F. Basarı et al.

elected to deploy either fixed cabled stations with real time data transfer
and continuous power, or autonomous systems that had to be retrieved
ntermittently for data download (see Table 2). The measurement cycle
varied between the stations, while some stations (mainly the cabled
ones) were measuring continuously, most stations recorded using duty
cycles (Fischer et al., 2021).
A variety of measurement systems were used (cf. Fischer et al.,
2021), as project partners could choose which devices they preferred or
already had available for the measurements. However, to ensure
comparability between the measurements, the hardware used to record
sound data had to meet technical standards set within the JOMOPANS
project and had to be deployed 2-3 m above the seafloor (Fischer et al.,
2021). Moreover, several quality assurance checks were performed on
the data following the project's standards for data processing — ie.
checks for missing data and data consistency, removal of contaminated
data (recordings before deployment, after recovery and during deployment,
 recovery and maintenance), checks for clipping, distortion and
spurious signals — i.e. low frequency periodicities, periodic impulsive
sounds and noise floor limitations (Ward et al., 2021).
Each partner ensured that the measured waveform data was converted
 to calibrated absolute power density spectra. The sound pressure
level (SPL) was defined in the JOMOPANS Terminology Standard
(‚Robinson and Wang, 2021) based on ISO 18405:2017 and represents
the mean-square sound pressure averaged over a specified time,
expressed as a level in decibels relative to a reference sound pressure
value of 1 uPa. Data with averaging times (T) of 1 s or 20 s were shared
among the project participants depending on the individual country's
data policy.
Making use of Parseval's theorem the sound pressure level is given
DY!

1 ,
— P
I = 10log1o (— > [Po] )

where N is the number of samples within the averaged time T, po is
the reference sound pressure 1 wPa and P,/ are the coefficients in the
discrete Fourier transform (DFT) of the sound pressure time series. The
sound pressure level in one-third octave frequency bands (base-10)
(TOL) were obtained by calculating the power sum of the individual
coefficients in the respective frequency band (Ward et al., 2021). These
requency bands are commonly used in acoustics (IEC 61260-1:2014)
and are also termed “one-tenth decade” or “decidecade” bands (ISO
18405:2017). It was agreed to measure and share TOLs for the 34 onethird
 octave bands between 10 Hz and 20 kHz (Merchant et al., 2018).
The acoustic metrics (presented in Table 1) were chosen to be 1)
realistic (in terms of practical constraints on monitoring), 2) ecologically
relevant (reflect the way marine species are affected by sound) and 3)
easily understood by non-specialists (Merchant et al., 2018).
To ensure a consistent application of the processing methods becween
 measurement partners, a benchmark exercise was carried out
where each partner processed synthetic noise signals generated with
known statistics in a benchmark test of their own analysis routines. The
“est verified that, notwithstanding small variations between the partaers'
 results (i.e. deviations of power averaged sound pressure levels
below 1 dB), reported spectral density and power spectrum could be

Table 1
Summary of acoustic metrics.
Attribute Specification
>hysical quantity Sound pressure level in dB re 1 uPa
Snapshot 1sor20s
duration
Analysis One-third octave (base 10)
bandwidth
Zenter From 10 Hz to 20 kHz, defined according to the base-ten
frequencies ronvention (ANSI, 2009; IEC 61260-1:20147

Marine Pollution Bulletin 198 (2024) 115891
processed consistently (Ward et al., 2021).
In total 4720 days of acoustic recordings were collected during the
project lifetime at the 19 measurement stations (see Fig. 5). Gaps in the
data coverage can in part be explained by the initial plan to only measure
 during the year 2019, although most countries continued
measuring during 2020 as part of their national monitoring activities.
Furthermore, some stations were not primarily JOMOPANS stations and
therefore coverage was not expected to be complete. As such, data from
the Swedish Vinga (01-SE-VIN) station, the two Norwegian stations (13-NO-LOV
 and 14-NO-NTR) and the Danish EDNA (18-DK-EDA) station
are only available for the year 2019. Excluding data that did not pass the
]uality assurance checks described above further introduced gaps in the
data coverage.
Sonsidering the ambitious aim to measure at 19 stations for one year
and the challenges associated with measuring at sea and during a
pandemic (lost equipment, irregular maintenance intervals, operational
difficulties for working at sea) the collected dataset during JOMOPANS
is the largest of its kind and provides a remarkable and unique asset.

2.2. Wind data

Hourly modelled wind data for the entire study area were retrieved
from the E.U. Copernicus Marine Service. These data were generated
based on the ERA5 climate reanalysis, which is produced by the European
 Centre for Medium-Range Weather Forecasts (ECMWF) (Hersbach
at al., 2018). The data were provided on a 0.25° x 0.25° spatial grid and
were subsequently interpolated to the respective positions of the
JOMOPANS measurement stations (de Jong et al., 2021). The hourly
amplitudes of the modelled wind speed were derived from the (zonal) Uınd
 (meridional) V- components by:
ha = vl tw?

Annual mean wind speeds of the year 2019 at the individual stations
are summarised in Table 2 and an overview map of mean wind speeds in
the North Sea for 2019 is provided in Fig. B in the supplementary
material.
2.3. Shipping data

To quantify the anthropogenic share of underwater noise related to
vessel traffic, AIS data and VMS data were utilised as sources for the
spatial distribution of ships in the North Sea. According to SOLAS re-]uirements
 (International Convention for the Safety of Life at Sea) all
ships of 300 or more gross tonnage (GT) and all passenger vessels must
be fitted with an AIS system. Under European Union legislation fishing
vessels exceeding 15 m are required to carry VMS transmitters in order
to track and monitor fishing activities.
In JOMOPANS, processed AIS and VMS (provided by Denmark,
Germany, Norway and Sweden) data for 2019 were acquired from the
“rench company Quiet Oceans, which fused AIS and VMS data, cleaned
autliers and duplicates, and interpolated the data onto a regular grid.
This merged dataset was further processed and checked for validity
and consistency by TNO before it was used as input parameter for the
sound maps of the North Sea that were also produced as part of the
JOMOPANS project (de Jong et al., 2021). The correlations presented in
this study, are based on the merged and processed dataset for the year
2019 as provided by TNO.
To give an overview of the average shipping density per acoustic
measurement station, the number of vessels present within a 35 km
radius (the estimated radius within which ships could be detected
acoustically) during 15 min periods (as explained in Section 3.2.2) were
;ummed over the year 2019 per station and referenced to the sum at the
3elgian Westhinder station (08-BE-WST), where the most vessel pas-;ages
 were registered (Table 2). An overview of the shipping density in