High resolution mapping of a cold water coral mound

Luis A Conti,Aaron Lim,Andrew J Wheeler

doi:10.1038/s41598-018-37725-x

Abstract

Cold-water coral (CWC) mounds are biogenic, long-lived morphostructures composed primarily by scleractinian CWC’s and hemipelagic sediments that form complex deep-sea microhabitats found globally but specifically along the European-Atlantic margin. In this work, high-resolution mapping was applied to identify individual organismal distribution and zonation across a CWC Piddington Mound within the Porcupine Seabight, Ireland Margin. Marine Object-Based Image Analysis (MOBIA) and different machine learning classification methods (decision tree, logistic regression, and deep neural network) were applied to a high-resolution (2 mm) reef-scale video mosaic and ROV-mounted multibeam data in order to provide new insights into the spatial organization of coral frameworks and environmental factors on CWC mounds. The results showed an accurate quantification of the amount of Coral Framework (14.5%; ~2% live and ~12.5% dead) and sponges (~3.5%) with heterogeneous distribution, restricted to a certain portion of the mound. This is the first object level quantification of live and dead coral framework facies and individual sponges across an entire CWC mound. This approach has application for habitat and conservation studies, provides a quantification tool for carbon budget assessments and can provide a baseline to assess CWC mound change. The approach can also be modified for application in other habitats.

Highlights

Cold-water coral (CWC) mounds are biogenic, long-lived morphostructures composed primarily by scleractinian CWC’s and hemipelagic sediments that form complex deep-sea microhabitats found globally but along the European-Atlantic margin
Given the inherent difficulties of seabed mapping, in relation to acquisition and analysis of marine information, new methods have been proposed to increase knowledge from this unmapped part of the planet[5,6]. These include the use of structure from motion derived 3D photogrammetric reconstructions of deep-water habitats and the use of multifrequency multibeam backscatter for improved interpretations of subtle seabed features[7,8]
Photogrammetry, ROV- and AUV-mounted multibeam mapping have revealed their local-scale distribution on relatively flat areas[17] to near-vertical canyon walls[18,19]. Despite this potential improvement of mapping seafloor habitats with coupled “video/sonar” data, seabed optical images derived from photo/video cameras mounted on Remote Operated Vehicles (ROVs) in many cases, remain limited by exploratory survey designs[20,21,22], ground truthing[23,24,25] and rapid ecological assessment[26]

Summary

Introduction

Cold-water coral (CWC) mounds are biogenic, long-lived morphostructures composed primarily by scleractinian CWC’s and hemipelagic sediments that form complex deep-sea microhabitats found globally but along the European-Atlantic margin. Given the inherent difficulties of seabed mapping, in relation to acquisition and analysis of marine information, new methods have been proposed to increase knowledge from this unmapped part of the planet[5,6] These include the use of structure from motion derived 3D photogrammetric reconstructions of deep-water habitats and the use of multifrequency multibeam backscatter for improved interpretations of subtle seabed features[7,8]. Photogrammetry, ROV- and AUV-mounted multibeam mapping have revealed their local-scale distribution on relatively flat areas[17] to near-vertical canyon walls[18,19] Despite this potential improvement of mapping seafloor habitats with coupled “video/sonar” data, seabed optical images derived from photo/video cameras mounted on Remote Operated Vehicles (ROVs) in many cases, remain limited by exploratory survey designs[20,21,22], ground truthing[23,24,25] and rapid ecological assessment[26]. Combining images of different resolution tends to increase internal variability and noise within classes and may decrease the classification accuracy of traditional per-pixel basis methods[30]

Objectives

Methods

Results

Discussion

Conclusion