You are on page 1of 7

Two-stream irradiance model for deep waters

Eyvind Aas

The two-stream model expresses the vertical attenuation coefficientK and the irradiance ratio R as functions
of the absorption coefficient a, the backward scattering coefficient bb, the downward and upward average
cosines ld and ,t,, and the normalized reflectance coefficients of downward and upward scalar irradiance, rd
and ru. While K/a andR are almost linear functions of bb/a when bb/a is small, they will approach asymptotic
values, which only depend on rd, r,, I'd, and ,,, when bb/a becomes large. The results agree well with oceanic
observations of K and R. They also agree with theoretical results derived by other methods. Still proper
testing of the model in turbid waters remains.

1. General Form of the Two-Stream Equations


The classical equation of transfer may be written
d J L(z,0,0) cosOdQ

dL cosO =cL
dz
+L (1) = d y L cosOdQ - LlcosO)dQ

where z is the vertical coordinate, positive downward,


and L = L(z,O,,b)is the monochromatic radiance at the = d (Ed - Ej)
depth z from the direction 0, 0. is the zenith distance
and k the azimuth angle. c is the attenuation coeffi-
cient, and L* may be termed the source function or
path function defined by
= -c J LdQ + J L~dQ
L.(z,O,0) = J (2)
= -cE 0 + J.[JL(z 0 , )(0,0,0 ,O )dQ' dQ (4)

d ' is an infinitesimal solid angle in the direction O',O'.


fl(OkO',k') is the scattering function for light scattered Here the index d denotes downward and u upward,
with direction 0, 0 from the direction 0', A'. The inte- so that 27rdmeans the hemisphere of downward flux,
gral of #dQ' over the forward hemisphere around the that is, the upper hemisphere. Ed and E&mean the
direction 0, 0 is the forward scattering coefficient bf, downward and upward irradiances, respectively. E is
while the integral over the backward hemisphere is the the scalar irradiance defined as
backward scattering coefficient bb. The integral of
#d ' for all directions is the total scattering coefficient
E0 = J LdQ'. (5)

b,so that By changing the sequence of integrations in the last


b = bf + bb. (3) double integral, we find that it simply becomes bEO.
The scalar irradiance can be divided into its contribu-
We let dQ denote a solid angle in the direction 0, 0. tions from the upper and lower hemispheres, Eod and
If we integrate Eq. (1) for all directions, we obtain Eon, respectively, so that
E = Eod + EO.- (6)

Equation (4) can now, since


c = a + b, (7)
were a is the absorption coefficient, be written
The author is with University of Oslo, Institute of Geophysics,
Post Office Box 1022 Blindern, 0315 Oslo 3, Norway. -d (Ed - En) = -cE 0 + bEO=-a(EOd + E0.). (8)
Received 29 October 1986.
0003-6935/87/112095-07$02.00/0. This is the classical Gershun equation for a vertical-
© 1987 Optical Society of America. ly stratified medium.

1 June 1987 / Vol. 26, No. 11 / APPLIEDOPTICS 2095


When the integration of Eq. (1) is limited to the Cd= ad + bd,
upper hemisphere, the result becomes
a = a/IU,(2
df L cosOd = d Ed
dz 27rd dz bu = rubblflu, (20)

cu = au + bu,
=-c J LdQ + L.dQ Eqs. (12) and (13) obtain the forms
dEd
= -CdEd
+bEu, (22)
= -CEOd + fId [JS"d #3(O0',0'')dQ]L(zO',O')dQ'

dE,
= -cuEu + bdEd- (23)
+AJ 2. [J, #(,0,0,0 )dQIL(zO
,O)dQ. (9)
These are two-stream equations, provided the coef-
The last two integrals will depend on and the ficients bd, b, Cd and c can be regarded as constants.
angular radiance distribution. If we define the quan- The same equations were obtained in a different way

bbI
dI
d
tities rd and ru by by Kozlyaninov and Pelevin,3 but the coefficients
d Er = fl[(0,so0
,ep)dRIL(z,O',w')d2' there had less precise definitions. A more primitive
1d f(rd [f.. form was applied by Joseph.4 The two-stream model
in a hydrologic context has also been discussed by
Preisendorfer,5 Prieur,6 and Preisendorfer and Mob-
1b [4 2 O , 0,00,s)dQ]L(z,0,O
fbE~dJ2,r[ )dQ'd
', (10) ley.7 Adaptions of two-stream equations to atmo-
spheric conditions are presented by, e.g., Chandrasek-
=bb o har, 8 Kondratyev, 9
and Sobolev.10' 11
r. = ubbE2J,J1(000'X')d
"'d#
2 L(z,0,0')WQ, (1 It is sometimes convenient to separate the contribu-
tion to the scattering coefficient due to particles bp
Eq. (9) can be written from the contribution due to the pure water bwso that
dEd b = bp + b. (24)
= -cEod+ (b- rdbb)EOd
+ rUbbEoU- (12)
Similarly # can be divided into fp and #3,,and if rdpis
The quantity rd represents the mean upward scat- defined by Eq. (10), only with index p at # and b, and
tering coefficient of the downward traveling photons, rd. is defined in the same way, the relation between rd,
rdp and rdWbecomes
while ru represents the mean downward scattering co-
efficient of the upward traveling photons, both coeffi- rdbb = rd(bPb+ bwb) = rdpbpb + rdwbwb, (25)
cients normalized with the backward scattering coeffi-
cient. rd and ru may also be considered as normalized which can be written
reflectance coefficients of downward and upward sca-
lar irradiance, respectively. rdP(bpb/bWb)+ rdW
If the downward irradiance is dominated by the (bpb/bwb)+ 1
vertical radiance, upward and backward directions be- The actual value of rd will then obtain a value between
come identical for most of the photons, and rd becomes rdpand rdWdepending on the ratio bpb/bwb. r can be
close to 1. In the same way ru becomes close to 1 if the defined as a function of rup and r in the same way.
upward irradiance is dominated by the vertical radi- It should be noted that so far Eqs. (22) and (23) are
ance. If 3(O)is symmetric around 0 = 7r/2,the upward, exact. To proceed, however, we have to make some
downward, forward, and backward scattering coeffi- approximations. The first will be that since the varia-
cients become equal, and rd and ru become 1. tion of Ed and Eu along the z axis is likely to be much
Subtraction of Eq. (12) from Eq. (8) gives greater than the variation of rd r, ,ud, andf g,, we can
dEu for a layer of water with constant a and bb regard bd, bu,
- -(a + rbb)Eou + rdbbEOd- (13) Cd, and c as constant too.
Elimination of Eu in Eq. (22) by means of Eq. (23)
By introducing the downward and upward average now gives
cosines,2 , d and ,ii defined as
d2 Ed dEd
/Id = Ed/EOd, (14) dz2 = (CU- cd) - + (CdCu- bdbu)Ed. (27)

(15) The general solution of this differential equation,


which requires that two boundary conditions are
and by writing known, can easily be found in textbooks. Eu is then
ad = ad, (16) uniquely determined by Eq. (22).
The advantage of two-stream models compared to
bd = rdbb/id, (17) single scattering models is that they contain the effects

2096 APPLIEDOPTICS / Vol. 26, No. 11 / 1 June 1987


of multiple scattering. The disadvantage is that they apply the definitions (14), (15), and (30), the result
cannot predict the change in the shape of the radiance becomes
distribution with depth and thus not the change in rd, K(1 - R) = a[(l/gd) + (R//i,)]. (32)
ru, ftd and , The solution of Eq. (27) should then
only be used in depth ranges where the relative shape Contrary to Eq. (31), this relation does not involve
of the radiance distribution can be regarded as fairly directly the backscattering coefficient bb.
constant, for example, in either a thin homogeneous
surface layer or the deep layers where the relative 11. Variation of R and K
radiance distribution is close to its asymptotic shape. If rd,ru, , d, and fu>remain fairly constant while a and
Preisendorfer and Mobley7 divide the entire water bb vary, the ratio bd/bu = rdfu /rutc d of Eq. (31) will be a
column into a series of horizontal slabs, where Cd, cu, bd, constant. R will then decrease monotonically as (Cd +
and bu of each slab can be assumed constant. The cu)/2bu increases, that is, when a/bb increases or bb/a
solutions of Eqs. (22) and (23) for each slab are nested decreases. The smallest and largest values of R will be
together, and Ed (z) and Eu(z) are obtained as functions obtained where bb/a has its smallest and largest values,
of Cd, cu, bd, and bu,when Ed and Eu are known at the respectively.
upper and lower boundary, respectively. The authors Equation (31) may be written
are then able to solve the inverse problem, that is, to Cd + C 4bdbU, 11/2~
find c, bd, and bu and ultimately a(z) and bb(z) as
Cd, 2bU {
RI -= 1 __
[ (Cd + C")2 }
~~~~~~~
functions of the observed irradiances Ed(Z), Eu(z),
Eod(Z), and Eou(z). To do so they apply the assump- When c or a is much greater than bb, so that Cd >>bd and
tion rd = ru {whichin their notation reads b(y,-)/ Cu>>bu, the square root may be expanded in a series:
D(y,-) = b(y,+)/[D(y,+)]}. However, as we shall see,
this assumption is not correct.
Cd+ Cu 2bdbu 1
The aim of the present work is to express Ed(z) and
2b, L (Cd + Cu)
2
J
Eu(z) as functions of a, b, rd, ru, , d, and ft u in the surface
bd bd
layer of deep waters and to discuss the characteristic
Cd + Cu +(
features of this relationship.
Cd)

II. Solution for the Upper Layer of Deep Waters


rd bb/a
Provided that the waters are so deep that reflected
light from the bottom does not influence the irradiance + Ad 1 + rabb/a 1 + rdbb/a
of the upper layer, all irradiance measurements show Mu,1 + rdbb/a
that a first approximation of Ed should be of the form
rd bb/a
Ed(z) = Ed(O) exp(-Kz), (28) (34)
1 + Id/Ma 1 + rdbb/a
where K is a positive constant vertical attenuation There is an almost linear proportionality between R
coefficient of irradiance. This function contains the and the small values of bb/a. In fact, the expression
twoboundaryconditionsthat Ed isEd () at the surface above can be approximated further to
and zero at infinite depth. Comparison to the general
solution of Eq. (27) or insertion of Eq. (28) in (27) R ~ rd bb (35)
shows that 1+ Ad/t 1 a

K =[( ) bdbu] - C (29) A modified form of this relation has been applied earli-
2
er' 2 to explain how brown algae can color the ocean
green.
The irradiance ratio R is defined as the ratio be- If, on the other hand, bb/a is a large number, Cd will
tween upward and downward irradiance be close to bd and cu close to bu. Equation (31) then

Ed(Z)
R=
E1 1(z) gives that
~~~~~~~~(30) bd + b br -bd 2 1/2 bd rd/1
(36)
2b1 L 2bu bu ruAd
By inserting Eq. (28) in Eq. (22), we see that 1

Cd K Cd + , Cd + u2 bdl1/2 This is the largest possible value that R may obtain


R II 1-I(31)
_
b11 2b1 L 2b1 b11 according to our two-stream theory, no matter how
large bb/a becomes.
The four equations above give Ed(z) and E11 (z) as The significance of the expressions.for R to remote
functions of Ed(O) when a, bb, rd, ru, ft d, and ftu are sensing should be pointed out. If a concentration of
known. absorbing particles becomes so high that the ratio bb/a
Another useful relation between R and K can be - bbplap,a further increase in concentration will not
obtained from Eq. (8). Equations (30) and (31) show increase the ratio bb/a, and it is clear from Eq. (31) or
that Eu(z) is a function of z similar to Ed (Z), as given by (33) that R will remain unchanged provided rd, ru, f d,
Eq. (28). If we insert forEd(z) and Eu(z) in Eq. (8) and and ft u remain constant. A high concentration of non-

1 June 1987 / Vol. 26, No. 11 / APPLIEDOPTICS 2097


absorbing particles may result in a large value of the The last result resembles Eq. (39), except that the right
ratio bb/a, so that R becomes close to its asymptotic side contains an additional factor depending on the
value given by Eq. (36), and a concentration increment shapes of the scattering function and the radiance
will not change the value of R. Whether the particles distribution. Some examples of the value of this fac-
are absorbing or nonabsorbing, their concentrations tor, as well as the factor rdftulrud [Eq. (36)], are given
must be below certain limits if variations in the con- in Table I. It should be noted, however, that the
centration shall be detected by the remote sensor. applied radiance distributions do not represent condi-
It can be shown that K will always increase, accord- tions where bb/a >>1. How this influences the results
ing to Eq. (29), if either a or bb increases and the other we do not know.
factors remain constant. Equation (29) may be writ- While K is a linear function of bb in Eq. (38), that is
ten when bb/a <<1, K becomes independent of bb in Eq.
K= Cd + Cu 1- 4bbdu 1
/_ C Cd
(41), where bb/a >>1.
(37)
2 L (Cd Cu j 2
IV. Value of the Constants
When bb/a is sufficiently small, the square root may be The values of K and R are functions of a, bb, rd ru, f d,
expanded in series, and K becomes and IIu.
rd and ru depend on the shapes of the volume scatter-
bdbu_ b2b2
3
ing function and the radiance distribution L(0,(p).
d Cd + Cu (Cd + CU)
Whitlock et al.' 3 have presented values of : for very
turbid waters. It is guessed here that their station A2
- Cd= (a + rdbb)//Pd- (38) represents a situation where a relatively large number
The expression may be further approximated to of small particles influences the scattering.
An analytic approximation to their scattering func-
K a/Aid. (39) tion can be
The vertical attenuation coefficient of irradiance is A(0) 0.092 0 0 10°
primarily a function of the absorption coefficient and (42a)
b (1.00002 - cosO)0 7 1
the downward average cosine when the ratio bb/a is
small. () _ 0.0113
If, on the other hand, bb/a is a large number, so that b (1 - CosO)'
7 10 <0 90, (42b)
bd Cd and bu c, Eq. (29) gives K as the difference
between two almost equal numbers. To find the mag- -() = 0.0256 + 0.0099 cosO
nitude of this difference we shall write Eq. (29) as b

K=u bd1+2bd + buad + au (ad+ a)2 1/2 Cu-Cd - 0.0143 sinO 900 • 0 S 1800. (42c)
2 l b, - bd bu -bd (bu -bd )2 2
Bauer and Morel 4 have given a scattering function
(40) where the forward scattering is more pronounced, and
When (ad + au)/(bu - bd) <<1, that is bbla >>1, the this is here thought to be due to a dominance of large
square root may be expanded in series so that particles. Their function can be represented by
K bu-bd (bd + b)(ad + a) Cu- Cd aubd + adbu ,(0) 0.00328
0 • 0 • 100, (43a)
l2 2(bU-bd) J 2 bu bd b (1.0006- cosO)' 4

a 1 + rd/rU 0(0) 0.00224


(41) 100 0 • 900, (43b)
ad 1 - (rdU/rk 1 d) b (1.017 - cosO)'8

Table 1. Optical Constantsfor Different Radiance and Scattering Conditions


Radiance 1 + rdr 1
distribution Ed/L(°) E/L(1800) R Ad P PdlPu rd ru rdU/rUj* 1 (rd/rud)

L(180)dQu Molecules 1 1
Vertical d 2d dQu° 1 1 1 Smaller particles 1 1
radiance L(00 )dQd Larger particles 1 1
Clear Molecules 1 1 0.512 4.10
sky 0.0532 3.74 0.033 0.86 0.44 1.96 Smaller particles 1.28 2.51 0.261 2.04
Larger particles 1.80 6.68 0.128 1.46
Overcast Molecules 1 1 0.474 3.80
sky 1.37 5.29 0.022 0.76 0.36 2.11 Smaller particles 1.42 2.94 0.229 1.92
Larger particles 2.16 8.63 0.119 1.42
Constant L(180 0 ) Molecules 1 1
hemispherical ir L 0° 0.5 0.5 1 Smaller particles 2.28 2.20
radiance (0) Larger particles 5.50 5.29

2098 APPLIEDOPTICS / Vol. 26, No. 11 / 1 June 1987


(0) = 0.00629 + 0.00272 cos0
b
- 0.00411 sinG 90° 0 • 1800. (43c)
z0
For pure water the relative scattering function will z
be given by the modified Rayleigh scattering function 0
2
b() 0.062 + 0.052 cos 0 (44) z

according to Morel.'5 The applied values for A(0)are v-)

given in Fig. 1, where they are normalized at 90°. w


Integration of the functions gives bb/b = 0.059 for the Iw
0U
smaller particles, 0.011 for the larger particles, and 0.5
for the molecular scattering of pure water, respective- ix

ly.
From observations by Smith' 6 it has been guessed
that a radiance distribution in the surface layer of clear
waters with a clear sky and zenith sun might be some-
thing like 0' 30' 60° g0 120' 150' 1P'
SCATTERING ANGLE

L(0) = 1000 Fig. 1. Examples of relative volume scattering functions for parti-
0 0 5, (45a)
L(90') cles and pure water as functions of the scattering angle.

L(O) 6
50 0 <490, (45b)
L(900 ) (1 - cos0) 04

5
L(0) = 1 + 25(cos0)' 490 0 90°, (45c)
L(90 0 )

L(O) 1 - 0.5 cos0 900 0 1800. (45d)


L(90 0 ) 1 - 2.2 cos 9 w

Tyler' 7has measured radiance distribution with an 0


4U
overcast sky. The mean radiance for each zenith dis-
tance in the surface layer can be approximated by ix
W.
L(O) = 31
0 0 100, (46a)
L(90 0 )

L(=) 71 cos0 - 39 100 ' 0 • 490, (46b)


L(90')

0 490 0 < 90, (46c) , ml-


o'
0U 30' 60' 90' 120' 150' 180'
L( ) = exp(3 cos0)
ZENITH ANGLE
Fig. 2. Examples of relative radiance distributions for a clear and
L(O) 1 + 0.3 cos0 overcast sky as functions of the zenith angle.
(46d)
L(90 0 ) 1 - 3 cos 9

These two different radiance distributions are shown become <1, and r may become closer to rd0 = 1 than to
in Fig. 2. Two extreme distributions which do not rdp.
occur in nature are the cases of only vertical downward Calculations by Tyler et al.' 8 give that ftd1itu for
and upward radiance and of constant radiance within clean natural water should vary from 2 in the blue part
each hemisphere. The integrated properties of these of the spectrum to 3 in the red part. If we take the
distributions, as expressed by Ed, R, d, and ft, are value 2 from Table I for the ratio , dift u, we see that Eq.
presented in Table I. The table also gives the calculat- (34) now becomes with rd 1
ed values of rd and ru with the scattering functions
given by Eqs. (42)-(44). 1 bb/a (47)
It is seen that while rd 0 and rw are always 1, the 3 1 + bb/a
values of rdp and rp will obtain higher values. As Similarly the expression for K for clear ocean waters,
explained in connection with Eq. (26), the resulting which is given by Eq. (38), reduces to
values of rd and rudepend on the ratio bpb/bwb. Even in
the clearest ocean waters bp is 2-10 times greater than K bb
(48)
b0 , but since bib always is b0 /2, while bpb may be 1-5% X a d a
of bp,the ratio bpbb,,b may in large parts of the ocean A crude estimate of ft d with a high solar altitude and

1 June 1987 / Vol. 26, No. 11 / APPLIEDOPTICS 2099


clear waters may be cosO,, where 0s is the zenith dis-
tance of the refracted sunrays. eq - c -
Equation (32) can be written -- K/a

eq 29

a = I R)d (1 + 3 R)/gd. (49) Cl

This result' 9 may also be deduced from Eqs. (47) and a


0o
eq 47
_ - _ .--- ---- -R -t
(48).
When the water is more turbid, however, rd and ru 10I1
11 = _ t= |-eq-31
become closer to rdp and rp, and K and R should be
calculated from Eqs. (29) and (31). They may also be
calculated from Eqs. (38) and (35) if the condition bb/a
<< 1 still is satisfied. ft d will probably obtain values r2 ,,
IC . , _,L,, , 1..L
10
o 1 10 102

between 0.6 and 0.8.20 - bb/a


Whitlock et al.13 observed for very turbid waters Fig. 3. Relative vertical attenuation coefficientK/a and irradiance
that, although both b and bb decreased with increasing ratio R as functions of the ratio bb/aaccording to different equations.
wavelength, the ratio bb/b remained fairly constant.
Their tabulated values of 13(0)show that the shape of
the scattering function in their case was fairly indepen-
dent of the wavelength. While lid, Muand d/Mu vary
with wavelength in clear water,18 the variation in tur- This expression corresponds fairly well with Eq. (47)
bid natural waters is not known. provided bb = 0.022b. Since Jerlov2 0 gives that bb =
The difference between the general expressions (29) 0.02b for ocean surface water, Eq. (47) may give a
and (31) and the clear water expressions (47) and (48) reasonable estimate of the irradiance reflection from
may be illustrated by example. Let us assume that we clean natural water. Further support for Eq. (47)
have a suspension of quartz particles in pure water. comes from Gordon et al. 23 and Kirk2 42 5 who applied a
Quartz is nonabsorbing in the visible part of the spec- Monte Carlo technique and from Morel and Prieur2 6
trum, sothat the only absorption comes from the water who used a successive order scattering method and by
itself. With increasing particle concentration the ra- that obtained expressions close to Eq. (47).
tio bb/a = (bob + bpb)/a, then increases. Let us further Whitlock et al.' 3 applied Eq. (47) on turbid waters
assume that, d= 0. 7 0 ,ftu = 0. 35, rdp = 2.1, and rp = 5.4 and found rather poor agreement. But Fig. 3 illus-
regardless of the particle concentration. This may trates how Eq. (47) may underestimate R for values of
correspond to a condition with overcast sky. We have bb/a between 0.1 and 0.5, and most of their observa-
set bb/a 0 = 0.0875, which corresponds to blue light tions of bb/a lie within this range. The ratio bpb/bwb
(475 nm). Both the ratio K/a and R can now be calcu- could be higher than 1000,so that rd rdp, and Table I
lated as functions of bb/a. suggests that the value of rd in their case (smaller
The results are presented in Fig. 3. We see that particles) might be 1.3-1.4 rather than 1. Comparison
when bb/a becomes greater than b b/a, deviations be- with Eq. (35) now shows that Eq. (47) still may under-
tween Eqs. (31) and (47) will take place. Also the estimate R even when bb/a is a small number. It
asymptotic values of the two functions are different, should also be pointed out that the reliability of their
being rdu/rutd = 0.194 and 1/3, respectively. estimates of R from observations in air is open to
The approximated value for K/a, given by Eq. (48), questions. They measured the nadir radiance from
agrees with the general expression (29)in a wider range the sea and assumed that the relation between sub-
of the variable bb/a, but eventually also these equa- surface nadir radiance and upward irradiance was
tions separate. Eu 7rL(180'). It can be seen from Table I that this
Table I shows that the assumption rd = ru made by relation may underestimate Eu by a factor of 0.6-0.8.
Preisendorfer and Mobley7 is not correct. However, The mean value of their underestimations, as rated
the approximated expressions (34) or (35) for R and from their figures, seems to be a factor of value of 0.6-
(38) or (39) for K do not contain the quantity r, and 0.7.
their method, therefore, probably yields reasonable Another result 9 from the data by Hojerslev was that
results for conditions where bb/a <<1, that is, for most K (a + 0.022 b)/iid. (51)
natural conditions. But when nonabsorbing or weakly
absorbing particles dominate the attenuation, their We see that this expression agrees with Eq. (48) if
assumption is likely to lead to erroneous results. again bb = 0.022b.
Kirk2 5 found that the expression
V. Comparison with Observations
K = a [1 + (0425 cos0 - 0.19) b 1/2 (52)
From data by Hjerslev from the Mediterranean 2 l coO 8 L, a]
and the Baltic2 2 it has earlier been found' 9 for blue
light that described his results obtained by the Monte Carlo
technique. Here K is the average coefficient between
b the surface and the depth where Ed is reduced to 1%of
R 0.00725 .(50)
a +0.145b its surface value. is the zenith distance of the re-
2100 APPLIEDOPTICS / Vol. 26, No. 11 / 1 June 1987
fracted solar rays. By expanding the square root in 9. K. Ya. Kondratyev, Radiation in the Atmosphere (Academic,
series K becomes New York, 1969).
10. V. V. Sobolev, A Treatise on Radiative Transfer (Van Nostrand,
K [a + (0.213 cos02 - 0.095)b] (53) New York, 1963).
cosOs, 11. V. V. Sobolev, Light Scattering in Planetary Atmospheres
(Pergamon, Oxford, 1975).
Although this expression resembles the empirical rela- 12. E. Aas, "The Wavelength Selectivity of Light Scattering in the
tion (51), it probably overestimates the contribution Barents Sea," Inst. Rep. Ser., Inst. Geofysikk, U. Oslo 54 (1984).
from b to K when Asis small. 13. C. H. Whitlock, L. R. Pool, J. W. Usry, W. M. Houghton, W. G.
The conclusion so far is then that the described two- Witte, W. D. Morris, and E. A. Gurganus, "Comparison of Re-
stream model seems to be valid for oceanic conditions, flectance with Backscatter and Absorption Parameters for Tur-
where bb/a <<1 and rd 1. The general expressions bid Waters," Appl. Opt. 20, 517 (1981).
(29) and (31) remain to be tested in conditions where 14. D. Bauer and A. Morel, "Etyde aux petit angles de l'indicatrice
bb/a is closer to 1 or even larger and where rd and ru are de diffusion de la lumiere par les eaux de mer," Ann. Geophys.
closer to rdp and rup. A problem which then should be 23, 109 (1967).
15. A. Morel, "Optical Properties of Pure Water and Pure Sea
solved is how to estimate the values of the important Water," in Optical Aspects of Oceanography, N. G. Jerlov and
quantities rd and ru. E. Steemann Nielsen, Eds. (Academic,London, 1974),p. 1.
16. R. C. Smith, "Structure of Solar Radiation in the Upper Layers
VI. Inverse Problem
of the Sea," in Optical Aspects of Oceanography, N. G. Jerlov
If Ed(z) and Eu(z) are observed, K and R can be and E. Steemann Nielsen, Eds. (Academic, London, 1974), p. 95.
calculated, and a and bb can be found from previous 17. J. E. Tyler, "Radiance Distribution as a Function of Depth in an
results provided rd, ru, f d, and , u can be estimated. Underwater Environment," Bull. Scripps Inst. Oceanogr.7,363
Equation (32) may be written (1960).
18. J. E. Tyler, R. C. Smith, and W. H. Wilson Jr., "Predicted
aKId -R (4 Optical Properties for Clean Natural Water," J. Opt. Soc. Am.
d 1 + Rdd/i5u
62, 83 (1972).
and insertion of this expression for a in Eq. (31) togeth- 19. E. Aas, "The Vertical Attenuation Coefficient of Submarine
Irradiance," Inst. Rep. Ser., Inst. Geofysikk U. Oslo 28 (1978).
er with definitions (16)-(20) gives 20. N. G. Jerlov, Marine Optics (Elsevier, Amsterdam, 1976).
(id + f'u)fZdMu 21. N. K. Hjerslev, "Inherent and Apparent Optical Properties of
bb = RK the Western Mediterranean and the Hardangerfjord," Rep.
(,a,+Rd)(rdldU- Rrjid)
Inst. Fysisk Oceanografi, U. Copenhagen 21 (1973).
When bb/a <<1, so that R <<1, the expressions may 22. N. K. Hojerslev, "Inherent and Apparent Optical Properties of
be approximated to the Baltic," Rep. Inst. Fysisk Oceanografi, U. Copenhagen 23
(1974).
a = Kid, (56) 23. H. R. Gordon, 0. B. Brown, and M. M. Jacobs, "Computed
RK~id Relationships Between the Inherent and Apparent Optical
bb RK (1 + HdIIu)- (57) Properties of a Flat Homogeneous Ocean," Appl. Opt. 14, 417
rd
(1975).
When bb/a >>1, so that R is close to rdu/rut d, a is 24. J.T.O. Kirk, "Monte Carlo Study of the Nature of Underwater
still expressed by Eq. (54), while bb cannot be estimat- Light Field in, and the Relationships Between Optical Proper-
ties of, Turbid Yellow Waters," Aust. J. Mar. Freshwater Res.
ed. 32, 517 (1981).
References 25. J.T.O. Kirk, "Dependence of Relationship Between Inherent
and Apparent Optical Properties of Water on Solar Altitude,"
1. A. Gershun, "The Light Field," J. Math. Phys. 18, 51 (1939).
Limnol. Oceanogr. 29, 350 (1984).
2. A. Morel and R. C. Smith "Terminology and Units in Optical
26. A. Morel, and L. Prieur, "Analysis of Variations in Ocean Color,"
Oceanography," Mar. Geod. 5, 335 (1981).
Limnol. Oceanogr. 22, 709 (1977).
3. M. V. Kozlyaninov and V. N. Pelevin, "On the Application of a
One-dimensional Approximation in the Investigation of the
Propagation of Optical Radiation in the Sea," Dept. Commerce,
J. Publ. Res. Ser. Rep. 36, 54 (1966).
4. J. Joseph, "Untersuchungen fiber Ober- and Unterlicht-mes-
sungen im Meere," Dtsch. Hydrogr. Z. 3, 324 (1950).
5. R. W. Preisendorfer, Hydrologic Optics, Vol. 5: Properties
(U.S. Department of Commerce National Oceanic and Atmo-
spheric Administration, Environmental Research Laboratory,
Honolulu, 1976).
6. L. Prieur, "Transfert radiatif dans les eaux de mer. Application
a la determination de parametres optiques caracterisant leur
teneur en substances dissoutes et leur contenu en particules,"
Thesis, U. P. et M. Curie, Paris (1976).
7. R. W. Preisendorfer and C. D. Mobley, "Direct and Inverse
Irradiance Models in Hydrologic Optics," Limnol. Oceanogr. 29,
903 (1984).
8. S. Chandrasekhar, Radiative Transfer (Dover, New York,
1960).

1 June 1987 / Vol. 26, No. 11 / APPLIEDOPTICS 2101

You might also like