A new algorithm for the downscaling of cloud fields

Abstract

We present a novel algorithm for the downscaling of three-dimensional cloud fields. The goal of the algorithm is to add realistic subscale variability to a coarse field taking the resolved variability into account. The method is tested by coarse graining high-resolution sparse cumulus and broken stratocumulus clouds in the horizontal plane, downscaling these coarse fields back to the high resolution and comparing the radiative and microphysical properties of these downscaled fields with the original high-resolution fields. The resolutions of the cumulus and stratocumulus clouds used for this purpose are increased by a factor of four and ten, respectively. The downscaling decreases the errors in the flux transmittance and reflectance of the cumulus and stratocumulus cloud fields by at least a factor of ten and three, respectively, compared to utilising the coarse cloud fields. A novel aspect of our algorithm is the fact that it constrains the high-resolution fields of cloud liquid water content as well as the subscale cloud fraction. An alternative version that does not include cloud fraction information is less accurate, but still significantly better than using the coarse fields. The latter downscaling algorithm can also be utilised for the disaggregation of geophysical fields for which fractional coverages are not defined. Furthermore, the downscaling algorithm can be combined with our other algorithms to generate surrogate fields with other constraints, for example, surrogate clouds with a prescribed liquid water content height distribution.

Downscaling of METEOSAT SEVIRI 0.6 and 0.8 μm channel radiances utilizing the high-resolution visible channel

“…Moreover, Deneke et al (2009b) showed that biases in satellite-estimated cloud climatologies can also be reduced by using estimates of the unresolved variance. Alternatively or complementary, these biases may also be reduced by using approaches that simulate realistic surrogate variability at sub-pixel scale (see e.g., Venema et al, 2010;Schutgens and Roebeling, 2009).…”

Multi-sensor cloud retrieval simulator and remote sensing from model parameters – Part 1: Synthetic sensor radiance formulation

“…This makes negligible difference to the ICA-calculated radiative properties of the subcolumn ensemble, since the subcolumns are not changed in any way, only resampled. So, like the downscaled fields of Venema et al (2010), our un-clumped ICA-generated subcolumns are still able to give vastly improved mean ID radiative properties for a subcolumn, by averaging over the non-linear radiative calculations on the subcolumn level. Clumping would make a difference if we had been using a 3-D radiation calculation, but this is currently not even close to being possible, because of computational expenses, for the applications we have in mind for our simulation tool.…”

“…Bugliaro et al (2011) discuss validation of SEVIRI retrieval algorithms in a manner similar to that of Jonkheid et al (2012), but using statistical downscaling, via a spectral extrapolation method, from a 7 km resolution central European NWP forecast. We will discuss spectral extrapolation methods later in the context of the Venema et al (2010) downscaling method. The Bugliaro et al (2012) method is similar, essentially using a k −5/3 power law to extrapolate the power spectrum of liquid water content from resolved 30 km scales down to scale of 2.33 km, comparable to the SEVIRI pixel scale while removing negative condensate amounts and using a partly randomized small-scale spectral phase to treat vertical correlations between layers.…”

“…Venema et al (2010) describe an elegant method for down-scaling coarse-resolution cloud fields (such as NWP models produce) to finer horizontal scales suitable as pixel "subcolumns" for input to 1-D or 3-D radiation calculations. Their method carefully extrapolates the spectrum of variance in the resolved cloud field down to pixel scales using various extrapolation assumptions.…”

“…Their method carefully extrapolates the spectrum of variance in the resolved cloud field down to pixel scales using various extrapolation assumptions. Venema et al (2010) do acknowledge, however, that the accuracy of their downscaling approach depends on the quality achieved by the spectral extrapolation algorithm being used, and ultimately on what assumptions are made regarding the statistical properties of the subresolved scale variance.…”