Twentieth century variability of radiative fluxes over global oceans from homogenized visually observed cloud cover

We analyze the applicability of cloud cover observations from Voluntary Observing Ships (VOS) for reconstruction of centennial time series of radiative fluxes over the global ocean. VOS provides the longest coverage of the Ocean by surface meteorological measurements. However, VOS cloud cover is characterized by space-time inhomogeneity of sampling as well as systematic biases in the first part of 20^th century implied by the changes in observational practices and coding systems in the late 1940s – early 1950s. For developing homogenized time series of cloud cover, we developed several correction algorithms accounting for the nature of biases and for inhomogeneous sampling. These algorithms are based upon (while not limited to) a critical inventory of early 20^th century reports and the developments of analytical probability distribution for fractional cloud cover.

Homogenized times series of cloud cover were used for computing short- and long-wave radiation fluxes for the period from 1900 onwards. To use instantaneous bulk parameterizations for shortwave radiation we developed a technique which allows avoiding the impact of uneven distribution of reports over daily and seasonal cycles. Also we developed a scheme for computing radiative fluxes directly from probability distributions of cloud cover. 

Finally, we obtained homogeneous global time series of radiative fluxes spanning the period from 1900 onwards with few regions being free of long period data gaps. These time series were used for estimation of linear trends. For the North Atlantic, we identified centennial increase in the cloud cover and associated declining trends in long-wave radiation with no robust basin-wide trend patterns in shortwave radiation. In the Indian monsoon region there upward trends in cloud cover result in the downward changes in shortwave radiation.