Research Report: Analysis of Cavum Signatures from Rainmaker's Field Operations

Executive Summary

Rainmaker has generated at least seven cavum signatures using seeding drones in Oregon. The detection of cava is exciting for both scientific and commercial reasons. Cava are the most obvious signal of a successful seeding event, visible to the naked eye and even from space. By gathering instances of drone-generated cava in diverse conditions, Rainmaker opens up new possibilities for research on cavum development, persistence, and relationship to precipitation. This could help us address larger questions in cloud microphysics. From a commercial standpoint, this development represents a novel application for cloud seeding: the ability to clear overcast skies in thin, supercooled stratiform regimes. This could support solar farms and other facilities dependent on clear skies.

Background: Cloud Seeding & Cloud Microphysics

Rainmaker uses glaciogenic cloud seeding to enhance precipitation from existing clouds. It targets cold clouds containing supercooled liquid water (SLW), or liquid water droplets below 0ºC that have not yet frozen. Such clouds require ice-nucleating particles (INPs) to precipitate efficiently, and INPs can be scarce in nature.

To enhance precipitation, Rainmaker uses custom-built drones to release silver iodide (AgI)—a potent INP—into cloud layers where SLW is present. In the right conditions, this triggers ice formation. The ice crystals grow by taking up surrounding water vapor and supercooled droplets or by combining with other ice crystals into larger snowflakes. When sufficiently large, these particles fall out of the cloud as precipitation. These processes, from ice formation through precipitation growth and fallout, have been documented in airborne cloud-seeding experiments (e.g., French et al. 2018).

These cloud seeding processes lead to changes in cloud microphysics and bulk properties that can be seen from space (Figure 1). Cloud seeding directly converts the SLW layers commonly found on top of mixed-phase clouds from liquid to ice, phase differences that carry corresponding infrared signatures in satellite imagery. When ice crystals fall out of a cloud as precipitation, absorbing cloud water along the way, they deplete the cloud of its pre-existing water that defined its volume and density. The result is a deep pit or trench, depending on the seeding strategy, along with microphysical changes of cloud-top phase that can be observed with remote sensing.

Cloud Thinning & Cava in Historic Operations

Seeding-induced cloud thinning has been documented since the earliest cloud seeding programs. Project Cirrus cleared a roughly 20-mile rectangular lane through a stratus deck over New York State in the late 1940s (Figure 3; Havens 1978), and AgI flares dropped along parallel flight lines opened holes in clouds over a 30–50 minute period at Elmendorf Air Force Base, Alaska in 1969 (Accola 1969). Modern satellite-based studies have also observed this response. In a seeding campaign in central China, glaciated cloud-top sinking became discernible on satellite approximately 22 minutes after seeding, descending at a rate consistent with ice-crystal fallout (Rosenfeld et al. 2005; Yu et al. 2005). Li et al. (2022) measured channel depths of 500–2,200 m across a seeded track. Peng et al. (2025) flew research aircraft directly through a seeded cloud trough, finding onset approximately one hour post-seeding, persistence of roughly three hours, and horizontal extents of 30–40 km.

When ice growth and fallout sufficiently deplete a supercooled liquid-cloud layer, a visible opening known as a cavum can form. The World Meteorological Organization defines cavum (plural: cava) as a well-defined, generally circular hole in a thin cloud layer containing supercooled liquid water droplets (WMO 2017). Often called hole-punch clouds, these features commonly occur when rapid expansion and cooling of air around aircraft initiate ice formation (Heymsfield et al. 2011). The Swiss CLOUDLAB project has documented their formation through deliberate drone-based glaciogenic seeding, providing direct observations of the underlying microphysical response (Omanovic et al. 2026). Three of 78 experiments produced visible holes, all during the same morning in a thin supercooled cloud layer. Measurements from instruments on a tethered balloon showed liquid-water depletion coinciding with the appearance of ice crystals. This work provides a foundation for investigating how seeding-induced cava develops under different cloud conditions and how their evolution relates to precipitation.

Cavum Identification in Rainmaker’s Field Campaigns

Cloud top deformation and cava have been confirmed in several Rainmaker field campaigns. Here, we present two examples: Intensive Operating Period A (“IOP A”), which provides a particularly clear view of cavum development within a relatively uniform stratiform cloud deck, and Intensive Operating Period A (“IOP B”), which shows larger cava within a more variable cloud field. Out of respect for Rainmaker’s local partners, we have anonymized the dates and locations of these operations. Ceilometer observations provide context for the cloud-base conditions and their evolution throughout each IOP (Figure 5).

Satellite observations can be used to identify cloud thinning and cava. Changes in shortwave-infrared reflectance and retrieved cloud properties provide evidence of changes in cloud phase and particle size (Rosenfeld et al. 2005; Li et al. 2022; Yan et al. 2024). We focus on two channels from the GOES-18 Advanced Baseline Imager (ABI): Band 2 visible reflectance (0.64 µm), which reveals cloud texture and the structure of openings during daylight, and Band 13 infrared brightness temperature (10.3 µm), which allows cloud features to be followed during both day and night. Localized increases in temperature can be consistent with cloud thinning or cava, although its interpretation also depends on overlying clouds and precipitation within the opening. We therefore interpret these satellite features alongside radar and ground observations.

Our attribution of a cloud response to our seeding efforts begins with radar observations of precipitation enhancements, evaluated based on timing and proximity to silver iodide release, observed wind conditions, and subsequent evolution of enhanced reflectivity echoes. Seeding signatures must exhibit coherent downwind motion and vertical development consistent with ice-particle growth and fallout; repeatability across successive releases provides further confidence. Satellite observations are then compared with radar-identified signatures to determine whether cloud thinning or clearing occurs in the same region and evolves alongside the precipitation response.

In IOP A, ABI Band 2 visible imagery captures the lifecycle of the first cloud depression and the beginning of a second before sunset (Figure 6). The relatively uniform stratiform cloud deck provides a clear background against which their development, expansion, and subsequent weakening can be followed. Both features coincide with a localized warming in ABI Band 13, consistent with cloud thinning or a depression in the cloud top. After sunset, two additional infrared signatures develop and move downwind in a similar manner. Their repeated appearance downstream of the seeding site and resemblance to earlier features support their interpretation as additional cavum-associated cloud deformation.

During IOP A, the satellite features are collocated with radar-identified precipitation signatures, assessed using the attribution framework described above. Figure 7 shows the associated radar reflectivity and quantitative precipitation estimates. These observations provide evidence of precipitation development and its vertical extent, complementing the satellite-observed cloud deformation. This distinction matters because a visible opening alone does not establish that precipitation reaches the surface: falling ice can sublimate, or melt and subsequently evaporate, before reaching the ground.

Photographs taken by surface operators during the first cavum provide a complementary view of the opening (Figure 8). The photograph as well as collocated radar echoes support the presence of precipitation within the cavum. This helps explain why the satellite brightness temperatures within the feature are only approximately 1–2 K warmer than the surrounding cloud. An unobscured opening large enough to fill an infrared pixel would be expected to produce a brightness temperature closer to that of the surface under clear skies. However, removal of the liquid-cloud layer does not necessarily leave an unobstructed path to the ground. Falling ice within the cavum can absorb and emit infrared radiation, limiting the satellite’s view of the warmer surface below. Contributions from surrounding clouds within partially filled pixels can further reduce the apparent warming, particularly near the irregular edges of the opening. Thus, the modest infrared warming does not rule out clearing of the liquid-cloud layer, as falling ice and surrounding clouds can continue to influence the satellite signal.

IOP B illustrates cavum development within a more complex cloud environment. Unlike IOP A, the stratiform deck has pronounced undulations and variations in cloud texture, while passing higher-level clouds intermittently obscure the underlying cloud deck. This makes it more difficult to identify individual depressions in a single image, increasing the necessity of following their evolution through successive frames. Despite this variable background, large depressions remain identifiable at first only through Band 13 with warmer infrared brightness temperatures, but as the sun rises, they are also identifiable through visible imagery. These cava extend over a larger area than those observed during IOP A, while retaining more irregular boundaries and internal cloud structure.

During the unobscured portions of the sequence, the visible depressions coincide with warmer infrared features. A representative frame from the animation highlights the complexity of these features (Figure 9). The cavum is most apparent along its edges, where distinct boundaries separate the depression from the surrounding textured cloud deck. A narrow band of higher clouds crosses the region, partially obscuring the underlying cava and interrupting their warmer infrared signatures. The central portion of the highlighted cavum also appears more filled in than its edges.

This may reflect falling crystals remaining within the opening, regeneration of cloud liquid, or a combination of both—the imagery alone cannot distinguish between these possibilities. Following successive frames helps trace the cavum despite these changes in its appearance. Its collocation with radar-identified seeding signatures and associated precipitation again gives us confidence that the feature is related to seeding efforts.

Together, the two IOPs demonstrate how complementary observations allow for identification of cavum development across both relatively uniform and more complicated stratiform cloud conditions.

Cavum Evolution & Life Cycle

Beyond identifying cavum occurrence, these observations allow us to examine how individual features expand, persist, and become less distinct over time. To examine the evolution of an individual cavum, we tracked the clearest case (first cavum from IOP A) through satellite imagery at five minute intervals over 100 minutes (Figure 10). Two intersecting transects, A–B and a perpendicular line through its midpoint, capture changes in the cavum’s structure using GOES 18 ABI Band 2 visible reflectance and Band 13 infrared brightness temperature. These observations were compared with a local cloud reference derived from the two flanking regions shown in cyan, updated at each time to account for changes in the surrounding cloud. Visible reflectance was corrected for solar illumination by dividing the ABI Band 2 reflectance factor by the cosine of the solar zenith angle (Schmit and Gunshor 2021). 

The clearest combined visible and infrared response occurred approximately 20–40 minutes into the tracked sequence. At 25 minutes, localized visible reflectance was approximately 36% lower than the reference cloud along both transects, accompanied by infrared brightness temperatures approximately 0.9 K warmer. The thermal anomaly peaked at approximately 1 K at 35 minutes. The time–distance diagrams show the anomalies expanding along both transects over roughly the first 45 minutes. Figure 10 also highlights their asymmetry: the negative reflectance anomaly extended approximately 8 km along A–B and 4–6 km along the perpendicular transect.

The central visible and infrared anomalies weakened over the following 20–30 minutes as the tracked region became less distinct from the surrounding cloud. This pattern is consistent with localized cloud thinning or clearing followed by recovery or regeneration. Towards the end of the cavum detectability, the strongest depression is biased towards the A side of the transect, with a small region of negative reflectance and warmer temperature anomalies still identified. However, these later observations do require additional care as the sun angle falls below 10°, and shadowing becomes more pronounced.

Conclusion

Cava form when seeded precipitation clears out a column of cloud water, producing a visually and radiometrically distinct hole in an otherwise continuous cloud deck (Omanovic et al. 2026). Rainmaker has observed cavum signatures on multiple occasions during its field campaigns, using a combination of satellite imagery, radar, and direct visual confirmation from field personnel. Because cavum formation depends on favorable cloud structure and is difficult to confirm without daytime satellite views, our current count likely underrepresents the true rate of occurrence.

We have highlighted two IOPs containing multiple cavum signatures, with at least one daytime signature per IOP to provide satellite attribution across all relevant channels. We also provide a preliminary analysis of the formation and dissipation of one of the clearest cavum signatures. Tracking this feature with GOES-18 ABI imagery at five-minute intervals over 100 minutes, we find visible reflectance reaching approximately 36% lower than the surrounding cloud reference and infrared brightness temperatures approximately 1 K warmer, with the strongest response 20–40 minutes after first detection. The feature expanded asymmetrically to roughly 8 km along its major axis before weakening over the following 20–30 minutes as the region became less distinct from the surrounding cloud.

As we gather more observations of cava in our field campaigns, we can learn more about these phenomena and their relationship to precipitation efficiency. We are also investigating opportunities that cava may provide for Rainmaker operations unrelated to precipitation generation, such as clearing clouds above solar cells or improving visibility for aircraft. Finally, we are excited to use these cava observations to contribute to broader cloud physics research problems. Rainmaker’s seeding operations provide a rare opportunity to conduct control/variable experiments on the natural atmosphere. As we conduct more research, we can potentially address more fundamental scientific questions related to cloud evolution and dissipation.