Influence of 3D thermal radiation on cloud development

This thesis aims to answer the question if 3D effects of thermal
radiative transfer need to be considered in cloud resolving
simulations and if an influence of 3D thermal heating and cooling
rates exists in contrast to common 1D approximations. To study this
question with the help of a cloud resolving model, an accurate, yet
fast parameterization of 3D radiative transfer is needed. First, an
accurate 3D Monte Carlo model was developed which was used as
benchmark for developing the fast `Neighboring Column
approximation' (NCA), which was then coupled to the UCLA-LES to
study the effects of 3D thermal heating and cooling rates in
comparison to common 1D radiative transfer approximations. First,
differences between common 1D radiative transfer approximations and
a correct 3D radiative transfer model were analyzed. For this,
efficient Monte Carlo variance reduction methods have been
developed and implemented in MYSTIC, a Monte Carlo radiative
transfer model. The dependence of 1D and 3D heating and cooling
rates on cloud geometry has been investigated by analyzing
idealized clouds such as cubes or half spheres. Further more, 1D
and 3D heating and cooling rates in realistic cloud fields were
simulated and compared. It could be shown that cooling rates reach
maximum values of several 100 K/d at cloud tops if the model
resolution was between 50 m to 200 m. Additional cloud side cooling
of several 10 to 100 K/d was found in 3D heating and cooling rate
simulations. At the cloud bottom, modest warming of a few 10 K/d
occurs. Heating and cooling rates depend on the vertical location
of the cloud in the atmosphere, the liquid water content of the
cloud, the shape of the cloud and the geometry of the cloud field
(for example the distance between clouds). Based on the results of
a detailed analysis of exact simulations of 3D thermal heating and
cooling rates, a fast, but still accurate 3D parameterization for
thermal heating and cooling rates has been developed. This
parameterization, the `Neighboring Column Approximation' (NCA), is
based on a 1D radiative transfer solution and uses the next
neighboring columns of a column to estimate the 3D heating or
cooling rate. The method can be used in parallelized models. With
the NCA, it is possible to simulate 3D cloud side cooling and
warming. It was shown that the NCA is a factor of 1.5 to 2 more
expensive in terms of computational time when used in a cloud
resolving model, compared to a 1D radiative transfer approximation.
The NCA was implemented in UCLA-LES, a cloud resolving, large-eddy
simulation model. With the UCLA-LES and the NCA it was possible for
the first time to study the effects of 3D interactive thermal
radiation on cloud development. Simulations without radiation, with
1D thermal radiation and 3D thermal (NCA) radiation have been
performed and differences have been analyzed. First, single,
isolated clouds were investigated. Depending on the cloud shape, 3D
thermal radiation changes cloud development in comparison to 1D
thermal radiation. Overall it could be shown that a thermal
radiation effect on cloud development exists in general. Whether
there is a differences between 1D and 3D thermal radiation on cloud
development seems to depend on the specific situation. One of the
main features of thermal radiation affecting a single cloud is a
change in the cloud circulation. Stronger updrafts in the cloud
core and stronger downdrafts at the cloud sides were found, causing
an enhanced cloud development at first, but a faster decay of the
cloud in the end. Second, large scale simulations of a shallow
cumulus cloud field in a 25 x 25 km^2 domain with 100~m horizontal
resolution were analyzed. To the authors knowledge, this is the
first time that a cloud field of this size and resolution was
simulated including 3D interactive thermal radiation. It was shown
that on average, updrafts, downdrafts and liquid water increases if
thermal radiation is accounted for. While most variables (for
example liquid water mixing ratio or cloud cover) did not show
significant systematic difference between no-radiation simulation
and the simulations with 1D and 3D thermal radiation, the cloud
size (or horizontal extent) was larger in the simulations with
interactive 3D thermal radiation. Convective organization set in
after a few hours already. This is a clear indication that 3D
thermal radiation could trigger convective organization.