Atmospheric immersion freezing (IF), a heterogeneous ice nucleation process where an ice nucleating particle (INP) is immersed in supercooled water, is a dominant ice formation pathway impacting the hydrological cycle and climate. Implementation of IF derived from field and laboratory data in cloud and climate models is difficult due to the high variability in spatio-temporal scales, INP composition, and morphological complexity. We demonstrate that IF can be consistently described by a stochastic nucleation process accounting for uncertainties in the INP surface area. This approach accounts for time-dependent freezing, a wide range of surface areas and challenges phenomenological descriptions typically used to interpret IF. The results have an immediate impact on the current description, interpretation, and experiments of IF and its implementation in models. The findings are in accord with nucleation theory, and thus should hold for any supercooled liquid material that nucleates in contact with a substrate.

大气浸没冷冻（IF）是一种非均质的冰核形成过程，其中，冰核颗粒（INP）浸入过冷水中，是影响水文循环和气候的主要成冰途径。由于时空尺度，INP组成和形态复杂性的高变异性，很难在云和气候模型中实施从野外和实验室数据得出的IF。我们证明，IF可以由考虑INP表面积不确定性的随机成核过程一致地描述。这种方法会导致时间依赖性冻结，大范围的表面积，并挑战通常用于解释IF的现象学描述。结果对当前的描述，解释，IF的实验及其在模型中的实现。该发现与成核理论相符，因此对于任何与基体接触成核的过冷液体材料都应成立。