Long-term variations in the occurrence and effective solar transmission of clouds as determined from surface-based total irradiance observations

Time series of cloud solar transmission and cloud occurrence frequency are developed for the past 27 years at four globally remote and climatically diverse surface locations. A new methodology is developed that objectively segregates times of cloud‐free conditions from those times when clouds are detected in high‐time‐resolution total solar irradiance observations that are obtained from pyranometers. The methodology for cloud detection depends on the magnitude and short‐term variability of observed departures from clear‐sky conditions. Expected clear‐sky irradiances are based on interpolated clear‐sky observations. Results of the new cloud detection methodology are compared to four independent cloud detection methods. An effective cloud transmission is determined as the ratio of observed irradiance in the presence of clouds to that expected in the absence of clouds. Selective forward scattering by clouds toward the observation site results in computed effective transmissions frequently being >1.0. It is shown that conditional temporal averaging of effective cloud transmission over periods of three days or more virtually eliminates the unrealistic cloud transmissions exceeding 1.0. Such temporal averaging of the surface measurements is advantageous for comparing against other area‐wide cloud transmission estimates, such as those determined from satellite or by numerical climate models. The cloud occurrence frequency and the effective solar transmission for long‐term observational records are summarized into monthly and annual averages, and their long‐term variability is investigated. Temporal variations in frequency distributions of transmission are used to determine which clouds are responsible for changes in mean cloudiness. A statistically significant upward trend in cloud occurrence frequency, from 76 to 82% between 1976 and 2001, is detected at Barrow, Alaska, where clouds having solar transmission near 0.2 exhibit the largest increase. At the South Pole, decadal timescale oscillations in both cloud characteristics are detected, but no particular cloud category is identified as the source of that oscillation.