Scene identification for ScaRaB data: the ISCCP approach

Determination of top‐of‐atmosphere longwave radiative fluxes: A comparison between two approaches using ScaRaB data

“…Table 1 shows the fraction of the ISCCP‐declared scene types that are labeled “unmixed” or “mixed” for May 1994. On average, about 65% of the matched clear (determined by ISCCP analysis) ScaRaB DX pixels are labeled “mixed,” indicating that a quite large portion of the ISCCP‐declared clear ScaRaB pixels may actually contain small amount of clouds due to the lower ScaRaB spatial resolution, as pointed out by Briand et al [1998]. A large majority (73%) of the ISCCP‐declared cloudy ScaRaB pixels are labeled “unmixed” rather than “mixed,” especially for the cloud types that are optically thick or of large size, such as deep convective or cirrostratus clouds (see Figure 2 for the ISCCP radiometric cloud classification).…”

“…In other words, the ScaRaB algorithms are as close as possible to the ERBE algorithms except for some parameters used in ERBE scene identification, which are missing in the published descriptions of ERBE [ Viollier et al , 1995], and the steps specific to the instrument characteristics, such as the spectral corrections. (Several improvements are being planned for the future version of ScaRaB data processing, including adapting ISCCP cloud detection algotithms to the coarse spatial resolution of the ScaRaB narrowband measurements for better scene/cloud identifications and thus improved clear‐sky fluxes and cloud radiative effects [ Briand et al , 1998; Stubenrauch et al , 2002] and applying different angular dependence models from recent studies for the conversion of radiances into fluxes [e.g., Stubenrauch et al , 1993]. ) Like all scanning radiometers, ScaRaB measures radiances at specific angles.…”

“…(1) The spatial and temporal variability of the surface characteristics between different pixels decreases relative to the effect of scattered low‐level cloudiness as the pixel size increases. This will cause some cloud‐contaminated ScaRaB pixels to be declared clear by the original ISCCP cloud detection algorithm (due to the nature of ISCCP algorithm itself), where the thresholds in the space/time variability test are selected for use with higher spatial and temporal resolutions [ Briand et al , 1998; Stubenrauch et al , 2002]. (2) ISCCP assumes that the cloud optical properties are uniform at pixel scales; hence the cloud cover of pixels is assumed to be either zero or 1.…”

“…It is worth pointing out that one of the limitations when applying the original ISCCP algorithm to ScaRaB narrowband radiances is the coarse resolution of ScaRaB. Previous studies, however, have shown that the ISCCP algorithm is rather stable with spatial resolution variations [ Briand et al , 1998; Stubenrauch et al , 2002], which implies that this is actually a smaller problem than one might have thought. Nevertheless, considerable efforts are devoted in this study to assess and mitigate the consequence of applying the original ISCCP algorithm to the coarser ScaRaB measurements.…”