The Response of Cloud Dynamic Structure and Microphysical Processes to Glaciogenic Seeding: A Numerical Study

Abstract

Stratocumulus clouds are cloud systems composed of stratiform clouds with embedded convective clouds, possessing strong catalytic potential and serving as key target cloud systems for weather modification operations. In this study, the parameterization of ice nucleation for silver iodide (AgI) particles was applied to the Thompson microphysics scheme in the WRF model. Numerical experiments were designed for a stratocumulus cloud that occurred over the Hulunbuir region, northeastern China, on 31 May 2021, to investigate how the structure and evolution of cloud macro- and microphysical properties and precipitation formation respond to glaciogenic seeding. The simulation results indicate that AgI nucleation increased ice concentrations at 4–5 km altitude, enhancing ice crystal formation through condensation–freezing and deposition nucleation and the growth of ice particles through auto-conversion and riming, leading to increased precipitation. The results also show that owing to the non-uniform distribution of supercooled water within this stratocumulus cloud system, the consumption of AgI and the enhanced ice nucleation release latent heat more strongly in regions with higher supercooled water content. This leads to more pronounced isolated updrafts, altering the structure of shear lines and subsequently influencing regional precipitation distribution after silver iodide seeding concludes. These findings reveal that seeding influences both the microphysical and dynamic structures within clouds and highlight the non-uniform seeding effects within cloud systems. This study contributes to a deeper understanding of the effects of artificial seeding on stratocumulus clouds in high-latitude regions and holds significant reference value for artificial weather modification efforts in mixed-phase stratiform clouds.

1. Introduction

Stratocumulus clouds are cloud systems composed of stratiform clouds with embedded convective clouds. They represent a significant part of precipitation systems in northern China. The outer stratiform layer exhibits relative stability and prolonged precipitation duration, whilst the embedded convective cells deliver intense downpours. This cloud system serves as a prime target for weather modification experiments [1,2].

A number of studies have been conducted on stratiform mixed-phase clouds using radar, satellite, aircraft observations, and numerical simulation methods. Observations reveal that the average liquid water content within convective cells in stratiform mixed-phase clouds is approximately twice that of surrounding cloud regions. These clouds also exhibit higher cloud particle concentration and greater ascent velocity and produce more ice crystals [3,4]. Herzegh and Hobbs observed that ice crystal particles within embedded convective regions of stratocumulus clouds primarily grow through riming processes, whereas in the stratus cloud region, deposition growth predominates [5]. Analysis of two mixed-layer convective cloud processes in the Beijing region by Guo et al., based on joint multi-aircraft observations, confirmed a higher prevalence of rime-coated ice crystals within the embedded convective cells [6]. In numerical simulation analyses, Frederic et al. showed that the more vigorous the convective cloud embedded within stratiform clouds becomes, the higher the precipitation efficiency [7]. Stratiform clouds primarily release latent heat through enhanced condensation processes, strengthening updrafts and promoting cloud development, thereby significantly intensifying precipitation. Within convective cloud regions, precipitation particles undergo multiple cycles of vertical growth, allowing cloud particles to grow substantially larger. Higher-level hydrometeor particles falling from cumulus clouds into stratus clouds enhance precipitation efficiency [8]. Consequently, precipitation intensity is higher beneath convective cells, giving rise to the phenomenon of rain cores [9]. Y. Li et al. observed that during cloud system development, convective cells may merge into supercell convection, replenishing the moisture and energy required for its sustained growth and further development [10,11]. The high moisture content of stratocumulus-nimbus hybrid clouds and the substantial upward velocities within embedded convective cells indicate their catalytic potential.

Artificial weather modification is a technique that alters natural precipitation processes by manipulating static or dynamic forces within clouds. For mixed-phase clouds, this primarily involves seeding clouds with artificial ice nuclei. These nuclei consume supercooled water within the cloud, converting it into ice crystals. These crystals then form precipitation particles through diffusion growth and accretion mechanisms [12]. Given that silver iodide (AgI) possesses a crystal structure similar to that of ice nuclei, it has been extensively employed as an ice nucleating agent [13,14]. D. Li et al. examined the results of an outside field seeding experiment conducted on 21 May 2018 in Shandong Province, China [15]. They found that seeding silver iodide within the convective zone enhanced the conversion of supercooled droplets into ice crystals, releasing latent heat that strengthened updrafts and deepened convection. Following silver iodide seeding within stratiform clouds, radar returns weakened near the seeded layer height, with echo tops diminishing. Concurrently, features resembling ‘ice-seeding trails’ emerged, indicating that partial seeding within the stratiform cloud accelerated the conversion of liquid droplets into ice crystals. However, from an observational standpoint, it is challenging to determine the relative contributions of various intra-cloud microphysical processes to the phenomena observed post-seeding. This necessitates further investigation through model simulations.

In recent years, researchers have developed numerous parameterization schemes, driving the continuous advancement and refinement of numerical models. Fang et al. incorporated AgI interaction simulations into the cloud microphysical processes of mesoscale numerical models and discovered that different seeding rates exert varying effects on precipitation [16]. Lower seeding rates yielded negligible rain enhancement, whereas excessively high rates paradoxically suppressed precipitation. Xue et al. incorporated a cloud seeding parameterization scheme into the Thompson microphysical scheme within the WRF (Weather Research and Forecasting) model [17,18]. Through simulation of a two-dimensional idealized moist airflow crossing a bell-shaped mountain, they analyzed the influence of the parameterization scheme to meteorological conditions, cloud characteristics, and seeding rates. Their findings indicated that aerial seeding proved more effective than ground seeding, with aerial seeding predominantly employing sublimation nucleation of silver iodide, increasing precipitation on the windward slope. He et al. incorporated silver iodide particle–cloud interaction processes into the Morrison two-parameter scheme within the mesoscale WRF model and conducted numerical seeding experiments for a layer cloud over Beijing [19]. Their results revealed that seeding initially reduced precipitation before inducing increased rainfall. The initial precipitation reduction phase resulted from decreased cloud water and increased snow crystals, reducing droplet collision–coalescence and auto-conversion processes while increasing droplet and snow riming. However, snow had not yet fallen and melted to form rain droplets. The subsequent precipitation enhancement phase occurred due to falling and melting snow crystals. Chen et al. employed LES simulations to investigate the effect of turbulence on glaciogenic seeding within mixed-phase stratiform clouds [20]. By enhancing turbulence through increased vertical wind shear, they observed that stronger turbulence intensifies the nucleation process of supercooled water via silver iodide consumption. This leads to a short-term increase in precipitation rates but reduces precipitation downwind, with precipitation locations shifting. Moreover, heightened turbulence promotes liquid water formation. These studies demonstrate that seeding protocols substantially influence outcomes, yet past simulations predominantly analyzed post-seeding macroscale characteristics qualitatively or quantitatively, with limited research on microphysical and dynamic changes within clouds following actual case-specific seeding operations. Case-specific AgI seeding simulations in high-latitude stratocumulus clouds under Mongolian High and cyclone conditions, with explicit analysis of shear line structural changes, have not yet been conducted. This study addresses this gap by isolating the role of non-uniform supercooled water distribution in shaping localized updraft enhancement and precipitation redistribution, providing new insights into the microphysical and dynamical impacts of AgI seeding under these conditions.

Hulunbuir, situated in the northeast of the Inner Mongolia Autonomous Region, forms part of the Mongolian Plateau in central Asia. The central region of Hulunbuir hosts the Greater Khingan Mountains, which are an important forest resource in China. The western part consists of the Hulunbuir Grassland, where animal husbandry is highly developed, while the eastern part is dominated by low hills and plains, supporting an agriculture-based economy. Therefore, conducting weather modification operations in the Hulunbuir region is highly relevant, both for promoting the development of animal husbandry and for protecting forest resources. In addition, this region lies on the southeastern flank of the Mongolian High during spring and autumn, where layered mixed-phase cloud systems commonly develop, making it a particularly suitable target for AgI seeding experiments. This study incorporates the ice nucleation parameterization process of AgI particles into the Thompson microphysical scheme of the WRF model. A specially designed seeding experiment was conducted on a stratocumulus cloud in the Hulunbuir region on 31 May 2021. This study analyzes the macro- and microstructures of the precipitation process, the microphysical mechanisms influencing precipitation, and the effects of non-uniform supercooled water distribution on local updraft enhancement and precipitation redistribution, providing new insights into the microphysical and dynamical impacts of AgI seeding under high-latitude conditions. It also aids in understanding the effects of artificial seeding operations on stratocumulus clouds in this region, offering valuable reference for weather modification efforts.

2. Synoptic Situation and Model Description

2.1. Synoptic Situation

From 30 to 31 May 2021 (UTC), a significant precipitation event occurred across the Hulunbuir region under the influence of a Mongolian cyclone. Figure 1a–c depict the upper-level weather patterns at 500 hPa, 700 hPa, and 850 hPa, respectively, as reconstructed from the ERA5 reanalysis data from the European Centre for Medium-Range Weather Forecasts for 00:00 on 31 May 2021. Figure 1a indicates that the study area was situated ahead of a trough and behind a ridge. Upper-level divergence favored the generation of updrafts, while the temperature trough lagged behind the height trough, suggesting that the system would continue developing cloud systems and precipitation. Combined with Figure 1b, the 700 hPa trough positioned ahead of the 850 hPa trough resulted in superimposed convergence at both levels, further enhancing ascent. A deep low-pressure system transported warm, moist air into the Hulunbuir region. Figure 1c indicates that the study area lies within a low-pressure zone experiencing strong convergence shear. The convergence of cold and warm air masses triggered precipitation processes over the region. Figure 1d, a cloud-top brightness temperature image from the Himawari-8 satellite, reveals extensive stratiform cloud cover over the study area. The cloud band extends from southwest to northeast, aligning with the low-level shear direction and demonstrating potential for further cloud development.

2.2. Model Description

This study employed the mesoscale numerical weather research and forecasting model WRF version 4.6.1 for numerical simulation experiments. The Thompson microphysical scheme within the model can forecast cloud droplet, raindrop, ice crystal, snow, and sleet content and has been extensively applied in previous research for simulating cold clouds [21,22]. To investigate the impact of silver iodide seeding on cold cloud precipitation processes, a silver iodide forecast equation was incorporated into the Thompson microphysics scheme. Given that contact freezing contributes minimally to ice crystal formation, yielding activation rates two orders of magnitude lower than other activation processes [17,18], this study primarily considers three activation mechanisms for seeded AgI particles: deposition nucleation, condensation–freezing nucleation, and immersion–freezing nucleation. The parameterization scheme draws upon methods proposed by DeMott and Meyers et al. [23,24].

For deposition nucleation,

$${\mathrm{F}}_{\mathrm{d}\mathrm{e}\mathrm{p}}=\mathrm{a}\left({\mathrm{S}}_{\mathrm{i}}-1\right)+\mathrm{b}\left({\displaystyle \frac{273.16-\mathrm{T}}{{\mathrm{T}}_0}}\right)+\mathrm{c}{\left({\mathrm{S}}_{\mathrm{i}}-1\right)}^2+\mathrm{d}{\left({\displaystyle \frac{273.16-\mathrm{T}}{{\mathrm{T}}_0}}\right)}^2+\mathrm{e}{\left({\mathrm{S}}_{\mathrm{i}}-1\right)}^3,$$

where F_dep is the nucleation rate of AgI particles, T₀ = 10 K, a = −3.25 × 10⁻³, b = 5.39 × 10⁻⁵, c = 4.35 × 10⁻², d = 1.55 × 10⁻⁴, e = −0.07, S_i is the ice surface saturation, and T is the Kelvin temperature. The formula is valid when S_i > 1.04 and T < 268.2 K.

For condensation–freezing nucleation,

$${\mathrm{F}}_{\mathrm{c}\mathrm{d}\mathrm{f}}=\mathrm{a}{\left({\displaystyle \frac{268.66-\mathrm{T}}{{\mathrm{T}}_0}}\right)}^3{\left({\mathrm{S}}_{\mathrm{w}}-1\right)}^2,$$

where F_cdf denotes the condensation–freezing nucleation rate of AgI particles, with T₀ = 10 K and a = 900. S_w represents the water vapor saturation level. The formula remains valid when T < 268.66 K and S_w > 1.

For immersion–freezing nucleation,

$${\mathrm{F}}_{\mathrm{i}\mathrm{m}\mathrm{f}}=\mathrm{a}{\left({\displaystyle \frac{268.2-\mathrm{T}}{{\mathrm{T}}_0}}\right)}^{\mathrm{b}},$$

where F_imf is the immersion–freezing nucleation rate of AgI, with T₀ = 10 K, a = 0.0274, and b = 3.3. The formula remains valid for temperatures below 268.2 K.

Since the calculated nucleation rates of each process are the ratios of silver iodide nucleated through different mechanisms to the total activated silver iodide, the sum of these three nucleation and activation rates does not exceed 1. In the mode, the mass of ice crystals produced after the nucleation of silver iodide is transformed into ice crystal mass. Condensation–freezing and deposition nucleation are constrained by available water vapor conditions, whilst immersion–freezing nucleation is limited by the quantity of available cloud droplets. Furthermore, according to the parameters set by Xue et al., these three activation rates are equal to their values at 248 K when the temperature falls below this threshold [17].

According to Deng et al., the YSU scheme demonstrated greater accuracy in delineating precipitation areas compared to the MRF scheme [25]. Tian et al. simulated eight distinct rainfall events across northern China to compare the performance of different boundary layer schemes, concluding that the YSU scheme outperformed the MYJ scheme in simulating the spatial distribution of precipitation at high resolutions [26].

In summary, this study employs a three-layer nested configuration with horizontal resolutions of 9 km, 3 km, and 1 km, respectively, and 51 vertical levels. The innermost nested horizontal grid is set at 451 × 451, simulating the region and terrain depicted in Figure 2. Model initial and boundary conditions were derived from ERA5 data with a spatial resolution of 1° × 1° and hourly temporal resolution. The simulation period spans from 12:00 on 30 May 2021 to 12:00 on 31 May 2021. The Kain–Fritsch scheme is employed for convective cumulus clouds in the first layer. The second and third convective units can be directly simulated within the grid without requiring a parameterization scheme, thus rendering a cumulus parameterization unnecessary. Specific model settings are detailed in

Table 1. WRF settings.

	Domain 1	Domain 2	Domain 3
Grid spacing	9 km	3 km	1 km
Lattice number	275 × 275	478 × 478	451 × 451
Model top height	100 hPa	100 hPa	100 hPa
Vertical layer	51	51	51
Cloud microphysics solution	Thompson	Thompson	Thompson
Boundary layer scheme	YSU	YSU	YSU
Cumulus parameterization scheme	Kain-Fritsch	\	\
Long/short wave radiation scheme	RRTMG	RRTMG	RRTMG
Land surface process scheme	Noah	Noah	Noah

.

3. Simulation Results Analysis

3.1. Comparison of Measured and Simulated Results

To validate the reliability of the model simulation results, a comparative analysis was conducted using Himawari-8 cloud-top temperature data as a reference against the model-simulated cloud-top temperatures. The cloud-top temperature data were segmented into intervals of 3 K. The number of grid points at each cloud-top temperature interval was counted, and the frequency distribution of cloud-top temperatures across different temperature intervals was obtained by dividing the number of grid points in each interval by the total number of grid points. The cloud system under study in this case study began forming around 04:00 on 31 May and exited the simulation domain around 10:00. The cloud top temperature and precipitation rate at 06:00, when the cloud region was most active, were selected for comparison. As shown in Figure 3a, analysis of Himawari-8 data (black solid line) reveals that cloud top temperatures primarily ranged between 215 K and 290 K, with a peak around 225 K, indicating the development of a relatively high cold cloud top. For the WRF model simulation results, regions with total water content exceeding 1 × 10⁻⁵ kg/kg were designated as cloud areas. The figure reveals that the peak cloud-top temperature is approximately one temperature band lower than the satellite data, with a higher frequency of lower temperatures. Himawari-8 data utilized three infrared channels from the AHI instrument, employing the 1DVAR optimal estimation method with Levenberg–Marquardt iteration to derive optimal solutions for cloud-top temperature generation. Conversely, our WRF cloud-top temperatures were derived from water content. Differences in these diagnostic approaches may introduce certain biases. Furthermore, prior research indicates that the Thompson parameterization scheme may simulate stronger cold clouds, potentially explaining the lower simulated cloud-top temperature peak. Nevertheless, the probability distribution trend of cloud-top temperatures aligns reasonably well with satellite data, broadly reflecting the primary characteristics of natural clouds. Figure 3b indicates that the modelled precipitation rates are slightly higher than those calculated from the GMCP dataset [27], correlating with the simulated stronger cold clouds, with a domain-mean accumulated precipitation bias of approximately 2.7 mm. The trend of precipitation rates increasing at 6:00 and decreasing after 10:00 is consistent. To further validate the reliability of the simulation results, comparisons were made between the simulated data and MICAPS4 ground-based observations for surface temperature and dew point temperature. Given the limited number of stations within the region, the comparison revealed consistent temperature distributions between the observed and simulated data: temperatures were lower in the west than in the east (Figure 4a and Figure S1). Furthermore, dew point temperatures at the two northeastern stations were higher than at the other three stations (Figure 4b), aligning with the simulated distribution. The model simulations thus corresponded well with actual conditions.

3.2. Seeding Test Design

To investigate the effect of aircraft seeding with AgI catalysts on the precipitation process within cold clouds, suitable seeding times and locations were selected through analysis of supercooled water paths (SLWPs) within the study region. To determine optimal seeding timing, regional supercooled water paths were calculated. The simulation revealed another cloud zone with substantial supercooled water content near the lower boundary of the study area. To mitigate the influence of this boundary cloud zone, only grid points with SLWP values exceeding 5 kg/m² and maximum SLWP values were calculated for regions north of 47.5° N latitude, as depicted in Figure 5. It can be observed that the first peak in the supercooled water path occurs around 4:30, while the number of grid points with SLWPs exceeding 5 kg/m² reaches its first peak around 4:55. Based on the temporal trend of SLWPs, this study selected the period from 4:30 to 5:00 as the seeding time.

Figure 6a depicts the spatial distribution of supercooled water pathways at 04:30 on 31 May. It reveals abundant supercooled water within the range of latitudes 48° N to 48.5° N and longitudes 119° E to 119.4° E during this period, with maximum values occurring at approximately 48.2° N and 119.1° E. Considering the prevailing southwesterly winds at 500 hPa, it was determined to conduct a longitudinal straight-line seeding operation along 48.2° N. Based on the operational aircraft’s cruising speed, approximately one degree of longitude could be covered along this latitude within half an hour. Consequently, the seeding range was set from 118.8° E to 119.8° E. In operational AgI seeding, burn-in-place (BIP) flares typically release 16.2 g of AgI per flare over a burn duration of about 4.5 min [28]. We conducted the experiment using nine flares; to reduce the calculation error caused by the decimal point, an approximate value of 150 g was used in the mode. Figure 6b presents a latitudinal cross-section of supercooled water mass concentration, wind field, and temperature along 48.2° N at this time. The maximum supercooled water content occurred at approximately 4.5 km altitude. Given that silver iodide exhibits higher activation rates within the −5 °C to −10 °C range [29], seeding at around 4 km altitude was deemed appropriate. The specific seeding test settings are shown in

Table 2. Seeding test settings.

Time	Altitude	Trajectory	AgI
4:30–5:00	4.5 km	48.2° N, 118.8–119.8° E	150 g

.

3.3. Diffusion and Nucleation of AgI Particles

Figure 7 shows the horizontal distributions of AgI particle concentration after 1 to 5 h of seeding, where the shaded values represent the total silver iodide particle concentration across the entire layer at each grid point. The cross-sections of silver iodide concentration at corresponding times, plotted along the red lines in Figure 7, are shown in Figure 8. It can be observed that silver iodide moves and diffuses downwind over time, with a horizontal diffusion range reaching approximately 200 km. Due to strong updrafts within convective cloud cells, a small portion of silver iodide can be carried to the upper cloud layers, reaching heights exceeding 10 km. Five hours after seeding cessation (Figure 8f), the silver iodide particles are carried downward by the descending airflow within the cloud. Within the two-hour period from seeding commencement to completion, silver iodide particles essentially covered the regions of maximum supercooled water content within the study cloud area. The silver iodide primarily resided below 6 km altitude, gradually diminishing in concentration as it interacted with cloud water and water vapor during its movement and diffusion.

The nucleation of AgI particles to form ice crystals in this study encompasses three mechanisms: vapor deposition, condensation–freezing, and immersion–freezing of droplets. Deposition and condensation–freezing nucleation are jointly influenced by temperature and humidity, whereas immersion–freezing is solely temperature-dependent. Figure 9 illustrates the temporal and altitudinal variations in the total nucleation rate of AgI and the activation rates of the three nucleation mechanisms within the study area after seeding trials. It is evident that nucleation of AgI commences immediately after the particles are dispersed, predominantly occurring within the first three hours post-seeding (Figure 9a). The nucleation mechanism is dominated by condensation–freezing (Figure 9c), followed by deposition nucleation (Figure 9b) at a level one order of magnitude lower, with immersion–freezing nucleation (Figure 9d) being the weakest. Condensation–freezing and immersion–freezing nucleation predominantly occur between 4 and 6 km altitude, whereas deposition nucleation occurs at higher altitudes, reaching a maximum height of approximately 10 km. Condensation–freezing and deposition nucleation constituted the two predominant nucleation processes in this seeding trial, accounting for approximately 60% and 40% of total nucleated AgI, respectively, while immersion–freezing represents a slow activation process [30]. Due to the supersaturation caused by the updrafts in the clouds, the majority of silver iodide particles are consumed by condensation–freezing and deposition nucleation processes, rendering the immersion–freezing mechanism comparatively weaker. During the initial seeding phase, silver iodide particles are transported to higher altitudes by updrafts, enabling condensation–freezing activation above 7 km. As silver iodide particles are consumed, their concentration decreases, weakening condensation–freezing activation and lowering its altitude threshold.

3.4. Changes in Precipitation

To investigate the impact of AgI seeding on precipitation, Figure 10 illustrates the cumulative precipitation differences between seeded and control runs. Within the black rectangular area in Figure 10a, precipitation increased within three hours post-seeding, with the maximum increment reaching 12.6 mm and the grid-average precipitation rising by approximately 0.3 mm. Precipitation changes three hours after seeding (8:00–12:00 UTC) within the region are depicted in Figure 10b. The location of precipitation within the black rectangular area changed markedly, though the total precipitation volume remained largely unchanged. The maximum change in grid precipitation reached 20 mm. This outcome differs somewhat from prior studies reporting a sequence of rain suppression followed by enhancement. Subsequent analysis will examine cloud water droplets, microphysical processes, and regional dynamic conditions to interpret these experimental results.

3.5. The Effect of AgI Seeding on Microphysical Processes in Clouds

For this simulation, we considered five types of hydrometeor particles, including cloud droplets, rain drops, ice crystals, snow, and graupel. Within the first 3.5 h after seeding commencement (4:30–8:00), a comparison of the water contents of various hydrometeor particles within the black rectangular region in Figure 10a between the control and seeded runs revealed that the water content of snow is one to two orders of magnitude higher than that of the other two types of ice-phase particles, ice and graupel, while the mixing ratios of liquid-phase particles were comparable. In subsequent analyses, grid cells with total water content exceeding 10⁻⁵ kg/kg were classified as cloud regions. Figure 11 depicts the regional average variations in water contents and differences (seeding trial minus control trial) for rain, cloud, ice, snow, and graupel over time and altitude within the black rectangular region of Figure 10a. The ice water content from the control experiment (Figure 11(c1)) shows that ice was predominantly distributed between 6 and 11 km altitude, peaking at around 8.5 km. Following AgI seeding (Figure 11(c2)), the ice water content increased below 6 km. The changes before and after seeding (Figure 11(c3)) indicate that the primary increase occurred between 4 and 6 km, corresponding to the nucleation altitude of AgI.

By examining the variations in the water contents of snow and graupel (Figure 11(d3,e3)), it can be seen that snow content exhibits a modest increment at altitudes between 3 and 5 km but decreases substantially above 5 km, while graupel’s ice water content increases almost continuously within the region, with a more pronounced increase after 07:00 and a simultaneous rise in the lower levels. Except for a brief increase around 07:00 at 5–6 km, the cloud water content generally exhibits a decreasing trend, as the activation of silver iodide consumes cloud water. The rainwater content increases within three hours after seeding, during which the snow content below 4 km also increases. The magnitude of the snow content change is one order of magnitude greater than that of the rainwater content, suggesting that melting precipitation dominates in the study area, and the low-level snow content has the greatest impact on precipitation.

After seeding, AgI nucleation consumed water vapor and cloud droplets and produced small ice crystals. By analyzing the differences between the seeded and control experiments in the physical processes of ice variation in Figure 12, it can be seen that the deposition (ide) process of ice is significantly enhanced at 3–6 km, indicating a stronger deposition process that consumes more water vapor and promotes ice crystal formation and growth. The sublimation and increased content of ice lead to a marked enhancement of the self-conversion process of ice around 4.5 km, which serves as the main sink of ice crystals during the entire process and is the primary reason for the increase in low-level snow. The enhancement of the auto-conversion of cloud ice to snow (iau) process is most evident within 3.5–6 km altitude, corresponding to the altitude where AgI nucleation and ice deposition growth take place, resulting in increased ice content, with the additional ice aggregating to form snow at this level. Changes in various processes are less pronounced in the upper layers; around 8 km, the deposition (ide) and freezing (wfz) processes show slight growth, while above 10 km, the deposition process weakens, corresponding to the slight increase in ice crystals at 8–10 km and the decrease above 10 km shown in Figure 11(c3).

From the source-sink terms for snow and graupel (Figure 12a,d), it is evident that increased snow formation at approximately 5 km altitude, driven by ice aggregation, enhances the rain–snow collection (scw) process. This process generates both snow and sleet, while the variation in graupel formation is negligible. Consequently, the process predominantly contributes to snow accumulation. This process consumes atmospheric water vapor, reducing its content and thereby enhancing snow sublimation (ssu). However, this increase is modest compared to the increase in volume. Snow deposition (sde) constitutes the primary sink for snow, and its continuous weakening leads to diminished upper-level snow content. Rain droplet content increases only for a limited period following seeding. Above the zero-degree layer and up to 6 km altitude, the mass content of graupel increases due to the collection of droplets by graupel (gcw) and ice–rain collision (rci). The enhanced melting of snow and graupel particles below the freezing level constitutes the primary cause of increased rainfall. By examining Figure 12c, it can be found that the main process causing the increase in rainwater content is the increase in the mass of low-level snow and graupel after seeding. Raindrops produced by the melting of snow and graupel contribute approximately 80% of the total raindrop source term, whereas the collision–coalescence of raindrops and cloud droplets accounts for only about 6%. Snow melting (sml) and graupel melting (gml) generate more raindrops, further indicating that this precipitation process is dominated by melting precipitation, with snow melting playing the most important role.

3.6. Changes in Cloud Dynamic Structure

During the study period, a Mongolian cyclone moved from west to east, and the seeded cloud area formed on the southeast side of the cyclone and moved northeastward with it. Oue et al. found that vertical wind shear can generate turbulent updrafts, with snowbands forming in the low- to mid-level frontal zones, where upward motion exists throughout the layer, positively influencing snow formation and snowfall [31]. Figure 13 shows the vertical velocity and horizontal wind fields at 850 hPa (a1–a3), 700 hPa (b1–b3), and 500 hPa (c1–c3) at 07:00 for the control experiment (a1–c1), the seeding experiment (a2–c2), and their differences (a3–c3), with shading indicating the distribution of vertical velocity. From Figure 13(a1–c1), it can be seen that at 07:00, a low-level wind shear formed in the study area due to the convergence of low-level southwesterly and southeasterly flows, producing convergence that forced air to rise. This uplift supported continuous cloud development, extended the cloud lifetime, and concentrated precipitation near the shear line. It is also an important reason for the higher snow content in this area. From Figure 13(a2–c2), it is evident that the low-level wind shear still exists in the seeding experiment, but its structure changes, breaking near 49.7° N, and the updraft is no longer as uniform and continuous as in the control experiment, showing localized enhancement instead. Figure 13(a3–c3) shows that after seeding, the updraft is enhanced downstream while it weakens upstream, indicating that the shear position shifts downwind. From the changes in the microphysical structure of the cloud, it can be inferred that AgI nucleation consumed supercooled water and enhanced the deposition process of ice, and the latent heat released by these processes strengthened the updraft. Because the supercooled water in the study area is not uniformly distributed, this effect is stronger in regions with higher supercooled water content, resulting in more pronounced local updraft enhancement and a structural change in the shear line, as shown in Figure S2. This further leads to stronger local circulations on both sides of the shear line and an increased vertical velocity gradient near the shear line. In the control experiment, the maximum vertical velocity gradient at 500 hPa is about 2.7 (m/s) km⁻¹, whereas in the seeding experiment, it reaches about 3.5 (m/s) km⁻¹. Strong updrafts cause rapid expansion and cooling of air with height and transport air with higher relative humidity to upper levels, thereby enhancing heterogeneous freezing nucleation and deposition growth of ice crystals and promoting the formation of small ice crystals. Compared with the continuous and uniform updraft in the control experiment, the increased updraft in the seeding experiment raises the snow mass concentration in local areas. However, the deposition growth of snow requires a certain time scale, and the change in the shear line structure leads to insufficient duration of the updraft, resulting in an overall weakening of the deposition process and a decrease in upper-level snow content.

After 08:00, silver iodide diffused to the edges of the region with the wind field, and the main nucleation process had essentially ended. With no additional input of ice nuclei, changes in precipitation were influenced by the distribution of hydrometeor particles and the dynamical structure within the region. Although the location of precipitation shifted, the total precipitation amount remained nearly unchanged.

4. Conclusions

This study used the mesoscale WRF model with the Thompson double-moment microphysics scheme coupled with a silver iodide (AgI) seeding parameterization to numerically simulate a stratocumulus mixed-phase precipitation event over the Hulunbuir region in northeastern China on 31 May 2021. Based on an analysis of the supercooled water path in the stratocumulus mixed-phase cloud, an appropriate seeding scheme was designed to conduct the cloud seeding simulation. By comparing the simulated results of the seeded and natural clouds, the main conclusions are as follows:

1. AgI diffused northeastward with the wind field, with a horizontal spread of about 200 km and maximum vertical transport reaching near the cloud top with the cloud updraft, though most remained below 5 km. Five hours after seeding ended, most AgI either moved out of the region with the wind field or settled to the ground with downdrafts. The nucleation of AgI mainly occurred within three hours after seeding, dominated by condensation–freezing and deposition nucleation, accounting for 57.81% and 38.09% of the total nucleated AgI particles, respectively, and primarily occurred within the 4–5 km layer.

2. Within three hours after the end of seeding, precipitation increased downstream, with local rainfall enhancement reaching up to 12.6 mm. After the main AgI nucleation process ended, the location of precipitation shifted within the region, but the total precipitation amount remained nearly unchanged.

3. Snow had the highest mixing ratio in the study area. After seeding, AgI nucleation produced ice crystals, with the ice water content increasing mainly within 4–6 km. Due to locally enhanced updrafts, small ice crystals increased near 8 km as a result of freezing nucleation and deposition growth. Snow content increased within 3–5 km during the three hours after seeding but decreased at higher levels. Changes in low-level snow content were consistent with trends in raindrop mass concentration. Enhanced collection of droplets by graupel and ice–raindrop collision processes led to an increase in graupel mass.

4. The main precipitation mechanism was the melting of snow and graupel, followed by the collision–coalescence process of raindrops and cloud droplets. Thus, precipitation was dominated by melting precipitation. The processes of ice crystal-to-snow conversion and snow–cloud droplet aggregation forming graupel were strengthened within 3–6 km, increasing the mass of snow and graupel falling below the freezing layer and melting into precipitation, with snow melting playing the dominant role.

5. A low-level wind shear formed in the study area due to the convergence of low-level southwesterly and southeasterly flows, forcing air to rise and supporting the continuous development of the cloud system. Silver iodide consumed supercooled water to produce ice crystals, and the enhanced deposition process of ice crystals released latent heat. Because the supercooled water was unevenly distributed in the study area, this effect was stronger in regions with higher supercooled water content, leading to more pronounced local updraft enhancement and structural changes in the shear line. The change in the shear line structure shortened the duration of the updraft, resulting in an overall weakening of the deposition process and a decrease in upper-level snow content.

The results of this study indicate that in high-latitude stratiform mixed-phase clouds, seeding with silver iodide in regions of high supercooled water content produces a more pronounced seeding effect. However, this effect does not necessarily lead to a significant increase in total rainfall; rather, it may manifest as a redistribution of precipitation in time or space, which is also an important factor in evaluating seeding effectiveness.

This study has some limitations. First, it focuses on a single case and a specific seeding configuration, which may limit the generality of the results. Second, uncertainties remain in the simulation of AgI nucleation parameters and the representation of turbulence. Future work will address these limitations by conducting sensitivity experiments on seeding rate, timing, and location, improving the parameterization of silver iodide nucleation in the model, and coupling with large-eddy simulation (LES) to better resolve turbulent mixing within embedded convective units.