The impact of subsampling on MODIS level-3 statistics of cloud optical thickness and effective radius

The Moderate Resolution Imaging Spectroradiometer (MODIS) Level-3 optical thickness and effective radius cloud product is a gridded 1°×1° dataset that is derived from aggregation and subsampling of every fifth pixel, along both spatial directions, of Level-2 orbital swath data (Level-2 granules). The present study examines the impact of this subsampling on the mean, standard deviation, and inhomogeneity parameter statistics of optical thickness and effective radius. The methodology is simple and consists of estimating mean errors for a large collection of Terra and Aqua Level-2 granules by taking the difference of the statistics at the original and subsampled resolutions. It is shown that the Level-3 subsampling does not affect the various quantities investigated to the same degree, with second-order moments suffering greater subsampling errors, as expected. Mean errors drop dramatically when averages over a sufficient number of regions (e.g., monthly and/or zonal averages) are taken, pointing to a dominance of errors that are of random nature. When histograms built from subsampled data with the same binning rules as in the Level-3 dataset are used to reconstruct the quantities of interest, the mean errors do not deteriorate significantly. The results in this paper provide guidance to users of MODIS Level-3 optical thickness and effective radius cloud products on the range of errors due to subsampling they should expect and perhaps account for, in scientific work with this dataset. In general, subsampling errors should not be a serious concern when moderate temporal (e.g., monthly) and/or spatial (e.g., zonal) averaging is performed.