ALBERT

Beschreibung: Cloud optical depth remains a difficult variable to represent in climate models, and hence there is a need for high-quality observations of cloud optical depth from locations around the world. Such observations could be readily obtained from Aerosol Robotic Network (AERONET) radiometers using a two-wavelength retrieval method. However, the method requires an assumption that all of the cloud in a profile is liquid, and this has the potential to introduce errors into long-term statistics of retrieved optical depth. Using a set of idealised cloud profiles, we find that the fractional error in retrieved optical depth is a linear function of the fraction of the optical depth that is due to the presence of ice cloud (“ice fraction”), with a magnitude of order 55 % to 70 % for clouds that are entirely ice. We derive a simple linear equation that could potentially be used as a correction at AERONET sites where ice fraction can be independently estimated. The greatest contribution to error statistics arises from optically thick cloud that is either mostly or entirely ice. Using this linear equation, we estimate the magnitude of the error for a set of cloud profiles measured at five sites of the Atmospheric Radiation Measurement programme. Instances of such clouds are not frequent, with less than 15 % of cloud profiles at each location showing an error of greater than 10. However, differences in the frequency of such clouds from one location to another affect the magnitude of the overall mean error, with sites dominated by deep tropical convection and thick frontal mixed-phase cloud showing greater errors than sites where deep clouds are less frequent. The mean optical depth error at the five locations spans the range 2.5 to 5.5, which we show to be small enough to allow calculation of top-of-atmosphere flux to within 10 %, and surface flux to about 15 %.