Abstract

The interaction between radiation and clouds represents a persistent source of uncertainty in numerical weather and climate prediction. Clouds are inherently complex meteorological phenomena, appearing in an immense variety of geometrical shapes and exhibiting highly variable degrees of heterogeneity. A physically consistent and computationally efficient coupling of three-dimensional cloud structures with the solar and thermal radiative field thereby remains one of the greatest challenges in the atmospheric science community. The present thesis aims to make progress towards an improved treatment of the unresolved cloud-radiation interchange for both regional and global modeling applications. The first dissertation objective is to quantify the radiative bias in regional models for a realistically evolving shallow cumulus cloud field. The bias dependence on various input parameters of radiation schemes such as solar zenith angle, surface albedo, cloud cover and liquid water path is examined. Nighttime and daytime biases within the cloud-layer and at the surface are thoroughly investigated and evaluated against a high-resolution three-dimensional benchmark computation. The focus is laid on quantifying the regional-scale model bias arising from two chief shortcomings. First, the poor representation of unresolved cloudiness, which is normally approximated as a series of horizontally homogeneous partially cloudy layers. Second, the intrinsic constraint of one-dimensional radiation schemes, employing merely two streams for capturing the upward and downward radiative flux, but entirely neglecting the grid- and subgrid-scale horizontal photon flow. Since it is unclear which error source is dominant at the scale of regional modeling where these multiple issues intersect, the bias stemming from the latter drawback is simultaneously assessed. The principal findings highlight the importance of an improved cloud representation even at the regional scale. The second dissertation objective is to advance the cloud-radiation interaction parameterization in coarse-resolution global models, focusing on the issues related to misrepresentation of cloud horizontal inhomogeneity. This subject is tackled with the state-of-the-art Tripleclouds radiative solver, the fundamental feature of which is the inclusion of the optically thicker and thinner cloud fraction. The research challenge is to optimally set the pair of cloud condensates characterizing the two cloudy regions and the corresponding geometrical split of layer cloudiness. A diverse cloud field data set was collected for the analysis, comprising case studies of cumulus, stratocumulus, cirrus and cumulonimbus. The primary goal is to test the validity of global cloud variability estimate along with various condensate distribution assumptions. More sophisticated parameterizations are subsequently explored, optimizing the treatment of overcast as well as extremely heterogeneous cloudiness. The radiative diagnostics including atmospheric heating rate and net surface flux are for the first time consistently studied using the Tripleclouds method. The performance of Tripleclouds mostly significantly surpasses the conventional calculation on horizontally homogeneous cloudiness. The effect of horizontal photon transport is further quantified. The overall conclusions are intrinsically different for each particular cloud type examined, encouraging endeavors to enhance the use of cloud regime dependent methodologies in next-generation atmospheric models. The major technical effort undertaken within the scope of this work was the design of the classic two-stream radiation scheme supporting homogeneous partial cloudiness and its subsequent extension to incorporate the Tripleclouds concept. Both algorithms were implemented in the libRadtran radiative library, promoted to be utilized for further unraveling of key scientific mysteries related to cloud-radiation interplay.