emissions can vary by a factor of up to three. In regions where the fire activity is largely controlled by human activity, the related dust emissions are subject to a rather small interannual variability. In contrast, in areas where fire occurrence is more strongly governed by changing weather and climate conditions, such as heat waves or long-lasting droughts (or even by the political framework conditions), fires represent a much more variable source of dust emission. Some of these rather irregularly occurring fires, that are affecting grass-, crop-, and shrublands and are common in parts of Australia, Eastern Europe and Central Asia, and the Midwestern US, have been identified as a significant source of the related fire-dust emissions. While fire occurrence is undoubtedly the most important precondition for fire-dust emission, it was shown that significant dust emission fluxes can occur despite rather weak fire intensities if the soil-surface conditions are supportive. This includes rather sparse and fragile vegetation that is supposed to be effectively burned away, so that the fire-induced winds can attack the soil surface, and is in accordance with the observed post-fire emissions even after moderate fires (Yu & Ginoux, 2022). If the soil type is moreover rather emissive, the related dust emissions can be quite significant. Such regions were identified north of the Black Sea, where the occurrence of dust emissions in the context of wildfires is in good agreement with the observations of Nisantzi et al. (2014). The same accounts for regions in the US, where Kavouras et al. (2012) identified soil-dust emissions during fires in the Great Basin desert. In contrast, the more intense fires within forest-dominated biomes were largely excluded from emitting soil-dust particles as the fire-induced winds can hardly interact with a soil layer that is possibly still covered by plants or vegetation remnants, and where the strength of fire-dust emission fluxes is strongly reduced by the large surface roughness that is typical for such landscapes. In essence, a more intense vegetation fire is not necessarily linked to large fire-dust emissions in the same way; it rather depends on the interplay of fire strength, the local soil-surface properties such as soil type, roughness length, and the burning vegetation type. Atmospheric winds can modulate the strength of fire-dust emissions even further on a local scale.

Once released and mobilized from the soil surface, the pyro-convective updrafts enable the dust particles to get injected into higher tropospheric levels. To simulate this behavior, a proven parameterization for fire emissions based on Val Martin et al. (2010) was applied. The emitted dust particles are injected within and partially even above the PBL. Thus, the additional fire-dust emissions do not only lead to an increasing total atmospheric dust load but also to an elevated height of the mean dust layer and to a possibly extended atmospheric lifetime enabling long-range transport. As the fire-dust injection process allows even for an entrainment of larger, coarse-mode particles into the free troposphere, fire-dust emission could contribute to the occasionally findings of mineral dust particles far away from the typical desert source regions (van der Does et al., 2018). Although the fire-based injection parameterization represents a realistic emission pathway for fire-dust emissions, even these injection heights tend to be an underestimation in the case of very severe fires where pyrocumulunimbus clouds form and combustion-related aerosols can reach even the stratosphere (Heinold et al., 2022). Reisner et al. (2023) suspected that also dust particles can be injected by pyroCbs in these altitudes, which would increase their atmospheric lifetime and climate impacts even further. An investigation of different injection parameterizations for the fire-dust emissions might help to resolve these uncertainties and presents an aim for further research activities. This is supported by the ongoing development of ICON-HAM, which enables multiple further options to test and refine the parameterization of fire-dust emissions. This includes nudged model versions with synchronized atmospheric conditions, a limited area model for regional studies (ICON-HAM-LAM), models with a reduced complexity allowing for more efficient and therefore higher resolved simulations (ICON-HAM-lite, Weiss et al. (2024)), and combinations of those (ICON-HAM-lite-LAM). These will allow for more detailed regional case studies, where a more case-sensitive approach can be pursued, for example, the representation of the diurnal cycle of fire activity, an application of in-situ data of fire-affected PSDs, and the investigation of the atmospheric impacts of the additional fire-dust emissions such as on radiation and cloud feedbacks or circulation changes.

As fires have and will always been part of the Earth system, the here presented fire-dust parameterization has the potential for many further research activities, such as an investigation of the fire-dust emission potential that arises from changing climate conditions and the related adjustments of the global vegetation distribution that are expected in the future (Gonzalez et al., 2010). A look back into the past could also provide interesting insights as not only the land use was drastically altered by human activity, but also the number and regional distribution of fires is nowadays much different than it was during pre-anthropogenic times and compared to how it will be in the future (Hamilton et al., 2018; Pechony & Shindell, 2010). This of course has altered the related fire-dust emissions and the current global dust load is nowadays likely very different than it used to be in a natural climate before human interventions.

Appendix

Appendix : List of Coefficients Used for the Calculation of the Fire-Dust Emission Fluxes - A

This section contains a table of the size-separated coefficents a_i(D),…,g_i(D) that were used for the calculation of the fire-dust emission fluxes as given by Equations 1–4 for all of the five soil types implemented in ICON-HAM.

Table A1 Size-Separated Coefficients Applied for Equations 1–4

Soil type	Particle size	$${a}_{i}(D)$$	$${b}_{i}(D)$$	$${c}_{i}(D)$$	$${d}_{i}(D)$$	$${e}_{i}(D)$$	$${f}_{i}(D)$$	$${g}_{i}(D)$$
Coarse	$$D\in [0.1,0.288]\,{\upmu }\mathrm{m}$$	$$2.765\cdot 1{0}^{-15}$$	$$2.431\cdot 1{0}^{-13}$$	$$0.0$$	$$2.092\cdot 1{0}^{-11}$$	1.374	$$0.0$$	0.0
$$D\in [0.288,0.87]\,\mu \mathrm{m}$$	$$1.580\cdot 1{0}^{-11}$$	$$1.389\cdot 1{0}^{-9}$$	$$0.0$$	$$1.195\cdot 1{0}^{-7}$$	1.374	$$0.0$$	0.0
$$D\in [0.87,2.63]\,\mu \mathrm{m}$$	$$3.655\cdot 1{0}^{-9}$$	$$3.212\cdot 1{0}^{-7}$$	$$0.0$$	$$2.765\cdot 1{0}^{-5}$$	1.374	$$0.0$$	0.0
$$D\in [2.63,7.94]\,\mu \mathrm{m}$$	$$3.655\cdot 1{0}^{-9}$$	$$3.212\cdot 1{0}^{-7}$$	$$4.071\cdot 1{0}^{-11}$$	$$6.338\cdot 1{0}^{-4}$$	1.374	$$1.579\cdot 1{0}^{-4}$$	0.3162
$$D\in [7.94,24.0]\,\mu \mathrm{m}$$	$$9.672\cdot 1{0}^{-8}$$	$$2.825\cdot 1{0}^{-5}$$	$$7.923\cdot 1{0}^{-6}$$	$$1.607\cdot 1{0}^{-3}$$	1.231	$$3.119\cdot 1{0}^{-3}$$	0.3214
Medium	$$D\in [0.1,0.288]\,\mu \mathrm{m}$$	$$9.015\cdot 1{0}^{-11}$$	$$1.042\cdot 1{0}^{-8}$$	$$5.049\cdot 1{0}^{-15}$$	$$7.774\cdot 1{0}^{-7}$$	1.359	$$2.013\cdot 1{0}^{-12}$$	0.336
$$D\in [0.288,0.87]\,\mu \mathrm{m}$$	$$5.410\cdot 1{0}^{-9}$$	$$6.250\cdot 1{0}^{-7}$$	$$3.318\cdot 1{0}^{-11}$$	$$4.665\cdot 1{0}^{-5}$$	1.359	$$1.306\cdot 1{0}^{-8}$$	0.3271
$$D\in [0.87,2.63]\,\mu \mathrm{m}$$	$$2.055\cdot 1{0}^{-8}$$	$$2.374\cdot 1{0}^{-6}$$	$$1.388\cdot 1{0}^{-8}$$	$$1.772\cdot 1{0}^{-4}$$	1.359	$$5.406\cdot 1{0}^{-6}$$	0.3201
$$D\in [2.63,7.94]\,\mu \mathrm{m}$$	$$1.831\cdot 1{0}^{-8}$$	$$2.115\cdot 1{0}^{-6}$$	$$1.839\cdot 1{0}^{-6}$$	$$1.579\cdot 1{0}^{-4}$$	1.359	$$7.130\cdot 1{0}^{-4}$$	0.3165
$$D\in [7.94,24.0]\,\mu \mathrm{m}$$	$$1.109\cdot 1{0}^{-8}$$	$$3.384\cdot 1{0}^{-6}$$	$$5.007\cdot 1{0}^{-5}$$	$$1.889\cdot 1{0}^{-4}$$	1.236	$$1.962\cdot 1{0}^{-2}$$	0.3202
Fine	$$D\in [0.1,0.288]\,\mu \mathrm{m}$$	$$4.726\cdot 1{0}^{-11}$$	$$2.282\cdot 1{0}^{-9}$$	$$6.104\cdot 1{0}^{-9}$$	$$1.877\cdot 1{0}^{-7}$$	1.754	$$2.438\cdot 1{0}^{-6}$$	0.3369
$$D\in [0.288,0.87]\,\mu \mathrm{m}$$	$$2.836\cdot 1{0}^{-9}$$	$$1.369\cdot 1{0}^{-7}$$	$$4.387\cdot 1{0}^{-7}$$	$$1.502\cdot 1{0}^{-5}$$	1.532	$$1.730\cdot 1{0}^{-4}$$	0.3282
$$D\in [0.87,2.63]\,\mu \mathrm{m}$$	$$1.067\cdot 1{0}^{-8}$$	$$5.149\cdot 1{0}^{-7}$$	$$6.331\cdot 1{0}^{-6}$$	$$5.648\cdot 1{0}^{-5}$$	1.532	$$2.469\cdot 1{0}^{-3}$$	0.3211
$$D\in [2.63,7.94]\,\mu \mathrm{m}$$	$$7.126\cdot 1{0}^{-9}$$	$$3.441\cdot 1{0}^{-7}$$	$$4.380\cdot 1{0}^{-5}$$	$$3.774\cdot 1{0}^{-5}$$	1.532	$$1.699\cdot 1{0}^{-2}$$	0.3169
$$D\in [7.94,24.0]\,\mu \mathrm{m}$$	$$2.744\cdot 1{0}^{-9}$$	$$4.403\cdot 1{0}^{-7}$$	$$1.343\cdot 1{0}^{-4}$$	$$2.703\cdot 1{0}^{-5}$$	1.388	$$5.247\cdot 1{0}^{-2}$$	0.3194
Coarse medium	$$D\in [0.1,0.288]\,\mu \mathrm{m}$$	$$5.826\cdot 1{0}^{-11}$$	$$1.110\cdot 1{0}^{-8}$$	$$0.0$$	$$7.172\cdot 1{0}^{-7}$$	1.262	$$0.0$$	0.0
$$D\in [0.288,0.87]\,\mu \mathrm{m}$$	$$3.496\cdot 1{0}^{-9}$$	$$6.662\cdot 1{0}^{-7}$$	$$5.644\cdot 1{0}^{-15}$$	$$4.303\cdot 1{0}^{-5}$$	1.262	$$2.221\cdot 1{0}^{-12}$$	0.3268
$$D\in [0.87,2.63]\,\mu \mathrm{m}$$	$$1.326\cdot 1{0}^{-8}$$	$$2.526\cdot 1{0}^{-6}$$	$$6.461\cdot 1{0}^{-11}$$	$$1.632\cdot 1{0}^{-4}$$	1.262	$$2.514\cdot 1{0}^{-8}$$	0.3197
$$D\in [2.63,7.94]\,\mu \mathrm{m}$$	$$1.133\cdot 1{0}^{-8}$$	$$2.159\cdot 1{0}^{-6}$$	$$1.276\cdot 1{0}^{-7}$$	$$1.395\cdot 1{0}^{-4}$$	1.262	$$4.947\cdot 1{0}^{-5}$$	0.3164
$$D\in [7.94,24.0]\,\mu \mathrm{m}$$	$$6.625\cdot 1{0}^{-9}$$	$$3.019\cdot 1{0}^{-6}$$	$$2.279\cdot 1{0}^{-5}$$	$$1.624\cdot 1{0}^{-4}$$	1.161	$$8.948\cdot 1{0}^{-3}$$	0.3207
Medium fine	$$D\in [0.1,0.288]\,\mu \mathrm{m}$$	$$4.115\cdot 1{0}^{-11}$$	$$5.396\cdot 1{0}^{-9}$$	$$5.472\cdot 1{0}^{-18}$$	$$3.833\cdot 1{0}^{-7}$$	1.339	$$2.181\cdot 1{0}^{-15}$$	0.3356
$$D\in [0.288,0.87]\,\mu \mathrm{m}$$	$$2.469\cdot 1{0}^{-9}$$	$$3.221\cdot 1{0}^{-7}$$	$$1.007\cdot 1{0}^{-12}$$	$$2.299\cdot 1{0}^{-5}$$	1.339	$$3.965\cdot 1{0}^{-10}$$	0.3265
$$D\in [0.87,2.63]\,\mu \mathrm{m}$$	$$9.290\cdot 1{0}^{-9}$$	$$1.212\cdot 1{0}^{-6}$$	$$3.449\cdot 1{0}^{-9}$$	$$8.653\cdot 1{0}^{-5}$$	1.339	$$1.342\cdot 1{0}^{-6}$$	0.3198
$$D\in [2.63,7.94]\,\mu \mathrm{m}$$	$$6.304\cdot 1{0}^{-9}$$	$$8.227\cdot 1{0}^{-7}$$	$$2.007\cdot 1{0}^{-6}$$	$$5.873\cdot 1{0}^{-5}$$	1.339	$$7.781\cdot 1{0}^{-4}$$	0.3164
$$D\in [7.94,24.0]\,\mu \mathrm{m}$$	$$2.533\cdot 1{0}^{-9}$$	$$8.201\cdot 1{0}^{-7}$$	$$1.173\cdot 1{0}^{-4}$$	$$4.542\cdot 1{0}^{-5}$$	1.225	$$4.598\cdot 1{0}^{-2}$$	0.3204

Acknowledgments

Robert Wagner and Kerstin Schepanski are funded by the Deutsche Forschungsgemeinschaft (German Research Foundation, DFG), project number 432456920. This work used resources of the Deutsches Klimarechenzentrum (DKRZ) granted by its Scientific Steering Committee (WLA) under project ID bb1242. We thank the AERONET principal investigators and their staff for establishing and maintaining the sites used in this study, and the authors of the MIDAS data set. Furthermore, we acknowledge the HAMMOZ community for providing the model framework and some analysis scripts used for model validation, in particular Anne Kubin for preparing the GFAS input data on the ICON grid. We also thank the three anonymous reviewers for their helpful recommendations that helped to improve the manuscript. Open Access funding enabled and organized by Projekt DEAL.

Data Availability Statement

The model data that were used to prepare the plots and on which the statements in this study are based are available under zenodo (Wagner, 2024). The data set contains for the 10 years simulation period 2004–2013 the monthly averaged GFAS input file of the FRP, and monthly, seasonal, or yearly averaged fields of the (dust) aerosol optical depth, the wind- and fire-driven dust emission fluxes, as well as the atmospheric dust concentration together with basic atmospheric information for both simulations with and without the additional fire-dust emission.

References

Achard, F., Eva, H. D., Mollicone, D., & Beuchle, R. (2008). The effect of climate anomalies and human ignition factor on wildfires in Russian boreal forests. Philosophical Transactions of the Royal Society B: Biological Sciences, 363(1501), 2329–2337. https://doi.org/10.1098/rstb.2007.2203

Albani, S., Mahowald, N. M., Perry, A. T., Scanza, R. A., Zender, C. S., Heavens, N. G., et al. (2014). Improved dust representation in the community atmosphere model. Journal of Advances in Modeling Earth Systems, 6(3), 541–570. https://doi.org/10.1002/2013MS000279

Alves, C., Gonãlves, C., Pio, C., Mirante, F., Caseiro, A., Tarelho, L., et al. (2010). Smoke emissions from biomass burning in a Mediterranean shrubland. Atmospheric Environment, 44(25), 3024–3033. https://doi.org/10.1016/j.atmosenv.2010.05.010

Andreae, M., & Rosenfeld, D. (2008). Aerosol‐cloud‐precipitation interactions. Part 1. The nature and sources of cloud‐active aerosols. Earth‐Science Reviews, 89(1), 13–41. https://doi.org/10.1016/j.earscirev.2008.03.001

Balch, J. K., Bradley, B. A., Abatzoglou, J. T., Nagy, R. C., Fusco, E. J., & Mahood, A. L. (2017). Human‐started wildfires expand the fire niche across the United States. Proceedings of the National Academy of Sciences of the United States of America, 114(11), 2946–2951. https://doi.org/10.1073/pnas.1617394114

Barry, K. R., Hill, T. C. J., Levin, E. J. T., Twohy, C. H., Moore, K. A., Weller, Z. D., et al. (2021). Observations of ice nucleating particles in the free troposphere from Western US wildfires. Journal of Geophysical Research: Atmospheres, 126(3), e2020JD033752. https://doi.org/10.1029/2020JD033752

Bellouin, N., Quaas, J., Gryspeerdt, E., Kinne, S., Stier, P., Watson‐Parris, D., et al. (2020). Bounding global aerosol radiative forcing of climate change. Reviews of Geophysics, 58(1), e2019RG000660. https://doi.org/10.1029/2019RG000660

Blank, R. R., Allen, F. L., & Young, J. A. (1996). Influence of simulated burning of soil‐litter from low sagebrush, squirreltail, cheatgrass, and medusahead on water‐soluble anions and cations. International Journal of Wildland Fire, 6(3), 137–143. https://doi.org/10.1071/WF9960137

Boening, C., Willis, J. K., Landerer, F. W., Nerem, R. S., & Fasullo, J. (2012). The 2011 La Niña: So strong, the oceans fell. Geophysical Research Letters, 39(19), L19602. https://doi.org/10.1029/2012GL053055

Cakmur, R. V., Miller, R. L., Perlwitz, J., Geogdzhayev, I. V., Ginoux, P., Koch, D., et al. (2006). Constraining the magnitude of the global dust cycle by minimizing the difference between a model and observations. Journal of Geophysical Research, 111(D6), D06207. https://doi.org/10.1029/2005JD005791

Chalbot, M.‐C., Nikolich, G., Etyemezian, V., Dubois, D., King, J., Shafer, D., et al. (2013). Soil humic‐like organic compounds in prescribed fire emissions using nuclear magnetic resonance spectroscopy. Environmental Pollution, 181, 167–171. https://doi.org/10.1016/j.envpol.2013.06.008

Chappell, A., Webb, N. P., Hennen, M., Zender, C. S., Ciais, P., Schepanski, K., et al. (2023). Elucidating hidden and enduring weaknesses in dust emission modeling. Journal of Geophysical Research: Atmospheres, 128(17), e2023JD038584. https://doi.org/10.1029/2023JD038584

Chen, S., Lu, X., Ma, X., Zhu, X., Gao, Q., Jiang, X., et al. (2023). Aerosol emission from soil during field burning. Atmospheric Environment, 309, 119924. https://doi.org/10.1016/j.atmosenv.2023.119924

Cheng, T., Peng, Y., Feichter, J., & Tegen, I. (2008). An improvement on the dust emission scheme in the global aerosol‐climate model ECHAM5‐HAM. Atmospheric Chemistry and Physics, 8(4), 1105–1117. https://doi.org/10.5194/acp‐8‐1105‐2008

Cook, B. I., Mankin, J. S., & Anchukaitis, K. J. (2018). Climate change and drought: From past to future. Current Climate Change Reports, 4(2), 164–179. https://doi.org/10.1007/s40641‐018‐0093‐2

Croft, B., Lohmann, U., Martin, R. V., Stier, P., Wurzler, S., Feichter, J., et al. (2010). Influences of in‐cloud aerosol scavenging parameterizations on aerosol concentrations and wet deposition in ECHAM5‐HAM. Atmospheric Chemistry and Physics, 10(4), 1511–1543. https://doi.org/10.5194/acp‐10‐1511‐2010

Croft, B., Lohmann, U., Martin, R. V., Stier, P., Wurzler, S., Feichter, J., et al. (2009). Aerosol size‐dependent below‐cloud scavenging by rain and snow in the ECHAM5‐HAM. Atmospheric Chemistry and Physics, 9(14), 4653–4675. https://doi.org/10.5194/acp‐9‐4653‐2009

Crueger, T., Giorgetta, M. A., Brokopf, R., Esch, M., Fiedler, S., Hohenegger, C., et al. (2018). ICON‐A, the atmosphere component of the ICON Earth system model: II. Model evaluation. Journal of Advances in Modeling Earth Systems, 10(7), 1638–1662. https://doi.org/10.1029/2017MS001233

Das, R., Wang, X., Khezri, B., Webster, R. D., Itoh, M., Shiodera, S., et al. (2023). Suspension of crustal materials from wildfire in Indonesia as revealed by Pb isotope analysis. ACS Earth and Space Chemistry, 7(2), 379–387. https://doi.org/10.1021/acsearthspacechem.2c00270

Dipankar, A., Stevens, B., Heinze, R., Moseley, C., Zängl, G., Giorgetta, M., & Brdar, S. (2015). Large eddy simulation using the general circulation model ICON. Journal of Advances in Modeling Earth Systems, 7(3), 963–986. https://doi.org/10.1002/2015MS000431

Doherty, O. M., Riemer, N., & Hameed, S. (2008). Saharan mineral dust transport into the Caribbean: Observed atmospheric controls and trends. Journal of Geophysical Research, 113(D7), D07211. https://doi.org/10.1029/2007JD009171

Dubovik, O., Smirnov, A., Holben, B. N., King, M. D., Kaufman, Y. J., Eck, T. F., & Slutsker, I. (2000). Accuracy assessments of aerosol optical properties retrieved from Aerosol Robotic Network (AERONET) Sun and sky radiance measurements. Journal of Geophysical Research, 105(D8), 9791–9806. https://doi.org/10.1029/2000JD900040

Dukes, D., Gonzales, H. B., Ravi, S., Grandstaff, D. E., Van Pelt, R. S., Li, J., et al. (2018). Quantifying postfire aeolian sediment transport using rare Earth element tracers. Journal of Geophysical Research: Biogeosciences, 123(1), 288–299. https://doi.org/10.1002/2017JG004284

Earl, N., & Simmonds, I. (2018). Spatial and temporal variability and trends in 2001–2016 global fire activity. Journal of Geophysical Research: Atmospheres, 123(5), 2524–2536. https://doi.org/10.1002/2017JD027749

Earl, N., Simmonds, I., & Tapper, N. (2015). Weekly cycles of global fires—Associations with religion, wealth and culture, and insights into anthropogenic influences on global climate. Geophysical Research Letters, 42(21), 9579–9589. https://doi.org/10.1002/2015GL066383

Flannigan, M. D., Krawchuk, M. A., de Groot, W. J., Wotton, B. M., & Gowman, L. M. (2009). Implications of changing climate for global wildland fire. International Journal of Wildland Fire, 18(5), 483–507. https://doi.org/10.1071/WF08187

Foken, T., & Napo, C. J. (2008). Micrometeorology (Vol. 2). Springer. https://doi.org/10.1007/978‐3‐642‐25440‐6

Gaudichet, A., Echalar, F., Chatenet, B., Quisefit, J. P., Malingre, G., Cachier, H., et al. (1995). Trace elements in tropical African savanna biomass burning aerosols. Journal of Atmospheric Chemistry, 22(1), 19–39. https://doi.org/10.1007/BF00708179

Giglio, L., Randerson, J. T., & van der Werf, G. R. (2013). Analysis of daily, monthly, and annual burned area using the fourth‐generation global fire emissions database (GFED4). Journal of Geophysical Research: Biogeosciences, 118(1), 317–328. https://doi.org/10.1002/jgrg.20042

Giglio, L., van der Werf, G. R., Randerson, J. T., Collatz, G. J., & Kasibhatla, P. (2006). Global estimation of burned area using MODIS active fire observations. Atmospheric Chemistry and Physics, 6(4), 957–974. https://doi.org/10.5194/acp‐6‐957‐2006

Gill, R. L., Fleck, R., Chau, K., Westerhausen, M. T., Lockwood, T. E., Violi, J. P., et al. (2024). Fine particle pollution during megafires contains potentially toxic elements. Environmental Pollution, 344, 123306. https://doi.org/10.1016/j.envpol.2024.123306

Gillette, D., & Walker, T. R. (1977). Characteristics of airborne particles produced by wind erosion of sandy soil, high plains of west Texas. Soil Science, 123(2), 97–110. https://doi.org/10.1097/00010694‐197702000‐00004

Giorgetta, M. A., Brokopf, R., Crueger, T., Esch, M., Fiedler, S., Helmert, J., et al. (2018). ICON‐A, the atmosphere component of the ICON Earth system model: I. Model description. Journal of Advances in Modeling Earth Systems, 10(7), 1613–1637. https://doi.org/10.1029/2017MS001242

Gkikas, A., Proestakis, E., Amiridis, V., Kazadzis, S., Di Tomaso, E., Tsekeri, A., et al. (2021). ModIs Dust AeroSol (MIDAS): A global fine‐resolution dust optical depth data set. Atmospheric Measurement Techniques, 14(1), 309–334. https://doi.org/10.5194/amt‐14‐309‐2021

Gonzalez, P., Neilson, R. P., Lenihan, J. M., & Drapek, R. J. (2010). Global patterns in the vulnerability of ecosystems to vegetation shifts due to climate change. Global Ecology and Biogeography, 19(6), 755–768. https://doi.org/10.1111/j.1466‐8238.2010.00558.x

Hagemann, S. (2002). An improved land surface parameter dataset for global and regional climate models (Tech. Rep.). https://doi.org/10.17617/2.2344576

Hamilton, D. S., Hantson, S., Scott, C., Kaplan, J. O., Pringle, K., Nieradzik, L. P., et al. (2018). Reassessment of pre‐industrial fire emissions strongly affects anthropogenic aerosol forcing. Nature Communications, 9(1), 3182. https://doi.org/10.1038/s41467‐018‐05592‐9

Hamilton, D. S., Perron, M. M., Bond, T. C., Bowie, A. R., Buchholz, R. R., Guieu, C., et al. (2022). Earth, wind, fire, and pollution: Aerosol nutrient sources and impacts on ocean biogeochemistry. Annual Review of Marine Science, 14(1), 303–330. https://doi.org/10.1146/annurev‐marine‐031921‐013612

Hand, V. L., Capes, G., Vaughan, D. J., Formenti, P., Haywood, J. M., & Coe, H. (2010). Evidence of internal mixing of African dust and biomass burning particles by individual particle analysis using electron beam techniques. Journal of Geophysical Research, 115(D13), D13301. https://doi.org/10.1029/2009JD012938

Haywood, J. M., Pelon, J., Formenti, P., Bharmal, N., Brooks, M., Capes, G., et al. (2008). Overview of the dust and biomass‐burning experiment and African monsoon multidisciplinary analysis special observing period‐0. Journal of Geophysical Research, 113(D23), D00C17. https://doi.org/10.1029/2008JD010077

Heinold, B., Baars, H., Barja, B., Christensen, M., Kubin, A., Ohneiser, K., et al. (2022). Important role of stratospheric injection height for the distribution and radiative forcing of smoke aerosol from the 2019–2020 Australian wildfires. Atmospheric Chemistry and Physics, 22(15), 9969–9985. https://doi.org/10.5194/acp‐22‐9969‐2022

Heinold, B., Tegen, I., Schepanski, K., & Banks, J. R. (2016). New developments in the representation of Saharan dust sources in the aerosol‐climate model ECHAM6‐HAM2. Geoscientific Model Development, 9(2), 765–777. https://doi.org/10.5194/gmd‐9‐765‐2016

Heinold, B., Tegen, I., Schepanski, K., & Hellmuth, O. (2008). Dust radiative feedback on Saharan boundary layer dynamics and dust mobilization. Geophysical Research Letters, 35(20). https://doi.org/10.1029/2008GL035319

Holben, B. N., Eck, T., Slutsker, I., Tanré, D., Buis, J., Setzer, A., et al. (1998). AERONET—A federated instrument network and data archive for aerosol characterization. Remote Sensing of Environment, 66(1), 1–16. https://doi.org/10.1016/S0034‐4257(98)00031‐5

Holben, B. N., Tanré, D., Smirnov, A., Eck, T. F., Slutsker, I., Abuhassan, N., et al. (2001). An emerging ground‐based aerosol climatology: Aerosol optical depth from AERONET. Journal of Geophysical Research, 106(D11), 12067–12097. https://doi.org/10.1029/2001JD900014

Huang, Y., Wu, S., & Kaplan, J. O. (2015). Sensitivity of global wildfire occurrences to various factors in the context of global change. Atmospheric Environment, 121, 86–92. https://doi.org/10.1016/j.atmosenv.2015.06.002

Huneeus, N., Schulz, M., Balkanski, Y., Griesfeller, J., Kinne, S., Prospero, J., et al. (2011). Global dust model intercomparison in AeroCom phase I. Atmospheric Chemistry and Physics, 11(15), 7781–7816. https://doi.org/10.5194/acp‐11‐7781‐2011

IPCC. (2023). H. Lee, Core Writing Team, & J. Romero (Eds.), In Climate Change 2023: Synthesis Report. Contribution of Working Groups I, II and III to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. IPCC. https://doi.org/10.59327/IPCC/AR6‐9789291691647

Johnson, B. T., Shine, K. P., & Forster, P. M. (2004). The semi‐direct aerosol effect: Impact of absorbing aerosols on marine stratocumulus. Quarterly Journal of the Royal Meteorological Society, 130(599), 1407–1422. https://doi.org/10.1256/qj.03.61

Jungclaus, J. H., Lorenz, S. J., Schmidt, H., Brovkin, V., Brüggemann, N., Chegini, F., et al. (2022). The ICON Earth system model version 1.0. Journal of Advances in Modeling Earth Systems, 14(4), e2021MS002813. https://doi.org/10.1029/2021MS002813

Justice, C., Giglio, L., Korontzi, S., Owens, J., Morisette, J., Roy, D., et al. (2002). The MODIS fire products. Remote Sensing of Environment, 83(1), 244–262. https://doi.org/10.1016/S0034‐4257(02)00076‐7

Kaiser, J. W., Heil, A., Andreae, M. O., Benedetti, A., Chubarova, N., Jones, L., et al. (2012). Biomass burning emissions estimated with a global fire assimilation system based on observed fire radiative power. Biogeosciences, 9(1), 527–554. https://doi.org/10.5194/bg‐9‐527‐2012

Kavouras, I. G., Nikolich, G., Etyemezian, V., DuBois, D. W., King, J., & Shafer, D. (2012). In situ observations of soil minerals and organic matter in the early phases of prescribed fires. Journal of Geophysical Research, 117(D12), D12313. https://doi.org/10.1029/2011JD017420

Klamt, J., Giersch, S., & Raasch, S. (2024). Saltation‐induced dust emission of dust devils in the convective boundary layer—An LES study on the meter scale. Journal of Geophysical Research: Atmospheres, 129(7), e2023JD040058. https://doi.org/10.1029/2023JD040058

Klimont, Z., Kupiainen, K., Heyes, C., Purohit, P., Cofala, J., Rafaj, P., et al. (2017). Global anthropogenic emissions of particulate matter including black carbon. Atmospheric Chemistry and Physics, 17(14), 8681–8723. https://doi.org/10.5194/acp‐17‐8681‐2017

Klose, M., Shao, Y., Li, X., Zhang, H., Ishizuka, M., Mikami, M., & Leys, J. F. (2014). Further development of a parameterization for convective turbulent dust emission and evaluation based on field observations. Journal of Geophysical Research: Atmospheres, 119(17), 10441–10457. https://doi.org/10.1002/2014JD021688

Knippertz, P., & Todd, M. C. (2012). Mineral dust aerosols over the Sahara: Meteorological controls on emission and transport and implications for modeling. Reviews of Geophysics, 50(RG1007). https://doi.org/10.1029/2011RG000362

Kok, J. F. (2011). Does the size distribution of mineral dust aerosols depend on the wind speed at emission? Atmospheric Chemistry and Physics, 11(19), 10149–10156. https://doi.org/10.5194/acp‐11‐10149‐2011

Kok, J. F., Mahowald, N. M., Fratini, G., Gillies, J. A., Ishizuka, M., Leys, J. F., et al. (2014). An improved dust emission model—Part 1: Model description and comparison against measurements. Atmospheric Chemistry and Physics, 14(23), 13023–13041. https://doi.org/10.5194/acp‐14‐13023‐2014

Kokkola, H., Korhonen, H., Lehtinen, K. E. J., Makkonen, R., Asmi, A., Järvenoja, S., et al. (2008). SALSA—A sectional aerosol module for large scale applications. Atmospheric Chemistry and Physics, 8(9), 2469–2483. https://doi.org/10.5194/acp‐8‐2469‐2008

Lamarque, J.‐F., Bond, T. C., Eyring, V., Granier, C., Heil, A., Klimont, Z., et al. (2010). Historical (1850–2000) gridded anthropogenic and biomass burning emissions of reactive gases and aerosols: Methodology and application. Atmospheric Chemistry and Physics, 10(15), 7017–7039. https://doi.org/10.5194/acp‐10‐7017‐2010

Leuenberger, D., Koller, M., Fuhrer, O., & Schär, C. (2010). A generalization of the SLEVE vertical coordinate. Monthly Weather Review, 138(9), 3683–3689. https://doi.org/10.1175/2010MWR3307.1

Li, J., Carlson, B. E., Yung, Y. L., Lv, D., Hansen, J., Penner, J. E., et al. (2022). Scattering and absorbing aerosols in the climate system. Nature Reviews Earth & Environment, 3(6), 363–379. https://doi.org/10.1038/s43017‐022‐00296‐7

Li, W., Xu, L., Liu, X., Zhang, J., Lin, Y., Yao, X., et al. (2017). Air pollution‐aerosol interactions produce more bioavailable iron for ocean ecosystems. Science Advances, 3(3), e1601749. https://doi.org/10.1126/sciadv.1601749

Li, Z., Rosenfeld, D., & Fan, J. (2017). Aerosols and their impact on radiation, clouds, precipitation, and severe weather events. In Oxford research encyclopedia of environmental science. Oxford University Press. https://doi.org/10.1093/acrefore/9780199389414.013.126

Loukas, A., Vasiliades, L., & Tzabiras, J. (2008). Climate change effects on drought severity. Advances in Geosciences, 17, 23–29. https://doi.org/10.5194/adgeo‐17‐23‐2008

Lu, H., & Shao, Y. (1999). A new model for dust emission by saltation bombardment. Journal of Geophysical Research, 104(D14), 16827–16842. https://doi.org/10.1029/1999JD900169

Maenhaut, W., Salma, I., Cafmeyer, J., Annegarn, H. J., & Andreae, M. O. (1996). Regional atmospheric aerosol composition and sources in the eastern Transvaal, South Africa, and impact of biomass burning. Journal of Geophysical Research, 101(D19), 23631–23650. https://doi.org/10.1029/95JD02930

Mahowald, N., Ward, D. S., Kloster, S., Flanner, M. G., Heald, C. L., Heavens, N. G., et al. (2011). Aerosol impacts on climate and biogeochemistry. Annual Review of Environment and Resources, 36(1), 45–74. https://doi.org/10.1146/annurev‐environ‐042009‐094507

Mamouri, R.‐E., Ansmann, A., Ohneiser, K., Knopf, D. A., Nisantzi, A., Bühl, J., et al. (2023). Wildfire smoke triggers cirrus formation: Lidar observations over the eastern Mediterranean. Atmospheric Chemistry and Physics, 23(22), 14097–14114. https://doi.org/10.5194/acp‐23‐14097‐2023

Mariani, M., Fletcher, M.‐S., Holz, A., & Nyman, P. (2016). ENSO controls interannual fire activity in southeast Australia. Geophysical Research Letters, 43(20), 10891–10900. https://doi.org/10.1002/2016GL070572

Marinou, E., Amiridis, V., Binietoglou, I., Tsikerdekis, A., Solomos, S., Proestakis, E., et al. (2017). Three‐dimensional evolution of Saharan dust transport towards Europe based on a 9‐year earlinet‐optimized calipso dataset. Atmospheric Chemistry and Physics, 17(9), 5893–5919. https://doi.org/10.5194/acp‐17‐5893‐2017

Marticorena, B., & Bergametti, G. (1995). Modeling the atmospheric dust cycle: 1. Design of a soil‐derived dust emission scheme. Journal of Geophysical Research, 100(D8), 16415–16430. https://doi.org/10.1029/95JD00690

Marticorena, B., Bergametti, G., Aumont, B., Callot, Y., N'Doumé, C., & Legrand, M. (1997). Modeling the atmospheric dust cycle: 2. Simulation of Saharan dust sources. Journal of Geophysical Research, 102(D4), 4387–4404. https://doi.org/10.1029/96JD02964

Maudlin, L., Wang, Z., Jonsson, H., & Sorooshian, A. (2015). Impact of wildfires on size‐resolved aerosol composition at a coastal California site. Atmospheric Environment, 119, 59–68. https://doi.org/10.1016/j.atmosenv.2015.08.039

McNabb, D., & Swanson, F. (1990). Effects of fire on soil erosion. In D. S. J. D. Walstad & S. R. Radosevich (Eds.), Natural and prescribed fire in Pacific Northwest forests (pp. 159–176). Oregon State University Press.

Meloni, D., di Sarra, A., Monteleone, F., Pace, G., Piacentino, S., & Sferlazzo, D. (2008). Seasonal transport patterns of intense Saharan dust events at the Mediterranean Island of Lampedusa. Atmospheric Research, 88(2), 134–148. https://doi.org/10.1016/j.atmosres.2007.10.007

Menut, L., Siour, G., Bessagnet, B., Cholakian, A., Pennel, R., & Mailler, S. (2022). Impact of wildfires on mineral dust emissions in Europe. Journal of Geophysical Research: Atmospheres, 127(24), e2022JD037395. https://doi.org/10.1029/2022JD037395

Nagy, R. C., Fusco, E., Bradley, B., Abatzoglou, J. T., & Balch, J. (2018). Human‐related ignitions increase the number of large wildfires across U.S. ecoregions. Fire, 1(1), 4. https://doi.org/10.3390/fire1010004

Nisantzi, A., Mamouri, R. E., Ansmann, A., & Hadjimitsis, D. (2014). Injection of mineral dust into the free troposphere during fire events observed with polarization lidar at Limassol, Cyprus. Atmospheric Chemistry and Physics, 14(22), 12155–12165. https://doi.org/10.5194/acp‐14‐12155‐2014

Oreopoulos, L., Cho, N., & Lee, D. (2020). A global survey of apparent aerosol‐cloud interaction signals. Journal of Geophysical Research: Atmospheres, 125(1), e2019JD031287. https://doi.org/10.1029/2019JD031287

Palmer, T. Y. (1981). Large fire winds, gases and smoke. Atmospheric Environment, 15(10–11), 2079–2090. https://doi.org/10.1016/0004‐6981(81)90241‐9

Paris, R., Desboeufs, K. V., Formenti, P., Nava, S., & Chou, C. (2010). Chemical characterisation of iron in dust and biomass burning aerosols during AMMA‐SOP0/DABEX: Implication for iron solubility. Atmospheric Chemistry and Physics, 10(9), 4273–4282. https://doi.org/10.5194/acp‐10‐4273‐2010

Pechony, O., & Shindell, D. T. (2010). Driving forces of global wildfires over the past millennium and the forthcoming century. Proceedings of the National Academy of Sciences of the United States of America, 107(45), 19167–19170. https://doi.org/10.1073/pnas.1003669107

Pérez, C., Haustein, K., Janjic, Z., Jorba, O., Huneeus, N., Baldasano, J. M., et al. (2011). Atmospheric dust modeling from meso to global scales with the online NMMB/BSC‐Dust model—Part 1: Model description, annual simulations and evaluation. Atmospheric Chemistry and Physics, 11(24), 13001–13027. https://doi.org/10.5194/acp‐11‐13001‐2011

Pérez‐Cabello, F., dela Riva Fernández, J., Montorio Llovería, R., & García‐Martín, A. (2006). Mapping erosion‐sensitive areas after wildfires using fieldwork, remote sensing, and geographic information systems techniques on a regional scale. Journal of Geophysical Research, 111(G4), G04S10. https://doi.org/10.1029/2005JG000148

Perron, M. M., Meyerink, S., Corkill, M., Strzelec, M., Proemse, B. C., Gault‐Ringold, M., et al. (2022). Trace elements and nutrients in wildfire plumes to the southeast of Australia. Atmospheric Research, 270, 106084. https://doi.org/10.1016/j.atmosres.2022.106084

Phillips, C., Nair, U., & Christopher, S. (2022). The influence of dust‐smoke mixtures on boundary layer processes and nocturnal warming in the Sahel. Journal of Geophysical Research: Atmospheres, 127(11), e2021JD036349. https://doi.org/10.1029/2021JD036349

Pio, C., Legrand, M., Alves, C., Oliveira, T., Afonso, J., Caseiro, A., et al. (2008). Chemical composition of atmospheric aerosols during the 2003 summer intense forest fire period. Atmospheric Environment, 42(32), 7530–7543. https://doi.org/10.1016/j.atmosenv.2008.05.032

Prospero, J. M. (1996). Saharan dust transport over the North Atlantic Ocean and Mediterranean: An overview. In S. Guerzoni & R. Chester (Eds.), The impact of desert dust across the Mediterranean (pp. 133–151). Springer Netherlands. https://doi.org/10.1007/978‐94‐017‐3354‐0_13

Raddatz, T., Reick, C., Knorr, W., Kattge, J., Roeckner, E., Schnur, R., et al. (2007). Will the tropical land biosphere dominate the climate‐carbon cycle feedback during the twenty‐first century? Climate Dynamics, 29(6), 565–574. https://doi.org/10.1007/s00382‐007‐0247‐8

Radke, L. F., Hegg, D. A., Hobbs, P. V., Nance, J. D., Lyons, J. H., Laursen, K. K., et al. (1991). Particulate and trace gas emissions from large biomass fires in North America. In J. Levin (Ed.), Global biomass burning: Atmospheric, climatic, and biospheric implications (pp. 209–216). The MIT Press. https://doi.org/10.7551/mitpress/3286.003.0032

Ratcliffe, N. G., Ryder, C. L., Bellouin, N., Woodward, S., Jones, A., Johnson, B., et al. (2024). Long range transport of coarse mineral dust: An evaluation of the Met Office Unified Model against aircraft observations. EGUsphere, 2024, 1–32. https://doi.org/10.5194/egusphere‐2024‐806

Raupach, M. R., Gillette, D. A., & Leys, J. F. (1993). The effect of roughness elements on wind erosion threshold. Journal of Geophysical Research, 98(D2), 3023–3029. https://doi.org/10.1029/92JD01922

Ravi, S., Baddock, M. C., Zobeck, T. M., & Hartman, J. (2012). Field evidence for differences in post‐fire aeolian transport related to vegetation type in semi‐arid grasslands. Aeolian Research, 7, 3–10. https://doi.org/10.1016/j.aeolia.2011.12.002

Reick, C. H., Raddatz, T., Brovkin, V., & Gayler, V. (2013). Representation of natural and anthropogenic land cover change in MPI‐ESM. Journal of Advances in Modeling Earth Systems, 5(3), 459–482. https://doi.org/10.1002/jame.20022

Reisner, J. M., Josephson, A. J., Gorkowski, K. J., Koo, E., Thompson, D. K., Schroeder, D., & Dubey, M. K. (2023). Informed multi‐scale approach applied to the British Columbia fires of late summer 2017. Journal of Geophysical Research: Atmospheres, 128(5), e2022JD037238. https://doi.org/10.1029/2022JD037238

Rieger, D., Bangert, M., Bischoff‐Gauss, I., Förstner, J., Lundgren, K., Reinert, D., et al. (2015). ICON‐ART 1.0—A new online‐coupled model system from the global to regional scale. Geoscientific Model Development, 8(6), 1659–1676. https://doi.org/10.5194/gmd‐8‐1659‐2015

Roberts, G., Wooster, M., & Lagoudakis, E. (2009). Annual and diurnal african biomass burning temporal dynamics. Biogeosciences, 6(5), 849–866. https://doi.org/10.5194/bg‐6‐849‐2009

Roesch, C., Roesch, M., Wolf, M. J., Zawadowicz, M. A., AlAloula, R., Awwad, Z., & Cziczo, D. J. (2021). CCN and INP activity of middle eastern soil dust. Aeolian Research, 52, 100729. https://doi.org/10.1016/j.aeolia.2021.100729

Sá, A. C., Pereira, J. M., Charlton, M. E., Mota, B., Barbosa, P. M., & Stewart Fotheringham, A. (2011). The pyrogeography of sub‐Saharan Africa: A study of the spatial non‐stationarity of fire‐environment relationships using GWR. Journal of Geographical Systems, 13(3), 227–248. https://doi.org/10.1007/s10109‐010‐0123‐7

Salzmann, M., Ferrachat, S., Tully, C., Münch, S., Watson‐Parris, D., Neubauer, D., et al. (2022). The global atmosphere‐aerosol model ICON‐A‐HAM2.3—Initial model evaluation and effects of radiation balance tuning on aerosol optical thickness. Journal of Advances in Modeling Earth Systems, 14(4), e2021MS002699. https://doi.org/10.1029/2021MS002699

Schepanski, K., Tegen, I., & Macke, A. (2009). Saharan dust transport and deposition towards the tropical northern Atlantic. Atmospheric Chemistry and Physics, 9(4), 1173–1189. https://doi.org/10.5194/acp‐9‐1173‐2009

Schepanski, K., Tegen, I., Todd, M. C., Heinold, B., Bönisch, G., Laurent, B., & Macke, A. (2009). Meteorological processes forcing Saharan dust emission inferred from MSG‐SEVIRI observations of subdaily dust source activation and numerical models. Journal of Geophysical Research, 114(D10), D10201. https://doi.org/10.1029/2008JD010325

Schlosser, J. S., Braun, R. A., Bradley, T., Dadashazar, H., MacDonald, A. B., Aldhaif, A. A., et al. (2017). Analysis of aerosol composition data for western United States wildfires between 2005 and 2015: Dust emissions, chloride depletion, and most enhanced aerosol constituents. Journal of Geophysical Research: Atmospheres, 122(16), 8951–8966. https://doi.org/10.1002/2017JD026547

Seinfeld, J. H., & Pandis, S. N. (2006). Atmospheric chemistry and physics: From air pollution to climate change. John Wiley & Sons, Inc.

Senande‐Rivera, M., Insua‐Costa, D., & Miguez‐Macho, G. (2022). Spatial and temporal expansion of global wildland fire activity in response to climate change. Nature Communications, 13(1), 1208. https://doi.org/10.1038/s41467‐022‐28835‐2

Stevens, B., Giorgetta, M., Esch, M., Mauritsen, T., Crueger, T., Rast, S., et al. (2013). Atmospheric component of the MPI‐M Earth system model: ECHAM6. Journal of Advances in Modeling Earth Systems, 5(2), 146–172. https://doi.org/10.1002/jame.20015

Stier, P., Feichter, J., Kinne, S., Kloster, S., Vignati, E., Wilson, J., et al. (2005). The aerosol‐climate model ECHAM5‐HAM. Atmospheric Chemistry and Physics, 5(4), 1125–1156. https://doi.org/10.5194/acp‐5‐1125‐2005

Stout, J. E., & Zobeck, T. M. (1997). Intermittent saltation. Sedimentology, 44(5), 959–970. https://doi.org/10.1046/j.1365‐3091.1997.d01‐55.x

Susott, R. A., Ward, D. E., Babbitt, R. E., & Latham, D. J. (1991). J. Levin (Ed.), The measurement of trace emissions and combustion characteristics for a mass fire [Chapter 32] (pp. 245–257). The MIT Press. https://doi.org/10.7551/mitpress/3286.003.0036

Tang, R., Mao, J., Jin, M., Chen, A., Yu, Y., Shi, X., et al. (2021). Interannual variability and climatic sensitivity of global wildfire activity. Advances in Climate Change Research, 12(5), 686–695. https://doi.org/10.1016/j.accre.2021.07.001

Tegen, I., Harrison, S. P., Kohfeld, K., Prentice, I. C., Coe, M., & Heimann, M. (2002). Impact of vegetation and preferential source areas on global dust aerosol: Results from a model study. Journal of Geophysical Research, 107(D21), 4576. https://doi.org/10.1029/2001JD000963

Tegen, I., Neubauer, D., Ferrachat, S., Siegenthaler‐Le Drian, C., Bey, I., Schutgens, N., et al. (2019). The global aerosol‐climate model ECHAM6.3‐HAM2.3—Part 1: Aerosol evaluation. Geoscientific Model Development, 12(4), 1643–1677. https://doi.org/10.5194/gmd‐12‐1643‐2019

Tegen, I., & Schepanski, K. (2009). The global distribution of mineral dust. IOP Conference Series: Earth and Environmental Science, 7(1), 012001. https://doi.org/10.1088/1755‐1307/7/1/012001

Tegen, I., & Schepanski, K. (2018). Climate feedback on aerosol emission and atmospheric concentrations. Current Climate Change Reports, 4(1), 1–10. https://doi.org/10.1007/s40641‐018‐0086‐1

Twohy, C. H., Toohey, D. W., Levin, E. J. T., DeMott, P. J., Rainwater, B., Garofalo, L. A., et al. (2021). Biomass burning smoke and its influence on clouds over the western U. S. Geophysical Research Letters, 48(15), e2021GL094224. https://doi.org/10.1029/2021GL094224

Twomey, S. (1977). The influence of pollution on the shortwave albedo of clouds. Journal of the Atmospheric Sciences, 34(7), 1149–1152. https://doi.org/10.1175/1520‐0469(1977)034<1149:tiopot>2.0.co;2

Val Martin, M., Logan, J. A., Kahn, R. A., Leung, F.‐Y., Nelson, D. L., & Diner, D. J. (2010). Smoke injection heights from fires in North America: Analysis of 5 years of satellite observations. Atmospheric Chemistry and Physics, 10(4), 1491–1510. https://doi.org/10.5194/acp‐10‐1491‐2010

van der Does, M., Knippertz, P., Zschenderlein, P., Harrison, R. G., & Stuut, J.‐B. (2018). The mysterious long‐range transport of giant mineral dust particles. Science Advances, 4(12), eaau2768. https://doi.org/10.1126/sciadv.aau2768

Vignati, E., Wilson, J., & Stier, P. (2004). M7: An efficient size‐resolved aerosol microphysics module for large‐scale aerosol transport models. Journal of Geophysical Research, 109(D22), D22202. https://doi.org/10.1029/2003JD004485

Wagner, R. (2024). “Dataset related to the manuscript Wagner and Schepanski (submitted to JAMES, 2024)”: Quantifying fire‐driven dust emissions using a global aerosol model. Zenodo. https://doi.org/10.5281/zenodo.11204182

Wagner, R., Jähn, M., & Schepanski, K. (2018). Wildfires as a source of airborne mineral dust—Revisiting a conceptual model using large‐eddy simulation (LES). Atmospheric Chemistry and Physics, 18(16), 11863–11884. https://doi.org/10.5194/acp‐18‐11863‐2018

Wagner, R., Schepanski, K., & Klose, M. (2021). The dust emission potential of agricultural‐like fires—Theoretical estimates from two conceptually different dust emission parameterizations. Journal of Geophysical Research: Atmospheres, 126(18), e2020JD034355. https://doi.org/10.1029/2020JD034355

Wang, R., Tao, S., Shen, H., Huang, Y., Chen, H., Balkanski, Y., et al. (2014). Trend in global black carbon emissions from 1960 to 2007. Environmental Science and Technology, 48(12), 6780–6787. https://doi.org/10.1021/es5021422

Ward, D. S., Shevliakova, E., Malyshev, S., & Rabin, S. (2018). Trends and variability of global fire emissions due to historical anthropogenic activities. Global Biogeochemical Cycles, 32(1), 122–142. https://doi.org/10.1002/2017GB005787

Wardell, A., Nielsen, T., Rasmussen, K., & Mbow, C. (2004). Fire history, fire regimes and fire management in West Africa: An overview. In J. Goldammer & C. de Ronde (Eds.), Wildland fire management handbook for Sub‐Sahara Africa (pp. 350–382). Global Fire Monitoring Center.

Weis, J., Schallenberg, C., Chase, Z., Bowie, A. R., Wojtasiewicz, B., Perron, M. M. G., et al. (2022). Southern Ocean phytoplankton stimulated by wildfire emissions and sustained by iron recycling. Geophysical Research Letters, 49(11), e2021GL097538. https://doi.org/10.1029/2021GL097538

Weiss, P., Herbert, R., & Stier, P. (2024). ICON‐HAM‐lite: Simulating the Earth system with interactive aerosols at kilometer scales. EGUsphere, 2024, 1–26. https://doi.org/10.5194/egusphere‐2024‐3325

Wesely, M. (1989). Parameterization of surface resistances to gaseous dry deposition in regional‐scale numerical models. Atmospheric Environment, 23(6), 1293–1304. (1967). https://doi.org/10.1016/0004‐6981(89)90153‐4

Wolfe, S. A., & Nickling, W. G. (1996). Shear stress partitioning in sparsely vegetated desert canopies. Earth Surface Processes and Landforms, 21(7), 607–619. https://doi.org/10.1002/(SICI)1096‐9837(199607)21:7⟨607::AID‐ESP660⟩3.0.CO;2‐1

Wooster, M. J., Roberts, G., Perry, G. L. W., & Kaufman, Y. J. (2005). Retrieval of biomass combustion rates and totals from fire radiative power observations: FRP derivation and calibration relationships between biomass consumption and fire radiative energy release. Journal of Geophysical Research, 110(D24), D24311. https://doi.org/10.1029/2005JD006318

Wu, C., Lin, Z., & Liu, X. (2020). The global dust cycle and uncertainty in CMIP5 (Coupled Model Intercomparison Project phase 5) models. Atmospheric Chemistry and Physics, 20(17), 10401–10425. https://doi.org/10.5194/acp‐20‐10401‐2020

Yu, Y., & Ginoux, P. (2021). Assessing the contribution of the ENSO and MJO to Australian dust activity based on satellite‐ and ground‐based observations. Atmospheric Chemistry and Physics, 21(11), 8511–8530. https://doi.org/10.5194/acp‐21‐8511‐2021

Yu, Y., & Ginoux, P. (2022). Enhanced dust emission following large wildfires due to vegetation disturbance. Nature Geoscience, 15(11), 878–884. https://doi.org/10.1038/s41561‐022‐01046‐6

Zängl, G., Reinert, D., Rípodas, P., & Baldauf, M. (2015). The ICON (ICOsahedral Non‐hydrostatic) modelling framework of DWD and MPI‐M: Description of the non‐hydrostatic dynamical core. Quarterly Journal of the Royal Meteorological Society, 141(687), 563–579. https://doi.org/10.1002/qj.2378

Zhang, K., O’Donnell, D., Kazil, J., Stier, P., Kinne, S., Lohmann, U., et al. (2012). The global aerosol‐climate model ECHAM‐HAM, version 2: Sensitivity to improvements in process representations. Atmospheric Chemistry and Physics, 12(19), 8911–8949. https://doi.org/10.5194/acp‐12‐8911‐2012

Zhao, A., Ryder, C. L., & Wilcox, L. J. (2022). How well do the CMIP6 models simulate dust aerosols? Atmospheric Chemistry and Physics, 22(3), 2095–2119. https://doi.org/10.5194/acp‐22‐2095‐2022