Sunteți pe pagina 1din 29

Applied Acoustics 37 (1992) 111-139

Heuristic Model for Outdoor Sound Propagation Based on an Extension of the Geometrical Ray Theory in the Case of a Linear Sound Speed Profile
A. L'Esp6rance, a P. Herzog, b G. A. Daigle c & J. R. N i c o l a s
* G.A.U.S., Universit6 de Sherbrooke, Qu6bec, Canada, J1K 2R1 bUniversit6 du Maine, Le Mans, France Clnstitute for Microstructural Sciences, National Research Council, Ottawa, Ontario, Canada, K1A 0R6 (Received 20 May 1991; revised version received 17 January 1992; accepted 20 January 1992

A BSTRA CT In the case of outdoor sound propagation at relatively short ranges, it is reasonable to assume that sound follows straight ray paths. A t longer ranges, refraction due to temperature and wind gradients results in ray paths that are curved, and calculations using straight rays are no longer valid and lead to erroneous results. In addition, atmospheric turbulence plays an everincreasing role by degrading the coherence of the soundfield. In this paper, the theory validfor propagation from a point source above an impedance plane is extended to include the effects of curved rays and the loss of coherence due to atmospheric turbulence. The model assumes a sound speed gradient that varies linearly with height above the ground. This assumption allows an analytical determination of the travel times of the curved rays and the modified angle of incidence of the ground-reflected paths. The additional reflected rays that may appear in the presence of a positive gradient are included in the calculation. The assumption also permits the determination of the position of the shadow zone in the presence of a negative gradient. The total sound pressure level is computed by summing the contribution from all the rays. In this summation, a fluctuating index of refraction is used to take into account the partial coherence between the rays caused by the effects of atmospheric turbulence. I f the receiver is in the shadow "zone, a diffraction theory based on a residue series solution is used to compute the sound levels. The results of calculations using the model are compared with experimental results. 111 Applied Acoustics 0003-682X/92/$05.00 1992 Elsevier Science Publishers Ltd, England. Printed in Great Britain

112

A. L'Espbrance et al.

INTRODUCTION For the last 30 years, outdoor sound propagation near the ground has been the subject of extensive research. Practical models that consider spherical spreading, atmospheric absorption and ground effects have been developed and it is now possible to predict sound pressure levels generated by a source over a finite impedance ground with relatively good accuracy at short distances. This field has been reviewed by Piercy e t al. 1 in 1977, and more recently by Attenborough. 2 However, most of the models cannot explain the lower or higher levels commonly observed when dealing with long-range sound propagation close to the ground. Two additional phenomena that contribute to these variations are refraction (related to temperature and/or wind gradients) and atmospheric turbulence. Although these phenomena and their interactions have been studied by many researchers (see, for example, Refs 3-5), it has not been possible to achieve a simple model for engineering purposes that incorporates all the combined physical phenomena. More recently, theories developed for microwaves and used in underwater acoustics have been adapted for outdoor sound propagation. These solutions can deal with arbitrary temperature and wind conditions, and even with slowly varying ground shapes. Examples are the fast field program described by Lee e t al. 6 and the parabolic equation described by White and Gilbert.7 (An interesting comparison of these different approaches for ocean acoustics has been made recently by Harrison. a) However, the disadvantages of these solutions are that they require excessive computation time and a detailed understanding of outdoor sound propagation to give a physical interpretation of the results obtained. The aim of the work described in this paper was to develop a practical but reliable model to predict outdoor sound propagation for general engineering purposes. The model attempts to be as complete as possible by incorporating most of the important phenomena involved in outdoor sound propagation. This model, based on geometrical ray theory, takes into account the effects of geometrical spreading, atmospheric absorption and the ground effect. The effect of atmospheric refraction is evaluted by assuming a linear sound speed profile which allows an analytical determination of all the possible ray paths between a source and a receiver close to the ground. The amplitudes and phase contribution of each ray are evaluated and summed by assuming partial coherence between the rays. The model therefore represents an extension of the ideas reported by Hidaka et aL, 9 but also includes the effects of turbulence and multiple ray paths. Moreover, when the receiver is in the shadow zone (i.e. when no ray can reach the receiver as the result of a negative gradient), the model uses the diffraction solution to calculate the sound pressure level at the receiver.

Heuristic modelfor outdoor sound propagation

113

In this paper, the model itself is described first, then a parametric analysis is presented and finally the results given by the model are compared with experimental and theoretical results found in the literature.

D E S C R I P T I O N OF T H E M O D E L This section describes the geometrical ray model components and their interactions. The ground reflection equation, the ray calculation equations and the turbulence calculation are discussed separately and the combined calculation is described. Spherical wave evaluation of ground effects without refraction The basic problem to be solved is described in Fig. 1. In the absence of atmospheric effects and/or for short source-receiver distance, the sound emitted by a monopole source S travels to a receiver R at an horizontal distance D following a direct ray path and a reflected ray path incident to the ground at an angle ~c. The acoustic pressure at the receiver can be computed by summing the contribution of the direct ray and the reflected ray. A useful approximation to this problem is 2 p(R) = exp ( j k R 0 + Q exp (jkR2) R1 R2 (1)

where R x and R 2 represent the geometrical path lengths of the direct and of the reflected ray, respectively, k is the wavenumber and Q is the so-called spherical reflection coefficient:
Q = Rp + (1 -

Rp)r(w)

(2)

where Rp is the plane wave reflection coefficient: sin ~,6 - (1/ZG) Rp = sin ~'c + (l/Z6) R1 R (3)

Fig. 1. Generalschemeof the sound propagation over a ground of finiteimpedanceZG.The sound pressure at the receiveris the sum of the contributions of the direct and reflectedrays.

114

A. L'Espbrance et ai.

where Z6 is the ground impedance normalized to pc, the characteristic air impedance. It is sufficient for our purpose to use the empirical model described by Delany and Bazley. 1 Many models exist to determine Z6, in particular, Attenborough 11 has developed a variety of acoustical impedance models for outdoor ground surfaces which appear to provide the best description of outdoor ground impedance. However, for the purpose of this paper, a single parameter model has been used x and is given by Z ~ = [ 1 + 9-08(f) -'7s + j l 1.9(f)'731 (4)

where tr is the effective flow resistivity; values of tr for several common types of ground can be found in Ref. 12. In eqn (2), w is the numerical distance and is written as w2 = ~/'kR2[sin ~k6 + (1/ZG)] 2 (5)

The numerical computation of F(w) in eqn (2) can be done in various ways. 13.14 In the present model, the following series expansions are used and calculated until the additional nth term is less than 1 x 10-6.15 If Iwl < 5,
22 (2w) F(w) = 1 +jx/(rOw exp ( - w2) -- 2w 2 + ~ +... +

( - 1)"(2w2)"
1 x 3 x ...(2n + 1) (6a)

If IwL> 5, F(w) .~ 2jx/(n)w exp ( - w2)H(- Im (w)) 1 2w2 lx3 (2w.2) 2 .... -1 x 3...x(2n+l) (2w2), (6b)

where H is the Heaviside step function, defined by H ( - Im (w)) = 1, I m ( w ) < 0 0, otherwise

(7)

For computational purposes, it is useful to rewrite eqn (1) as 1 IQI2 21QI cos p2 = R--~-+ - ~ z + R---~ [k(Rz-R1)+Arg(Q)] (8)

In eqn (8), k(R 1 - R 2 ) represents the phase shift due to the path-length difference between R x and R z, and Arg(Q) is the phase shift due to the reflection on the finite impedance ground. Considering this expression, the average squared relative pressure can be considered as the sum of the incoherent field (the first two terms of eqn (8)) and of an algebraic term

Heuristic modelfor outdoor sound propagation

115

describing the partial interaction between the direct and the reflection rays (third term). This is the basic equation which will be used in the following sections.

Effect of atmospheric absorption


When a sound wave travels in air, viscosity effects, thermal diffusion, and vibrational and rotational relaxation of the molecules of the air cause the sound energy to be absorbed by the air. This absorption is principally a function of the temperature and humidity of the air, and it increases rapidly with frequency. To calculate this atmospheric attenuation, the Standard ANSI S1.2616 is used. We note that atmospheric attenuation is given in dB/i00 m, i.e. the decibel attenuation is constant for each meter of propagation, unlike geometrical spreading where the attenuation is 6 dB for each doubling distance. Thus, the atmospheric absorption tends to become important at medium to long distances. For computing purposes, an atmospheric absorption coefficient is defined in the model as A ( R ) = 10 AT(f)/2OOOR (9) where A T(f) is the atmospheric absorption in dB/100 m for the frequencyf given by the standard. Thus, eqn (8) can be written as

p2
-

A(R1) 2 A(R2)2IQI 2
2 X A(R1) x

A(R2)

IQI

R1R2

cos [k(R2 - R1) + Arg (Q)]

(10)

Effects of refraction
As a result of variations in wind and temperature with height, the sound speed also changes as a function of height above ground. According to Snelrs law, this variation of the sound speed, called the sound speed gradient, will change the direction of the sound r a y : The shape of the sound speed gradient above the ground can be very complicated in certain meteorological conditions, and at a specific height the acoustical rays may be more or less bent downward or upward depending on the sign and amplitude of the sound speed profile at this height. The general shape of the sound speed profile can be represented by a quadratic polynomial function, s'17 However, the use of such a function requires complex prediction models such as the parabolic equation, ray-tracing and other models, which is incompatible with the objective of the present work.

116

A. L'Espbrance et al.

F r o m a practical point of view, close to the ground for a typical sunny day, the air temperature generally decreases with height, whereas it increases with height during a clear night. In the same manner, the wind speed always decreases with height close to the ground. This causes the overall sound speed profile to decrease or increase with height depending on the time o f day, the relative strength of temperature and wind speed profiles, and/or the source-receiver orientation with respect to the wind direction. Thus, as a first approximation, it should be possible to assume a linear sound speed profile, expressed as

c(z) = c(OX 1 + az)

(11)

where c(0) is the reference sound velocity on the ground, and z is the height of the point at which the velocity is to be determined. The quantity a is the velocity gradient (it is assumed that the velocity is not a function o f x and y). Negative values for a correspond to a negative temperature gradient (lapse conditions), and/or upwind conditions, whereas positive values for a correspond to inversion and/or downwind conditions. We will now study the case o f a so-called 'moderate' sound speed gradient, a 'strongly positive' gradient and a 'strongly negative' gradient.

Effect o f a moderate sound speed gradient This paragraph follows closely the presentation made by Hidaka, 9 who showed the possible usefulness o f linear sound speed profiles in outdoor sound propagation as proposed by Brekhovskikh Is in an earlier work. In these studies, it has been shown that for a linear sound speed profile acoustical rays are arcs o f circles (see Fig. 2) with a radius o f curvature of R c = 1/(a cos (~'G)) and whose center o f curvature lies on an horizontal line at a height z c = - 1/a. (A curvature of the ground may also be taken into account using an adequate value o f a, i.e. a = 1/Rc, where Rc is the effective radius o f curvature, with a positive for concave cases and negative for convex ones

R-r-

i R

T
Z \

D \
"e

/')zG
lla C O S ~vG

Fig. 2. Schemeof an acoustic ray emitted by a source on the ground in the presence of a positive sound speed gradient. The ray is an arc of a circle of radius R== 1/(a cos ~bG).

Heuristic model for outdoor sound propagation

117

R
s

Fig. 3. When the ray path passes through its maximum height in the case of a positive gradient (or through its minimum height in the case of a negative gradient), two points on the arc have the same height.

(see Ref. 19).) Thus, for a source on the ground, the angle of reflection at the ground ~G, the length of the curved ray R(ZR) and the travel time Z(Za) can be calculated using the following expressions:
R(ZR) = acos(~kG) sin-a {(1 +aZR)COS~OG} - - ~ + ~ b C

(12)

Z(ZR) ~ 1 where

= 2acG

lnFf(O) l kf(z.)J

(13)

f ( z ) = 1 + x/J1 -- (1 + az) 2 cos 2 ~G] 1 - x/J1 -- (1 + az) 2 cos 2 ~b~]

(14)

with
aD

tan ffa = ~

-+

ZR(2+ azR) 2D

(15)

where D is the horizontal distance. When the ray path passes through its maximum height in the case of a positive gradient or through its minimttm height in the case of a negative gradient, two points on the arc of the circle have the s a m e height (see Fig. 3), so that eqns (12)-(15) cannot be used directly. As shown by Hidaka et al., 9 it is possible to remove this ambiguity by comparing the horizontal distance from the source to the receiver (D) with the horizontal distance from the source to the extremum of the ray (Din). Thus if z R, is used to represent the receiver position in Fig. 3 when D > Dm where Dm -tan ~ 6
a

(16)

(17)

118

A. L'Esp&ance et al.

then the path length and the travel time can be calculated using R(zR,) = 2 x R(zm) - R(zR) and z(zR,) = 2 x T(Zm) -- T(ZR) where
R(zm) = - (18)

(19)

a c o s ~/G

(20)

and T(Zm) = 2 a ' ~ In ]- _ sin with

[l+sin

(21)

Zm 1(CO . 1)

'22,

In eqns (18) and (19), R(zI~), R(zm) , T(ZR) and T(Zm)are calculated using eqns (12)-(15). In the previous analysis, it is assumed that the source is on the ground and z R > Zs. One could easily exchange zR and z s if the receiver is on the ground and Zs > zR. By extension, it is possible to calculate the total length and travel time z of the reflected ray as the sum of the separate parts. Although not explicitly discussed by Hidaka et al., 9 it is also possible to calculate the direct curved ray from S to R (or R to S, depending which is the higher point) by changing the reference of the sound speed gradient from e(0) to e(Zs) in eqn (11). Thus eqns (12)-(15) could be used with the relative sound speed gradient a, defined as a' = a

c(0) C(Zs)

(23)

In this case, one should use zh (or z~) instead of zR (or Zs): zh = ZR -- ZS (24)

In the case of a 'moderate' sound speed gradient, the refraction will cause only a deformation of the direct and reflected rays (Fig. 4) and one can calculate the effective geometrical parameters (lengths, traveling times and angle of reflection) of these two rays. F o r a 'moderate positive' sound speed gradient (downwind condition a n d / o r during the night), the rays are bent d o w n w a r d . The grazing angle increases, which reduces the excess attenuation. Also, the time difference between the direct and the reflected ray

Heuristic model for outdoor sound propagation

119

~/G/////////~~///////,
Fig. 4.

o~

R1

~oR

R
-__ R I . ~ 0

:/'////////////////////

In the presence of a moderate sound speed gradient and/or at short distance, the refracted case is a deformation of the non-refracted case.

paths is increased, and therefore their geometrical interference is shifted to higher frequencies. Inversely, in the case of a 'moderate negative' sound speed gradient (during the day or in upwind condition), the rays are curved upward. The grazing angle is reduced and thus the ground effect will be increased and the geometrical interference is shifted to higher frequencies. In eqn (10), this geometrical interference is represented by k ( R 2 -- R t)- However, because the sound speed varies with height, the wavenumber k is not constant for all the rays. To take into account this phenomenon, the model uses the difference of the travel times between the rays instead of the difference of travel paths. Therefore, eqn (10) is rewritten as
p2
=

A(RI) 2

A(R2)2"IQ[ 2
R

2 "A(R1)'A(R2)'[Q [
RxR2

cos [2nf(z 2 - zl) + Arg (Q)]

(25)

where "t1 and z2 are the effective travel times of R 1 and R 2 computed using eqns (13)-(24). Equation (25) could be associated with the case of a moderate sound speed gradient, i.e. when two rays reach the receiver. This equation is thus equivalent to eqn (47) of Hidaka et al. 9 and represents the case of a linear 'moderate' sound speed profile to which Hidaka et al. restricted'their analysis. This case is also equivalent to the cases studied recently by Li et
al. 2

Effect o f a strong positive gradient: a >>0

In the presence of a strong positive sound speed gradient, and/or for larger propagation distances, additional rays that go through n reflections at the ground appear between the source and the receiver (Fig. 5). It has been shown 2x that these additional rays can be determined using the following fourth-order equation
n(n + 1)x 4 -- (2n + 1)Dx a + [bR 2 + (2n 2 -- 1)b2 + O2]x 2

-- (2n - 1)b2Dx + n(n - 1)b~ = 0

(26)

120

A. L'Espkrance et al.

R(z s) vO~2.R(zr.) V~//"


2"/////////////////////2
k ( N - 1 ) * 2*D m

D
Fig. 5. Under a strong positive gradient (and/or at long distance), additional rays could reach the receiver after n reflections at the ground.

where b~=Z/(2+azi)
a

fori=SorR

In eqn (26), n is the number of reflections and the u n k n o w n is the horizontal distance between the source and the first reflection at the ground. This equation must be solved for n = 0, 1, 2, 3,... (the n u m b e r of reflections at the ground), until there is no real solution for x, i.e. as long as 0 < x < D. In the presence of a positive gradient a, it can be shown 21 that there is at least one direct ray (for n = 0) and one reflected ray (for n = 1). When the receiver moves closer to the ground and further from the source, or when the gradient increases, the complex conjugate o f the roots may become real, which means that additional reflected rays have to be considered. For n = 2, two additional reflected rays m a y appear, and for n > 2, four additional reflected rays may appear for each n. For each additional reflected ray, the geometrical and temporal parameters can be determined using R~ = 2 (n - 1) R(zm) + R(zs) + R(zR) zi = 2 x (n -- 1) x z(zm) + Z(Zs)+ z(zR) (27) (28)

where R(zs), R(zR), R(zm), ~(Zs), v(zR) and V(Zm)can be determined from eqns (12)-(24). It should be noted that although the above expressions are assumed to be correct, they lead to numerical problems when small reflecting angles are involved. For computational purposes, one should avoid the u s e of trigonometric functions. We replace all the trigonometric functions involved by an equivalent polynomial expression o f tan (w~). This technique leads to more precise results and improves the overall numerical efficiency. Defining N as the total n u m b e r of rays reaching the receiver (including the direct ray), the total sound pressure at the receiver can be calculated by

Heuristic modelfor outdoor sound propagation

121

summing the contributions of all the N rays involved. This is done in the model using the following expression:
N N i-1

i=1

t=2 ~=1

(29)

where i = 1 denotes the direct ray, and therefore Q~ = 1. At represents the standard attenuation of a single ray due to atmospheric absorption, computed using the refracted length of the path, R i, zi is the travel time of the ray and Q~ is the equivalent reflection coefficient at the ground of this ray, calculated with
at = a(~t)"'

(30)

where ~,~ is the angle of reflection at the ground and n i is the number of reflections. In the case of a moderate sound speed gradient, there is only one direct and one reflected ray. Therefore N = 2, and eqn {29) reduces to eqn (25). This case corresponds to the one to which Hidaka et al. limited their analysis. 9
Effect o f strong negative gradient: a <<0 When the negative sound speed gradient increases (or/and for large propagation distances), eqn (26) may not have a real solution. This means that no ray reaches the receiver, and thus results in an acoustical shadow zone (see Fig. 6). This arises when a < 0 and when

D > ( 2 ~ ) '/2 + ( 2 ~ )

'/z

(see Ref. 22,

In this case, the model uses the diffraction solution proposed by Ber,;y and

R =lla

~:4A~A e ~ w

//////////////////////~/////~//////////////,
Fig. 6. Under a strong negative sound speed gradient {and/or at long distance), the receiver can be in the shadow zone, which no ray enters.

122

A. L'Espbrance et al.

Daigle 22 instead of the geometrical acoustic solution. Thus, in the model the sound pressure in the shadow zone is evaluated with
rt eJ"/6 V H~ol)(k.r) x Ai[ b" -- (zsfl) e2i"/3]A,[ b. -- ( z d l) e 2i"/3] P(r,z) - - l /'~ A'~[b.]2-b.[Ai(b.)] 2
"

(31a) b. =
z e 2i"/3 = (k2n -

where
k 2 ) l 2 e 21n/3

(31b)

are the zeros of the expression


A'i(b.) + q eU'/3A+(b.) = 0

(31 c)

where
q =jkoplC/ZG z = (k 2 - k2)l 2 l = (R/2k~) ~/3 R = 1/a

with
ko = oge(o)

where c(0) is the local sound speed at the ground surface. In eqn (31), A i denotes the Airy function, A'i is its derivative, and H (1) o is the Hankel function of the first kind and of order zero. The task of evaluating the solutions b, of eqn (31c) is simple only in the limiting cases of a rigid surface (Z G ~ 0) for which asymptotic expressions exist. In the intermediate range of ground surfaces haying a finite impedance, eqn (31c) must be solved numerically for b,, for each value ofn required in the series. This involves solving the zeros of a complex function which depends on the complex variable b. A m o n g the many numerical methods that exist to solve such problems, the N e w t o n - R a p h s o n algorithm was found to be simple and accurate. In most situations of outdoor sound propagation, the Hankel function and the Airy functions that appear in the numerator of eqn (31a) may be represented by their asymptotic expressions. This simplification cuts down the computation time. Furthermore, if the receiver is deep within the shadow, the infinite sum is well approximated by its first term. In the general case, two terms are sufficient to obtain convergence within the whole shadow zone, and, to some extent, in the zone near the shadow boundary. Finally, one may note that this diffraction solution remains valid above the shadow boundary and can therefore be compared with the geometrical solution. When comparing this diffraction solution with the geometrical ray solution

Heuristic model for outdoor sound propagation

123

near the limiting ray, slight differences are, however, observed. This point will be discussed in detail below, in the section on appearance of additional rays.

Effects of atmospheric turbulence As discussed previously, under downwind propagation the number of rays reaching the receiver will vary from two to N rays for a given amplitude gradient and source-receiver geometry. In the presence of atmospheric turbulence, the resulting partial coherence between the rays must be taken into account when they are summed. The influence of atmospheric turbulence on the propagation of sound has been studied by Chernov 23 and Tatarski. 24 Their work has been adapted to the case of direct and reflected acoustical rays in the presence of the ground by Daigle and coworkers. 3'25'26 A further extension by Clifford and Lataitis 27 leads to the following practical formulation:
--=1 IQI2+ 21QI

2)=Rf+

RIR2

COS

[k(R2- R1)+ Arg(Q)]F

(32)

where F is an estimation of the coherence between the two rays and ( ) expresses a time average. Considering two adjacent rays with a maximum separation between them equal to p, F can be expressed as their mutual coherence function, and can be written as F(R, p) = exp [ - - x/(n)2 <#2>k2RL(1 = ~p/L e-t~ dt

c~(p/L)/(p/L))]

(33) (34)

Jo

The constants (/~2> and L are respectively the fluctuating index of refraction and the outer scale of the turbulence. Their values can be determined from meteorological measurements, although data corresponding to typical cases have been given by Daigle et al. 26 Equation (33) displays a slight problem when determining the separation p. Although it should be the maximum transverse separation, comparison with experiments shows discrepancies which can be avoided through the use of half the expected separation (see Ref. 26). As a consequence, it is believed that the effective separation should be calculated as half the actual one in a non-refractive case. For the refractive case, exact determination of the separation between rays may become tedious when numerous rays are involved. Therefore, an estimate of an effective separation would achieve only slight difference from an exact calculation, but would permit a far better numerical efficiency. The main fact concerning the separation is that it

124

A. L'Espbrance et al.

should be relatively small between rays belonging to the same path and larger between other rays. A rough estimation of an effective separation can be given by the difference of the average heights of the two rays, which can be computed very easily for a given reflection angle ~b c. It should be noted that this estimation leadsto half the maximum separation in the non-refractive case, and can be computed very easily for a given reflection angle ~ .

General algorithm
The general algorithm of the geometrical ray model proposed is shown in Fig. 7. The first step is to solve eqn (26) and find the total number of ray paths (N) that may exist between the source and the receiver in the case of a positive gradient, or if the receiver is in the shadow zone in the case of a negative gradient. If N > 0, the total field can be calculated with the following generalized expression:
N N i-1

i=1

/=2 j = l

(35) A 'IQ,I
C , - - -

with R,

where i = 1 denotes the direct ray and F,~ is the mutual coherence function between rays i and j, computed using eqn (33), with an effective separation p estimated as discussed in the previous section. If the fluctuating index (/~2) is zero, F o = 1; and eqn (35) is reduced to eqn (29), and if N = 2 (in the case of a moderate sound speed gradient), eqn (35) is reduced to eqn (24). On the other hand, if N = 0, the sound pressure is calculated according to eqn (30). Apart from bending the rays and possibly increasing their number, the refraction may also change the geometrical spreading of the rays. This phenomenon has been studied by Anderson et a l ) s and Jacobson. 29 In addition, one may also take into account the phase shift occurring at a caustic, as discussed by Brekhovskikh. is For the moment, these last two aspects are neglected in the model. The following sections will show that these simplifications can be used without reducing the overall usefulness of the model. Finally, one may note that the effect of turbulence on the sound level in the shadow zone is not taken into account in this model, as there is no simple solution available that could be introduced in the model to take such a phenomenon into consideration. However, research on turbulence effects is

Heuristic model for outdoor sound propagation

125

HEURISTIC MODEL OF SOUND PROPAGATION Is the receiver in the shadow zone (for a < 0)7 (D > (2 zs/a) 1/2 + {2 zR/a)1/2)

1 No
Determination of the N ray paths and associated geometrical parameters for the given geometry speed gradient 'a' (see eqn 26) Yes

II
Evaluation of the complex amplitude of each ray, Ci (see eqns 9 and 30) Computation of the diffracted pressure in the shadow zone ~see et n 31)

ii Ii
Computation of the total sound pressure considering the propagation time (xi)and turbulence between each ray~i (Fij) (see eqn 29)

Computation of the sound pressure level

Fig. 7.

General algorithm o f the proposed model.

taking place, 3 and it should be possible in the near future to take these effects into account. PARAMETRIC ANALYSIS Effect of turbulence and moderate gradients Figure 8 gives the attenuation due to the ground as a function o f the frequency for a source and a receiver over a finite ground impedance

126

A. L'Espbrance et al.

) o : I x l O-5

-20

-'il
< = -6

u~
N -I0
-20

z ~
~

o
-20 -30

) o =_ i= 10-5

10z

103
FREOUENCY (Hz }

104

Fig. 8. A t t e n u a t i o n versus frequency for a source and a receiver on a finite impedance g r o u n d (z s = ]-8 m; z~ = !'5 m, DsR = 1000 m and a = 3 10 s Rayls M KS). Curves a--c give the a t t e n u a t i o n for three sound speed gradients (a = + ! x ] 0 - ~, 0 and - ! x 10- 5 m - 1) and f o u r values o f fluctuating refractive index in each case ( ( p 2 ) = 0 , 1 x 10 -7, 5 ]0 -7 and

1 10-6).

(Zs= 1"8m zR= ].5m, DsR= 1000m and a = 3 x 102 Rayls MKS). The attenuations (sound pressure levels relative to free field) are given for three sound speed gradients, a = + l x l 0 - S m - 1 , a=0 and a = - l x 10- 5 m - 1 (Fig. 8(a)-(c)), and for four values of fluctuating refractive index ( p 2 ) (0, 1 x 10 -7, 5 x 10 -~ and 1 x 10 - 6 ) for each of these sound speed gradients. In every case, the outer scale of the turbulence L has been fixed to 1 m. These cases correspond to the cases of moderate sound speed gradients studied by Hidaka et al., 9 but also include the effects of turbulence. Under a moderate positive sound speed gradient (a = 1 x 10- 5), the ground attenuation is reduced because of the increase of the reflection angle. On the other hand, the time difference between the direct and the reflected rays shifts the geometrical interference pattern towards low frequencies. For a moderate negative gradient, the opposite phenomena are observed. When the effect of turbulence is included, the coherence between the direct and the reflected rays is reduced. Thus for each case of gradient (positive, neutral and negative), the ground attenuation which is due to a phase shift is reduced as the fluctuating refractive index increases. At high frequencies, the amplitude of the attenuation due to geometrical interference patterns between the direct and the reflected rays is also reduced. For large values of the fluctuating refractive index and/or large distances, the ground wave

Heuristic mode/for outdoor sound propagation

127

attenuation and the geometrical interference patterns will disappear. The resulting attenuation will be a constant + 3 dB, corresponding to the quadratic sum of the direct and reflected rays.

Appearance of additional rays


Figure 9 gives the attenuation at 1000 Hz as a function of distance for strong positive gradient, a = 1 x 1 0 - 4 m -1, for a source and a receiver on hard ground (Fig. 9(a), with tr = 1 x l0 s Rayls MKS) and on ground of finite impedance (Fig. 9(b), with tr = 3 x 105 Rayls MKS). As before, the source and receiver heights are 1.8m and 1.5m, respectively. The solid curve represents the calculation with the geometrical ray model for which the fluctuating refractive index (#2> = 1 x 106 is used. The dashed curve is obtained by the fast field program for the same sound speed profile, in the case of a non-fluctuating refractive index. For this configuration, the first two additional rays in the geometrical ray model appear at about 450 m, causing a sudden increase in the sound level (solid curve). As the distance increases, further additional rays appear, resulting in increased sound pressure level. The discontinuities caused by the

10 0 -10 -20 "

. . . . .

10

11111

1000

-10

b) 0"= 3 x 10s Rayls m.k.s.


-2010 100 1000

RANGE (m)
Fig, 9. Attenuation calculated with the last field program method ( - - - - ) a n d with the geometrical ray approach ( ): (a) for hard ground (a = l x 10 a Rayls (MKS); (b) for soft ground (a = 3 x l0 s Rayls MKS).

128

A. L'Espbrance et al.

sudden a p p e a r a n c e o f these additional rays result in some differences between the two curves. At distances b e y o n d 1 k m the appearance at a b o u t 450 m o f yet other rays causes additional differences which can restrict the validity o f the proposed model for very large distances. The a g r e e m e n t between the dashed a n d solid curves also suggests t h a t for this case the fluctuating refractive index can be neglected.

Beginning of the shadow zone


Figures 10(a) and (b) give for h a r d a n d finite i m p e d a n c e g r o u n d respectively (a = 1 x l0 s Rayls M K S and tr = 3 x 105 Rayls MKS), the a t t e n u a t i o n as a function o f distance for a negative gradient (a = - 1 x 10- 5 m - 1), a n d the c o r r e s p o n d i n g fluctuating refractive index used is ( # 2 ) = 5 10-7. In b o t h cases, the a t t e n u a t i o n is presented for four frequencies, f = 100, 500, 1000 and 2000 Hz. Once again, the source a n d receiver heights are 1.8 m a n d 1"5 m respectively. F o r these values, the s h a d o w zone is detected n e a r 1100m. B e y o n d that distance, the p r o g r a m uses the diffraction solution, eqn (32), instead o f the geometrical ray solution. There exist small discontinuities between the geometrical ray solution a n d the diffraction solution in the case for h a r d ground. These discontinuities seem to be more i m p o r t a n t in the case o f finite g r o u n d impedance. These differences are possibly due to the geometrical acoustic a p p r o x i m a t i o n ,
IO

P ~

-" - : - ~ - -

~looo

5 0 0 Hz

H,

2 0 0 0 Hz

,oF-

'~ - 3 0 -40 -50 i0 z i = a n inlll 103 RANGE (m) i S O ~) ...... 104

Fig. 10. Attenuation versus distance at 1130,500, 1000 and 2000 ttz in presence of a negative sound speed gradient a = - 1 x 10- s m- l (zs = 1"8m, za = 1"5m). (a) a = l x 10s Rayls MKS; (b) a = 3 x 105 Rayls MKS. Near 1100m, the receiver reaches the shadow zone and then the diffraction solution is used. The dashed curves ( - - - - ) show the results given by the diffraction solution in the illuminated zone.

Heuristic model for outdoor sound propagation

129

which may not be appropriate or valid when the receiver moves nearer the shadow zone. The importance of these discontinuities will be evaluated in the following sections, where theoretical and experimental results are compared. We note that the geometrical configuration used in these figures corresponds to the one used by Parkin and Scholes in their experiments. 3x Therefore, in the case of a negative gradient, a = - 1 x 10- 5 m - 1 (Fig. 10(b)), discrepancies between experimental and theoretical results could be expected at high frequencies from 400 m, as will be observed in the section on comparison with the results of Parkin and Scholes. However, as mentioned by Berry and Daigle, 22 it could be possible to minimize these discontinuities by using the diffraction solution even before the shadow zone. The results obtained in this manner are illustrated by the dashed curves in Fig. 10(a) and (b). This palliative solution is, however, limited by the computation time of the actual diffraction solution, which increases rapidly when the receiver is in the illuminated zone and moves away from the shadow zone. Moreover, this diffraction solution stops converging when the receiver is too far from the shadow zone (in Fig. 10(b), the beginnings of the broken lines are the first converging points).

COMPARISON WITH OTHER THEORIES AND EXPERIMENTAL DATA Comparison with the results of Raspet
Recently, White and Gilbert 7 have compared the parabolic equation with the experimental data obtained by Raspet et al. 32 In the latter experiment, the source and receiver heights were 30 m and 1-5 m respectively, and the ground impedance could be roughly estimated with t r = 3 x 105 Rays MKS. Figures 11 and 12 compare some of the results presented by White and Gilbert with those obtained using the proposed geometrical ray model. These figures give the attenuation at 40 Hz and 630 Hz respectively, as a function of the distance for five different sound speed gradients: a = - 2 x 10 -4, - 1 x 10 -4, 0, 1 x 10 -4 and 2 x 1 0 - 4 m -x. These values have been determined by best fit to the experimental curves presented by White and Gilbert. 7 Figures 13(a) and (b) compare these estimated linear sound speed profiles with the logarithmic gradient used by White and Gilbert. 7 It is interesting to note that the slopes of the two sound speed profiles are similar at 15 m, which is approximately the mean height between the source and the receiver (15 m). Thus, as a first approximation, we may suggest that the linear sound speed profile a can be evaluated according to the slope of the

130

A. L'Espbranceet al.

oy
~

o:
-----

+
_
=o

+
,

---\1

-,o[-,,)o,-,,,o;"

-I0 I,-

~C) O =

O"

I
IOI

I
+

t-I0 H 0 )o'2 10- 4 I 400

"-.*---~I 800 RANGE (m)

---';I

4.1 1200

I 1600

Fig. I 1. Attenuation versus distance at 40 Hz for five values of sound speed gradient: a = - 2 x 1 0 -4 , - l x 10-4 , 0, + 1 x 1 0 -4 and + 2 x 10-4m -1. +, and + represent maximum, mean and minimum experimental results of White and Gilbert;26 parabolic equation method; . . . . , geometrical ray model. logarithmic profile at the mean height between the source and the receiver when the sound speed profile is known or defined by a logarithmic profile:

a~C[ 6z [(=s+=.)/2

(36)

However, when studying the attenuation as a function o f distance from the source, because the rays are bent upward (or downward), the linear sound speed gradient that we need to consider for a receiver near the source may be greater (or smaller) than that used for a receiver far from the source. Nevertheless, the following comparisons will show that, as a first approximation and for propagation distance o f less than 1 kin, the use o f a constant linear sound speed profile over the propagation range studied appears to be acceptable. Figures 11 and 12 show that the results obtained with the geometrical ray model (dashed curves) are in good agreement with the experimental data and

Heuristic model for outdoor sound propagation


20 '

131

-20 -4( -6( "'x.

-So-l'-2xlO
-I00
I

"4
I

"~/

~
I r~.

i orarv~:-.~-J_ __~ +
f-C +, "

-k ~'~-'''~'~ -SO/ I

"~,-~1
i

~ -2op~.o
bJ

-20

i
--i--

: ....

I ..... I
-.....

-1=

-2

400

I 800
RANGE (m)

i 1200

1600

Fig. 12. As Fig. 11 except for 630 H z . - . . . . , calculation without turbulence (i.e. (#2) = 0).

80 70
6O
E

I-Im

50 40

/
/,~'-- ct, )~cotr,,i,,io % )
SOUND SPEED C(m s "I)

8O

3O

zo

20

2xlO -4 z

I0 'jk~'--C(z) Co+0.4 In ( z ) f 5 0 I I I I L 340 341 342 3 4 3 3 4 4

,o/./--c,,,.~o.OS
0 ~.,-"I.. I I 341 342 343 I =

=,,,

SOUND SPEED C ( m I "1]

Fig. 13. Comparison between the logarithmic sound speed gradient used by White and Gilbert and the estimated linear sound speed profiles used in the geometrical ray model.

132

A. L'Espbrance et al.

with the parabolic equation (unbroken line). In the case of the strongest positive gradient, a = 2 x 10 -4, we can observe a rise in sound levels near 1600m, especially in Fig. ll(e), indicating the appearance of new rays. However, the appearance of these new rays is sudden for the case of the geometrical ray solution, compared with the progressive increase in sound level for the parabolic equation. This difference is due to the high-frequency approximation ofthe geometrical ray theory which makes the appearance of new rays sudden, and thus without effect on the receiver below the distance at which these rays appear (see Fig. 9). This point indicates a limitation of the model, and theoretical developments are in progress to introduce a transition solution. In the case of negative sound speed gradients ( a = - 2 x 10 -4 and a = - 1 x 10- 4), the shadow zone is detected at 600 m and 900 m respectively (see Figs 1 l(a), (b) and 12(a), (b)). Beyond these distances, the receiver is in the shadow zone and the diffraction solution replaces the geometrical solution. In Figs 12(a) and (b) the results show the geometrical solution with and without the effects of turbulence (dashed curve and d a s h - d o t - d a s h curves respectively). The values of the fluctuating refractive index used are (#2) = 1 x 10 - 6 and ( # 2 ) = 2 x 10 - 6 for la[ = 1 x 10 -4 and lal = 2 x 10 -4 respectively. As discussed in the section on the beginning of the shadow zone, there exist some discontinuities between the geometrical and the diffraction solution, especially at medium and high frequencies in the presence of a finite impedance ground, but in Figs ll(a), ll(b), 12(a) and 12(b), the discontinuities are more related to the fact that the diffraction solution actually used in the model does not include the effect of turbulence. As discussed by White and Gilbert, 7 turbulence may scatter acoustical energy from the illuminated zone to the shadow zone and thus increase the sound levels in the shadow, a phenomenon which is currently under study. 3 The previous comparisons show that, in general, the parabolic equation and the geometrical ray model are in reasonable agreement and could provide good estimations of the sound pressure level under wind speed gradient, except far into the shadow zone when scattering effects are present. Compared with the parabolic equation solution, the model proposed still has some discontinuities when new rays appear but, because it requires much shorter computation times (a few seconds on a personal computer for a set of receiver positions or for a full frequency range), this model appears to be a useful tool for engineering purposes.

Comparison with the experimental data of Hawkins


In 1987, Hawkins 33 compared the sound propagation generated by a wind turbine at long distance (up to 10km) with a ray-tracing program. The

Heuristic mode~for outdoor sound propagation

133

experimental data used in that study had been obtained by Willshire. 34 The estimated source and receiver heights were 40 m and 0"5 m respectively, and the ground impedance was evaluated with a = 106 Rayls MKS. Figure 14(a) gives the propagation loss as calculated by the ray-tracing model used by Hawkins (continuous line) and by the geometrical ray model (dashed line). The points represent the experimental data and the bold dashed line represents the free field propagation loss. For this case, the estimated linear sound speed gradient (evaluated by best fit to the experimental data) is a = 1 x 10 -4 m - ~, which is in agreement with the relatively high wind speed considered: v = + 16 m s - ~ at 40 m. The fluctuating refractive index used for this high wind was (/z 2) = 5 x 1 0 - 6 . Both theory and measured results show that up to 2 km the propagation loss is almost spherical until new rays appear. When the additional rays appear the sound levels suddenly increase, which is in good agreement with
-40:
m DRUN6,

-50'
O,
= o

IOHz i0"4

-60:
-70

~
n~0.

-go

,,

,ll

,-,-,";"

I0

15

20

25

R A N G E ( km )

(a)
-40
m 'o U') U) 0 --I Z p. ,< -50

=-

IxlO-4

_o

-60
-70

"\ \

-80102

1 I I I [

I~

I I

iO s

104

RANGE ( m )

(b)
Fig. 14. (a) Propagation loss versus distance o f the 1 0 H z generated by a wind turbine (Zs = 40 m, za = 0'5 m) under a positive sound speed gradient o f a = 1 1 0 - 4 m - 1 . - - , Ray-tracing model; . . . . -, geometrical ray model; . . . . -, free field propagation loss; e , experimental results. (b) A s (a), but for the corresponding upwind condition a = - 1 x 10 - 4 m - 1 . , Geometrical ray model; . . . . . , free field attenuation; S , experimental results.

134

A. L'Espbrance et al.

the experimental results. When further additional rays appear, the propagation loss predicted by both models becomes almost cylindrical. 35 Hawkins also presented experimental data for the corresponding upwind conditions but he did not show theoretical results for the upwind situation because the ray-tracing model used could not predict the sound level in the shadow zone. In Fig. 14(b), these experimental data are compared with the results given by the model proposed for the corresponding negative gradient a = - 1 x 10 -4 m - 1. For this upwind situation, the shadow zone is detected at about 900 m, and the results predicted by the diffraction solution agree reasonably well with the experimental values beyond the shadow zone. It should be noted that at these distances, and up to 2 km, the effect of turbulence appears to be negligible for the frequency considered (10 Hz).
C o m p a r i s o n with the results o f P a r k i n and S c h o l e s

Although these general tendencies are well simulated by the geometrical ray calculation, there exist some discrepancies between the theoretical and the experimental results. These have been investigated in relation to the noise generated by a jet engine over grass-covered ground in various meteorological conditions. These measurements thus included the effect of the ground, the effect of the wind and temperature gradient and the effect of turbulence. These measurements were made in a third-octave band and were presented in terms of excess attenuation relative to the level at 19.5m. These measurements have been extensively reported in the past, mainly in analyses of the ground effects in neural atmospheric conditions. 12 In a few cases, the additional effects of turbulence or refraction were studied, but not all three simultaneously. It is interesting to make some comparisons of these acoustical measurements, using the general atmospheric conditions reported during these measurements, with the model proposed. Figures 15(a)-(c) give the experimental excess attenuation as a function of frequency for seven receiver distances (31.2, 62.5, 110, 195, 315, 620 and 1100m) and for three different wind conditions as classified by Parkin and Scholes.31 Figures 15(a), (b) and (c) are respectively for a positive vector wind (v> 5 m s -1 at 10m above ground level), a zero vector wind ( S m s -1 > v > - 5 m s- 1), and a negative vector wind (v < - 5 m s- 1) in a so-called 'neutral' temperature condition. F r o m Fig. 2 of Ref. 31, however, it is possible to estimate that the effective mean wind speeds (at 10-m height) corresponding to these three classes of wind vector are v = 7-0 m s- 1, v = 1.5 m s - x a n d o = - 5 - 0 m s-1. For the simulation, the following sound speed gradients were used: a = + l ' 5 x l 0 -5, a = l x l 0 -6 and a = - l x l 0 - S m - l , and the corresponding fluctuating refractive index values used were (#2) = 7.5 x 10 -7, ( / a 2 ) = 1 x 10 -7 and ( # 2 ) = 5 x 10 -7 respectively. It should be

Heuristic model for outdoor sound propagation


EXPERIMENTAL RESULTS THEORETICAL RESULTS

135

IC 0

-10 - ' \ - ~ 7 ~ F ' 1 0 ) ~ ~ T.O m $" in


E
-2o

-IC -2C

15 5

~.,

_~
-~
o i m

I0

io!
0

e) o , l x l ( ~ s ,.~.,~]

'./

,o

-I0

I0

\
z

_o -20 I-<[ =)

~1

b)-v~l.sms"

-20

'\...;~.~.~," ~,'/ ~ /

.. .,,~-

.,/',j"

<,=>.l=lO "T

~-~
I---

-30~

~ 0
o x

"' -I0
-20

...\ ~., ...... .:..~ .,

-IC

,.- ,, / /

\,

\
-20
_

:
,

.
, . ,.,,,] i , 1,1=1

\'"--..X.../..."
,
~

;;o. = o-'
* /-.

<j,=>=sxio-7
,,

" -3C

-30 IOO

,~ho~., ,.,o~/v

in th'e"""

IOOO

IOOOO IOO

FREQUENCY ( HZ )

IOOO FREQUENCY (Hz)

IO OOO

Fig. 15. Experimental and theoretical excess attenuation as a function of frequency for seven receiver distances (3]-2, 62.5, ] 10, 195, 315, 620 and 110Ore) for a positive vector wind ((a) and (d)), for a zero vector wind ((b) and (e)), and for a negative vector wind ((c) and (f)). For the theoretical results (d)--(f), the sound speed gradients were respectively a = + 1.5 x 10-s a = ] x 10 -6 and a = - 1 x lO-Sm -1 and the fluctuating index of refraction </a2) = 7"5 x lO -~, ( / z 2 ) = l x 10 -~ and ( / z ' ) = S x 10 -~.

noted that these values were estimated by best fit according to the experimental measurements. In all cases, the overall ground impedance was estimated with a flow resistivity of 3 x 105 Rayls MKS. Figure 15(d)--(f) gives the theoretical attenuation obtained in the three cases. In the absence of a sound speed gradient (Fig. 15(b) and (e)), the ground effect dip is centered near 500Hz and increases with distance, which corresponds to the results published elsewhere. 1'2 It should be noted, however, that the agreement between theoretical and experimental results is better when turbulence is considered. In fact, the effect of turbulence (even for a low turbulence index, (/z 2) = 1 x 10-7, associated with the low wind speed considered) is to reduce the ground wave attenuation at large distances (620 and 1100 m) for frequencies higher than 500 Hz by decreasing the coherence between the direct and the reflected rays. This reduction of the

136

A. L'Espbrance et al.

ground effect at high frequencies and large distances creates the double dip at 300 and 800 Hz. For a positive sound speed gradient (Fig. 15(a)), experimental results show that the medium- and high-frequency attenuations are reduced and, inversely, for a negative sound speed gradient, these attenuations are increased. Globally, these comparisons are very illuminating. They show that increase and decrease of the angle of reflection at the ground and change in travel time between the direct and the reflected rays could provide a physical explanation for the decrease and increase of the experimental attenuations measured by Parkin and Scholes. Although these general tendencies are well simulated by the geometrical ray calculation, there exist some discrepancies between the theoretical and experimental results. For example, in the case of the negative sound speed gradient, the experimental results show lower sound levels than the theoretical values at 315m and 620m. As shown in the section on the beginning of the shadow zone, these differences may occur because, even at these distances, the geometrical solution may be inappropriate when approaching the shadow zone. On the other hand, at 1100 m the receiver is in the shadow zone and the model uses the diffraction solution (see Fig. 10(b)), and the resulting theoretical attenuation appears to be larger than the experimental values. One could expect an underestimation of the experimental attenuation at high frequencies as a result of background noise, as these attenuations are large and almost the same for 315, 620 and 1100m. Also, some scattering effects from the illuminated to the shadow zone at 1100 m may have been present during the experimental measurements. One could also hypothesize that higher sound speed gradients and/or turbulence index should be used at these distances. The meteorological data recorded during these measurements are, however, insufficient to clarify these points, and detailed comparisons between the theoretical and the experimental results appear to be speculative. Thus, only the general tendencies should be considered in these results, and these show that the model is useful in explaining the various attenuation curves measured by Parkin and Scholes 20 years ago.

DISCUSSION The objective of the work presented in this paper was to develop a model that could predict, in a short computation time, the outdoor sound propagation in general atmospheric conditions. This model is based on the assumptions of geometrical acoustics and combines analytical solutions developed earlier for the characterization of the principal phenomena

Heuristic model for outdoor sound propagation

137

involved in outdoor sound propagation: the ground effect, the atmospheric absorption, atmospheric refraction and turbulence. To obtain a simple and useful model, some restrictions have been imposed, such that each of the phenomena involved is represented by a single parameter: a for the ground impedance, AT for the atmospheric absorption, a for the atmospheric refraction and (p2) for the atmospheric turbulence. The resulting heuristic model obtained assumes a constant linear sound speed gradient, which allows an analytical determination of all ray paths (and associated quantities) that exist between a source and a receiver. The total sound pressure at the receiver is computed by summing the contribution of each ray. When the receiver is in the shadow zone, the sound pressure is evaluated with a diffraction solution. Comparisons with other models and with experimental results have shown that the proposed model can give a good estimate of the sound pressure levels generated by a source in general meteorological conditions. Moreover, the model has the advantages of providing physical insight despite the numerous interacting mechanisms. Thus, it appears to be an interesting tool for engineering purposes. Compared with other models, the main advantages of the geometrical ray model are that it is easily implemented on a personal computer and that it requires little computation time. For a given source and receiver position, computation time is typically a few seconds for a frequency range of 10010000Hz in the case of a moderate sound speed gradient. For a strong positive gradient (when additional rays appear) or a strong negative gradient (when the receiver is in the shadow zone), less than I min is needed in most cases. This model is, however, restricted to the case o f a constant linear sound speed gradient, which is not strictly justified and may not be appropriate or even acceptable in some cases. However, the comparisons between experimental and theoretical results have shown that the use of a constant linear sound speed profile over the entire propagation range appears to be acceptable as a first approximation for propagation distances of up to 1 km. A preliminary procedure is provided to find a realistic gradient by using a linear sound speed profile. More experimental results and theoretical development will be necessary, however, to improve the model, mainly: (1) to establish the effective values of the linear sound speed gradient and of the refractive turbulence index to be used, according to the meteorological conditions considered; (2) to evaluate the acoustical scattering expected from the illuminated to the shadow zone in the presence of turbulence; (3) to introduce a transition solution when new rays appear; and (4) to correct for phase changes for propagation through caustics.

138

A. L'Espbrance et al.

A C K N O W L E D G E M ENTS The authors are grateful to Hydro-Qu6bec and the C.R.S.N.G. for their financial support.

REFERENCES 1. Piercy, J. E., Embleton, T. F. W. & Sutherland, L. C., Review of noise propagation in the atmosphere. J. Acoust. Soc. Am., 61(6) (1977) 1403-18. 2, Attenborough, K., Review of ground effects on outdoor sound propagation from continuous broadband sources. Applied Acoustics, 24 (1988) 289-319. 3. Daigle, G. A., Embleton, T. F. W. & Piercy, J. E., Propagation of noise in presence of gradient and turbulence near the ground. J. Acoust. Soc. Am., 73(3) (1986) 613-27. 4, Johnson, M. A., Raspet, R. & Bobak, M. T., A turbulence model for sound propagation from an elevated source above the ground. J. Acoust. Soc. Am., 81(3) (1987) 638-46. 5. Hallberg, B., Larsson, C. & Israelsson, S., Numerical ray-tracing in the atmospheric surface layer. J. Acoust. Soc. Am., 83(6) (1988) 2059-68. 6. Lee, S. W., Bong, N., Richards, W. F. & Rasper, R., Impedance formulation of the fast-field program for acoustic wave propagation in the atmosphere. J. Acoust. Soc. Am., 79 (1986) 628-34. 7. White, M. J. & Gilbert, K. E., Application of the parabolic equation to the outdoor propagation of sound. Applied Acoustics, 27 (1989) 227-38. 8. Harrison, C. H., Ocean propagation models. Applied Acoustics, 27 (1989) 163-201. 9. Hidaka, T., Kageyama, K. & Masuda, S., Sound propagation in the rest atmosphere with linear sound velocity profile. J. Acoust. Soc. Jpn (E), 6(2) (1985) 117-25. 10. Delany, M. E. & Bazley, E. N., Acoustical properties of fibrous absorbent materials..Applied Acoustics, 3 (1970) 105-16. 11. Attenborough, K., Acoustical impedance models for outdoor ground surfaces. J. Sound Vib., 99(4) (1985) 521--44. 12. Embleton, T. F. W., Piercy, J. E. & Daigle, G. A., Effective flow resistivity of ground surface determined by acoustical measurements. J. A coust. Soc. Am., 74 (1983) 1239-44. ! 3. Chien, C. F. & Soroka, W. W., A note on the calculation of sound propagation along an impedance surface. J. Sound Vib., 69 (1980) 340--3. 14. Nobile, M. A. & Hayek, S. I., Acoustic propagation over an impedance plane. J. Acoust. Soc. Am., 78(4) (1985) 1325-36. 15. Nicolas J. & Berry, J. L., Propagation du son et effet de sol. Rev. Acoust., 71(4) (1984 191-200. 16. Acoustical Society of America, Method for the calculation of the absorption of sound by atmosphere, ANSI S 1.26. Acoustical Society of America, New York, 1978. ! 7. Hallberg, B., Larsson, C. & Israelsson, S., Some aspects on sound propagation outdoors. Acoustica, 66 (1988) 109-12.

Heuristic modelfor outdoor sound propagation

139

18. Brekhovskikh, L. M., Waves in Layered Media. Academic Press, New York, 1960. 19. Embleton, T. F. W., Analogies between non-fiat ground and non-uniform meteorological profiles in outdoor sound propagation. J. Acoust. Soc. Am., Suppl. 1, 78 (1985) s86. 20. Li, K. M., Attenborough, K. & Heap, N. W., Source height determination by ground effect inversion in the presence of a sound velocity gradient. J. Sound Vib., 145(1) (1991) 111-28. 21. Embleton, T. F. W., Thiessen, G. J. & Piercy, J. E., Propagation in an inversion and reflections at the ground. J. Acoust. Soc. Am., 59(2) (1976) 278-82. 22. Berry, A. & Daigle, G. A., Controlled experiments on the diffraction of sound by a curved surface. J. Acoust. Soc. Am., 83(6) (1988) 2059-68. 23. Chernov, L. A., Wave Propagation in a Random Medium. McGraw-Hill, New York, 1960. 24. Tatarski, V. I., Wave Propagation in a Turbulent Medium. McGraw-Hill, New York, 1961. 25. Daigle, G. A., Effects of atmospheric turbulence on the interference of sound waves above a finite impedance boundary. J. A coust. Soc. Am., 65(1) (1979) 45-9. 26. Daigle, G. A., Piercy, J. E. & Embleton, T. F. W., Effects of atmospheric turbulence on the interference of sound waves near a hard boundary. J. Acoust. Soc. Am., 64(2) (1978) 622=30. 27. Clifford, S. F. & Lataitis, R. J., Turbulence effects on acoustic wave propagation over smooth surface. J. Acoust. Soc. Am., 73(5) (1983) 1545-50. 28. Anderson, G., Gocht, R. & Sirota, D., Spreading loss of sound in an inhomogeneous medium. J. Acoust. Soc. Am., 36(1) (1964) 140-5. 29. Jacobson, M. J., Analysis of spreading loss for refracted/reflected rays in constant-velocity-gradient media. J. Acoust. Soc. Am., 36(12) (1964) 2298-305. 30. Gilbert, K. E., Raspet, R. & Di, X., Calculation of turbulence effects in an upward-refracting atmosphere. J. Acoust. Soc. Am., 87(6) (1990) 2428-37. 31. Parkin, P. H. & Scholes, W. E., The horizontal propagation of sound from a jet engine close to the ground at Hatfield. J. Sound Vib., 2(4) (1975) 353-74. 32. Raspet, R., Schomer, P. D. & Hottman, S. D., The measurement of sound propagation from an elevated source. J. Acoust. Soc. Am., Suppl. 1, 73 (1983) s50. 33. Hawkins, J. A., Application of ray theory to propagation oflow frequency noise from wind turbines. NASA CR-178367, 1987, 105 pp. 34. Willshire, W. L., Jr., Long-range downwind propagation of low-frequency sound. NASA, TM 86409, 1985. 35. Willshire, W. L. & Zorumski, W. E., Low-frequency acoustic propagation in high winds. Proceeding Noise-Con 87, ed. J. Tichy. & S. Hayek. The Pennsylvania State University, State College, Pennsylvania, 1987, pp. 275-88.

S-ar putea să vă placă și