Cloud Detection

Here I present results of the cloud detection method used to identify clear-sky HIRS observations. The cloud detection method utilizes the technique used by ISCCP in that it identifies cloudy observations by spatial and temporal constrast testing. The following figure gives histograms of the HIRS channel 8 surface channel observations for N12 observations for January 2, 1992. The spatial contrast test compares HIRS 8 observations from adjacent observations and identifies cloudy observations for those with relative cold values. Temporal contrast test compares observations from the same location from the day previous and following the current observation and assigns the observation as clear, cloud, mixed cloud or undecided. Combining the two tests through a logic table gives the final cloud/no cloud result. Only observations assigned as clear are saved as clear-sky.
Cloud detection short-term test

A long-term cloud detection test is needed to remove regions with persistent cloud cover. Long-term tests require gather statistics on HIRS 8 over 15 and 30 days. Further comparison of the clear-sky observations from the short-term test yield remove the persistent cloud cover regions. The following figure gives the results for N12 on January 30, 1992.
Cloud detection long-term test

Validation of the cloud detection algorithm with ISCCP AVHRR data indicate good agreement between the two methods. Using ISCCP DX pixel-grid data, we made direct comparisons with N12 observations from January 8, 1992 with the Pathfinder clear-sky data. The first figure gives the all-sky HIRS 8 observations over Western Africa for 3-9Z. High clouds can be seen by the grey shading. Since N12 is a morning satellite, the desert regions appear relatively cool. The second figure gives the clear-sky results from the Pathfinder data. White areas indicate clear skies. They correspond reasonably well with the warmest HIRS 8 observations; however, large clear-sky regions are still identified over the relatively cool desert regions. The third figure gives the clear-sky regions as is identified from ISCPP AVHRR data. The clear-sky regions are larger than identified by Pathfinder. The convervative cloud test used in Pathfinder results in 23% of the cloudy areas in Pathfinder to be identified as clear with the ISCCP data. However, only 2% of region for the Pathfinder data identified as clear are found to be cloudy with ISCCP, thus indicating very good agreement between the methods.
N12 HIRS 8 all-sky observations
N12 HIRS 8 clear-sky observations
N12 ISCCP AVHRR clear-sky observations

Last modified: Wed May 2 11:51:20 MDT 2001