There is general consensus that the cloud feedback introduces the largest uncertainty in climate model projections. One way to constrain the sign and the amplitude of the feedback is to use observations to evaluate trends in cloud radiative forcing (CRF) at the top of the atmosphere (TOA) over the 20th century. However, retrievals of radiation fluxes from satellites are available only from 1980s, when the first satellites for atmospheric measurements were launched. This issue can be overcome by using surface based observations of cloud cover which extend back to 1954. We compute a regression coefficient representative of cloud effect on radiation fluxes, and then multiply this factor by the longer surface based cloud observations data set (Hahn and Warren archive) to develop a long-term radiation flux product that spans more than five decades.

However, surface based observations of clouds are also affected by spurious trends. Therefore, we analyze long-term trends in low-clouds and total clouds observations over oceanic stratus regions from the Hahn and Warren archive and compare with clouds retrievals from satellites (Isccp corrected low plus mid level clouds and Patmos-x low clouds). There are extensive regions where surface based observations are not reliable due to poor sampling, but in regions in which archives from ships provided a continuous record of observations (e.g. SE Atlantic, NE Pacific) we see a good degree of agreement with satellites data. Looking at well-known correlations of low and total cloud cover with other variables, such as SST, SLP and EIS (Estimated Inversion Strength), we evaluate which category of data sets (surface observations rather than satellites) appears to be more reliable when there is discrepancy in the sign of cloud anomaly.

This analysis will help understand how cloud cover has changed over the last sixty years and if there has been a significant increase in the amount of incoming solar radiation at the top of the atmosphere due to decreased cloud cover.