1

Original Scientific Article • DOI: 10.2478/rmzmag-2021-0003

Received: July 12, 2021
Accepted: July 15, 2021

Regression methods for evaluation of the underwater
noise levels in the Slovenian Sea
Andreja Popit*
Institute for Water of the Republic of Slovenia
*Corresponding author: E-mail: andreja.popit@izvrs.si

Abstract in English

Abstract in Slovene

Anthropogenic underwater noise pollution of seas and
oceans caused by shipping can have negative effects
on marine animals. The aim of this study was to evaluate quantitatively how much the underwater noise
levels in the Slovenian Sea were influenced by anthropogenic pressures and meteorological parameters in
the period from 2015 until 2018. For this purpose,
correlation method and least squares multiple linear
regression analysis were used. The results of this
study show that the correlation of underwater noise
levels with the dredging activity is significant but low,
while correlation with the ship densities is insignificant, which could be due to reduced sound wave propagation in the shallow sea levels. Correlation of the
underwater noise levels with the wind speed was significant but low to medium, which could be explained
by the breaking waves generated by the wind that
produced sound.

Antropogeno onesnaževanje morij in oceanov
s podvodnim hrupom zaradi ladijskega prometa ima
lahko negativen vpliv na morske živali. Cilj te študije je
bil kvantitativno oceniti, koliko so bile vrednosti ravni
podvodnega hrupa v slovenskem morju odvisne od
antropogenih pritiskov in meteoroloških parametrov
v obdobju od 2015 do 2018. V ta namen sta bili
uporabljeni korelacijska metoda in multipla linearna
regresijska analiza z metodo najmanjših kvadratov.
Rezultati te študije so pokazali, da je bila korelacija
ravni podvodnega hrupa z aktivnostjo poglabljanja
morskega dna signifikantna, toda nizka, medtem ko je
bila korelacija z gostoto ladij zanemarljiva, kar je lahko
posledica zmanjšanega širjenja zvočnega valovanja v
plitvem morju. Korelacija ravni podvodnega hrupa s
hitrostjo vetra je bila signifikantna, vendar nizka do
srednja, kar je mogoče razložiti z zvokom nastalim
zaradi lomljenja valov, ki jih je generiral veter.

Keywords: Underwater noise, correlation, multiple
linear regression, anthropogenic pressures, meteorological parameters

Ključne besede: Podvodni hrup, korelacija, multipla linearna regresija, antropogeni pritiski,
meteorološki parametri

Introduction

To study the correlations of underwater
noise with ship traffic that is representative
for shallow waters worldwide, examples were
chosen in the Celtic and North Seas of the UK,
the Scotian Shelf of Canada, the Guanabara Bay
in Southeast Brazil and in the shallow waters
of the Indian Ocean. An example of the underwater noise of biological origin was chosen in
the lagoon of a coral atoll in the Indian Ocean.
For correlations of underwater noise with the

Anthropogenic noise pollution is present in a
number of world’s the seas and oceans, caused,
in particular, by marine traffic. Continuous
underwater noise from ships disturbs communication of marine mammals, such as dolphins
and whales, which causes problems in their orientation, mating, and feeding that are critical to
their survival [1, 2].

2

ship traffic representative of deep waters examples were chosen in the Northeast Pacific Ocean
in the USA; the Ramsey Sound, Pembrokeshire,
West Wales in the UK; and the Gulf of Mexico
(Table 1).
An analysis of shallow-water ambient noise
levels collected during 14 cruises was reported
for the Scotian Shelf on the eastern Canadian
coast over the period from 1972 to 1985 [3].
The frequency range covered was 30 Hz to 900
Hz. It was found that the average ambient noise
levels (Table 1) were characteristic of shallow
water areas with high shipping densities.
The results of an analysis of temporal fluctuations in the noise power spectrum level of
shallow water ambient noise in the bandwidth
of 100 Hz to 4 kHz were presented in India [4].
The results showed that variation in the average
noise spectrum level was higher in the lower
frequency level (100 Hz to 1 kHz) (Table 1) and
assumed a constant level from there onwards. It
was concluded that the temporal fluctuations of
the noise levels were due mostly to ship traffic
and changes in the weather conditions [4].
Frequency spectra of the underwater ambient noise were measured in the lagoon of a coral atoll, outside the reef and on shallow-water
banks in the tropical zone of the Indian
Ocean [5]. The measurements were performed
in the frequency range of 0.003–9 kHz. In all the
regions studied, continuous underwater noise
(Table 1) of biological origin was observed,
attributed to croaker fish [5].
Continuous measurements of underwater noise levels in the Northeast Pacific Ocean
west of San Nicolas Island, California, USA over
138 days, spanning 2003–2004 were compared with measurements made during the
1960s at the same site (Table 1) [6]. Ambient
noise levels at 30–50 Hz were 10–12 dB higher in 2003–2004 than in 1964–1966 (Table 1),
suggesting an average noise increase rate of
2.5–3 dB per decade. Low frequency (10–50 Hz)
ocean ambient noise levels were closely related
to the shipping vessel traffic. Increases in commercial shipping were believed to account for
the observed low-frequency ambient noise increase. Above 50 Hz the noise level differences
between recording periods gradually diminished to only 1–3 dB at 100–300 Hz [6].
RMZ – M&G | 2021 | Vol. 67[4] | pp. 1–11

Higher underwater noise levels were
measured in the area of Ramsey Sound,
Pembrokeshire, West Wales, UK, in the summer
of 2009 during the season with increased boat
traffic (Table 1) compared with the underwater noise levels measured in the spring of 2009
during the season with decreased boat traffic [7].
Below 2 kHz, the noise was thought to come
from boat traffic. The peak in sound pressure
level around 10 Hz in both seasons was thought
to originate from a variety of shipping-related
sources, including propeller-excited hull resonance, hull pressure, or propeller blades which
could appear from 4 to 70 Hz. These lower frequencies could originate from surface noise
due to waves associated with shipping or from
breaking on the rocks. The survey taken in the
summer, during periods of increased tourist
boat activity, showed a distinct peak around
100 Hz that was not apparent during the low
season survey. The origin of this noise was
likely to be diesel engines, propeller cavitation,
engine harmonics and gearboxes [7].
The first study in Brazil to characterise noise
levels in the coastal zone was carried out in
Guanabara Bay (Southeast Brazil). It showed
underwater noise pollution related to the
ship traffic and small vessel traffic (Table 1)
[8]. Locations with ship traffic had the highest
noise levels, while locations with small vessel
traffic had the lowest noise levels.
Elevated noise conditions were observed
across the Gulf of Mexico, where anthropogenic
marine activities were prominent [9]. The LF
(low frequency) band was selected to include
the environmental, meteorological, biological,
and anthropogenic sounds that occur primarily
between 10 Hz and 500 Hz. At recording sites
positioned nearest to high-density shipping
lanes that lead to the Port of South Louisiana
(H-3) and the Port of Houston (H-1) the highest
L01 values (T = 1 h) were recorded (Table 1) [9].
Nationally coordinated efforts to quantify
underwater noise levels in the UK were presented, in support of UK policy objectives under
the EU Marine Strategy Framework Directive
(MSFD) [10]. Field measurements were made
during 2013 and 2014 at twelve sites around
the UK (Table 1). Noise exposure varied considerably, with little anthropogenic influence
Andreja Popit

1972–1985

2006

2005

1966 – 2006

India, shallow water,
hydrophone depth of
5 m and water depth
of 25 m [4]

Indian Ocean, the
lagoon of a coral
atoll, hydrophone
depth of 2–15 m and
water depth of 50
m [5]

Northeast Pacific
Ocean, SW of the
San Nicolas Island,
California, USA,
hydrophone depth of
1,090–1,106 m [6]

Year of
study

Scotian Shelf, Canada,
hydrophone depth
of 31 m and water
depth of 79 m [3]

Location, depth of
the hydrophone
and depth of the
sea water

Regression methods for evaluation of the underwater noise levels in the Slovenian Sea

Mean ambient noise levels:
-30–50 Hz in 1964–1966:
73–75 dB re 1 Pa2/Hz
-30–50 Hz in 2003–2004:
85 dB re 1 Pa2/Hz [6]

Spectral maxima observed at
1 kHz and 7.5 kHz [5]:
1,000 Hz: 72 dB re µPa/√Hz
2,000 Hz: 65 dB re µPa/√Hz
7,500 Hz: 80 dB re µPa/√Hz

Average noise levels [4]:
98 Hz: 130.3 dB re µPa/√Hz
195 Hz: 128.6 dB re µPa/√Hz
293 Hz: 127.3 dB re µPa/√Hz
586 Hz: 127.6 dB re µPa/√Hz
977 Hz: 127.4 dB re µPa/√Hz
3027 Hz: 127.1 dB re µPa/√Hz
4023 Hz: 127.1 dB re µPa/√Hz

Winter average noise levels at
zero wind speed [3]:
30 Hz: 90.8 dB re µPa2/Hz
45 Hz: 90.5 dB re µPa2/Hz
80 Hz: 87.9 dB re µPa2/Hz
150 Hz: 81.9 dB re µPa2/Hz
300 Hz: 74.1 dB re µPa2/Hz
600 Hz: 64.4 dB re µPa2/Hz
900 Hz: 64.3 dB re µPa2/Hz

Underwater noise levels at
different frequencies

Celtic Sea, northern and
southern North Sea in
UK, hydrophone depth of
100 m and a water depth
of 200 m [10]

Port of South Louisiana
(H-3), Port of Houston
(H-1), NE Gulf of Mexico,
hydrophone depth from
250–1,370 m [9]

Guanabara Bay, SE
Brazil, hydrophone
depth of 2 m and a water
depth of 4 m [8]

Ramsey Sound,
Pembrokeshire, West
Wales, UK, hydrophone
depth of 20 m and a
water depth of 460 m [7]

Location, depth of the
hydrophone and depth
of the sea water

Table 1: Examples of underwater noise levels related mainly to the ship traffic in the global seas and oceans.

2013– 2014

2010–2012

2011–2012

2009

Year of
study

Median noise levels for 1/3 octave
bands from 63 Hz to 500 Hz [10]:
81.5–95.5 dB re 1μPa

The highest values L01 (T = 1 h) at
two sites (H-3 and H-1) nearest to
high-shipping lanes [9] in LF band
(10–500 Hz) were:
130 dB re 1 μPa and 128 dB re 1 μPa,
respectively
(L01 = sound levels that were
exceeded 1% of the time)

The highest mean sound pressure
level [8]):
At 187 Hz:
111.6 ± 9.0 dB re 1 µPa
At 15.89 kHz:
76.2 ± 8.3 dB re 1 µPa

Sound pressure levels [7]:
Between 5 Hz and 10 Hz:
- 92.6 dB re 1 µPa ± 3.9 dB re 1 µPa
during spring
- 104.7 dB re 1 µPa ± 2.2 dB re 1 µPa
during summer
Around 100 Hz:
- 85.9 dB re 1 µPa ± 2.4 dB re 1 µPa
during spring
- 104.6 dB re 1 µPa ± 2.4dB re 1 µPa
during summer

Underwater noise
levels at different
frequencies

3

4

at the Celtic Sea site to several North Sea sites
with persistent vessel noise [10].
The above studies showed that man-made
noise pollution is present in a number of world
seas and oceans caused, in particular, by the
maritime traffic (Table 1).
In January 2019, a group of NGOs specialising in the protection of marine life warned that
most EU member states were probably not going to honour their commitment to reduce marine noise pollution by 2020 [11]. In this regard,
methodological guidance on the underwater
noise mitigation measures were prepared by
ACCOBAMS in 2019 [12]. These are very useful to be taken into consideration in order to
successfully mitigate underwater noise sources
in the worlds’ seas and oceans.
The aim of this study was to prepare a methodology for quantitative determination of the
relationship of the measured underwater noise
levels in the Slovenian Sea with anthropogenic
pressures and meteorological parameters.
The content of the study is important primarily from the point of view of properly recognising the correlation between the pressures
and the status of the marine environment and
of determining the optimal mitigating measures for achieving the objective of the Marine
Directive (2008/56/EC) [13].

Materials and methods

Underwater noise is monitored near the
lighthouse foundation 300 m off the coast at
Debeli rtič, Slovenia (lat.: 45°35’ 28.2’’ N, lon.:
13°41’ 59.1’’ E), beginning in February 2015.
Hydrophone of type Bruel and Kjaer 8,105 was
installed 1 m away from the lighthouse foundation at a depth of 4 m (sea depth is 5 m) and
connected to a sound analyser of type Bruel and
Kjaer 2,250, with a sound level meter and an
octave-based frequency analyser that operates
in the frequency range of 6.3 Hz to 20 kHz [14].
A sound analyser is closed inside the lighthouse.
Measuring unit of the underwater noise levels
is dB re µPa.
Dependent variables were continuous underwater noise levels (in dB) in 1/3 octave
bands with center frequencies of 63 Hz and
125 Hz, Leq,63Hz and Leq,125Hz (dB). Independent
RMZ – M&G | 2021 | Vol. 67[4] | pp. 1–11

variables were ship densities in the four different areas of 2 NM and 5 NM from the measuring station, in the Gulf of Trieste and in the Gulf
of Venice (ρL,2 NM, ρL, 5NM, ρL,Trieste, ρL,Venice), dredging activities, cleaning of the sea floor, wind
speed at vv (m/s) and precipitation at hp (mm).
Wind speed data from the Piran buoy and precipitation data from the meteorological station in the Port of Koper were obtained from
the Environmental Agency of the Republic of
Slovenia. Ship densities were provided from
Automatic Information System data [15].
Measuring periods were the following: from
13 February 2015 to 5 May 2015; from 26
September 2015 to 31 December 2015; from 18
August 2016 to 1 November 2016; from 6 July
2017 to 27 August 2017; and from 18 August
2018 to 31 December 2018, in which measured
underwater noise levels were available.
For the analysis of the correlation between
the dependent variables and the independent
variables, the Pearson correlation coefficient
r was used [16], which is a measure of the linear correlation between the two variables that
gives information about the degree, or magnitude, of the association or correlation, as well
as of the direction (+/-) of the relationship between the variables. It is most suitable for use
if the distribution of variables is normal (or at
least symmetric and unimodal). With r we can
examine whether two variables tend to move
together. If the large values ​​of one variable are
related to the large values ​​of the other variable,
the correlation is positive. If the small values ​​of
one variable are related to the large values ​​of
the other variable, the correlation is negative.
If the values ​​of the two variables are not related,
the correlation is close to or equal to 0.
The results of the analysis of the correlation
between dependent and independent variables are presented in tables, the results being
interpreted and the weights of the influence of
the anthropogenic and meteorological noise
sources on the measured underwater noise
levels estimated.
With this correlation analysis, the interdependence of dependent and independent variables was investigated, while multiple linear
regression analysis was used to study how strongly the values of the dependent variables were
affected by the values ​​of independent variables.
Andreja Popit

5

To predict the dependent variables from independent variables, a linear regression model
was used, including the least squares method,
in which the sum of the squares of deviations
is minimal.
The multiple correlation coefficient
r between the dependent variables and the
independent variables was analysed, and the
results are presented in the next section.
The quality of the regression model was
evaluated based on the coefficient of determination R2, which tells how much of
the variance of the dependent variables is
explained by the variability of the independent
variable – ie explained variance. In the case of
the linear regression, it is equal to the square
of the Pearson correlation coefficient, R2.
However, 1 - R2 is an unexplained variance and
its root is a standard prediction error.
The test of statistical significance of the
correlation was performed based on p-values​​
(significance level). In the regression equation,
those variables that have a p-value of less than
0.05 are significant, which means that there is
less than a 5 % probability that the correlation
is due to an error, which means that there is a
95 % probability that the variables are related
to each other.
If the p-value is less than or equal to 0.05
(p ≤ 0.05), then the results can be generalised
with high certainty (95%) from the sample to
the population, ie that there are differences between the two variables or that the two variables are interrelated. If the p-value is greater
than 0.05 (p > 0.05), then the results cannot be

generalised from the sample to the population,
and we must interpret the results only at the
sample level.
From the regression analysis, we estimated
the weights of the influence of the anthropogenic and meteorological noise sources on the
levels of underwater noise measured.

Results and discussion

The average Leq,63Hz and Leq,125Hz levels measured in the Slovenian Sea in the period
between 2015 and 2018 were 82.8–101.1
dB re 1 μPa and 83.9–98.1 dB re 1 μPa,
respectively. The average ship densities
were 2–252. The average wind speed was
1.8–4.6 m/s and the average precipitation
was 0.02–0.07 mm [14].

Results of the correlation analysis

Interdependence between the dependent
(Leq,63Hz and Leq,125Hz) and independent (ρL,2NM,
ρL,5NM, ρL,Trieste, ρL,Venice, dreadg. act., clean. act.,
vv and hp) variables using the Pearson correlation coefficients is presented in Table 2
(for Leq,63Hz) and Table 3 (for Leq,125Hz).
The highest correlation between the underwater noise and the anthropogenic noise
sources was obtained between the Leq,63Hz and
the dredging activities (r = 0.31) (Table 2).
The highest correlation between the underwater noise (Leq,63Hz and Leq,125Hz) and the ship
density (ρL,2NM, ρL,5NM, ρL,Trieste, ρL,Venice) was 0.18.
Correlation between the underwater noise

Table 2. The Pearson correlation coefficients between the dependent variable (Leq,63Hz) and the independent variables
(ρL,2NM, ρL,5NM, ρL,Trieste, ρL,Venice., dreadg. act., clean. act., vv and hp) in all measuring periods.

Pearson’s
correlation
coefficient

From
13.02.2015 to
05.05.2015

From
26.09.2015 to
31.12.2015

From
18.08.2016 to
01.11.2016

From
06.07.2017 to
07.08.2017

From
18.08.2018 to
31.12.2018

r (vv)

0.58

0.51

0.54

0.39

0.35

-0.06

0.02

0.09

0.08

r (hp)

-0.02

r (ρL,Trieste)

-0.02

r (clean. act.)

n.a.

r (ρL,2NM)
r (ρL,5NM)

r (ρL,Venice)

r (dredg. act.)

0.06

0.13

-0.06

-0.10

-0.05

0.06
n.a.

-0.03

0.31
n.a.

0.08

0.08

0.05

-0.04

-0.05

0.01

0.08
n.a.

-0.10

Regression methods for evaluation of the underwater noise levels in the Slovenian Sea

0.12

n.a.
n.a.

0.07
0.05
0.12

-0.11
n.a.
n.a.

6

Table 3: The Pearson correlation coefficients between the dependent variable (Leq,125Hz) and the independent variables
(ρL,2NM, ρL,5NM, ρL,Trieste, ρL,Venice, dreadg. act., clean. act., vv and hp) in all measuring periods.

Pearsons
correlation
coefficients
r (vv)

From
13.02.2015 to
05.05.2015

From
26.09.2015 to
31.12.2015

From
18.08.2016 to
01.11.2016

From
06.07.2017 to
07.08.2017

From
18.08.2018 to
31.12.2018

0.39

0.18

0.55

0.24

0.15

0.05

0.05

-0.04

0.17

0.02

0.10

n.a.

r (hp)

-0.02

r (ρL,Trieste)

-0.02

r (clean. act.)

n.a.

r (ρL,2NM)
r (ρL,5NM)

r (ρL,Venice)

r (dredg. act.)

0.11

-0.08
n.a.

0.01
0.06
0.02

-0.01
n.a.

(Leq,63Hz and Leq,125Hz) and the cleaning of the sea
floor was -0.10.
Correlation analyses showed that the relationship between the Leq,63Hz and the vv was
between 0.35 and 0.58, while that between
Leq,125Hz and the vv, was between 0.15 and 0.55.
The relation between the underwater noise
(Leq,63Hz and Leq,125Hz) and the meteorological parameter – precipitation, hp – was between -0.02
and 0.09 (Tables 2 and 3). These results of the
correlation analyses are interpreted in the subsection discussion.

Results of the multiple linear regression
analysis

Furthermore, the least squares multiple linear regression analysis in Table 4 shows how
much the values of the dependent variables
were affected by the values ​​of the independent
variables. The multiple correlation coefficients
between the dependent variable Leq,63Hz and the
independent variables in all measuring periods
were moderate (r = 0.40 – 0.59), while those
between the dependent variable Leq,125Hz and the
independent variables were low to moderate
(r = 0.21 – 0.59) (Table 4).
In the regression analysis, 16%–35%
of the variance of the dependent variable Leq,63Hz and 5%–34% of the variance
of the dependent variable Leq,125Hz can be
explained by the variability of the independent variables (ρL,2NM, ρL,5NM, ρL,Trieste,
ρL,Venice, dreadg. act., clean. act., vv and hp),

RMZ – M&G | 2021 | Vol. 67[4] | pp. 1–11

0.09
0.07
0.01

-0.09
-0.10

0.04
0.02
0.18

-0.01
n.a.
n.a.

0.02
0.12
0.04

-0.10
n.a.
n.a.

with which the predictive power of the multiple regression model is shown (Table 5).
The statistical significance test of the
correlation based on the p-value indicated
that there was only one independent variable, wind speed vv, that had a significant
correlation with both dependent variables
(Leq,63Hz and Leq,125Hz) in all the measuring periods (Tables 6 and 7). Dredging and cleaning
activities also had a significant correlation with
Leq,63Hz and Leq,125Hz. Precipitation, hp, was mostly
insignificant. Ship densities (ρL,2NM, ρL,5NM, ρL,Trieste, ρL,Venice) were in some cases significant
and in some insignificant.
Subsequently, the multiple regression analysis was repeated without the insignificant
independent variables in each measuring
period. Regression equations in which the
dependent variables (Leq,63Hz and Leq,125Hz) were
predicted by the significant independent
variables are presented in Tables 8 and 9.

Discussion of the results

The average continuous underwater noise
levels (Leq,63Hz and Leq,125Hz) measured in the
Slovenian Sea [13] were similar to those
reported in the literature, which were related
to the shipping noise (Table 1).
Multiple correlation coefficients between
the dependent variables (Leq,63Hz and Leq,125Hz)
and independent variables (anthropogenic
Andreja Popit

7

Table 4: Multiple correlation coefficients (r) between dependent variables (Leq,63Hz and Leq,125Hz) and independent variables
(ρL,2NM, ρL,5NM, ρL,Trieste, ρL,Venice, dreadg. act., clean. act., vv and hp) in all measuring periods.

Multiple
correlation
coefficients

From
13.02.2015 to
05.05.2015

From
26.09.2015 to
31.12.2015

From
18.08.2016 to
01.11.2016

From
06.07.2017 to
07.08.2017

From
18.08.2018 to
31.12.2018

r (Leq,63Hz)

0.59

0.55

0.57

0.40

0.40

r (Leq,125Hz)

0.41

0.21

0.59

0.29

0.23

Table 5: Coefficients of determination (R2) between the dependent variables (Leq,63Hz and Leq,125Hz) and the independent variables
(ρL,2NM, ρL,5NM, ρL,Trieste, ρL,Venice, dreadg. act., clean. act., vv and hp) in all measuring periods.

Coefficients of
determination
R2 (Leq,63Hz)

R2 (Leq,125Hz)

From
13.02.2015 to
05.05.2015

From
26.09.2015 to
31.12.2015

From
18.08.2016 to
01.11.2016

From
06.07.2017 to
07.08.2017

From
18.08.2018 to
31.12.2018

0.35

0.30

0.33

0.16

0.16

0.17

0.05

0.34

0.09

0.05

Table 6: Significant independent variables with p-values lower than 0.05, meaning that there was a more than 95 % probability
that these variables were related to the dependent variable Leq,63Hz. The grey fields indicate insignificant independent variables,
with p-values greater than 0.05. The abbreviation n.a. means not applicable.

p-values

From
13.02.2015 to
05.05.2015

From
26.09.2015 to
31.12.2015

From
18.08.2016 to
01.11.2016

From
06.07.2017 to
07.08.2017

From
18.08.2018 to
31.12.2018

p (vv)

0.000

0.000

0.000

0.000

0.000

p (ρL,5NM)

0.007

0.015

0.000

0.556

0.000

p (hp)

p (ρL,2NM)

p (ρL,Trieste)
p (ρL,Venice)

p (dredg. act.)
p (clean. act.)

0.001
0.005
0.327
0.037
n.a.
n.a.

0.061
0.691
0.794
0.087
0.000
n.a.

0.240
0.048
0.000
0.023
n.a.

0.002

0.923
0.003
0.119
0.342
n.a.
n.a.

0.010
0.036
0.000
0.000
n.a.
n.a.

Table 7: Significant independent variables with p-values less than 0.05, meaning that there was a more than 95 % probability
that these variables were related to the dependent variable Leq,125Hz. The grey fields indicate insignificant independent variables
with p-values greater than 0.05. The abbreviation n.a. means not applicable.

p-values

From
13.02.2015 to
05.05.2015

From
26.09.2015 to
31.12.2015

From
18.08.2016 to
01.11.2016

From
06.07.2017 to
07.08.2017

From
18.08.2018 to
31.12.2018

p (vv)

0.000

0.000

0.000

0.000

0.000

p (ρL,5NM)

0.000

0.000

0.000

0.215

0.002

p (hp)

p (ρL,2NM)

p (ρL,Tireste)
p (ρL,Venice)

p (dredg. act.)
p (clean. act.)

0.019
0.097
0.001
0.256
n.a.
n.a.

0.509
0.022
0.000
0.110
0.000
n.a.

0.111
0.000
0.000
0.036
n.a.

0.021

Regression methods for evaluation of the underwater noise levels in the Slovenian Sea

0.867
0.033
0.033
0.000
n.a.
n.a.

0.599
0.000
0.000
0.000
n.a.
n.a.

8

Table 8: Regression equations in which the dependent variable Leq,63Hz was predicted by the significant independent variables
(ρL,2NM, ρL,5NM, ρL,Trieste, ρL,Venice, dreadg. act., clean. act., vv and hp).

Measuring periods

Regression equations

13.02.2015–05.05.2015

Leq,63Hz = 1.378*vv - 3.898*hp + 0.357*ρL,2NM + 0.103*ρL,5NM – 0.014*ρL,Venice +
63.450

26.09.2015–31.12.2015
18.08.2016–01.11.2016
06.07.2017–07.08.2017
18.08.2018–31.12.2018

Leq,63Hz = 1.117*vv - 0.093*ρL,5NM + 4.304*dred.act. + 67.800

Leq,63Hz = 2.227*vv + 0.165*ρL,2NM – 0.627*ρL,5NM + 0.528*ρL,Trieste – 0.012*ρL,Venice
– 3.1*clean.a. + 81.77
Leq,63Hz = 2.762*vv - 0.196*ρL,2NM + 70.065

Leq,63Hz = 1.555*vv + 1.116*hp + 0.112*ρL,2NM - 0.216*ρL,5NM + 0.268*ρL,Trieste –
0.022*ρL,Venice + 73.576

Table 9: Regression equations in which the dependent variable Leq,125Hz was predicted by the significant independent variables
(ρL,2NM, ρL,5NM, ρL,Trieste, ρL,Venice, dreadg. act., clean. act., vv and hp).

Measuring periods

Regression equations

13.02.2015–05.05.2015

Leq,125Hz = 0.752*vv - 2.659*hp + 0.263*ρL,5NM – 0.153*ρL,Trieste + 77.786

18.08.2016–01.11.2016

Leq,125Hz = 1.869*vv + 0.265*ρL,2NM – 0.551*ρL,5NM + 0.368*ρL,Trieste –
0.009*ρL,Venice – 1.87* clean.a. + 84.6

26.09.2015–31.12.2015
06.07.2017–07.08.2017
18.08.2018–31.12.2018

Leq,125Hz = 0.248*vv + 0.192*ρL,2NM + 0.167*ρL,5NM – 0.117*ρL,Trieste + 1.181*dred.
+ 77.375
Leq,125Hz = 1.063*vv - 0.122*ρL,2NM + 0.151*ρL,Triestre – 0.017*ρL,Venice + 71.912

Leq,125Hz = 0.562*vv + 0.364*ρL,2NM - 0.126*ρL,5NM + 0.121*ρL,Trieste – 0.017*ρL,Venice
+ 91.705

pressures and meteorological parameters)
were low to moderate (r = 0.21 – 0.59)
(Table 4). In the regression analysis up to
35 % of the variance of the dependent variable
can be explained by the independent variables
(Table 5).
The correlation coefficient between the
measured underwater noise levels (Leq,63Hz)
and dredging activity as an anthropogenic
noise source was significant but low (r = 0.31)
(Table 2). The highest correlation between the
underwater noise levels (Leq,63Hz and Leq,125Hz)
and the ship densities was up to 0.13 and
0.18, respectively (Tables 2 and 3), which
could be explained by reduced sound wave
propagation in the shallow sea [10,17–19].
Low frequency sound waves below the cutoff frequency do not propagate, because the
sound propagates into the sea bed [20, 21].
The correlation between underwater noise
and the cleaning of the sea floor was negligible (Tables 2 and 3), which was expected,
because cleaning was performed with an
excavator from the mainland.
RMZ – M&G | 2021 | Vol. 67[4] | pp. 1–11

The
relation
between
underwater
noise and the meteorological parameter –
precipitation – was insignificant. Correlation
between the Leq,63Hz and the wind speed was low
to medium and correlation between the Leq,125Hz
and the wind speed was negligible to medium
(Tables 2 and 3), which could be explained by
the wind-generated waves that break when
they are large enough and produce sound
[22–28]. That is, underwater ambient sound
measurements were, in some cases, used to
estimate wind speed over the seas and oceans.
The results of these cases showed a very strong
correlation between estimated wind speed,
provided by a passive acoustic recorder algorithm and in situ measurements of the wind
speed [29, 30].
Relevant to our study is the problem of
selection of frequencies in the 1/3 octave
band as indicators of shipping noise. We based
our measurements on the European Marine
Strategy Framework Directive which focuses on
low frequency vessel noise, in 1/3 octave bands
with centre frequencies of 63 Hz and 125 Hz,
Andreja Popit

9

as pressure indicators for ship noise [31].
The two bands were selected based on recordings of ship noise in deep-water areas where,
in general, these bands are most powerful [22].
A Danish study of ship noise in shallow water
has shown that the aforementioned MSFD indicators of shipping noise turned out to be
poor proxies for the impact of noise on small
cetaceans at higher frequencies. Thus, higher
frequencies were proposed to be included in the
assessment of good environmental status, ie in
the 1/3 octave band with a centre frequency of
10 kHz, which was chosen as a compromise between the range of hearing of small cetaceans
and frequency dependent absorption [32].
For this reason, higher frequencies should also
be included in further studies of underwater noise in the northern, shallow part of the
Adriatic Sea.

Conclusions

Given the multiple linear regression analysis
we conclude that dependent variables were
affected by the values of independent variables
to a low to moderate degree. The correlation
of the underwater noise levels with the dredging activity was significant but low, and the
one with the ship densities was insignificant,
which could be explained by reduced sound
wave propagation in the shallow sea. The correlation between underwater noise and the
cleaning of the sea floor was negligible, which
could be explained by the fact that cleaning was
performed with an excavator from the mainland. The relation between the underwater
noise and the meteorological parameter – precipitation was insignificant, while correlation
between underwater noise and the wind speed
was significant but low to medium, which could
be explained by the breaking waves generated
by the wind that produced sound.
In the near future, the methodology presented will be used to evaluate underwater noise
data measured in the years 2019 and 2020. In
addition to MSFD pressure indicators for ship
noise in 1/3 octave bands with center frequencies of 63 Hz and 125 Hz higher frequencies
will be included.

Acknowledgements
This research was funded by the Slovenian
Ministry for the Environment and Spatial
Planning. AIS data for the year 2015 were
obtained during implementation of the
European project BALMAS (2013–2016)
from the Italian Coast Guard Headquarters,
who gave us permission to use the AIS data
for our analyses. AIS data from 2016 to 2018
were obtained by the Slovenian Maritime
Administration in the frame of the Ministry
for Infrastructure.

References

[1] Southall, B.L. et al., ‘Underwater Noise from
Large Commercial Ships —International

Collaboration for Noise Reduction’, Encyclopedia

of Maritime and Offshore Engineering, (LOCATION
OF PUBLISHER?), John Wiley & Sons, 2017
DOI:10.1002/9781118476406.emoe056.

[2] Marley, S.A., Salgado-Kent, C.P., Erbe, C., Parnum, I.M.
‘Effects of Vessel Traffic and Underwater Noise

on the Movement, Behavior and Vocalizations of
Bottlenose Dolphins in an Urbanized Estuary’,

Scientific Reports, vol. 7, no. 13,437, 2017, pp. 1–14,
DOI:10.1038/s41598-017-13252-z.

[3] Zakarauskas, P., Chapman, D.M.F., Staal, P.R.

‘Underwater Acoustic Ambient Noise Levels on the
eastern Canadian continental shelf. Journal of the
Acoustical Society of America, vol. 87, no. 5, 1990,
pp. 2064–2071, DOI:10.1121/1.399333.

[4] Ramji, R., Ramakrishanan, S. ‘Analysis of Short Term
Temporal Fluctuations in Noise Power Spectrum
of Shallow Water Ambient Noise’, Symposium on

Underwater Technology and Workshop on Scientific
Use of Submarine Cables and Related Technologies,
Tokyo, IEEE, 2007, pp. 113–119.

[5] Bardyshev, V.I. (2007): ‘Underwater Ambient Noise in

Shallow Water Areas of the Indian Ocean Within the
Tropical Zone’, Acoustical Physics, vol. 53, no. 2,

pp. 167–171, DOI:10.1134/S1063771007020091.

[6] McDonald, M.A., Hildebrand, J.A., Wiggins, S.M.
‘Increases in Deep Ocean Ambient Noise in
the Northeast Pacific West of San Nicolas

Island, California’, Journal of the Acoustical

Society of America, no. 120, 2006, pp. 711–718,
DOI:10.1121/1.2216565.

Regression methods for evaluation of the underwater noise levels in the Slovenian Sea

10

[7] Willis, M.R. et al., ‘Measuring Underwater Background
Noise in High Tidal Flow Environments’, Renewable
Energy, no. 49, 2013, pp. 255–258, DOI:10.1016/
j.renene.2012.01.020.

[8] Bittencourt, L. et al., ‘Underwater Noise Pollution in
a coastal tropical environment’, Marine Pollution
Bulletin, vol. 83, no. 1, 2014, pp. 331–336,
DOI:10.1016/j.marpolbul.2014.04.026n.

[9] Estabrook, B.J. et al., ‘Widespread Spatial and

Temporal Extent of Anthropogenic Noise across

the Northeastern Gulf of Mexico Shelf Ecosystem’,
Endangered Species Research, no. 30, 2016
pp. 267–282, DOI:10.3354/esr00743.

[10] Merchant, N.D. et al., ) ‘Underwater Noise Levels

in UK Waters’, Scientific Reports, vol. 6, no. 36,942,
2016, (PAGE REF?) DOI:10.1038/srep36942.

[11] European Parliament, ‘Parliamentary Questions’,
https://www.europarl.europa.eu/doceo/document/

E-8-2019-001209_EN.html, accessed 15 June 2021.

[12] ACCOBAMS, ‘Methodological Guide – Guidance on
Underwater Noise Mitigation Measures, Seventh
Meeting of the Parties to ACCOBAMS’, Istanbul,
Republic of Turkey, 2019, p. 31.

[13] Directive of the European Parliament and of the

Council (EC/2008/56), ‘Establishing a Framework
for Community Action in the Field of Marine

Environmental Policy (Marine Strategy Framework
Directive)’, Official Journal of the European Union,
L164, pp. 19–40.

[14] Popit, A., ‘Underwater Noise in the Slovenian Sea’,

Materials and Geoenvironment – RMZ, vol. 67, no. 3,
2020, pp. 1–15, DOI:10.2478/rmzmag-2020-0018.

[15] Perkovič, M., Batista, M., Luin, B., ‘Prikaz in Analiza

Gostot Ladijskega Prometa‘, Fakulteta Za Pomorstvo
In Promet, Lyublyana, Univerza v Ljubljani, 2019,
(PAGE REF?)

[16] Davis, J.C., Statistics and Data Analysis in geology.

John Wiley & Sons, Inc., 1986New York. (PAGE REF?)

[17] Kozaczka, E., Grelowska, G., ‘Propagation of

Ship-Generated Noise in Shallow Sea’, Polish

Maritime Research, vol. 25, no. 2, 2018, pp. 37–46,
DOI:10.2478/pomr-2018-0052.

[18] Meyer, V., Audoly, C.A., ‘Comparison Between

Experiments and Simulation for Shallow Water

Short-Range Acoustic Propagation’, International

Congress on Sound and Vibration, ICSV 24th, London,
United Kingdom, 2017. (PAGE REF?)

[19] Duncan, A.J. et al., ‘Characteristics of Sound

Propagation in Shallow Water over An Elastic

Seabed with a Thin Cap-Rock Layer’, Journal of the

RMZ – M&G | 2021 | Vol. 67[4] | pp. 1–11

Acoustical Society of America, vol. 134, no. 1, 2013,
pp. 207–205, DOI:10.1121/1.4809723.

[20] Urick, R.J., Principles of Underwater Sound,
New York, McGraw-Hill, 1983 (PAGE REF?)

[21] Robinson, S.P., Lepper, P.A. Scoping Study: Review of
Current Knowledge of Underwater Noise Emissions

from Wave and Tidal Stream Energy Devices, London,
Crown Estate, 2013, (PAGE REF?)

[22] National Research Council, Ocean Noise and Marine

Mammals. Washington DC, The National Academies
Press, 2003, DOI:10.17226/10564. (PAGE REF?)

[23] Banner, M.L., Cato, D.H. ‘Physical Mechanisms of

Noise Generation by Breaking Waves – A Laboratory
Study’, in Natural Physical Sources of Underwater
Sound – Sea Surface Sound, Kerman, B.R. (ed.),

Dordrecht, Kluwer Academic, 1993, pp. 429–436,

[24] Carey, W.M., Fitzgerald, J.W., Browning, D.G., ‘Low
Frequency Noise Form Breaking Waves’, In, NUSC

Technical Report 8783, New London, Connecticut,
Naval Underwater System Center 1990,
(PAGE REF?)

[25] Kerman. B.R. ‘Underwater Sound Generation by
Breaking Wind Waves’, Journal of the Acoustical

Society of America, vol. 75, no. 1, 1984, pp. 149–165,
DOI:10.1121/1.390409.

[26] Stewart, M.S., Shallow water ambient noise caused

by breaking waves in the surf zone. Ph.D. diss., Naval
Postgraduate School, Monterey, California, 1994,
(PAGE REF?)

[27] Cuomo, G. et al., ‘Breaking Wave Loads at Vertical
Seawalls and Breakwaters’, Coastal Engineering,
vol. 57, no. 4, 2010, pp. 424–439, DOI:10.1016/j.
coastaleng.2009.11.005.

[28] Ravindar, R., Sriram, V., Schimmels, S., Stagonas,

D. ‘Characterization of Breaking Wave Impact on

Vertical Wall with Recurve’, ISH Journal of Hydraulic
Engineering, vol. 25, no. 2, 2019, pp. 153–161,
DOI:10.1080/09715010.2017.1391132.

[29] Vagle, S., Large, W.G., Farmer, D.M. An Evaluation of the
WOTAN Technique of Inferring Oceanic Winds from

Underwater Ambient Sound’, Journal of Atmospheric
and Oceanic Technology, vol. 7, no. 4, 1990,

pp. 576–595, DOI:10.1175/1520-0426(1990)007<0576:
AEOTWT>2.0.CO;2.

[30] Pensieri, S. et al., ‘Underwater Acoustic Measurements
to Estimate Wind and Rainfall in the Mediterranean
Sea’, Advances in Meteorology, no. 612,512, 2015,
DOI:10.1155/2015/612512, (PAGE REF?)

[31] Commission Decision, ‘Laying Down Criteria
and Methodological Standards on Good

Andreja Popit

11

Environmental Status of Marine Waters and

Specifications and Standardized Methods for
Monitoring and Assessment, and Repealing

Decision 2010/477/EU’, Official Journal of the
European Union, L125, 2017/848/EU,
pp. 43–74.

[32] Hermannsen, L’ et al., ‘High Frequency Components
of Ship Noise in Shallow Water with a Discussion
of Implications for Harbor Porpoises (Phocoena
Phocoena), Journal of the Acoustical Society of

America, vol. 136, no. 4, 2013, pp. 1,640–1,653,
DOI:10.1121/1.4893908.

Regression methods for evaluation of the underwater noise levels in the Slovenian Sea