See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/343006618

Field Manual for Imagery Based Surveys using Remotely Operated Vehicles
(ROVs)

Chapter · July 2020

CITATIONS READS

5 1,546

10 authors, including:

Jacquomo Monk Neville S. Barrett

University of Tasmania University of Tasmania
73 PUBLICATIONS 1,738 CITATIONS 200 PUBLICATIONS 11,890 CITATIONS

SEE PROFILE SEE PROFILE

Todd Bond Ashley M. Fowler

University of Western Australia New South Wales Department of Primary Industries
51 PUBLICATIONS 1,249 CITATIONS 86 PUBLICATIONS 2,503 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Joel Williams on 17 July 2020.

The user has requested enhancement of the downloaded file.

10. Field Manual for Imagery Based Surveys
using Remotely Operated Vehicles (ROVs)
Jacquomo Monk*, Neville Barrett, Todd Bond, Ashley Fowler, Dianne
McLean, Julian Partridge, Nicholas Perkins, Rachel Przeslawski, Paul
G Thomson, Joel Williams
* ​[email protected]

Chapter citation:
Monk J, Barrett N, Bond T, Fowler A, McLean D, Partridge J, Perkins N, Przeslawski R, Thomson P.G, Williams J.
2020. Field manual for imagery based surveys using remotely operated vehicles (ROVs). In ​Field Manuals for
Marine Sampling to Monitor Australian Waters, Version 2​. Przeslawski R, Foster S (Eds). National Environmental
Science Programme (NESP).
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

Platform Description
Remotely Operated Vehicles (ROVs) are piloted, tethered unmanned submersibles typically
controlled from a vessel (sometimes from other fixed structures such as oil and gas platform
jackets) via a reinforced umbilical cable as the main tethering device. The tether historically
provided electrical power and also allowed the real-time transfer of data between the vessel and
ROV to be transmitted. With advancements in battery technology, smaller ROVs can now be
powered by onboard battery systems, which reduces the diameter of the tether, decreasing drag
and improving ROV maneuverability. The motion of ROVs are controlled by multiple thrusters that
allow movement and manipulation in all directions and speeds up to 3 knots. Onboard cameras and
sensors provide data and visual information that is relayed back to the surface personnel to observe
the seabed or other structures and control the ROV. Onboard sensors typically provide feedback on
water depth, temperature, currents, orientation and location of the ROV. The attachment of
manipulator arms can also allow for specimens and samples to be collected (including on in the
water column).

ROVs were originally designed in the mid-1980s to complement manned scientific submersibles.
With the increase in technology since, ROVs have gained acceptance because of their distinct
advantages over manned submersibles in many areas, notably reduced risk to pilots and
researchers. For instance, they can remain on the seafloor for extended periods efficiently
performing large surveys, running extended time series observations, and conducting
multidisciplinary operations ​(Shepherd 2001, Macreadie et al. 2018, Sward et al. 2019)​. A large
volume of data is transmitted to the surface, via multiple channels including real time video, sonar,
CTD (conductivity–temperature–depth) data, real time location and other information.

ROVs are available in a range of sizes and configurations from smaller observation-class vehicles
(~3-20 kg for mini and ~30-120 kg for regular-sized models) to larger work-class systems
(100-1,500 kg for light- and up to 5,000 kg for heavy-duty models), which vary in power, depth
rating, accessibility, and additional payload capabilities ​(Baker et al. 2012, Capocci et al. 2017,
Huvenne et al. 2018)​. As a result of the versatility, ROVs are increasingly being used for deep-water
surveys, with published examples of using ROVs for physical sampling via manipulator and grabber
arms, scanning sonars and high-definition cameras to provide researchers with still or video images
of the physical environment ​(Shepherd 2001, Leckie et al. 2015, Robert et al. 2017, Macreadie et al.
2018) and associated sessile mega-benthic taxa ​(Salvati et al. 2010; Thresher et al. 2014; Lacharité
et al. 2015; Cánovas-Molina et al. 2016; Price et al. 2019; López-Garrido et al. 2020) as well as
mobile organisms (such as fish; ​Karpov et al. 2006, Pradella et al. 2014, McLean et al. 2017,
Thomson et al. 2018)​. With advances in technology, a wider range of ROV models are becoming
available, including many low-cost systems, resulting in a greater uptake by researchers.

For further information on the advantages and disadvantages of ROVs compared to other benthic
imagery and sampling platforms, refer to ​Comparative assessment of seafloor sampling platforms
Przeslawski et al. 2018 and review by Sward et al. 2019).

Scope
The primary aim of this field manual is to establish a consistent sampling protocol for marine benthic
assemblages using ROVs and to facilitate statistically sound research to allow comparisons
between studies. This manual will focus on the use of ROVs for the collection of still and video

Page | ​2
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

imagery of fish and associated seabed habitats but consider researchers may use them for other
purposes as detailed in Table 10.1. We also consider all ROV classes here and provide some
guidance around the limitations associated with each class. The document leverages the expertise
of the working group focusing on still and video imagery (Chapters 4 and 7 for example, but see
Table 10.1 for a brief summary of additional uses for ROVs). ​The scope of the manual covers
equipment, pre-survey preparation, field procedures, and post-survey procedure for using ROVs to
photographically and videographically survey seabed assemblages (including fishes) found within
Australia’s vast marine estate.

Table 10.1:​ Additional uses of ROVs in monitoring the marine environment that are not covered in this manual (modified
from McLean et al. 2020)​.

Payload Description
CTD
Seawater temperature and salinity depth profiles

Bio-optical sensors
Fluorescence and backscatter (turbidity) sensor

Light meter
Upwelling and downwelling light, photosynthetically active radiation (PAR)

Dissolved oxygen sensor Dissolved oxygen concentrations

pH sensor Water column pH

Water sampler Water column samples for microbes, nutrients, pollutants, chlorophyll using
bottle samplers

Acoustic telemetry, Detection of tagged and untagged animals, migration patterns, connectivity
Hydrophones/passive acoustics

Scanning/Imaging sonar
Bathymetry, structural complexity

Sediment Corers/grabs Sedimentology or biogeochemistry e.g. particle size, sediment chemistry

Faunal traps Deployment and retrieval of baited traps for sampling of mobile fauna, including
fish and invertebrates
Faunal sampling In situ​ sampling of sessile and mobile fauna, including pelagic and demersal
fish and benthic invertebrates

Page | ​3
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

ROVs in Marine Monitoring

Using ROVs ​to visually monitor ​marine ​ecosystems ​has ​experienced ​a ​rapid ​increase ​over ​the ​past
two ​decades as a ​result of cheaper, smaller ROVs becoming available as well as improved access
to oil and gas sector ROVs (e.g. through the ​SERPENT initiative​; Macreadie et al. 2018) and
philanthropic ROVs (e.g. ​Schmidt Ocean Institute​). Researchers ​have ​used ​RO​Vs ​in ​monitoring ​the
impacts ​of ​invasive ​species ​(Whitfield et al. 2007)​, ​assessing marine protected areas ​(Dauble 2006,
Torriente et al. 2019) ​assessing ​population ​trends ​in ​demersal ​fishes ​(reviewed in Sward et al.
2019)​, ​mapping ​of ​benthic ​habitats ​(García-Alegre et al. 2014, Torriente et al. 2019)​, ​examining
diversity ​in ​reef ​communities ​(including on vertical walls; ​(Robert et al. 2017, Price et al. 2019)​,
detecting marine litter (GESAMP 2019), ​and ​assessing spatial and temporal changes in fish and
sessile benthos associated with artificial structures (such as oil and gas infrastructure; ​McLean et al.
2017, Bond et al. 2018)​.

While ROVs can be used for deploying a variety of sensors, as well as taking samples of substrata
and organisms (Table 10.1) they are also used to ​generate ​spatially ​accurate ​photomosaics ​and
finescale ​digital ​elevation ​models. Multibeam data which is often available with accurate
georeferencing can provide important information regarding habitat types and structural complexity
but is often limited to cell resolutions of 50 cm to 5 m. Finescale ​digital ​elevation ​models from ROV
photomosaics can be done at 1-10 cm cell resolution, and on vertical structures (something AUVs
currently struggle to achieve), thus enabling extremely detailed structural information to be extracted
(Robert et al. 2017)​. Additionally, and perhaps more importantly, the benefits of using ROV to
provide digital elevation models is that they also provide colour information (via the photomosaics),
which is crucial for identification of species and evaluation of condition (e.g. live vs. dead coral).

ROVs are not without their limitations when visually monitoring organisms. Different classes of
ROVs are better suited to certain situations and components of a species assemblage (Table 10.2).
There is generally a trade-off with high-quality macro-imagery and ROV functionality associated with
high costs and technical requirements (Figure 10.1). When using ROVs for visually monitoring
marine organisms, researchers should consider the potential effects of differing light intensity and
wavelength, impacts of sound intensity and frequencies (for example, large hydraulic ROVs are
noisy), and consequences of vehicle speed, size, altitude on survey bias particularly on mobile
organisms. Research suggests that a combination of these factors can have substantial effects on
the data collected ​(Stoner et al. 2008, Ryer et al. 2009, Rountree & Juanes 2010)​. While all
sampling platforms have associated biases, the limited access to work-class ROVs and a steady
uptake of cheaper smaller vehicles may make ROVs particularly prone to this bias. This is
particularly important if different vehicles are used between regions (e.g. inside vs outside no-take
reserves) or across time series sampling.

A key advantage that ROVs ​have in a monitoring context is their ability to be dynamically controlled
in ‘real time’ across a range of depths and habitats. This is because data are streamed real time
which means that the vehicle can survey vast areas with constant supervision and can be easily
focused on areas of interest. ROVs are the only marine imagery systems available in Australia that
are able to readily collect quality imagery from highly rugose environments, including vertical rock
walls, steep slopes, and overhangs. These environments are prevalent in many marine parks, along
the continental slope and offshore reefs. Similar to AUVs, when equipped with acoustic positioning
(e.g., ultra-short baseline, USBL), ROVs can be piloted along precisely ​defined ​transects, at a
constant altitude, ​with ​the ​geolocation ​of ​individual still ​images ​along ​this ​path as well as forward
facing stereo-video (along with other sensors if required/fitted). The geolocation of imagery and

Page | ​4
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

flight paths allows relatively precise repeat transects to be conducted for monitoring purposes, and
also for the imagery to be used to ground-truth multibeam sonar ​(Ierodiaconou et al. 2011)​,
assessing the effectiveness of marine protected areas ​(Torriente et al. 2019)​, as well as for
modelling ​the ​environmental ​factors ​driving ​species’ ​distributions ​(Salvati et al. 2010, García-Alegre
et al. 2014, Lastras et al. 2016)​. Although ROVs have been shown to collect comparable reef fish
assemblage data as diver-operated video and slow towed video (Shchramm et al 2019), they are
uniquely suited to collect data in environments that are otherwise challenging to other sampling
platforms.

Figure 10.1:​ Sample images showing the tradeoffs for different ROVs: [left]: sessile invertebrates from Hunter Marine
Park from a BlueRobotics BlueROV (with a heavy kit upgrade) fitted with stereo GoPro HERO7 Black cameras, [middle]
limestone outcrops along a canyon slope in the Gascoyne Marine Park from the ROV SuBastien’s situational camera, and
[right] brittlestars entwined around a black coral from the ROV SuBastien’s 4K camera.

Page | ​5
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

Table 10.2:​ Summary of ROV classes and considerations associated with each when used for monitoring Australia’s
marine estate (table modified from JNCC, 2018).

ROV class Class I: Class II: Class III:

Observation Observation (with Work

payload option)

Definition and Typically < 40kg in Larger vehicles than Weighing <~5000kg, these
capability weight these vehicles Class I, weighing vehicles have a broad
are primarily intended ~100-150kg, are carrying capability and
for observation only. capable of basic operational conditions (e.g.
Fitted with inbuilt physical sampling and depth and currents). Usually
camera and lights, may observations. Capable used in deeper waters (i.e. off
be able to handle one of carrying multiple continental shelf) these are
additional sensor (such cameras and sensors as the most complex and
as USBL), simple well as simple gabber versatile of ROVs used. They
grabber claws, as well claws. are often used in the Oil and
as an additional Gas sector.
stereo-video camera.

Examples BlueROV, Boxfish, Ocean Modules V8 Argus Mariner XL/Worker,

DeepTrekker, M500, Hercules; Holland, Isis; Jason
Fusion, Ocean Modules Pollox, Phantom, Saab 2; Kiel6000; Ocean Modules
V4 S300, OpenROV, Seaeye Falcon V8 L3000, SuBastian
Seabotix LBV300, (DR)/Cougar XT
Trident, VideoRay Pro4

Scale of Fine (<20m) - Meso Meso - Macro (>1km) Meso - Macro

operation^ (200m - 1km)

Max. Depth: <100m Depth: 0 - 300m​#​, Depth: >300m,

operational Sea state: <2m Current: Sea state: <3m Sea state: <4m
conditions <1.5kt Current: <3kt Current: <4kt

Deployment Manual Manual (<300m depth) LARS package or

type or vessel A Frame/crane vessel A-frame/crane (for
and winch or Launch shallow deployment). A
And Recovery System moonpool is a further option.
(LARS) package.

Tether Free swimming - tether Single body on main Single body on main umbilical
management connected to ROV. umbilical (live boating) (live boating) or TMS.
Clump weight or Tether Management
recommended in System (TMS).
deep/high current
deployments.

Page | ​6
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

Approx. survey AUD 2,000 - 10,000 AUD 5,000 - 40,000 AUD 50,000 - 120,000
cost per day*

Approx. AUD 10,000- AUD 200,000- AUD 1,000,000-

purchase 250,000 1,000,000 6,000,000+
cost^^

Vessel Fixed platform Shallow draught vessels Large vessel (~>50m) with
requirements (jetty/pontoon/oil/gas suitable for inshore Dynamic Positioning (DP),
platform), small vessel waters (10-30m), for deck capacity for container
(<10m) (with or without extended offshore storage and LARS.
power supply) or other surveys larger (~>30m)
small vessel. vessels will be used.

^ Ability to navigate across distance

#​
Deep Rated vehicles are available for >300m but have limited mobility at these depths.
* Planning and field work only. Purchase of ROV, consumables, processing of samples and reporting are not included.
^^ Estimates include basic positioning systems (such as USBL), grabber/manipulator and depth rated stereo cameras.
Based on quotes from the companies as well as catalogue entries.

Pre-Survey Preparations
Ensure ​all ​permits, ​safety ​plans ​and ​approvals ​have ​been ​obtained. ​Any ​research ​undertaken ​within
Australian Marine Parks (​AMPs) ​requires ​a ​research ​permit ​issued ​from ​Parks ​Australia. ​See
Appendix B for ​a ​list ​of ​potential ​permits ​needed. The observation of animals should be undertaken
in an ethical manner and in many cases, surveys may require approval from an Animal Ethics
Committee.

Define the aim of the project​. This is a mandatory step in any marine monitoring project, but with
their multiple capabilities (imagery, sampling, sensors), projects using ROVs may be particularly
vulnerable to competing research interests or distractions during a dive. A clearly defined aim or
hypothesis ensures the ROV pilot stays on task and is not distracted.

Confirm ​sampling ​design ​is ​statistically ​sound ​with ​adequate ​spatial ​coverage ​and ​replication, and
addresses the aim or hypothesis. ​This ​is ​generally ​achieved ​through ​the ​use ​of ​an ​explicit
randomization ​procedure ​to ​ensure ​that a sufficient number of ​independent ​replicates ​are ​obtained
(Foster et al. 2017, 2019, Smith et al. 2017)​.​ ​See​ ​Chapter​ ​2​ ​for​ ​further​ ​details​ ​on​ ​sampling​ ​design.

Select appropriate ​transect ​design ​for ​ROV deployment (Foster et al. 2019). ​The decision to which
transect design is most appropriate is driven by the question being addressed, the applied
capabilities of the ROV (i.e. sampling may be applied concurrently with image acquisition), the
environment, available time and logistics of ROV deployment and retrieval (e.g. size of system). For
example, tether and vessel drag within environments exposed to strong currents makes piloting an
ROV along a predetermined transect difficult if not impossible. In such situations ROVs (particularly
small ROVs) may not be the best system for temporal monitoring purposes because of the
challenges with maintaining physical position to enable sufficient overlap between repeat surveys
(i.e., within 20 m) (e.g. Przeslawski et al. 2012 in northern Australia). In addition, some
consideration must be given to the unique capability of ROVs to traverse steep slopes, including
vertical deployments, when undertaking quantitative image transects of a set distance. For these
situations, calculated distance cannot be ‘as the crow flies’ and will rely on high-resolution

Page | ​7
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

bathymetry as well as continuous monitoring by the ROV crew during deployment to determine
actual distance traversed.

For ​marine monitoring demersal fishes on the continental shelf a transect of ~150-200 m is
sufficient. ​Monk et ​al. ​(Unpublished) ​contrasted three transect lengths (50, 100, 150 m) finding that
at least 150 m was a ​generally ​sufficient design for monitoring purposes of demersal fish diversity (<
200 m)​. ​For surveys aiming to collect imagery of the epibenthos, or in deeper environments, then
longer transects are possibly required to gather sufficient imagery to characterise the focal regions.

For surveys that include fauna of mixed mobility, for example fish and invertebrates, a dual transect
approach may be suitable. The transect area can first be surveyed rapidly to ensure individuals of
highly mobile taxa are included, and then again at a slower speed to ensure observation of smaller
and more cryptic species.

For survey of fauna associated with topographical features, for example seamounts, vertical reef
structures or oil and gas facilities, transects conducted in an arc around the feature may be more
suitable than linear transects. The ROV can be thrusted laterally, allowing cameras to be
consistently oriented toward the feature throughout the transect.

Stereo-cameras specifications and calibration (​must ​be ​pre- ​and post-calibrated) ​in ​shallow ​water
using ​the ​techniques ​similar ​to ​those ​outlined ​in ​Boutros et al. (2015)​. ​We recommend cameras with
full, high-definition resolution of at least 1920 x 1080 pixels and a capture rate of at least 30 frames
per second. Higher camera resolution will improve identification of fish, and the pixel selection
required for measurement, but some models of action cameras can overheat at high resolution.
Higher frame rates reduce blur on fast-moving species. To maintain stereo-calibrations, cameras
must have video stabilisation disabled, and a fixed focal length can facilitate measurements both
close to and far from the camera systems when correctly calibrated ​(Boutros et al. 2015)​. ​The field
of view should be standardised and chosen to limit distortion in the image (e.g. no more than a
medium angle, ~95° H-FOV). When sampling demersal fish assemblages at typical maximum range
(6 m) from the cameras, Boutros et al. ​(2015) suggested a separation < 450 mm will result in a
decrease in the accuracy of measurements. Cameras are fixed to a rigid base bar to preserve the
stereo-calibration required to calculate accurate length and range measurements ​(Boutros et al.
2015)​. ​As outlined in Chapter 5 for stereo-BRUVs, SeaGIS software and 3D calibration hardware is
recommended for calibration of stereo video imagery. For stereo still imagery then a similar
approach documented in Chapter 4 for AUVs, with consistent lighting and adequate base separation
~ 300 mm are important to obtain well-lit and calibrated stereo imagery ​(Boutros et al. 2015)​.

Decide on appropriate navigational systems (e.g. USBL) and required spatial precision of imagery​.
In many cases a USBL should be used for both navigation and georeferencing imagery. However,
other methods can be employed such as doppler velocity logging or simple timed directional
transects for navigation and calibrated stereo imagery or stereo lasers for image scaling. For many
ROV studies the choice of navigational and georeferencing of imagery is often limited to what is
fitted to the unit available. However, appropriate effort must be given to this during the survey
planning phase as it may limit the questions sought to be answered by the imagery. For example,
spatial precision is very important for fine scale analysis whereas navigational accuracy is important
for temporal replication. Some alternative navigational methods, simple timed directional transects
are sometimes used if a USBL is not used, are not well suited to temporal replication as the exact
spatial location of the track cannot be determined. This results in resultant data needing to be
pooled to transect level. This reduces a key advantage of ROVs that individual observations can be

Page | ​8
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

co-location with finescale covariates (such as from multibeam sonar). This makes that data
collected in this fashion more akin to stereo BRUVs or underwater visual census which essentially
aggregate individuals to a sample. We suggest that both accuracy and spatial precision need to be
addressed for distance and swept area determination.

Ensure appropriate software is installed on onboard laptops (e.g. ROV navigation software platform,
GIS, etc), and potential users are familiar with it so that the ROV can be tracked and its mission
success monitored while underway. It is worth setting all equipment up in the laboratory or at dock
to ensure everything is operational and no software updates are required.

Ensure a trained technical team. For the work-class ROVs, a professional technical and piloting
team with training specific to the designated ROV will be required. For the smaller ROVs, training on
piloting and technical issues is still highly recommended during the pre-survey planning stage.

Field Procedures
Many of the steps in this section are designed for smaller class ROVs and are to be managed by
researchers or general marine technicians. Work-class ROVs will have their own deployment
protocols based on the technical capabilities and logistic requirements for the particular ROV and
associated professional team, and these may supersede the specific steps below.

Onboard sample acquisition

Complete​ ​an​ ​on-site​ ​briefing.

Prior to deployment, a ​deployment ​briefing ​should ​always ​be ​completed ​to ​ensure ​the ​operation ​can
be ​completed ​safely. ​Always ​take ​a ​precautionary ​approach ​to ​risks ​associated ​with ​vehicle
deployment.​ ​See Chapter 1 for further information about risk assessments.

Set up and test the ROV system.

Allow ​sufficient ​time ​during ​survey ​mobilisation ​to ​undertake ​system ​checks, ​calibrations ​and ​testing
of ​equipment ​and ​account ​for ​unforeseen ​problems; ​in ​most ​cases ​it ​will ​be ​possible ​to ​complete ​all
system ​setup ​and ​tests ​within ​half ​a ​day. ​The ​conduct ​of ​pre-start checks should be noted in the trip
log and any test failures specifically recorded for later-reference. Detailed settings ​for ​each
component​ ​should​ ​be​ ​made​ ​using​ ​relevant​ ​operations​ ​manuals​ ​(e.g.​ ​USBL​ ​operations​ ​manual​ ​etc.).

Acoustic​ t​ racking​ ​setup

• Set​ ​position​ ​of​ ​GPS​ ​receiver.​ ​Differential​ G

​ PS​ ​is​ m
​ andatory​ f​ or​ ​repeat​ s​ ite​ m
​ onitoring.

• Measure ​offsets ​of ​USBL ​transceiver ​head ​to ​GPS ​receiver ​and ​put ​offsets ​into ​navigation
systems.

• Deploy​ ​USBL​ ​transceiver​ ​(e.g.​ ​pole​ ​or​ ​vessel​ ​mounted).

• USBL calibration dockside is a good idea as well to verify that range and bearing (and depth
if estimated by USBL) are within expected tolerances. Understanding the selection and
recording of filtering/smoothing settings of the USBL system should also be noted.

On-deck​ t​ ests​ s​ hould​ i​ nclude,​ ​but​ n

​ ot​ l​ imited​ t​ o,​ t​ he​ ​following​ ​checks:
• on-board​ ​data​ ​storage

Page | ​9
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

• on-board​ ​power (if fitted)

• cameras

• tether management system (including assessing for nicks in tether)

• strobe​ ​lighting

• thrusters (assessing for fouling and operation)

• Manipulator arm(s) and sample container(s) (if fitted)

• all​ ​blanking​ ​plugs​ ​are​ ​installed

• crane​ ​and​ ​associated​ ​shackles​ ​are​ ​working​ ​order

• check​ ​all​ ​seals/o-rings​ ​and​ ​blanking​ ​plugs​ ​are​ ​good​ ​working​ ​order

• check​ ​all​ ​surface​ ​communications

Wet​ t​ esting​ ​should​ i​ nclude​ c​ hecks​ o

​ f​ ​the​ f​ ollowing:
• Thrusters (including all directions)

• USBL​ ​and​ ​internal​ ​navigation​ ​(e.g.​ ​compass​ ​and​ ​avoidance​ ​sonar)

• cameras​ ​and​ ​strobes

• avoidance/scanning sonar (if fitted)

• through-water​ ​communications

Conduct​ ​ROV​ ​transects

Pre-deployment

• Transects ​should ​only ​be ​undertaken ​in ​areas ​where ​the ​substratum ​is ​known, preferably ​i​n
the ​form ​of ​multibeam ​mapping, so as ​to ​avoid ​entrapment ​and ​potential ​loss ​of ​ROV. ​Do not
deploy ​blind​, ​as ​this ​increases ​the ​risk ​of ​equipment ​loss ​and ​damage, ​as ​well ​as
unnecessary​ ​impact​ ​on​ ​potentially​ ​vulnerable​ ​ecosystems.

• Once ​final ​transect ​locations ​have ​been ​determined, ​provide ​the ​locations ​of ​the ​transects
(usually ​in ​ESRI ​shapefile ​format or start and end waypoints) ​and ​associated ​multibeam
maps ​(in ​geotif ​format) ​to ​the ​ROV crew responsible ​for piloting ​missions. ​Cross-check ​the
uploaded ​transect ​corresponds ​to ​the ​correct ​area ​on ​the ​geotif ​(i.e. ensure ​the ​geographic
coordinates​ ​are​ ​defined​ ​for​ ​all​ ​spatial​ ​data).

• Discuss the desired target location and the feasibility of deploying at that location. Main
items to take into account are:

o Terrain. To minimise the risk of a deployment in highly rugose seafloor (e.g. walls) it
is recommended that transects should be conducted up or along walls. Also consider
the water visibility. If there are any large ridges, boulders, drop-offs, etc. along the

Page | ​10
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

proposed transect with minimal forward vision (< 10 m) there may not be a large
margin for avoidance.

o Currents/weather/sea state. During the transect, the USBL display will show the boat
and ROV position, allowing the skipper and ROV pilot to discuss tracks and adjust
speed if required. This can limit the manoeuvrability of the ship and depending on the
direction of the prevailing wind and sea, is not always possible on a particular
heading. As the sea-state and swell can affect the ships manoeuvrability when
travelling at low speeds it is essential to regularly check the weather forecast to
ensure the sea state is acceptable and the platform can be safely deployed and
retrieved.

o Depth. Be aware of the depth limitations of the ROV and the length of the tether.

o Entanglement procedure. Discuss potential entanglement procedure (detailed below)

making sure each person is familiar with their role.

• Prepare for ​ROV ​launch ​and ​recovery ​on ​deck ​and ​ensure ​only ​essential ​personnel
participate​ ​in​ ​its​ ​preparation​ ​and​ ​deployment.

• Place USBL​ ​transceiver​ ​in​ ​water​ ​and​ ensure ​functionality.

• Ensure tether is connected, turn on ROV and run all surface checks of the ROV as per
manufacturer's requirements.

• Check camera settings (if external cameras are being used).

• Check data sheet is ready (note site, camera numbers and memory card numbers).

• Turn external cameras on, check there is battery and storage space available.

• Insert cameras into housings, check that the housing is dry and that there is no sand, hair or
other objects obstructing the o-rings, and ensure there is a good seal and the o-ring is not
pinched.

• Film data sheet or clapper board so that the site/location is identifiable at the beginning of
the video (only needed if cameras are external to ROV).

• Film diode, or use clapper board, or alternative device to synchronise video footage.

• Correctly insert the deployment​ ​release​ ​pin (if using).

ROV deployment

1. Vessel​ ​master​ ​must​ ​ensure​ ​the​ ​vessel​ ​is​ ​positioned​ ​at​ ​the​ ​start​ ​of​ ​the​ ​transect​ ​location.

2. Following ​the ​signal ​to ​deploy ​from ​the ​vessel ​Master, ​use ​the ​crane ​and/or ​A-Frame ​to
lift ​and ​guide ​the RO​V ​from ​the ​deck ​into ​the ​water. Or, if using a small observation class
ROV signal to the deckhand to gently place the ROV in the water ensuring the thrusters
are disabled or unarmed.

3. Minimise ​the ​time ​taken f​ rom ​when ​the ​ROV i​ s ​out ​of ​reach, ​to ​when ​it ​is ​lowered ​in t​ he
water, ​so ​as ​to ​reduce p ​ otential ​swing ​and i​ mpact ​against ​the ​vessel. As soon as the

Page | ​11
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

ROV enters the water, pilot it below or away from the vessel to avoid drifting into or over
the ROV.

4. Using appropriate software (see Pre-Survey Preparations), rapidly pilot the ROV ​to the
seabed​ ​and​ at the ​start​ ​of​ ​transect​ ​location to avoid drifting off the starting point.

5. Confirm imagery and positional data ​are being ​recorded ​where ​possible ​(e.g. ​recording
indicators,​ ​hard​ ​drive​ ​operating).

ROV maneuvering

1. At the start of the transect flash lights or something similar should be used to indicate the
start of the transect. This is important to be able to sync footage with a USBL track (if
used) when the cameras are not integrated into the ROV.

2. The ROV should be positioned so that it is on course for the transect trajectory before the
transect start-point, so that movements are stable when it reaches the start of the
transect. Once the ROV is following the planned transect track the pilot can switch to
‘auto-heading’ to hold course (if available).

3. The ​flight ​elevation ​of the ​ROV ​should be ​set (either manually or automatically) ​and
maintained ​at ​~ 1 ​m ​from ​the ​seafloor ​to ​facilitate ​a ​consistent ​field ​of ​view (i.e. ~5 m
width transect for mobile organisms with this width being measurable if calibrated stereo
cameras are fitted). Try to maintain a constant forward momentum of ~ 0.5-1 ms​-1 ​(1-2 kt).
Avoid stopping or chasing fish/organisms off the transect. Also avoid disturbing the
substratum as sediment clouds will obscure the image ​(Hitchin et al. 2015)​. However, if
elevation is too high then fish observations are likely to be reduced. ​These factors need to
be informed by the 'survey question', camera type and performance, illumination type and
output power, etc.

4. Ask the vessel's ​Master to ​follow ​the ROV during ​transects. If current/wind is too strong
then the vessel may need to anchor. A sea anchor or drop/clump weight can be used to
reduce the effects of vessel and tether drag, respectively. If survey designs require
live-boat procedures it is more likely that operations would cease if weather conditions
deteriorate too much, unless there was an alternate survey objective that could be
accomplished at anchor.

5. Make sure that the tether is kept away from vessel propellers at all times. A crew member
must maintain tether management at all times. Clear and uninterrupted communication
between ROV pilot, tether crew and vessel master must be maintained at all times.

6. Monitor ​weather ​forecast ​conditions ​prior ​to ​and ​during ​deployment ​to ​maintain a safe
working ​environment. ​Consider ​aborting ​operations ​if ​local ​weather ​and ​forecast
conditions​ ​are​ ​marginal.

7. Vessel/ROV maneuvering is a nuanced topic, with most work class ROV teams having
their own protocols. Importantly, planning a transect in a fashion that avoids positioning
the ROV between the vessel and known entanglement risks (ledges, pinnacles, fishing
gear, etc) is the most important general protocol. The goal being to avoid a situation
where the vessel drags the tether into the entanglement because the vessel is typically
less maneuverable and has less situational awareness of the terrain. Current direction
and speed become forces that influence how easy this is to accomplish but many other

Page | ​12
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

factors may dictate how a team chooses to mitigate this risk. Before each transect
operators should discuss with vessel master if the entrapment risks associated with the
seafloor are low enough for the transect to be completed successfully.

ROV retrieval

1. When ​the ​transect ​is ​complete ​or ​if ​the ​transect ​is ​being ​aborted, ​advise the ​vessel
Master​ ​of​ the ​intention​ to ​retrieve​ ​the​ ​ROV.

2. Watch ​for the ​ROV to ​resurface, ​ensuring ​only ​required ​personnel are ​near ​open ​transom.
Avoid approach​ing the ​ROV looking​ ​into​ ​the​ ​sun​ ​as​ ​this​ ​increases​ ​the​ ​risks​ ​of​ ​collision.

3. Use a ​grapple ​hook ​to ​connect the ​lift ​line ​to the ​ROV for ​retrieval. Depending on the size
of the ROV, a​t ​least ​three ​personnel ​should be present ​with ​hooks ​to ​avoid the ​ROV
colliding​ ​with​ the ​vessel ​[Recommended]​.

4. Shut​ ​down​ the ​ROV. (Dis)connect​ relevant ​tether or​ ​data​ transfer ​cables.

5. For the last ​transect ​of the ​day, if available, ​wash ​down the ​ROV ​with ​freshwater and
unplug​ the ​USBL.

6. Raise the ​USBL ​transducer ​(if ​pole ​mounted) before moving the vessel to the next
location.

Procedures​ f​ or​ ​seabed​ e

​ ntanglement​ o
​ r​ l​ oss​ o
​ f​ ​communications​ ​with​ ​ROV
Potential ​entanglement ​of ​the ​ROV ​is ​always ​a ​possibility. ​The ​following ​procedures ​should ​be
followed​ ​upon​ ​entanglement/loss:

1. Log​ the ​last​ ​known​ ​position​ ​of​ the ​ROV.

2. If the ROV appears entangled (i.e. not moving) try to maneuver the vehicle so as to be able
to follow back along the tether to see if and where the tether has become snared. If the ROV
is trapped under a ledge/cave, or ensnared in a fishing line or kelp, a dive team or additional
ROV may be required. It may be required that the tether is disconnected from the vessel
before recovery equipment is launched. In such circumstances, the tether end should be
temporally sealed and attached to surface floats which will reduce water damage to the
tether.

3. Ensure the vessel is maintaining position and is not adding increased tension to entangled
tether.

4. Ensure that you check ROV thoroughly for damage before redeployment.

Completion​ ​of​ o
​ perations

Prior​ ​to​ ​any​ ​vessel​ ​movement​ ​or​ ​engine​ ​start-up,​ ​operators​ ​should​ ​check the following:

● All​ ​equipment​ ​is​ ​clear​ ​of​ ​the​ ​water, including the USBL transducer pole.

● ROV is shut down.

● All gear is safely stowed.

Page | ​13
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

● All power and data cables are (dis)connected.

● External cameras are turned off.

● An “All Clear to Move” command is given to the vessel Master when the ROV team is
satisfied it is OK for the vessel to move on.

Onboard data processing and storage

5. Once ​the ​ROV ​transect ​is ​complete, ​it ​is ​good ​practice ​to ​download ​associated ​raw ​imagery
and ​associated ​positional ​data. ​Imagery ​and ​associated ​positional ​data ​should ​be ​checked ​to
ensure​ ​no​ ​failures​ ​have​ ​occurred,​ ​including​ ​but​ ​not​ ​limited​ ​to​ ​the​ ​following:

● Miss-timing between image capture and strobes (i.e. dark/black imagery)

● Failure of one of the stereo cameras

● Failure of positional logging

2. Name data files according to established conventions. File naming conventions are vital for
ensuring both efficient and effective management of field data and its integration into
appropriate data management repositories. It is important to note that these conventions will
differ among agencies and academic institutions. Examples of stereo imagery naming
conventions are provided in Chapter 5 for benthic stereo-BRUVs.

3. Ensure ​accurate ​recording ​of ​metadata. ​Metadata ​are ​descriptive ​data ​sources ​composed ​of
information ​that ​may ​be ​used ​to ​process ​the ​images ​or ​information ​therein and for archiving
data on data portals (​Durden ​et al. ​2016). ​While ​it ​is ​important ​to ​follow ​agency ​specific
protocols ​for ​capturing ​metadata, ​it ​is ​also ​essential ​that ​metadata ​are ​sufficient ​enough ​in
detail ​to ​satisfy ​conformance ​checks ​for ​subsequent ​data ​release ​via ​AODN. ​Minimum ​data
for​ ​each​ ​transect​ ​should​ ​contain​ ​as​ ​follows:

● Campaign (i.e. Survey identifier)

● Station/event number

● Platform

● Latitude and longitude (WGS 1984 in decimal degrees with a minimum of 6 decimal
places ​[Recommend]​)

● Altitude in m

● Depth in m

● Time and date stamp in UTC

● AUV orientation (roll, pitch, heading) in degrees

● Precision details (e.g. type of navigation system used and its associated errors)

Page | ​14
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

● Data provenance

4. Backup data. This is necessary to ensure all data collected in the field is safely returned and
securely backed-up at host facilities, prior to quality control and public release. Onboard
copies of data should be made as soon as practical following acquisition. When operating
external to a network, it is recommended that all data be backed up on a RAID or a NAS that
contain built-in storage redundancy in case of hard-drive failure. A duplicate copy of all data
can be copied onto external hard drives for transportation back to host facilities
[Recommended].​

Post-Survey Procedures
Imagery collected by ROV can be either in the form of video footage or still imagery. What type of
imagery is collected and annotated is dependent on the aims or hypothesis. Each has its
advantages and disadvantages. Below outlines the workflow for both video and still imagery.

Processing and annotation of video footage

The annotation of ROV imagery will vary according to survey aims and hypotheses, as well as
availability of staff and time for this activity. Below we provide standards for annotating ROV
imagery for fish based on stereo imagery and habitat and communities based on downward-facing
stills.

ROV based stereo-video should be treated similar to stereo-DOV footage ​(Goetze et al. 2019)​.
Where possible and in line with survey aims and hypotheses, species composition, abundance and
length data for all species should be recorded.

For studies focussing on fish or overall community composition, every fish along a transect should
be measured (where possible). However, fish that occur in large schools, and are of similar size,
can be attributed to binned length measurement using the Number field associated with each length
in EventMeasure (or equivalent if analysed using other softwares). It is important to document the
range from camera as this is likely to change between regions/ecosystems. This information is
included in the standard outputs of EventMeasure and is imported by default into GlobalArchive
(see below).

There are several software packages available, but it is important the output from the analysis of
data is in the same or similar formats to facilitate comparison of data between campaigns, studies,
and organisations. The most commonly used annotation software is EventMeasure from SeaGIS
(​https://www.seagis.com.au​). If afforded, then the EventMeasure software is recommended, unless
your organisation already has an alternative established stereo-video annotation workflow (e.g.
AIMS). The essential information produced by such annotation software includes three main
outputs:

● Point information

● Length measurements

● 3-D point information

Page | ​15
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

Point information is typically used to calculate abundance values, while length and 3D point
information is used to calculate length and biomass metrics. EventMeasure has established queries
built-in to produce typical metrics over a user defined period within the footage. Periods can be used
to define the start and end transects if multiple are conducted in the same deployment. In addition,
EventMeasure annotation datasets held within GlobalArchive (​http://globalarchive.org/​) can be
queried in a similar fashion to produce such metrics (see the manual for ​GlobalArchive​).

Type of fish length (e.g., fork length or total length for fish and disc length for rays) should be clearly
indicated as part of the adequate annotation information for each transect/campaign.

Processing and annotation of downward facing still imagery

A​ ​general​ ​workflow​ ​for processing and annotating ​epibenthos still imagery ​can​ ​be​ ​found​ ​in​ ​Williams
et al. (2012)​.​ ​Key​ ​requirements​ ​for​ ​raw​ ​image​ ​processing​ ​and​ ​positional​ ​data​ ​are​ ​as​ ​follows:

• It ​is recommended ​that ​at ​least ​one ​of ​the ​stereo ​images ​is ​in ​colour ​and ​enhanced ​following
similar​ ​procedures​ ​as​ ​outlined​ ​by​ ​Bryson et al. (2016)​.

• Ideally all stereo images should be georectified similar to ​Williams et al. (2012)​. If not stereo
then processing routines can be found in ​Morris et al. (2014)​.

• Positional ​data ​should ​be ​post-processed. This could include ​using ​Simultaneous
Localisation ​and ​Mapping ​(SLAM) ​as ​demonstrated ​in ​(Barkby et al. 2009) and ​(Palomer et
al. 2013)​ for AUV imagery.

Annotation of ​individual ​images ​can ​be ​done ​using ​a ​number ​of ​annotation ​software ​tools. ​Examples
include, ​Transect ​Measure, BenthoBox, ​Coral ​Point ​Count, ​CoralNet ​and ​Squidle+. ​For ​national
consistency ​Squidle+ ​(​http://squidle.org​) is ​recommended ​as ​it ​is free and allows ​for ​different
approaches ​in image subsampling (such as a spatially balanced selection), ​which is important to
minimise spatial autocorrelation and ​influence ​inferences ​from ​data ​(Monk et al. unpublished data),
as ​well ​as ​stratified ​and ​random ​point ​count ​distribution ​on ​images. ​Squidle+ will also ​automatically
import ​the ​ROV data ​once ​it ​is ​linked to a data portal (such as IMAS data repository) making ​it
ready ​for ​analysis. ​Squidle+ also ​has ​tools ​for ​exploring ​survey ​data ​as ​well ​as ​analysis. ​In ​addition,
it ​supports ​multiple ​annotation ​schemes, ​and ​will ​provide ​consistency ​through ​translation ​between
schemes,​ ​which​ ​is​ ​an​ ​important​ ​point​ ​that​ ​differentiates​ ​Squidle+.

There​ ​are​ ​three​ ​approaches​ ​recommended​ ​for​ ​annotating​ ​imagery​ ​from​ ​ROVs:

• Annotation​ ​of​ ​individual​ ​images

• Annotation of​ ​photomosaics

• Extracting structural​ ​complexity​ ​from​ ​orthomosaics

Annotation of individual images or photomosaics can be undertaken using three methods:

• Full ​assemblage ​scoring ​of ​imagery across ​space ​and ​time. ​It ​is ​important ​to ​note ​that ​this ​is
a ​time-consuming process​, ​requiring ​a ​lot ​of ​replicate ​images ​to ​be ​scored ​to ​enable
sufficient ​power ​to ​detect ​biologically ​meaningful ​change ​as ​most ​morphospecies ​cover ​< ​10
% ​of an image. ​This ​approach ​appears ​to ​be ​good ​for ​delineating ​bioregional ​and ​cross-shelf
patterns ​at ​a ​morphospecies ​(Monk et al. unpublished data) and ​CATAMI ​(Althaus et al.

Page | ​16
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

2015) level ​(Monk et al. 2016, James et al. 2017)​. ​This ​approach ​will ​no ​doubt ​be ​effective ​in
choosing an initial​ ​suite​ ​of​ ​indicators​ ​for​ ​national​ ​level​ ​monitoring​ ​and​ ​reporting.

As a general guideline, and dependent on the survey question, we recommend that 25

random points per image from at least 50 images per transect are a good starting point for
recording most morphospecies present within images (based on Perkins et al. 2016). It is
important to note that the properties of the organism themselves will also influence the
number of points/images to score. Obviously morphospecies that are less abundant require
more effort, but also the 'clumpiness' of species will affect the scoring effort needed ​(Perkins
et al. 2016)​. ​(Van Rein et al. 2011) and Perkins et al. (2016) suggest that, while a higher
number of points per image can increase the detection rate of more organisms within an
image, increasing the number of scored images using fewer points is likely to have a similar
(or greater) effect. Ideally, increasing both the number of images scored and the number of
points scored within an image would result in greater power ​(Roelfsema et al. 2006)​, but
preference is usually for increasing the number of images ​(Perkins et al. 2016)​.
Unfortunately, the adoption of this approach is likely to result in substantial increases in
processing time and thus cost.

• Targeted ​scoring ​of ​indicators ​or ​proxies ​(such ​as ​grouping ​fine ​level ​morphospecies ​into
broader ​level ​CATAMI ​classes; Monk et al. unpublished data). ​This approach ​has ​been
shown ​to ​work ​very ​well ​at ​an ​indicator ​morphospecies level ​for ​detecting ​change ​at ​a
regional ​level ​(e.g. AUV imagery used by ​Perkins et al. 2017) ​as ​well ​as ​for ​detecting
invasive ​species ​trends ​(Whitfield et al. 2007)​. ​Since ​this ​approach ​requires ​substantially
less ​effort ​to ​score ​each ​image, ​more ​images ​(i.e. ​often ​all ​images) ​can ​be ​scored ​and, ​thus,
increasing ​statistical ​power. The drawback is that narrower understanding of the
environment is produced.

• Automated analysis of imagery potentially provides a cost-effective alternative to annotating

imagery from ROVs. It is important to note that automated imagery analysis is a relatively
new, and largely developmental, way of annotating images. Despite this, some studies
suggest that coral and macroalgae can be reliably identified using automated image analysis
(Table 4.1 in Chapter 4 AUV).

• The ​last ​approach ​to ​annotating ​ROV imagery ​involves ​the ​extraction ​of ​3D ​structural
information ​from ​stereo ​images ​using ​structure ​from ​motion ​techniques ​(Marcon 2014)​. ​This
approach ​works ​particularly ​well ​for ​sessile ​species ​to ​track ​changes ​in ​growth ​form ​through
time ​at ​a ​fine scale ​(Price et al. 2019)​. It also has application for vertical structure such as
reef walls or artificial structures​ ​(Robert et al. 2017)​.

Data curation and quality control

Data quality control at both the collection and annotation stage is critical. For fish datasets we
suggest that the same protocols outlined in section 5.7.3 in Chapter 5 (benthic stereo-BRUVs) be
followed, whereby strict training of new annotators is undertaken and thorough checks of species
IDs are done by trained taxonomists. It is crucial to include the salary or in-kind contribution of
taxonomists into project budgets. For epibenthic sessile communities we recommend that the same
protocols outlined in section 4.6.3 in Chapter 4 (AUV) be followed, with, most importantly, the
annotation schema needs to be consistent between studies. Where possible morphospecies and
associated CATAMI parent classes should be used ​[Recommended].​ An ​initial ​morphospecies
catalogue for southeastern shelf waters ​is ​currently ​held ​and ​maintained ​at ​the ​Institute ​for ​Marine

Page | ​17
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

and ​Antarctic ​Studies ​(IMAS) ​(contact ​Assoc. Prof. Neville ​Barrett ​or ​Dr ​Jacquomo ​Monk). ​Clearly,
other annotation schemas are available and can be applied. Where existing protocols prevent the
adoption of this approach the alternative schema must be mapped to CATAMI so that comparisons
can be made with previous studies or between regions. Translations between schema can be
readily applied within Squidle+. The quality control of all annotations of epibenthic sessile organisms
undertaken by novice scorers should be assessed against an experienced analyst or machine
learning algorithm (e.g. using confusion matrices; see Figure 4.4 in Chapter 4). Similarly, all
datasets annotated by multiple people, even skilled scorers, should be tested for observer bias. If
there are significant differences among annotators it is important to correct discrepancies. This can
be done by re-examining the images to ensure an agreement can be reached between annotators.
Alternatively, if an agreement cannot be reached, then the miss-classified item could be potentially
grouped into a higher level CATAMI class.

Data release

Many national marine observing programs (for example IMOS through the Australian Ocean Data
Network (AODN), or the Marine Geoscience Data System (MGDS) in the USA) routinely store
imagery online in an openly accessible location. ​Squidle+ is a centralised online platform for
standardised analysis and annotation of georeferenced imagery and video. Squidle+ operates
based on flexible distributed data storage facilities (i.e. imagery can be stored anywhere in an
openly accessible online location) to reduce data duplication and inconsistencies, and provides a
flexible annotation system with the capability to translate between different annotation schemes.

Following the steps listed below will ensure the timely release of imagery and associated annotation
data in a standardised, highly discoverable format.

1. Create a metadata record describing the data collection. Provide as much detail as possible on
the deployment (either directly in the metadata record itself, or in the form of attached field
sheets as .csv, .txt or similar). Details of minimum metadata requirements are provided in the
On-board Data Storage section above. Publish metadata record(s) to the ​Australian Ocean Data
Network (AODN) catalogue as soon as possible after metadata has been QC-d. This can be
done in one of two ways:

● If metadata from your agency is regularly harvested by the AODN, follow agency-specific
protocols for metadata and data release.

● Otherwise, metadata records can be created and submitted via the ​AODN Data Submission
Tool​. ​Note that user registration is required, but this is free and immediate.

Lodging metadata with AODN in advance of annotation data being available is an important step
in documenting the methods and location of acquired imagery and enhancing future
discoverability of the data.

2. Upload raw imagery from the survey to a secure, publicly accessible online repository (​contact
AODN​ if you require assistance in locating a suitable repository).

3. Create a ​Squidle+ campaign as soon as possible after imagery is uploaded, choose the most
appropriate annotation schema, and commence annotation of imagery.

Page | ​18
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

4. Add links to the location of the Squidle+ campaign to the previously published metadata record.
You may also wish to attach or link a copy of the annotation data directly to the record.

5. Produce a technical or post-survey report documenting the purpose of the survey, sampling
design, sampling locations, sampling equipment specifications, annotation schema (e.g.
morphospecies, CATAMI, etc.), and any challenges or limitations encountered. Provide links to
this report in all associated metadata ​[Recommended]​.

Data analysis

The ​breadth ​of ​research ​questions ​precludes ​any ​detailed ​advice ​on ​the ​analysis ​of ​data ​from ​ROV
transects. ​However, ​one ​common ​attribute ​of ​the ​image-based ​data ​that ​will ​have ​to ​be ​considered
for ​all ​analyses ​is ​spatial ​proximity. ​The ​closeness ​of ​images, ​within ​and ​sometimes ​between
transects (for example if triangle or clover-leaf transect designs or subsets of longer transects are
used), ​means ​that ​image ​data ​are ​unlikely ​to ​be ​independent ​(due ​to ​spatial ​autocorrelation). ​Yet,
this ​is ​an ​assumption ​that ​many ​statistical ​methods ​rely ​upon. ​The ​failure ​to ​meet ​this ​assumption
means ​that ​the ​inferences ​from ​the ​statistical ​analysis ​may ​be: ​(i) ​over-confident, ​e.g. ​having ​a
p-value ​that ​is ​too ​small; ​(ii) ​biased, ​i.e. ​the ​estimates ​do ​not ​reflect ​the ​truth; ​(iii) ​both, ​or; ​(iv) ​no
effect. ​Obviously, ​the ​fourth ​category ​is ​what ​a ​researcher ​hopes ​for, ​but ​it ​is ​improbable ​and ​must
be ​validated. ​However, ​if ​it ​is ​known ​that ​the ​study ​organism ​exhibits ​particularly ​low ​autocorrelation
then​ ​the​ ​analysis​ ​need​ ​not​ ​consider​ ​it​ ​explicitly.

Methods ​to ​analyse ​data, ​accounting ​for ​autocorrelation ​are ​available. ​These ​include ​geostatistical
models ​(Foster et al. 2014)​. ​However, ​in ​certain ​situations ​subsampling ​images ​will ​help ​(Mitchell et
al. 2017)​, ​but ​not ​necessarily ​alleviate ​completely. ​Further, ​if ​the ​study ​is ​for ​a ​broad ​area, ​where
transects ​are ​small ​and ​are ​well-separated, ​then ​amalgamating ​data ​to ​transect ​level ​may ​also ​be
appropriate. The issues of spatial auto-correlation should also be considered if longer transects are
being broken up into smaller sections for analysis (as is commonly done in the oil and gas sector).

Some effort should be made to estimate sources of error inherent in navigational (USBL) systems
(and/or other geo-referencing methods) and understand how these errors affect the overall target
parameter estimation and variability (see Karpov 2006, Rattray et al., 2017, Mitchell et al. 2017).

Field​ ​Manual​ ​Maintenance

At the time of writing this manual, there is currently no support for future versions of this manual.
However, in accordance with the universal field manual maintenance protocol described in Chapter
1 of the Field Manual package, if such support arises, this manual will be updated in the future as
Version 3. Updates will reflect user feedback and new developments (e.g. data discoverability and
accessibility). Version 3 will also detail subsequent version control and maintenance.

​The​ ​version​ ​control​ ​for​ ​Chapter​ ​10​ ​(field​ ​manual​ ​for​ ROV​s)​ ​is​ ​below:

Version Number Description Date

0 Submitted for review (NESP Marine Hub, external 25 May 2020

reviewer as listed Acknowledgements.

Page | ​19
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

1 There was no ROV manual included in Version 1 of the n/a

field manual package

2 Publicly released as Chapter 10 on July 2020

www.nespmarine.edu.au​ through online portal

Acknowledgements
The authors are grateful to Michael Prall for reviewing this chapter. Alex Ingle and Schmidt Ocean
Institute provided images from the ROV SuBastien. Darryn Sward (IMAS) is thanked for images of
observation class ROVs. Front cover images (left to right) supplied by Darryn Sward, Joel Williams,
Schmidt Ocean Institute.

References
Althaus F, Hill N, Ferrari R, Edwards L, Przeslawski R, Schönberg CHL, Stuart-Smith R, Barrett N, Edgar G, Colquhoun J,
Tran M, Jordan A, Rees T, Gowlett-Holmes K (2015) A standardised vocabulary for identifying benthic biota
and substrata from underwater imagery: The CATAMI classification scheme. PLoS One 10:e0141039.
Baker KD, Haedrich RL, Snelgrove PVR, Wareham VE, Edinger EN, Gilkinson KD (2012) Small-scale patterns of
deep-sea fish distributions and assemblages of the Grand Banks, Newfoundland continental slope. Deep Sea
Res Part I 65:171–188.
Barkby S, Williams S, Pizarro O, Jakuba M (2009) An efficient approach to bathymetric SLAM. In: ​2009 IEEE/RSJ
International Conference on Intelligent Robots and Systems​. p 219–224
Bond T, Partridge JC, Taylor MD, Cooper TF, McLean DL (2018) The influence of depth and a subsea pipeline on fish
assemblages and commercially fished species. PLoS One 13:e0207703.
Boutros N, Shortis MR, Harvey ES (2015) A comparison of calibration methods and system configurations of underwater
stereo-video systems for applications in marine ecology. Limnol Oceanogr Methods 13:224–236.
Bryson M, Johnson-Roberson M, Pizarro O, Williams SB (2016) True Color Correction of Autonomous Underwater Vehicle
Imagery. J Field Robotics​ 33:853​–​874.
Cánovas-Molina A, Montefalcone M, Bavestrello G, Cau A, Bianchi CN, Morri C, Canese S, Bo M (2016) Research
papers. Cont Shelf Res C:13–20.
Capocci R, Dooly G, Omerdić E, Coleman J, Newe T, Toal D (2017) Inspection-Class Remotely Operated Vehicles—A
Review. J Mar Sci Eng 5:13.
Dauble A (2006) Characterization of Copper Rockfish (Sebastes caurinus) Habitat in Marine Protected Areas in the San
Juan Islands. Friday Harbor Laboratory.
Foster SD, Hosack GR, Hill NA, Barrett NS, Lucieer VL (2014) Choosing between strategies for designing surveys:
autonomous underwater vehicles. Methods Ecol Evol 5:287–297.
Foster SD, Hosack GR, Lawrence E, Przeslawski R, Hedge P, Caley MJ, Barrett NS, Williams A, Li J, Lynch T,
Dambacher JM, Sweatman HPA, Hayes KR (2017) Spatially balanced designs that incorporate legacy sites.
Methods Ecol Evol 8:1433–1442.
Foster SD, Hosack GR, Monk J, Lawrence E, Barrett NS, Williams A, Przeslawski R (2019) Spatially-Balanced Designs
for Transect-Based Surveys. Methods Ecol Evol​ 11: 95-105.
García-Alegre A, Sánchez F, Gómez-Ballesteros M, Hinz H, Serrano A, Parra S (2014) Modelling and mapping the local
distribution of representative species on the Le Danois Bank, El Cachucho Marine Protected Area (Cantabrian
Sea). Deep Sea Res Part 2 Top Stud Oceanogr 106:151–164.
GESAMP (2019) Guidelines or the monitoring and assessment of plastic litter and microplastics in the ocean (Kershaw
P.J., Turra A. and Galgani F. editors), (IMO/FAO/UNESCO-IOC/UNIDO/WMO/IAEA/UN/UNEP/UNDP/ISA Joint
Group of Experts on the Scientific Aspects of Marine Environmental Protection). Rep. Stud. GESAMP No. 99,
130p.
Goetze JS, Bond T, McLean DL, Saunders BJ, Langlois TJ, Lindfield S, Fullwood LAF, Driessen D, Shedrawi G, Harvey
ES (2019) A field and video analysis guide for diver operated stereo-video. Methods Ecol Evol 10:1083–1090.
Hitchin R, Turner JA, Verling E (2015) Epibiota Remote Monitoring from Digital Imagery: Operational Guidelines.

Page | ​20
Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

Huvenne VAI, Robert K, Marsh L, Lo Iacono C, Le Bas T, Wynn RB (2018) ROVs and AUVs. In: ​Submarine
Geomorphology​. Springer Geology, Micallef A, Krastel S, Savini A (eds) Springer, Cham, Switzerland, p 572
Ierodiaconou D, Monk J, Rattray A, Laurenson L, Versace VL (2011) Comparison of automated classification techniques
for predicting benthic biological communities using hydroacoustics and video observations. Cont Shelf Res
31:S28–S38.
James LC, Marzloff MP, Barrett N, Friedman A, Johnson CR (2017) Changes in deep reef benthic community composition
across a latitudinal and environmental gradient in temperate Eastern Australia. Mar Ecol Prog Ser 565:35–52.
JNCC (2018) Remotely Operated Vehicles for use in marine benthic monitoring. Marine Monitoring Platform Guidelines
No.1. JNCC, Peterborough. ISSN 2517-7605.
Karpov KA, Lauermann A, Bergen M, Prall M (2006) Accuracy and Precision of Measurements of Transect Length and
Width Made with a Remotely Operated Vehicle. Mar Technol Soc J 40:79–85.
Lacharité M, Metaxas A, Lawton P (2015) Using object-based image analysis to determine seafloor fine-scale features
and complexity: Computer vision to estimate seafloor complexity. Limnol Oceanogr Methods 13:553–567.
Lastras G, Canals M, Ballesteros E, Gili J-M, Sanchez-Vidal A (2016) Cold-Water Corals and Anthropogenic Impacts in La
Fonera Submarine Canyon Head, Northwestern Mediterranean Sea. PLoS One 11:e0155729.
Leckie SHF, Draper S, White DJ, Cheng L, Fogliani A (2015) Lifelong embedment and spanning of a pipeline on a mobile
seabed. Coast Eng 95:130–146.
Macreadie PI, McLean DL, Thomson PG, Partridge JC, Jones DOB, Gates AR, Benfield MC, Collin SP, Booth DJ, Smith
LL, Techera E, Skropeta D, Horton T, Pattiaratchi C, Bond T, Fowler AM (2018) Eyes in the sea: Unlocking the
mysteries of the ocean using industrial, remotely operated vehicles (ROVs). Sci Total Environ 634:1077–1091.
Marcon Y (2014) LAPMv2: An improved tool for underwater large-area photo-mosaicking. In: ​2014 Oceans - St. John’s.​ p
1–10
McLean DL, Partridge JC, Bond T, Birt MJ, Bornt KR, Langlois TJ (2017) Using industry ROV videos to assess fish
associations with subsea pipelines. Cont Shelf Res 141:76–97.
McLean DL, Parsons MJG, Gates AR, Benfield MC, Bond T, Booth D, Bunce M, Fowler AM, Harvey ES, Macreadie PI,
Pattiaratchi CB, Rouse S, Partridge JC, Thomson PG, Todd VLGT, Jones DOB (2020). Enhancing the
scientific value of industry remotely operated vehicles (ROVs) in our oceans. Front in Mar Sci 7: 220.
Mitchell PJ, Monk J, Laurenson L (2017) Sensitivity of fine-scale species distribution models to locational uncertainty in
occurrence data across multiple sample sizes. Methods Ecol Evol 8:12–21.
Monk J, Barrett NS, Hill NA, Lucieer VL, Nichol SL, Siwabessy PJW, Williams SB (2016) Outcropping reef ledges drive
patterns of epibenthic assemblage diversity on cross-shelf habitats. Biodivers Conserv 25:485–502.
Morris KJ, Bett BJ, Durden JM, Huvenne VAI, Milligan R, Jones DOB, McPhail S, Robert K, Bailey DM, Ruhl HA (2014) A
new method for ecological surveying of the abyss using autonomous underwater vehicle photography. Limnol
Oceanogr 12:795–809.
Palomer A, Ridao P, Ribas D, Mallios A, Vallicrosa G (2013) A Comparison of G2o Graph SLAM and EKF Pose Based
SLAM with Bathymetry Grids. IFAC Proceedings Volumes 46:286–291.
Perkins NR, Foster SD, Hill NA, Barrett NS (2016) Image subsampling and point scoring approaches for large-scale
marine benthic monitoring programs. Estuar Coast Shelf Sci 176:36–46.
Perkins NR, Foster SD, Hill NA, Marzloff MP, Barrett NS (2017) Temporal and spatial variability in the cover of deep reef
species: Implications for monitoring. Ecol Indic 77:337–347.
Pradella N, Fowler AM, Booth DJ, Macreadie PI (2014) Fish assemblages associated with oil industry structures on the
continental shelf of north-western Australia. J Fish Biol 84:247–255.
Price DM, Robert K, Callaway A, Lo lacono C, Hall RA, Huvenne VAI (2019) Using 3D photogrammetry from ROV video to
quantify cold-water coral reef structural complexity and investigate its influence on biodiversity and community
assemblage. Coral Reefs 38:1007-1021.
Przeslawski R, Alvarez de Glasby B, Smit N, Evans-Illidge L, Dethmers K (2013) Benthic Biota of Northern Australia:
SS2012t07 Post-survey Report. Record 2013/07. Geoscience Australia: Canberra
Przeslawski R, Foster S, Monk J, Langlois T, Lucieer V, Stuart-Smith R (2018) Comparative Assessment of Seafloor
Sampling Platforms. Report to the National Environmental Science Programme, Marine Biodiversity Hub.
Geoscience Australia. 57 pp.
Rattray A, Ierodiaconou D, Monk J, Laurenson LJB, Kennedy P (2014) Quantification of Spatial and Thematic Uncertainty
in the Application of Underwater Video for Benthic Habitat Mapping.” ​Marine Geodesy ​37: 315–36.
Robert K, Huvenne VAI, Georgiopoulou A, Jones DOB, Marsh L, D O Carter G, Chaumillon L (2017) New approaches to
high-resolution mapping of marine vertical structures. Sci Rep 7:9005.
Roelfsema C, Phinn S, Joyce K (2006) Evaluating benthic survey techniques for validating maps of coral reefs derived
from remotely sensed images.​ Proceedings of 10th International Coral Reef Symposium, 1771-1780.
Rountree RA, Juanes F (2010) First attempt to use a remotely operated vehicle to observe soniferous fish behavior in the
Gulf of Maine, Western Atlantic Ocean. Curr Zool 56:90–99.
Ryer CH, Stoner AW, Iseri PJ, Spencer ML (2009) Effects of simulated underwater vehicle lighting on fish behavior. Mar
Ecol Prog Ser 391:97–106.
Salvati E, Angiolillo M, Bo M, Bavestrello G, Giusti M, Cardinali A, Puce S, Spaggiari C, Greco S, Canese S (2010) The
population of Errina aspera (Hydrozoa: Stylasteridae) of the Messina Strait (Mediterranean Sea). J Mar Biol
Assoc U K 90:1331–1336.

Page | ​21
View publication stats

Marine Sampling Field Manuals for Monitoring Australia’s Commonwealth Waters Version 2

Schramm KD, Harvey E, Travers MJ, Goetze J, Warnock B, Sanders BJ (2020) A comparison of stereo-BRUV, diver
operated and remote stereo-video transects for assessing reef fish assemblages. J Exp Mar Biol Ecol 524:
151273.
Shepherd K (2001) Remotely Operated Vehicles (ROVs). In: ​Encyclopedia of Ocean Sciences (Second Edition).​ Steele
JH (ed) Academic Press, Oxford, p 742–747
Smith ANH, Anderson MJ, Pawley MDM (2017) Could ecologists be more random? Straightforward alternatives to
haphazard spatial sampling. Ecography 40:1251–1255.
Stoner AW, Ryer CH, Parker SJ, Auster PJ, Wakefield WW (2008) Evaluating the role of fish behavior in surveys
conducted with underwater vehicles. Can J Fish Aquat Sci 65:1230–1243.
Sward D, Monk J, Barrett N (2019) A Systematic Review of Remotely Operated Vehicle Surveys for Visually Assessing
Fish Assemblages. Front in Mar Sci 6:134.
Thomson PG, Fowler AM, Davis AR, Pattiaratchi CB, Booth DJ (2018) Some Old Movies Become Classics – A Case
Study Determining the Scientific Value of ROV Inspection Footage on a Platform on Australia’s North West
Shelf.​ ​Front in Mar Sci 5:471.
Thresher R, Althaus F, Adkins J, Gowlett-Holmes K, Alderslade P, Dowdney J, Cho W, Gagnon A, Staples D, McEnnulty
F, Williams A (2014) Strong depth-related zonation of megabenthos on a rocky continental margin (∼700-4000
m) off southern Tasmania, Australia. PLoS One 9:e85872.
Torriente A, González-Irusta JM, Aguilar R, Fernández-Salas LM, Punzón A, Serrano A (2019) Benthic habitat modelling
and mapping as a conservation tool for marine protected areas: A seamount in the western Mediterranean.
Aquat Conserv 15:263.
Van Rein H, Schoeman DS, Brown CJ, Quinn R, Breen J (2011) Development of benthic monitoring methods using
photoquadrats and scuba on heterogeneous hard-substrata: a boulder-slope community case study. Aquat
Conserv 21:676–689.
Whitfield PE, Hare JA, David AW, Harter SL, Muñoz RC, Addison CM (2007) Abundance estimates of the Indo-Pacific
lionfish Pterois volitans/miles complex in the Western North Atlantic. Biol Invasions 9:53–64.
Williams SB, Pizarro OR, Jakuba MV, Johnson CR, Barrett NS, Babcock RC, Kendrick GA, Steinberg PD, Heyward AJ,
Doherty PJ, Mahon I, Johnson-Roberson M, Steinberg D, Friedman A (2012) Monitoring of Benthic Reference
Sites: Using an Autonomous Underwater Vehicle. IEEE Robot Autom Mag 19:73–84.

Page | ​22