Condensate variability - mining the ARM archive

 

Authors

Maike Ahlgrimm — Deutscher Wetterdienst
Richard M Forbes — European Centre for Medium-Range Weather Forecasts

Category

General Topics – Cloud

Description

The grid box size of global models remains large compared to the size of individual cloud features. Condensate mass is a typical prognostic variable in these models, yet radiative transfer calculations and microphysical processes depend on local condensate amounts rather than the grid box mean values typically forecast. The ARM archive with its long-term observational record at sites in different climatic regions contains a wealth of information that can be used to inform parametric descriptions of condensate variability. Yet, condensate retrievals are not uniform across sites and time periods, but rather form a patchwork. To take advantage of the full length of the record, we need to gain confidence that variability observed at one site with a particular retrieval method is comparable to variability at another site, derived with a different algorithm. To establish how retrieval methods compare, the condensate variability is characterised in the form of the Fractional Standard Deviation (FSD) from several products and compared where products overlap. To ensure that different retrieval algorithms observe substantially the same cloud scene, the ARM Best Estimate product is used to classify scenes based on cloud boundaries, presence of precipitation and water path. As might be expected, retrievals are more often successful in the absence of surface precipitation, and tenuous clouds with low water path are more likely to be missed. For example, boundary layer clouds with a liquid water path exceeding 50 g m-2 prove to be reliably retrieved by most algorithms, lending confidence to a comparison of FSD for this cloud type. For deeper precipitating clouds, differences in retrieval success can be large. Apart from questions on retrieval quality in precipitation-affected scenes, retrievals are likely performed on substantially different cloud scenes, making it difficult to draw reliable conclusions from a comparison. With consistency checks in place, robust variations in FSD become apparent. For example, the Mace, Microbase and Cloudnet retrievals all indicate that condensate variability in boundary layer clouds at Darwin is greater than that of similar clouds observed at SGP. Having established which retrievals supply similar information for certain cloud types will allow a wider use of available observations. Early results indicate that parametric descriptions of FSD from literature may not fully capture regional differences in condensate variability.