The Role of Experimental and Statistical Uncertainty in Interpretation of Immersion Freezing: A Case for Classical Nucleation Theory

Ice nucleation is the initial step in forming mixed-phase and cirrus clouds, and is well established as an important influence on global climate. Laboratory studies investigate at which cloud relevant conditions of temperature (T) and relative humidity (RH) ice nucleation occurs and as a result, numerous fundamentally different ice nucleation descriptions have been proposed for implementation in cloud and climate models. We introduce a new immersion freezing model based on first principles of statistics to simulate individual droplet freezing requiring only three experimental parameters, which are the total number of droplets, the uncertainty of applied surface area per droplet, and the heterogeneous ice nucleation rate coefficient, Jhet, as a function as a function of T and water activity (aw), where in equilibrium RH=aw. Previous studies reporting frozen fractions (f) or Jhet for a droplet population are described by our model for mineral, inorganic, organic, and biological ice nuclei and different techniques including cold stage, oil-immersion, continuous flow diffusion chamber, flow tube, cloud chamber, acoustic levitation and wind levitation experiments. Taking advantage of the physically based parameterization of Jhet by Knopf and Alpert (Faraday Discuss., 165, 513-534, 2013), our model can predict immersion freezing for the entire atmospherically relevant range of T, RH, particle surface area, and time scales, even for conditions unattainable in a laboratory setting. Lastly, we present a rigorous experimental uncertainty analysis using a Monte Carlo method of laboratory derived Jhet and f. These results imply that classical nucleation theory is universal for immersion freezing. In combination with a aw based description of Jhet, this approach allows for a physically based and computational little demanding implementation in climate and cloud models.