Using artificial neural networks and citizen science data to assess jellyfish presence along coastal areas

• Published in: The Journal of applied ecology 2024-09, Vol.61 (9), p.2244-2257
• Main Authors: Castro‐Gutiérrez, J., Gutiérrez‐Estrada, J. C., Báez, J. C.
• Format: Article
• Language: English
• Subjects: Adaptive management; Artificial neural networks; beach management; Beaches; citizen science; Cnidaria; Coastal management; Coastal tourism; Coastal zone; Coasts; Environmental factors; Environmental management; Environmental monitoring; Infomedusa APP; jellyfish; Machine learning; multi‐layer perceptron; Neural networks; Noise prediction; Playas; Public health; Sea surface temperature; Temperature perception; Tourism; Wind; wind conditions; Wind direction; Wind speed
• ISSN: 0021-8901; 1365-2664
• DOI: 10.1111/1365-2664.14734

Jellyfish blooms along coastal areas can pose significant challenges for beach users and local authorities. Understanding the factors influencing jellyfish presence is crucial for effective management and mitigation strategies. In this study, citizen science data from the Andalusian coast (232 beaches, in 40 different localities) and machine learning techniques are used to investigate if the presence and absence of jellyfish along coastal areas can be predicted. A multi‐layer perceptron (MLP) neural network was employed to classify user comments regarding jellyfish presence or absence, achieving an accuracy of approximately 96%. The MLP model demonstrated robustness in handling non‐linear classification problems and noise, although it showed lower precision for predicting jellyfish presence, likely due to an imbalance in the dataset. Environmental data were also incorporated to characterise the influence of sea surface temperature, wind direction and wind speed on jellyfish distribution. The results align with previous studies, suggesting these environmental factors significantly impact jellyfish presence. Synthesis and applications. This research provides actionable recommendations for beach management. The implementation of continuous monitoring of sea surface temperature and wind conditions will enable more accurate predictions of jellyfish distribution. Adaptive management strategies that respond dynamically to environmental data will help mitigate the impact of jellyfish blooms on coastal tourism and public health. Resumen Las proliferaciones de medusas en áreas costeras pueden representar desafíos significativos para los usuarios de playas y las autoridades locales. Comprender los factores que influyen en la presencia de medusas es crucial para estrategias efectivas de gestión y mitigación. En este estudio, se utilizaron datos de ciencia ciudadana de la costa andaluza (232 playas, en 40 localidades diferentes) y técnicas de aprendizaje automático para investigar si se puede predecir la presencia y ausencia de medusas en las áreas costeras. Se empleó una red neuronal Perceptrón Multicapa (MLP) para clasificar los comentarios de los usuarios sobre la presencia o ausencia de medusas, logrando una precisión de aproximadamente el 96%. El modelo MLP demostró robustez al manejar problemas de clasificación no lineales y ruido, aunque mostró menor precisión para predecir la presencia de medusas, probablemente debido a un desequilibrio en el conjunto de datos.