Hole-punch clouds are typically associated with circular or linear voids after an aircraft penetrates through a supercooled liquid cloud layer . Prior to the formation of the hole, this cloud is void of any ice because of missing aerosol particles, i.e., ice-nucleating particles (INPs), that help form ice crystals. However, the passage of an aircraft results in an inadvertent seeding effect of the cloud. The expansion of air behind an aircraft can lead to a cooling by 20–30 °C, often enough to kick-start homogeneous ice nucleation, bypassing the need for any INPs . Upon the formation of ice crystals in a supercooled liquid cloud, the subsequent evolution of a hole-punch cloud depends on the interactions between cloud droplets and ice crystals. One such interaction is the Wegener–Bergeron–Findeisen (WBF) process and it is known to be crucial for the glaciation of a cloud as well as for continental precipitation formation. It takes place because of different saturation water vapor pressures over water (es,w) and ice (es,i). At water-subsaturated conditions cloud droplets will evaporate, while ice crystals continue to grow as long as they experience ice supersaturated conditions (es,i<e<es,w, where e is the ambient water vapor pressure). Moreover, the evaporating cloud droplets serve as a reservoir of water vapor, which can be taken up by the ice crystals. In the case of e>es,w, both hydrometeors grow while in case of e<es,i, cloud droplets evaporate and ice crystals sublimate. During the WBF process, ice crystals can grow large enough to start to sediment and form fallstreaks below the cloud. This terminates the extent of the hole, because no more interactions with the cloud droplets are possible. Exactly this process-chain was also observed downstream of industrial areas, which emit anthropogenic aerosols . They observed a reduction in shortwave reflection as well as in cloud cover and cloud optical thickness.

The importance of the conversion of liquid to ice through the WBF process was early-on recognized in the weather modification community. conducted first experiments of ice crystals in supercooled clouds and identified the importance of silver iodide acting as an INP to form ice crystals already at -5 °C. At these temperatures, very few natural INPs exist, such that silver iodide is typically used in targeted cloud seeding operations. Across the globe cloud seeding programs emerged with focus on scientific understanding as well as operational usage for precipitation enhancement . Within the scientific community, a project called “Seeded and Natural Orographic Wintertime Clouds: The Idaho Experiment” demonstrated the microphysical process-chain upon seeding orographic clouds using remote sensing instruments . They also highlighted fundamental knowledge gaps in our understanding of mixed-phase cloud processes. Building on their findings, the CLOUDLAB project aims to address these knowledge gaps by employing the methods of weather modification, i.e., glaciogenic seeding, in a controlled and local manner . Instead of targeting orographic cloud systems, we chose a stable low-level stratus cloud to conduct our seeding experiments. This way, confounding factors, such as orographic forcing, are minimized. The focus was mainly on the quantification of ice growth rates within supercooled clouds in a natural setting . We also created a numerical modeling framework that can be used to reproduce the seeding experiments . This framework forms the foundation for this study, such that we can assess the interplay of the liquid and ice phases across a range of cloud states. We aim to answer how the background state, in particular the liquid water content and path, of a supercooled liquid cloud defines the effectiveness of forming a hole-punch cloud by glaciogenic seeding. Additionally, we perturb the turbulent mixing length to assess the effect of enhanced and reduced subgrid-scale mixing within the model simulations. The targeted glaciogenic seeding experiments during CLOUDLAB closely follow the underlying mechanism of hole-punch clouds, as a local perturbation within a liquid cloud triggers the WBF process and changes the cloud characteristics.

We first show single time steps during the evolution of the seeding plume in L25 (Fig. ) to give an overview of the seeding simulations. We chose the time steps 1, 4, and 7 min after the seeding start, such that the first changes in the cloud can be discussed as well as after the seeding concluded (6 min after seeding start). The full evolution of the changes are discussed in the next section. From Fig. a–c it is evident that the reductions in LWC are slightly delayed compared to the nucleation of ice crystals. The ice crystals first need to grow to a certain size to be effective in reducing the LWC as we showed in a previous study, where the ice crystal size was underestimated by almost a factor 2 compared to observations . Nonetheless, the reductions highlight the WBF mechanism. After 1 min of the seeding start, a fine line appears downstream of the highest ICNC, along which LWC is reduced up to -100 %, and this signal spreads with time. These stronger reductions can be visible across a few hundred meters, both in latitude and longitude direction. Note that we chose to show the changes along a constant height, but it is evident that the signal also spreads vertically as shown in Fig. d–f. The minimum of the mean change at 1 min coincides with the seeding level as here the highest concentration of ICNC is to be expected. These strong changes in LWC are also visible across time as shown in the last row of Fig. h–i, which we discuss in the next section. Here, we first want to assess the impact of varying turbulent mixing lengths acting on the subgrid-scale, expressed by the Smagorinsky constant Cs. A value of Cs = 0.17 is identified as a minimum value and indicates more turbulent flows, while a value of 0.3 creates a more homogeneous field by enhancing the mixing. The default value in the model is 0.23 (highlighted in Fig. g). The simulated ICNC and absolute and relative changes in LWC are rather insensitive to Cs, i.e., their response over time is unchanged. Hence, the subgrid-scale mixing seems not to be a crucial parameter in changing cloud microphysical processes for this particular cloud type. This is in contrast to other studies , which found a dependence of Cs, but also employed coarser horizontal grid resolutions. We hypothesize that in our case the explicitly resolved turbulence (grid-scale at 65 m) is dominant enough, such that the sub-grid scale effect plays a minor role. A more thorough analysis of horizontal grid spacing is the subject of a future study building on the work in CLOUDLAB. In the next section, we computed averages across these three simulations for each seeding simulation. Given the low spread between the perturbations, we are only showing the average and no shading. We also want to note here that the absolute changes in LWC are approximately one order of magnitude smaller than what was observed (see Fig. c). While we selected the seeding simulations based on the background LWP (= 25 gm-3), we did not achieve an agreement between observed and simulated LWCs. In our simulations, the simulated clouds have a higher vertical extent (see Fig. ).

Figure  summarizes the stronger changes in LWC and cloud optical thickness by computing the 5th percentile for each seeding simulation. The 5th percentile represents the lower tail of the distribution of LWC, i.e., the 5 % of grid points which experience the strongest changes. The cloud optical thickness was calculated as follows: τ≈∫32h(z)LWC(z)ρlr(z)dz, where h(z) refers to the layer thickness in m, ρl is the density of liquid water in kgm-3, and r is the cloud droplet radius in m. In absolute terms, the reductions are stronger for seeding simulations with higher LWPs, because there is more liquid that can be consumed. This is also the case for the timing of the strongest reductions, which scale in most cases with the initial LWP. We also computed the maximum hole area at these minima by calculating the cell area of the associated grid points. We find a similar pattern: The area of the hole, i.e., strong reductions in LWC, is larger with increasing LWP (see Table ). At first this seems counterintuitive, however, our hypothesis is that there are two mechanisms defining the hole-punch cloud. (1) Ice crystals form, grow, and lead to the evaporation of cloud droplets through the WBF mechanism creating the hole. (2) As soon as maximum changes of -100 % LWC are achieved, no more liquid water is present to be consumed, such that the extent of the hole is limited. To achieve faster and more wide-spread reductions, a stronger seeding perturbation across a larger area (i.e., longer seeding legs with higher seeding concentrations and/or during a longer seeding time period to achieve higher ICNC) would be required. Hence, ICNC as well as the surrounding LWC matter for the size of a hole-punch cloud. Furthermore, we identify two clusters: For a low background LWP, we see the first -100 % changes after 7–8 min, e.g., for L25, L40, and L50. The other simulations form a cluster with strongest LWC reductions after 11 ± 1 min. Hence, a delay in the appearance of strong LWC reductions is notable. It also notable that for seven out of 9 seeding simulations a second local minimum is achieved (between 20 and 25 min), which is also associated with 100 % LWC reductions. As the seeding plume is advected in space, it encounters background states with first lower and then again higher LWCs, such that a renewed reduction in LWC is possible at later times. The cloud optical thickness evolves in accordance with the LWC reductions, where we see an almost complete dissipation of the cloud.

During the CLOUDLAB field campaigns, we observed the emergence of a hole-punch cloud upon glaciogenic seeding of a low-level stratus cloud. Observations show that strong reductions in LWC coincided with high ICNC, following the WBF mechanism. These findings prompted an investigation into the sensitivity of the seeding-induced hole-punch clouds using numerical simulations. We employed the numerical weather prediction model ICON in large-eddy mode at a horizontal resolution of 65 m and simulated one seeding experiment with different starting LWPs, while the temperature and seeding concentration were kept constant. A total of 9 seeding simulations were performed with LWPs ranging from 25–120 gm-2. For each of the 9 simulations, two additional simulations with different turbulent mixing lengths (Smagorinsky constant) were performed.

The first reductions in LWC were associated with the emission of seeding particles at the seeding level, where the first ice crystals formed. With the dispersion and sedimentation of the ice crystals, reductions in LWC were identified across the entire cloud. Overall, areas of strong LWC reductions were up to 1 km2. Focusing on the strong reductions in LWC (absolute and relative) and cloud optical thickness, we identified a delaying effect with increasing LWP: the strongest reductions appeared later in time, with maximum differences of up to 4 min. This behavior was consequently also reflected in the changes in cloud optical thickness, with stronger changes (< -60 %) appearing only 10 min after the seeding start. We also found that the turbulent mixing length has little to no impact on the changes in LWC. This raises the question if the explicitly resolved turbulence is dominant enough, such that no additional subgrid-scale turbulence is required. and both found a strong dependence on horizontal resolution when simulating glaciogenic cloud seeding. With increasing resolution, the efficiency through the WBF process to form precipitation increased. Hence, if one wants to successfully simulate glaciogenic seeding at coarser horizontal resolutions, the subgrid-scale turbulence may be of higher importance.

This study took on the interesting phenomena of hole-punch clouds, which are known from aircraft flying through a supercooled cloud layer, and investigated them in the context of glaciogenic cloud seeding. Our findings highlighted the interplay of ice crystal growth and the liquid background state of a cloud, i.e., the WBF mechanism. We believe our study provides a foundation for improving our understanding of mixed-phase processes and can inform future cloud seeding strategies.