Generación de un mapa dinámico de zonas propensas a blanqueamiento en el arrecife de Varadero’s Hope Spot, Cartagena de Indias, Colombia, aplicando algoritmos de machine learning en imágenes de PlanetScope y datos de temperatura de la superficie del mar de NOAA en el periodo 2023-2024.

Abstract

The project focuses on creating a dynamic map of bleaching-prone areas in Varadero's Hope Spot Reef, located in the Bay of Cartagena, Colombia, during the 2023-2024 period. Advanced remote sensing techniques and machine learning algorithms (Random Forest, ANN, and U-Net) were applied to PlanetScope satellite images and NOAA sea temperature data to identify and classify high-risk areas. The U-Net model demonstrated the best performance in segmentation, achieving an overall accuracy of 93.5% and a Kappa index of 0.90. Field validations were conducted through reef dives to confirm the results. The final map serves as a key tool for environmental management and supports the development of conservation strategies aimed at protecting this unique and resilient ecosystem.