In this paper we present a cloud detection algorithm developed for the Arctic region using AVHRR data. Our approach is a simplified version of the Ebert (1987) method to discriminate between clouds, ice and open water in the Arctic Sea. The algorithm is tuned to work on an AVHRR scene typical of the winter to spring transition period. The algorithm has been applied to one month (154 scenes) of NOAA14-AVHRR images (from 16 March to 15 April 1998) covering the region of the Arctic Sea near the Svalbard Islands. The cloud detection results are analyzed using various check procedures. The algorithms pixel classification performance was verified by a satellite image expert. The misclassified pixels were digitalized on the image and counted by the expert in order to quantify the algorithms accuracy. The cloud classification results are quite accurate: 70% of the images (109) have an error less than 5% and only 11% of the images results to have an error greater than 10%. The methods performance is also tested against independent cloud and ice observations obtained respectively from the Ny-Ålesund meteorological base and from the Special Sensor Microwave/Imager (SSM/I) data set. The comparison with these independent sources of data confirms the algorithms good performance.
Remote sensing of cloud cover in the Arctic region from AVHRR data during ARTIST experiment
2003
Abstract
In this paper we present a cloud detection algorithm developed for the Arctic region using AVHRR data. Our approach is a simplified version of the Ebert (1987) method to discriminate between clouds, ice and open water in the Arctic Sea. The algorithm is tuned to work on an AVHRR scene typical of the winter to spring transition period. The algorithm has been applied to one month (154 scenes) of NOAA14-AVHRR images (from 16 March to 15 April 1998) covering the region of the Arctic Sea near the Svalbard Islands. The cloud detection results are analyzed using various check procedures. The algorithms pixel classification performance was verified by a satellite image expert. The misclassified pixels were digitalized on the image and counted by the expert in order to quantify the algorithms accuracy. The cloud classification results are quite accurate: 70% of the images (109) have an error less than 5% and only 11% of the images results to have an error greater than 10%. The methods performance is also tested against independent cloud and ice observations obtained respectively from the Ny-Ålesund meteorological base and from the Special Sensor Microwave/Imager (SSM/I) data set. The comparison with these independent sources of data confirms the algorithms good performance.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
