A cloud screening algorithm containing variability test and coarse mode test was developed in order to eliminate cloud-affected data in the skyradiometer measurements taken under partly cloudy conditions. The results compared to cloud amount from weather station reports and lidar measurements show that the variability test appears to effectively remove thick low level cloud while the use of size distribution is effective for removing thin high level cloud. The new algorithm demonstrated that cloud screening is more effectively performed in comparison to the method currently used for SKYNET data processing. The developed cloud screening method has been applied for the dust detection from skyradiometer measurements. The performance of dust detection for skyradiometer was validated on the basis of SYNOP dust reports and the yellow sand index from lidar measurements. It is shown that the developed cloud screening methods helps to detect dust cases, effectively removing cloud-contaminated signals from the dust signals. In turn this algorithm contributes to improving the accuracy of the dust influence on radiative forcing and its efficiency by reducing uncertainties in the AOT and SSA retrievals.