Search for: All records

Holographic cloud probes provide unprecedented information on cloud particle density, size and position. Each laser shot captures particles within a large volume, where images can be computationally refocused to determine particle size and location. However, processing these holograms with standard methods or machine learning (ML) models requires considerable computational resources, time and occasional human intervention. ML models are trained on simulated holograms obtained from the physical model of the probe since real holograms have no absolute truth labels. Using another processing method to produce labels would be subject to errors that the ML model would subsequently inherit. Models perform well on real holograms only when image corruption is performed on the simulated images during training, thereby mimicking non-ideal conditions in the actual probe. Optimizing image corruption requires a cumbersome manual labeling effort. Here we demonstrate the application of the neural style translation approach to the simulated holograms. With a pre-trained convolutional neural network, the simulated holograms are “stylized” to resemble the real ones obtained from the probe, while at the same time preserving the simulated image “content” (e.g. the particle locations and sizes). With an ML model trained to predict particle locations and shapes on the stylized data sets, we observed comparable performance on both simulated and real holograms, obviating the need to perform manual labeling. The described approach is not specific to holograms and could be applied in other domains for capturing noise and imperfections in observational instruments to make simulated data more like real world observations.

Airborne observations in deep stratiform and anvil clouds showed extensive layers of 10‐ to 40‐kV/m electric fields colocated with highly stratified uniform radar reflectivity of 20 to 25 dBZ from 5‐ to 9‐km altitude. Size distributions with numerous small‐ and intermediate‐sized ice crystals (mostly aggregates) and large aggregates were observed in these regions. We infer that the uniform electric field, radar reflectivity, and broad particle size distributions were the result of mesoscale updrafts, confirmed by high‐resolution images of particles showing diffusional growth and no riming. No measurable supercooled liquid water was found in these regions from −10 to −45 °C. Calculated particle collision rates from observed distributions were >5 × 10³collisions per cubic meter per second in this volume. Laboratory results show that weak charge separation occurs when ice particles collide and separate even in the absence of supercooled water. We infer that charge separation occurred in the mesoscale updrafts via a noninductive mechanism in which ice particles growing by diffusion collide and transfer charge without supercooled water being present. These regions with strong, uniform fields, stratified radar reflectivity, and broad size distributions also occurred in anvils that barely reached the melting zone. Thus, we deduce that the nonriming collisional mechanism acts at middle to upper cloud levels and is not dependent upon electrification occurring near the melting zone. This mechanism should produce two oppositely charged layers of charge with the top layer residing on smaller particles often existing near the top of the cloud.

Maritime boundary‐layer clouds over the Southern Ocean (SO) have a large shortwave radiative effect. Yet, climate models have difficulties in representing these clouds and, especially, their phase in this observationally sparse region. This study aims to increase the knowledge of SO cloud phase by presenting in‐situ cloud microphysical observations from the Southern Ocean Clouds, Radiation, Aerosol, Transport Experimental Study (SOCRATES). We investigate the occurrence of ice in summertime marine stratocumulus and cumulus clouds in the temperature range between 6 and −25°C. Our observations show that in ice‐containing clouds, maximum ice number concentrations of up to several hundreds per liter were found. The observed ice crystal concentrations were on average one to two orders of magnitude higher than the simultaneously measured ice nucleating particle (INP) concentrations in the temperature range below −10°C and up to five orders of magnitude higher than estimated INP concentrations in the temperature range above −10°C. These results highlight the importance of secondary ice production (SIP) in SO summertime marine boundary‐layer clouds. Evidence for rime splintering was found in the Hallett‐Mossop (HM) temperature range but the exact SIP mechanism active at lower temperatures remains unclear. Finally, instrument simulators were used to assess simulated co‐located cloud ice concentrations and the role of modeled HM rime‐splintering. We found that CAM6 is deficient in simulating number concentrations across the HM temperature range with little sensitivity to the model HM process, which is inconsistent with the aforementioned observational evidence of highly active SIP processes in SO low‐level clouds.