The new Seafloor Observatory (OBSEA) for remote and long-term coastal ecosystem monitoring

Abstract

A suitable sampling technology to identify species and to estimate population dynamics based on individual counts at different temporal levels in relation to habitat variations is increasingly important for fishery management and biodiversity studies. In the past two decades, as interest in exploring the oceans for valuable resources and in protecting these resources from overexploitation have grown, the number of cabled (permanent) submarine multiparametric platforms with video stations has increased. Prior to the development of seafloor observatories, the majority of autonomous stations were battery powered and stored data locally. The recently installed low-cost, multiparametric, expandable, cabled coastal Seafloor Observatory (OBSEA), located 4 km off of Vilanova i la Gertrú, Barcelona, at a depth of 20 m, is directly connected to a ground station by a telecommunication cable; thus, it is not affected by the limitations associated with previous observation technologies. OBSEA is part of the European Multidisciplinary Seafloor Observatory (EMSO) infrastructure, and its activities are included among the Network of Excellence of the European Seas Observatory NETwork (ESONET). OBSEA enables remote, long-term, and continuous surveys of the local ecosystem by acquiring synchronous multiparametric habitat data and bio-data with the following sensors: Conductivity-Temperature-Depth (CTD) sensors for salinity, temperature, and pressure; Acoustic Doppler Current Profilers (ADCP) for current speed and direction, including a turbidity meter and a fluorometer (for the determination of chlorophyll concentration); a hydrophone; a seismometer; and finally, a video camera for automated image analysis in relation to species classification and tracking. Images can be monitored in real time, and all data can be stored for future studies. In this article, the various components of OBSEA are described, including its hardware (the sensors and the network of marine and land nodes), software (data acquisition, transmission, processing, and storage), and multiparametric measurement (habitat and bio-data time series) capabilities. A one-month multiparametric survey of habitat parameters was conducted during 2009 and 2010 to demonstrate these functions. An automated video image analysis protocol was also developed for fish counting in the water column, a method that can be used with cabled coastal observatories working with still images. Finally, bio-data time series were coupled with data from other oceanographic sensors to demonstrate the utility of OBSEA in studies of ecosystem dynamics.

Keywords: EMSO ESONET; OBSEA; activity rhythms; automated video image analysis; cabled observatories; fish community; multidisciplinary observation; remote ecosystem monitoring.

Figure 1.

The location of the OBSEA seafloor cabled observatory in the western Mediterranean Sea (A), with details of its localisation off the Catalan coast (B). The cable routes in the sea and on the land (B) are indicated by green and red lines, respectively. OBSEA is located in a protected fishing area (i.e., the “a–e” polygon shown in B).

Figure 2.

Oblique, vertical, and close lateral views of the OBSEA cabled seafloor observatory showing three structural elements: (A) the cable powering the platform instruments; (B) the junction box within the cylinder holding the installed sensors; and (C) the video imaging system. The concrete column used for studies on faunal colonisation is also visible close the observatory.

Figure 3.

The architecture of the main cylinder (indicated by B in Figure 2) holding the junction box and the sensors within the OBSEA seafloor cabled observatory. Element sizes and distances are reported in millimetres.

Figure 4.

One-month time series of processed hourly oceanographic data, as recorded by CTD (A) and ADCP (B) sensors installed on the OBSEA seafloor cabled observatory (see Table 1 for references on their specifications).

Figure 5.

Fish species commonly detected in the OBSEA seafloor cabled observatory area by the installed video imaging system (indicated by C in Figure 2). Given the capability of the imaging system to rotate on its axis by 360°, different views of the water column and the concrete column for colonisations studies are presented. (A) Diplodus vulgaris and D. annularis; (B) Diplodus vulgaris and D. annularis; (C) D. sargus, D. annularis, and Chromis chromis; (D) C. chromis and Dentex dentex.

Figure 6.

Diagram of the processing steps in the automated video image analysis protocol used to count fish in digital frames acquired by the OBSEA seafloor cabled observatory.

Figure 7.

Time series of visual fish counts (black line) as produced by the automated video imaging analysis of still frames acquired by the OBSEA seafloor cabled observatory video imaging system. (A) The whole three-month time series of fish visual fish observations. (B) A 15-day subset (from 18 September to 3 October 2009) is presented (black line) together with corresponding CTD measurements (for a period not included in the previous oceanographic measurements). The latter comparison was performed as an example of potential interdisciplinary multiparametric coupled habitat and bio-data acquisition.

Figure 8.

A two-week graphical comparison (from 8–23 September 2009) of the time series obtained by manual counting (i.e., visual inspection) and automatic fish identification from images taken by the camera of the OBSEA seafloor cabled observatory. This manual versus automatic time series comparison shows the efficiency of the elaborated automated processing protocol.

Figure 9.

Examples of different types of errors (see Table 2 for references on their typologies) encountered during the automated processing of images taken by the OBSEA cabled seafloor observatory video camera: (A) is an example of an image with correctly classified fishes; (B) is an example of an image with a fish school (>20 animals) classified as containing less than 10 individuals (Img > 20-Class < 10 in Table 2); (C) is an example of an image in which a distant school is not fully differentiated from the background; and (D) is an example of an image where the resampling of the same object occurred (i.e., M > N in Table 2).