F. Basarı et al.

at 10 m above the sea surface and range corrected point estimates of
ships' contribution to ambient sound levels. TOLs were further corre-‘'ated
 with bottom currents to evaluate the contaminating effect of flow
noise on the measurements.
Since the averaging time (1 s or 20 s) and the duty cycles of the
submitted acoustic data varied from partner to partner, simple correlations
 provided a straightforward method to distinguish the impact of
-:hese three parameters on the individual measurements.
To allow for the non-parametric distributions of the compared parameters,
 all presented correlation coefficients are Spearman Correla-‚on
 Coefficients (Spearman, 1904).
[Linear regressions allowed deriving a linear scaling law between the
measured TOLs and wind speeds and current velocities (example in
Ag. 2) for each station. These scaling laws allow for comparison with
existing studies and provide a tool to estimate sound levels in areas
without available measurements.
3.2.1. Wind induced sound component
The strength and direction of association between model-generated
wind velocity and each TOL was measured using the non-parametric,
Spearman correlation test.
As in Hildebrand et al. (2021) wind speeds below 5 m/s were
aeglected in the analysis. Modelled wind speed data were correlated
with sound pressure levels for 17 project monitoring stations (the Norwegian
 station 13-NO-LOV and the Swedish 1b-SE-HON station were
excluded as these stations were, respectively, not located in the area of
interest or had no data available from 2019) to verify local wind-related
ambient sound correlation at specific positions for different frequency
pands. Wind noise at the Norwegian LoVe station has been studied by
Ödegaard et al. (2019).
3.2.2. Shipping induced sound component
To quantify shipping density, a distance-weighted measure of vessels
oaresent is introduced and Spearman correlation coefficients are calculated
 with all TOLs.
Unique vessel presence within a 35 km radius around each station
was determined in time steps of 15 min, including vessels 7 min before
and after each time step. The basic assumption is that the received sound
pressure level (L„;) for a ship is proportional to the logarithmic distance
from a ship (Ainslee et al., 2014).
Lpi = Lsi — (A + Blog10 (2) ) dB, with ro = 1m.
Where Ls;; is the source level of a ship, A is the propagation constant,
B is the transmission loss factor and r; is the distance between the ith ship
and the hydrophone. For each time step (+/—- 7 min), the received sound
pressure level at a hydrophone is thus proportional to the level sum of all
Ep 1.e. the sum of all contributions from individual ships:

N E
m | dB
L = 10080 ( 57 10%)
‘p,sum /

In order to simplify the relationship between received sound pressure
levels and vessels present we assume that source levels are the same for
all ships and that the propagation constant A is also the same for all
stations. The distance weighted metric for shipping CS (contributing
shipping). is then given bv:

N —0.1 Blog1o (*)
CS = Losm — Ls +A = 10logyo (© 10 ” \ dB
which reduces to:
N r; —0.1 8
CS = 10logo (© (2) ) dB
As B is not known around the stations and for all freauenecies. and it

Marine Pollution Bulletin 198 (2024) 115891
:ould be shown that the resulting correlation coefficients are not very
sensitive to changes in B, it was decided to assume B to be 20 (transmission
 loss factor for spherical spreading and unit of propagation loss).
CS is thus a simplified index of the contribution from shipping to
averall received sound levels. It is distance weighted and responsive to
presence of multiple ships at different distances.
This simplified approach was found to be suitable for the intended
analysis of identifying frequency bands correlated with shipping, but not
for estimating the energy received from ships.
Linear scaling laws were found to be too simplistic and would not
reflect the complex relationship between measured TOLs and shipping,
which depends on many different factors.
Only vessels that were in direct line of sight of the hydrophone (i.e.
not shadowed by land) and travelling at speeds above 0.1 knots were
included. A time series of CS along with a spectrogram of a 5-day period
is presented in Fig. 4.

3.2.3. Ocean current induced noise component (flow noise)
To identify frequency bands affected by flow noise, Spearman correlation
 coefficients of TOLs and modelled bottom current velocity were
2alculated.
Due to the highly variable current velocities across the stations no
absolute lower cut-off velocity was defined (as was done in (van Geel
at al., 2020)). Instead, to avoid periods of very low sound pressures,
which can be close to the noise floor of the hydrophone, and to account
for the uncertainty of the hydrodynamic model, only current velocities
above the median current velocity per station were considered for correlations.
 An example of a regular flow noise pattern below 50 Hz,
induced by tidal currents is shown in Fig. 3.
4. Results

4.1. Acoustical characteristics of JOMOPANS measurement stations

4.1.1. Spectrograms
The spectrograms, depicted in Fig. 5, show the temporal course of the
FTOLs at all JOMOPANS stations. Seasonal fluctuations appear to be
negligible compared to the differences between stations, indicating high
spatial variability in contrast to low temporal variability of the sound
field in the North Sea.
Stations close to shipping lanes such as 06-DE-FN1, 07-NL-TEX, 08-3E-WST
 or 16-DK-TN1 show relatively high TOLs over large parts of
‘he frequency range and during most of the time. The TOLs at the 09-UK-DOW
 station were further elevated due to construction noise at the
Triton Knoll wind farm from January to August 2020, which is only 18
km away from the station. Although this posed by far the longest
recorded period of construction noise in our dataset, other stations were
ılso impacted by shorter periods of construction activities or seismic
zurveys (e.g. 05-DE-ES1 and 18-DK-EDA).
Lower TOLs were measured at 16-DK-TN1 following the rerouting of
che major shipping lanes in the Kattegat from the 1st of July 2020. The
data at the neighbouring 17-DK-TN4 does not show a clear change at this
date. The associated effects of the rerouting are analysed in more detail
in (Lalander et al., 2022),
The spectrograms show clearly that the deep Scottish and Norwegian
stations (10-SC-ARB, 11-SC-HEL, 12-SC-MOR, 13-NO-LOV, 14-NO-NTR
and 15-SC-CNS) were measuring relatively low TOLs compared to
shallower, southern JOMOPANS stations (e.g. 06-DE-FN1, 07-NL-TEX,
I8-BE-WST and 09-UK-DOW).
Surprisingly high TOLs were measured at station 05-DE-ES1 in the
:;entral part of the North Sea at the Doggerbank. Relatively low
anthropogenic noise levels were expected at this station, given the
remoteness from shipping lanes. Some stations were also affected
zonstantly by tonal sounds (visible as horizontal lines throughout the
;pectrogram). At 03-DK-HRF and 04-DE-FN3 this can be linked to the
mains hum of adjacent Offshore Wind installations.