Survey report for CRIMAC SFI 2023

CRIMAC is a center of research-based innovation funded by the research council of Norway through their program for research-based innovation (SFI). Sustainable, healthy food production and clean energy production for a growing population are important global goals, and CRIMAC will contribute to these by obtaining accurate underwater observations of gas, fish, nekton and other targets. The data will be used in conjunction with CRIMAC data from other surveys to build a reference data set for optical and acoustic target classification. The classification libraries will be used for developing methods and products toward the fishing industry and marine science.

1.1 - Time period and area

The first leg was conducted between November 15^th (Tromsø) and November 21^st (Tromsø). The work was performed in the fjords around Tromsø, from Malangen to Kvænangen. The main objective was to test the Kongsberg Sounder capabilities, including noise testing, weather window testing and safe operations. The second leg was conducted between November 15^th (Tromsø) and November 21^st (Tromsø).

1.2 - Timeline

1.3 - Vessel details

The cruise was conducted with RV G.O. Sars ( Figure 2 a) and the USV Frigg ( Figure 2 b) operated by the Institute of Marine Research and Kongsberg Discovery, respectively.

RV G.O. Sars is 77.5 m length overall, has a m aximum speed of 17 knots and a c rew of 15 in addition to space for 30 scientific crew members including instrument technicians. The vessel is equipped with Kongsberg Maritime EK80 scientific broadband echosounders (operating at 18, 38, 70, 120, 200, and 333 kHz centre frequency) and a range of other sensors (sonars, ADCPs). The vessel is equipped to deploy a wide range of additional equipment (e.g. probes, towed vehicles, pelagic and demersal trawls). More information about the vessel can be found online ( https://www.hi.no/resources/brosjyre-g.o.sars.pdf ).

USV Frigg is a uncrewed surface vehicle developed by Kongsberg Discovery (https://www.kongsberg.com/maritime/products/marine-robotics/uncrewed-surface-vehicle-sounder/). The USV is a multi-purpose unmanned surface vehicle with overall length 8 m and beam of 2.2m. The vessel weight is 4200kg. The vessel is equipped with a 125hp Steyr diesel engine. Means of communications through Maritime Broadband Radio and Iridium. The control system is delivered by maritime robotics (https://www.maritimerobotics.com/).

1.4 - Cruise participants

Prior to the experiments the EK80 on RV G.O. Sars and USV Sounder were calibrated with IMR standard survey settings for both CW and FM ( Table 1 ). Preliminary recommendations suggest FM pulses be calibrated using 2-3 spheres, where needed; and the results are merged using the EK80 software procedures. Calibrations were performed using the combination of calibration spheres and settings given in Table 2 , with some modifications as described under each sub-chapter.

The standard survey data directory ‘EK80_CALIBRATION’ was established containing three types of data:  The EK80 calibration *.raw files , the calibration results *xml files  and the post-calibration copy of the “ConfigSettings” folder containing the “TrList_calibration.xml”.  

The target strength probe was not used during the survey, and a calibration of this equipment was therefore not performed.

2.1 - EK80 on RV G.O. Sars

G.O. Sars is equipped with six drop-keel mounted echosounders (Simrad EK80) capable of continuous wave (CW)/narrowband or frequency modulated (FM)/broadbanded pulse generation. These have nominal frequencies at 18, 38, 70, 120, 200, and 333 kHz.

Both the 18 and 333 kHz transducers were new and installed prior to the survey and had not been calibrated previously.

Ship echosounders were operated with both CW and FM acoustic pulses. Settings for these were chosen to conform with standard IMR settings, to avoid undesirable effects such as acoustic “crosstalk” in broadband data. This influenced the choice of the acoustic bandwidth, power, and pulse duration settings (Table 2). Standard CW pulse settings (Korneliussen et al., 2008) were used, but with reduced power to match the power setting of alternating CW / FM pulses.

Calibration plan and calibration data collection progress  

1. Two anchors and bottom depth ≥40 m

2. CTD cast. Update EK80 ‘environment’ (average sound speed, temperature, and salinity between transducer and the calibration target depth).

3. Calibration starts with the largest sphere (WC57.2).

4. When using multiple spheres to calibrate the EK80 channel use largest sphere first (or, update TrList_calibration.xml with calibration data starting with the largest sphere),  

5. Use 500 g weight to stabilise the smaller spheres (WC35-25-22-20), 4m between sphere and the weight (0.5mm nylon),  

6. Sphere netting should have minimum 1 m long netting loop. (i.e., ≥1m between the triple winch suspension lines knots/loops and the calibration target),  

7. Make a post-calibration copy of the TrList_calibration.xml file (save whole folder ‘ConfigSettings’ to the survey directory ‘EK80_CALIBRATION’),  

8. The Table 3 top-down progression is based on the order of the calibration target deployment.  

Initially the ES18 mk2 was not calibrated (either in FM or CW) due to the high noise levels. The original plan was to calibrate the ES18 mk2 according to Table 1 and Table 2 , and perform additional calibrations with 1 and 4 ms FM up pulses in addition to 2 ms (which is assumed to be future standard setting), and full allowed bandwidth. In addition, two power settings would have been used for 2 ms, “low” and “high” power (400 and 800). Further we planned to compared SNR and noise at those combinations of settings and investigate the transducer performance drop at the outskirts of the band. The ES18 mk2 was calibrated later during the survey (25.11.2023) in a narrow frequency band (FM mode, 22-28 kHz) as this band was found to be reasonably clean (see chapter Modified 333 & 18 kHz ).

Additionally, all echosounders were run simultaneous and sequentially in survey settings with a calibration target located within the acoustic beams. This was performed to collect a data set for evaluating whether simultaneous calibration of all channels could be performed, and to investigate the crosstalk between the ES18 mk2 (at 400 and 800 W) as the 18 kHz power will likely be the driving factors for crosstalk.

2.2 - EK80 on Sounder

The EK80 on the Sounder “Frigg” was calibrated in port (Tromsø, November 15-16). The USV was docked alongside GO Sars, and similar settings and spheres to the research vessel was used. The spheres were deployed with two lines (as opposed to three for the vessel) and manually moved along the vehicle to move the spheres through the beam. The echosounders were powered by the Sounder and a remote desktop was used to perform the calibrations.

Calibration plan and suggested calibration data collection progress (Table 4):  

1. CTD cast. Update EK80 ‘environment’ (the average sound speed, temperature, and salinity between transducer and the calibration target depth),  

2. Start with largest, WC57.2 sphere,  

3. When using multiple spheres to calibrate the EK80 channel use largest sphere first (or, update TrList_calibration.xml with calibration data starting with the largest sphere),  

4. Sphere netting should have minimum 1 m long netting loop (i.e., ≥1m between the triple winch suspension lines knots/loops and the calibration target), 

5. Make a post-calibration copy of the TrList_calibration.xml file (save whole folder ‘ConfigSettings’ to the survey directory ‘EK80_CALIBRATION’),  

6. Table 4 top-down progression is based on the order of calibration target deployment.  

Due to challenging environment, i.e. other scatterers in the water column interfering with the signal from the sphere, all raw files had to be manually cleaned before updating the calibration values.

The echosounder beam of the EC150-3C was not calibrated, this was not possible due to the narrow beam in combination with short range to the sphere and seafloor.

3.1 - Objectives

A new broad banded 18 kHz transducer and a modified 333 kHz transducer were mounted in the instrument keel of RV G.O. Sars prior to the 2023001016 survey, and the objective was to test the performance of the two systems. The introduction of a new broadband 18 kHz transducer is important for, e.g., identifying mesopelagic fish, where the higher echosounder frequencies, like the 333 kHz, are useful for detection and categorisation of smaller organisms like zooplankton, such as the copepod Calanus finmarcicus.

3.2 - Methods

The new broadband 18 kHz transducer was the second in a pilot series of 3 units. The first 18-kHz transducer in this series performed well onboard the Kongsberg vessel RV “Echo”. The EK80/18kHz system was going to be extensively tested and verified during this survey.

The 333 kHz transducer had a modified grounding where the objective was to increase the EMC immunity. A shield termination to sea was introduced on the transducer to improve EMC immunity against interference and noise along the cable routing.

Both transducers were tested during the survey. Noise estimates and echosounder data were compared to the earlier versions of the transducers. The comparisons were both visual and based on noise estimation from the LSSS system.

3.3 - Preliminary results

The impression was that the new EK80/18kHz system had severe challenges appearing as ringing. The performance in CW was worse than the previous narrowband 18 kHz transducer when comparing data from this survey with the data from last year’s CRIMAC survey ( Figure 3 ). The figure shows that the system performed worse than the original from the 2022 survey. We suspected that the apparent ringing could be caused by the tight fit between the hull and the transducer, and we lowered the transducer below the drop keel by using longer bolts allowing the transducer to hang freely below the keel. The problem remained. More work to investigate the causes are needed. The apparent ringing was less prominent at the higher frequency ranges of the transducer.

The increase in performance for the modified EK80/333kHz system was encouraging ( Figure 3 ), with much less noise compared to the original installation. The probability density functions for the noise estimated from the LSSS system demonstrate a clear improvement in performance for the modified system ( Figure 4 ). This is also seen from the echograms for the original ( Figure 5 ) and modified ( Figure 6 ) 333 kHz data, respectively. Although the maximum range of the 2022 data appears to be similar between the 2023 and 2022 examples, the 2022 example school is much stronger. There is also some noise that is not removed from the FM data, which is likely due to narrowband noise in the broadband signal.

4.1 - Objectives

The objective of this task was to assess the echosounder data quality from the USV Frigg in variable sea states and headings relative to the wind. Both noise from passive recordings and data quality based on masking from bubbles under the transducer were assessed.

4.2 - Methods

We collected both passive and active acoustic data to assess data quality at various sea states, different headings to the wind, and variable vessel speeds. Three different sites were chosen and ranged in sea state from sea state 2 to 4. The depth in the experiment areas ranged from 350 to 420 m ( Table 5 ).

For sea state 2-3 and 4 an octagon pattern was chosen to cover 8 different headings to the wind ( Figure 7 ab). For the sea state 2, a transect in and out the fjord was chosen ( Figure 7 c), assuming that the wind did not affect the data quality at this sea state. An additional treatment was introduced for the latter, where the engine rpm was kept constant for a 5 minute-interval followed by a 5 minute-interval where the rpm was manually varied. The varying rpm range was similar to the rpm range during autonomous operation for the current speed control settings. The latter was included to assess whether there was any effect of varying the rpm on the noise levels, and particularly from propeller cavitation. We also measured the noise levels during the tow test, simulating the effect of operating the USV with the engine disengaged. We also tested the noise levels with the vessel stationary and running the engine with the propeller disengaged using RPMs like normal operation.

Each transect or octagon pattern was repeated with four different vessel speeds (3, 5, 7 and 9 knots), and each leg of the octagon pattern had a duration of 4 minutes each. For the data collection in the calm state 5 minutes were used for each speed, allowing time for the speed to stabilize when changing speed. Each sequence covering the different headings to the wind and speed was repeated 3 times per sea state. After the completion of the octagon sequence in the outer region, Frigg was drifting while logging acoustic data (Austerhola, Frigg, Speed=0). The purpose was to estimate the amount of bubbles in the upper water column not caused by the platform. RV G.O. Sars covered the same transects as USV Frigg. This allowed us to compare the data quality metrics between RV G.O. Sars and USV Frigg.

Sensor data were logged at the platform during acquisition. These included GPS position, engine RPM, heave, tilt and roll, wind direction and wind speed. The sea state was taken from the RV G.O. Sars logbook, where the sea state is manually assessed by the personnel on the bridge. All systems were time synchronized, and echosounders on USV Frigg and RV G.O. Sars were calibrated in CW and FM using IMR standard settings (c.f. calibration table). The echosounder data were acquired using the EK80 “advanced sequencing”, where four ping groups were set up covering combinations of passive/active and CW/FM transmissions. Each group included all frequencies, but the 18 kHz transducer was included in the CW groups only since it lacks broadband capabilities. The combinations were (CW+FM)*4 + (2*CWpassive+2*FMpassive).

The passive noise estimates were estimated using the Korona module in LSSS for the different frequencies ( Figure 9 and Figure 7 ). The noise estimates were provided for each frequency channel and platform as a function of time, and the time interval for each treatment ( Figure 8 ) was used to assign the noise to the corresponding treatment. The data was merged using python and imported in R as a tibble and plotted using the tidyverse packages.

To compare the effect of the reduction in bottom echoes due to potential bubble sweep down caused by vessel motion, the bottom echoes were integrated for the transects in each location for both RV G.O. Sars and USV Frigg. The ratio between the transects are used as a proxy for bubble attenuation while noting that the echo integrated bottom signal is affected by tilt and roll of the transducer. The preliminary results must be interpreted with this in mind.

4.3 - Preliminary results

The raw noise estimates were exported from the LSSS system ( Figure 9 ). The initial results show that the noise estimates are substantially higher and varies more for USV Frigg than for RV G.O. Sars. A more detailed view is provided when dividing the data based on sea state (location), vessel speeds, and platform ( Figure 10 ). Much of the variability is explained by platform speed for USV Frigg. Towing the vehicle at the same speed intervals results in lower noise estimates ( Figure 11 ). This suggests that the cause of the noise is associated with the propulsion system including the propeller, and the characteristics are typical for propeller cavitation. It is therefore likely that the propeller cavitation is causing the variability in noise that is related to the changing speeds.

The data quality is also affected by the sea state. The echogram from the USV during the Austerhola experiment in sea state 4 ( Figure 12 ) show indications of bubble damping as well as increased noise for sea state 4 ( Figure 12 ) compared to sea state 2-3 ( Figure 13 ). The bottom echo recorded from G.O. Sars and USV Frigg for both areas was integrated and compared ( Figure 14 ). For the sea state 4 the integrated backscatter from the bottom signal was lower from Frigg compared to G.O. Sars, and the difference was higher for the lower speed. Note that the echo integration was not corrected for transducer motion, and without adjusting for transducer motion we cannot conclude on the cause of the difference. For sea state 2-3 the difference was lower and consistent between different platform speeds.

This task investigated whether the quality of the collected acoustic data could be increased by adjusting the navigation of the USV during operation. A controller was proposed that takes in bubble induced ping drop out as an input to adjust heading and speed of the vehicle USV Frigg.

5.1 - Objectives

Objective 1 : Quantify correlation between sea state, data quality and navigation.

To provide a justified control architecture, the model of correlation between input/observed parameters (sea state and data quality) and navigational control outputs (speed and heading) must be established.

Objective 2: Evaluate controller performance.

The objective was to evaluate the controller’s performance in terms of reduced ping drop out and time increase. Further the metrics used for navigational accuracy (position), data quality estimation (ping drop out) and sea state (peak wave frequency, mean wind velocity) were evaluated in terms of robustness to tuning changes.

Objective 3: Identify platform-related challenges and weaknesses in the data sampling approach.

Since this work package takes an iterative approach, the purpose of this data set is also to identify issues in fundamental algorithm design, data logging and platform interface as early as possible.

5.2 - Methods

Algorithm The purpose of the algorithm is to provide a real-time feedback control loop, that outputs a desired speed and heading based on the observed ping dropout rate. The speed control is a PID controller, which scales a defined error signal at every timestep. The error signal is defined either as the difference of (1) the current ping dropout rate and a desired percentage drop out or (2) the sum of roll and pitch movement and a set threshold.

Platform Interface The command of the Sounder platform is executed via the Maritime Robotics backseat driver interface. The backseat driver is a command abstraction layer running on the on-board computer (OBS). It can be used to receive basic navigational information and transmit command signals to the platform via a TCP connection (cf. Maritime Robotics document MR2USV/D101).

5.3 - Preliminary Results

When testing the controller, two versions were used: One that took ping drop out as a control input, and one that acted on roll and pitch. No dropouts were observed during the experiments, so that the threshold for dropouts would either result in a constant maximum velocity (algorithm always detected no dropouts at high threshold) or minimum velocity (constant dropouts due to low threshold), cf. Figure 15 . When based on roll/pitch magnitude, the platform reacted reliably, reducing the attitude motion by velocity adaption as compared to no controller active, cf. Figure 16 .

The main weaknesses in the controller design are observed to be: (1) Ping drop out as control input. The drop out threshold was observed to have a large impact and depend on the sea state, velocity range and platform. It further is a delayed input with an observed update delay of 5s, whereas roll and pitch information delay were less than a second. (2) Real-time echogram processing: Access to the echograms can (for now) only be facilitated on the same local machine as the EK80 software.

It was further evaluated whether taking in passive noise will benefit the stability of “data quality” as an indicator but was deemed to not be useful as a control input, since it does not depend on sea state or platform motion but rather on electrical and propeller noise.

6.1 - Objective

The original objective was to use the 18 kHz broadband echosounder to measure mesopelagic species and validate species composition with image data from the Deep Vision trawl camera. The hypothesis was that the wide band covers the resonance frequencies of pearlside, Maurolicus muelleri , and can be used for sizing and distinction between species. The lantern fish Benthosema glaciale replaces its gas filled cavities with oil sacks, and we would like to see if we could detect this using BB sweeps around the operating frequency of this transducer. A new broadband 18 kHz echosounder from Kongsberg Discovery was installed at cruise start, but high noise levels (ringing) in the whole range prevented us from using it in broadband. The objective was then changed to collect data to describe the species composition in the fjords combining broadband acoustic methods and image data from trawl camera, focusing mainly on Maurolicus and Benthosema . Additional objectives were to test a newly developed image detection algorithm for mesopelagic species, implement it into the data pipeline on GO Sars and collect new training data.

6.2 - Method

Data were collected in Ullsfjord on the 23. November (between 09:46 and 15:20 UTC) and in Lyngsfjorden on the 24. November (between 13:00 and 16:00 UTC). This time of the year dawn is at 08:30 and dusk is at 14:20. Acoustic data were collected using the ship based Simrad EK80 wideband transceiver and six wideband split beam transducers ES18-11 (14 kHz single frequency), ES38B (34 – 45 kHz), ES70-7C (50 – 85 kHz), ES120-7C (95 – 165 kHz), ES200-7C (170 – 260 kHz) and ES330-7C (280 – 380) in transects along the mid fjord ( Figure 17 ). Trawl samples were collected with the Harstad sampling trawl. The trawl is 84 m long, has 200 mm mesh in the wings and forepart and 60 mm mesh in the belly and aft sections. Thyborøn type 7a trawl doors were used. The trawl was equipped with a Scantrol Deep Vision camera system mounted in front of the cod end. Trawling was made with open codend and the Deep Vision recorded stereo images at a rate of 10 frames/second. Trawl geometry was monitored with sensors on the doors and headline and the trawl opening height was 17 – 19 m and door spread about 65 m. An external depth sensor was attached to the Deep Vision system. Long trawl hauls were made along the centre of the fjord, covering the same area as the acoustic transects, with stepwise trawling at different depths. The aim was to trawl minimum 15 minutes in each depth layer. The depths were selected based on the scattering layer depths as observed on the echosounder varying between 30 and 200 m.

The deep vision data were analysed using a newly developed machine learning algorithm for mesopelagic species. The detector is trained on Maurolicus, Gonatus, krill, Benthosema, Sergestes and Periphylla. The model uses a smaller scaling factor for predictions than the previous version developed for pelagic species (min image size=1000 instead of 800)  (Allken et al. , 2021) . Different optimal score thresholds for different species were also used. These were determined empirically on a validation data set.

6.3 - Preliminary results

We aimed for distinct layers of mainly, Maurolicus sp. and Benthosema to investigate if broadband data can be used to distinguish species and sizes. Small Maurolicus tend to be distributed near surface at daytime, while larger Maurolicus slightly deeper. Benthosema is typically at depths deeper than 300 m. At nighttime the Benthosema tend to migrate to shallower water. The scattering layers in Ullsfjord or Lyngenfjord were not distinct, but loosely distributed in the whole water column with highest densities from about 150 to seabed (~250 m). Balsfjord and Fugløysund west of Arnøya were also covered, but with no observations of clear mesopelagic layers. However, no clear layers of mesopelagic fish were discovered during the survey.

Echosounder data indicated that large quantities of krill were present in both fjords. This was confirmed by the results from deep vision image analyses. Krill dominated in the images ( Figure 18 ). Maurolicus was also present in both fjords and perphylla and sergestes were present in Ullsfjord in smaller quantities. Other species not included in the algorithm were likely also present. A thermocline with rapid change in temperature and salinity was present at about 100 m in Ullsfjord and 150 m in Lyngenfjord ( Figure 19 ).

7.1 - Objective

The objective of this task is to compare the measured acoustic density of herring during night between G.O. Sars and USV Frigg.

7.2 - Material and methods

An area of Kvænangen fjord that contained high densities of overwintering herring was chosen for the experiment. The area was chosen within the herring distribution area to avoid area conflicts with the commercial herring fishery and the gill net fishing activities in the fjord. The website “sildelaget.no” was used to study the herring fisheries, whereas the gill net activity information was downloaded from “fishinfo.no” as a redskap.gpx layer. Based on this we defined one stratum polygon for the first run ( Figure 20 ), and another for the second run ( Figure 21 ) using the OpenCPN software with a high resolution map downloaded from https://www.kartverket.no/. The surveys were carried out during night when the herring are distributed closer to the sea surface.

For each coverage, the R function surveyPlanner from the Rstox package was used to generate evenly distributed transects with random start position in a zig-zag pattern in the strata for RV G.O. Sars and USV “Frigg”. Table 8 shows the parameters used for the transect planning for each stratum and coverage by platform.

The echosounders on RV G.O. Sars and USV Frigg was operated using sequential pinging between FM and CW using IMR standard settings. The acoustic 38 kHz data was scrutinized using the LSSS system, and all backscatter above the integration threshold was assigned to herring. The data was exported to the ICES standard file format and imported into StoX  (Johnsen et al. , 2019) . StoX 3.6.2 was used to make StoX projects for each of the surveys. The NMD Echosounder xml files produced in LSSS were imported into their respective StoX project. For the preliminary results the length distribution from 2020 was used, but biological information were collected from the catch sample lottery (https://www.sildelaget.no/no/kvoter-og-fangst/fangst/fangstproevelotteriet) that will be used to update the estimate after the cruise. The biomass was calculated using the standard acoustic target strength for herring  (Foote, 1987) : TS = 20log L – 71.9. Based on these data, mean and 90%- confidence interval of biomass were estimated using standard baseline and bootstrap ( runs = 1000 ) settings in StoX.

7.3 - Preliminary results

Visual inspection of the echograms shows that the depth distribution of the upper herring layer was distributed shallower below USV Frigg ( Figure 22 ) than G.O. Sars ( Figure 23 ). Also, the acoustic densities appear higher from the USV Frigg echograms compared to RV G.O. Sars. The NASC estimates from StoX were significantly larger using USV Frigg than RV G.O. Sars (two sample weighted t-test, p<0.005), and about 3.5 times higher for both for experiments ( Figure 24 ).

8.1 - Objective

Norwegian catches of the pelagic species herring, mackerel and capelin are about 1 mill tons annually, and about 80% of these are caught using purse seines (Fiskeridirektoratet). This approach make the fish easily accessible for marine mammals and birds, and some species, such as killer whales, may follow fishing vessels in the search of “a free meal”  (e.g. Vogel et al. , 2021) . This may create unfortunate incidents with whales getting entangled in the nets; with potential for both injury and even death for the animal and lost catches and gear for the fishermen  (Bjørge et al. , 2023) . In the ongoing project “Kartlegging og testing av metoder for å redusere interaksjoner mellom fiskeri og hval”, it is tested whether sound can be used to keep whales away from ongoing fishing operations. The project results show a good deterrent effect on killer whales  (Langstein, 2023) .

These sounds can be detected by herring, a species with good hearing due to a duct connecting its swimbladder to the inner ear  (Enger, 1967) . If the herring respond to the sound signals, it may negatively affect the fishing operation. The objective of this task is to test whether the whale deterrent sounds affect the behaviour of herring or not.

8.2 - Methods

The TAST (target specific startle technology) system, developed by Genuswave LDT, transmits sound pulses that are of short duration and rise time with random intervals between transmitted pulses. Based on the published literature on this system  (Götz and Janik, 2015; Langstein, 2023) , we generated sound sequences with the following properties:

• Each signal was 200 ms in duration

• Signal rise time was less than 5 ms

• Interval between pulses was random, varying from 1 to 40 seconds

• Signals were band passed white noise

The TAST system also randomizes the frequency bands within the sound sequence, however, in our experiment, we aimed at testing different frequency bands separately to assess if any frequency bands influenced herring behaviour. Consequently, we separated the sound clips into 4 different frequency band, each covering one octave:

Frequency band 1: 200-400 Hz

Frequency band 2: 400-800 Hz

Frequency band 3: 800-1600 Hz

Frequency band 4: 1600-3200 Hz

In addition to the TAST signals, we also played back recordings of killer whales. These sounds were recorded in Vestfjorden, Northern Norway in 2006, of killer whales feeding on overwintering herring. Recordings consisted both of calls and echolocation signals.

The playbacks were done with an underwater speaker ( Lubell Labs LL1424HP) designed for Underwater Acoustic Deterrent Systems. It transmits within the frequency range of 200Hz - 9kHz, with maximum power at 600 Hz. Sound recordings, to monitor the sound level and proper output, were done with a sound trap 300 hydrophone from ocean instruments.

The sound source was lowered from a crane from the side of the ship to a depth of 15 m, limited by its cable length. A hydrophone was deployed 10 m below the source to measure the sound source level close to the source. A second hydrophone was lowered from a crane to 100 m depth, which was the assumed depth of the main herring layer, to measure the sound level received by the herring (c.f. Figure 25 ).

The experiment was conducted as a block design, where each block consisted of a 10 min sequence of each of the 4 frequency bands. The order of the frequency bands within each block were randomized prior to start of the experiment. Also, a random selection of each of the 20 available clips were made for each frequency band.

Prior to start, we surveyed the area with echosounder and sonar to map herring distribution in the area. Once a suitable location was selected, the sound source and hydrophones were lowered. After deploying the instrument, we allowed the fish to acclimatize after any potential disturbance from the operation. The ship’s main engine was turned off and the ship were drifting to avoid sound from the ship to disturb the fish. The 10 min sound sequences were played with 10 min no-sound control between, and we always started with a 10 min control.

After each block, we waited 1 h before starting the next exposure. During this period, the equipment was brought back on deck, and the ship relocated before the next block.

Observations of herring behaviour were conducted by the ships echosounder (EK80), measuring the density and depth structure of the herring layer, hence detecting potential vertical or horizontal avoidance behaviour.

8.3 - Preliminary results

The experiments were conducted over 3 days in November 2023, and a total of 10 blocks were conducted, and 4 playbacks of killer whale vocalisations ( Table 9 ). All experiments were conducted in Kvænangen, which is the main overwintering area for the Norwegian Spring Spawning herring. Due to differences in wind conditions, the distance drifted during a block varied from 3500 m (block 2) to less than 100 m (block 6).

Before starting the experiments, the sound source was tested with all the frequency bands and levels measured with the hydrophones 10 m below the source and at 100 m depth. Based on these measurements the source level for the 4 frequency bands were calculated ( Figure 26 ).

A preliminary screening of the echograms did not show any apparent reaction of the herring (c.f. Figure 27 ). Further, no apparent indication of a change in acoustic density could be determined from the preliminary screening, c.f. block 1 in Figure 28 .

The data is organized in accordance with the IMR data organization procedure. In this section the placement of each data set is described as well as a short description of each individual data set. The headings are equal to the folders in the data structure.

The data from G.O. Sars is stored at IMRs secure data storage system under \cruise_data\2023\S2023001016_PGOSARS_4174

The USV Frigg uses a separate cruise number, and the data files from USV Frigg are placed under \cruise_data\2023\S2023301002_PFRIGG_10318\.

9.1 - ACOUSTIC DATA

The shipboard EK80 echosounder channels were calibrated 2023.11.15-11 prior to data acquisition. The EK80 echosounder channels onboard USV Sounder Frigg were also calibrated 2023.11.15 prior to data acquisition.

Ship-borne unprocessed EK80 data from FF G.O. Sars and USV Frigg was stored in accordance with the IMR data storage structure at \S2023001016_PGOSARS_4174\ACOUSTIC\EK80\EK80_RAWDATA and \S2023001016_PFRIGG_10318\ACOUSTIC\EK80\EK80_RAWDATA. There were several different strategies for collecting data, so that the content of the EK80 data files were not all the same: some contained CW-data only, some contained a mixture of CW and FM-data. They were, however, still all stored in the specified directory for each of the two surveys.

Due to the different content of the data, the processed data were not placed in a single directory, but were placed in a cluster of directories like …\ACOUSTIC\LSSS\KORONA_, i.e. directories like KORONA_Octagon1_GOSars_noise. Each of those directories contain the processing setup in the sub-directory “\copiedConfigFiles”.

The overall organizing of data from the survey is stored in the survey file localized at \ACOUSTIC\LSSS\LSSS_FILES. A survey-file keeps track of how the directories are organized, e.g. which Work-files to use, which KORONA-files to use and which preprocessing setup to use.

Files exported from LSSS fro RV G.O. Sars are placed in \S2023001016_PGOSARS_4174\ACOUSTIC\LSSS\EXPORT\EchogramPlot. Each of the data-files contain a prefix to identify its origin, and also a time-tag to identify which raw-data they are based on.

There are similar files from USV Frigg, but there is an additional directory \S2023301002_PFRIGG_10318\ACOUSTIC\LSSS\EXPORT\EchosounderData. The data in that directory are all at 38 kHz merged from its original raw datafiles.

Finally, data from the database are placed in directories like Reports_Survey1_Frigg, Reports_Survey2_Frigg or Reports_Survey1_GOSars, Reports_Survey2_GOSars. There would on an ordinary survey only be one directory per platform (named REPORTS), but here it was necessary to distinguish between the abundance estimation surveys of G.O. Sars and Frigg.

9.2 - StoX project

The Stox projects are placed under

\cruise_data\2023\S2023001016_PGOSARS_4174\CRUISE_DOCUMENTS

9.3 - BIOLOGY

9.3.1 - DEEP_VISION

The deep vision data is stored at:

\cruise_data\2023\S2023001016_PGOSARS_4174\BIOLOGY\CATCH_MEASUREMENTS\DEEP_VISION

9.4 - USV Frigg operator notes

During the USV Frigg operation, an operator_notes.txt file was used as notes between operators, handling the USV in 8-hour shifts. The document is located here:

\CRUISE_DOCUMENTS\OTHER_DOCUMENTS\operator_notes.txt

12.3 - USV Frigg Backseat control software

The backseat control software used to allow the USV to modify its course based on feedback from echosounder, as described earlier in this report, is stored here:

\CRUISE_DOCUMENTS\OTHER_DOCUMENTS\CONTROLSW\*

12.4 - USV Frigg Vehicle control system

The topside operating software vehicle control system files are stored and includes mission execution files. Files are named “Sounder003.csv*” and are modified during the cruise:

 \CRUISE_DOCUMENTS\OTHER_DOCUMENTS\VCS\VAR_LOG_VCS\Sounder003.csv*

These files does not contain a descriptive header; but contain the timestamp, latitude and longitude of the USV, as well as other meta-data transferred from the USV to the topside VCS software while operating.

Logfiles are also as generated by the Maritime Broadband Radio web interface (MBR). The files can be found here:

\CRUISE_DOCUMENTS\OTHER_DOCUMENTS\VCS\MBR_LOGS\*.log

With the following headers:

Field 1: [SN] Unit serial number

Field 2: [S] Unix time in seconds either from boot or GPS

Field 3: [ms] Milliseconds after seconds

Field 4: [lat] GPS lat

Field 5: [lon] GPS lon

Field 6: [C] Temperature

Field 7: [mA] Input current

Field 8: [mV] Input voltage

Field 9: [MHz] Frequency MHz part

Field 10: [kHz] Frequency kHz part

Field 11: [Type] Product variant number

Field 12: [dB] Antenna element gain

Field 13: [Frames] RX Frames received OK

Field 14: [Frames] RX Frame errors

Field 15: [Frames] RX CRC 32 Errors

Field 16: [Frames] Frames relayed

Field 17: [Frames] TX Frames OK

Field 18: [Frames] TX frames dropped due to buffer overrun

Field 19: [Frames] TX rames dropped due to no link

Field 20: [Frames] TX frames dropped due to MAC Busy

Field 21: [Frames] TX Frames with missed ACK

Field 22: [kb/s] Total TX bandwidth to mac

Field 23: [kb/s] Total RX bandwidth from mac

Field 24: [Sites] Sites in following site table

Field 25+21*(n-1): Extra information about connection to designated site

Note that leaving the MBR to log indefinitely will result in the MBR web interface crashing with the firmware in use at the time. The MBR link will still work.

 Allken, V., Rosen, S., Handegard, N. O., and Malde, K. 2021. A deep learning-based method to identify and count pelagic and mesopelagic fishes from trawl camera images. ICES Journal of Marine Science. https://doi.org/10.1093/icesjms/fsab227 (Accessed 26 November 2021).

Bjørge, A., Moan, A., Ryeng, K. A., and Wiig, J. R. 2023. Low anthropogenic mortality of humpback (Megaptera novaeangliae) and killer (Orcinus orca) whales in Norwegian purse seine fisheries despite frequent entrapments. Marine Mammal Science, 39: 481–491.

Enger, P. S. 1967. Hearing in herring. Comparative Biochemistry and Physiology, 22: 527–538.

Foote, K. G. 1987. Fish target strengths for use in echo integrator surveys. Journal of the Acoustical Society of America, 82: 981–987.

Götz, T., and Janik, V. M. 2015. Target-specific acoustic predator deterrence in the marine environment. Animal Conservation, 18: 102–111.

Johnsen, E., Totland, A., Skålevik, Å., Holmin, A. J., Dingsør, G. E., Fuglebakk, E., and Handegard, N. O. 2019. StoX: An open source software for marine survey analyses. Methods in Ecology and Evolution, 10: 1523–1528.

Langstein, E. 2023, May 15. Assessing the deterrence effect of target-specific acoustic startle technology on killer whales and humpback whales during interactions with Norwegian purse seine herring fishery. UiT Norges arktiske universitet. https://munin.uit.no/handle/10037/29479 (Accessed 5 January 2024).

Vogel, E. F., Biuw, M., Blanchet, M.-A., Jonsen, I. D., Mul, E., Johnsen, E., Hjøllo, S. S., et al. 2021. Killer whale movements on the Norwegian shelf are associated with herring density. Marine Ecology Progress Series, 665: 217–231.