Search for: All records

Abstract During freezing rain, secondary ice produced by the fragmentation of freezing drops (FFD) can initiate a chain reaction, potentially transitioning freezing rain into ice pellets. Including this process in numerical weather prediction models is challenging due to the uncertainty of this mechanism. To bridge this gap, this study aims to evaluate the efficiency of the FFD process during ice pellet precipitation using measurements collected onboard the National Research Council Canada (NRC) Convair-580 research aircraft during the 2022 Winter Precipitation Type Research Multiscale Experiment (WINTRE-MIX). Below the supercooled raindrops freezing altitude, in situ probes measured a bimodal particle size distribution. Observations from imaging and optical-array probes show that most particles smaller than 500μm in diameter were nonspherical ice crystals, in the concentration of ∼500 L⁻¹. In contrast, most particles larger than 500μm were identified as fractured ice pellets and ice pellets with bulges, which suggested the occurrence of the FFD process. A conceptual model is then developed to show that five–eight fragments of ice were produced for each freezing drop. Two existing parameterizations of the FFD process are also tested. It is shown that one parameterization would result in less ice crystals than the measured number concentration, while the second one would result in too many ice crystals. Adjustments to these parameterizations are computed based on the collected observations. This analysis will be valuable for including the FFD process into simulations of freezing rain, ice pellets, and other weather phenomena where this process plays a significant role. Significance StatementThis study presents unique measurements from a winter storm recorded by state-of-the-art instruments onboard a research aircraft at the altitude where ice pellets are formed. The collected data suggest that the freezing of a few initial raindrops at an altitude of around 250 m above the ground resulted in the production of ice crystals. These ice crystals led to the freezing of additional raindrops in a feedback loop that can be referred to as ice multiplication. This process is quantified in the current study. The results will be valuable in improving the representation of ice pellets and freezing rain in computer simulations of winter storms. 

We propose Deep Feature Interpolation (DFI), a new data-driven baseline for automatic high-resolution image transformation. As the name suggests, it relies only on simple linear interpolation of deep convolutional features from pre-trained convnets. We show that despite its simplicity, DFI can perform high-level semantic transformations like "make older/younger", "make bespectacled", "add smile", among others, surprisingly well - sometimes even matching or outperforming the state-of-the-art. This is particularly unexpected as DFI requires no specialized network architecture or even any deep network to be trained for these tasks. DFI therefore can be used as a new baseline to evaluate more complex algorithms and provides a practical answer to the question of which image transformation tasks are still challenging in the rise of deep learning.