Research Report 1225: Review of underwater video data ... · i . ANNEX 1 – TASK 3, THE REVIEW OF EFFECTIVENESS OF UNDERWATER VIDEO MONITORING TECHNIQUES, EQUIPMENT AND PROCESSES

i

ANNEX 1 – TASK 3, THE REVIEW OF EFFECTIVENESS OF UNDERWATER VIDEO MONITORING TECHNIQUES, EQUIPMENT AND PROCESSES TO DATE Table of Contents Page 1. REVIEW OF EFFECTIVENESS OF UNDERWATER VIDEO MONITORING

TECHNIQUES, EQUIPMENT AND PROCESSES - ANDRITZ HYDRO HAMMERFEST TURBINE: DETAILED TECHNICAL REVIEW 1 1.1 Introduction 1 1.2 Approach 1 1.3 Overview of dataset 1 1.4 Assessment 2 1.4.1 Planning and design 2 1.4.2 Data collection 4 1.4.3 Powering of equipment, transmission of data and data storage 6 1.4.4 Video data review and analysis 6 1.5 Summary of findings 7

2. ANDRITZ HYDRO HAMMERFEST TURBINE: ESTIMATED CAMERA POSITIONS OF VIEW 11 2.1 Camera schematic layout 11 2.1.1 Camera 1 (on TTG 3 only as Camera 1 on TTG 2 was faulty): 11 2.1.2 Camera 2 11 2.1.3 Camera 3 11

3. REVIEW OF EFFECTIVENESS OF UNDERWATER VIDEO MONITORING TECHNIQUES, EQUIPMENT AND PROCESSES - SCOTRENEWABLES SR2000 TIDAL TURBINE: DETAILED TECHNICAL REVIEW 15 3.1 Introduction 15 3.2 Approach 15 3.3 Overview of dataset 15 3.4 Assessment 16 3.4.1 Planning and design 16 3.4.2 Data collection 16 3.4.3 Powering of equipment, transmission of data and data storage 24 3.4.4 Video data review and analysis 24 3.5 Summary of findings 26

4. REVIEW OF EFFECTIVENESS OF UNDERWATER VIDEO MONITORING TECHNIQUES, EQUIPMENT AND PROCESSES - VOITH HYDRO HYTIDE TIDAL TURBINE: DETAILED TECHNICAL REVIEW 30 4.1 Introduction 30 4.2 Approach 30 4.3 Overview of dataset 30 4.4 Assessment 31 4.4.1 Planning and design 31 4.4.2 Data collection 32 4.4.3 Powering of equipment, transmission of data and data storage 37 4.4.4 Video data review and analysis 38 4.5 Summary of findings 39

ii

List of Figures Page Figure 1.1 Estimated positions of cameras 4 Figure 1.2 Field-of-view of underwater video cameras from TTG 3 (top left = camera

1, top right = camera 2, bottom left = camera 3, bottom right = no camera) 5 Figure 2.1 Field-of-view of underwater video cameras from TTG 3 (top left = camera

1, top right = camera 2, bottom left = camera 3, bottom right = no camera) 11 Figure 2.2 Field-of-view from each camera on turbine, Camera 1 is below and

considerably closer to the rotor blades than Cameras 2 and 3 12 Figure 2.3 Estimated position of cameras on turbine 13 Figure 2.4 Direction cameras are pointing 13 Figure 2.5 Field-of-view of cameras on turbine 14 Figure 3.1 Schematics of the SR2000 device with camera placements 17 Figure 3.2 Sample fields of view from cameras 2, 3 and 5 (left to right respectively) 18 Figure 3.3 Camera 3 footage with dead pixels circled in red 19 Figure 3.4 ADCP adjacent to SR200 deployment 20 Figure 3.5 Position of turbine blades in field-of-view (left: port side nacelle; right:

middle leg brace) 21 Figure 3.6 Progression of biofouling on underwater camera lenses 23 Figure 4.1 Schematic of Voith HyTide turbine and support structure 32 Figure 4.2 Positions of cameras 33 Figure 4.3 Field-of-view from cameras (top left = camera 1 on monopile, top right =

camera 2 on nacelle, bottom left = camera 3 on nacelle, bottom right = no camera). Video still is from May when biofouling coverage was at its minimum 35

Figure 4.4 Video still taken from July when biofouling was at its greatest extent (top left = camera 1 on monopile, top right = camera 2 on nacelle, bottom left = camera 3 on nacelle, bottom right = no camera) 36

List of Tables Page Table 1.1 Summary of MeyGen data reviewed to inform the assessment of

effectiveness of techniques, equipment and processes 7 Table 1.2 Summary of the effectiveness of video monitoring techniques, equipment

and processes 7 Table 3.1 Summary of Scotrenewables data reviewed to inform the assessment of

effectiveness of techniques, equipment and processes 25 Table 3.2 Summary of the effectiveness of video monitoring techniques, equipment

and processes 26 Table 4.1 Summary of Voith HyTide data reviewed to inform the assessment of

effectiveness of techniques, equipment and processes 39 Table 4.2 Summary of the assessment effectiveness of techniques, equipment and

processes used in the Voith HyTide video data 39

1

1. REVIEW OF EFFECTIVENESS OF UNDERWATER VIDEO MONITORING TECHNIQUES, EQUIPMENT AND PROCESSES - ANDRITZ HYDRO HAMMERFEST TURBINE: DETAILED TECHNICAL REVIEW

1.1 Introduction

This section provides the detailed results for Task 3 in relation to underwater video monitoring undertaken around two of the Andritz Hydro Hammerfest (AHH) turbines deployed as part of Phase 1A of the MeyGen Tidal Array. The Phase 1A array is comprised of three Andritz Hydro Hammerfest (AHH) turbines and one Atlantis Resources Ltd (ARL) turbine. The aim of this task was to assess the effectiveness of underwater video monitoring techniques, equipment and processes used during underwater video monitoring activities. Note: The environmental monitoring activities being carried out as part of the Scottish Government Demonstration Strategy (SGDS) and Knowledge Transfer Partnership (KTP) with University of Aberdeen (UoA) are associated with the ARL turbine at the site and are not considered within this Briefing Note. The aim of this assessment is to learn from the experience of past deployments to identify recommendations that can be used by developers and regulators involved in planning future monitoring activities of tidal stream developments. The results also provide information to help guide future technology development and associated strategic funding streams. It should be noted that all comments and recommendations drawn in this report are in relation to the monitoring systems’ effectiveness in monitoring environmental effects. It should be recognised that many monitoring programmes around first deployments and arrays are focussed on technical monitoring and therefore not yet optimised for environmental monitoring purposes. The lessons learned and recommendations presented in this document should not therefore be taken as a critique of previous/current monitoring systems, but as a series of recommendations for optimising future environmental monitoring systems. 1.2 Approach

Several aspects of this video dataset have been reviewed to assess the effectiveness of the techniques, equipment and processes used at each stage in the process with the aim of identifying key challenges and lessons learned that can be used make recommendations to inform future environmental monitoring around tidal energy devices and arrays. The following stages of the monitoring programme were considered during the assessment:

• Planning and design; • Data collection (equipment, duration, performance, etc.); • Powering of equipment, transmission of data and data storage; and • Video data review and analysis.

1.3 Overview of dataset

The available video monitoring data is from two of the AHH turbines, namely Tidal Turbine Generator (TTG) 2 and TTG 3. The time period and format of the data differed depending on which TTG the data came from. For TTG 3, individual camera feed data were available for 22nd February 2017-04 March 2017 while an incorporated feed of all three cameras (as in Figure 1.2, see below) was available from 10:00 on 13th July 2017-11:00 on 14th July 2017. For TTG 2, data incorporating all three cameras was provided from 14:00 on 12th July 2017-11:00 on 14th July 2017.

2

1.4 Assessment

The following sections provide the results of the assessment undertaken in relation to the effectiveness of the equipment, techniques and processes implemented during each stage of the monitoring programme. 1.4.1 Planning and design

1.4.1.1 Details

In order to comply with the MeyGen Section 36 Consent and to begin to build an evidence base to reduce scientific uncertainty around key potential impacts, a Project Environmental Monitoring Plan (PEMP) (Rollings, Donovan and Eastham, 2016) 1 was designed in consultation with Marine Scotland, NatureScot and other statutory and non-statutory bodies in advance of the deployment of the MeyGen array which commenced in September 2016. The activities set out in this plan covered mitigation and monitoring activities that would attempt to collect empirical data from various types of equipment. The monitoring activities set out in the PEMP attempt to satisfy the requirements of Condition 12 of the Project’s Section 36 consent which was issued in September 2013. Condition 12 sets out the four main elements to be monitored:

• Hydro dynamics / benthic surveys, export cable route and turbine locations and modelling to validate EIA predictions;

• Collision / encounter interactions with the tidal turbines for diving birds, marine mammals and fish of conservation concern;

• Disturbance and displacement of birds, marine mammals and basking sharks during construction and operation; and

• Migratory salmonids. A Monitoring Steering Report (see Rollings, Donovan and Eastham, 2016) was produced and agreed on in May 2014 by an Advisory Group which included representation from the following organisations:

• MeyGen Ltd.; • Marine Scotland Licensing Operations Team (MS-LOT); • Marine Scotland Science (MSS); • Marine Scotland Planning and Policy (MSPP); • NatureScot; and • The Crown Estate.

The Monitoring Steering Report provides details of the prioritisation of both receptor and potential impact and refined the above four elements of Condition 12 to the objectives and concepts for the development of the PEMP. The Steering report took the lessons learned from previous research and identified where current (at the time of writing) and planned research would contribute to meeting the needs of the consent conditions and wider industry questions. Bringing these together, the report recommended more specific objectives and monitoring which were taken forward and developed for the PEMP and were appropriate for addressing the consent conditions.

1 Rollings, E., Donovan, C. and Eastham, C. 2016. MeyGen Tidal Energy Project Phase 1 Project Environmental Monitoring Programme. Available at: https://tethys.pnnl.gov/sites/default/files/publications/MeyGen-Environmental-Monitoring-2016.pdf

https://tethys.pnnl.gov/sites/default/files/publications/MeyGen-Environmental-Monitoring-2016.pdf

3

A number of monitoring activities were planned for the MeyGen Phase 1a deployment. It should be noted that the main sources of environmental monitoring data gathered at the MeyGen Tidal Array centre around the Atlantis turbine and not those covered in this report. MeyGen planned to gather environmental monitoring datasets in association with the Atlantis turbine by using two independent monitoring platforms, underwater video cameras and passive acoustic monitoring (PAM) equipment mounted on the support structure of the Atlantis turbine. Independent monitoring would then be produced which detail the results of these monitoring activities. The PEMP (Rollings, Donovan and Eastham, 2016) provided a comprehensive description of the planning and design of these activities. The video cameras mounted on the AHH turbines, to which this Briefing Note relates, provided additional monitoring data. 1.4.1.2 Lessons learned during planning and design

An accurate and detailed PEMP, such as that prepared for MeyGen Phase 1a, which clearly describes the objectives of the monitoring campaign and full technical details of the equipment set-up, including positions and orientation of the monitoring equipment, is necessary to enable accurate analysis of the gathered data and cataloguing of metadata. Consultation and sign off from the relevant regulatory bodies during PEMP development is a legal requirement in most relevant Section 36 Consents and is considered fundamental to ensure that the objectives of the monitoring activities are clearly identified and serve to address key consenting risks to the tidal stream energy industry. Considering this, the process undertaken by MeyGen Ltd included quarterly meetings with the MeyGen Advisory Group (AG). The AG facilitates discussion around the current state of knowledge of potential environmental impacts, which in turn informs discussion around the design of integrated monitoring solutions for multiple receptors, the potential opportunities and constraints of these and the various types of monitoring equipment required. This promotes consensus on priority monitoring activities i.e. what key questions related to consent need to be answered, and how monitoring activities could be designed to optimise performance and to effectively reduce scientific uncertainty with regard to the key questions. Note: the nascent state of the tidal stream energy industry means it is important to maximise learning opportunities that could help reduce scientific uncertainty around key impacts. Therefore, at an early stage, regulators and statutory nature conservation bodies (SNCBs) should encourage developers to gather environmental monitoring data with clear objectives that feed into the consenting process and provide direction on the relevance of design of developers’ monitoring plans. The regulator could suggest that any video data gathered are shared with them for research/data analysis purposes however this would require the regulator to have significant data storage capacity. An alternative would be for developers to provide monitoring reports to the regulator, with the developer storing the data in house. If there are commercial sensitivities associated with releasing datasets, clauses to delay their publication until a later date should be encouraged as it would enable their future analysis and subsequent reporting in scientific journals by academics and researchers. This would help to contribute to the body of evidence required to reduce scientific uncertainty.

4

1.4.2 Data collection

1.4.2.1 Details

Underwater video Three Seacam Ultra Wide Angle Monochrome UV cameras (two on the nacelle and one on the foundation) were installed on each of the AHH turbines primarily to monitor blade conditioning and also to monitor behaviour/presence of marine wildlife (marine mammals, fish and birds around the turbine). No lighting was installed to facilitate video camera monitoring outside daylight hours. The idea of using a lamp or flash was considered by the Advisory Group (see Section 1.4.1), however as it was thought there may be a risk of receptors being attracted to the light source, it was rejected. The PEMP does not indicate exactly where on the AHH turbine’s the cameras are located, only that they will be positioned at 120° around the nacelle to capture a 360° view of the turbine rotor. Through analysis of the video footage, camera positions have been estimated (see Figure 1.12). For a full description of the estimated positions of the cameras see Section 2.

Figure 1.1. Estimated positions of cameras

1.4.2.2 Lessons learned during data collection

Underwater video Positioning of cameras The three cameras, mounted on the nacelle and foundation, provide only partial coverage of the rotor swept area giving very narrow field-of-view and only the portion of the blades closest to the nacelle visible, as they rotate. The positions of the cameras are such that approximately 50% of the field of view from Camera 2 and approximately 40% of Camera 3 are taken up by the nacelle (see Figure 1.2), thereby not capturing the rotor swept area of the turbine blades. 2 Note: the term ‘foreshortening effect’ describes the visual effect that causes an object to appear closer than it is or having less depth or distance, as a result perspective or the angle of view.

Camera 2 position: Positioned further back on nacelle. Large foreshortening effect observed.

Camera 3 position: Positioned further back on nacelle. Large ‘foreshortening effect’ observed.

Camera 1 position: On TTG2 this camera was faulty and never provided any data. TTG 3 data shows the camera is positioned very close to turbine blades on underside of nacelle. Therefore much less light as shadowed from above. The camera is angled in such a way that it is looking down the length of the turbine blade.

5

• Camera 1 (which only functioned on TTG 3) appears to be positioned very close to the turbine blades on the underside of the nacelle. Therefore, there is much less light in the image compared to Cameras 2 and 3 as the camera is shadowed from above by the nacelle (see Figure 1.2).

• Cameras 2 and 3 appear to be positioned further back on the top side of the nacelle. There appears to be a large foreshortening effect with the blades actually being much further away than they appear.

Cameras 2 and 3 are pointing forwards towards the blades while Camera 1 (on TTG 3) appears to be facing down, looking down the length (side) of the blades and capturing the triangular section of the rotor blade that is closest to the nacelle without showing the component that attaches the blade to the nacelle. This results in a somewhat confusing image for Camera 1 with the turbine blades seemingly floating, unattached from the nacelle (see Figure 1.2). Camera operation For TTG 3 the cameras performed well based on the footage that was available for analysis (see Section 1.4.4). However, for TTG 2 Camera 1 (the camera on the foundation) failed and did not produce data.

Figure 1.2. Field-of-view of underwater video cameras from TTG 3 (top left = camera 1, top right = camera 2, bottom left = camera 3, bottom right = no camera)

The underwater video cameras were installed primarily for blade condition monitoring purposes and to monitor marine wildlife behaviour around the turbine. For the purposes of monitoring marine wildlife, it would have been more effective if the full extent of each turbine blade was visible. It is possible that a slight adjustment to the angle of the cameras would ensure that the full rotor swept

6

area is covered. This will be determined by the objectives of installing the cameras i.e. whether they are for environmental or technical monitoring. For further details on the effectiveness of the video monitoring equipment, techniques and processes learned from the video analysis process, see Table 1.2. 1.4.3 Powering of equipment, transmission of data and data storage

1.4.3.1 Details

The cameras are powered by the turbine and feed data to the onshore substation at Ness of Quoys, Caithness via a fibre optic cable. The video footage is subsequently stored on the AHH server. 1.4.3.2 Lessons learned regarding power of equipment, transmission of data and data storage

The main challenge associated with the analysis of video data was the ability to store and access it in an efficient manner. Video data has burdensome storage requirements; the data that were made available to this project suggested that each turbine generated approximately 2 GB of data in a 24-hour period (this would total approximately 60 GB / per month / per turbine). The storage of these data requires equipment, the cost of which can be significant. This emphasises the need for a solution to the data storage and transmission issues related to underwater video monitoring. All available data had timestamps detailing the date and time to which the data relates. Timestamping of data greatly reduces the amount of time and effort associated with analysis as it allows specific times of data to be quickly identified and then analysed. Moreover, each filename clearly indicated the date and time to which it related. This allowed specific times of data to be easily sourced and analysed. If data is not timestamped then it is critical that the filenames accurately indicate the day, month, year and time to which the footage relates. 1.4.4 Video data review and analysis

An overview of the generic methodology used to assess the effectiveness of video monitoring techniques, equipment and processes suitable for video monitoring datasets is provided in Section 1. Details specific to this dataset are presented here. 1.4.4.1 Available data

The available dataset incorporated data from two turbines each with three cameras; two on each nacelle and one on each foundation (see Figure 1.1 and Figure 1.2). The usability, time period and format of the data that were provided differed depending on which TTG the data came from. For TTG 3, individual camera feed data were provided for 22nd February 2017-04 March 2017, however only the camera located on the foundation was operational for this period of the dataset. An incorporated feed of all three cameras (as in Figure 1.2 was provided from 10:00 on 13th July 2017-11:00 on 14th July 2017. For TTG 2, data incorporating cameras one and two were provided from 14:00 on 12th July 2017-11:00 on 14th July 2017. Camera 1 was faulty and never provided any data. 1.4.4.2 Sampling protocol for video data review and analysis of AHH MeyGen dataset

The total number of hours of video data obtained from the two turbines was relatively limited (approximately 56 hours including the hours of darkness). However, the data that was made available was more than sufficient to support the assessment of the effectiveness of techniques, equipment and processes undertaken during Task 3. The sampling strategy was as follows:

7

• Data from 13th July 2017 were selected as this was the day in which there was the greatest amount of available footage, across the two datasets; and

• Two-minute samples from the first two minutes in each hour of available data were selected for review.

Data was continually collected during the hours of darkness, however no lighting was used to assist with environmental monitoring and therefore, footage during the hours of darkness was unsuitable for analysis (see Section 1.4.2). As a result, footage between the hours of 23:00 and 04:00 (inclusive) was not analysed. A summary of the AHH video data that was reviewed to inform Task 3 (Assessment of the effectiveness of techniques, equipment and processes to date) is shown in Table 1.1. Table 1.1. Summary of MeyGen data reviewed to inform the assessment of effectiveness of techniques, equipment and processes

AHH turbine number

Date Hours of available footage

No. of usable 2-min samples

No. of 2-min samples actually analysed

TTG 2 13th July 2017 Full 24 hour period 17 (05:00 – 22:00)* 17

TTG 3 13th July 2017 10:00-23:00 12 (11:00 – 22:00) 12

*Excluding 10:00, for which there was no footage available on the hour. Trial period Three trial clips were randomly selected and analysed using Protocol 1 (Monitoring system review), which was developed as part of Task 2 ‘Establishment of protocols for analysing underwater video data’ (see Section 1.3 of the Main Report)3. The associated spreadsheet was populated by two different reviewers, which provided an opportunity to discuss any discrepancies between reviewing methods and allowed any limitations of the spreadsheet, together with other items requiring consideration to be identified and the spreadsheet adapted to take account of this. 1.5 Summary of findings

A full commentary of the effectiveness of techniques, equipment and processes has been recorded and collated in Microsoft Excel for this dataset. This information may be made available upon request to Aquatera. A summary of the key lessons learned from the review process is shown in Table 1.2. Table 1.2. Summary of the effectiveness of video monitoring techniques, equipment and processes

Effectiveness Category

Factor Lessons Learned

Planning and design

Camera positioning – Three Cameras: (Camera 1 on underside of nacelle facing downwards along length of blades (only on TTG 3, Camera 1 on TTG 2 was faulty so its

• The orientation of Camera 1 (viewing downwards along the turbine blade) has limited use for monitoring marine wildlife activity as most of the field-of-view is taken up by the base of the rotor blade.

• Camera 1 is positioned on the underside of the nacelle which results in a significantly reduced light level compared to Cameras 2 and 3, particularly during late

3 Protocol 1 (Monitoring system review) can be found via the wave and tidal energy page of the NatureScot website; it is due, subsequently, to be available also via the ORJIP Ocean Energy website.

https://www.nature.scot/professional-advice/planning-and-development/advice-planners-and-developers/renewable-energy-development/marine-renewables/advice-marine-renewables

http://www.orjip.org.uk/eia-tools

8



orientation could not be determined); Cameras 2 & 3 on the upper sides of nacelle and positioned further back facing the blades (see Section 1).

afternoon and evening hours, as the sunlight is blocked by the nacelle.

• The proximity of Camera 1 to the blades means that light is reflected off the blade making the image over-exposed.

• A 360° view of the rotor swept area is essential for collision risk detection and monitoring of near-field behaviour. There is a recommendation to ensure the field-of-view covers as much of the rotor swept area as possible.

Field-of-view • There is only partial coverage of the rotor swept area. All cameras are positioned close to the blades with very narrow fields-of-view with only the portion of the blades closest to the nacelle visible as they rotate. There is no coverage of the rest of the turbine blades or blade tips. This means it would not be possible to observe any near field interactions (including any collision events) with the unseen portions of the blades.

• Approximately 40% (Camera 3) and 50% (Camera 2) of the fields-of-view are taken up by the nacelle.

Consultation • Consultation with regulatory bodies is essential during PEMP development to ensure that lessons learned from previous monitoring activities can be incorporated into the plan and that the activities provide the best chance of reducing scientific uncertainty around potential impacts.

• The establishment of an advisory group such as that implemented during this project provides a good example of how this can be managed to inform the development of a robust PEMP.

Data collection

Video equipment – Seacam Ultra Wide Angle Monochrome UV cameras

• Video footage is not of sufficient quality to accurately monitor collision risk potential and near-field behaviour as it is difficult to confidently identify objects using standard visual monitoring techniques. It is challenging to differentiate between marine fauna, flora and inanimate objects, particularly during fast-moving tide.

• During conditions of sufficient daylight and high water clarity, colour cameras would better aid in the identification of marine wildlife to species level and help to distinguish between marine fauna, flora and inanimate objects when using standard visual monitoring techniques.

• The use of High Definition (HD) cameras would provide greater certainty around potential sightings and, again, would enable marine fauna, flora and inanimate objects to be more easily distinguished, the former to species level.

• The use of wide angle cameras would enable more accurate analysis of potential behavioural responses, however wide angle cameras may limit footage quality/sharpness.

9



• There is one potential sighting of interest (see https://youtu.be/6o3l9UMwBDo) of a fish (flatfish or perhaps a ray) which demonstrates the relative difficulty associated with identification. The very brief view of the underside of the fish/object shows that it is lighter in colour than its topside which suggests that it is a fish.

Camera operational performance

• All three cameras on TTG 3 were operational however Camera 1 on TTG 2 was not operational.

Visibility (light levels) • The video footage from all three cameras appears dark even before mid-day. This is likely related to the depth at which the turbines are located i.e. approximately 30m.

Visibility (picture quality) • The image is grainy. • Passing objects are blurry and unidentifiable when

manually checking frame-by-frame. • There is a notable lag effect (ghost pixel trail) visible in

Camera 2 and occasionally in Camera 3. It is possible that this lag effect means that some activity could be obscured.

• Video resolution is low meaning there is low confidence that small, fast moving receptors could be positively identified using standard manual visual analysis techniques.

Visibility (lens) • No biofouling visible however only a single day of footage was analysed and the turbines had only recently been redeployed following a programme of turbine improvements.

Size of objects • It is not possible to determine the scale of objects in comparison to the turbine nacelle and blades when using standard manual visual analysis techniques.

Power of equipment, transmission of data and data storage

Turbine operational periods

• If available, a log of the turbine rotational data would have helped refine the data to be analysed by allowing periods of footage to be selected which have the greatest chance of reducing scientific uncertainty around key impacts.

Timestamping • Timestamping of data, such as that applied during the collection of this dataset, greatly reduces manual processing time as it allows specific times of data to be quickly identified and then analysed.

Video data review and analysis

Access to technical information

• MeyGen provided some general data on the camera positioning in the PEMP; 3 cameras positioned at 120° around the nacelle to capture the full 360° of the rotor blades. However, it was still necessary to approximate camera positioning from this and from interpretation of the video data. Full schematics of camera positioning should be provided in PEMPs to avoid any confusion or misinterpretation of data.

Object distances • During manual analysis it was very difficult to determine the distance of objects from the camera.

• It is difficult to judge what, if any, distance is visible beyond the turbine blades.

https://youtu.be/6o3l9UMwBDo

10



Size of objects • It is not possible to accurately estimate the size of objects using standard visual monitoring techniques. This is demonstrated in the below clip of a jellyfish (see https://youtu.be/1oeAX1JqMoQ).

Direction of water flow • It is often difficult to determine the direction of water flow from video footage alone. Full schematics of the cameras in relation to the turbine in addition to the turbine positions geographically and in relation to water flow direction in the PEMP would have helped data analysis if they were available; they should be provided as standard practice in developmental PEMPs.

Estimate of time taken to review footage

• It is difficult to review footage from three cameras simultaneously due to the different orientation of Camera 1 compared to Cameras 2 and 3.

• The very fast-moving water flow means that objects are visible for fractions of a second only (perhaps only for a single frame). It is very difficult to identify objects even when viewed frame by frame.

• The time taken to review clips increases considerably when receptors (or possible receptors) are observed.

• Extracting the two-minute clips in the first instance saves time overall and makes review and analysis more accurate and efficient.

• It took approximately one day in total to review the datasets (32 two-minute clips in total) from TTG 2 and TTG 3.

Trial period • A dataset specific trial period is essential to ensure that reviewers are familiar enough with the particular dataset and have confidence in correctly identifying marine wildlife.

• If more than one reviewer is involved it would be important to ensure there is consistency between them.

• Three two-minute trial clips from this dataset were found to be more than sufficient to familiarise reviewers with the data and ensure effective analysis.

Reviewer analysis capability/proficiency

• During analysis it became apparent that it will take many hours of training for reviewers to confidently identify marine wildlife and there is potential for false-positive results, particularly at the initial stage.

Reviewer fatigue • Due to the relatively lengthy analysis periods there is potential for a reduction in the reviewer’s visual acuity.

• Reviewer fatigue was noted to occur when reviewing samples from this dataset in succession for more than one hour but is specific to individuals.

Speed of tide • In fast-flowing tide it is challenging to determine between animate and inanimate objects even when checking frame by frame.

• MeyGen data proved challenging to allow accurate identification of passing objects and marine wildlife due to the speed of the tide.

https://youtu.be/1oeAX1JqMoQ

11

2. ANDRITZ HYDRO HAMMERFEST TURBINE: ESTIMATED CAMERA POSITIONS OF VIEW

2.1 Camera schematic layout

In the absence of a description or schematic of the camera layout on the Andritz Hydro Hammerfest (AHH) turbines deployed at the MeyGen Tidal Array, an attempt has been made to estimate their positions through analysis of the available video datasets (see Figure 2.1 to Figure 2.5 below). 2.1.1 Camera 1 (on TTG 3 only as Camera 1 on TTG 2 was faulty):

Camera 1 appears to be positioned very close to turbine blades on underside of the nacelle. Therefore, there is much less light in the image compared to cameras 2 and 3 as the camera is shadowed from above by the nacelle. 2.1.2 Camera 2

Camera 2 appears to be positioned further back on one of the top sides of the nacelle. There appears to be large foreshortening effect with the blades actually being much further away than they appear. 2.1.3 Camera 3

Camera 3 appears to be positioned further back on one of the top sides of the nacelle. There appears to be large foreshortening effect with the blades actually being much further away than they appear.

Figure 2.1. Field-of-view of underwater video cameras from TTG 3 (top left = camera 1, top right = camera 2, bottom left = camera 3, bottom right = no camera)

12

Figure 2.2. Field-of-view from each camera on turbine, Camera 1 is below and considerably closer to the rotor blades than Cameras 2 and 3

Camera 3

Camera 1

Camera 2

13

Figure 2.3. Estimated position of cameras on turbine

Figure 2.4. Direction cameras are pointing

NB Text for cameras 2 and 3 refers to blade tips not being visible, but if schematic is correct the remaining 75% is not seen.



Camera 1 position: Positioned very close to turbine blades on underside of nacelle. Therefore much less light as shadowed from above. The camera is angled in such a way that it is looking down the length of the turbine blade.

Camera 2 direction: Camera 2 is facing out, looking through the blades.

Camera 3 direction: Camera 3 is facing out, looking through the blades.

Camera 1 position: Camera 1 is facing down, looking down the length (side) of the blades and capturing the top section of triangular part of turbine blade that is closest the nacelle without showing the part that attaches the blade to the nacelle. Turbine structural support is visible in some footage. This supports the idea that this camera is facing down.

14

Camera 2 field-of-view: View shows section of turbine blades close to the nacelle rotating past in clockwise direction, including section that attaches blades to nacelle. Blades come into view from left of screen. The tips of blades are not visible.

Camera 3 field-of-view: View shows section of turbine blades close to the nacelle rotating past in clockwise direction, including section that attaches blades to nacelle. Blades come into view from left of screen. The tips of blades are not visible.

Camera 1 field-of-view: view shows very top section of triangular part of turbine blade that is closest the nacelle without showing the part that attaches the blade to the nacelle, therefore giving the appearance that the blade is floating. Blades come into view from the right of the screen as the camera is position below the nacelle, however the blades are still turning in a clockwise direction.

Figure 2.5. Field-of-view of cameras on turbine

15

3. REVIEW OF EFFECTIVENESS OF UNDERWATER VIDEO MONITORING TECHNIQUES, EQUIPMENT AND PROCESSES - SCOTRENEWABLES SR2000 TIDAL TURBINE: DETAILED TECHNICAL REVIEW

3.1 Introduction

This section provides the detailed results for Task 3 in relation to underwater video monitoring undertaken around Scotrenewables’ full-scale 2MW prototype floating tidal turbine (SR2000). The SR2000 was deployed at the European Marine Energy Centre (EMEC) from October 2016 to July 2018. In October 2018, Scotrenewables announced their rebranding to Orbital Marine Power Ltd. The company and device shall hereafter be referred to as Scotrenewables and SR2000 respectively given that this is accurate for this deployment. The aim of this task was to assess the effectiveness of underwater video monitoring techniques, equipment and processes used during underwater video monitoring activities. The aim of this assessment is to learn from the experience of past deployments to identify recommendations that can be used by developers and regulators involved in planning future monitoring activities of tidal stream developments. The results also provide information to help guide future technology development and associated strategic R&D funding streams. It should be noted that all comments and recommendations drawn in this report are in relation to the monitoring systems’ effectiveness in monitoring environmental effects. It should be recognised that many monitoring programmes around first deployments and arrays are focussed on technical monitoring and therefore not yet optimised for environmental monitoring purposes. The lessons learned and recommendations presented in this document should not therefore be taken as a critique of previous/current monitoring systems, but as a series of recommendations for optimising future environmental monitoring systems. 3.2 Approach

Several aspects of this video dataset have been reviewed to assess the effectiveness of the techniques, equipment and processes used at each stage in the process with the aim of identifying key challenges and lessons learned that can be used to make recommendations to inform future environmental monitoring around tidal energy devices and arrays. The following stages of the monitoring programme were considered during the assessment:



The Scotrenewables dataset consists of 19 TB of continuous video monitoring data filed as five minute segments under each hour and day. Data were gathered from 17 January 2018 – 2 July 2018 (equating to 167 days or 4008 hours). The total dataset encompasses footage from 13 different cameras of which five provided footage of interest from an environmental monitoring perspective. Three of these cameras were situated underwater and a further two on the surface mast. Three additional cameras were present underwater but remained non-functional during deployment. For the purposes of reviewing the effectiveness of underwater monitoring techniques, only footage from cameras which were positioned to observe the rotors was analysed, rather than other aspects of the device. From this, the review focuses on video footage from two underwater cameras considered suitable for analysis, one on the port side nacelle (camera 3), and one on the leg brace (camera 5). Camera 4, on the starboard side nacelle, was not operational during deployment (as were cameras

16

6 & 7) and camera 2 did not capture footage of the turbine blades. All cameras, operational status and fields of view are detailed in Figure 3.1. 3.4 Assessment


3.4.1.1 Details

At the time of writing this technical briefing note, an Environmental Monitoring Plan (EMP) was not available for the SR2000, which presents a gap in discussing details and lessons learned from the planning and design of the monitoring programme undertaken for the SR2000 deployment, and remains to be addressed. 3.4.1.2 Lessons learned during planning and design

An accurate and detailed EMP which clearly describes the objectives of the monitoring campaign and full technical details of the equipment set-up including positions and orientation of the monitoring equipment is necessary to enable accurate analysis of the gathered data and cataloguing of metadata. As discussed above, an EMP was not available at the time of writing this technical briefing note, therefore there is a current gap in addressing recommendations on the planning and design of monitoring programmes specific to the SR2000. Note: he nascent state of the tidal stream energy industry means it is important to maximise learning opportunities that could help reduce scientific uncertainty around key impacts. Therefore, at an early stage, regulators and statutory nature conservation bodies (SNCBs) should encourage developers to gather environmental monitoring data with clear objectives that feed into the consenting process and provide direction on the relevance of design of developers’ monitoring plans. The regulator could suggest that any video data gathered are shared with them for research / data analysis purposes however this would require the regulator to have significant data storage capacity. An alternative would be for developers to provide monitoring reports to the regulator, with the developer storing the data in house. If there are commercial sensitivities associated with releasing datasets, clauses to delay their publication until a later date should be encouraged as it would enable their future analysis and subsequent reporting in scientific journals by academics and researchers. This would help to contribute to the body of evidence required to reduce scientific uncertainty. 3.4.2 Data collection

3.4.2.1 Details

Underwater video Six Vivotek colour cameras were installed to provide underwater footage of the SR2000 (Figure 3.1), four of which were Vivotek domes (FE8174). In that figure, cameras have been labelled with the corresponding channel number of the data as recorded on the Network Video Recorder (NVR). One was placed on the leg brace (camera 5), pointing towards both turbine blades. Another single dome was placed under the turret to monitor the connection to the moorings (camera 2). The remaining two domes were positioned on the respective port and starboard nacelles of the turbine, and were aimed towards the tips of the blades (cameras 3 & 4). Another two Vivotek bullets (IP8332) were originally placed next to each other around the centre of the hull, each pointing towards the starboard and port side turbine blades, but failed to function and were subsequently removed from the plans

17

(cameras 6 & 7). The starboard side nacelle camera was also not functional during deployment (camera 4). Two lights next to the Vivotek bullet cameras on the underside of the hull were additionally installed with the same directional aim but were not used to faciliate video camera monitoring during deployment. There were concerns related to the risk of receptors being attracted to the light source and possibly invalidating normality of data. Underwater cameras alternatively employed a black and white mode during low light conditions (dawn and dusk).

Key

Vivotek dome camera, functional during deployment but footage not used as part of this review as field of view covered turret, not turbine blades

Vivotek dome camera, functional during deployment and footage used as part of this review

Vivotek dome camera, not functional during deployment, therefore footage not used as part of this review

Vivotek dome camera, functional during deployment and footage used as part of this review

Vivotek bullet cameras, not functional during deployment, therefore footage not used as part of this review

Vivotek bullet cameras, not functional during deployment, therefore footage not used as part of this review

Acoustic Doppler Current Profiler (ADCP)

Figure 3.1. Schematics of the SR2000 device with camera placements

18

The top and bottom schematics as indicated in Figure 3.1 show the SR2000 in different operational modes; the top schematic illustrates the device as it would be in operation whilst the bottom schematic illustrates the device with legs brought up to the hull as would be the case during transport and other non-operational modes. It should also be noted that the fields of view as detailed in the figure are from best understanding of the footage. Cameras 6 & 7 were not operational, therefore it is not known if the field of view extended to the entirety of the turbine’s blades. Similarly, with cameras 3 & 4, it is believed that the cameras were positioned on the underside of the nacelles (from consultation with Scotrenewables and available footage) and therefore capture the lower half of blade rotation rather than the upper half. Camera 5 would appear to capture the farthest lateral edges of the blades but, due to visibility (and reduced visibility with increased depth), it is believed the lower halves of the blades’ rotations were not captured by this camera. Sample fields of view from each camera are provided below in Figure 3.2. From camera 2 the turret connection and start of the mooring lines can be seen on the right of the image (with the correct perspective the mooring lines should be at the top of the image).

Figure 3.2. Sample fields of view from cameras 2, 3 and 5 (left to right respectively)

In camera 3 & 5 footage, the blades are visible but are hard to recognise from a snapshot. When reviewing the footage the motion of the blades and the blades themselves are more apparent. What is thought to be dead pixels can also be seen in camera 3 footage and are highlighted below in Figure 3.3.

19

Figure 3.3. Camera 3 footage with dead pixels circled in red

Sonar No sonar device was installed during the monitoring programme. Accelerometers and strain gauges Accelerometers were installed in the blade tips of both turbine rotors with the intention of detecting any anomalies in blade speed that could be attributed to collision events. However, in the revised EMP for Orbital Marine Power’s upcoming Orbital O2 2MW demonstration (Orbital O2 EMP, 20184), it was noted that during blade testing of the SR2000 a fully grown adult could not get a signal to register on the accelerometers due to the heavy and stiff nature of the blades. Strain gauges were additionally installed in the blades for collision detection, however, it was also noticed in the O2 2MW EMP (Orbital O2 EMP, 20182) that for the SR2000, the vast volume of data and the high number of ‘spikes’ produced each day by the strain gauge meant it was difficult to distinguish between spikes caused by swirls and buffets of turbulence and those that may potentially have been caused by an object colliding with the rotor blade. Changes in hydrodynamics An Acoustic Doppler Current Profiler (ADCP) was installed on the underside of the device to monitor any changes in tidal flow but was not set to monitor continuously. A second ADCP device was deployed for two months on a separate frame downstream (to the south-southeast) of the device (Figure 3.4). A Motion Response Unit (MRU) was also installed in the device to monitor pitch, roll, and general stability of the device.

4 Orbital Marine Power Ltd, 2018. Orbital O2 Environmental Monitoring Programme. Available from developer upon request.

20

Figure 3.4. ADCP adjacent to SR200 deployment

3.4.2.2 Lessons learned during data collection

Underwater video Positioning of cameras The three underwater cameras mounted on the turbine nacelles and leg brace provide only partial coverage of the rotor swept area, with only the tips of the blades visible as they rotate. The position of the port side nacelle camera is such that the nacelle is not visible and the partial rotor swept area covers approximately 50% of the field-of-view, whereas the leg brace camera was positioned further back from the blades causing them to be distant and out-of-focus (Figure 3.5). The third camera (camera 4) was non-operational, however the cause of its malfunction remains unknown. The remainder of this review therefore focuses on cameras 3 and 5, respectively positioned on the port side nacelle and middle of the leg brace.

Key

Green circles Anchor towers

Blue lines SR2000 mooring lines

Red lines EMEC cables, adjacent lines from other berths

Red and green line SR2000 cable to EMEC hub – circles represent points used in a supporting ROV survey

21

Figure 3.5. Position of turbine blades in field-of-view (left: port side nacelle; right: middle leg brace)

The position of the cameras and resulting field-of-view are not optimum for environmental monitoring purposes. Future configurations should provide a clear view of the rotor swept area where possible.

Camera operation The fisheye configuration of the dome lenses created a distorted field-of-view which made it difficult to estimate distances or discern the nature of objects that passed close to the camera. Throughout the period of data capture there was a black shape noted in all footage from the port side nacelle (camera 3; see Figure 3.3), the cause of which was not identified but ascribed to a technical malfunction or “dead pixels”. The lens of the leg brace camera (camera 5) was additionally subject to water leakage, however this did not appear to affect its functionality. The port side nacelle camera was operational each day between the hours of 05:00 – 21:00, therefore potential wildlife interactions with the turbine blades beyond these hours in the summer months would not be captured. Conversely, the leg brace camera operated for a continuous 24 hours each day. However during hours of darkness, the image was so dark that any marine wildlife around the turbine would not have been visible. This has implications for underwater video monitoring during winter months, with the amount of usable data being limited by the number of hours of daylight using this system. The camera on the starboard side nacelle was non-functional and did not produce, thereby creating a reliance on using camera 3 on the port side nacelle for views of the rotor swept area.

Position of turbine blades

22

Biofouling of lenses Biofouling was apparent on the lenses of each of the cameras selected for review from the earliest footage of the available dataset, on 17 and 18 January for respective leg brace and port side nacelle cameras. The SR2000 was first installed on site in October 2016 and gauging from the amount of biofouling build up observed over 6 months (see Figure 3.6), the camera lenses were undoubtedly cleaned prior to the start of data collection in January 2017. Unfortunately, it is not known when this occurred or if a regular schedule was maintained. However, it is important to note that algal coverage was always more prevalent on the lens of camera 3 on the port side nacelle than camera 5 on the leg brace, which was not as quickly or severely affected. This may have been due to the position of the lens of camera 5 facing into the current, preventing the settling and adherence of fouling organisms. In January, approximately 5-10% of the field-of-view was obscured by algae on camera 3, whereas colonisation was minimal on camera 5 which presented around eight small dark patches thought to be barnacles. Biofouling became progressively worse into the late spring / early summer months (Figure 3.6), until approximately over 50% of the port side camera’s field-of-view was obscured by July. For further details on the effectiveness of the video monitoring equipment, techniques and processes learned from the video analysis process, see Table 3.2.

23

Figure 3.6. Progression of biofouling on underwater camera lenses

Port side nacelle camera (channel3) Leg brace camera (channel 5)

18/01, 2pm

18/01, 2pm

15/04, 12pm

15/04, 12pm

02/07, 8 am

02/07, 8am

24

3.4.3 Powering of equipment, transmission of data and data storage

3.4.3.1 Details

The SR2000 incorporates a battery bank into its design allowing the cameras to be powered from the turbine. Data were recorded on a hard drive in the SR2000 itself with transmission to an onshore Network Video Recorder (NVR) achieved through Satellite and Wi-Fi transmission. 3.4.3.2 Lessons learned regarding power of equipment, transmission of data and data storage

The use of an NVR and satellite transmission allows large data to be transferred and stored for comprehensive review as and when required. Investment into such technology ensured that all requirements related to data transmission and storage were met. The use of a battery bank, powered by the turbine itself and the immediate storage of data on on-board hard drives also provided an effective method for equipment power, transmission and storage. All data were timestamped, detailing the date and time to which the data relates. Timestamping of data greatly reduces the amount of time and effort associated with analysis as it allows specific times of data to be quickly identified and then analysed. Moreover, each filename clearly indicated the date and time to which it related. This allowed specific times of data to be easily picked out and analysed. If data are not timestamped then it is critical that the filenames accurately indicate the day, month, year and time to which the footage relates. Minor data gaps are present from conditions (possibly weather related) which prevented the transmission of stored data from the device to the onshore NVR, with that data being subsequently overwritten when the on-board hard drive ran out of available space. Other data gaps are present from loss of camera functionality. The main challenge associated with the analysis of video data was the ability to store and access it in an efficient manner. Video data has burdensome storage requirements with the data that were made available to this project having a total requirement of 19TB. The storage of these data requires equipment, the cost of which can be significant. This emphasises the need for a solution to the data storage and transmission issue related to underwater video monitoring. 3.4.4 Video data review and analysis


The available Scotrenewables dataset incorporated data from the single SR2000 turbine with three cameras; one on each nacelle (cameras 3 and 4) and one on the middle of the leg brace separating the nacelles (camera 5; see Figure 3.1 and Figure 3.5). The footage from each camera was viewed on a corresponding channel accessed from the NVR device. The usability and time period of the data that were provided differed depending on the camera. For cameras 3 and 4 positioned on the nacelles data were gathered from 18 January 2018 – 2 July 2018 for 16 hours each day (05:00 – 21:00), however only camera 3 located on the port side nacelle was functional for this period of the dataset. Data was continuously gathered from camera 5 on the leg brace from 17 January 2018 (00:00) – 2 July 2018 (23:00). Cameras 3 and 5 recorded a respective 160 and 146 days of footage in total.

25

3.4.4.2 Sampling protocol for video data review and analysis of the Scotrenewables dataset

The total number of hours of video data obtained from the two functional cameras was collectively estimated to be 6064 hours (2560 and 3504 hours from respective cameras 3 and 5), including the hours of darkness. A sampling strategy was developed for the Scotrenewables SR2000 dataset to inform the assessment of the effectiveness of techniques, equipment and processes undertaken during Task 3. The sampling strategy was as follows:

• One 24 hour period was selected from each of the following periods in the available dataset for analysis: at the start (18 January), approximately mid-way (15 April), and at the end (2 July); and

• Two- minute samples from the first two minutes in each hour of available data were selected for review5.

Data was continually collected during the hours of darkness, however no lighting was used to assist with environmental monitoring and therefore, footage during the hours of darkness was unsuitable for analysis (see Section 3.4.2). As a result, footage between the hours of 23:00 and 04:00 (inclusive) was not analysed. Fewer samples of footage could be analysed for January due to the reduced number of daylight hours in the winter season. A summary of the Scotrenewables video data that was reviewed to inform Task 3 (Assessment of the effectiveness of techniques, equipment and processes to date) is shown in Table 3.1. Table 3.1. Summary of Scotrenewables data reviewed to inform the assessment of effectiveness of techniques, equipment and processes

Camera channel number




3 18 January 2018 05:00 – 20:00 8 (09:00 – 16:00) 8

15 April 2018 05:00 – 20:00 15 (06:00 – 20:00) 15

2 July 2018 05:00 – 20:00 16 (05:00 – 20:00) 16

5 18 January 2018 Full 24 hour period 8 (09:00 – 16:00) 8

15 April 2018 Full 24 hour period 15 (06:00 – 20:00) 15

2 July 2018 Full 24 hour period 18 (05:00 – 22:00) 18 Trial period Three trial clips were randomly selected and analysed using the aforementioned Protocol 1 (Monitoring system review). The associated spreadsheet was populated by two different reviewers, which provided an opportunity to discuss any discrepancies between reviewing methods and allowed any limitations of the recording template, together with other items requiring consideration to be identified and the recording template adapted to take account of this.

5 Note: In order to maximise the efficiency of the data review process, all two-minute samples from channels 3 and 5, which were recorded on the same day and time, were paired together for reviewing and test footage reviews.

26

3.5 Summary of findings

A full commentary of the effective of techniques, and processes was recorded and collated in Microsoft Excel for this dataset. This information may be made available upon request to Aquatera. A summary of the key lessons learned from the review process is shown in Table 3.2. Table 3.2. Summary of the effectiveness of video monitoring techniques, equipment and processes



Planning and design

Camera positioning – Three cameras with different channel outputs. Channel 3: camera 3 on port side nacelle facing blades. Channel 5: camera 5 on leg brace looking towards turbine blades. Starboard side nacelle: camera was non-functional and had no channel output.

• The port side nacelle camera (camera 3) provides only a partial view of the rotor swept area, showing only the tips of the blades due to their size. Cameras positioned on the hull downwards at the blades may have provided a wider field-of-view encompassing more of each blade.

• The position of the camera on the leg brace between nacelles has limited use for monitoring marine wildlife activity as it is distanced from the rotor-swept area.

• A 360° view of the rotor swept area is essential for collision risk detection and monitoring of near-field behaviour. There is a recommendation to ensure the field-of-view covers as much of the rotor swept area as possible.

Field-of-view • There is only partial coverage of the rotor swept area. The camera on the port nacelle is positioned close to the blades with only tips visible as they rotate, which cover approximately 50% of the field-of-view. There is no coverage of the rest of the turbine blades. This means it would not be possible to observe any near field interactions (including any collision events) with the unseen portions of the blades.

• Only the distant tips of the blades are visible on either side of the leg brace camera’s field-of-view, and are most noticeable as they rotate otherwise they are difficult to view.

Consultation • Consultation with regulatory bodies is essential during PEMP development to ensure that lessons learned from previous monitoring activities can be incorporated into the plan and that the activities provide the best chance of reducing scientific uncertainty around potential impacts.

Data collection

Video equipment – Vivotek colour cameras (FE8174) with black and white low-light mode

• Video footage is not of sufficient quality to accurately monitor collision risk potential and near-field behaviour as it is difficult to confidently identify objects using standard visual monitoring techniques, particularly during fast-moving tide. No objects could be confidently classified as marine fauna, flora, or debris.

• Although colour camera was used, the reduced sensitivity in low light corresponded to a lower image quality. Automatic switching to black and white mode in low light periods (dusk and dawn) was effective in expanding the

27



number of hours available for monitoring potential marine interactions.

• The combination of low image quality and monochromatic blue-green hues of footage did not facilitate the identification of passing objects, nor species if a potential wildlife receptor was recorded. This is demonstrated in this clip of a possible bird passing near the lens of camera 5 (see https://youtu.be/Tz8wztd8pm4).


• Fisheye lenses made it difficult to estimate distances or identify objects that came relatively close to the camera due to its distorting effect. The use of wide angle cameras would enable more accurate analysis of potential behavioural responses, however wide angle cameras may limit footage quality/sharpness.


• The camera on the starboard nacelle was non-functional throughout the data collection period, preventing views of the right-hand portion of the SR2000’s rotor-swept area, therefore any potential interactions with this side of the turbine would not be captured.

• Water was noted to leak inside the lens of the leg brace camera; although it remained functional throughout the data capture period, this matter should be addressed to prevent further costly damage and potential loss of functionality of the camera.

Visibility (water clarity)

• Visibility varies with plankton concentration. Water appeared murkier in July in comparison to January.

Visibility (picture quality)

• Passing objects are blurry and unidentifiable when manually checking frame-by-frame. This is demonstrated in this clip of an unidentified object passing the blades on the port side (see https://youtu.be/YBCnhPrkABA).

• Picture quality continued to decrease when black and white mode was enabled in low light conditions.

• The low image quality reduces confidence that small, fast moving receptors could be positively identified using standard manual visual analysis techniques.

Visibility (lens) • “Broken pixels” were noted in footage from the port nacelle camera which permanently obscured a specific part of the field-of-view (see Figure 3.3)

• Biofouling obscured approximately 5-10% of the field-of-view on the port side nacelle camera in January, continuing to intensify until July where over 50% of the view was obscured. In several instances, flickering of light / potential movement was ascribed to algal growth on the lens.

• The leg brace camera was affected by biofouling to a lesser degree than the port nacelle camera, which could

https://youtu.be/Tz8wztd8pm4

https://youtu.be/YBCnhPrkABA

28



have been due to the lens facing into the current, preventing settling and adherence of fouling organisms.

Size of objects • The fisheye lens created a distorting effect on passing objects making it difficult to identify or estimate its size.

• It is not possible to determine the scale of objects in comparison to the turbine nacelle and blades when using standard manual visual analysis techniques.

Power of equipment, transmission of data and data storage


• If available, a log of the SR2000 turbine’s rotational data would have helped refine the data to be analysed, and then further helped with the analysis to reduce scientific uncertainty around key impacts.

Timestamping • Data were filed in 5-minute segments and accurately labelled to date and hour. Timestamping of data, such as that applied during the collection of this dataset, greatly reduces manual processing time as it allows specific times of data to be quickly identified and then analysed.

Data storage • The use of a Network Video Recorder (NVR) and satellite transmission allow large amounts of data to be transferred and stored for comprehensive study as and when required.

• Transmission of stored data to the onshore NVR was interrupted on occasion by loss of camera functionality or loss of Wi-Fi signal possibly due to poor weather conditions.

• A battery bank powered by the turbine itself and the immediate storage of data on an on-board hard drive provided an effective method for power and storage.

• The current video dataset had burdensome storage requirements, the cost of which can be significant.



• On request Scotrenewables provided some general data on the camera positioning but nothing specific. It was still necessary to approximate camera positioning from interpretation of the video data. Full schematics of camera positioning should be provided by the developers to avoid any confusion or misinterpretation of data.

Object distances • During manual analysis it was very difficult to determine the distance of objects from the camera.

• The distance of the leg brace camera from the turbine blades prevented any visibility beyond the turbine blades.

Size of objects • It is not possible to accurately estimate the size of objects using standard visual monitoring techniques, partially due to the distorting effect of the fisheye lens. This is demonstrated in this clip of a possible jellyfish (see https://youtu.be/Qj2EbD1KRdw).

Direction of water flow • It is often difficult to determine the direction of water flow from video footage alone.


• It is difficult to review footage from three cameras simultaneously due to the different orientation of the port side camera to that of the leg brace camera.

https://youtu.be/Qj2EbD1KRdw

29



• The very fast-moving water flow means that objects are visible for fractions of a second only (perhaps only for a single frame). It is very difficult to identify objects even when viewed frame by frame.

• The time taken to review clips increases considerably when receptors (or possible receptors) are observed. This time increases when multiple objects are viewed in the same clip.

• Extracting the two-minute clips in the first instance saves time overall and make review and analysis more accurate and efficient.

• It took approximately two-three days in total to review the sampled datasets (44 two-minute clips in total) from camera 3 and 5.

Trial period • A dataset specific trial period is essential to ensure that reviewers are familiar enough with viewing the particular dataset and has confidence in correctly identifying marine wildlife.


• Three two-minute trial clips from this dataset were found to be more than sufficient to familiarise reviewers with the data and ensure effective analysis.

Reviewer analysis capability/proficiency


Reviewer fatigue • Due to the relatively lengthy analysis periods there is potential for a reduction in the reviewer’s visual acuity.

• Reviewer fatigue was noted to occur when reviewing samples from this dataset in succession for more than one hour but is specific to individuals.


• Scotrenewables data proved challenging to allow accurate identification of passing objects and marine wildlife due to the speed of tide.

30

4. REVIEW OF EFFECTIVENESS OF UNDERWATER VIDEO MONITORING TECHNIQUES, EQUIPMENT AND PROCESSES - VOITH HYDRO HYTIDE TIDAL TURBINE: DETAILED TECHNICAL REVIEW

4.1 Introduction

This section provides the detailed results for Task 3 in relation to underwater video monitoring undertaken around Voith Hydro’s HyTide tidal stream turbine which was deployed at the European Marine Energy Centre (EMEC) in the summer of 2014. The aim of this task was to assess the effectiveness of underwater video monitoring techniques, equipment and processes used during underwater video monitoring activities. The aim of this assessment is to learn from the experience of past deployments to identify recommendations that can be used by developers and regulators involved in planning future monitoring activities of tidal stream developments. The results also provide information to help guide future technology development and associated strategic funding streams. It should be noted that all comments and recommendations drawn in this report are in relation to the monitoring systems’ effectiveness in monitoring environmental effects. It should be recognised that many monitoring programmes around first deployments and arrays are focussed on technical monitoring and therefore not yet optimised for environmental monitoring purposes. The lessons learned and recommendations presented in this document should not therefore be taken as a critique of previous/current monitoring systems, but as a series of recommendations for optimising future environmental monitoring systems. 4.2 Approach

Several aspects of this video dataset have been reviewed to assess the effectiveness of the techniques, equipment and processes used at each stage in the process with the aim of identifying key challenges and lessons learned that can be used to make recommendations to inform future monitoring around tidal energy devices and arrays. The following stages of the monitoring programme were considered during the assessment:



The Voith dataset incorporates three cameras; two on the nacelle and one on the monopile (see Figure 4.1, Figure 4.2 and Figure 4.3). Data were gathered from 12-18 May 2014, 20 May and 28 May-13 July 2014. Between the 12 May and 1 June, video footage was recorded during discrete periods only whereas footage was recorded continuously from 2 June to 13 July.

31

4.4 Assessment


4.4.1.1 Details

In order to comply with the project’s Marine Licence and to begin to build an evidence base to reduce scientific uncertainty around key potential impacts, an Environmental Mitigation and Monitoring Plan (EMMP)6 was designed in consultation with Marine Scotland, NatureScot and EMEC prior to the turbine’s deployment in March 2014. The activities set out in this plan covered mitigation and monitoring activities that would attempt to collect empirical data from various types of equipment. Supplementary monitoring activities were also outlined in the EMMP. The monitoring activities set out in the EMMP were as follows:

• Monitor effects on marine mammals, fish and birds from turbine presence and operation using: an integrated system of underwater video, active acoustics (sonar), strain gauge and accelerometer data to monitor potential interactions with the turbine’s moving parts; and hydrophones to characterise the operational noise signature of the device.

• Monitor effects on hydrography and seabed dynamics by using acoustic Doppler current profilers (ADCPs) to monitor current profiles up and downstream of the device.

Supplementary monitoring activities included:

• Monitoring of nearby seal haul-out sites; and • Incidental recording of notable wildlife observations during site operations.

4.4.1.2 Lessons learned during planning and design

An accurate and detailed EMMP or Project Environmental Management Plan (PEMP), such as that prepared for the Voith HyTide deployment, which clearly describes the objectives of the monitoring project and full technical details of the equipment set-up is essential to enable accurate analysis of the gathered data and cataloguing of metadata. Note: the nascent state of the tidal stream energy industry means it is important to maximise learning opportunities that could help reduce scientific uncertainty around key impacts. Therefore, at an early stage, regulators and statutory nature conservation bodies (SNCBs) should encourage developers to gather environmental monitoring data with clear objectives that feed into the consenting process and provide direction on the relevance of design of developers’ monitoring plans. The regulator could suggest that any video data gathered are shared with them for research/data analysis purposes however this would require the regulator to have significant data storage capacity. An alternative would be for developers to provide monitoring reports to the regulator, with the developer storing the data in house. If there are commercial sensitivities associated with releasing datasets, clauses to delay their publication until a later date should be encouraged as it would enable their future analysis and subsequent reporting in scientific journals by academics and researchers. This would help to contribute to the body of evidence required to reduce scientific uncertainty. 6 Aquatera Ltd, 2013. Voith Hydro’s deployment of a tidal energy converter at EMEC’s tidal test site – Environmental Mitigation and Monitoring Plan 2013. Available upon request

32

4.4.2 Data collection

4.4.2.1 Details

The planned monitoring strategy was comprised of an integrated system using multiple instruments. While each component of the system was intended to work together, if any one component of the equipment failed to work, meaningful data could still be gathered from the other components. The monitoring equipment was connected to and powered by the turbine. Data from the monitoring equipment was fed from the turbine to an onshore control centre where it was stored for later analysis. Underwater video Three Tritech Tornado low light monochrome video cameras (two on the nacelle and one on the monopile) were installed to monitor behaviour/presence of marine wildlife (marine mammals, fish and birds around the turbine) (see Figure 4.1 and Figure 4.2). No lighting was installed to facilitate video camera monitoring outside daylight hours.

Figure 4.1. Schematic of Voith HyTide turbine and support structure

33

Figure 4.2. Positions of cameras

Sonar One Tritech Gemini 720id multibeam sonar device was installed with bespoke movement detection software (Gemini SeaTec software) to detect and track movement of marine mammals. The Gemini SeaTec software works by identifying moving targets. The size, shape and swimming behaviour of these targets is then analysed in real time to determine the likelihood that a moving target is a marine mammal. The SeaTec software then classifies each target as a ‘possible’, ‘potential’ or ‘probable’ observation of a marine mammal. The data can be stored to allow detailed post-hoc analysis of target data for analysis of marine mammal behaviour to inform future environmental impact assessments. Strain gauge and accelerometer A single strain gauge and accelerometer were fitted to one of the three rotor blades. It was proposed that the strain gauge and accelerometer data would be analysed to identify any notable events or ‘spikes’ that may indicate a collision event between an object and the blade. These possible collision events would then be cross-checked with available video footage to check if any collisions with marine wildlife had occurred. Underwater operational noise A noise characterisation survey to determine the noise signature of the device was also planned; however the turbine’s operational capacity had to be limited due to a malfunction of an auxiliary device which was a safety relevant component inside the turbine. The noise characterisation survey was abandoned as this would only have been useful if the turbine was able to operate at full capacity. Changes in hydrodynamics Two ADCPs were installed with the aim of measuring current profiles upstream and downstream from the device to monitor any changes in tidal flow given different rotational speeds at each current

Camera 3 Camera 2

Camera 2

Camera 1

34

velocity. However, one of the ADCPs was defective and it was not possible to undertake a comparison of current profiles upstream and downstream of the turbine. 4.4.2.2 Limitations / challenges identified during data collection

Underwater video Positioning of cameras The three cameras were mounted on the nacelle and monopile at a spacing of 120° to provide a 360° view of the turbine rotors. The positions of the cameras were such that approximately 40% of the fields of view from cameras 2 and 3 were taken up by the nacelle (see Figure 4.3), thereby reducing the fields of view of the rotor swept area of the turbine blades. Additionally, their position approximately 5.5m behind the rotor blades meant the ability to identify objects beyond the blades was limited, particularly in June and July when visibility was reduced due to high levels of plankton present in the water column. While maximising the field-of-view is considered to be one of the main priorities in underwater video monitoring, the positioning of cameras 2 and 3 was however useful for investigating biofouling on the nacelle. Camera 1 was situated on the monopile underneath the nacelle (see Figure 4.3). As the main objective of the underwater video cameras was to monitor marine wildlife behaviour around the turbine, maximising the view of the rotor swept area should have been given greater consideration when planning the positions of the cameras. However, it should be noted that it has not been possible to ascertain whether the final positioning of the cameras was limited by technical considerations, which may have prevented their positioning elsewhere on the turbine. Camera operation Overall, the video cameras performed well with all three cameras remaining operational throughout the data collection period from 12 May to 13 July 2014. During the first few days of data collection, video data were collected during discrete periods only. In June and July data were collected continuously for 24 hours a day. However, during hours of darkness, the image was so dark that any marine wildlife around the turbine would not have been visible. This has implications for underwater video monitoring during winter months, with the amount of usable data being limited by the number of hours of daylight.

35

Figure 4.3. Field-of-view from cameras (top left = camera 1 on monopile, top right = camera 2 on nacelle, bottom left = camera 3 on nacelle, bottom right = no camera). Video still is from May when biofouling coverage was at its minimum

Biofouling of lenses Initially, all three cameras provided clear images of the turbine rotors both when the turbine was at standstill and when it was in operation (see Figure 4.3). Some level of interference was visible on all cameras in the form of vertical lines constantly moving horizontally across the screen. Towards the end of May (approximately two weeks after deployment), biofouling began to appear on the lens of Camera 3. Two small dark areas appeared on the top of the lens of Camera 3 around 21 June. It is thought that these were barnacles, as an alternating patch of light and dark grey in their centre is consistent with the opening and closing of the shell plates which occurs during the feeding behaviour of barnacles. This biofouling got progressively worse until nothing could be seen from Camera 3 by mid-June. Biofouling began to appear on Camera 1 on 24 June (92 days after installation of the turbine), which got progressively worse; however biofouling on this camera did not appear to occur as quickly or severely as it did on Camera 3 and a relatively clear image is still visible by 13 July (see Figure 4.4) Camera 2 stayed free of biofouling and a clear image is available throughout the data gathering period.

36

Figure 4.4. Video still taken from July when biofouling was at its greatest extent (top left = camera 1 on monopile, top right = camera 2 on nacelle, bottom left = camera 3 on nacelle, bottom right = no camera)

For further details on the effectiveness of the video monitoring equipment, techniques and processes learned from the video analysis process, see Table 4.2. Sonar The sonar camera was deployed but due to a technical issue with the power supply, it was never operational and no data were gathered. Strain gauge and accelerometer A strain gauge and accelerometer was installed in one of the turbine’s three rotor blades. While data were gathered it was considered impractical to carry out analysis, because the vast volume of data and the high number of ‘spikes’ produced each day by the strain gauge meant it was difficult to distinguish between spikes caused by natural changes in turbulence and those that may potentially have been caused by an object colliding with the rotor blade (Aquatera, 2015)7. Regardless of the data’s usability, the fact that a strain gauge and accelerometer were only included in one of the three turbine rotor blades would have prevented a collision from being recorded had it occurred with one of the other two rotor blades. 7 Aquatera Ltd, 2015. Voith Hydro’s Deployment of a tidal energy converter at EMEC’s tidal test site – Environmental monitoring report. Available upon request

37

Changes in hydrodynamics One of the ADCPs that was deployed was faulty which prevented changes in hydrodynamics up and downstream of the device from being analysed. 4.4.3 Powering of equipment, transmission of data and data storage

4.4.3.1 Details

The set-up for the monitoring equipment included a pair of data cables, extending from the monopile, along the seafloor, to two separate gravity-based mounts housing the instrumentation boxes. The instrumentation boxes received power directly from the turbine. For ease of installation, the data cables were each divided into two segments: one that reached from the turbine to the support structure base and another that extended from the support structure base along the seafloor. A third, short data cable ran from midway up the support structure to the top of the support structure. The turbine was connected to the pre-installed EMEC export cable using a jumper cable that spanned the gap between the termination box on the existing cable and the electrical ports on the turbine itself. It contained the main and auxiliary power conductors, as well as fibre optic threads. Video data passed directly via the cable to the onshore substation facility on the island of Eday and were stored on a hard-drive. It is estimated that approximately one month’s worth of data could be stored on the hard-drive before the data would be overwritten. Therefore, permanent storage of the video data required a member of staff to manually copy across the data onto Compact Disc (CD) every couple of weeks and delete the copied data from the hard-drive to free up space. 4.4.3.2 Lessons learned regarding power of equipment, transmission of data and data storage

The main challenge associated with the analysis of video data was the ability to store and access it in an efficient manner. The video data has very burdensome storage requirements with each 24-hour period generating approximately 8GB of data. The storage of these data requires equipment, the cost of which can be significant. This emphasises the need for a solution to the data storage and transmission issue experienced during underwater video monitoring. Additionally, the process of storing these data was a surprisingly labour intensive process, requiring someone to travel from the Orkney Mainland to the substation on the island of Eday every couple of weeks to manually copy across the data onto CD to enable the video footage to continue to be recorded and stored on the hard-drive. This method of manually copying across to CD was a very high risk strategy with a high risk of data loss and corruption of the data. During analysis of the video data it became apparent that some of it was timestamped while other sections were not. Timestamping of data greatly reduces the amount of time and effort associated with analysis as it allows specific times of data to be quickly identified and then analysed. If data is not time-stamped then it is critical that the filenames accurately indicate the day, month, year and time to which the footage relates. Timestamping is critical if data is to be correlated with other monitoring outputs. A log of the operational periods of the turbine was available for the Voith HyTide dataset (refer to Table 4.1). This was greatly beneficial as it allowed portions of the data of greatest interest to be easily identified and then analysed. Had this not been available then operational periods would have had to have been identified manually. This would have been an extremely onerous task and would likely have been subject to human error. By being able to easily pick out and analyse periods of data when the turbine was operational over the full daily tidal cycle, a sampling protocol can be selected which provides the opportunity to investigate marine fauna activity around the periods of greatest interest and therefore provides the greatest potential to reduce scientific uncertainty around key impacts.

38

Another issue associated with the environmental monitoring of the Voith HyTide turbine was the failure of the sonar equipment. If this had been working, any ‘target’ incidents could have been identified from these data and then the video footage looked at in order to validate what the sonar data had indicated. Secure and resilient power connectors are required for remote monitoring equipment to limit the potential for technical faults. 4.4.4 Video data review and analysis


The Voith dataset incorporated three cameras; two on the nacelle and one on the monopile (see Figure 4.1, Figure 4.2 and Figure 4.3). Data were gathered from 12 - 18 May 2014, 20 May and 28 May-13 July 2014. Between the 12 May and 1 June, video footage was recorded during discrete periods only whereas footage was recorded continuously from 2 June to 13 July. The video data are in two different formats depending on which period it was recorded. The data from 12 to 20 May are in .asf format while the remaining data are in .mkv format. The .mkv files include timestamps whereas the .asf files do not. The .asf files are significantly larger in size (per length of footage) than the .mkv files and therefore have more burdensome data storage requirements. 4.4.4.2 Sampling protocol for video data review and analysis of Voith Hydro Hytide dataset

The initial sampling strategy for the Voith HyTide dataset to inform the assessment of the effectiveness of techniques, equipment and processes undertaken during Task 3. The sampling strategy was as follows:

• One 24-hour period from each of the following periods in the available dataset: near the start (12 May), mid-way (8 June) and near the end (7 July) was selected for analysis;

• Information on the operational status of the turbine was used to identify days that had periods when the turbine was operational and non-operational; and

• Two-minute samples from the first two minutes in each hour of available data were selected for review.

Data were continually collected during the hours of darkness, however no lighting was used to assist with environmental monitoring and therefore, footage during the hours of darkness was unsuitable for analysis. As a result, footage between the hours of 23:00 and 04:00 (inclusive) was not analysed. A summary of the Voith data that was reviewed to inform Task 3 (Assessment of the effectiveness of techniques, equipment and processes to date) is shown in Table 4.1. Depending on the findings of this initial review, additional periods could be reviewed if required.

39

Table 4.1. Summary of Voith HyTide data reviewed to inform the assessment of effectiveness of techniques, equipment and processes




Samples that cover turbine operation

No. of samples that cover turbine operation

13 May 2014 08:17:12 – 20:40:16

12 (09:00 – 20:00) 12 08:00; 09:00; 14:00; 18:00; 19:00;

5

8 June 2014 02:02:45 – 20:47:47

16 (05:00 – 20:00) 16 10:00; 11:00; 12:00; 13:00; 17:00; 18:00

6

7 July 2014 Full 24 hour period

24 18 09:00; 10:00; 11:00; 12:00; 13:00; 16:00; 17:00; 18:00

8

Trial period Six trial clips on days from May, June and July were selected and analysed using Protocol 1 (Monitoring system review). The associated spreadsheet populated by three different reviewers. This provided an opportunity to discuss any discrepancies between reviewing methods and allowed any limitations of the spreadsheet, together with other items requiring consideration to be identified and the spreadsheet adapted to take account of this. 4.5 Summary of findings

A full commentary of the effective of techniques, and processes was recorded and collated in Microsoft Excel for this dataset. This information may be made available upon request to Aquatera. A summary of the key challenges identified during the review process is shown in Table 4.2. Table 4.2. Summary of the assessment effectiveness of techniques, equipment and processes used in the Voith HyTide video data

Effectiveness Category Factor Lessons Learned

Planning and design

Camera positioning – Three cameras mounted on nacelle and / or monopile at spacing of 120° providing a 360° view of the turbine rotor.

• A 360° view of the rotor swept area is essential for collision risk detection and monitoring of near-field behaviour. There is a recommendation to ensure field-of view covers as much of the rotor swept area as possible.

• At times it is challenging to identify objects close to the blades because of the positions of camera 1 and 2 quite far back on the nacelle (see Figure 4.3).

Field-of-view • Fields of view were reduced through camera positioning which then captured approximately 40% of the nacelle. The cameras need to be positioned for optimal coverage of the rotor swept area to then allow accurate analysis of any near-field behaviour events observed.

40


Data collection

Video equipment – Tritech Tornado low light monochrome video cameras

• Video footage is not of sufficient quality to accurately monitor collision risk potential and near-field behaviour using standard visual monitoring techniques as it is difficult to confidently identify objects in close proximity to the turbine blades. It is challenging to differentiate between marine fauna, flora and inanimate objects, particularly during fast-moving tide.

• During conditions of sufficient daylight and high water clarity, colour cameras may aid in the identification of marine wildlife to species level and help to distinguish between marine fauna, flora and inanimate objects but their use will need to be considered with regard to depth and turbidity limiting their functionality


• The use of wide angle cameras would enable more accurate analysis of potential behavioural responses, however wide angle cameras may limit footage quality / sharpness.


• All cameras remained operational throughout the deployment; however biofouling obscured the view significantly on two of the cameras, by the end of the 55 day monitoring programme.

Visibility (water clarity) • Visibility varies with plankton concentration. The highest concentrations were seen in June and July.

Visibility (picture quality) • Each camera feed had vertical lines moving horizontally across the screen.

Visibility (lens) • Biofouling completely obscured the lens on Camera 3 after approximately two weeks and partially obscured the lens of Camera 2 after approximately six weeks.

• The extent and speed of growth of biofouling on the turbine’s nacelle can also be tracked.

• A small piece of rope attached to the nacelle of the turbine is directly in line with Camera 2. As it moved in the tide it ‘flicked’ in close proximity to the camera lens. This appears to have deterred biofouling organisms from settling on the lens of Camera 2.

Size of objects • It is challenging to determine the scale of objects in comparison to the turbine nacelle and blades.

Power of equipment, transmission

Power connectors • Resilient and fit-for-purpose power connectors are required for monitoring equipment to limit the potential for malfunction.

41

Effectiveness Category Factor Lessons Learned of data and data storage

Data storage • Using CDs to store data is considered a high risk strategy with a high risk of data loss or corruption.

• Manual swapping-out of storage devices is not an efficient use of man-hours.


• The log of the Voith HyTide turbine’s rotational data helped refine the data to be analysed, and then further helped with the analysis to reduce scientific uncertainty around key impacts

Format of data • This dataset came in a combination of .asf and .mkv files. The file type doesn’t appear to affect picture quality; however .mkv files are smaller in size and therefore have less burdensome data storage requirements.

Timestamping • Data were saved and filed in an assortment of time lengths but were accurately labelled to date and time. Timestamping of data, such as that applied during the collection of this dataset greatly reduces manual processing time as it allows specific times of data to be quickly identified and then analysed.



• The Voith EMP provided some general data on the camera positioning but nothing specific. It was still necessary to approximate camera positioning from interpretation of the video data. Full schematics of camera positioning should be provided in PEMPS to avoid any confusion or misinterpretation of data.

Object distances • During manual analysis it was very difficult to determine the distance of objects from the camera. This is compounded by the low visibility and biofouling on two of the camera lenses.


• Extracting the two-minute clips in the first instance is thought to save time overall and make review and analysis more accurate and efficient.

• It took approximately two days to review the three days’ worth of samples (52 two-minute clips in total).

Trial period • A dataset specific trial period is essential to ensure reviewers are familiar enough with the particular dataset and has confidence in correctly identifying marine wildlife.


• Six two-minute trial clips from this dataset were found to be more than sufficient to familiarise reviewers with the data and ensure effective analysis.

Determining number of animals

• In clips where fish are present the low resolution of the footage makes it difficult to accurately count the numbers of individuals present in the two-minute clips.

42


Reviewer analysis capability / proficiency


Reviewer fatigue • Due to the lengthy analysis periods there is potential for a reduction in the reviewer’s visual acuity.

• Reviewer fatigue was noted to occur when reviewing samples from this dataset in succession for more an hour but is specific to individuals.


• Voith data proved challenging to allow accurate identification of passing objects and marine wildlife due to the speed of tide.

Unknown video fault • During analysis it became apparent that some of the footage was faulty and therefore couldn’t be analysed.

Documents

Research Report 1225: Review of underwater video data ... · i . ANNEX 1 – TASK 3, THE REVIEW OF EFFECTIVENESS OF UNDERWATER VIDEO MONITORING TECHNIQUES, EQUIPMENT AND PROCESSES