Cloud observations from land stations and from ships in the ocean are used to investigate the frequency of observation and the co-occurrence of different cloud types, and the geographical and seasonal variations of these co-occurrences. Ground-based observations are used because they provide a more definitive identification of clouds by type than do satellite observations. The clouds are grouped into six types (cirrus + cirrostratus + cirrocumulus, altostratus + altocumulus, stratus + stratocumulus, nimbostratus, cumulus, and cumulonimbus). The results are expressed as frequency of occurrence of different cloud types and as contingency probabilities; that is, given that one cloud type is present, the probability that another particular type is also present is computed. Several sources of bias are identified, and their effects on the results are estimated. It is found that, on the average at all latitudes and in all seasons, clear skies occur more frequently, by a factor of about 4, over land than... Abstract Cloud observations from land stations and from ships in the ocean are used to investigate the frequency of observation and the co-occurrence of different cloud types, and the geographical and seasonal variations of these co-occurrences. Ground-based observations are used because they provide a more definitive identification of clouds by type than do satellite observations. The clouds are grouped into six types (cirrus + cirrostratus + cirrocumulus, altostratus + altocumulus, stratus + stratocumulus, nimbostratus, cumulus, and cumulonimbus). The results are expressed as frequency of occurrence of different cloud types and as contingency probabilities; that is, given that one cloud type is present, the probability that another particular type is also present is computed. Several sources of bias are identified, and their effects on the results are estimated. It is found that, on the average at all latitudes and in all seasons, clear skies occur more frequently, by a factor of about 4, over land than...