Three-Dimensional Radiative Transfer - Do We Need it and Can We Afford it? (Invited Presentation)

Three-dimensional radiative transfer has been studied since the 1970s, and with increasing computational power and memory capacity, applications became more and more sophisticated. Radiative transfer has been traditionally done in one-dimensional approximation, since it is generally assumed that atmospheric optical properties vary faster in the vertical than in the horizontal dimensions. Clouds, however, are highly variable in all spatial dimensions. Yet, for the calculation of radiative transfer in climate models, numerical weather models, cloud-resolving models, and remote sensing algorithms clouds are generally considered horizontally homogeneous or plane-parallel. Many studies have shown that the neglect of horizontal inhomogeneities may cause random errors or biases in atmospheric heating rates, irradiances, or radiances. The huge computational cost of three-dimensional models, however, prevents their direct application in atmospheric models and remote sensing algorithms. Fast three-dimensional radiation parameterizations are needed for such applications. The talk will cover emerging accurate yet fast three-dimensional approximations for cloud-resolving models and remote sensing algorithms. Clear indications of three-dimensional radiative transfer effects on cloud formation have been found, e.g. radiation-driven formation of cloud streets or organization of convective clouds.