ICES CM 2011/F:05
Not to be cited without prior reference to the author
Integrating benthic surveys and community analysis through image management: a case study from a biodiversity special area in the Lower St. Lawrence Estuary, Québec, Canada
Claude Nozères1*, Lizon Provencher2, François Roy2, and Jean-Sébastien Lauzon-Guay2
1 611 route de la mer, Ste-Flave, QC, Canada, G0J 2L0
2 Fisheries and Oceans Canada, Maurice Lamontagne Institute, 850 route de la mer, Mont-Joli, QC, Canada G5H 3Z4
*Corresponding author. e-mail address: [email protected] tel: +1418-750-3786
Abstract
The Manicouagan peninsula in the lower St. Lawrence Estuary is under consideration as a marine conservation zone, notable for its rich productivity and species diversity. To assist with the evaluation of habitats, consumer-level cameras and image management software were used to make data from different surveys readily available for ecological analyses. Surveys of habitats from sand shoals to the deepwater channel were conducted at stations from 5 to 320 m depth using both physical sampling and a towed camera sled. An image catalogue database was used to compile field photos of grab and dredge samples (infauna) with laboratory images (conserved specimens) and the underwater tow photos (epifauna and demersal species). Software tools were used to leverage photo metadata and embed record information including date, station, coordinates, and taxonomic names into the catalogue. Browsing images and filtering for metadata proved to be an efficient means for validating identification and location data. Catalogue records were exported for archiving to a network geodatabase, and subsequently used in analyses of community composition. Multivariate analyses performed with data from underwater photos and benthic grabs suggested similar zones of communities, even as the groups of species used in the classifications differed between the two methods. Underwater photo surveys in conjunction with image cataloguing tools are suggested as a low-cost complement to physical sampling for the investigation of benthic habitats and their communities while greatly expanding the total area sampled.
Introduction
The lower St. Lawrence Estuary encompasses a marine region with primarily rocky shores and a deep channel (>300 m). A special area in this region is the Manicouagan peninsula on the north shore, where the estuaries of three rivers (Betsiamites, aux Outardes, and Manicouagan) empty out onto intertidal shoals and the main estuary of the St. Lawrence (Fig. 1). The particular hydrographic and physical features of this area result in high levels of primary productivity (Therriault and Levasseur 1985) and species diversity that has lead to interest in the establishment of a marine conservation area. A full review of geography and biology of the Manicouagan study area is presented in Mark et al. 2010.
Figure 1. Manicouagan study area (red line) as investigated in the St. Lawrence Lower Estuary, Québec, Canada. White dots indicate survey station locations. Apart from a few surveys in the estuary in the past (e.g., Massad and Brunel 1979, Robert 1979), little was known about the benthos in the area. As part of the development of an ecological monitoring plan for the Manicouagan (DFO 2011), surveys were undertaken between 2006 and 2008 to document the benthic communities (Mark et al. 2010). To sample organisms ranging from the intertidal (5-30 m) down to the channel (>300 m), several techniques were employed. A Smith-McIntyre benthic grab (0.1 m2) was used in 2006 for soft-sediment infauna, while a larger volume (1 m2) IKU grab was used in 2008 for sampling larger organisms (e.g., bivalves) in both sand and soft sediment (Fig. 2).
Figure 2. Example of an IKU benthic grab (above) and sorted organisms on a sample tray (below).
Additional sampling was conducted with a trawl, and by using a hydraulic dredge in the 5-30 m zone. The biological samples were complemented with underwater photo transects, using a towed benthic sled equipped with a Nikon D90 DSLR, taking photographs at 10 s. intervals. Photo transects were conducted at stations throughout the area in 2006, covering an average of 3.7 m2 by transect (Fig. 3). In 2007, additional photo transects were carried out perpendicular to the shore, down to 30 m (Mark et al. 2010).
Figure 3. Example of an underwater photo transect of images that are examined for the presence and abundance of epifaunal (surface) species. Yellow indicates frames during descent and ascent, not counted in species analysis.
Image management
Organizing the data and biological samples across the different surveys was assisted by the use of images and digital asset management (DAM) software. Image management tools, or DAM software, are used by photographers and
agencies to organize image data, leveraging the metadata in files and collections to allow for quick and efficient searches and data exports (Krogh 2009). While intended for commercial use, the development of metadata standards in consumer digital still camera has led to the widespread use of metadata viewing and editing tools. This popularity has lead to some degree of consistency, notably with the standards of EXIF (data properties originating with the camera) and IPTC-XMP (user-added data fields). Even as this popularity has increased, the use and definition of metadata in images is continually being revised, with fields being expanded or trimmed to better adapt their use for commercial photographers (IPTC 2010, MWG 2010). Despite these underpinnings for commercial artists, several of the data fields are useful for marine biologists working with images, especially in taxonomy (Kennedy et al. 2011, Nozères 2011).
In the Manicouagan project, an image catalog using Adobe Lightroom software was initially established to organize the images from the underwater photo transects. Photos to document biological samples of interest from the benthic grabs and dredge were subsequently also added to the catalog. The image catalog became a valuable resource once the automatic metadata, i.e., capture data and original file names as viewed from EXIF, were combined with added metadata in the IPTC fields such as keyword, title, and location. Examples of important data fields that were used in the project are listed in Table 1.
Table 1. Image metadata fields used in the Manicouagan catalog. Data
field Use Example Note
File name
A unique label to trace image file
date-file
(YYYYMMDD-file = 20080812_IGP8616)
Naming strategy may vary with the project and the # of cameras in use Title Intended to
describe image, but often used for original file name
20080812_IGP8616 If file name is changed, searches can be conducted using title Capture date Marker for organizing by survey date and location
08-08-12 7:48:23
(may include indication of time zone)
Capture date is crucial for tagging images with the GPS coordinates from that moment Keyword Labeling the
subject in view, origin of image
Taxonomic name (Anonyx), survey label (grab2008)
Important tool for data searches and exporting by subject Location A separate,
keyword-like tag for places
Station 70 Intended for
geographical locations
GPS Latitude-Longitude coordinates
49°2'34"N, 68°12'51"W May also input for depth (using negative altitude values), but not always understood by software
Rating 1-5 stars, to indicate quality
* = 1 star, ** = 2 stars Sometimes repeated as keywords: '1star' Label Group theme or
subject In this project: yellow=caution/not to use, red=unwanted or error/not to use, blue: stationary duplicates/not to use Colour alternative to text keywords; proprietary to some systems--not easily exportable as a data field Flag Picks and
Rejects Quickly marking images of interest, or for deletion Field proprietary to Adobe systems Several commercial DAM software packages are available to catalog image files, in particular, Microsoft Expression Media (now PhaseOne Media Pro), Canto Cumulus, ID Imager, Apple Aperture, and Adobe Bridge--although this last example is actually a file browser and not a cataloging program. One of the benefits of cataloging with DAM software is to efficiently apply and edit the data fields, as shown in Table 1. Thus, file names, capture dates, keywords and other data fields can be filled or modified (for example, to correct capture date in case of camera clock errors), individually or for thousands of files simultaneously. Such a task may difficult to accomplish efficiently using some file browsers.
Tagging with metadata in the Manicouagan catalog was performed for several purposes. For spatial analysis, images needed to have associated spatial coordinates, which was accomplished using either capture dates or station names as locations (Table 2). For underwater transect photos, the available GPS track logs were used to tag images with coordinates based on time stamps, after compensating for depth and cable length between the ship and the towed sled. For biological samples, the station name was used as a location tag; all images with that station name were then tagged in bulk with the known coordinates of that station.
Table 2. Spatial tagging of images: GPS tracks and station coordinates
Image source Geotagging method
Transects along a point station Image capture date field synchronized with GPS time-date and coordinates Biological samples from a point station Station coordinates applied in bulk to
all image files tagged with the name in the location field, e.g., 'Station 70'
After location (station name and GPS coordinates), the most important task in tagging was the application of keywords, principally as taxonomic names referring to organisms in view. Currently valid scientific names, according to marine species registers (Van Guelpen and Kennedy 2011), served as name lists. Keywords were also valuable for marking images by the objects in view (e.g., shells, debris, rocks), type (blurry photos, information/label tag photos), and survey mission (grab2006, grab2008, underwater2006, etc).
For a catalog of over 10 000 images, the tagging by location and keyword represented a significant effort. Fortunately, the Lightroom interface incorporates several organizational features, including 'smart collections' (rule-based collections) that dynamically update with the addition and changes in metadata tags to image files, thereby greatly improving the efficiency of tagging work (Krogh 2009). The completion of tagging resulted in a easy-to-use graphical database that enabled instant searches and viewing of files, for example by date, station, or taxon, across all surveys with image data (Figure 3).
Figure 3. An example showing one of several filtering/browsing tools of Lightroom, with consecutive filters for the IPTC standard tags of keyword (species), date, and location (station). Additional filters are shown for label (colour) and attribute (white flag).
This ease of searching proved to be very useful in verifying uncertainties, such as taxonomic identifications and locations or dates stored on paper records and transcribed into data files. Additionally, underwater images could be readily compared with images of laboratory-identified samples. While species identification based solely on an underwater image was not always possible, this comparison with images of biological samples from the same or adjacent stations helped to highlight and even resolve issues in several cases (Figure 4). Examples of taxons corrected through the comparison of tagged image files
included amphipods (Neohela monstrosa, Ceradocus torelli, Maera loveni), bivalves (Panomya norvegica, Mya truncata, Mya pseudoarenaria, Mesodesma arctatum), echinoderms (Pentamera calcigera, Ophiacantha bidentata), sea pens (Anthoptilum grandiflorum), and sea anemones (Actinauge cristata, Actinostola callosa).
Figure 4. Case example of complementary searches and identifications obtained through quick review of cataloged images. Uncertain ophiuroids (brittlestars) in underwater photos (A) were similarly obtained in grab samples at a nearby station (B, C), and were identified to the species (Ophiacantha bidentata) by biologists examining dredge samples in the laboratory (D).
The variety of image filtering and viewing options is a powerful feature of Adobe Lightroom and was one reason for choosing the software for this project. Of additional interest was the potential flexibility of the Lightroom software, as it is comprised of a graphical interface based on the Lua programming language, overlying a SQLite3 database backend. The use of Lua, and SQLite3 in
particular, has enabled a number of enthusiast programmers to create metadata tools that enhance the filtering and the data exporting capabilities (Table 3). It is this ability to export (or even import) sets of textual data by using plugins that rewards the effort invested in tagging image files.
Table 3. Examples of Lightroom shareware plugins for database use.
Source Website Plugin example
Jeffrey Friedl's Lightroom Goodies
http://regex.info/blog/lightroom-goodies/ Metadata Wrangler : selecting certain data fields for export Photographer's
Toolbox
http://www.photographers-toolbox.com/index.php
LR/Transporter : import, export data as text files; modify data fields
For the Manicouagan project catalog, once the image files were validated through visual review with the browsing filter tools of Lightroom, the metadata of individual files (e.g., filename, date, station, keywords) were exported as tab-delimited text files, for incorporation into an Oracle database with an ESRI Spatial Database Engine (SDE). Any remaining errors in data records (e.g., inconsistencies in tag labeling or spatial coordinates) were quickly brought into evidence using the Oracle SDE database and readily corrected by referring to the unique filenames in the Lightroom catalog. Upon completion of metadata tagging in Lightroom, and the subsequent archiving of records into an Oracle database, the project was ready for the final stage: investigating the benthic communities in the Manicouagan through multivariate analyses of species abundances.
Multivariate analyses
As part of a proposed ecological monitoring plan, comparative analyses using two of the surveys (underwater transects 2006, IKU grab 2008) were undertaken to examine the community assemblages as found in the Manicouagan study area. The results of the analyses are presented in a research document (Provencher and Nozères 2011, in press) and an upcoming article, with summary information also available in Mark et al. 2010 and DFO 2011.
Using Primer software (Clark and Warwick 2001), group-average hierarchical cluster analyses of species abundances by station were undertaken for both the grab and underwater photo survey. The same Bray-Curtis distance similarity matrix was subsequently used in multidimensional analyses (MDS) to visually examine the cluster groupings. The importance of species contributions to each cluster grouping was calculated using SIMPER.
The analyses from both surveys indicated a series of benthic assemblages, leading principally from inshore down to the channel, but also with some
groupings along the shore, and also out onto the sandy plateau to the east (Figure 5). While there was considerable spatial overlap in the cluster groupings between the grab and the photo surveys, some differences were also in evidence. The infauna sampled with the grab (Fig. 5A) generally produced less distinctive groupings (MDS 2D stress: 0.16), with species such as polychaetes found in several clusters, although their abundance tended to diminish with increasing depth. The inshore stations also varied with the presence of different communities of bivalves (e.g., Serripes groenlandicus, Macoma calcarea, Astarte. By contrast, epifauna such as echinoderms, crustaceans, and cnidarians (sea anemones and sea pens) as seen in the photo transects (Fig. 5B) tended to be discrete in their species distributions across the zones from the inshore to the eastern plateau and down to the channel, as was also evidenced by the lower MDS stress score of 0.11.
Conclusion
The use of managed image data, from a variety of survey techniques, was shown to be of value for documenting habitats and species assemblages in this special area. The comparison of results from the different surveys, both physical samples and transect photos, helped to validate the underwater images that would otherwise have fewer recorded species. This may lead to further use of underwater imagery in catalogs to monitor species of interest, as in commercial fisheries (e.g., pink shrimp, Pandalus borealis or snow crab, Chionoecetes opilio) or for regionally-important prey species that are poorly documented in traditional survey captures in the St. Lawrence Estuary, as in the example of sand lance (Ammodytes sp.).
Figure 5. Cluster results for a) grab samples and b) photo transects. Insets show MDS analyses and list the principal species contributions, coloured by cluster. Figures adapted from Provencher and Nozères 2011.
A.
References
Clarke, K.R. and Warwick, R.M. 2001. Change in marine communities: an approach to statistical analysis and interpretation, 2nd edition. PRIMER-E: Plymouth, UK.
DFO. 2011. Review of the Manicouagan Marine Protected Area (MPA) Ecological Monitoring Plan. DFO Can. Sci. Advis. Sec. Sci. Advis. Rep. 2010/075.
IPTC (International Press Telecommunications Council). 2010. IPTC Standard Photo Metadata. http://www.iptc.org/std/photometadata/specification
Kennedy, M.K., Nozères, C., Miller, R, Vanhoorne, B. and Appeltans, W. 2011. The Canadian Register of Marine Species Photo Gallery: A User's Guide, Version 1. Can. Tech. Rep. Fish. Aquat. Sci. 2933: v + 47 pp.
Krogh, P. 2009. The DAM Book. Digital Asset Management for Photographers. 2ndEd. O’Reilly Media, Sebastopol, CA. 477 p.
Mark, S., Provencher, L., Albert, E. et Nozères, C. 2010. Cadre de suivi écologique de la zone de protection marine Manicouagan (Québec): bilan des connaissances et identification des composantes écologiques à suivre. Rapp. tech. can. sci. halieut. aquat. 2914 : xi + 121 p.
Massad, R. and Brunel, P. 1979. Associations par stations, densités et diversité des polychètes du benthos circalittoral et bathyal de I'estuaire maritime du Saint-Laurent. Nat. can. 106: 229-253.
MWG (Metadata Working Group). 2010. Guidelines for handling image metadata. Version 2.0. November 2010. 73 p. http://www.metadataworkinggroup.org
Nozères, C. 2011. Best practices guide for managing image data in marine sciences. Can. Tech. Rep. Fish. Aquat. Sci: v + 160 p. (in review).
Provencher, L. and Nozères, C. 2011. Monitoring plan for benthic communities of the Manicouagan MPA. DFO Can. Sci. Advis. Sec. Res. Doc. 2011/xx (in press). Robert, G. 1979. Benthic molluscan fauna of the St. Lawrence estuary and its ecology as assessed by numerical methods. Nat. can. 106: 211-217.
Therriault, J.-C. and Levasseur, M. 1985. Control of phytoplankton production in the lower St. Lawrence Estuary: Light and freshwater runoff. Nat. can. 112:77-96. Van Guelpen, L. and Kennedy, M. K. 2011. The history of Canadian registers of marine species 2001-2009. Can. Tech. Rep. Fish. Aquat. Sci. 2906: iv + 20 pp.
