Observations of orographic mixed-phase clouds (MPCs) have long shown that measured ice crystal number concentrations (ICNCs) can exceed the concentration of ice nucleating particles by orders of magnitude. Additionally, model simulations of alpine clouds are frequently found to underestimate the amount of ice compared with observations. Surface-based blowing snow, hoar frost, and secondary ice production processes have been suggested as potential causes, but their relative importance and persistence remains highly uncertain. Here we study ice production mechanisms in wintertime orographic MPCs observed during the Cloud and Aerosol Characterization Experiment (CLACE) 2014 campaign at the Jungfraujoch site in the Swiss Alps with the Weather Research and Forecasting model (WRF). Simulations suggest that droplet shattering is not a significant source of ice crystals at this specific location, but breakups upon collisions between ice particles are quite active, elevating the predicted ICNCs by up to 3 orders of magnitude, which is consistent with observations. The initiation of the ice–ice collisional breakup mechanism is primarily associated with the occurrence of seeder–feeder events from higher precipitating cloud layers. The enhanced aggregation of snowflakes is found to drive secondary ice formation in the simulated clouds, the role of which is strengthened when the large hydrometeors interact with the primary ice crystals formed in the feeder cloud. Including a constant source of cloud ice crystals from blowing snow, through the action of the breakup mechanism, can episodically enhance ICNCs. Increases in secondary ice fragment generation can be counterbalanced by enhanced orographic precipitation, which seems to prevent explosive multiplication and cloud dissipation. These findings highlight the importance of secondary ice and seeding mechanisms – primarily falling ice from above and, to a lesser degree, blowing ice from the surface – which frequently enhance primary ice and determine the phase state and properties of MPCs.
Understanding orographic precipitation is one of the most critical aspects of weather forecasting in mountainous regions (Roe, 2005; Rotunno and Houze, 2007; Chow et al., 2013). Orographic clouds are often mixed-phase clouds (MPCs) containing simultaneously supercooled liquid water droplets and ice crystals (Lloyd et al., 2015; Lohmann et al., 2016; Henneberg et al., 2017). MPCs are persistent in complex mountainous terrain because the high updraft velocity conditions generate supercooled liquid droplets faster than can be depleted by ice production mechanisms (Korolev and Isaac, 2003; Lohmann et al., 2016). In mid- and high-latitude environments, almost all precipitation originates from the ice phase (Field and Heymsfield, 2015; Mülmenstädt et al., 2015), emphasizing the necessity of correctly simulating the amount and distribution of both liquid water and ice (i.e., the liquid-ice-phase partitioning) in MPCs (Korolev et al., 2017).
Our understanding of MPCs remains incomplete, owing to the numerous and highly nonlinear cloud microphysical pathways driving their properties and evolution (Morrison et al., 2012). MPCs tend to glaciate over time through the Wegener–Bergeron–Findeisen (WBF) process, which is the rapid ice crystal growth at the expense of the surrounding evaporating cloud droplets (Bergeron, 1935; Findeisen, 1938). Ice crystals falling from a high-level seeder cloud into a lower-level cloud (external seeder–feeder event) or a lower-lying part of the same cloud (in-cloud seeder–feeder event) can trigger cloud glaciation and enhance precipitation over mountains (e.g., Roe, 2005; Reinking et al., 2000; Purdy et al., 2005; Mott et al., 2014; Ramelli et al., 2021). Analysis of satellite remote sensing over the 11-year period, between April 2006 and October 2017, suggests that seeding events are widespread over Switzerland, occurring with a frequency of 31 % of the total observations in which cirrus clouds seed lower mixed-phase cloud layers (Proske et al., 2021).
Primary ice formation in MPCs is catalyzed by the action of ice nucleating particles (INPs; e.g., Hoose and Möhler, 2012; Kanji et al., 2017). However, in situ observations of MPCs in orographic environments regularly reveal that measured ice crystal number concentrations (ICNCs) are several orders of magnitude more abundant than INPs (Rogers and Vali, 1987; Geerts et al., 2015; Lloyd et al., 2015; Beck et al., 2018; Lowenthal et al., 2019; Mignani et al., 2019). Model simulations of alpine MPCs frequently fail to reproduce the elevated ICNCs dictated by observations (Farrington et al., 2016; Henneberg et al., 2017; Dedekind et al., 2021).
The inability of primary ice to reproduce the observed ICNCs in orographic MPCs has often been attributed to the influence of surface processes, including lofting of snowflakes (i.e., blowing snow; Rogers and Vali, 1987; Geerts et al., 2015), detachment of surface hoar frost (Lloyd et al., 2015), turbulence near the mountain surface or convergence of ice particles due to orographic lifting (Beck et al., 2018), and riming on snow-covered surfaces (Rogers and Vali, 1987). The impact of blowing snow ice particles (BIPs) has been studied thoroughly, either using observations collected in mountainous regions (e.g., Lloyd et al., 2015; Beck et al., 2018; Lowenthal et al., 2019), remote sensing (e.g., Rogers and Vali, 1987; Vali et al., 2012; Geerts et al., 2015), or detailed snow-cover models (e.g., Lehning et al., 2006; Krinner et al., 2018) coupled with atmospheric models (e.g., Vionnet et al., 2014; Sharma et al., 2021). The extent to which BIPs can affect ICNCs in MPCs remains poorly understood.
In-cloud secondary ice production (SIP) – or ice multiplication –
processes may also enhance ice production above the concentration of INPs
(Field et al., 2017; Korolev and Leisner, 2020). A total of three mechanisms are thought to be responsible for most of the SIP. The first, known as the Hallett–Mossop (HM) process (Hallett and Mossop, 1974), refers to the ejection of small secondary ice splinters after a supercooled droplet with a diameter larger than
Collisional fracturing and breakup (BR) of delicate ice particles with other ice particles (Vardiman, 1978; Griggs and Choularton, 1986; Takahashi et al., 1995) is another important SIP mechanism. Several field studies in the Arctic (Rangno and Hobbs, 2001; Schwarzenboeck et al., 2009), the Alps (Mignani et al., 2019; Ramelli et al., 2021), and laboratory investigations (Vardiman 1978; Takahashi et al. 1995) all show the importance of BR. The latter two studies created the basis for a mechanistic description of BR (e.g., Phillips et al., 2017a; Sullivan et al., 2018a; Sotiropoulou et al., 2020). Parameterizations of BR have recently been implemented in small-scale (Fridlind et al., 2007; Phillips et al., 2017a, b; Sotiropoulou et al., 2020, 2021b; Sullivan et al., 2018a; Yano and Phillips, 2011; Yano et al., 2016), mesoscale (Hoarau et al., 2018; Sullivan et al., 2018b; Qu et al., 2020; Sotiropoulou et al., 2021a; Dedekind et al., 2021), and global climate models (Zhao and Liu, 2021a), each with their own approach towards BR description.
Droplet freezing and shattering (DS) is a third SIP mechanism that can
produce significant amounts of ice crystals. It occurs when drizzle-sized
drops (diameter exceeding 50
Orographic ICNCs in MPCs exceeded the predicted INPs by 3 orders of magnitude, reaching up to
Although surface-originated processes have been frequently invoked to explain the disparity between ICNCs and INPs, the role of SIP processes – especially the BR and the DS mechanism – has received far less attention and is addressed in this study. We utilize the Weather Research and Forecasting model (WRF) to conduct simulations of two case studies observed in winter during the CLACE 2014 campaign. Our primary objective is to investigate if the implementation of two SIP parameterizations that account for the effect of BR and DS can reduce the discrepancies between observed and simulated ICNCs. Additionally, we aim to identify the conditions favoring the initiation of SIP in the orographic terrain and explore the synergistic influence of SIP with windblown ice.
CLACE is a long-established series of campaigns taking place for over 2 decades at the mountain-top station of JFJ, located in the Bernese Alps, in
Switzerland, at an altitude of
Map of synoptic conditions around JFJ station at
Shadowgraphs of cloud particles were produced by the two-dimensional stereo
hydrometeor spectrometer (2D-S; Lawson et al., 2006), part of a three-view cloud particle imager (3V-CPI) instrument. The 2D-S products have been used to provide information on the number concentration and size distribution of particles in the size range of 10–1280
WRF version 4.0.1, with augmented cloud microphysics to include the effects
of additional SIP mechanisms (Sotiropoulou et al., 2021a) is used for non-hydrostatic cloud-resolving simulations. The model has been run with
three two-way nested domains (Fig. 1), with a respective horizontal resolution of 12, 3, and 1 km. A two-way grid nesting is generally found to
improve the model performance in the inner domain (e.g., Harris and Durran, 2010), although the sensitivity of the results to the applied nesting technique has been shown to be negligible (not shown). The parent domain consists of 148
The fifth generation of the European Centre for Medium-Range Weather
Forecasts (ECMWF) atmospheric reanalyses dataset (ERA5; Hersbach et al., 2020) is used to initialize the model and provide the lateral forcing at the edge of the 12 km resolution domain every 6 h. Static fields at each model grid point come from default WRF pre-processing system datasets, with a
resolution of 30
The Morrison two-moment scheme (Morrison et al., 2005; hereafter M05) is
used to parameterize the cloud microphysics, following the alpine cloud
study of Farrington et al. (2016). The scheme includes double-moment
representations of rain, cloud ice, snow, and graupel species, while cloud
droplets are treated with a single-moment approach, and therefore, the cloud
droplet number concentration (
In total, three primary ice production mechanisms through heterogeneous nucleation are described in the default version of the M05 scheme, namely immersion freezing, contact freezing, and deposition/condensation freezing nucleation. Immersion freezing of cloud droplets and raindrops is described by the probabilistic approach of Bigg (1953). Contact freezing is parameterized following Meyers et al. (1992). Finally, deposition and condensation freezing is represented by the temperature-dependent equation derived by Rasmussen et al. (2002), based on the in situ measurements of Cooper (1986) collected from different locations at different temperatures. Following Thompson et al. (2004), this parameterization is activated either when there is saturation with respect to liquid water and the simulated temperatures are below
Apart from primary ice production, the HM process is the only SIP mechanism
included in the default version of the M05 scheme. This parameterization
adapted from Reisner et al. (1998), based on the laboratory findings of Hallett and Mossop (1974), allows for splinter production after cloud- or raindrops are collected by rimed snow particles or graupel. The efficiency of this process is zero outside the temperature range between
In addition to the HM process, we have also included two parameterizations
to represent the BR mechanism. An extensive description of the implementation method is provided in Sotiropoulou et al. (2021a; see their Appendix B). Among the three ice particle types included in the M05 scheme (i.e., cloud ice, snow, and graupel), we assume that only the collisions between cloud ice–snow, cloud ice–graupel, graupel–snow, snow–snow, and graupel–graupel can result in ice multiplication. The first parameterization
tested here follows the simplified methodology proposed by Sullivan et al. (2018a), which is based on the laboratory work of Takahashi et al. (1995). Their findings revealed a strong temperature dependence of the fragment numbers generated per collision (
Phillips et al. (2017a) proposed a more physically based formulation, developing an energy-based interpretation of the experimental results conducted by Vardiman (1978) and Takahashi et al. (1995). The initial collisional kinetic energy is considered as being the governing constraint driving the BR process. Moreover, the predicted
In the M05 scheme, two different parameterizations are implemented to investigate the potential efficiency of the DS mechanism in producing secondary ice splinters (
The second mode of raindrop–ice collisions includes the accretion of raindrops on impact with more massive ice particles, such as snow or graupel, the description of which in the M05 scheme is adapted from Ikawa and Saito (1991). While there is only one dedicated laboratory study of this SIP mode (James et al., 2021), it was also indirectly investigated in the experimental study of Latham and Warwicker (1980), who reported that the collision of supercooled raindrops with hailstones can generate secondary ice. Phillips et al. (2018) proposed an empirical, energy-based formulation to account for the tiny splinters ejected after collisions between raindrops and large ice particles as follows:
The second DS parameterization tested in this study was developed by Sullivan et al. (2018a) and is a function of the freezing droplet diameter (
The diameter dependence describing the fragment numbers generated per fractured frozen droplet is derived by nudging the liquid water and ice particle size distributions in one-dimensional cloud model simulations towards aircraft observations collected in tropical cumulus clouds (Lawson et al., 2015). The
The control simulation (CNTRL), performed with the standard M05 scheme, sets
the basis for assessing the validity of the model against available
meteorological observations. Temperature, relative humidity, wind speed, and
wind direction are obtained from the MeteoSwiss weather station at JFJ. The
comparison of each meteorological variable with the results from the nearest
model grid point of the CNTRL simulation is shown in Fig. 2. Note that the
outputs are from the first atmospheric level of the innermost domain at
Time series of
Throughout the two case studies, the WRF simulations seem to closely follow
the observed temperatures (Fig. 2a), which is also indicated by the high IoA
in Table 1. The synoptic situation occurring on 26 January, with a deep
trough extending to western Europe (Fig. 1), has been associated with
intense snowfall in the alpine regions (Panziera and Hoskins, 2008). The passage of the cold front was followed by a sharp temperature decrease, with the simulated temperatures fluctuating between
Mean modeled values (
The 1 km resolution domain can sufficiently capture the local wind systems to a certain extent (Fig. 2c, d). During the NW flow, the
horizontal wind speeds are reproduced better by the CNTRL simulation
(IoA
Given the good representation of the atmospheric conditions at JFJ, the CNTRL simulation of WRF is further accompanied by five sensitivity simulations aiming to investigate the contribution of BR and DS mechanisms. Here we also perform three additional sensitivity experiments to explore the potential impact of blowing ice and the synergistic interaction with SIP on the development of the simulated MPCs. A detailed list of the sensitivity experiments is provided in Table 2.
The contribution of the DS mechanism is addressed in two sensitivity experiments, DS_PHILL and DS_SULL, where the parameterizations of Phillips et al. (2018) and Sullivan et al. (2018a) were applied, respectively (Sect. 2.2.4). Both sensitivity simulations yield predictions that coincide with the CNTRL simulation (see Fig. S1 in the Supplement), suggesting that the DS mechanism is hardly ever activated, and fail to produce realistic total ice number concentrations (
The effect of secondarily formed ice particles through BR is then examined in the following three sensitivity simulations. First, the TAKAH simulation adopts the temperature-dependent formula of Takahashi et al. (1995) scaled with the size of particles that undergo fragmentation (Sotiropoulou et al. 2021a). Applying Eq. (2) to collisions between all ice categories considered in the M05 scheme (except collisions between cloud ice particles; Sect. 2.2.3) inserts a caveat to our approach. The laboratory results of Takahashi et al. (1995) suggest that it is mostly the collisions between rimed particles and graupel that are more conducive to SIP through BR. Vardiman (1978) also reported that ice crystal growth through riming is essential to boost fragmentation. Applying the Takahashi breakup scheme for unrimed ice particles might, therefore, overestimate the number of secondary ice fragments. To test this hypothesis, we performed the TAKAHrim sensitivity simulation, where we enabled ice multiplication through BR only after collisions between rimed cloud ice/snow and graupel particles. To diagnose the presence of rime on ice particles, we used the amount of cloud droplets or raindrops accreted by snow and cloud ice, which is predicted in the M05 scheme.
List of sensitivity simulations conducted with WRF. Note: NBIPs is the number concentrations of BIPs.
Finally, the performance of the more advanced Phillips et al. (2017a)
parameterization is tested in the PHILL simulation. Parameters involved in
the Phillips parameterization that are not explicitly resolved in the M05
microphysics scheme are the rimed fraction and the ice habit of colliding
ice particles. The choice of ice habit is based on particle images collected
during the CLACE 2014 campaign, showing the presence of non-dendritic
sectored plates and oblate particles at temperatures
The remaining sensitivity simulations focus on the potential impact of BIPs.
A recently developed blowing snow scheme, used to simulate alpine snowpacks,
reported significant mass and number mixing ratios of BIPs that can be found
up to
The applied concentrations of BIPs varied between 10
As SIP through BR and blowing snow are both important when trying to explain
the high ICNCs observed in alpine environments, their combined effect is
addressed in our last simulation, BIPs100_PHILL (Table 2). In this sensitivity simulation, the effect of BR is parameterized after Phillips
et al. (2017a), while a constant ice crystal concentration of 100 L
The temporal evolution of
Time series of
During the NW flow, between 26 and 28 January, the measured ICNCs exceed 100 L
The 25th, 50th (median), and 75th percentiles of ICNC (per liter) time series.
Activating the BR process in TAKAH, TAKAHrim, and PHILL simulations is found
to produce
It is worth noting that, despite the fact that the Takahashi parameterization (Eq. 2) is applied to both TAKAH and TAKAHrim simulations, the former seems to systematically overestimate the number of secondary ice fragments, while the latter produces ICNCs that are more consistent with the observations. Hence, the Takahashi parameterization predicts reasonable results if it is allowed to generate fragments from collisions between rimed ice particles only (Sect. 2.4).
The 25th, 50th (median), and 75th percentiles of IWC
(in grams per cubic meter; hereafter gm
The observed IWC time series (Fig. 3b) are frequently reaching
The comparison of the simulated cloud LWC with the concurrent CDP observations at JFJ is presented in Fig. 3c. The LWC values recorded during the NW case are highly variant, reaching up to 0.75 gm
The modeled LWC in the 75th percentile is decreased by a factor of 1.5–5 in the simulations that account for the BR process (Table 5), improving the agreement with observations (Fig. 3c). The reduction in LWC is expected, considering that the higher
The 25th, 50th (median), and 75th percentiles of LWC
(in gm
The time-averaged vertical profiles of cloud ice (
Mean vertical profiles of
Graupel number concentrations (Fig. 4b) do not contribute much to the modeled ice phase, especially during the SE case when the simulated
The mean vertical profile of
The temporal evolution of the vertical profiles of
Time–height plots of total N
To illustrate the processes taking place during the two cases of interest,
Fig. 6 displays the tendency of primary and secondary ice production, as well
as the growth of ice particles through deposition, riming, and aggregation
from the CNTRL and PHILL simulations at 17:00 (19:00) UTC on 26 (30) January. The vertical profiles on 26 January are taken within the seeder–feeder event, while those on 30 January are taken when the high-level
cloud associated with the warm front has already passed the region of
interest. Upon arrival of the frontal system on 26 January, CNTRL indicates
a rapid increase in the total
Vertical profiles of
Focusing on the ice-seeding event of 26 January, the enhanced aggregation
rate observed at heights above
Activating the BR mechanism along with the seeding of precipitating hydrometeors in PHILL shifts the simulated
The key difference between CNTRL and PHILL simulations is that the latter
takes advantage of the enhanced ice particle growth through aggregation,
while falling to the feeder cloud below
A classification of the dominant type of precipitation was applied to the
polarimetric data collected by a weather radar deployed at the Kleine
Scheidegg station (2061 m a.s.l) during the SE case between 30 and 31 January (Fig. S6). In the derived time series, we can identify
periods when individual ice crystals (not aggregated and not significantly
rimed) dominate over the entire precipitation column, followed by periods
when a clear stratification is present with ice crystals aloft and mostly
aggregates and rimed ice particles below. This stratification is observed on
30 January at 19:00 UTC when the model tendencies are extracted (dashed
lines in Fig. 6). Allowing for the BR process in PHILL results in an enhancement of 2 orders of magnitude in the aggregation rates close to the ground, which can better reproduce the signatures observed in the hydrometeor
classification at that time. An increase in the simulated aggregates and
rimed particles is expected to increase orographic precipitation, which is
important given that these low-level feeder clouds are incapable of
producing significant amounts of precipitation. Indeed, the mean surface
precipitation produced by PHILL is increased by 30 % (10 %) during the NW (SE) case compared with CNTRL (Fig. S7), which is in contrast to
Dedekind et al. (2021), where the activation of the BR process is found to
suppress the regions of strong surface precipitation. This was attributed to
the limited efficiency of the small secondary ice particles to grow sufficiently to precipitation sizes when the local updrafts lift them to the
upper parts of the cloud that were glaciated. The radar-based hydrometeor
classification reveals also the predominance of ice crystals at the
beginning and the end of the precipitating periods (e.g., on 30 January at
15:00–17:30 UTC or 31 January at 04:30–06:00 UTC), which is again more consistent with the vertical profile of
In this section, we examine if the surface-originating small ice particles
could have the potential to initiate and enhance ice particle growth in the
near-surface MPCs present in our case studies. Figure 7 illustrates two
additional WRF simulations – BIPs10 and BIPs100 – where the ice crystal
source applied to the first model level is equal to 10 and 100 L
Time series of
As indicated in Fig. 7b, during the NW flow the simulated LWC at the first
model level in BIPs10 and BIPs100 almost coincides with the CNTRL simulation
of WRF. The three sensitivity simulations are producing comparable median
and quartile LWC values (Table 5), with BIPs10 and BIPs100 producing median
LWC values closer to the observed ones during the SE flow. When comparing
against the LWC values in the third quartile though, the two simulations
lead to an overestimation up to a factor of
Our findings are in contrast with the modeling study of Farrington et al. (2016), where a different approach was proposed to include the surface effect on the ICNCs simulated with WRF. In this study, a single model domain was used with a horizontal resolution of 1 km. To account for the flux of hoar crystals being detached from the surface by mechanical fracturing, Farrington et al. (2016) included a wind-dependent surface flux of frost flowers adapted from Xu et al. (2013). Despite the improved performance of WRF in terms of predicted ICNCs and LWC, the wind-dependent formulation of the surface flux caused the modeled ICNCs to become strongly correlated with the simulated horizontal wind speed – a behavior that was not confirmed by the observations of Lloyd et al. (2015). Nonetheless, the highest observed ICNCs at the beginning of the NW case correspond to the time when both the observed and modeled wind speed is the strongest (Fig. 2c), implying that a wind-dependent surface flux of BIPs could potentially elevate the simulated
It is deducible from the above discussion that the sole inclusion of a constant source of BIPs in our simulations cannot efficiently bridge the gap between modeled and measured ICNCs. Our aim in this section is to explore the combined effect of SIP through BR and blowing snow on the simulated orographic MPCs, since both processes are deemed to be important when trying to explain the high ICNCs observed in alpine environments. This is addressed in the final sensitivity simulation, BIPs100_PHILL, the results of which are compared with the CNTRL and PHILL simulations in Fig. 8.
Time series of
In terms of the modeled ice particle concentrations, the combination of the
simplified blowing snow treatment and BR parameterization can account for
most of the discrepancy between modeled and observed ICNCs, particularly
during the SE case (Fig. 8a), when the simulation leads to a best agreement
with the observed interquartile values (Table 3). BIPs100_PHILL and CNTRL generally differ by an average factor of
As the mixed-phase conditions are sustained throughout both case studies
(Fig. 8c), the plume of ice crystals is mixed into an ice-supersaturated
environment and, thus, BIPs are expected to promote ice growth through their
interaction with the surrounding supercooled liquid droplets and (ice) supersaturated air. The number of BIPs reaching the cloud base might not be
large, but their presence is expected to further facilitate the action of the BR mechanism, considering the depositional growth they will undergo within the supercooled boundary layer cloud. This is illustrated, for example, with the concurrent increase in
A discrepancy between modeled and observed IWC was also highlighted in the
study of Farrington et al. (2016) and was attributed to the small sizes of
the hoar frost particles assumed (i.e., 10
One final point that is worth noting here is that there are still certain periods when BIPs100_PHILL fails to reproduce the observed range of ICNCs. This could imply the potential contribution of additional ice multiplication processes to the observed ice particle concentrations. Indeed, the seeder–feeder configuration observed in the examined case studies could favor the fragmentation of sublimating hydrometeors while falling through an subsaturated environment before entering the feeder cloud (e.g., Bacon et al., 1998; Deshmukh et al., 2022). The so-called
sublimational breakup is an overlooked SIP process which is not yet described in the M05 scheme. Also, note that the periods when the modeled ICNCs remain below the observed ice number levels are mainly identified when the simulated temperature drops below
This study employs the mesoscale model WRF to explore the potential impact of ice multiplication processes on the liquid-ice-phase partitioning in the orographic MPCs observed during the CLACE 2014 campaign at the mountain-top site of JFJ in the Swiss Alps. The orography surrounding JFJ channels the direction of the horizontal wind speed, giving us the opportunity to analyze two frontal cases occurring under NW and SE conditions.
DS and BR mechanisms were implemented in the default M05 scheme in WRF, in
addition to the HM parameterization, which, however, remained inactive in the
simulated temperature range (
To parameterize the number of fragments generated per ice–ice collision, we
followed again two different approaches. We used either the simplified temperature-dependent formulation of Takahashi et al. (1995) scaled for the size of the particle that undergo fragmentation (Sotiropoulou et al., 2021a) or the more advanced physically based Phillips et al. (2017a) parameterization. It is important to apply the Takahashi parameterization only to consider collisions between rimed ice particles, otherwise the number of generated fragments is significantly overestimated. Including a description of the BR mechanism is essential for reproducing the ICNCs observed in the simulated orographic clouds, especially at temperatures higher than
One of the most interesting outcomes of this study is the association of the enhanced BR efficiency with the occurrence of in-cloud seeder–feeder events, which are commonly found in Switzerland (Proske et al., 2021). While ice-seeding situations are associated with enhanced orographic precipitation in the alpine region, the CNTRL simulation fails to reproduce the elevated ICNCs reaching the ground. The falling ice hydrometeors experience efficient growth through aggregation in the seeder part of the cloud, which is enhanced when reaching the feeder cloud at altitudes below 2 km, where primary ice crystals form and grow through vapor deposition and riming. Aggregation of snowflakes seems to be the major driver of secondary ice formation in the examined seeder–feeder events. SIP through BR is initiated already within the seeder cloud, while it becomes immensely important in the feeder cloud, where its production rate exceeds the one of primary ice formation. The increased generation of secondary ice fragments does not always lead to ice explosion and cloud glaciation, as it is followed by an enhancement in the precipitation sink owing to a shift in the ice particle spectrum. Including a description of the BR mechanism is, therefore, crucial for explaining the ice particle concentration and mass observed in the low-level feeder clouds.
To assess the potential role of blowing snow in the simulated orographic
clouds, a constant source of ice crystals was introduced in the first atmospheric level of WRF. Our results indicate that blowing snow alone cannot explain the high ICNCs observed at JFJ, but when this source is combined with the BR mechanism then the gap between modeled and measured ICNCs is sufficiently bridged. The biggest influence of blowing snow is mainly detected at times when the simulated temperatures are low enough (
Overall, our findings indicate that, outside the HM temperature range, a correct representation of both secondary ice (through BR) and an external ice-seeding mechanism, which is primarily precipitating ice particles formed aloft and, to a lesser degree, windblown ice from the surface, is fundamentally important for accurately predicting the liquid-ice-phase partitioning and properties of MPCs. Given the high frequency of seeder–feeder events in orographic environments, including the new physics of BR may address a large source of predictive bias in atmospheric models.
The WRF outputs and the microphysical observations presented in this study can be downloaded from the EnviDat data portal at
The supplement related to this article is available online at:
PG and AN conceived and led this study, with input from GS. EV helped with the WRF configuration and setup. GS provided the updated microphysics scheme with the detailed BR parameterizations. PG implemented the DS parameterizations, with help from GS, conducted the simulations, analyzed the results, and, together with AN, wrote the main paper. AB and ACBR provided the processed radar data and created Fig. S6. All authors contributed to the scientific interpretation and writing of the paper.
The contact author has declared that neither they nor their co-authors have any competing interests.
Publisher’s note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The authors would like to thank Gary Lloyd, for providing the microphysical measurements, and Jacopo Grazioli, for collecting and pre-processing the radar data. The authors are also thankful to Varun Sharma and Michael Lehning, for the fruitful discussions on the contribution of blowing snow in the alpine region.
This research has been supported by the European Research Council, H2020 European Research Council (grant nos. FORCeS (821205) and PyroTRACH (726165)) and the Svenska Forskningsrådet Formas (grant no. 2018-01760). Georgia Sotiropoulou received funding from the Swedish Research Council for Sustainable Development FORMAS (project ID 2018-01760). Georgia Sotiropoulou and Athanasios Nenes have been supported by the European Union under the H2020 Marie Skłodowska-Curie Actions project SIMPHAC (grant no. 898568).
This paper was edited by Johannes Quaas and reviewed by two anonymous referees.