Application of dynamic threshold in sea and lake ice mapping and monitoring
Ice detection and monitoring algorithms using visible and infrared images are generally founded on thresholds-based approaches. Classification of features over ice covered sea requires a series of reliable thresholds. The change in surface conditions throughout the season affects these thresholds and makes their adjustment necessary. This study proposes an operational method based on a set of dynamic thresholds for ice and water identification using data from a geostationary satellite. The Spinning Enhanced Visible and Infrared Imager (SEVIRI) on the Meteosat Second Generation (MSG) satellite data has been used. The proposed approach has been tested and validated over the Caspian Sea. Visible, near infrared and thermal infrared channels are being used to automatically create a cloud mask for a single image. The dynamic threshold is being developed to clarify the misclassification of ice and water pixels. The constant and dynamic thresholds have been used in comparison and applied to classification model. Dynamic threshold is used with reflectance channels R01 (0.6μ) and R02 (0.8μ) and the near infrared channel R03 (1.6μ).
Keywords: sea ice, dynamic threshold, Goes-R, remote sensing, lake ice, ice mapping, ice monitoring, ice detection, surface conditions, feature classification, geostationary satellites, Caspian Sea, reflectance channels, near infrared channels, satellite data, hydrology