Parametrizing cloud geometry and its application in a subgrid cloud‐edge erosion scheme

Abstract

To represent the effects of unresolved cloud processes in numerical weather prediction and climate models, parametrizations of the subgrid properties of clouds are required. In this paper, we describe a method for specifying the “cloud‐edge length” within a model grid‐box, which is an important parameter for approximating the subgrid mixing of air at cloud boundaries. We begin by proposing three conceptual models that predict the cloud‐edge length using the grid‐box cloud fraction and a length‐scale to be derived empirically. The conceptual models are then evaluated using a wide range of observations and cloud‐resolving models. Based on the finding that the “effective cloud spacing” approach fits both these data best, we parametrize the effective cloud spacing as a function of pressure and model resolution. An application of this parametrization to the cloud erosion scheme in the ECMWF forecast model is then demonstrated. The effective cloud spacing approach is compared to the “effective cloud scale” approach and is shown to increase cloud fraction in stratocumulus regions, while decreasing cloud fraction in cumulus regions. These cloud changes have the overall effect of decreasing the error of the modelled top‐of‐atmosphere net short‐wave irradiance when compared to CERES observations by around 3%. Additionally, the cloud‐edge length is an important parameter for approximating subgrid radiative transfer and it is hoped that this parametrization will be useful to quantify the effect of representing 3D cloud radiative transfer in global models.

Process‐Based Climate Model Development Harnessing Machine Learning: III. The Representation of Cumulus Geometry and Their 3D Radiative Effects

“…Vertical profiles of diagnosed C s are illustrated in Figure 1c. C s ranges from 50 to 600 m with some variability both in height and between the different cloud fields, with an average value of 249 m. They are slightly smaller than those found by Hogan et al (2016) and Fielding et al (2020) in the I3RC LES cumulus cloud field of Hinkelman et al (2005). Their simulation is also based on the ARMCu case, with the same forcings and domain size, but their larger resolution of ( 67) 2 × 40 m explains the differences.…”

“…This would mean that a sub-parameterization should be developed to make this parameter depend on atmospheric conditions. Such parameterizations exist for example to predict cloud perimeter length in Fielding et al (2020), or the degree of overlap in for example, Sulak et al (2020). New parameters appear in these formulations, which can in turn be calibrated using the same procedure as described in this work.…”

Performance of Cloud 3D Solvers in Ice Cloud Shortwave Radiation Closure Over the Equatorial Western Pacific Ocean

“…The lateral exchange rate in Hogan and Shonk (2013) is assumed to be proportional to the normalized cloud perimeter length ( L c ; unit: m −1 ), which is defined as the ratio of the total cloud edge length in the horizontal plane in a model layer to the area of the selected domain. Some empirical approaches to estimate L c have been suggested (Fielding et al., 2020; Hogan et al., 2018; Schäfer et al., 2016). The method developed in Hogan and Shonk (2013) was later called the SPeedy Algorithm for Radiative TrAnsfer through CloUd Sides (SPARTACUS), after some further improvements were made, including (a) extending the method to include SW and LW (Schäfer et al., 2016); and (b) facilitating horizontal radiation exchange calculations in the Tripleclouds framework (Hogan et al., 2016).…”

Process-based climate model development harnessing machine learning: III. The Representation of Cumulus Geometry and their 3D Radiative Effects

“…Vertical profiles of diagnosed C s are illustrated inFigure1c. C s ranges from 50 to 600 meters with some variability both in height and between the different cloud fields, with an average value of 249 m. They are slightly smaller than those found byHogan et al (2016) andFielding et al (2020) in the I3RC LES cumulus cloud field ofHinkelman et al (2005). Their simulation is also based on the ARMCu case but their larger resolution of (67 m) 2 × 40 m explains the differences.…”

“…Given observed values for c and p, C s is the size of the cloud that is such that when a virtual layer is filled randomly with instances of this cloud until the cloud fraction of the layer is c, then the total cloud perimeter in the virtual scene is p. Other choices of representation for p have been explored, in particular the recent work of Fielding et al (2020) has led to a new parameterization for the cloud perimeter in SPARTACUS.…”