Abstract
Transmissivities are determined for different cloud types using nine years of hourly irradiance measurements under overcast skies at six Canadian stations. Values for individual stations and for pooled data using irradiances uncorrected for multiple reflections are similar to values for Blue Hill, Massachusetts but 1arger than values for Hamburg, West Germany. It is argued that transmissivities used in numerical models which utilize surface observations of cloud layer amounts and types should be determined from irradiances without correction for multiple reflections. This would ensure at least partial compensation for attenuation by undetected cloud above overcast. The superior performance of transmissivities calculated in this manner is demonstrated in numerical model calculations of irradiance. It is also shown that there is no need to replace Blue Hill transmissivities with either the new values for Canada or the values proposed by Atwater and Ball for such models. There is also no indication in the Canadian results that cloud transmissivity varies with cloud amount as suggested by Atwater and Ball. Regional and seasonal variations in the Canadian transmissivities have a negligible effect on calculated irradiance. Irradiance calculations can be simplified with little loss in accuracy using an average transmissivity for each cloud layer; 78, 42 and 32% for high, middle and low cloud, respectively.
