a comprehensive examination of polar ice dynamics integrating amsr e dataset to analyze spatiotemporal variability in arctic ice concentration extent and types in response to temperature changes

Abstract

newline his research focuses on investigating sea ice characteristics, encompassing concentration at varying time intervals, analyzing extent over diverse periods in a specific region, and classifying types based on associated temperatures. The primary goal is to enhance meteorological predictions concerning temperature, precipitation, and atmospheric conditions. newlineIn the pursuit of these objectives, three studies are conducted. Each study follows a consistent methodology involving the normalization of original sea ice images, clustering, segmentation, noise reduction using techniques like DTCWT and DDDTDWT, and subsequent feature extraction based on texture properties by using GLCM. The features are then subjected to selection using the infinite feature selection algorithm. newlineIn the first study, a Multi-class Support Vector Machine, leveraging a suitable kernel function, significantly improves prediction accuracy. The second study employs the k-Nearest Neighbor method, valuable for capturing local patterns and relationships, essential in predicting ice types with spatial coherence. In the third study, Convolutional Neural Networks (CNNs) are utilized for learning complex spatial patterns in Polar Regions, providing data on sea ice concentration, extent, and surface temperatures.