|
This paper presents a quantitative comparison oftwo cloud detection techniques using satellite
observations. The AVHRR (Advanced Very High Resolution Radiometer) Processing scheme
Over cLouds Land and Ocean (APOLLO) makes use of five spectral channels with a spatial
resolution of I .I km. The Collocated BIRS/2 and AVHRR ProductS (CHAPS) operates with
more spectral channels but a lower spatial resolution. To reference the satellite derived cloud
amounts, APOLLO results are compared with surface observations of cloud amount. The
APOLLO cloud amount and surface observations of cloud cover are generally within over
vegetated surfaces.
Over oceans, the agreement in total cloud cover between the two satellite techniques is very
good (r=O.92). Application of a dependent sample i-test to the two cloud amount data sets
indicates that there is a greater than 99.9% probability that the two samples were drawn from the
same population. This demonstrates that the subsampling of AVHRR pixels in the CHAPS
processing is appropriate for deriving cloud amounts over a 2.5degree oceanic region. For such
a region there is a tendency for CHAPS to derive higher cloud amounts than APOLLO. This is
attributed to differences in clear-sky radiance thresholds derived from the CHAPS spatial
variability test.
Over land, the derived cloud amount products from the two methods are considerably different.
The CHAPS product is an effective cloud amount defined for each HIRS field ofview which is
the product of cloud fraction and cloud emissivity rather than a simple areal percentage. Also,
the HIRS/2 footprint size (17 km at nadir) is much larger than that of the AVHRR. There is a
good correlation of the two cloud products (r=O.82); however, a t-test indicates the two
techniques are deriving fundamentally different parameters. This is consistent with the above
differences. Recommendations for improving the two cloud retrieval techniques are suggested.
|
