Discriminating clear sky from clouds with MODIS

Abstract

The MODIS cloud mask uses several cloud detection tests to indicate a level of confidence that the MODIS is observing clear skies. It will be produced globally at single‐pixel resolution; the algorithm uses as many as 14 of the MODIS 36 spectral bands to maximize reliable cloud detection and to mitigate past difficulties experienced by sensors with coarser spatial resolution or fewer spectral bands. The MODIS cloud mask is ancillary input to MODIS land, ocean, and atmosphere science algorithms to suggest processing options. The MODIS cloud mask algorithm will operate in near real time in a limited computer processing and storage facility with simple easy‐to‐follow algorithm paths. The MODIS cloud mask algorithm identifies several conceptual domains according to surface type and solar illumination, including land, water, snow/ice, desert, and coast for both day and night. Once a pixel has been assigned to a particular domain (defining an algorithm path), a series of threshold tests attempts to detect the presence of clouds in the instrument field of view. Each cloud detection test returns a confidence level that the pixel is clear ranging in value from 1 (high) to zero (low). There are several types of tests, where detection of different cloud conditions relies on different tests. Tests capable of detecting similar cloud conditions are grouped together. While these groups are arranged so that independence between them is maximized, few, if any, spectral tests are completely independent. The minimum confidence from all tests within a group is taken to be representative of that group. These confidences indicate absence of particular cloud types. The product of all the group confidences is used to determine the confidence of finding clear‐sky conditions. This paper outlines the MODIS cloud masking algorithm. While no present sensor has all of the spectral bands necessary for testing the complete MODIS cloud mask, initial validation of some of the individual cloud tests is presented using existing remote sensing data sets.