Supervised Contrastive Learning for Coastal Resilience: Enhancing Climate Adaptation and Sustainability through Representation Learning

Climate change, rising sea levels, and stronger storms are threatening coastal areas, causing significant risks to ecosystems, tourism, and communities. Monitoring coastal change is vital to ensure that appropriate protection policies are established. Current deep learning approaches that monitor coastal changes face significant challenges in generalization, transferability, and adaptability in varied environments. These approaches often rely on large amounts of labelled data, which can be resource intensive and difficult to obtain on a scale. This can limit the effectiveness of the model in downstream tasks, such as coastline change classification. To address these limitations, this work proposes a supervised contrastive learning approach, an advanced deep learning technique that can enhance model performance by grouping embeddings from the same class, while differentiating them from other classes. These embeddings improve the performance for downstream classification tasks. The proposed approach demonstrates a 75.97% accuracy for classifying shoreline change compared to traditional supervised deep learning approaches at 61.04%. This development could improve the monitoring of coastal change, allowing governments and policy makers to establish more effective strategies to protect coastal areas from environmental challenges.