Machine Learning-Based Supervised Classification of Sentinel-2 MSI and Landsat-8 OLI Imagery in Marguerite Bay of Antarctic Peninsula

Aug 4, 2026

Upload by Sage

Machine learning algorithms are applied to classify Sentinel-2A MSI and Landsat-8 OLI satellite imagery of the Marguerite Bay region on the Antarctic Peninsula. Four supervised approaches are used for glacier-field mapping, combining pixel-based and object-based workflows. Random Forest and Decision Tree, as well as Support Vector Machines and k-nearest neighbor, support object-based image analysis, while SVM, LightGBM and XGBoost are used for pixel-based classification. Each dataset is labeled into glacier, water and soil, and accuracy is assessed through comparative error-matrix evaluation. Overall accuracies reach 97.31% and 96.28% using SVM.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages8

Reading Time20 min

AccessFree to download

Introduction

Study Area and Data

Machine Learning-Based Supervised Classification of Sentinel-2 MSI and Landsat-8 OLI Imagery in Marguerite Bay of Antarctic Peninsula Mehmet Arkalı 1, Muhammed Enes Atik 1, Şaziye Özge Atik 1 1 ITU, Department of Geomatics Engineering, Civil Engineering Faculty, 34469 Maslak Istanbul, Türkiye – [markali@itu.edu.tr](markali@itu.edu.tr); [atikm@itu.edu.tr](atikm@itu.edu.tr); [donmezsaz@itu.edu.tr](donmezsaz@itu.edu.tr) Keywords: Antarctica, Machine Learning, Sentinel-2, Landsat-8, Object-based Classification, Pixel-based Classification. Abstract Especially in the last decade, many innovative advantages of machine learning algorithms have been known, and their use in places where the effects of climate change are closely monitored, such as the polar regions, has introduced revolutionary scientific breakthroughs. In this study, machine learning methods were used to classify Sentinel-2A and Landsat-8 OLI satellite images of Marguerite Bay of Antarctic Peninsula. Four supervised classification algorithms were applied for pixel-based and object-based classification. Random Forest (RF), Decision Tree (DT), Support Vector Machines (SVM), k-nearest neighbor (kNN) are the algorithms selected for object-based image analysis (OBIA) . SVM, RF, Light Gradient Boosting Machine (LightGBM) and Extreme Gradient Boosting (XGBoost) were used for pixel-based classification. Each image is labelled into three classes: glacier, water and soil. The classification methods were analysed comparatively for each data set. In both Sentinel-2 and Landsat-8 images, 97.31% and 96.28% overall accuracy were achieved with SVM, respectively. 1. Introduction Antarctica's ecosystem records atmospheric events from the past to the present. It offers the best environment for scientific studies aimed at monitoring the effects of global warming. Global warming, also known as the greenhouse effect, refers to the increase in the Earth's average surface temperature due to greenhouse gases trapping solar heat in the atmosphere. This effect occurs when greenhouse gases in the atmosphere absorb thermal radiation emitted from the Earth's surface, acting as a blanket over the surface (Houghton, 2005) . The cryosphere consists of areas of snow or ice exposed to temperatures below 0°C for at least part of the year. Glaciers, also a cryosphere component, hold more than 70% of the world's freshwater reserves. The largest parts of the cryosphere are located in Greenland and Antarctica. The Antarctic ice sheet contains approximately 91% of the world's ice (Baumhoer et al., 2018) . Due to this feature, it is an important research area to examine the changes in glacier areas caused by global warming. Satellite remote sensing has facilitated significant advancementsin comprehending the climate system and its changes by quantifying the processes and spatiotemporal states of the atmosphere, land, and oceans (Yang et al., 2013) . Remote sensing allows the examination of characteristics and phenomena that are difficult to access or suitable for direct observation. Remote sensing satellites with many different sensors and measurement techniques are important for monitoring changes in glacial areas. Detecting changes in glacier areas contributes to future scientific studies by characterizing glacier balance and modelling climatic behaviour. The study involved using Sentinel-2 and Landsat-8 satellite images to perform feature extraction with classification in glacier fields. Satellite data were selected in February and March 2024. The selection of these dates considered important criteria, including the summer season in the southern hemisphere and the absence of cloud cover. Following the classification step, an accuracy assessment utilizing error matrices was used. The outcomes demonstrated that satellite data and pixel-based and object-based classification techniques were adequate for the classification of the region. Several machine learning algorithms are used for the class

FAQs

Which satellite images and sensors are used in the study?

The study uses Sentinel-2A MSI and Landsat-8 OLI satellite images of Marguerite Bay on the Antarctic Peninsula.

How many supervised classification algorithms are tested, and what are they?

Four supervised classification algorithms are applied in total, including Random Forest, Decision Tree, Support Vector Machines, and k-nearest neighbor for object-based analysis, and SVM, LightGBM, and XGBoost for pixel-based classification.

What classes are the images labeled into and how is accuracy evaluated?

Images are labeled into three classes: glacier, water and soil. Accuracy is evaluated comparatively using error matrices after classification.

Aug 4, 2026

Upload by Sage