Cartographie de la biophonie des écosystèmes côtiers - Archive ouverte HAL Accéder directement au contenu
Article Dans Une Revue Traitement du Signal Année : 2016

Cartographie de la biophonie des écosystèmes côtiers

Résumé

1922 10s-long snapshots of sounds have been measured in a small 4km 2 coastal area (depth from 0 m to 50 m) near the pointe of Revelatta, (Calvi, Corse). Our aim is to build the soundscape of this complex area hosting 4 interleaved habitats (rocks, seagrass, sand, coralligenous). Wide Band [2 kHz, 40 kHz] SPL is maximum at the rocks and decrease offshore with the regression law: SPL (dB re.1µPa) = 147 – 14log10(r) with r the distance to the coastline. The range of detection of the biophony of the rocks extends from 1800m to 8300m. This regression law based on energetic criteria only allows to think that the biophony of the rocks is loud and propagates to the others habitats. We assess the diversity of the spectral shape of the biophony by computing the PCA of 2 million of measured spectra. The two first eigenvalues explain 50% of the spectral variability. The spectra and their scores 1 and 2 create a continuous cloud in {score 1, score 2} plan. Then, we propose to segment the data in 4 quadrants depending of the sign of score 1 and score 2 of the PCA. Each quadrant produces a family of spectral signatures (spectrum with high peak at 3 kHz and 4 KHz, spectrum with wide bump between 5 kHz and 12 kHz, spectrum with maximum at high frequency more than 20 kHz). The 4 proportions of impulses from a given habitat in each of the 4 quadrants are proposed to describe the diversity of spectrum shape of this habitat. For our data base, each habitat hosted the four families of spectral signature but with different proportions. The coralligenous hosts its own sound production with high frequency impulses. The rocks host its own sound production with low frequency impulses. Seagrass habitats listen to the rocks and sand habitats listen to the rocks and the coralligenous. MOTS-CLES : paysages acoustiques sous-marins, biophonie, habitats benthiques, analyse en composantes principales, monitorage par acoustique passive. Extended abstract From 7 days of seatrial in August 2014, 1922 10s-long snapshots of sounds have been measured with one acoustics drifting buoy and fixed acoustics moorings in a small 4km 2 coastal area (depth from 0 m to 50 m) near the pointe of Revelatta, Calvi, Corse (lat = 42°34'49.44''N, long = 8°43'27.15''E). Our aim is to build the soundscape of this complex area hosting 4 interleaved habitats (rocks, sea grass, sand, and coralligenous) and a well-conserved marine biodiversity and to assess if each habitat has its own biophony or if they share the same one. Wide Band SPL ([2 kHz, 40 kHz], bandwidth of biophony) is maximum at the rocks and decreases offshore with the regression law: SPL (dB re.1µPa) = 147 – 14log10(r) with r the distance to the coastline. The range of detection of the biophony of the rocks (CFAR energy detector, Pfa = 10-4 , ambient noise level equal to Wenz's model) extends from 1800 m to 8300 m depending on the wind speed. This regression law based on " energetic criteria only " allows to think that the biophony of the rocks is loud and propagates to the others habitats. We assess the diversity of the spectral shape of the biophony by computing the PCA of 2 million of the measured spectra. The two first eigenvalues explain 50% of the spectral variability (including variability due to the multi-path propagation and to the spectral shape of the emissions) and 80% of the spectral shape of the emissions. The spectra and their scores 1 and 2 create a continuous cloud in {score 1, score 2} plan without well distinct patches associated to a specific habitat. Then, we propose to segment the data in 4 quadrants depending on the sign of score 1 and score 2 of the PCA. Each quadrant produces a family of spectral signatures (spectrums with high peak at 3 kHz and 4 kHz, spectrums with wide bump between 5 and 12 kHz, spectrums with maximum at high frequency more than 20 kHz). The 4 proportions of the impulses from a given habitat in the 4 quadrants are proposed to describe the diversity of spectrum shape of this habitat. For our data base, each habitat hosted the 4 families of spectral signatures but with different proportions. The coralligenous hosts its own sound production with high frequency and weak impulses. The rocks host its own sound production with low frequency load impulses. Seagrass habitats " listen " to the rocks and sand habitats " listen " to the rocks and the coralligenous. For our studied area, to account for the spectral shapes of impulses from one habitat and their diversity allows to complement the blurred conclusions obtained using " SPL only considerations ". We were able to distinguish the sound production of coralligenous that is weak compared to the sound production of rocks but differ from it by the spectral shape of its impulses.
1922 segments de 10 secondes de mesures sonores ont été enregistrés dans une zone de 4 km² comprise entre le trait de côte et l'isobathe 50 m à la pointe de la Revelatta (Calvi, Corse). Nous caractérisons le paysage acoustique de cette zone complexe qui présente 4 habitats (roche, herbier, sable et coralligène) entrelacés. Le SPL calculé sur la bande biophonique [2 kHz, 40 kHz] est maximum sur la frange rocheuse et décroît vers le large en suivant la régression SPL (Sound pressure Level, dB re. 1µPa) = 147 – 14log10(r) où r est la distance à la côte. Ce modèle de régression s'appuyant sur des considérations énergétiques uniquement laisse penser que les émissions sonores de la roche sont fortes et qu'elles se propagent vers les autres habitats dont les biophonies ne seraient que des captations de celle de la roche. Pour évaluer la validité de cette hypothèse, nous avons étudié la diversité des formes spectrales mesurées. Pour cela, nous avons calculé une analyse en composantes principales (ACP) des 2 millions de spectres collectés. Les deux premières valeurs propres de l'ACP résument plus de 50 % de la variabilité totale des spectres et 80 % de la variabilité de la forme des impulsions. Les scores 1 et 2 de l'ACP forment un nuage continu que nous avons segmenté en 4 familles en fonction du quadrant du plan {score 1, score 2} auquel ils appartiennent. Ces 4 familles définissent 3 signatures spectrales de la biophonie du site étudié (les signatures des quadrants 1 et 2 étant proches) : des impulsions présentant un maximum marqué vers 3 et 4 kHz, des impulsions présentant une bosse entre 5 et 12 kHz, des impulsions présentant un maximum haute fréquence au-delà de 20 kHz.
Fichier principal
Vignette du fichier
Lossent_2015_RTS_Mise_page.pdf (1.06 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-01666317 , version 1 (29-03-2018)

Identifiants

Citer

Julie Lossent, Cedric Gervaise, Lucia Di Iorio, Pierre Boissery. Cartographie de la biophonie des écosystèmes côtiers. Traitement du Signal, 2016, 33 (1), pp.131 - 151. ⟨10.3166/TS.33.131-151⟩. ⟨hal-01666317⟩
195 Consultations
256 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More