Cloud Masking Techniques for AVHRR Data for Oceanographic Applications in Malaysia
Khiruddin Abdullah *, Nasirun Mohd. Saleh *, Alui Bahari+, Ruslan Raints * and Yusuff Mahamod*
*University Sains Malaysia,
+Malaysian Metrological Service
Abstract
The objective of the present work is to apply AVHRR cloud masking producers to full resolution AVHRR images over sea water areas around Malaysia. The techniques used are (1) gross test-uses channel 5 brightness temperature, (ii) spatial coherence technique - consist of defining 3x3 pixels ( local array ) and calculating local mean and standard deviation using channel 4 data over each local array of brightness temperature, (iii) dynamic reflectance threshold test-channel 2 reflectance is use sine it is less affected by aerosol and molecular scattering, (iv) ration of near infrared ( channel 2) to visible ( channel ) reflectance's -cloud and cloud free pixels are identified form the histogram of this ratio and (v) channel difference-by examining the difference in brightness temperatures between thermal channels. All the above methods are used for day time test. The technique have been applied to AVHRR scenes have indicated that the threshold values obtained are consistent for each technique. A program incorporating the above techniques using the determined threshold values has been written for performing cloud masking for full resolution AVHRR images. A pixel will only be considered cloud free if it passes all the tests.
Introduction
AVHRR data from NOAA satellites have been widely use in oceanographic applications such as sea surface temperature ( SST) studies and other oceanic phenomena. Due to its high repetitive coverage AVHRR can become a use full tool for monitoring the oceanic processes. The atmosphere introduces errors in satellite measurements of SST through cloud cover, aerosols, water vapour, reflected sky and the presence of other gases. Clouds dominate the errors introduced into satellite observations of SST. Therefore the presence of clouds hinders their application especially in the equatorial region. Successful application of remote sensing data for oceanographic studies such as sea surface tem0perature measurements depends largely on the accuracy of the cloud removal techniques to eliminate cloud-contaminated pixels for m the analysis.
Methods of cloud removal for satellite data vary form radioactive thresholds to rather complicated procedures involving analysis of texture ( Saunders 1986, Gelloegos et al. 1993, Simposen and Humphrey 1990, Simpson and Gobat 1996). The techniques use by Saunders and Kriebel (1988) which are based on the rediative thresholds had been successfully applied in Western Europe and the North Atlantic have been adopted in the present study. But their application in the South East Asia equatorial region such as Malaysia still require some modification in terms of their threshold values which have been obtained by investigators using data sets form other geographical regions. A study by Franca ( 1994) has indicated that the thresholds values obtained for the equatorial region of Brazil differ form that obtained by Saunders and Kriebl ( 1988). Knowledge of cloud type within the image is not needed for this procedure.
