DeutschClear Cookie - decide language by browser settings
Jakub, Fabian (2016): On the impact of three dimensional radiative transfer on cloud evolution. Dissertation, LMU München: Faculty of Physics



The goal of this study is to gain insight into cloud-radiative feedback mechanisms and what role three-dimensional radiative transfer effects play in the evolution of convective clouds. The usually employed one-dimensional radiative transfer solvers neglect any horizontal energy transfer and thereby introduce considerable errors in surface and atmospheric heating rates. While fully three-dimensional radiative transfer solvers exist, they are several orders of magnitude too slow. In conclusion, so far, there is no straightforward solution that would solve the task at hand — namely, compute accurate three-dimensional radiative heating rates in the atmosphere — fast enough to be coupled interactively to a cloud resolving model. This thesis presents a new method — the TenStream solver — that provides a fast yet accurate approximation for three-dimensional heating rates. The TenStream is furthermore integrated into the University of California, Los Angeles large-eddy simulation (UCLA-LES) cloud-resolving model. This setup allows to study the effects of three-dimensional radiative heating on the evolution of clouds. The TenStream method extends the well-known one-dimensional two-stream theory to 10 streams. The new solver significantly reduces the root mean square error for atmospheric heating and surface heating rates when compared to traditionally employed one-dimensional solvers. In the case of a cumulus cloud field and the solar zenith angle being 60 ◦ , the error is reduced from 178 % to 31 %. Parallel scalability was a primary concern developing the TenStream solver. This thesis documents the overall performance of the solver as well as the technical challenges of migrating from 1-D schemes to 3-D schemes. To understand the performance characteristics of the TenStream solver, weak as well as strong-scaling experiments are conducted. In this context, two matrix preconditioner are investigated: geometric algebraic multigrid preconditioning (GAMG) and block Jacobi incomplete LU (ILU) factorization and it is found that algebraic multigrid preconditioning performs well for complex scenes and highly parallelized simulations. The TenStream solver is tested on several state of the art super-computers for up to 4096 cores and shows a parallel scaling efficiency of 80 % to 90 %. The central part of this thesis examines the influence of three-dimensional radiative transfer effects on the development of convective cumulus clouds. The influence is tested on short time scales of a single convective warm-bubble and over a longer period of time and a reasonably large domain for shallow cumulus clouds. The directionality of the direct solar beam introduces an asymmetry in the atmospheric heating of the convective motion and tilts the updraft. While a cloud’s shadow is always directly beneath itself in a one-dimensional radiative transfer solver. In contrast, the TenStream solver correctly displaces the shadowy region according to the sun’s zenith angle. The constant supply of warm and moist air due to the local heating in the updraft region beneath the cloud, prolongs the cloud’s lifetime by a factor of two and generally increases cloud development. The influence of three-dimensional heating on the evolution of clouds shows to be persistent even in the presence of a horizontal wind. The results presented here motivate further research in the field of cloud- radiative feedbacks and their role in weather and climate prediction simulations.