An expeditious low-cost method for the acoustic characterization of seabeds in a Mediterranean coastal protected area

• Autores: Sanchez Carnero, Noela Belen; Rodríguez Pérez, Daniel; Llorens, S.; Orenes Salazar, V.; Ortolano, A.; García Charton, J. A.
• Año de publicación: 2023
• Idioma: inglés
• Tipo de recurso: artículo
• Materia: ACOUSTIC CLASSIFICATION; MEDITERRANEAN SEA; POSIDONIA OCEANICA; RANDOM FOREST; SEABED MAPPING
• Institución: Consejo Nacional de Investigaciones Científicas y Técnicas

Posidonia oceanica meadows are ecosystem engineers which, despite their ecological relevance, are experiencing habitat fragmentation and area decrease. Cartography and information on the ecological status of these habitats is key to an effective maritime spatial planning and management for habitat conservation. In this work we apply an acoustic methodology to map benthic habitats (substrate and vegetation) in an archipelago of the Natura 2000 Network close to the coast of Murcia (SE Spain) where dense and sparse areas of P. oceanica, and patches of Cymodocea nodosa appear over a sandy and had bottom. The methodology uses dual frequency information (200 kHz and 38 kHz) acquired with a single-beam echosounder to develop a bathymetry, and performs sea bottom and vegetation supervised classifications, using video and scuba diver observations as groundtruthing data. Sea bottom was classified from acoustic features of the first and second 200 kHz echoes into 5 substrate classes using a random forest classifier: sand, fine sand, coarse sand, hard bottoms and hard bottoms with sandy patches. The vegetation was classified from features extracted from the “above-bottom” part of the echo (height and backscattering intensity) in both frequencies, resulting also in a 5 class classification: C. nodosa meadows, dense P. oceanica meadows, dispersed P. oceanica meadows, dense P. oceanica with sand patches, and no-vegetation; according to the random-forest Gini index, 38 kHz features were the most informational variables for this classification. The validation accuracies of both classifications were 85% (substrates) and 70% (vegetation), close to accuracies reported in the literature when using a similar number of classes. The results of this article (including bathymetric, and substrate and vegetation thematic maps), together with the acoustic methodology described and used, are contributions that can improve the continuous monitoring of Mediterranean seagrasses.
Fil: Sanchez Carnero, Noela Belen. Consejo Nacional de Investigaciones Científicas y Técnicas. Centro Científico Tecnológico Conicet - Centro Nacional Patagónico. Centro para el Estudio de Sistemas Marinos; Argentina
Fil: Rodríguez Pérez, Daniel. Universidad Nacional de Educación a Distancia; España
Fil: Llorens, S.. Universidad Politécnica de Valencia; España
Fil: Orenes Salazar, V.. Universidad de Murcia; España
Fil: Ortolano, A.. Universidad de Murcia; España
Fil: García Charton, J. A.. Universidad de Murcia; España