Optical Remote Sensing of Oil Spills by using Machine Learning Methods in the Persian Gulf - A Multi-Class Approach

Aug 3, 2026

Upload by Himbo

Marine oil spills cause significant environmental harm and impose high social costs, especially in coastal regions where ecosystems and habitats depend on stable conditions. This thesis applies supervised machine learning to classify thick oil, using remotely sensed multi-spectral data from the Persian Gulf. The study extracts labeled data from 10 Sentinel-2 tiles across six spectral bands (492–2202 nm) and trains multiple model families, including decision trees, KNN, and artificial neural networks. After selecting top performers, robustness is tested on an independent dataset, supporting multi-class supervised monitoring with multi-spectral optical sensing.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages108

Reading Time4 h 32 min

AccessFree to download

Abstract

Acknowledgements

Contents

Introduction

Outline of the thesis

Theory

Background physics

Passive optical remote sensing of oceans

Passive remote sensing of marine oil slicks

Pattern recognition

Experimental setup

Data

Image selection

Data collection

Data specification

Spectral response

Methodology

Annotation

Band selection

Upsampling

Dataset

Training and testing

Prediction

Classifier setup

Software

Results

Phase 1

Phase 2

Phase 3

Prediction

Faculty of Science and Technology Department of Physics and Technology Optical Remote Sensing of Oil Spills by using Machine Learning Methods in the Persian Gulf: A Multi-Class Approach Martin H. Evenseth EOM-3901 Master’s thesis in energy, climate and environment 30 SP June 2023 This thesis document was typeset using the UiT Thesis LaTEX Template.© 2023 – [http://github.com/egraff/uit-thesis](http://github.com/egraff/uit-thesis) Abstract Marine oil spills are harmful for the environment and costly for society. Coastal areas are particularly vulnerable since they provide habitats for organisms, animals and marine ecosystems. This thesis studied machine learning methods to classify thick oil in a multi-class case, using remotely sensed multi-spectral data in the Persian Gulf. The study area covers a large area between United Arab Emirates (UAE) and Iran. The dataset is extracted from 10 Sentinel-2 tiles on six spectral bands between 492 nm to 2202 nm. These images were annotated for four classes, namely thick oil, thin oil, ocean water and turbid water by using the Bonn Agreement to analyse true color composite images. A variety of machine learning methods were trained and evaluated using this dataset. Then a robustness evaluation was done by using selected machine learning methods on an independent dataset. Initially multiple machine learning methods were included; three decision trees, six K-Nearest Neighbor (KNN) models, two Artificial Neural Network (ANN) models, two Naive bayes models, and two discriminant models. Two KNN models and two ANN models were then picked for further evaluation. The results show that the fine KNN approach with two nearest neighbors had the best performance based on the computed statistical measures. However, the robustness evaluation showed that the tri-layered NN performed better. This thesis has shown that supervised machine learning with a multi-class approach can be used for oil spill monitoring using multi-spectral remote sensing data in the Persian Gulf. Acknowledgements Without the help and support from fantastic people I don’t know what would have become of this thesis, thank you! First, I would like to express my greatest gratitude to my supervisor at UiT, Katalin Blix. Thank you for your extensive knowledge about the subject, foryour guidance with encouraging feedback, for your enthusiasm trying to recruit me to different seminars, for your continuous good mood, for proofreading my thesis and for your availability, always having time for questions and meetings. I would like to thank Martine Espeseth and Hugo Isaksen at Kongsberg Satellite Services for their expertise and help composing ideas and putting together a framework for the thesis, as well as helping to identify oil slicks using the Bonn Agreement Oil Appearance Code. Thanks to Silje Birgitte Segrem Grue for help and for proofreading. Thanks to my colleagues at TEOS who the last two months willingly have traded shifts with me, which made it easier to juggle between work and studies. Finally I would like to thank my girlfriend, Julie Høie Nygård, for her emotional support and for taking extra care of me and our home in the final stages of the writing process, and our cat Arja for always reminding me to take breaks. Martin H. Evenseth Tromsø, June 2023 . Contents Abstract i Acknowledgements iii List of Figures vii List of Tables xi Abbreviations xiii 1 Introduction 1 1.1 Outline of the thesis ...................... 3 2 Theory 5 2.1 Background physics ...................... 5 2.1.1 Electromagnetic radiation ............... 5 2.1.2 Atmospheric interactions ............... 9 2.1.3 Ocean interactions ................... 11 2.2 Passive optical remote sensing of oceans ........... 14 2.2.1 Radiation-From the sun to data products ...... 14 2.2.2 Passive optical sensing ................. 15 2.2.3 Multispectral imaging ................. 16 2.2.4 Affects on detection capabilitie

FAQs

What problem does the thesis address and where is the study area?

The thesis focuses on classifying marine oil spills, particularly thick oil, using remotely sensed multi-spectral data. The study area covers a region in the Persian Gulf between the United Arab Emirates and Iran.

How is the dataset built and what classes are used?

The dataset is extracted from 10 Sentinel-2 tiles across six spectral bands from 492 nm to 2202 nm. Images are annotated into four classes: thick oil, thin oil, ocean water, and turbid water.

Which machine learning approaches are evaluated and how are models selected?

Multiple methods are initially trained, including decision trees, K-Nearest Neighbor models, Artificial Neural Networks, Naive Bayes, and discriminant models. The study then selects two KNN and two ANN models for further evaluation before conducting robustness testing on an independent dataset.

Aug 3, 2026

Upload by Himbo