Bayesian cloud detection for 37 Years of Advanced Very High Resolution Radiometer (AVHRR) Global Area Coverage (GAC) data

Bulgin, C. E. ORCID: https://orcid.org/0000-0003-4368-7386, Mittaz, J. P.D., Embury, O. ORCID: https://orcid.org/0000-0002-1661-7828, Eastwood, S. and Merchant, C. J. ORCID: https://orcid.org/0000-0003-4687-9850 (2018) Bayesian cloud detection for 37 Years of Advanced Very High Resolution Radiometer (AVHRR) Global Area Coverage (GAC) data. Remote Sensing, 10 (1). 97. ISSN 2072-4292 doi: 10.3390/rs10010097

Abstract/Summary

Cloud detection is a source of significant errors in retrieval of sea surface temperature (SST). We apply a Bayesian cloud detection scheme to 37 years of Advanced Very High Resolution Radiometer (AVHRR) Global Area Coverage (GAC) data, which is an important source of multi-decadal global SST information. The Bayesian scheme calculates a probability of clear-sky for each image pixel, conditional on the satellite observations and prior probablity. We compare the cloud detection performance to the operational Clouds from AVHRR Extended algorithm (CLAVR-x), as a measure of improvement from reduced cloud-related errors. To do this we use sea surface temperature differences between satellite retrievals and in-situ observations from drifting buoys and the Global Tropical Moored Buoy Array (GTMBA). The Bayesian scheme reduces the absolute difference between the mean and median SST biases and reduces the standard deviation of the SST differences by ~10 % for both daytime and nighttime retrievals. These reductions are indicative of removing cloud contaminated outliers in the distribution, as these fall only on one side of the distribution forming a cold tail. At a probability threshold of 0.9 typically used to determine a binary cloud mask for SST retrieval, the Bayesian mask also reduces the robust standard deviation by ~5-10 % during the day, in comparison with the operational cloud mask. This shows an improvement in the central distribution of SST differences for daytime retrievals.

Item Type	Article
URI	https://reading-clone.eprints-hosting.org/id/eprint/74770
Identification Number/DOI	10.3390/rs10010097
Refereed	Yes
Divisions	Science > School of Mathematical, Physical and Computational Sciences > Department of Meteorology
Publisher	MDPI
Download/View statistics	View download statistics for this item