A Bayesian Cloud Mask for Sea Surface Temperature Retrieval

Abstract

Bayesian methods are used to develop a cloud mask classification algorithm for use in an operational sea surface temperature (SST) retrieval processing system for Advanced Very High Resolution Radiometer (AVHRR) local area coverage (LAC) resolution data. Both radiative and spatial features are incorporated in the resulting discriminant functions, which are determined from a large training sample of cloudy and clear observations. This approach obviates the need to specify the arbitrary thresholds used by hierarchical cloud-clearing methods, provides an estimate of the probability that an instantaneous field of view is cloudy (clear), and allows the skill of different cloud discriminant models to be objectively analyzed. Results show that spatial information is of particular importance in reducing the false alarm rate of the cloudy class. However, while the use of complex textural measures such as gray-level difference statistics?as opposed to simple statistics such as the standard deviation?improves the skill of nighttime cloud-masking algorithms, they are of little advantage during daytime hours. Cloud mask discriminant models having similar high Kuipers? performance index scores (i.e., 0.935) are developed for both day and night satellite data from the Southern Hemisphere midlatitudes. Applied to LAC orbital (i.e., operational) data, the characteristics of the cloud masks appear to be similar to those derived from analysis of the training sample data. However, in this case, to enhance processing performance, a hybrid algorithm is employed?obviously cloudy instantaneous fields of view (IFOVs) are first removed via a gross threshold check and the Bayesian method applied only to the remaining IFOVs. This same (hybrid) algorithm is also applied to an ensemble of 30 days of AVHRR LAC data from the New Zealand region. Analysis of the resulting time-composited SST data (means and standard deviations) shows there is little evidence of a day?night bias in the performance of the Bayesian cloud-masking algorithm and that the resulting SST data may be used to determine the variability of oceanographic features. Although this paper uses AVHRR data to demonstrate the principles of the Bayesian cloud-masking algorithm, there is no reason why the approach could not be used with other instruments.