Tests of Monte Carlo Indepedent Column Approximation in ECHAM5

Radiative fluxes averaged over a general circulation model (GCM) grid
cell can depend substantially on assumptions about unresolved cloud
structure (i.e., cloud features smaller than GCM grid-spacing). To
address this issue, the Monte Carlo Independent Column Approximation
(McICA) was introduced a few years ago as a new approach for
parametrizing broadband radiative fluxes in climate models. A key
feature of McICA is that it separates the description of unresolved
cloud structure from the radiative transfer solver. This allows a much
more flexible, and potentially more realistic, description of
unresolved cloud structure for radiation calculations. However, McICA
is a stochastic approach, which means that radiative flux and heating
rate results produced by McICA contain random errors ("McICA noise".)
In this seminar, I will discuss tests of McICA in ECHAM5 conducted at
the Finnish Meteorological Institute since 2005.

Firstly, I will focus on the single potential pitfall of McICA: the
possibility that McICA's unbiased random errors might have a
systematic impact on simulated climate. It turns out that in ECHAM5,
the primary impact of "McICA noise" is a slight reduction in low cloud
fraction. This feature develops very rapidly, within a couple of model
days, and was traced back to a non-linear response of precipitation
formation to radiative heating rate errors.  When the sea surface
temperature is allowed to adjust, this leads to an increase in
global-mean temperature. For a typical implementation of McICA, the
impact is, however small: roughly 0.3 K, which is similar to the
impact of increasing the cloud droplet effective radius by 0.1
microns. Therefore, McICA's random errors are not a serious concern
for ECHAM5.

Secondly, I will consider the influence of McICA on simulated climate,
as compared with the standard version of ECHAM5.  At the face value,
the main impact is that without model retuning, use of McICA makes
shortwave cloud forcing stronger, thereby increasing a model bias. The
irony here is that this is related to a major benefit of McICA. Use of
McICA makes the model more self-consistent, as it allows to use
cloud subgrid-scale variability derived directly from the output of the
Tompkins cloud scheme in ECHAM5. This, of course, eliminates
the justification for a tuning factor related to cloud inhomogeneity.
There are, however, some indications that clouds derived from the
beta distribution of total water content in the Tompkins scheme may
not possess sufficient subgrid-scale variability.

Räisänen, P., S. Järvenoja, H. Järvinen, M. Giorgetta,
  E. Roeckner, K. Jylhä and K. Ruosteenoja, 2007: Tests of Monte Carlo
  Independent Column Approximation in the ECHAM5 atmospheric GCM.
  J. Climate, 20, 4995-5011.




13:30 h


Bundesstr. 53, room 022/023
Seminar Room 022/023, Ground Floor, Bundesstrasse 53, 20146 Hamburg, Hamburg


Petri Räisänen, Finnish Meteorological Institute, FMI


Back to listing