HR: 0800h
AN: OS51C-1313 [Abstracts]
TI: Efficient and Statistically Valid Method of Textural Sea Floor Characterization in Benthic Habitat
Mapping
AU: Kostylev, V E
EM: vkostylev@nrcan.gc.ca
AF: Geological Survey of Canada (Atlantic), Natural Resources Canada
PO Box 1006, Dartmouth, NS B2Y 4A2
Canada
AU: * Orpin, A R
EM: aorpin@nrcan.gc.ca
AF: Geological Survey of Canada (Atlantic), Natural Resources Canada
PO Box 1006, Dartmouth, NS B2Y 4A2
Canada
AB:
The advent of multibeam bathymetric sonar technology and the thematic development of benthic habitat research have spawned
renewed interest in the systematic characterization and mapping of the seafloor. This necessitates the application of
reliable and accurate sea floor descriptors in combination with a robust means to statistically assess descriptor
associations. Traditionally, geoscientific sea floor mapping was comprised primarily of identifying the spatial extent and
relationship of geological units, broadly following chronostratigraphic conventions. Classifying seafloor sediments using
geological facies may not be meaningful biologically because they incorporate temporal elements that stem from a
geochronological qualifier. Textural properties of geological facies are typically reliant on the application of
distribution-dependent statistics, which have been shown to be inappropriate with multimodal marine sediments. While the
relationship between grain size and biota appears self-evident, there is a compelling argument that granulometric properites
alone are not a determinant of species distribution or community composition. The classification process is problematic
because most statistical clustering techniques will, by their very nature, form clusters which may or may not represent
meaningful and discernable differences. Moreover, as habitat mapping is aimed at boundary definition, the boundaries between
clusters in such cases could be based on very subtle differences, or noise (e.g. sampling bias). An independent measure of
the appropriate number of groups in a dataset is required. Therefore, we examine a statistical approach pioneered by Calinski
& Harabasz (C-H), which was implemented by a computer routine to work in partnership with information entropy analysis of
grain size data. We utilize a 30-year legacy of grain size data collected from the Scotian Shelf, Canadian Atlantic
continental margin, and show that considerable improvements in textural zonation are obtained using a combined entropy-C-H
technique, and these new zones better match known benthic divisions. The statistical validity of the clustering outcomes is
tested at each step of the classification and the effects of data reduction resulting from archiving or data processing is
examined.
DE: 4804 Benthic processes/benthos
DE: 4219 Continental shelf processes
DE: 3022 Marine sediments--processes and transport
DE: 3099 General or miscellaneous
DE: 0400 Biogeosciences
SC: Ocean Sciences [OS]
MN: 2004 AGU Fall Meeting