To obtain accurate estimates of surface and cloud parameters from satellite radiance data a scheme has to be devised which identifies cloud-free and cloud-filled pixels (i.e. fields of view). Such a scheme has been developed for application to high resolution (1·1 km pixel) images recorded over Western Europe and the North Atlantic by the AVHRR on the TIROS-N/NOAA polar orbiters. The scheme consists of five daytime or five night-time tests applied to each individual pixel to determine whether that pixel is cloud-free, partly cloudy or cloud-filled. The pixel is only identified as cloud-free or cloud-filled if it passes all the tests to identify that condition; otherwise it is assumed to be partly cloudy. Surface parameters (e.g. skin temperature, reflectance, vegetation index, snow cover) can then be inferred from the cloud-free radiances, and cloud parameters (e.g. cloud top temperature, optical depth and liquid water content) from the cloud-filled radiances. Only fractional cloud cover is derived from the partly cloudy pixels which, together with the number of cloud-filled pixels, gives total cloud cover over a given area. The scheme has been successfully applied to data for all seasons, including images with unusually cold or warm surface temperatures. To assess the method both daytime and night-time NOAA-9 passes over the U.K. were obtained for a week in April 1985 and some results from this data set are presented here.