2.4.5.4 Uncertainties in Satellite Estimates

The improvements in the retrievals and satellite instrumentation have provided valuable data to begin observation-motivated assessments of the effect of aerosols on cloud properties, even though satellite measurements cannot unambiguously distinguish natural from anthropogenic aerosols. Nevertheless, an obvious advantage of the satellite data is their global coverage, and such extensive coverage can be analysed to determine the relationships between aerosol and cloud properties at a number of locations around the globe. Using these data some studies (Sekiguchi et al., 2003; Quaas et al., 2004) indicate that the magnitude of the RF is resolution dependent, since the representation of convection and clouds in the GCMs and the simulation of updraft velocity that affects activation themselves are resolution dependent. The rather low spatial and temporal resolution of some of the satellite data sets can introduce biases by failing to distinguish aerosol species with different properties. This, together with the absence of coincident LWP measurements in several instances, handicaps the inferences from such studies, and hinders an accurate analysis and estimate of the RF. Furthermore, the ability to separate meteorological from chemical influences in satellite observations depends on the understanding of how clouds respond to meteorological conditions.

Retrievals involve a variety of assumptions that introduce uncertainties in the relationships. As mentioned above, the retrievals for aerosol and cloud properties are not coincident and the assumption is made that the aerosol optical depth can be linked to the aerosol concentration below the cloud. The POLDER instrument may underestimate the mean cloud-top droplet radius due to uncertainties in the sampling of clouds (Rosenfeld and Feingold, 2003). The retrieval of the aerosol index over land may be less reliable and lead to an underestimate of the cloud albedo effect over land. There is an indication of a systematic bias between MODIS-derived cloud droplet radius and that derived from POLDER (Breon and Doutriaux-Boucher, 2005), as well as differences in the aerosol optical depth retrieved from those instruments (Myhre et al., 2004a) that need to be resolved.