Evaluación de la ocurrencia de medusas en la costa andaluza mediante datos de ciencia ciudadana
Loading...
Publication date
Authors
Department
Research group
Center
Related publication
Abstract
En la presente Tesis Doctoral se ha llevado a cabo un análisis integral sobre las proliferaciones de medusas en la costa andaluza mediterránea a partir de datos de ciencia ciudadana no dirigida y técnicas de inteligencia artificial y modelización ecológica. El primer capítulo ofrece una revisión del conocimiento científico disponible sobre las medusas en el mar de Alborán, con especial atención a la especie Pelagia noctiluca, centrándose en los principales aspectos biológicos y ecológicos, los factores que influyen en sus eventos de aparición, así como en los impactos ecológicos y socioeconómicos que generan. El segundo capítulo aplica redes neuronales artificiales Perceptrón Multicapa, para clasificar automáticamente los comentarios de los usuarios de la aplicación móvil Infomedusa con información sobre presencia o ausencia de medusas durante 2019. En estos análisis se alcanzó una precisión del 96% en la clasificación, destacando la efectividad del aprendizaje automático en el procesamiento de comentarios basados en lenguaje natural procedentes de ciencia ciudadana. En el tercer capítulo se utiliza el modelo de distribución de especies MaxEnt para predecir las zonas con mayor y menor probabilidad de presencia de medusas, combinando los reportes de los ciudadanos con variables ambientales con y sin desfase temporal. Los resultados muestran que la profundidad de la capa de mezcla en abril es el factor ambiental más determinante, con contribuciones superiores al 70%.En el cuarto capítulo se evalúa la utilidad de la lógica difusa como herramienta de inteligencia artificial explicable para modelar la aparición de enjambres de medusas en las playas de la Costa del Sol analizando el efecto a corto plazo de la dirección y velocidad del viento. Los resultados muestran que los vientos perpendiculares favorecen los varamientos en la zona centro-oriental, mientras que los paralelos lo hacen en la zona occidental, ofreciendo una base sólida para una gestión costera adaptativa. En conjunto, los hallazgos presentados ofrecen un punto de partida para el diseño de sistemas de monitoreo más robustos y estrategias de gestión que integren la variabilidad ambiental y la participación ciudadana en la planificación costera.
------------------------------------------------------------------
In the present Doctoral Thesis, a comprehensive analysis has been conducted on jellyfish proliferations along the Mediterranean coast of Andalusia, based on unstructured citizen science data and techniques from artificial intelligence and ecological modelling. The first chapter provides a review of the available scientific knowledge on jellyfish in the Alboran Sea, with particular emphasis on the species Pelagia noctiluca. It focuses on key biological and ecological aspects, the drivers influencing their appearance events, as well as the ecological and socio-economic impacts they cause. The second chapter applies Multilayer Perceptron artificial neural networks to automatically classify user comments from the mobile application Infomedusa containing information on jellyfish presence or absence during 2019. These analyses achieved a classification accuracy of 96%, highlighting the effectiveness of machine learning in processing natural language-based comments derived from citizen science. The third chapter employs the MaxEnt species distribution model to predict areas with higher and lower probabilities of jellyfish presence, combining citizen reports with environmental variables both with and without time lag. Results indicate that the depth of the mixed layer in April is the most influential environmental factor, contributing to over 70%. The fourth chapter assesses the usefulness of fuzzy logic as an explainable artificial intelligence tool to model the occurrence of jellyfish swarms on the beaches of the Costa del Sol, analysing the short-term effects of wind direction and speed. The results show that perpendicular winds favour strandings in the central-eastern area, while parallel winds do so in the western area, providing a solid foundation for adaptive coastal management. Taken together, the findings presented offer a starting point for the design of more robust monitoring systems and management strategies that integrate environmental variability and citizen engagement in coastal planning.
In the present Doctoral Thesis, a comprehensive analysis has been conducted on jellyfish proliferations along the Mediterranean coast of Andalusia, based on unstructured citizen science data and techniques from artificial intelligence and ecological modelling. The first chapter provides a review of the available scientific knowledge on jellyfish in the Alboran Sea, with particular emphasis on the species Pelagia noctiluca. It focuses on key biological and ecological aspects, the drivers influencing their appearance events, as well as the ecological and socio-economic impacts they cause. The second chapter applies Multilayer Perceptron artificial neural networks to automatically classify user comments from the mobile application Infomedusa containing information on jellyfish presence or absence during 2019. These analyses achieved a classification accuracy of 96%, highlighting the effectiveness of machine learning in processing natural language-based comments derived from citizen science. The third chapter employs the MaxEnt species distribution model to predict areas with higher and lower probabilities of jellyfish presence, combining citizen reports with environmental variables both with and without time lag. Results indicate that the depth of the mixed layer in April is the most influential environmental factor, contributing to over 70%. The fourth chapter assesses the usefulness of fuzzy logic as an explainable artificial intelligence tool to model the occurrence of jellyfish swarms on the beaches of the Costa del Sol, analysing the short-term effects of wind direction and speed. The results show that perpendicular winds favour strandings in the central-eastern area, while parallel winds do so in the western area, providing a solid foundation for adaptive coastal management. Taken together, the findings presented offer a starting point for the design of more robust monitoring systems and management strategies that integrate environmental variability and citizen engagement in coastal planning.














