Spatial Vertical Directionality and Correlation of Low ... - MDPI

sensors Article

Spatial Vertical Directionality and Correlation of Low-Frequency Ambient Noise in Deep Ocean Direct-Arrival Zones Qiulong Yang 1,2 , Kunde Yang 1,2, *, Ran Cao 1,2 and Shunli Duan 1,2 1 2

*

School of Marine Science and Technology, Northwestern Polytechnical University, Xi’an 710072, China; [email protected] (Q.Y.); [email protected] (R.C.); [email protected] (S.D.) Key Laboratory of Ocean Acoustics and Sensing, Northwestern Polytechnical University, Ministry of Industry and Information Technology, Xi’an 710072, China Correspondence: [email protected]; Tel.: +86-132-2801-8268

Received: 4 November 2017; Accepted: 19 January 2018; Published: 23 January 2018

Abstract: Wind-driven and distant shipping noise sources contribute to the total noise field in the deep ocean direct-arrival zones. Wind-driven and distant shipping noise sources may significantly and simultaneously affect the spatial characteristics of the total noise field to some extent. In this work, a ray approach and parabolic equation solution method were jointly utilized to model the low-frequency ambient noise field in a range-dependent deep ocean environment by considering their calculation accuracy and efficiency in near-field wind-driven and far-field distant shipping noise fields. The reanalysis databases of National Center of Environment Prediction (NCEP) and Volunteer Observation System (VOS) were used to model the ambient noise source intensity and distribution. Spatial vertical directionality and correlation were analyzed in three scenarios that correspond to three wind speed conditions. The noise field was dominated by distant shipping noise sources when the wind speed was less than 3 m/s, and then the spatial vertical directionality and vertical correlation of the total noise field were nearly consistent with those of distant shipping noise field. The total noise field was completely dominated by near field wind generated noise sources when the wind speed was greater than 12 m/s at 150 Hz, and then the spatial vertical correlation coefficient and directionality pattern of the total noise field was approximately consistent with that of the wind-driven noise field. The spatial characteristics of the total noise field for wind speeds between 3 m/s and 12 m/s were the weighted results of wind-driven and distant shipping noise fields. Furthermore, the spatial characteristics of low-frequency ambient noise field were compared with the classical Cron/Sherman deep water noise field coherence function. Simulation results with the described modeling method showed good agreement with the experimental measurement results based on the vertical line array deployed near the bottom in deep ocean direct-arrival zones. Keywords: vertical directionality; correlation; low frequency; ambient noise; deep ocean

1. Introduction Ocean ambient noise is a kind of acoustic background, which constantly exists in the ocean and is produced by a number of different types of noise sources, including natural and man-made ones. Numerous previous studies have described the ambient noise spatial characteristics since it is a main parameter in designing sonar equipment and determining the sonar performance. Furthermore, measurements of ambient noise could also be used to infer information about the ocean acoustic environment [1–4]. Several theoretical expressions for wind-generated noise field had been derived in previous studies. Kuperman derived the expressions for the intensity and spatial correlation of the noise

Sensors 2018, 18, 319; doi:10.3390/s18020319

www.mdpi.com/journal/sensors

Sensors 2018, 18, 319

2 of 18

field produced in a stratified ocean by the action of wind at the surface through normal-mode representation [5]. Lee described a ray-based noise model to calculate the two-point spatial coherence function of an ocean surface generated in a hydrophone triplet array with a Maclaurin series [6]. Buckingham developed theoretical models for vertical directionality and spatial coherence based on a uniform distribution of wind-generated ambient noise surface sources in a semi-finite homogenous ocean [7,8]. Furthermore, the coherence function was derived for a sensor pair considering the effects of vertical anisotropy for three ambient noise fields of the azimuthally uniform deep ocean [9]. On the assumption that the overall directionality of the field was the product of the individual directionalities in the horizontal and vertical directions Walker conducted an analysis of the coherence and cross correlation of the noise at two sensors in 3D noise fields [10]. Furthermore, a large number of experiments and noise field models had been conducted to measure and simulate the spatial characteristics of noise field. The spatial coherence functions of low-frequency shipping noise derived on the basis of the Von Mises distribution showed good agreement with the experimental results in the deep ocean [11]. The spatial coherence and cross-spectral density recovered from available hydrophone pairs mounted on a deep-diving platform were consistent with a local wind-driven surface source distribution and matched that predicted by a simple model of deep water, wind-generated noise in the Tonga Trench and Philippine Sea [12,13]. Naughton presented the results where the decrease in amplitude of the noise correlation function with increased separation followed the power law on a set of free-floating oceanic receivers whose relative positions varied with time [14]. The method of frequency-wavenumber diagrams was used to examine the statistic and directional properties of ambient noise in North Pacific [15,16]. The parabolic equation solution method was applied to build the model of the vertical correlation and vertical directionality of ocean ambient noise field under the slope, seamount and varying sound speed profile environments [17]. The various approximations for noise field models based on a simple ray approach were investigated; the ray approach can produce the same accuracy as full wave treatments and can be extended to a range-dependent inhomogeneous field in the presence of non-uniform horizontal distribution of noise sources [18]. However, only one type of noise sources was considered in the theoretical derivation and simulation of spatial properties of noise field in previous researches. In this work, the wind-generated and distant shipping noise fields may significantly and simultaneously affect the spatial characteristics of the total noise field to some extent. The spatial properties of low-frequency ambient noise field in the deep ocean direct-arrival zone were modeled and analyzed by utilizing a ray approach and parabolic equation solution method jointly. The simulation results under several wind speed conditions were consistent with experimental results measured in the South China Sea. The remainder of this paper is organized as follows: Section 2 describes the method for modeling ambient noise and theory of ambient noise spatial coherence function and directionality; Section 3 presents the numerical simulation results in the typical deep ocean for the Munk sound speed profile. Section 4 presents the experimental verification results in the South China Sea. Section 5 summarizes the discussion and conclusions. 2. Model and Spatial Characteristics of Ambient Noise Field 2.1. Methods for Modeling Ambient Noise Field The monopole noise source is assumed to be on a plane beneath the sea surface, where the source-image pairs act as dipoles. The monopole noise sources intensity per unit area is expressed as follows: n2j,l = 10 NSL j,l /10 (1) where, the monopole noise source intensity level NSLj,l depends on the location and is specified in units of decibels referenced to 1 µPa2 /Hz at 1 m per square meter of the surface area.

Sensors 2018, 18, 319

3 of 18

Figure 1 illustrates the schematic of modeling ambient noise field that is induced by wind agitation and distant shipping. All noise sources were scaled and combined with a random phase at a receiver Sensors 2018, 18, x 3 of 18 position to derive the realization of the noise as follows [19,20]: PPnoise (z) = noise ( z ) =

JJ

L

exp((iiψ ψ l ))∑ iψ jj )ns nsj,lj ,l ∑ exp   exp (iψ


ll= =11

p

j= j =11

(


area areaj pj pz, zr ,j ,rzj ,s z, sβ,lβ l

)

(2) (2)


j and ψl are the uniformly distributed random numbers on the interval [0, 2π]; whereψψand where ψl are the uniformly distributed random numbers on the interval [0, 2π]; p z,( r̅,j , z, s , β, l ) j thecomplex complexacoustic acousticpressure pressureatatreceiver receiverposition position z because ̅ becauseofofthe thenear-surface near-surfacesource sourceatatrange range isisthe +j∆r, ∆ ,j= and bearing bearing β l == l∆β, ∆ ,l = = 1, 1, L,, and and can can be be obtained obtained through throughthe theray ray r j ==r0 + = 1, 1, J,, depth depth zs ,, and approach and parabolic equation solution method for near-field wind-driven and far-field distant approach and parabolic equation solution method for near-field wind-driven and far-field distant shippingnoise noisefields fieldsby byconsidering consideringtheir theircalculation calculationaccuracy accuracyand andefficiency, efficiency,respectively; respectively;The Thetime time shipping dependenceexp( exp(−iωt) is neglected; neglected;ω ω == 2πf 2πf is and ff is dependence −iωt) is is the the angular angular frequency; frequency; and is the the frequency frequency in in Hertz. Hertz.

Figure Figure1.1.Schematic Schematicofofambient ambientnoise noisefield fieldmodeling. modeling.

Theproduct productof ofaarealization realization of of the the noise noise and its complex conjugate The conjugate at at two twopositions positions z̅ and z0̅ isis expressedasasfollows: follows: expressed L

J

L

J

* J J L L exp i (hψ l − ψ l ' ) + i (ψ j 
( z ' ) =  − ψ j ' )  i P (∗z ) Pnoise  exp i ψ − ψ + i ψ j − ψ j0 Pnoise (znoise 0 ) Pnoise z0 = ∑ ( ) ∑ ∑ ∑ l l l ' =1 j ' =1 l =1 j =1 l 0 =1 j 0 =1 l =1 j =1 √ j area j ' p ( z , rj , zs β l ) p
* (∗z ', r0 j , zs β l )
××nsnsj , l nsnsj ', l 0' 0 area area j area j0 p z, r j , zs β l p z , r j , zs β l j,l j ,l

(3) (3)

wherethe thesuperscript superscriptasterisk asteriskstands standsfor forcomplex complexconjugation. conjugation.The Theterm termwith withl l==l’l’and andj j==j’j’are arethe the where same in each realization of the product. Assuming that the noise arriving from different cells (l ≠ l’ same in each realization of the product. Assuming that the noise arriving from different cells (l 6= l’ and and j ≠ j’) is uncorrelated, and then the average of realizations of products of the noise j 6= j’) is uncorrelated, and then the average of realizations of products of the noise is approximatedis as follows: asapproximated follows: LL JJ



∗ 0
∗* 0 ∼ PP z ns22j,larea area j pp z,z r, jr, z, sz, β, β , z,sz, β,lβ (4) = noise ((zz) )PP ∑ ∑ ns l p p * z z, r',j r noise z ' ≅ ) (4) noise noise ( j ,l j j s l j s l



ll= =11 jj= =11

(

) (

)

where, where,the theangular angularbrackets bracketsrepresent representan anaverage averageover overmultiple multiplerealizations realizationsofofthe theproduct productand andare are called an “ensemble average”. When the noise is stationary in the statistical sense, the ensemble called an “ensemble average”. When the noise is stationary in the statistical sense, the ensemble average averagecan canbe bereplaced replacedby byaatime timeaverage. average.

The plane wave noise response can be defined briefly in terms of column vector ̅ = ( ), = 1, … , which contains the complex pressure at each receiver. Then the crosscorrelation or covariance matrix of complex pressure on the array can be expressed as: H Rnoise = pnoise pnoise

(5)

sources and distant shipping noise sources are uncorrelated sources (where Rwind and Rship represent the covariance matrices of wind-driven and distant shipping noise fields, respectively). 2.2. Vertical Directionality of Ambient Noise 4 of , 18) = , , , ( = 1,2, … , respectively. Thus, the array manifold vector can be expressed

SensorsThe 2018,coordinate 18, 319 positions of each element and the unit vector are

and = cos cos , cos sin , sin as follows: The plane wave noise response can be defined briefly in terms of column vector T complex pressure at each receiver. Then the pnoise = col [ pnoise (zn ), n = 1, . . . , N ] which the  v1 (contains k ) exp ( − jk q1 )   v k  1( )   on the cross-correlation or covariance matrix of complex pressure array can be expressed as: T  v2 ( k ) exp ( − jk q2 )   v2 ( k )    = v (k ) =   (6) H      Rnoise = pnoise pnoise (5)     v ( k ) T N  vN ( k ) exp ( − jk qN )   where, Rnoise denotes the covariance matrix of noise field on the vertical line array; the superscript H denotes where complexisconjugate a constanttranspose. if all elements are identical; N is the number of array elements; k = −ωu/c T depicts thethe flowchart for modeling the low-frequency noise the field. In addition and Figure τn = k p2n/ω denote wavenumber and relative time delay, ambient respectively; superscript T to the environment parameters, bottom parameters and bathymetry, denotes transposition. The beamincluding scanning sound spatial speed powerprofiles, spectrum as a function of azimuth is defined wind as: speed and ship density were input into underwater sound propagation models based on ray approach and parabolic equation solution method Hfor modeling noise field. The covariance matrices P (θ , φ ) = v ( k ) Rnoise v ( k ) (7) of the total noise field can be written asbeam Rtotal = Rwind + Rship on the assumption that the wind noise sources andsuperscript distant shipping noise sources are transposition; uncorrelated sources (where Rwind anddata Rshipcovariance represent noise where the H denotes conjugate Rnoise represents the covariance matrices of wind-driven and distant shipping noise fields, respectively). matrix; θ and φ are the polar angle and azimuthal angle, respectively.

Figure 2. 2. Flowchart Flowchart of of modeling modeling the the low-frequency low-frequency ambient ambient noise noise field. field. Figure

2.3. Vertical Coherence of Ambient Noise Field 2.2. Vertical Directionality of Ambient Noise Two sensors were considered at positions z1 and z2 in the plane wave noise field, where the The coordinate positions of each element and the unit vector are qn = [ xn , yn , zn ], (n = 1, 2, . . . , N ) fluctuations were represented by theTtime series x1(t) and x2(t), respectively. The Fourier transforms and u = [cos φ cos θ, cos φ sin θ, sin φ] , respectively. Thus, the array manifold vector can be expressed of the time series were assumed as X1(ω) and X2(ω). Then, the bilateral power spectral density of the as follows:  
  noise at the two sensors could be written [9,13]: ve1 (kas − jk T q1 v1 ( k ) ) exp
   e  2  v2 (k) exp − jk T q2   v2 ( k )  X ω    ( ) j v(k) =  = (6) .. ) = .. (8)     S jj (ω . .     T
veN (k) exp − jk T q N v N (k) where {ven }nN=1 is a constant if all elements are identical; N is the number of array elements; k = −ωu/c and τ n = kT pn /ω denote the wavenumber and relative time delay, respectively; the superscript T denotes transposition. The beam scanning spatial power spectrum as a function of azimuth is defined as: Pbeam (θ, φ) = v H (k) Rnoise v(k) (7) where the superscript H denotes conjugate transposition; Rnoise represents noise data covariance matrix; θ and ϕ are the polar angle and azimuthal angle, respectively. 2.3. Vertical Coherence of Ambient Noise Field Two sensors were considered at positions z1 and z2 in the plane wave noise field, where the fluctuations were represented by the time series x1 (t) and x2 (t), respectively. The Fourier transforms of

Sensors 2018, 18, 319

5 of 18

the time series were assumed as X1 (ω) and X2 (ω). Then, the bilateral power spectral density of the noise at the two sensors could be written as [9,13]: S jj (ω ) =

X j ( ω ) 2

(8)

T

where, T is the observation time used to create the Fourier transforms; the overbar denotes an ensemble average. Similarly, the cross-spectral density of the noise fluctuations is: S12 (ω ) =

X1 (ω ) X2∗ (ω ) T

(9)

where the asterisk (*) denotes complex conjugation. The coherence function Γ12 (ω) is defined as the cross-spectral density that is normalized to the geometric mean of the power spectral densities at the two sensors: S12 (ω ) (10) Γ12 (ω ) = p S11 (ω )S22 (ω ) In general, the directional density function represents the noise power incident at the receiver in the ocean as a function of arrival angle. A convenient normalization of the directional density function, which is obtained from the total noise power integrated over all angular spaces in spherical-polar coordinates, can be written as Z Z 2π

0

π

0

F (θ, φ) sin θdθdφ = 4π

(11)

where, F(θ,ϕ) is the directional density function and can be obtained from Equation (7). Cox derived an elegant expression which leads to an expression for the coherence function in term of directional density function, valid for arbitrary orientation of the two sensors in the plane wave noise field [21]. The expressions for coherence function are expressed as follows [9]:   Z Z d d 1 2π π F (θ, φ)e−i2π λ cos θ sin θdθdφ Γ = λ 2 0 0     Z Z d 1 2π π d Γ = F (θ, φ) J0 2π sin θ sin θdθdφ λ 2 0 λ 0

(vertical)

(12)

(horizontal)

(13)

where J0 (·) is the Bessel function of the first kind of order zero; d and λ denote the distance separating sensors and the wavelength. Cron and Sherman proposed a model of deep ocean ambient noise in which independent point sources were distributed uniformly in a horizontal plane beneath the sea surface by considering the ocean itself to be a semi-infinite, homogeneous half space. Cron and Sherman assumed straight line propagation in infinite deep water (ideal waveguide) and derived correlation coefficients for vertically and horizontally separated hydrophones The noise sources radiate sound with vertical directional density function F(θ,φ) = cosm θ (where m is a positive integer, usually taken as 1 or 2, and m = 1 corresponds to dipole sources). Then the coherence functions of the noise at the two aligned sensors are [1,9,10,13,21,22]:     R π/2 d −i2πd/λ −i2πd/λ −1 Γ λd = 2 0 e−i2π λ cos θ cos θ sin θdθ = 2 −ie2πd/λ + e (2πd/λ)2     (14) cos(2πd/λ) 1 sin(2πd/λ) (2πd/λ) + 2i ( vertical ) = 2 sin2πd/λ + cos(2πd/λ)− − 2 2 2πd/λ (2πd/λ)

Γ

(2πd/λ)

  Z π/2 d 2 =2 J0 (2πd/λ sin θ ) cos θ sin θdθ = J (2πd/λ) λ 2πd/λ 1 0

where, J1 (·) is the Bessel function of the first kind of order unity.

(horizontal)

(15)

Sensors 2018, 18, 319

6 of 18

Sensors 2018, 18, x

3. Numerical Results of the Simulation

6 of 18

Simulations performed for the Munk sound speed profile as presented in Figure 3. The 3.1. Wind Speedwere Dependence center depth of the receiver array and surface conjugate depth were assumed to be 3800 m and 4000 Simulations were performed for the Munk sound speed profile as presented in Figure 3. The center m, respectively. A two-layer bottom model was used to model low frequency ambient noise field. depth of the receiver array and surface conjugate depth were assumed to be 3800 m and 4000 m,3 The sediment sound speed, density and attenuation were assumed to be 1550 m/s, 1.30 g/cm , and respectively. A two-layer bottom model was used to model low frequency ambient noise field. 0.15The dB/λ, correspondingly. The basement sound speed, density, and sediment sound speed, density and attenuation were assumed to attenuation be 1550 m/s,were 1.30 assumed g/cm3 , to 3, and 0.10 dB/λ, respectively. The thickness of the sediment was approximately be 1650 m/s, 1.80 g/cm and 0.15 dB/λ, correspondingly. The basement sound speed, density, and attenuation were assumed to 3 , and 12 m. agitation and distant shipping were assumed to beofthe dominant noise sources in the be Wind 1650 m/s, 1.80 g/cm 0.10 dB/λ, respectively. The thickness the sediment was approximately noise The other and noise sources werewere disregarded work. The noise source 12 field. m. Wind agitation distant shipping assumed toin be this the dominant noise sources in thedepth noise was field. The other noise sources were disregarded in this work. Thebeneath noise source was assumed to be assumed to be on the plane of one-quarter of the wavelength the depth sea surface for wind-driven on[19,20,23], the plane ofwhereas one-quarter the wavelength beneath theassumed sea surfacetofor noisethe [19,20,23], noise theofnoise source depth was bewind-driven 10 m beneath surface for whereas the noise source depth was assumed to be 10 m beneath the surface for distant shipping noise. distant shipping noise. The ray approach and parabolic equation solution method could be utilized The ray approach and parabolic equation solution method could be utilized to model the wind-driven to model the wind-driven and distant shipping ambient noise fields correspondingly based on their and distant shipping ambient noise fields correspondingly based on their calculation accuracy and calculation accuracy and efficiency. Therefore, the ray approach and parabolic equation solution efficiency. Therefore, the ray approach and parabolic equation solution method were jointly utilized to method were jointly utilized to model the low-frequency ambient noise field on the basis of the model the low-frequency ambient noise field on the basis of the BELLHOP [24] and RAM [25] acoustic BELLHOP [24] andsource RAMintensities [25] acoustic Theand noise source intensities ofacquired wind agitation codes. The noise of windcodes. agitation distant merchant ship were through and distant merchant ship were acquired through Research Ambient Noise Directionality (RANDI) Research Ambient Noise Directionality (RANDI) and Ambient Noise Directionality Estimation System and Ambient Noise Directionality Estimation System [19,23,26]. the sake simplicity, the (ANDES) [19,23,26]. For the sake of simplicity, the (ANDES) distant shipping noise For intensity was of uncorrelated distant noise intensity was uncorrelated with wind speed condition in this Section. withshipping wind speed condition in this Section.

Figure soundspeed speedprofile. profile. Figure 3. 3. Munk Munk sound

Figures 4–64–6 present verticaldirectionality directionality and correlation coefficient of total Figures presentthe theambient ambient noise noise vertical and correlation coefficient of total noise field for wind speeds of 3, 7 and 12 m/s at 50, 100 and 150 Hz, respectively. Positive and negative noise field for wind speeds of 3, 7 and 12 m/s at 50, 100 and 150 Hz, respectively. Positive and negative angles represent thethe downward and upward acoustic angles represent downward and upward acousticrays raysarriving arrivingatatthe thearray, array,correspondingly. correspondingly. The distant noise reaching the vertical lineline array (VLA) from anglewas was the The shipping distant shipping noise reaching the vertical array (VLA) fromshallow shallow grazing grazing angle the dominant the noise source when the wind speed was 3 m/s. Dual peaks structure, which are dominant the noise source when the wind speed 3 m/s. Dual peaks structure, which are approximately symmetricalwith withrespect respect to to the cancan be observed in the approximately symmetrical the horizontal horizontaldirection, direction, be observed invertical the vertical directionality pattern of distant shipping noise field. The directionality results were induced by upward directionality pattern of distant shipping noise field. The directionality results were induced by and downward acousticacoustic rays from the from deep sound channel and channel along theand deepalong soundthe channel upward and downward rays the deep sound deep orsound convergence zone path [27], as acoustic rays could completely reverse at the surface conjugate depth. channel or convergence zone path [27], as acoustic rays could completely reverse at the surface Currently, vertical correlation and directionality result of the total noise field were consistent with conjugate depth. Currently, vertical correlation and directionality result of the total noise field were those of the distant shipping noise field. The first zero radius of the shipping noise field also showed consistent with those of the distant shipping noise field. The first zero radius of the shipping noise good agreement with that of the total noise field. The wind-driven noise that reached the array from fieldrelatively also showed good agreement with that of the total field. The noise that steep grazing angles significantly contributed to the noise total ambient noisewind-driven field as wind speed reached the array from relatively steep grazing angles significantly contributed to the total ambient increased. The wind-generated noise sources induced by wind agitation and breaking waves could noise fieldatasthewind speed increased. wind-generated noise sources by the wind agitation arrive receiver array along theThe direct or reliable acoustic path [28–30].induced Currently, vertical coefficient of the total at noise became inconsistent with that or of the shipping noise path field [28– andcorrelation breaking waves could arrive thefield receiver array along the direct reliable acoustic 30]. Currently, the vertical correlation coefficient of the total noise field became inconsistent with that of the shipping noise field but tended toward that of the wind-driven noise field as frequency increased. The first zero radius in the vertical correlation coefficient demonstrated a complex relationship with frequency. The vertical directionality and correlation coefficient of the total noise

Sensors 2018, 18, 319

7 of 18

but tended toward that of the wind-driven noise field as frequency increased. The first zero radius in the vertical correlation coefficient demonstrated a complex relationship with frequency. The vertical directionality and correlation coefficient of the total noise field were completely consistent with that of the wind noise field at 150 Hz for the wind speed of 12 m/s as illustrated in Figure 6c,f. Currently, the first zero Sensors 2018, 18, xlocation occurs at nearly half wavelength approximately. What’s more, the vertical 7 of 18 Sensors 2018, 18, x 7 of 18 directionality and correlation coefficient were the weighed results of wind-driven and distant shipping noise field for wind speeds 7 m/s. The weighted results have relationship with the proportion have close relationship withofthe proportion of wind-driven andclose distant shipping noise sources in the have close relationship with the proportion of wind-driven and distant shipping noise sources in the of wind-driven total noise field.and distant shipping noise sources in the total noise field. total noise field.

(a) (a)

(b) (b)

(c) (c)

(d) (d)

(e) (e)

(f) (f)

Figure 4. 4. Vertical correlation correlation and directionality directionality at wind speed speed of 3 m/s for frequencies of (a,d) 50 Hz Figure Figure 4.Vertical Vertical correlationand and directionalityatatwind wind speedofof3 3m/s m/sfor forfrequencies frequenciesof of(a,d) (a,d)50 50Hz Hz (b,e) 100 Hz and (c,f) 150 Hz. (b,e) 100 Hz and (c,f) 150 Hz. (b,e) 100 Hz and (c,f) 150 Hz.

(a) (a)

(b) (b)

(c) (c)

(d) (d)

(e) (e)

(f) (f)

Figure 5. Vertical correlation and directionality at wind speed of 7 m/s for frequencies of (a,d) 50 Hz Figure5.5.Vertical Verticalcorrelation correlationand anddirectionality directionalityatatwind windspeed speedofof7 7m/s m/sfor forfrequencies frequenciesof of(a,d) (a,d)50 50Hz Hz Figure (b,e) 100 Hz and (c,f) 150 Hz. (b,e) 100 Hz and (c,f) 150 Hz. (b,e) 100 Hz and (c,f) 150 Hz.

(d)

(e)

(f)

Figure 5. Vertical correlation and directionality at wind speed of 7 m/s for frequencies of (a,d) 50 Hz (b,e)18, 100319 Hz and (c,f) 150 Hz. Sensors 2018,

Sensors 2018, 18, x

8 of 18

(a)

(b)

(c)

(d)

(e)

(f)

8 of 18

Figure6.6.Vertical Verticalcorrelation correlationand anddirectionality directionalityatatwind wind speed m/sfor forfrequencies frequenciesofof(a,d) (a,d)5050Hz Hz Figure speed ofof 1212 m/s (b,e)100 100Hz Hzand and(c,f) (c,f)150 150Hz. Hz. (b,e)

The spatial correlation coefficients in the numerical calculation results of the total noise field The spatial correlation coefficients in the numerical calculation results of the total noise field were were compared with spatial correlation results described by Cron/Sherman (C/S). The spatial compared with spatial correlation results described by Cron/Sherman (C/S). The spatial correlation correlation coefficients of the wind-driven noise field were approximately consistent with the C/S coefficients of the wind-driven noise field were approximately consistent with the C/S results based on results based on surface noise source distribution. The first zero location occurred at a half surface noise source distribution. The first zero location occurred at a half wavelength for wind-driven wavelength for wind-driven noise field and C/S coherence function. However, the spatial correlation noise field and C/S coherence function. However, the spatial correlation coefficients of the wind coefficients of the wind noise field deviated a little from the C/S results because of the presentation noise field deviated a little from the C/S results because of the presentation of bottom reflection and of bottom reflection and sound speed gradient. Furthermore, at low wind speed condition the first sound speed gradient. Furthermore, at low wind speed condition the first zeros may have occurred zeros may have occurred at many times a half wavelength because of the presentence of distant at many times a half wavelength because of the presentence of distant shipping noise field. Finally, shipping noise field. Finally, the wind-driven and distant shipping noise fields significantly and the wind-driven and distant shipping noise fields significantly and simultaneously influenced the simultaneously influenced the vertical correlation coefficient and directionality pattern of the total vertical correlation coefficient and directionality pattern of the total noise field. The first zero location noise field. The first zero location was determined by the competition of wind-driven and distant was determined by the competition of wind-driven and distant shipping noises in the total noise field. shipping noises in the total noise field. 3.2. Frequency Dependence 3.2. Frequency Dependence Figure 7a–c depicts the frequency-wavenumber spectra of the total noise field for wind speeds the frequency-wavenumber the utilized total noise field for wind speeds of 3, 7Figure and 127a–c m/s.depicts Frequency-wavenumber diagramsspectra [15,16]of were to examine the vertical of 3, 7 and 12 Frequency-wavenumber diagramsnoise [15,16] were reaching utilized to theshallow vertical directionality ofm/s. the ambient noise field. The dominant sources theexamine VLA with directionality of the ambient noise field. The dominant noise sources reaching the VLA with shallow grazing angle close to horizontal direction are originated by distant ships when the wind speed was angle close tonoise horizontal arefrom originated distant ships when the wind speed was 3grazing m/s. The dominant powerdirection originated noise by sources with relatively minimal absolute 3 m/s. wavenumber The dominantalong noisethe power from path. noiseThe sources with relatively minimal absolute vertical deeporiginated sound channel near-field wind-driven noise sources vertical wavenumber along the deep sound channel path. The near-field wind-driven noise may become the dominant sources when the wind speed was 12 m/s. The wind noise arrivedsources at the may become the dominant sources when the wind speed was 12 m/s. The wind noise arrived at the array through downward acoustic rays with relatively greater positive vertical wavenumber and steep array through downward acoustic withacoustic relatively greater positive vertical and grazing angles along the direct path orrays reliable path. The wind-driven noisewavenumber field under high steep grazing angles along the direct path or reliable acoustic path. The wind-driven noise field under wind speed condition significantly influenced the spatial vertical directionality pattern. What’s more, high wind speed condition significantly influenced the spatial vertical directionality pattern. What’s several diagonal lines (black dash line) were superimposed to aid the conversion of stripes with vertical more, several diagonal lines (black dash line) were superimposed to aid the conversion of stripes wavenumber into grazing angles (±90◦ ,±60◦ ,±30◦ ,0◦ ). The general formula is given by [15]: with vertical wavenumber into grazing angles (±90°,±60°,±30°,0°). The general formula is given by   [15]: −1 d f /dk z (16) Θ = 90 − cos c  array −1 df dk z Θ=90-cos  (16)  carray   

where lines with a positive slope correspond to noise traveling downward the array; ±90° denotes the vertical endfire directions; f, kz and carray represent the frequency, vertical wavenumber and sound speed at the array, respectively. The location of the first zeros is the critical parameter in determining the spacing separating sensors and the signal-to-noise gain of an array, and must be determined in the total noise field for

about a half wavelength under high wind speed conditions for frequencies higher than 150 Hz. But for frequencies lower than 150 Hz, the first zero radiuses is many times a half wavelength and shows different characteristics with that in isotropic noise field where the first zero occurs at d = λ/2. Furthermore, as wind speed increase, the first zero radius of correlation coefficient becomes 9short Sensors 2018, 18, 319 of 18 and tends to a half wavelength. The first frequency at which the correlation coefficient of total noise field is equal to zeros becomes lower. At high wind speed condition the vertical correlation coefficient where lines positive slope to noiseinfluenced traveling downward the array; ±90◦ The denotes structure of with windagenerated noisecorrespond field significantly that of the total noise field. first the vertical endfire directions; f, k and c represent the frequency, vertical wavenumber and sound zero location at the wind speed of differs from that at the wind speed of 12 m/s for frequencies z 3 m/sarray speed at the from 100 Hzarray, to 160respectively. Hz as shown in Figure 7d,f.

(a)

(b)

(c)

(d)

(e)

(f)

Figure 7. 7. Frequency-wavenumber Frequency-wavenumberspectra spectraand andvertical vertical correlation coefficient of the total noise Figure correlation coefficient of the total noise fieldfield for for wind speeds of (a,d) 3 m/s and (b,e) 7 m/s (c,f) 12 m/s. wind speeds of (a,d) 3 m/s and (b,e) 7 m/s (c,f) 12 m/s.

4. The Experiment Results The location of the first zeros is the critical parameter in determining the spacing separating sensors and theDescription signal-to-noise gain of an array, and must be determined in the total noise field for 4.1. Experiment surface noise source distribution. Figure 7d–f displays the vertical correlation coefficient of the total experiment conducted in a7deep area of the on 5–21 August 2016. noiseAn field for wind was speeds of 3 m/s, m/socean and 12 m/s. TheSouth whiteChina linesSea in the figures represent Figure 8a demonstrates the measured sound speed profile with Conductivity-Temperature-Depth the contour of the vertical correlation coefficient, which is equal to zero. The clear light and shade (CTD)in inthe thevertical experiment. The water depth and mixing layers were approximately and 40 by m, stripe correlation coefficient diagram of frequency-depth may have3910 beenm induced correspondingly. The acoustic an zero incomplete to theshort lack and of surface near-field wind-generated noise waveguide sources. Theisfirst location channel obviouslydue becomes occurs conjugate depth, because the sound speed at the bottom is less than that on the sea surface. A 150 vertical at about a half wavelength under high wind speed conditions for frequencies higher than Hz. line for array, which consists 16 150 elements, was deployed near aismain ship laneathat Malacca But frequencies lower of than Hz, the first zero radiuses many times halfconnects wavelength and and thedifferent Bashi orcharacteristics Taiwan Strait.with Figure 8bin depicts the noise real-time distribution from the shows that isotropic fieldshipping where the first zero occurs at dsatellite = λ/2. Automatic Identification System (AIS) and receiver position of experiment. The green dots Furthermore, as wind speed increase, the first zero radius of correlation coefficient becomesdenote short the real time position of the ships. A large number of ships traversed the VLA during the experiment and tends to a half wavelength. The first frequency at which the correlation coefficient of total noise measurement, then the contribution of the shipping field the within 50 km should becoefficient regarded field is equal toand zeros becomes lower. At high wind speednoise condition vertical correlation as interference and should be eliminated in modeling the low-frequency noise field. The element structure of wind generated noise field significantly influenced that of the total noise field. The first separation was approximately 4 m, and the VLA center depth was approximately 3718 m based on zero location at the wind speed of 3 m/s differs from that at the wind speed of 12 m/s for frequencies depth sensors that are fixed on the VLA. The ambient noise data during the measurements were from 100 Hz to 160 Hz as shown in Figure 7d,f. recorded continuously. The distant shipping and wind-driven noise source depths were assumed to 4. Experiment Results beThe 10 m and a quarter of the wavelength beneath the sea surface in modeling the low-frequency ambient noise field, respectively [19,20,23]. 4.1. Experiment Description An experiment was conducted in a deep ocean area of the South China Sea on 5–21 August 2016. Figure 8a demonstrates the measured sound speed profile with Conductivity-Temperature-Depth (CTD) in the experiment. The water depth and mixing layers were approximately 3910 m and 40 m, correspondingly. The acoustic waveguide is an incomplete channel due to the lack of surface conjugate depth, because the sound speed at the bottom is less than that on the sea surface. A vertical line array, which consists of 16 elements, was deployed near a main ship lane that connects Malacca and

Sensors 2018, 18, 319

10 of 18

the Bashi or Taiwan Strait. Figure 8b depicts the real-time shipping distribution from the satellite Automatic Identification System (AIS) and receiver position of experiment. The green dots denote the real time position of the ships. A large number of ships traversed the VLA during the experiment measurement, and then the contribution of the shipping noise field within 50 km should be regarded as interference and should be eliminated in modeling the low-frequency noise field. The element separation was approximately 4 m, and the VLA center depth was approximately 3718 m based on depth sensors that are fixed on the VLA. The ambient noise data during the measurements were recorded continuously. The distant shipping and wind-driven noise source depths were assumed to be 10 m and a quarter of the wavelength beneath the sea surface in modeling the low-frequency ambient Sensors 2018, 18, xrespectively [19,20,23]. 10 of 18 noise field,

(a)

(b)

(c)

Figure 8. Experimental environment andand signal waveform of of explosive sources. Measured Figure 8. Experimental environment signal waveform explosive sources.(a)(a) Measuredsound speed profile; (b) real-time shipping distribution from the satellite AIS system in experiment; sound speed profile; (b) real-time shipping distribution from the satellite AIS system in experiment; (c) (c) received waveform of brand explosive sources. received signalsignal waveform of brand explosive sources.

Furthermore, a sound wasconducted conducted during ambient noise recording. Furthermore, a soundpropagation propagation experiment experiment was during ambient noise recording. Broadband explosive sources were dropped from the research vessel for transmission losses Broadband explosive sources were dropped from the research vessel for transmission losses measurement and geoacousticinversion. inversion. Figure the received signal waveform of explosive measurement and geoacoustic Figure8c8cdepicts depicts the received signal waveform of explosive charges for a source depth of 50 m in the experiment. The signal was received at range of 6.5 km charges for a source depth of 50 m in the experiment. The signal was received at range of 6.5and km and at receiver depth of 3715 m. The objective function, which is a function of the bottom reflection losses at receiver depth of 3715 m. The objective function, which is a function of the bottom reflection losses (BL) extracted from experimental data and calculated by underwater acoustic model, was utilized for (BL) extracted from experimental data and calculated by underwater acoustic model, was utilized for geoacoustic parameters inversion. A two-layer bottom model was obtained from geoacoustic inversion geoacoustic parameters inversion. A two-layer bottom model was obtained from geoacoustic and optimization process [31,32]. And then the two-layer bottom model was used to model the inversion optimization [31,32]. And then the two-layer model wasparameters used to model ambientand noise field. Table 1process summarizes the values of the sediment andbottom basement acoustic the in ambient noise field. Table 1 summarizes valuesand of basement the sediment and basementwere acoustic modeling the ambient noise field. The bottomthe sediment acoustic parameters parameters in modeling the ambient noise field. The bottom sediment and basement acoustic inverted from experimental sound propagation data with explosive sources. Figure 9 illustrates the parameters werelosses inverted from experimental sound propagation datathe with explosive sources. Figure 9 transmission comparison between simulations results using bottom parameters listed in Table the 1 and the experimental measurement results. The experimental were measured illustrates transmission losses comparison between simulationsresults results using the at bottom a receiver listed depth of m. The simulation results were calculated using the inverted parameters parameters in3715 Table 1 and the experimental measurement results. Thebottom experimental results based on the RAM acoustic code and were consistent with experimental measurements for source were measured at a receiver depth of 3715 m. The simulation results were calculated using the depthsbottom of 50 mparameters and 300 m atbased a frequency 100 Hz. Finally,code the bottom acoustic parameters were valid inverted on theofRAM acoustic and were consistent with experimental and could be used to model the low-frequency ambient noise field. measurements for source depths of 50 m and 300 m at a frequency of 100 Hz. Finally, the bottom acoustic parameters were valid could be to model low-frequency ambient noise field. Tableand 1. Sediment andused basement acousticthe parameters. Bottom

Sound Speed (m/s)

Density (g/cm3 )

Attenuation (dB/λ)

Thickness (m)

Sediment basement

1535–1545 1560–1650

1.32–1.35 1.80

0.15 0.10

0–15 15–200

(a)

(b)

illustrates the transmission losses comparison between simulations results using the bottom parameters listed in Table 1 and the experimental measurement results. The experimental results were measured at a receiver depth of 3715 m. The simulation results were calculated using the inverted bottom parameters based on the RAM acoustic code and were consistent with experimental measurements for source depths of 50 m and 300 m at a frequency of 100 Hz. Finally, the bottom Sensors 2018, 18, 319 11 of 18 acoustic parameters were valid and could be used to model the low-frequency ambient noise field.

(a)

(b)

Figure 9. 9. Transmission Transmission losses losses comparison comparison between between simulation simulation and and experiment experiment at 100 Hz Hz for for source source Figure at 100 depths of (a) 50 m and (b) 300 m. depths of (a) 50 m and (b) 300 m. Table 1. Sediment and basement acoustic parameters.

4.2. Auxiliary Database for the Modeling Noise Field Bottom

SensorsFigure 2018, 18,10 x Sediment Sensors 2018, 18, x

Sound Speed (m/s)

Density (g/cm3)

Attenuation (dB/λ)

Thickness (m)

displays the sea surface wind speed as a function of time above the VLA. The11wind of 18 1535–1545 1.32–1.35 0.15 0–15 11 of 18 speed data were acquired from the National Center of Environment Prediction (NCEP) reanalysis basement 1560–1650 1.80 0.10 15–200 database [33]. The The wind-driven wind-driven ambient ambient noise noise may may have have significantly significantly contributed to the total total ambient ambient database [33]. The wind-driven ambient noise may have significantly contributed to the total ambient noise field on 14–20 August because of the high wind speed and sea state. The numbers in brackets 4.2. Auxiliary for the because ModelingofNoise Field wind speed and sea state. The numbers in brackets noise field onDatabase 14–20 August the high represent three wind speed conditions under which experimental measurement measurement data were selected represent three wind speed conditions under whichas experimental measurement data wereThe selected Figure 10 displays the sea surface wind speed a function of time above the VLA. wind to demonstrate the effectiveness of the modeling method. The wind speed conditions at demonstrate The wind speed conditions at the the three to demonstrate the effectiveness of the modeling method. The wind speed conditions at the three speed data were acquired from the National Center of Environment Prediction (NCEP) reanalysis moments corresponded correspondedtotothose those presented in Figure 11, which thefield wind field moments at three presented in Figure 11, which depictsdepicts the wind at three moments corresponded to those presented in Figure 11, which depicts the wind field at three moments in the experiment; blackdenotes trianglethe denotes theofposition of the VLA. According to the in the experiment; the black the triangle position the VLA. According to the historical moments in the experiment; the black triangle denotes the position of the VLA. According to the historical typhoon information the China Meteorology Administration (CMA), the high typhoon information of the ChinaofMeteorology Administration (CMA), the high wind speed maywind have historical typhoon information of the China Meteorology Administration (CMA), the high wind speed may have induced by typhoon wasinathe local cyclone inSouth the north of Sea the been induced by been typhoon “Dianmu”, which“Dianmu”, was a localwhich cyclone north of the China speed may have been induced by typhoon “Dianmu”, which was a local cyclone in the north of the South China Sea and appeared by accident. Then, the wind field data from NCEP database was used and appeared by accident. Then, the wind field data from NCEP database was used for modeling South China Sea and appeared by accident. Then, the wind field data from NCEP database was used for modeling noise wind-driven noise source intensity and distribution. wind-driven source intensity and distribution. for modeling wind-driven noise source intensity and distribution.

Figure function Figure 10. 10. The The wind wind speed speed as as a function of of time time during during experimental experimental measurement. measurement. Figure 10. The wind speed as aa function of time during experimental measurement.

(a) (a)

(b) (b)

(c) (c)

Figure 11. Wind speed field in experiment (unit: m/s) at (a) 3:00, 13 August (b) 22:00, 16 August (c) Figure 11. Wind at at (a)(a) 3:00, 13 13 August (b)(b) 22:00, 16 August (c) Figure Wind speed speedfield fieldininexperiment experiment(unit: (unit:m/s) m/s) 3:00, August 22:00, 16 August 3:00, 20 11. August. 3:00, 20 August. (c) 3:00, 20 August. )

)

Figure 12 demonstrates the average shipping noise source intensity based on the volunteer Figure 12 demonstrates the average shipping noise source intensity based on the volunteer observation system (VOS) database for frequencies of 50, 100 and 150 Hz. VOS database, which observation system (VOS) database for frequencies of 50, 100 and 150 Hz. VOS database, which provides information on the waypoints of global merchant ships, was used for modeling distant provides information on the waypoints of global merchant ships, was used for modeling distant shipping noise source distribution. Approximately 10–12% of the global fleet provides their ship shipping noise source distribution. Approximately 10–12% of the global fleet provides their ship position to the VOS. The VOS ships were calculated to consist of about 82% ANDES type “merchant position to the VOS. The VOS ships were calculated to consist of about 82% ANDES type “merchant

Sensors 2018, 18, 319

12 of 18

Figure 12 demonstrates the average shipping noise source intensity based on the volunteer observation system (VOS) database for frequencies of 50, 100 and 150 Hz. VOS database, which provides information on the waypoints of global merchant ships, was used for modeling distant shipping noise source distribution. Approximately 10–12% of the global fleet provides their ship position to the VOS. The VOS ships were calculated to consist of about 82% ANDES type “merchant vessel” and 18% “large tanker” and determine the source levels (SL) of modern global fleet [34]. The global shipping data were provided in terms of kilometers of track per 10 × 10 km2 and then were reprocessed into the SL density maps by assuming a fixed speed to calculate the dwell or residence time for a ship in each cell. The ship count is the instantaneous number of ships instead of the number of ships that traverse the cell during a time of period. The number of new ships entering in a cell and the numbers of ships leaving the cell during a short snapshot length are assumed to be equal, thereby indicating that the average ship count in a cell remains constant. The Poisson-distribution based VOS database was used to generate 10,000 realizations, which represent 10,000 moments. For the Poisson distribution, the mean number in a cell is determined on the basis of the VOS database. Monte Carlo method based on Poisson distribution was applied to model the distant shipping noise source intensity and distribution [35]. And then the shipping noise source intensity and distribution was used to model2018, the 18, low-frequency ambient noise field. Furthermore, ocean bathymetry data were obtained Sensors x 12 of 18 from ETOPO1, which is a 1 arc-min global relief model developed by the National Geophysical Data Center (NGDC) bathymetry from database was compared with in aboard the depthometer in [36]. the The experiment and was approximately accurate the depthometer deep ocean.inThe experiment and was approximately in the deepthe ocean. The abovementioned parameters abovementioned parameters were accurate applied to model low-frequency ambient noise field inwere the applied to model the low-frequency ambient noise field in the experiment. experiment.

(a)

(b)

(c)

Figure source intensity based onon VOS forfor frequency of (a) Figure 12. 12. The Theaverage averageshipping shippingnoise noise source intensity based VOS frequency of 50 (a) Hz 50 (b) Hz 2/Hz/m 2 @ 1 m). 100 Hz and (c) 150 Hz (unit: dB re 1 μPa 2 2 (b) 100 Hz and (c) 150 Hz (unit: dB re 1 µPa /Hz/m @ 1 m).

4.2. Experiment Verification 4.3. Experiment Verification The acoustic field was calculated along a fan of radials with a 3° angle at each position of the The acoustic field was calculated along a fan of radials with a 3◦ angle at each position of receiver hydrophones in the experiments. The transmission losses were calculated along the 120 slices the receiver hydrophones in the experiments. The transmission losses were calculated along the for frequencies of 50, 100 and 150 Hz by utilizing the standard ray approach and parabolic equation 120 slices for frequencies of 50, 100 and 150 Hz by utilizing the standard ray approach and parabolic solution method jointly for low-frequency ambient noise field. The near-field wind-driven and farequation solution method jointly for low-frequency ambient noise field. The near-field wind-driven field distant shipping noise fields could be calculated through the ray approach and parabolic and far-field distant shipping noise fields could be calculated through the ray approach and parabolic equation solution method accurately and effectively in the range-dependent deep ocean equation solution method accurately and effectively in the range-dependent deep ocean environment, environment, respectively. The radii along each radial were 100 and 500 km for the wind-generated respectively. The radii along each radial were 100 and 500 km for the wind-generated and distant and distant shipping noise fields correspondingly. The contribution of the shipping noise field within shipping noise fields correspondingly. The contribution of the shipping noise field within 50 km was 50 km was disregarded in this work but was considered as interference in modeling the spatial disregarded in this work but was considered as interference in modeling the spatial properties of the properties of the noise field. The experimental measurement data at three moments, which were 3:00 noise field. The experimental measurement data at three moments, which were 3:00 13 August, 22:00 13 August, 22:00 16 August and 3:00 20 August, were selected to verify the previously discussed 16 August and 3:00 20 August, were selected to verify the previously discussed theory considering theory considering the interference of near-field discrete ship and the wind speed above the vertical the interference of near-field discrete ship and the wind speed above the vertical line array. At each line array. At each moment, 15 min noise data that were measured with the deployed VLA were cut moment, 15 min noise data that were measured with the deployed VLA were cut into several segments into several segments of 30 s data. Each segment was used to calculate the spatial vertical correlation of 30 s data. Each segment was used to calculate the spatial vertical correlation coefficient and vertical coefficient and vertical directionality of the ambient noise field. Then, the spatial property results directionality of the ambient noise field. Then, the spatial property results with all of the segments with all of the segments were averaged to eliminate the transient noise interference and reduce the error. The wind speeds above the VLA at three times were approximately equal to 7, 12 and 3 m/s and are illustrated in Figures 10 and 11. Figure 13 shows the time frequency spectrogram in the experiment for wind speed of 3, 7 and 12 m/s which corresponded to the three moments. When the wind speed was 3 m/s, the noise field

Sensors 2018, 18, 319

13 of 18

were averaged to eliminate the transient noise interference and reduce the error. The wind speeds above the VLA at three times were approximately equal to 7, 12 and 3 m/s and are illustrated in Figures 10 and 11. Figure 13 shows the time frequency spectrogram in the experiment for wind speed of 3, 7 and 12 m/s which corresponded to the three moments. When the wind speed was 3 m/s, the noise field was dominated by distant shipping noise completely; when the wind speed was 12 m/s, wind driven noise had significant effects on the total noise field, received levels increased greater than 5 dB for frequencies up to 500 Hz when compared with those at wind speed of 3 m/s; when the wind speed was 7 m/s, the total noise field may be the weighted results of wind-driven and distant shipping noise field. Furthermore, the shipping density may decrease due to high sea state and high wind speed condition. Therefore, the received levels for frequency lower than 150 Hz at wind speed of 7 m/s were less than those at wind speed of 3 m/s. As the wind speed decreased, the shipping density may increase, and distant shipping noise field may have significant effects on the total noise field and take up a greater in the total noise field. Sensors 2018, 18,proportion x 13 of 18

(a)

(b)

(c)

Figure 13. 13. Time Time frequency frequencyspectrogram spectrogram(unit: (unit: dB dBre re11µPa μPa22/Hz) /Hz) in the experiment experiment for for wind wind speed speed of of Figure in the (a) 7 m/s at 3:00, 13 August, (b) 12 m/s at 22:00, 16 August and (c) 3 m/s at 3:00, 20 August. (a) 7 m/s at 3:00, 13 August, (b) 12 m/s at 22:00, 16 August and (c) 3 m/s at 3:00, 20 August.

Figure 14–Figure 16 depict the vertical correlation coefficient and directionality at three times Figures 14–16 depict the vertical correlation coefficient and directionality at three times that that corresponded to the three wind speed conditions in the experiment. Here the received level was corresponded to the three wind speed conditions in the experiment. Here the received level was replaced by the normalized power for some reasons. The reference value was the maximum of replaced by the normalized power for some reasons. The reference value was the maximum of original original vertical directionality of total noise field. The normalized power is defined as a power vertical directionality of total noise field. The normalized power is defined as a power relative to relative to reference value, and the normalized power is defined as the logarithm of a ratio which is reference value, and the normalized power is defined as the logarithm of a ratio which is output of output of beam former divided by reference value in angular space. The wind-driven noise induced beam former divided by reference value in angular space. The wind-driven noise induced by wind by wind agitation and breaking waves significantly influenced the spatial vertical correlation agitation and breaking waves significantly influenced the spatial vertical correlation coefficient and coefficient and vertical directionality under the high wind speed condition. In Figure 16, the distant vertical directionality under the high wind speed condition. In Figure 16, the distant shipping noise shipping noise field was the dominant noise field when the wind speed was 3 m/s, and the spatial field was the dominant noise field when the wind speed was 3 m/s, and the spatial characteristics of the characteristics of the total ambient noise field were consistent with those of the shipping noise field total ambient noise field were consistent with those of the shipping noise field completely. In Figure 15, completely. In Figure 15, the near field wind generated noise field was the dominant noise field when the near field wind generated noise field was the dominant noise field when the wind speed was the wind speed was 12 m/s, and the spatial vertical directionality and correlation coefficient of the 12 m/s, and the spatial vertical directionality and correlation coefficient of the total ambient noise field total ambient noise field were nearly consistent with those of wind driven noise field at frequency of were nearly consistent with those of wind driven noise field at frequency of 150 Hz. For frequencies 150 Hz. For frequencies of 50 and 100 Hz, the first zero radii were many times a half wavelength but of 50 and 100 Hz, the first zero radii were many times a half wavelength but tended to the results of tended to the results of the wind-driven noise field. At high wind speed condition the vertical the wind-driven noise field. At high wind speed condition the vertical structure of wind-generated structure of wind-generated noise field significantly influenced that of the total noise field. In Figure 14, noise field significantly influenced that of the total noise field. In Figure 14, the spatial directionality the spatial directionality and correlation coefficient results for wind speed between 3 m/s and 12 m/s and correlation coefficient results for wind speed between 3 m/s and 12 m/s were the weighed results were the weighed results of the wind-generated and distant shipping noise fields. The spatial results of the wind-generated and distant shipping noise fields. The spatial results had close relationship had close relationship with the proportion of wind-driven noise and distant shipping noise source in with the proportion of wind-driven noise and distant shipping noise source in the total noise field. the total noise field. Wind-driven noise and distant shipping noise arrived at the VLA through Wind-driven noise and distant shipping noise arrived at the VLA through different propagation paths different propagation paths in the experimental oceanic environment. The wind-driven noise can in the experimental oceanic environment. The wind-driven noise can reach the receiver VLA by rays reach the receiver VLA by rays along the reliable acoustic path with low transmission loss and steep along the reliable acoustic path with low transmission loss and steep grazing angle or great vertical grazing angle or great vertical wavenumber, whereas the distant shipping noise can reach the wavenumber, whereas the distant shipping noise can reach the receiver VLA with high transmission receiver VLA with high transmission losses and shallow grazing angle or little vertical wavenumber. The ocean acoustic channel was an incomplete channel because of the lack of surface conjugate depth. The acoustic rays could not reverse completely but were reflected by the bottom with reflection losses. Therefore, the power is less along the upward rays than along the downward rays in the noise directionality pattern of noise field. The vertical directionality of distant shipping noise exhibited an approximately symmetrical dual peak structure because of the downward and upward rays along

Sensors 2018, 18, 319

14 of 18

losses and18, shallow grazing angle or little vertical wavenumber. The ocean acoustic channel Sensors 2018, x 14 was of 18 Sensors 2018, 18, xchannel because of the lack of surface conjugate depth. The acoustic rays could 14not of 18 an incomplete reverse completely but were reflected bythe thetotal bottom with reflection losses. Therefore, thelevels power less coefficient in the simulation results of noise field. Second, the noise source of iswind coefficient in the simulation results of the total noise field. Second, the noise source levels of wind along the upward rays than along the downward rays in the noise directionality pattern of noise field. and distant shipping were not accurate completely but were the empirical results. Third, the and distant shipping were not accurate completely but were the empirical results. Third, the The vertical directionality of distant shipping noise exhibited an approximately dual peak geoacoustic parameters may deviate from the inversion results far from the symmetrical VLA position; besides geoacoustic parameters may deviate from the inversion results far from the VLA position; besides structure because of the downward andspeed upward rays along the paths of from the deep soundmay channel or the geoacoustic parameters, the sound profiles that were distant the VLA also be the geoacoustic parameters, the sound speed profiles that were distant from the VLA may also be convergence zone. inconsistent with measured result at the position of VLA in range-dependent waveguide. inconsistent with measured result at the position of VLA in range-dependent waveguide.

(a) (a)

(b) (b)

(c) (c)

(d) (d)

(e) (e)

(f) (f)

Figure 14. Vertical correlation coefficient and directionality at 3:00, 13 August for frequencies of (a,d) Figure correlation coefficient coefficientand anddirectionality directionalityatat 3:00, August frequencies of Figure14. 14. Vertical correlation 3:00, 13 13 August for for frequencies of (a,d) 50 Hz, (b,e) 100 Hz and (c,f) 150 Hz. (a,d) 50 Hz, (b,e) 100 Hz and (c,f) 150 Hz. 50 Hz, (b,e) 100 Hz and (c,f) 150 Hz.

(a) (a)

(b) (b)

(c) (c)

(d) (d)

(e) (e)

(f) (f)

Figure 15. Vertical correlation coefficient and directionality at 22:00, 16 August for frequencies of (a,d) Figure15. 15. Vertical Vertical correlation coefficient and 22:00, 16 16 August for for frequencies of (a,d) Figure correlation anddirectionality directionalityatat 22:00, August frequencies of 50 Hz, (b,e) 100 Hz and (c,f) 150coefficient Hz. 50 Hz, (b,e)(b,e) 100 100 Hz and (c,f)(c,f) 150 150 Hz. Hz. (a,d) 50 Hz, Hz and

Sensors 2018, 18, 319 Sensors 2018, 18, x

15 of 18 15 of 18

(a)

(b)

(c)

(d)

(e)

(f)

Figure16. 16. Vertical 3:00, 20 20 August for for frequencies of (a,d) Figure Vertical correlation correlation coefficient coefficientand anddirectionality directionalityatat 3:00, August frequencies of 50 Hz, (b,e) 100 Hz and (c,f) 150 Hz. (a,d) 50 Hz, (b,e) 100 Hz and (c,f) 150 Hz.

5. Conclusions The vertical correlation coefficient results in three wind conditions were compared with the typical The array deep gain exhibits the enhancement of function. received signal-to-noise ratio (SNR) in a sonarofarray Cron/Sherman water noise field coherence The vertical correlation coefficients the system. In were addition, the array gain is directly the spatial correlation coefficient of the C/S theory approximately consistent with related those oftowind-generated noise field in numerical ambient noise field. measurements of spatial of ambient noise fieldgradient. include calculation because of The the presentence of the bottom properties reflection and the sound speed considerable the information on the ocean acoustic environment. modeling the spatial Furthermore, first zero location occurred at a half wavelength Therefore, when the wind generated noise correlation of the ambient noise field necessary for sonar prediction sources wereand thedirectionality dominant noise source. However, theisfirst zeros radius wasperformance multiple times a half and ocean when environment this work, the resultsnoise of the total noise field were wavelength the totalinversion. noise field In was dominated bysimulation distant shipping sources. The simulation consistent experimental results with thethe described modeling method in the direct-arrival results of thewith totalthe noise field were consistent with experimental results of the proposed modeling zone in the deep ocean. Therefore, several conclusions are summarized as follows: method. However, several factors may induce errors in modeling the noise field spatial properties. First, database may describe the ship noise distribution VLA completely and accurately, (1) the TheVOS wind-driven andnot distant shipping field near may the significantly and simultaneously thereby inducing errors in the vertical directionality and correlation coefficient in the simulation results influence the spatial characteristics of the total noise field. The ray approach and parabolic of the equation total noisesolution field. Second, noise source levels to of wind shippingambient were notnoise accurate methodthe were jointly utilized modeland thedistant low frequency field completely but were the empirical results. Third, the geoacoustic parameters may deviate from in the range-dependent deep ocean environment by considering their calculation accuracy the and inversion resultsinfar from the VLA besides the geoacoustic parameters, theVOS sound speed efficiency wind-driven andposition; distant shipping noise fields. The NCEP and reanalysis profiles that werewere distant from the VLA may also besource inconsistent with measured result at the position databases used to model ambient noise intensity and distribution. of(2)VLA in range-dependent waveguide. The spatial vertical directionality and correlation of total noise field were analyzed in three

scenarios that corresponded to three wind speed conditions. The total noise field was dominated 5. Conclusions by distant shipping noise when the wind speed was less than 3 m/s. The spatial vertical The array gain enhancement of received signal-to-noise ratio (SNR) inconsistent a sonar array correlation andexhibits verticalthe directionality of the total noise field were approximately with system. In addition, the array gain is directly related to the spatial correlation coefficient of that of the shipping noise field; The near-field wind-generated noise source became the the ambient noise field. Thesource measurements spatial properties of ambient field include considerable dominant noise when theofwind speed was larger than noise 12 m/s, and the spatial vertical information on the ocean acoustic environment. Therefore,consistent modelingwith the spatial and correlation of the total noise field was approximately that of correlation the wind-driven directionality of the ambient noise field is necessary for sonar performance prediction and ocean noise field at 150 Hz; Furthermore, the spatial correlation coefficient results were the weighted environment In this work, the simulation results of the total noise field werefor consistent with results inversion. of the wind-generated noise field and distant shipping noise fields wind speeds the experimental results with the described modeling method in the direct-arrival zone in the deep between 3 and 12 m/s. ocean. Therefore, conclusions areof summarized as follows: (3) The verticalseveral directionality pattern total noise field was the hybrid result of the wind-driven and distant shipping noise fields because of their different arrival paths. The wind-generated (1) The wind-driven and distant shipping noise field maythe significantly and simultaneously noise sources reaching the received VLA were along direct or reliable acoustic pathinfluence with low the spatial characteristics of grazing the totalangle. noiseBy field. The ray and parabolic equationat transmission loss and steep contrast, the approach distant shipping noises arriving solution method jointly utilized model thechannel low frequency ambientzone noise field in the the receiver VLA were were through paths oftodeep sound or convergence with relatively high transmission loss and shallow grazing angle. The vertical directionality of the shipping

Sensors 2018, 18, 319

(2)

(3)

(4)

(5)

16 of 18

range-dependent deep ocean environment by considering their calculation accuracy and efficiency in wind-driven and distant shipping noise fields. The NCEP and VOS reanalysis databases were used to model ambient noise source intensity and distribution. The spatial vertical directionality and correlation of total noise field were analyzed in three scenarios that corresponded to three wind speed conditions. The total noise field was dominated by distant shipping noise when the wind speed was less than 3 m/s. The spatial vertical correlation and vertical directionality of the total noise field were approximately consistent with that of the shipping noise field; The near-field wind-generated noise source became the dominant noise source when the wind speed was larger than 12 m/s, and the spatial vertical correlation of the total noise field was approximately consistent with that of the wind-driven noise field at 150 Hz; Furthermore, the spatial correlation coefficient results were the weighted results of the wind-generated noise field and distant shipping noise fields for wind speeds between 3 and 12 m/s. The vertical directionality pattern of total noise field was the hybrid result of the wind-driven and distant shipping noise fields because of their different arrival paths. The wind-generated noise sources reaching the received VLA were along the direct or reliable acoustic path with low transmission loss and steep grazing angle. By contrast, the distant shipping noises arriving at the receiver VLA were through paths of deep sound channel or convergence zone with relatively high transmission loss and shallow grazing angle. The vertical directionality of the shipping noise field exhibited a symmetrical dual-peak structure along a pair of upgoing and downgoing rays along convergence zone path. The spatial correlation coefficients of the numerical results of each type of noise field were compared with those proposed by Cron/Sherman. The vertical correlation coefficient of the wind-generated noise field was nearly consistent with the C/S result. The first zero location occurred at a half wavelength when the wind generated noise sources were the dominant noise source. However, the first zeros radius was multiple times larger than a half wavelength when the total noise field was dominated by the distant shipping noise sources. The first zero location demonstrated a complex relationship with the competition of wind-driven and distant shipping noise sources in the total noise field. Several factors may induce errors in modeling the noise field spatial properties. First, the VOS database may not describe the ship distribution near the VLA completely and accurately, thereby inducing errors in the vertical directionality and correlation coefficient in the simulation results of the total noise field. Second, the noise source levels of wind and distant shipping were inaccurate but were the empirical results. Third, the geoacoustic parameters may deviate from the inversion results far from the VLA position; besides, the sound speed profiles that were distant from the VLA may be inconsistent with measured result at the position of VLA.

Acknowledgments: This project was supported by the National Natural Science Foundation of China (Grant No. 11174235, Grant No. 11574251, Grant No. 11704313 and Grant No. 61701405). Author Contributions: Qiulong Yang (Q.Y.) contributed to the concept of the research, performed the data analysis and wrote the manuscript; Kunde Yang (K.Y.) revised the manuscript, approved the final version, and provided the necessary support for the research; Ran Cao (R.C.) helped perform the data analysis with constructive discussions and helped performed the experiment. Shunli Duan (S. D.) conducted the experiment and collected the data. Conflicts of Interest: The authors declare no conflict of interest.

References 1.

Buckingham, M.J. Cross-correlation in band-limited ocean ambient noise fields. J. Acoust. Soc. Am. 2012, 131, 2643–2657. [CrossRef] [PubMed]

Sensors 2018, 18, 319

2.

3. 4. 5. 6. 7. 8. 9.

10. 11. 12. 13. 14. 15.

16. 17. 18. 19. 20. 21. 22. 23. 24.

25.

17 of 18

Fried, S.E.; Walker, S.C.; Hodgkiss, W.S.; Kuperman, W.A. Measuring the effect of ambient noise directionality and split-beam processing on the convergence of the cross-correlation function. J. Acoust. Soc. Am. 2013, 134, 1824–1832. [CrossRef] [PubMed] Roux, P.; Sabra, K.G.; Kuperman, W.A.; Roux, A. Ambient noise cross correlation in free space: Theoretical approach. J. Acoust. Soc. Am. 2005, 117, 79–84. [CrossRef] [PubMed] Xia, H.J.; Yang, K.D.; Ma, Y.L. A noise removal method for uniform circular arrays in complex underwater noise environments with low SNR. Sensors 2017, 17, 1345. [CrossRef] [PubMed] Kuperman, W.A. Spatial correlation of surface generated noise in a stratified ocean. J. Acoust. Soc. Am. 1980, 67, 1988–1996. [CrossRef] Lee, K.; Seong, W. Ray-based model for spatial coherence of ocean surface generated noise and its approximation in a triplet array. J. Ocean Eng. 2017, 42, 199–207. [CrossRef] Buckingham, M.J. Theory of the directionality and spatial coherence of wind driven ambient noise in a deep ocean with attenuation. J. Acoust. Soc. Am. 2013, 134, 950–958. [CrossRef] [PubMed] Buckingham, M.J.; Carbone, N.M. Source depth and the spatial coherence of ambient noise in the ocean. J. Acoust. Soc. Am. 1997, 102, 2637–2644. [CrossRef] Buckingham, M.J. On the two-point cross-correlation function of anisotropic, spatially homogeneous ambient noise in the ocean and its relationship to the Green’s function. J. Acoust. Soc. Am. 2011, 129, 3562–3576. [CrossRef] [PubMed] Walker, S.C.; Buckingham, M.J. Spatial coherence and cross correlation of three dimensional ambient noise fields in the ocean. J. Acoust. Soc. Am. 2012, 131, 1079–1086. [CrossRef] [PubMed] Xiao, P.; Yang, K.D.; Lei, B. Model of shipping noise in deep water: Directional density and spatial coherence functions. China Ocean Eng. 2017, 30, 591–601. [CrossRef] Barclay, D.R.; Buckingham, M.J. On the spatial properties of ambient noise in the Tonga Trench including effects of bathymetric shadowing. J. Acoust. Soc. Am. 2014, 136, 2497–2511. [CrossRef] [PubMed] Barclay, D.R.; Buckingham, M.J. Depth dependence of wind-driven, broadband ambient noise in the Philippine Sea. J. Acoust. Soc. Am. 2013, 133, 62–71. [CrossRef] [PubMed] Naughton, P.; Roux, P.; Yeakle, R.; Schurgers, C.; Kastner, R.; Jaffe, J.S.; Roberts, P.L.D. Ambient noise correlations on a mobile deformable array. J. Acoust. Soc. Am. 2016, 140, 4260–4270. [CrossRef] [PubMed] Baggeroer, A.B.; Scheer, E.K.; Colosi, J.A.; Cornuelle, B.D.; Dushaw, B.D.; Dzieciuch, M.A.; Howe, B.M.; Mercer, J.A.; Munk, W.H.; Spindel, R.C.; et al. Statistics and vertical directionality of low-frequency ambient noise at the North Pacific Acoustic Laboratory site. J. Acoust. Soc. Am. 2007, 117, 1643–1665. [CrossRef] Farrokhrooz, M.; Wage, K.E.; Dziciuch, M.A.; Worcester, P.F. Vertical line array meaments of ambient noise in the North Pacific. J. Acoust. Soc. Am. 2017, 141, 1571–1581. [CrossRef] [PubMed] Zhang, Q.Z.; Guo, X.Y.; Ma, L. The research of the characteristics of the ambient noise under varying environment. J. Comput. Acoust. 2017, 25, 1750021. [CrossRef] Harrison, C.H. Formulas for ambient noise level and coherence. J. Acoust. Soc. Am. 1996, 99, 2055–2066. [CrossRef] Carey, W.M.; Evans, R.B. Ocean Ambient Noise Measurement and Theory; AIP Press: New York, NY, USA, 2011; pp. 102–116. Yang, Q.L.; Yang, K.D.; Duan, S.L. A method for noise source levels inversion with underwater ambient noise generated by typhoon in Deep Ocean. J. Comput. Acoust. 2018, 26, 1850007. [CrossRef] Cox, H. Spatial correlation in arbitrary noise fields with application to ambient sea noise. J. Acoust. Soc. Am. 1973, 54, 1289–1301. [CrossRef] Cron, B.F.; Sherman, C.H. Spatial correlation functions for various noise models. J. Acoust. Soc. Am. 1961, 34, 1732–1736. [CrossRef] Hamson, R.M. The modeling of ambient noise due to shipping and wind sources in complex environments. Appl. Acoust. 1997, 51, 251–287. [CrossRef] Porter, M.B. The BELLHOP Manual and User’s Guide: Preliminary Draft; Heat, Light, and Sound Research, Inc.: La Jolla, CA, USA, 2011. Available online: http://oalib.hlsresearch.com/Rays/HLS-2010-1.pdf (accessed on 31 January 2011). Collins, M.D. User’s Guide for RAM Versions 1.0 and 1.0p; Naval Research Lab.: Washington, DC, USA. Available online: http://oalib.hlsresearch.com/PE/RAM/ (accessed on 3 September 2017).

Sensors 2018, 18, 319

26. 27. 28. 29. 30. 31. 32. 33. 34. 35.

36.

18 of 18

Harrison, C.H. CANARY: A simple model of ambient noise and coherence. Appl. Acoust. 1997, 51, 289–315. [CrossRef] Li, J.; Li, Z.L.; Ren, Y.; Li, W.; Zhang, R.H. Horizontal-longitudinal correlations of acoustic field in deep water. Chin. Phys. Lett. 2015, 32, 107–110. [CrossRef] Duan, R.; Yang, K.D.; Ma, Y.L. A reliable acoustic path: Physical properties and a source localization method. Chin. Phys. B 2012, 21, 276–289. [CrossRef] Duan, R.; Yang, K.D.; Ma, Y.L.; Yang, Q.L.; Li, H. Moving source localization with a single hydrophone using multipath time delays in the deep ocean. J. Acoust. Soc. Am. 2014, 136, EL159–EL165. [CrossRef] [PubMed] Lei, Z.X.; Yang, K.D.; Ma, Y.L. Passive localization in the deep ocean based on the cross-correlation function matching. J. Acoust. Soc. Am. 2016, 139, EL196–EL201. [CrossRef] [PubMed] Yang, K.D.; Xiao, P.; Duan, R.; Ma, Y.L. Bayesian Inversion for Geoacoustic Parameters from Ocean Bottom Reflection Loss. J. Comput. Acoust. 2017, 25, 1750019. [CrossRef] Dong, H.F.; Chapman, N.R.; Hannay, D.E.; Dosso, S.E. Estimation of seismic velocities of upper oceanic crust from ocean bottom reflection loss data. J. Acoust. Soc. Am. 2010, 127, 2182–2192. [CrossRef] [PubMed] National Center of Environment Prediction (NCEP). Available online: https://rda.ucar.edu/ (accessed on 3 September 2017). Porter, M.B.; Henderson, L.J. Modeling ocean noise on the global scale. In Proceedings of the INTERNOISE, Melbourne, Australia, 16–19 November 2014; pp. 1–7. Li, Z.; Zurk, L.M.; Ma, B. Vertical arrival structure of shipping noise in deep water channels. In Proceedings of the OCEANS 2010, Seattle, WA, USA, 20–23 September 2010; IEEE: Piscataway, NJ, USA, 2010; pp. 1–8. [CrossRef] National Geophysical Data Center. Available online: https://www.ngdc.noaa.gov/mgg/global/global.html (accessed on 3 September 2017). © 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).