Evaluating and Improving a PDF Cloud Scheme Using High‐Resolution Super Large Domain Simulations

Abstract

AbstractCloud schemes, which use probability density functions (PDFs) to describe the subgrid‐scale variability of thermodynamic state variables, rely heavily on large eddy simulation (LES) data to provide information on the distributions of state variables in a turbulent domain. Until recently, LES simulations were commonly limited to small domains and highly idealized prototype regimes, but super large domain realistically forced simulations are now available, which contain previously excluded synoptic scale variability and orographic effects. Thanks to these advances PDF cloud schemes can be confronted with much more realistic distributions of thermodynamic states and clouds on the horizontal scales they were developed for. One example of super large domain LES is a library of recent Germany‐wide simulations with a 100‐m resolution. We use these simulations to study how a PDF cloud scheme, which represents variability through a beta function of total water, represents cloud fraction and the total and liquid water distribution. Special emphasis is placed on studying how various closures, which limit the possible shape of the beta function, affect the resulting cloud fraction. Our analysis shows that the PDF cloud scheme leads to much smaller cloud fraction errors than a classic parameterization that relates cloud fraction to relative humidity and that the PDF scheme is suitable for model resolutions from 5 to 100 km. The PDF cloud scheme performance is increased by introducing a zero‐buoyancy relationship between water vapor and temperature to approximate the orientation of their joint PDF.