Effects of Pollen on Hydrometeors and Precipitation in a Convective System

Anemophilous (wind‐driven) pollen is one type of primary biological aerosol particle, which can rupture under high humidity conditions and form smaller sub‐pollen particles (SPPs). Both pollen and SPPs can reach the upper troposphere under convective conditions, acting as cloud condensation nuclei (CCN) and ice nucleating particles (INPs), thus influencing cloud formation and precipitation. However, the impacts of these biological aerosols on cold cloud formation and local climate remain unclear as there are large uncertainties on their emission flux and ice nucleating abilities. Here, we incorporate pollen emission and rupture processes in the Weather Research and Forecasting Model with Chemistry (WRF‐Chem) simulations and update the Morrison microphysics scheme within WRF‐Chem using aerosol‐aware INP parameterizations to account for pollen in addition to other anthropogenic and biogenic aerosol. INP parameterizations for pollen and SPP are derived from laboratory experiments. When including pollen rupture rates as observed in a series of chamber studies, SPP concentrations increase, leading to an increase of cloud ice and water by up to 50% and potentially extending the duration of the convective system. Among all simulated hydrometeors, graupel and raindrops exhibit the largest enhancements from the inclusion of SPPs, with intensifying precipitation at the backside of the convective system and a greater spatial extent. Sensitivity simulations indicate that SPPs have a greater effect on cloud microphysical processes than whole pollen grains, and further observational evidence is needed to constrain these processes.

INPs are crucial in initiating cloud ice formation, and can enable heterogeneous freezing processes for atmospheric temperatures between approximately 38 and 0°C and influence ice crystal formation in cirrus clouds when temperatures are lower than 38°C (Burrows et al., 2022;Kanji et al., 2017).Variations in INP concentration have been found to alter cloud ice content and ice particle radius, thereby impacting atmospheric radiation (X.Zeng et al., 2009aZeng et al., , 2009b;;B. Zhao et al., 2018).In previous studies, the impacts of INP on clouds have been investigated, considering sources such as dust, black carbon, and sea salt (McGraw et al., 2020;Shi et al., 2022;X. Zeng & Li, 2022;Y. Zeng et al., 2023).In comparison to other INPs (e.g., dust, black carbon, soot, sea salt, and organic aerosol), pollen and other PBAPs can facilitate ice nucleation at warmer temperatures (up to 5°C) (e.g., Kanji et al., 2017;Möhler et al., 2007;Patade et al., 2021;Schnell & Vali, 1973;Testa et al., 2021), potentially acting as an important INP source in PBAP abundant environments such as forest ecosystems (Prenni et al., 2009;Testa et al., 2021;Tobo et al., 2013).Recent long-term observations by Schneider et al. (2021a) show that the abundance of biological particles (e.g., PBAP) is closely correlated with the variations in INP concentrations, with seasonal peaks of INPs attributed to peaks in pollen and other PBAP.Moreover, the co-occurrence of the INP and PBAP maxima has been observed in other studies (Schumacher et al., 2013).
Despite their importance, PBAPs, including pollen, are typically not included in the modeling studies due to uncertainties in emission fluxes and their ice nucleating abilities (Hoose, Kristjánsson, & Burrows, 2010;Huang et al., 2021).The exclusion of PBAPs in models further limits our understanding of their effects on cloud formation and precipitation.There are two main gaps that need to be addressed for this issue.The first gap involves a lack of understanding of the concentration of pollen and SPPs in the atmosphere.Currently, pollen observational data are sparse due to the limitation of measurement approaches (Buters et al., 2018;Ziska et al., 2019), and existing observations are typically limited to small spatial scales (Y.Zhang et al., 2015;Ziska et al., 2019) and/or short time periods (Grewling et al., 2012;Ziello et al., 2012).To overcome spatial limitations, models have been developed based on observed pollen-meteorology correlations to simulate long-term pollen data over large spatial scales (e.g., continental).However, these models often only include a subset of the total types of pollen species emitted, leading to an underestimate of regional pollen concentrations (e.g., Anenberg et al., 2017;Hamaoui-Laguel et al., 2015;Neumann et al., 2019;Prank et al., 2013;Sofiev et al., 2013;Y. Zhang et al., 2013).To address this, we developed the Pollen Emissions for Climate Models version 2.0 (PECMv2.0), a climate-flexible pollen emission model that simulates daily pollen emission of 13 of the most prevalent pollen emitting species over the Unites States (Wozniak & Steiner, 2017;Y. Zhang & Steiner, 2022c).Subba et al. (2023) coupled this model with the Chemistry version of Weather Research and Forecast model (WRF-Chem 3.9.1)(Grell et al., 2005), enabling the simulation of atmospheric pollen and SPPs concentrations with meteorology and providing a foundation for future studies on pollen-cloud interactions.
The second knowledge gap relates to a lack of representation of the freezing mechanisms by pollen and SPPs in the atmosphere.Previous studies have developed INP parameterizations for specific types of PBAPs, including pollen, using deterministic and classic nucleation theory-based approaches (e.g., Hoose, Kristjánsson, Chen, & Hazra, 2010;Patade et al., 2021;Phillips et al., 2009Phillips et al., , 2013;;Schneider et al., 2021a;Tobo et al., 2013).However, for studies based on field experiments, the attribution of INP activities to different PBAP types are ambiguous (e.g., Phillips et al., 2009Phillips et al., , 2013;;Schneider et al., 2021a;Tobo et al., 2013).Furthermore, most field experiments do not include smaller PBAP fragments (e.g., <1 μm such as SPPs), which may underestimate the role of PBAP as INP (Patade et al., 2021;Tobo et al., 2013).Pollen and SPPs could be relevant for different pathways in heterogenous freezing, including immersion, contact, and deposition freezing (Gute & Abbatt, 2020).To date, several laboratory experiments have examined the ice nucleating ability of pollen and SPPs (e.g., Augustin et al., 2013;Burkert-Kohn et al., 2017;Diehl et al., 2001Diehl et al., , 2002;;Gute & Abbatt, 2020;O'Sullivan et al., 2015;Pummer et al., 2012Pummer et al., , 2015;;Tong et al., 2015;von Blohn et al., 2005).Most of these studies have focused on either whole pollen grains or SPPs, but not both.Additionally, most prior work examined the freezing of pollen based on experimental techniques that rely on using SPPs generated from wash water solutions (i.e., where pollen is placed in the solution and the resulting "wash water" analyzed, as in Pummer et al., 2012) rather than ambient SPPs, which may alter the SPPs ice nucleating behavior.A more recent laboratory study using an environmental chamber to generate the SPPs examined the immersion freezing temperature of pollen and SPPs and found that the SPPs were not as effective for immersion freezing as compared to prior studies (Matthews et al., 2023).
Prior modeling approaches to investigate the importance of PBAPs in cloud formation processes have incorporated INP parameterizations for PBAPs into parcel models or climate models (e.g., Diehl & Mitra, 2015;Hoose, Kristjánsson, & Burrows, 2010;Hummel et al., 2018;Patade et al., 2022;Phillips et al., 2009;Werchner et al., 2022).Several studies concluded that the inclusion of PBAPs (including pollen) in cloud microphysical processes is not sufficient to have significant influence on cloud ice nucleation (Diehl & Mitra, 2015;Hoose, Kristjánsson, & Burrows, 2010;Patade et al., 2022).However, other studies show that PBAPs are able to alter cloud and precipitation processes when there are sufficiently high concentrations (Phillips et al., 2009), when the cloud top temperature is below 15°C (Hummel et al., 2018), or when their ice nucleation efficiency is doubled (Werchner et al., 2022).The discrepancies of these studies could be attributed to different model parameterization approaches, which can strongly impact results (Burrows et al., 2022;Sahyoun et al., 2016).Additionally, the extrapolation to pollen and other PBAPs' importance to cloud formation is hampered by the fact that previous studies have not included realistic pollen emissions in the model framework, as many of these studies used constant pollen emissions.For example, Hoose, Kristjánsson, and Burrows (2010) did not account for environmental factors or pollen daily variations, and Diehl and Mitra (2015) used a fixed number concentration of pollen per droplet based on previous literature.Patade et al. (2022) assumed a fixed fraction of insoluble organic carbon as PBAPs and derived the fraction of PBAP composition from a previous study located in the Amazon (Patade et al., 2021).Finally, the generation of SPPs has not been included in prior simulations (Diehl & Mitra, 2015;Hoose, Kristjánsson, & Burrows, 2010;Patade et al., 2022), which may be important as SPPs could have more impacts on cloud formation processes due to their smaller size and longer atmospheric residence time.Werchner et al. (2022) investigated the impacts of SPPs on cloud ice nucleation processes, yet the INP parametrization incorporated (Phillips et al., 2013) is not specifically developed for SPPs.This parameterization treats all the PBAP types (e.g., pollen, fungi, bacteria, and virus) as the same and does not account for the large ice nucleation variability between different PBAP types, which may potentially underestimate or inaccurately estimate their influence.
In this work, we present new simulations of PBAPs and their interaction with clouds using pollen-WRF-Chem coupled modeling framework, including comprehensive pollen emissions (PECMv2.0;Wozniak & Steiner, 2017;Y. Zhang & Steiner, 2022c) and the transport, rupture, and fate of pollen in the atmosphere (Subba et al., 2023).The pollen rupture rate is measured by a realistic laboratory experiment using an environmental chamber (Matthews et al., 2023).Additionally, we update the Morrison two-moment bulk scheme inside WRF-Chem to include aerosol-aware INP parameterizations for both heterogeneous freezing (immersion, contact, and deposition freezing) and homogeneous freezing.The updated microphysics scheme simulates ice nucleation from anthropogenic and natural aerosol (e.g., dust, soot, sea salt, sulfate) as well as pollen and SPP, where the INP parametrizations for pollen and SPP are obtained from laboratory experiments (Matthews et al., 2023).All aerosol included in the microphysics scheme are coupled to the chemistry scheme and interact with meteorology within the model framework.
Using the updated model framework, we focus our analysis on a convective event in the United States Southern Great Plains (SGP) during April 2013, a time period with simultaneously high pollen emissions and intense thunderstorm activity (Subba et al., 2023).This event involves a mesoscale convective system triggered by a cold front and presents an ideal case for studying the impacts of pollen and SPP on both warm and ice cloud processes due to the convective nature and size of the system.Through a series of sensitivity experiments, we aim to address three objectives of this work: (a) to evaluate the sensitivity of simulated cloud properties to different INP parameterizations used for pollen and SPPs, (b) to compare the effects of pollen and SPPs on cloud microphysical properties and hydrometeor evolution, and (c) to assess when and where pollen and SPPs are most impactful to cloud processes and precipitation.This work aims to shed light on biological aerosol-cloud interactions and their impact on Earth's radiative budget and hydrological cycle.

Pollen Emissions
Pollen emissions are simulated by the Pollen Emissions for Climate Models version 2.0 (PECMv2.0)(Wozniak & Steiner, 2017;Y. Zhang & Steiner, 2022c), a climate-flexible model that simulates gridded daily pollen emission over the United States.The model includes 13 of the most prevalent wind-pollinating taxa (pollen emitting vegetations) over the United States: Acer, Alnus, Ambrosia, Betula, Cupressaceae, Fraxinus, Poaceae, Morus, Pinaceae, Platanus, Populus, Quercus, and Ulmus, which account for 77% of the total pollen counts in the Unites States from 2003 to 2010 (Wozniak & Steiner, 2017).Total pollen production and pollen phenology are simulated based on an empirical relationship with prior-year average temperatures, and the daily pollen emission is calculated offline based on a Gaussian distribution within the pollen emitting period (Figure S1 in Supporting Information S1).We scale the total daily pollen emission potential (grains m 2 day 1 ) of all 13 pollen emitting taxa into a total pollen hourly emission flux that is incorporated into the modal aerosol module in WRF-Chem, the Modal Aerosol Dynamics Model for Europe (MADE) (Ackermann et al., 1998) with the Secondary Organic Aerosol Model (SORGAM) (Schell et al., 2001) (MADE/SORGAM) (Subba et al., 2023).We simulate two pollen tracers in the model: (a) total whole grain pollen, as tracking individual pollen types would include a large number of tracers, and (b) ruptured pollen or SPPs (Figures 1a and 1b).MADE/SORGAM is a modal aerosol model with three aerosol log-normal size distribution modes: Aitken (0.01-0.10 μm), accumulation (0.1-1.0 μm), and coarse (1-10 μm) (Schell et al., 2001).The Aitken and accumulation modes include 16 chemical species: ammonium, nitrate, sulfate, sodium, chloride, elemental carbon, primary organic aerosol, and nine secondary organic aerosol types simulated by SORGAM.The coarse mode includes three chemical constituents: sea salt, unspeciated anthropogenic aerosol, and soil dust.The whole pollen grain is added in the coarse mode and the ruptured pollen or SPPs are added in the accumulation mode (Subba et al., 2023).Within each mode, aerosols are separately tracked as interstitial (particles suspended in the air) and cloud-borne (particles embedded in cloud droplets).Interstitial aerosol can be converted to cloud-borne aerosol through aerosol activation, and this determines the number of cloud droplets nucleated (Abdul-Razzak & Ghan, 2000, 2002).
The coupled pollen-WRF model accounts for the impact of meteorology, including wind, precipitation, and relative humidity on the daily total pollen emission potential (Subba et al., 2023;Wozniak et al., 2018).Once in the atmosphere, pollen rupture occurs when relative humidity exceeds 80%, and a large amount (70%) of whole pollen grains will be converted into SPPs (e.g., EXP4 in Subba et al., 2023).The default rupture rate is estimated to be 1,000 SPPs grain 1 based on previous studies (Stone et al., 2021;Suphioglu et al., 1992), and we increase this rate to 1.25 × 10 5 SPPs grain 1 for a sensitivity analysis based on laboratory experiments (Matthews et al., 2023).Primary pollen and SPPs concentrations are dynamic in the model, and interact with the main 12:00 to April 18 12:00, 2013) spatial concentration of (a) pollen (grains/m 3 ) and (b) sub-pollen particles (SPPs) (grains/m 3 ) within the planetary boundary layer (PBL).Blue dashed square bounds the area of interests in this study, which also has the highest ice/liquid water path and precipitation (see Figure 4).The simulated pollen and SPPs concentration are from experiment M23.atmospheric processes, including rupture, transport, gas-aqueous-aerosol phase chemistry, influence on radiation, interaction with clouds and wet/dry deposition.The CCN activation is impacted by hygroscopicity of different aerosol types.In this work, we use an average hygroscopicity value of 0.08 for pollen and 0.16 for SPPs (M.Zhang et al., 2021).
In the atmosphere, heterogeneous freezing occurs through three predominant pathways: immersion, contact, and deposition freezing (Kanji et al., 2017).Immersion freezing occurs when an INP is fully immersed in a water droplet, while contact freezing occurs following a collision between an interstitial particle and a supercooled droplet (Burrows et al., 2022;Kanji et al., 2017).Deposition freezing occurs when water vapor deposits directly on a particle to form ice (Burrows et al., 2022).Within MMS, these freezing processes are simulated by different INP parameterizations.When the atmospheric temperature is below 4°C, cloud drops freeze through contact and immersion freezing in the model.Immersion freezing follows the Bigg (1953) parametrization, while the contact freezing is simulated based on a flux of contact nuclei to the droplets due to Brownian motion, diffusiophoresis, and thermophoresis, with an effective diffusion coefficient given by Young (1974).The contact nuclei number is calculated based on Meyers et al. (1992) (Morrison et al., 2005).Deposition freezing occurs when the atmospheric temperature is lower than 8°C and the supersaturation ratio is larger than 0.999 (Cooper, 1986).However, because the ice number concentration simulated by the Cooper parameterization becomes unrealistically large with colder temperatures, the original MMS sets the upper limit of formed ice number concentration as 500 L 1 for deposition nucleation.For these three freezing pathways, immersion and contact freezing mechanisms involve freezing cloud droplets to form ice particles.Consequently, in the model simulations, the total number of simulated ice particles generated through these two pathways cannot exceed the cloud droplet number concentration.Additionally, the instant homogeneous freezing of cloud and rain droplets is assumed to occur below 40°C in the model (Morrison et al., 2005).Secondary ice formation is simulated through rime-splintering (Hallett & Mossop, 1974).

Inclusion of Aerosol-Aware Microphysics in the MMS
The heterogenous freezing parameterizations (e.g., Bigg, 1953;Cooper, 1986;Meyers et al., 1992) in the original MMS solely depend on the temperature or ice supersaturation, and some (e.g., Cooper, 1986;Meyers et al., 1992) fail to reproduce observed ice nuclei concentration (DeMott et al., 2010).Here we update the MMS by replacing these parameterizations with aerosol-aware parameterizations from literature for both mixed phase cloud (Section 2.3.1) and cirrus cloud (Section 2.3.2),which show improved agreement of ice number-temperature correlations observed in natural environment (DeMott et al., 2010).

Mixed-Phase Cloud Microphysics
For mixed-phase cloud, we have replaced Meyers et al. (1992) and Bigg (1953) with the DeMott et al. ( 2010) parameterization (Equation 1; hereafter D10) to simulate contact and immersion freezing.Deposition freezing is not included in the revised mixed-phase cloud microphysics, as we would expect this to contribute much less to the total freezing than other processes (Hoose et al., 2008).In D10, the ice nuclei number concentration (n IN,T k ; L 1 ) depends on temperature (T k ; K) and the aerosol number concentration with diameters larger than 0.5 μm (n aer,0.5 ; cm 3 ): a-d are constant parameters, where a = 0.0000594, b = 3.33, c = 0.0264, d = 0.0033.Immersion freezing ice is simulated by D10 when the saturation ratio respect to water is larger than 0.999 and the temperature is between 40 and 9°C.We incorporate cloud-borne aerosol as n aer,0.5 for immersion freezing, because immersion freezing is initiated by INPs in cloud droplets.The onset of contact freezing occurs at slightly warmer temperature (4-5°C higher) than immersion freezing (e.g., Diehl et al., 2002;Durant & Shaw, 2005;Fornea et al., 2009;Matthews et al., 2023;Shaw et al., 2005) and we set the temperature range for contact freezing from 35 to 4°C.The formula for contact nuclei number concentration is assumed to be the same as immersion freezing (D10), but the temperature is shifted higher from T k into T k T 0 (Barahona et al., 2014) with T 0 equal to 5 K, consistent with lab studies (Fornea et al., 2009;Ladino et al., 2011).In the model, contact freezing occurs through the collision between contact nuclei and cloud droplet, and the contact nuclei number concentration is simulated by D10 using interstitial aerosol as n aer,0.5 .The simulated contacted nuclei number concentration will be further implemented in Young (1974) scheme to simulate the ice number concentration.
All aerosols in the aerosol-aware microphysics are determined prognostically from the aerosol chemistry component of the model.Atmospheric aerosol concentrations reflect all atmospheric processes, including emission, transport, transformation and by wet/dry deposition.All aerosols with diameters larger than 0.5μm can be INPs as part of the D10 scheme, including coarse mode aerosol such as sea salt, soil dust, and anthropogenic aerosol.Part of the accumulation mode aerosol has diameter larger than 0.5 μm and we include this subset of the accumulation mode in the D10 scheme (Ackermann et al., 1998).
To examine the impacts of pollen on mixed-phase cloud microphysics, we implement new pollen INP parametrizations based on recent laboratory experiments that generated SPPs in an environmental chamber that controls wind, humidity, and light.Pollen rupture rates are measured for three types of pollen (live oak, ryegrass and giant ragweed; Matthews et al., 2023).The emitted SPPs and pollen collected from the chamber are used in ice nucleating experiments using a cold stage (Linkam Scientific Instruments Model LTS 120) mounted onto an optical microscope (Olympus Model BX43F), which uses repeated freeze and thaw cycles to quantify the immersion and contact freezing temperature of pollen grains and immersion freezing temperature of SPPs.The contact freezing temperature of SPP was not examined because SPP is too small to conduct the experiments.Whole pollen grains and SPPs show different freezing characteristics, with SPPs freezing at colder temperatures than intact pollen grains.This suggests pollen grains can act as more effective INPs than SPPs.Because we do not track individual pollen types within the model, we use an average of three pollen types determined experimentally and consider the effects of three different freezing mechanisms: pollen immersion freezing, pollen contact freezing, and SPP immersion freezing.We incorporate these measured fractional freezing parameterizations (hereafter M23) in the MMS with the same formula (Equations 2 and 3) and revised parameters: where F freezing is pollen freezing fraction at different temperatures T (K), n freezing is the ice number concentration formed by pollen and SPP immersion/contact freezing, n pollen is the number concentration of pollen or SPPs.For pollen immersion freezing, l = 1.005,T 0 = 246.23,m = 0.692, n = 1.009; for pollen contact freezing, l = 1.009,T 0 = 249.35,m = 0.684, n = 1.010; and for SPP immersion freezing, l = 1.015,T 0 = 242.4,m = 0.819, n = 1.016.
Note that we assume one freezing droplet contains one pollen grain per SPP when applying the fractional freezing curve in the model, however the number of SPPs immersed in one droplet could be greater.As a result, model simulations may overestimate SPPs immersion freezing efficiency.As contact freezing data for SPPs is not available, we group interstitial SPPs with diameters >0.5 μm with other aerosol to input in the D10 and Young (1974) scheme for contact freezing.

Ice Cirrus Cloud Microphysics
In the original MMS, all cloud and rain droplets with cloud-borne aerosol freeze homogeneously when temperatures are < 40°C.However, heterogenous freezing can also occur at these low temperatures (Cziczo et al., 2013;Heymsfield et al., 2017;Kanji et al., 2017;Koop et al., 2000), and the conditions of homogenous freezing can vary depending on solutes contained within droplets (Koop et al., 2000).At higher altitudes which are colder and drier, deposition (not considered for mixed phase clouds) becomes a more important heterogeneous nucleation mechanism.The first revision for cirrus formation is to replace the Cooper (1986) scheme that simulates deposition freezing in original MMS with a mechanism developed by Ullrich et al. (2017), hereafter U17, to simulate deposition freezing for dust and soot, which has been implemented and evaluated by previous studies (e.g., Marinou et al., 2019;Y. Zeng et al., 2023).Modeled interstitial dust and element carbon (soot) in Aitken and accumulation modes are included in U17 to represent heterogeneous freezing.Because not all the soot particles are effective INPs (Zhu & Penner, 2020), we assume only 10% of soot act as INPs (Schill et al., 2020).Additionally, we include the heterogeneous freezing of whole pollen grains in cirrus using the pollen contact freezing formula described above (Equations 2 and 3; Matthews et al., 2023).The second revision in the cirrus formation processes modifies the homogeneous freezing to account for solutes within droplets based on water activity criterion (Koop et al., 2000;Thompson & Eidhammer, 2014), which depends on temperature and partial vapor pressure of water.The homogeneous nucleation rate is only a function of water activity, which in thermodynamic equilibrium equals the relative humidity over ice of the surrounding environment (Schneider et al., 2021b).We use interstitial aerosol in both the Aitken and accumulation modes as aqueous aerosol (including SPPs) for the Koop scheme.
Prior work suggests the competition between homogeneous and heterogenous freezing within cirrus clouds is important (Liu & Penner, 2005;Penner et al., 2018).In cirrus clouds, supercooled aqueous aerosol generally freeze first via homogeneous freezing, which can be the dominant mechanism of cirrus cloud ice formation and produce many small ice droplets.Heterogenous freezing requires lower relative humidity and could happen in advance of homogeneous freezing, forming a smaller number of large ice droplets and potentially prohibiting homogeneous freezing.Here, we simulate the competition between heterogeneous and homogeneous ice formation with the parameterization of Liu and Penner (2005).While the initial freezing temperature for different freezing mechanisms varies (e.g., the deposition freezing initial temperature for dust is 33°C and for soot is 38°C for U17, while the homogeneous freezing onset temperature for aqueous aerosol is 33°C for Koop et al. (2000)), we assume the heterogeneous and homogeneous ice formation competition starts at 35°C in the model following Liu and Penner (2005) and Zhu & Penner (2020).

Domain and Configuration
The WRF-Chem model simulation domain is centered on the US SGP region bounded by 33.4-39.6°Nand 93.5-101.5°W(Figure 1).The domain has 224 × 224 grid cells horizontally with 3 km spacing and 45 unevenly spaced vertical layers ranging from 1,000-50 mb, with a meteorological model timestep of 18 s and output saved every 2 hr.We select this domain to evaluate a high-impact convective event from 17-18 April 2013.Prior work (Subba et al., 2023) analyzed this event and found the model can represent the convective system and regional aerosol concentrations.Simulations of this event transported pollen to the upper troposphere (up to 12 km) and, assuming pollen can act as a CCN only, found that pollen has the potential to impact warm cloud formation (Subba et al., 2023).Initial and boundary meteorological condition are produced from the National Centers for Environment Prediction (NECP) North America Mesoscale Forecast System (NAM) data with 12 km resolution.The gas-phase and aerosol chemistry are simulated with the Regional Acid Deposition Model v2 chemical mechanism (RADM2, Stockwell et al., 1990) and aerosol chemistry with the MADE-SORGAM model (Ackermann et al., 1998;Schell et al., 2001).Anthropogenic emissions are derived from the 2011 US EPA National Emissions Inventory (NEI, 2011) and biogenic emissions are simulated by the Model of Emissions of Gases and Aerosols from Nature version MEGAN v2.1 biogenic emissions (Guenther et al., 2012).The chemical initial and boundary conditions are provided by Model for Ozone and Related chemical Tracers (MOZART) model simulations (Emmons et al., 2010) for major gas and aerosol species, and these do not include pollen or SPP.Other relevant model configurations include using the Rapid Radiation Transfer Model for Global climate model (RRTMG) for shortwave and longwave radiation (Iacono et al., 2008), the Noah land surface model (Chen & Dudhia, 2001), Monin-Obukhov for the surface layer (Janić, 2001;Monin & Obukhov, 1954), and the YSU scheme for the planetary boundary layer (PBL) (Hong et al., 2006).As described in Sections 2.2 and 2.3, the microphysical parameterization is based on Morrison et al. (2005) and because of the 3 km resolution, a convective parameterization is not activated.

Simulation Design
To test the effects of pollen and SPP on mixed-phase and cold cloud formation, we designed five sensitivity experiments (Table 1).Five ensemble members are conducted for each sensitivity experiment from 11-20 April 2013, with a 6-hr difference in the start time to alter initial conditions.All results presented below represent the multi-member average to reduce noise common in cloud-aerosol simulations.A simulation with the original MMS is also included to evaluate the changes to the ice nucleation parameterizations.Experiments include: 1. "Original": WRF-Chem simulation with original MMS (Morrison et al., 2005), where pollen and other aerosol are not included in the heterogeneous ice nucleation scheme.2. "Control": updates MMS as described in Section 2, where the heterogeneous ice nucleation includes all anthropogenic and natural aerosol except pollen and SPPs.3. "POL p ": adds primary pollen only (no pollen rupture) to the anthropogenic and natural aerosol in the D10 scheme to simulate immersion and contact freezing from pollen in mixed-phase clouds.Additionally, we also utilize the deposition freezing scheme for dust (U17) to simulate primary pollen heterogeneous freezing in cirrus clouds.4. "POL p+s ": includes primary pollen and pollen rupture to generate SPPs.Both pollen and SPPs are included in the D10 scheme for immersion and contact freezing in mixed-phase clouds.In cirrus clouds, SPPs can homogeneously freeze as cloud-borne aqueous aerosol (Koop et al., 2000), along with other aerosol in Aitken and accumulation modes. 5. "M23": Includes primary pollen and SPPs with contact and immersion freezing from pollen and immersion freezing for SPPs using the newly developed pollen INP parametrization (M23; Matthews et al., 2023) in mixed-phase clouds.Contact freezing of SPPs is simulated by the D10 scheme as lab-based data is not available.In cirrus clouds, the heterogeneous freezing of primary pollen is simulated using the lab-developed contact freezing parameterization (M23).6. "M23 + rup": Includes primary pollen and SPPs with the M23 scheme to simulate pollen and SPP ice nucleation.Additionally, this experiment increases pollen rupture rate from 1,000 SPPs grain 1 to 1.25 × 10 5 SPPs grain 1 based on Matthews et al. (2023).
Overall, when comparing the results of these sensitivity experiments, we find that the simulated cloud structure and convective system spatial distribution are very similar between experiment "POL p+s " (results not shown) and "M23."This indicates that the revised laboratory freezing parameterization does not change our results substantially.Therefore, we exclude the "POL p+s " experiments in the subsequent discussion and analyze the "M23" experiment in the following sections.To separate the effects of aerosol ice nucleation from the thermodynamic impacts of aerosol on cloud microphysics resulting from changes in solar radiation, pollen emissions are activated in the "Control" experiment to maintain consistent thermodynamic conditions with the "POL p " experiment, but pollen is not included in the heterogeneous ice nucleation scheme.

Observations From Atmospheric Radiation Measurement (ARM) SGP
We use observational data from the Improved Continuous Baseline Microphysics Retrieval (MICROBASE) Product with Uncertainties (MICROBASEKAPLUS; https://www.arm.gov/capabilities/science-data-products/vaps/microbasekaplus) at the Atmospheric Radiation Measurement (ARM) SGP site to evaluate the modeled vertical ice and liquid water profiles.MICROBASEKAPLUS is a value-added product (VAP) that provides continuous, high-time-resolution profiles of cloud microphysical properties such as the liquid water content (LWC), ice water content (IWC) and cloud effective radius (Dunn & Jensen, 2011).This product provides daily data with a time resolution of 4 s and vertical resolution of 30 m.The algorithms used in MICROBASEKAPLUS are identical to those used in MICROBASE, with many based on empirical relationships (Dunn & Jensen, 2011).
Additionally, MICROBASEKAPLUS adds uncertainties to the derived quantities using a perturbation method first applied through the Atmospheric System Research Quantifying Uncertainty in Cloud Retrievals science focus group (C.Zhao et al., 2014).The MICROBASEKAPLUS product combines three data products including Active Remote Sensing of Clouds (ARSCL) product using Ka-Band ARM Zenith Radars (KAZRARSCL), the Interpolated Sonde (INTERPSONDE) VAP, and the Microwave Radiometer Retrievals (MWRRET) VAP.ARSCL detects cloud characteristics including hydrometeors' vertical distribution and cloud boundary information, INTERPSONDE gives the information of environment profile that is used to define the phase of cloud water, and MWRRET retrieves cloud liquid water path (LWP) and column precipitable water (Dunn & Jensen, 2011).

Ice Nucleation Parameterization Evaluation
We compare observed and simulated vertical IWC and LWC from experiments "Original," "Control," and "M23 + rup" (Figure 2).The model output has a coarser time resolution (2 hr) compared to the observational product (4 s) due to computational limitations.In the initial model configuration ("Original"), it underestimates the IWC (with a maximum mean value of 0.11 g m 3 ) in the cold cloud region (>4 km above ground level [AGL]) compared to the observation (up to 0.18 g m 3 ) (Figure 2a).After updating the MMS with aerosol-aware ice microphysics in the "Control" experiment, the simulated IWC increases (by around 0.02 g m 3 ) and aligns more to the observation (Figure 2a).Both the "Original" and "Control" experiments simulate less ice in the cirrus cloud regime (>9 km) compared to the observations, indicating that neither of the ice nucleation schemes accurately represent the observed ice concentration within the cirrus clouds.This underestimation could be due to either weak ice nucleation of cirrus clouds or detrainment from the convective clouds in the simulation, but it is not possible to attribute the discrepancy to a specific process.After incorporating pollen and SPPs with a lab rupture rate (experiment "M23 + rup"), the IWC increases and improves agreement with observations.The simulated IWC, up to 0.2 g m 3 , is slightly higher than the observation in the mixed cloud region but lower than observation in the cirrus cloud region (around 10 km AGL).The standard deviation of simulated IWC by experiment "M23 + rup" also improves agreement with the observed results, especially in the mixed-phase cloud region.
Overall, these results indicate that compared to the original MMS, the updated Morrison scheme with the labgenerated SPP rupture rate better represents the observed IWC over the atmosphere (Figure 2a).
The simulated LWC from all experiments is lower than then observations in the lower atmosphere (<1.5 km AGL) but compares well at higher altitudes (Figure 2b).The very high LWC observed in the lower atmosphere is likely impacted by heavy precipitation, as the presence of heavy precipitation can lead to attenuation of the lidar signal and prevent the lidar from detecting reliable LWC (Ahlgrimm & Forbes, 2016).

Simulated Pollen and SPP
Atmospheric concentrations of primary pollen and SPP averaged over the PBL are shown for experiment "M23" (Figure 1), largely driven by greater pollen emissions in the eastern and southeastern part of the domain (Figure S1 in Supporting Information S1).During the convective event, instabilities are observed in the southeastern part of the domain that lead to a wave cloud structure at the front edge of the storm and generate minor vertical "striping" patterns in the simulated pollen concentrations.When only primary pollen is included in the model and no rupture is included (experiment "POL p "), pollen grains can reach the upper troposphere (∼12 km) with number concentration of 12 grains m 3 (Figure 3a) during deep convection.After allowing pollen rupture (experiment "M23"), most whole pollen grains are ruptured and converted to SPPs in the convective region (blue box; Figure 1) due to the high humidity during the convective event, leading to high simulated SPP concentrations throughout the atmosphere (Figure 3e) and reduced primary pollen present only in the lower atmosphere (Figure 3b).The generated SPPs are lofted to the upper troposphere, with concentrations up to 10 5 m 3 (Figure 3e).By increasing the rupture rate from 10 3 SPPs grain 1 to 1.25 × 10 5 SPPs grain 1 (experiment "M23 + rup"), the concentration of SPPs increases by two orders of magnitude, reaching up to 10 7 SPP m 3 in the upper troposphere (Figure 3f).We note that in the upper atmosphere, pollen and SPPs are transported to the northwestern part of the domain, leading to different spatial patterns compared to Figures 1a and 1b.In the upper troposphere with the highest pollen rupture rate, the combined number concentration of pollen and SPPs accounts for up to 2% of the total aerosol number concentration in the accumulation and coarse modes.Notably, the simulated magnitude of SPPs in the upper troposphere is much larger than that of pollen, suggesting that SPPs may potentially have larger impacts on the cold cloud formation processes.

Impacts of Pollen on the Convective System
To examine pollen and SPP impacts on cloud microphysics and precipitation, we compare simulation results from the three sensitivity experiments that include pollen ("POLp", "M23", and "M23 + rup") with the control experiment ("Control").Figure 4 illustrates the spatial differences in ice water path (IWP), LWP, and precipitation due to the presence of pollen (Figure 3).The convective region selected, denoted by the dashed blue boxes, exhibits the highest time-averaged IWP (Figure 4a), LWP (Figure 4e), and precipitation (Figure 4i) from the "Control" experiment.This convective region also demonstrates a high average concentration of pollen (Figure 3a) or SPPs if the pollen rupture is turned on (Figures 1b,3e,and 3f).As the convective system (a squall line triggered by the cold front) proceeds from the northwestern to the southeastern part of the domain, we define the eastern boundary as the leading front of convective system and the western boundary as the back of the convective system (Figure 4).Compared to the "Control" experiment (Figures 4a, 4e, and 4i), primary pollen grains alone have a minimal impact on the time-averaged IWP, LWP, and precipitation (experiment POL p ; Figures 4b, 4f, and 4j).The differences in IWP and LWP exhibit both increases and decreases across portions of the domain (Figures 4b and 4f), but they are diffuse and near zero.When pollen rupture is introduced and SPPs are added to the simulation (experiment "M23" and "M23 + rup"), IWP (Figures 4c and 4d) and LWP (Figures 4g and 4h) both increase.SPPs aloft in the "M23" simulation increases IWP up to 140% at the back of the convective system (western side of the box) (Figure 4c), because the additional SPPs provide extra ice nucleation material that might not otherwise be present, further perturbating the cloud processes.In general, spatial changes in hydrometeors do not correlate well with the pollen emissions, and hydrometeor perturbations are more diffuse due to aerosol circulation in the cloud system.Increasing the rupture rate and subsequent concentration of SPPs ("M23 + rup") further intensifies these differences, with IWP increases exceeding 200% at back of the convective system (Figure 4d).Similar to IWP, the increases in LWP reach up to 80% for "M23" and up to 140% for "M23 + rup" with the addition of SPPs, primarily located at the back of the convective system (Figures 4g and  4h).One feature in the time-averaged LWP are several narrow bands of enhanced LWP along the leading edge of the convective system in the "Control" experiment (Figure 4e) that are enhanced by SPPs (Figures 4g and 4h).These bands represent the front edge of the convective system at various time steps in the time average, influenced by the presence of anthropogenic aerosol emissions, which facilitate condensation of water vapor and increase cloud liquid water droplet formation.When the cold front advances from the northwest to the southeast, the tail of the convective system extends toward the east.Because of the model's 2 hr output time resolution, changes in the timing of the system advancing are evident in the sensitivity runs (Figures 4g and 4h), indicating that the presence of SPPs in the model simulation influence the progression of the storm.
When the northwest-southeast-oriented squall line reaches its mature stage (April 18 06:00 UTC), it forms a long and straight convective line with model simulated radar reflectively up to 55 dBZ and a strong surface cold pool (Figures 5a-5d).When comparing the experiment with whole pollen only ("POL p ") to the "Control" experiment, the spatial distribution and magnitude of the simulated radar reflectivity are similar.After adding SPPs, the radar reflectivity strengthens at the leading edge of the storm and the cold pool weakens (Figures 5c and 5d, black contours).Additionally, the radar reflectivity has a slightly larger spatial extent, with its location extended to the northwest.The vertical cross section perpendicular to the squall line reveals details about the vertical structure of the system (Figures 5e-5l).Before including pollen in the simulation (experiment "Control"), the convective system is simulated from 50 to 200 km horizontally along the cross section, with the maximum IWC reaching 3.6 g m 3 (Figure 5e).After adding pollen and SPPs (experiment "M23"), the convective system widens slightly and the region with high IWC (>3.6 g m 3 ) is extended both vertically and horizontally (Figure 5g).When increasing the pollen rupture rate (experiment "M23 + rup"), the system extends further westward 30 km, and the magnitude of the IWC slightly decreases (Figure 5h).Compared to the IWC, the vertical cross sections of LWC have smaller changes (Figures 5j-5l).Overall, this analysis suggests that pollen and SPPs can widen the spatial extent of the convective system, and these modifications slow the progression of the convective system across the domain, enhance precipitation at the back of the system, and increase the spatial extent of IWP (Figure 4).

Ice Phase Hydrometeors
Focusing on the region with highest time-averaged IWP during the convective event, the spatially averaged vertical changes in total ice number and mass concentration over the convective region (blue box; Figure 4) are depicted in Figures 6a-6h, where the total ice number and mass concentration are the sum of three ice phase hydrometeors (ice, snow, and graupel).Primary pollen grains slightly increase (<10%) both the total ice number and mass within the convective system (experiment "POL p ," Figure 6b).Adding SPP (experiment "M23") has a more substantial impact on the hydrometeors than primary pollen alone, increasing the total ice mass up to 20% at the beginning of the convective system (April 17 18:00-21:00 UTC) and 35% at the end (April 18 06:00-09:00 UTC) (Figure 6g), prolonging its duration.With a higher SPP rupture rate (experiment "M23 + rup"), these changes are amplified and total ice mass increases >50% at the end of the convective system (Figure 6h).The increases in total ice mass correspond with changes in vertical wind velocity (Figures 6k and 6l), suggesting that additional SPPs invigorate convective cloud updrafts through modifications to water-phase processes that condense water vapor into cloud droplets and produce more latent heat, and further strengthen the updraft velocity.The maximum differences in updraft velocity exceed 7 m s 1 in the convective region after adding SPPs, while spatial averaging over the entire convective region reduces the updraft velocities, for example, "M23" increases up to 0.08 m s 1 (Figure 6k) and "M23 + rup" increases 0.1 m s 1 (Figure 6l).
Graupel exhibits the largest increases of the ice-phase hydrometeors due to pollen and SPPs (Figure 7f).Strengthened updrafts enhance riming, deposition, and ice-rain collection processes (Figure S2 in Supporting Information S1), leading to an increase in graupel mass mixing ratio of 10%-20% for "M23" and 15%-30% for "M23 + rup" (Figures 7e and 7f).Graupel nucleation also increases slightly, but this process does not have a substantial effect.Ice mass mixing ratio decreases in the mixed-phase cloud region (above 40°C) by ∼10% (Figures 7a and 7b), and is partially attributed to increased cloud ice-rain collision processes that form snow and graupel (Figures 7d and 7f).Additionally, as a significant amount of cloud droplets collide with and are collected by rain and graupel (Figure S4a in Supporting Information S1), fewer cloud droplets are available to participate in immersion and contact ice nucleation in the mixed-phase cloud (Figures S5a and S5b in Supporting Information S1), reducing ice formation.This simultaneous decrease in cloud droplets and cloud ice have been observed in a previous study focused on dust impacts (Shi et al., 2022).In contrast, in the cirrus cloud region, the ice mass mixing ratio increases up to 5% for experiment M23% and 15% for experiment M23 + rup (Figure 7b), as deposition ice nucleation dominates the heterogeneous freezing and is not dependent on the cloud droplet number concentration (Ansmann et al., 2019;Cziczo et al., 2013).With additional SPPs, more ice is formed through homogeneous freezing of aqueous aerosol in ice cirrus cloud.Over most of the vertical profile, the snow mass mixing ratio exhibits relatively small changes compared to other ice-phase hydrometeors (Figures 7c and 7d).In the lower atmosphere, the amount of precipitating ice (including snow and graupel) increases (Figures 7d and 7f), further increasing precipitation (see Section 4.5).
The changes in total ice number concentration are relatively small compared to the changes in total ice mass concentration, with increases of ∼10% in the mixed-phase region (about 15 to 40°C) for "M23" (Figure 6c) and slightly larger increases in "M23 + rup" (up to 25%) at the end of the convective system (Figure 6d).In the cirrus region (around 60°C from 12.5 to 15 km), total ice number increases 15%-25% for "M23" and up to 50% for "M23 + rup" (Figures 6c and 6d).These cirrus enhancements in total ice number concentration are due to increased homogeneous and heterogeneous ice nucleation with the addition of SPPs.Additionally, the acceleration of updraft velocity from 2 to 12 km in "M23" and "M23 + rup" has the potential to transport more ice particles to the upper troposphere (Figures 6k and 6l), resulting in a higher total ice number concentration at the top of the convective system (around 12 km) (Figures 6c and 6d).Similar to changes in the mass mixing ratio (Figure 7), graupel number concentration has the largest changes among all ice-phased hydrometeors (around 35% for "M23 + rup"; Figures S3e-S3f in Supporting Information S1) due to the enhanced riming and ice-rain collection processes.The differences between ice mass mixing ratio (Figures 7a and 7b) and ice number concentration (Figures S3a and S3b in Supporting Information S1) changes suggest variations in ice particle sizes.With a decreased ice mass mixing ratio (Figures 7a and 7b) and increased ice number concentration (Figures S3a and S3b in Supporting Information S1), ice particles become smaller in the mixed-phase clouds when pollen is added to the atmosphere and this effect is further enhanced with the addition of SPP (e.g., M23 vs. M23 + rup).Within the cirrus cloud regime (temperature < 40°C), the ice mass mixing ratio slightly increases with the addition of SPPs (Figures 7a and 7b; experiment M23 and M23 + rup); however, the increase is smaller than the increase in ice number concentration (Figures S3a and S3b in Supporting Information S1), indicating a smaller ice particle size.This result aligns with conclusions from previous studies (Jiang et al., 2011;B. Zhao et al., 2019), where higher aerosol loading decreases the ice particle effective radius for a strong convective system, as homogeneous freezing competes with heterogeneous freezing and dominates the ice formation (B.Zhao et al., 2019).

Liquid Phase Hydrometeors
In the convective region (blue box; Figure 4), the liquid water mass does not exhibit substantial changes after including pollen in the ice nucleation ("POL p "; Figure 8f).However, after the inclusion of SPPs, liquid water mass increases >35% for "M23" (Figure 8g), and >50% for "M23 + rup" (Figure 8h).These changes are primarily driven by increases in rain drop mass mixing ratio (Figure 9), which increases from 10% to 80% for "M23" and 20%-130% for "M23 + rup" in the lower atmosphere (warmer than 15°C) (Figures 9c and 9d) and is associated with the increased collection and auto conversion of cloud droplets to rain drops (Figures S4a and S4b in Supporting Information S1).Although the processes that reduce the mass of rain drops (e.g., rain evaporation, riming, ice-rain collection, and snow-rain collection; Figure S4b in Supporting Information S1) are enhanced with more SPPs added in the simulation, the gain from cloud droplet conversion dominates and increases rain mass.Moreover, the addition of SPPs increases the condensation of cloud droplets, releasing more latent heat and increasing the updraft velocities.Therefore, cloud droplet collision and collection by rain, snow, and graupel are also increased.The loss in cloud droplets is offset by the increased cloud droplet condensation (Figure S4a in Supporting Information S1), resulting in small changes in cloud droplet mass mixing ratio, with a decrease of 4% and 10% in the lower atmosphere and an increase of 25% and 35% around 20°C for "M23" and "M23 + rup," respectively (Figure 9b).These findings regarding changes in cloud and rain mass mixing ratio are consistent with previous studies that only included pollen and SPP as CCN (e.g., Subba et al., 2023), which also found higher rain drop mass mixing ratio and lower cloud droplet mass mixing ratio in the lower atmosphere (with ∼10% decrease in cloud drop mass mixing ratio and 1% increase in rain drop mass mixing ratio).After adding pollen and SPPs as INPs here, the changes in rain drop mass mixing ratio increase compared to the prior study, up to 10% in "M23" (with the same rupture rate of SPPs as Subba et al., 2023) (Figure 9d), further increasing precipitation rates (Figure 10).Overall, our findings indicate that the addition of pollen and SPPs increases the deep convection in the system, enhance the collision and coalescence processes in the cloud, and increase rain drop number in the lower atmosphere compared to simulations considering the effects of pollen and SPPs effects as CCNs alone.
The total liquid water number concentration (rain drops and cloud droplets) decreases in the lower atmosphere (temperatures >-15°C) after including SPPs (Figures 8c and 8d), primarily driven by the reduction in cloud droplet number concentration (Figures S5a and S5b in Supporting Information S1).Since cloud droplet number concentration (∼1e 7 kg 1 ) is significantly higher than rain drop number concentration (∼1e 5 kg 1 ), it dominates the total liquid water number concentration.In "M23," cloud droplet number concentration decreases by ∼40% while rain drop number concentration increases by 40% (Figures S5b and S5d in Supporting Information S1).This is amplified when more SPPs are simulated ("M23 + rup"), with cloud droplet number concentrations decreasing by around 60% and rain drop number concentration increasing by 60% (Figures S5b and S5d in Supporting Information S1).This increase in rain drop number concentration, along with decrease in cloud droplet number concentration, is attributed to intensified deep convection, which facilitates the collection and autoconversion of cloud droplets to rain drops, where one raindrop efficiently collect multiple smaller cloud drops.The higher rain drop mass mixing ratio and number concentration in the lower atmosphere further increases the precipitation rate (Figure 10).

Precipitation
Changes in both cold and warm phase hydrometeors will influence the precipitation amount and timing (Figures 4i-4l).In the region of highest precipitation rates produced by the squall line (blue box; Figure 4), the inclusion of pollen and SPPs increases intense precipitation (Figure 10).Experiments "M23" and "M23 + rup" exhibit two time periods with relatively large precipitation changes, with increases of up to 0.4 mm hr 1 for "M23" and up to 0.9 mm hr 1 for "M23 + rup."The two peaks occur at the beginning of the storm (April 17 22:00 UTC to April 18 04:00 UTC) and in the second half of the storm (April 18 04:00 UTC to April 18 10:00 UTC), respectively (Figure 10b).These two time periods align with the highest increases in the total liquid mass mixing ratio in the lower atmosphere (Figures 8g and 8h).As a result, they extend the duration of heavy precipitation (rate >7 mm hr 1 ) by an additional 2 hr (Figure 10a).Furthermore, spatial changes in precipitation correspond to increases in the LWP, with greatest increases at the back of the convective system (Figures 4k and 4l), potentially extending the spatial range of the convective system.

Discussion and Conclusions
This study incorporates pollen emissions and their atmospheric processes in a coupled meteorology-chemistry model (WRF-Chem), utilizing pollen and SPP INP parametrizations and rupture rates derived from lab experiments (Matthews et al., 2023) to examine their impacts on cold cloud formation and precipitation.We incorporate aerosol-aware INP parameterizations for all aerosol types, including the newly added pollen and SPP as well as other anthropogenic and biogenic aerosols.We evaluate the model simulations with radar products for a convective event, specifically a squall line triggered by a cold front, and find that the addition of aerosol-aware INP parameterizations that include pollen and SPPs better represents the observed convective system timing, duration, and vertical structure.To compare the relative importance of pollen and SPPs effects on the convective event, we conduct several sensitivity experiments.Our findings demonstrate that the addition of PBAP such as pollen and SPPs can influence the vertical structure and timing of the convective system.SPPs, compared to whole pollen grains, have a greater impact on cloud microphysics due to their higher concentrations and smaller sizes, which enable them to be easily transported to the upper troposphere and influence mixed-phase and cold cloud processes.
In the model simulations, the inclusion of SPPs in cloud microphysics invigorates the convective system by triggering more cloud droplet formation that releases latent heat, and therefore enhances the system updraft velocity.While the concept of convective invigoration by aerosols has been recently debated (Fan & Khain, 2021;Fan et al., 2018;Grabowski & Morrison, 2020), our analysis indicates that these processes are occurring in the model.These changes intensify atmospheric processes such as ice-rain collection, riming, and the collection of cloud droplets by rain and graupel.One rain drop (the collector drop) can collect multiple cloud droplets, therefore the total liquid water number concentration (cloud droplet + rain drop) decreases.The reduction in cloud droplet number constrains the immersion and contact ice nucleation processes and limits ice formation in the mixed phase cloud region.Consequently, the simulated ice number concentration slightly increases (∼10%) and ice mass mixing ratio decreases in the mixed phase cloud region when considering the effects of SPPs, leading to a smaller ice particle radius.In the cirrus cloud region, increases in ice number concentration are larger, reaching up to 50% with a high SPPs rupture rate.This is because the main heterogeneous freezing mechanism in cirrus clouds (deposition freezing) is unconstrained by cloud droplets (Ansmann et al., 2019;Cziczo et al., 2013).Compared to other ice phase hydrometeors, graupel exhibits the largest changes throughout the entire atmosphere with the addition of SPPs, due to the increased riming processes.Focusing on the region with the highest time-averaged precipitation and ice and liquid water paths during the convective event, the presence of SPPs leads to an increase in the vertical ice and liquid water mixing ratio up to 35%, particularly at the end stage of the convective system.When SPPs rupture rates from lab experiments (i.e., Matthews et al., 2023) are included, these changes are further amplified, with increases in ice and liquid mass mixing ratio exceeding 50%, enhancing precipitation rates.The peak precipitation increases up to 0.9 mm hr 1 in the convective region at the end of the storm.These changes have the potential to broaden the spatial extent and increase the duration of the convective system.
There are several limitations in the model simulations presented here.First, we only include one type of PBAP, pollen, in our model simulations as PBAP emission model development for other types (e.g., fungal spores, bacteria, plant debris) is still in a nascent phase (Janssen et al., 2021;Thiel et al., 2020;Vélez-Pereira et al., 2022).Therefore, the impact of PBAPs on cloud formation may be underestimated in models.Furthermore, the pollen tracer included in the model represents the total emission of 13 pollen species rather than individual species, and for different freezing mechanisms, we utilize a single INP parameterization for all pollen.This configuration does not account for the varying ice nucleating ability of different pollen types as shown in Matthews et al. (2023).However, incorporating different pollen types in the model is computationally expensive, and there is limited data on the ice nucleating ability of pollen and SPP for different species.Second, the inclusion of secondary ice formation in the model is incomplete (Morrison et al., 2005).Besides primary ice nucleation, secondary ice formation processes, which generate ice from pre-existing ice, can further increase the cloud ice number concentration and affect the cloud structure and lifetime (Field et al., 2017;Korolev & Leisner, 2020;Luke et al., 2021;X. Zhao & Liu, 2022).In the simulations presented, the only secondary ice formation process included is rime-splintering (Hallett & Mossop, 1974), while other processes such as collision fragmentation, droplet shattering, and sublimation fragmentation (Field et al., 2017) can also occur and play a significant role in cloud ice and snow formation (Patade et al., 2022).The addition of pollen and SPPs can potentially trigger more secondary ice formation by increasing primary ice nucleation as well as enhancing riming process and graupel concentration.The lack of inclusion for these processes may underestimate the cloud ice concentration and the overall effects of pollen and SPPs.Third, it is important to note that the CCN parameterizations within MMS are likely not comprehensive.For example, processes such as condensational conversion may not be accurately included which would make aerosol activation processes overly sensitive to CCN (X.Zeng & Li, 2022).This would affect the ability of pollen and SPP to activate as CCN, and will also influence microphysical processes that lead to the increase in vertical velocities.Lastly, our analysis focuses on one case study of a convective event in a specific region of the Unites States.These conclusions may differ when investigating convective systems with different characteristics in different regions that have varying land cover and pollen emissions.
Previous modeling studies investigating the impacts of pollen on cold clouds as INPs have suggested that pollen has minimal influences on ice microphysics compared to other aerosols or ice freezing mechanisms, such as secondary ice production (Diehl & Mitra, 2015;Hoose, Kristjánsson, & Burrows, 2010;Patade et al., 2022).However, when we simulate pollen rupture processes and include the resulting SPPs, the impacts are substantial, leading to an increase in the ice/liquid water mass mixing ratio up to 35% and over 50% when using a higher rupture rate (1.25 × 10 5 SPPs grain 1 ).Our results demonstrate that SPPs have a greater impact than pollen grains on the convective system, and not accounting for SPPs effects in previous studies may have underestimated the overall impacts of pollen (Diehl & Mitra, 2015;Hoose, Kristjánsson, & Burrows, 2010;Patade et al., 2022).One recent study included SPP impacts on ice formation processes (Werchner et al., 2022), concluding that SPPs can have a large effect on ice formation when considered as highly efficient ice nuclei.This study utilized INP parametrizations for general biological aerosol particles (Phillips et al., 2013), which cannot capture the variations in ice nucleating abilities among different types of biological aerosols.Here we employ INP parameterizations specifically developed for pollen and SPPs based on laboratory experiments (Matthews et al., 2023), where the measured ice nucleation efficiency for pollen and SPPs is higher than that used in Werchner et al. (2022).However, when Werchner et al. (2022) increased the ice nucleation efficiency by factors of 1,000 times to 100,000 times in their ice nucleating parametrizations, their results are comparable to those presented here.Using a pollen rupture rate of 8,260 SPPs grain 1 based on turgor pressure calculations, they simulated an increase in precipitation up to 6.2%, decreases in cloud droplet number concentration of up to 50%, and increases in ice number concentration in the lower atmosphere (∼2 km at 5°C) by 25%.In our sensitivity experiment using a pollen rupture rate of 1,000 SPPs grain 1 , we found an increase in precipitation rate of up to approximately 10%, a decrease in cloud droplet number concentration of up to 40%, and an increase in ice number concentration of around 20% around 4 km at 5°C.These changes are further intensified when we include lab-measured pollen rupture rate of 1.25 × 10 5 SPPs grain 1 , with precipitation increase up to 15%, and the simulated increase in cloud droplet concentration up to 60%.Finally, we note that Werchner et al. (2022) averaged over a relatively large spatial domain (approximately 40°latitude) and long time period (10 days), while our study focuses on a convective core (spanning ∼4°latitude) and period (1 day), and this may contribute to the larger changes observed in our simulations.
The importance of pollen and other primary biological aerosols in ice nucleation processes in the atmosphere has been uncertain given the lack of observational data and modeling studies.In this work, we have developed new model methods to assess the impacts of these processes by combining realistic pollen emission modeling and adding new pollen INP parameterizations.Our findings demonstrate that SPPs can have a larger effect on cloud microphysics compared to whole pollen grains, influencing the spatial extent and vertical structure of the convective system.This work emphasizes the importance of better understanding of biological aerosol emission, rupture, and ice nucleation process in the atmosphere, as their inclusion in the model can improve the simulation accuracy of the convective system timing and characteristics.Additionally, our study provides a novel tool to better understand the climate effects of biological aerosols and compare them with other aerosol types such as dust.

Figure 1 .
Figure 1.Weather Research and Forecasting Model with Chemistry simulation domain with the time averaged (April 17 12:00 to April 18 12:00, 2013) spatial concentration of (a) pollen (grains/m 3 ) and (b) sub-pollen particles (SPPs) (grains/m 3 ) within the planetary boundary layer (PBL).Blue dashed square bounds the area of interests in this study, which also has the highest ice/liquid water path and precipitation (see Figure4).The simulated pollen and SPPs concentration are from experiment M23.

Figure 2 .
Figure 2. Model evaluation for (a) simulated ice water content (IWC) and (b) simulated liquid water content using the MICROBASEKAPLUS observation product over the Atmospheric Radiation Measurement (ARM) Southern Great Plains (SGP) site.Lines represent the averaged observation/model simulation results from April 17 12:00 UTC, to April 18 12:00 UTC, 2013, with shading representing the standard deviation.We cap the maximum value of observed IWC from ARM SGP site as 3 g/m 3 .

Figure 3 .
Figure 3. Timeseries of area averaged (blue dashed box in Figure 1) (a-c) pollen (grains/m 3 ) and (b-f) sub-pollen particles (SPPs) (#/m 3 ) vertical concentration from April 17 12:00 UTC to April 18 12:00 UTC, 2013.The pollen and SPPs concentration are simulated by different experiments including (a, d) experiment POL p , (b, e) experiment M23, and (c, f) experiment M23 + rup.Experiment Control and experiment POL p+s are not included here, since the simulation results from experiment Control and POL p+s are similar with experiment POL p and M23, respectively.Contour lines represent temperature (K).The results showing here are five-ensemble member average.

Figure 4 .
Figure 4. Time averaged (April 17 12:00 UTC to April 18 12:00 UTC, 2013) spatial distribution of (a) ice water path (IWP) (kg/m 2 ) (e) liquid water path (LWP) (kg/m 2 ), and (i) precipitation (mm/hr) simulated by the experiment Control.The second, third, and fourth column show the difference between experiment Control and experiment POLp, M23, and M23 + rup, respectively (Δ experiment = experiment-Control), displayed as a five-ensemble member average.Blue dashed square bounds the area of interest centered over the Southern Great Plains with high values of IWP, LWP and precipitation.

Figure 5 .
Figure 5. (a-d) Spatial distribution of the simulated radar reflectivity at 500 m at one timestep (April 18 06:00 UTC, 2013) above the ground level for different experiments, where the radar reflectively is simulated based on precipitating particles and water vapor.Black contours show the temperature change between two time steps (06:00 UTC to 04:00 UTC), indicating the surface cold pool.Blue dashed squares show the averaged vertical cross section for (e-l).The box location is perpendicular to the convective system and crosses through the strongest convection and cold pool of the front.(e-l) Show the averaged vertical cross section of ice water content (e-h) and liquid water content (i-l) for four sensitivity experiments, where the first, second, third, and fourth column show the results from experiment Control, POL p , M23, and M23 + rup, respectively.The x-axis represents the distance alongside the cross sections in (a)-(d) from west to east.All results display a fiveensemble member average.

Figure 7 .
Figure 7. Time (April 17 12:00 UTC to April 18 12:00 UTC, 2013) and spatial (blue dashed box in Figure 4) average vertical distribution of ice phase hydrometeors' mass mixing ratio (g kg 1 ).Left column: (a) ice, (c) snow, and (e) graupel mass mixing ratio simulated by different experiments.Right column: percentage changes of different experiments compared to experiment Control for (b) ice, (d) snow, and (f) graupel mass mixing ratio.All results display the five-ensemble member average.

Figure 8 .
Figure 8. Same with Figure 6 but for (a-d) liquid water number concentration (/g) and (e-h) liquid water content (g/m 3 ) concentration (a, e; simulated by experiment Control) as well as the difference between experiment "Control" and experiment POL p (b, f), M23 (c, g), and M23 + rup (d, h), respectively (Δ experiment = experiment-Control).The results showing here are five-ensemble member average.

Figure 10 .
Figure 10.Spatial (center of the convective system; blue dashed box in Figure 2) averaged (a) precipitation rate (unit: mm/hr) for four experiments as well as the (b) differences between each experiment with experiment Control from April 17 12:00 UTC to April 18 12:00 UTC, 2013.The results showing here are five-ensemble member average.