The Advanced Baseline Imager (ABI) instrument onboard the GOES-R series satellites, which is expected to be launched in the year 2014, has considerable potential for providing accurate retrievals of earth Land Surface Temperature (LST). At NOAA/NESDIS/STAR, Yu et al. (2007) have developed and refined a number of independent unique split window LST retrieval algorithms applicable for the ABI sensor. These algorithms are being evaluated through radiative transfer model simulations as well as through ground-truth data. For the ground-truth based evaluation, the LST algorithms have been applied to one year of GOES-8 and GOES-10 measurements in the year 2001, and then compared with coincident LST estimates from Surface Radiation Network (SURFRAD) irradiance measurements. This paper describes a manual cloud filtering method which we developed for the SUFRAD-GOES data comparison purpose.

The manual cloud filtering technique is applied to the spatially closest GOES pixel to the SURFRAD site, to determine if the pixel is cloudy or clear, in order to filter out the cloud-contaminated LST retrievals. This way of pixel cloud filtering employs visual determination of cloudiness based on visible channel 1 reflectance image, IR channel 4 brightness temperature image, daily time series curves of solar irradiance provided by the SURFRAD pyranometer, the broadband sky irradiance provided by the SURFRAD PIR instrument and a number of channel differences. While the solar irradiance curve provides much help during the day time, the down-welling sky irradiance curve is shown to be the most useful tool along with the channel 4 brightness temperature images in cloudiness determination irrespective of whether it is day or night. The combined use of GOES data in five different channels and the paired SURFRAD data of solar irradiance and down-welling sky radiance ensured high quality cloud filtered data during both day and night timings. The details of the manual cloud filtering criteria are discussed. The challenges in effective cloud filtering of GOES-data and how the combined use of GOES and SURFRAD data can be successfully utilized to address them in the context of this study are demonstrated with examples. A future generalized automated cloud detection scheme for satellite based multi-band passive radiometer data is discussed.