Prasow-Émond, Myriam, ‘Impacts of climate change on small island nations: a data science framework using satellite imagery and observational time series’
Abstract:
Small Island Developing States (SIDS) comprise a group of 58 nations identified by the United Nations as facing unique sustainability challenges. These challenges include high exposure to climate change, lack of data, and limited resources. The effects of climate change are already observed in SIDS, notably an increase in the magnitude and frequency of natural disasters, biodiversity loss, ocean acidification, coral bleaching, sea-level rise, and coastal erosion. The coastal zone is considered to be the main economic, environmental, and cultural resource of SIDS, making them particularly vulnerable to the adverse effects of climate change. This project focuses on quantifying and disentangling coastal changes, including erosion, accretion and coastline stability. Existing literature lacks a comprehensive understanding of the patterns of coastal changes, as well as the main anthropogenic and environmental drivers involved. We address this research gap by quantifying the challenges that SIDS encounter, with a particular emphasis on coastal changes.
The approach is data-driven, relying on observational time series extracted from satellite imagery, in situ measurements, and open-access databases. We have developed a robust method based on image segmentation to extract the island’s shape over time, enabling us to illustrate the island’s dynamics and obtain reliable time series of the coastline position.
The main drivers of coastal changes are then identified and quantified using time series analysis methods, including causal inference and discovery methods, for SIDS worldwide. We place a specific focus on the Maldives (Indian Ocean) due to its low elevation and high human activity. Additionally, the methodology expands to investigate a spectrum of issues, including the impacts of human activities (e.g., land reclamation, sand mining, shoreline armouring) on the natural responses of coastlines, as well as the effects of confounding factors or common drivers (e.g., Indian monsoon, tropical cyclones, and El Niño/Southern Oscillation). The ultimate goal is to develop a spatiotemporal variable coastline vulnerability index by integrating socioeconomic and environmental time series data, facilitating the assessment of environmental policies in SIDS.
Bio:
Myriam is a second-year PhD student in the department of Earth Science & Engineering at Imperial College London and is part of the Science and Solutions for a Changing Planet DTP from the Grantham Institute – Climate Change and the Environment. She has a background in physics and astronomy and has extensive experience of using telescope data. She is now working with remote sensing data to study the impacts of climate change on small island nations.
AlZayer, Zayad B, ‘Monitoring of shallow water environments with unsupervised learning‘
Abstract:
Monitoring the vast and dynamic coral reefs of Eastern Australia presents unique challenges. This talk explores the application of unsupervised learning techniques to streamline and enhance the analysis of large-scale remote sensing data. By leveraging Principal Component Analysis (PCA), K-means clustering, and Simple Linear Iterative Clustering (SLIC) superpixels, we demonstrate how these methods can be harnessed to automate image selection, identify water bodies, create depth invariant indices.
PCA serves as a powerful tool for reducing the dimensionality of complex remote sensing data, making it more tractable and interpretable. K-means clustering further organizes the simplified data into distinct clusters, revealing intricate spatial patterns and conditions within the coral reef ecosystems.
Through the integration of these techniques, we showcase the development of a robust and efficient pipeline for objective assessment of coral reef health and dynamics. By automating key processes, we significantly enhance the scalability and accuracy of our monitoring efforts, ensuring comprehensive data analysis while maintaining manageability.
Bio:
Zayad is a 3rd year student in the department of Earth Science & Engineering at Imperial College London, affiliated with John Lab, he has previously worked as a Geologist in both Oil & Gas and Mining as a micropalaentologist and exploration geologist. Currently working on a much larger scale, using satellite imagery to study change in shallow marine environments using machine learning.