microwave emission model of layered snowpacks...1 microwave emission model of layered snowpacks...
TRANSCRIPT
1
Microwave Emission Model of Layered Snowpacks
Andreas Wiesmann and Christian Mätzler
Institute of Applied Physics - University of Bern
Sidlerstr. 5, 3012 Bern, Switzerland
Phone ++ 41 31 631 45 90 - Fax ++ 41 31 631 37 65
[email protected], [email protected]
Paper 1, submitted to Remote Sensing of Environment, December, 1998
Revised, March 1999
Abstract
A thermal Microwave Emission Model of Layered Snowpacks (MEMLS) was developed for
the frequency range, 5 to 100 GHz. It is based on radiative transfer, using six-flux theory to
describe multiple volume scattering and absorption, including radiation trapping due to total
reflection and a combination of coherent and incoherent superpositions of reflections between
layer interfaces. The scattering coefficient is determined empirically from measured snow
samples, whereas the absorption coefficient, the effective permittivity, refraction and reflection
at layer interfaces are based on physical models and on measured ice dielectric properties. The
number of layers is only limited by computer time and memory. A limitation of the empirical
fits and thus of MEMLS is in the range of observed frequencies and correlation lengths (a
measure of grain size). First model validation for dry winter snow was successful. An extension
to larger grains is given in Paper 2 (Mätzler and Wiesmann, 1999). The objective of the present
paper is to describe and illustrate the model and to pave the way for further improvements.
MEMLS has been coded in MATLAB. It forms part of a combined land-surface-atmosphere
microwave emission model for radiometry from satellites (Pulliainen et al., 1998).
2
1. Introduction
The observation of passive microwave signatures of different types of snowpacks [e.g. Mätzler
1994] and the need to simulate their microwave emission for future applications was the moti-
vation for the development of a Microwave Emission Model of Layered Snowpacks
(MEMLS). Since the penetration depth of microwaves and the scattering coefficient depend on
frequency and snow properties, a reasonable model should account for realistic snow-profiles of
the sensitive parameters, such as correlation length, density and temperature, and in case of wet
snow also the liquid-water content. A successful model might then be used for the development
of remote sensing tools for observing processes on and within the snowpack. An important
internal process, being responsible for the release of large avalanches, is temperature-gradient
metamorphism, often taking place near the bottom of the snowpack. This process is based on
large water-vapor gradients; it leads to the formation of cohesionless depth hoar (Armstrong,
1977, 1985). Microwave emission of depth-hoar layers are very special due to the enhanced
volume-scattering by the large crystals (Weise, 1996). On the other hand, snowpack
stratification has important implications for physical processes, such as avalanche formation by
weak layers (Föhn et al. 1998; Fierz, 1998), reduced vertical diffusion of water vapor and heat
transfer, and percolation of liquid water in case of wet snow (Arons and Colbeck, 1995, and
references therein). Snow layers also mark meteorological events in the history of snowpacks,
and thus they contain information about past weather conditions. Specific microwave signatures
related to stratification are expressed, for instance, by polarization features of microwave
emission, and sometimes by special spectral properties. Since the earliest radiometer observa-
tions, interest has been paid to layering of polar ice sheets (Gurvich et al. 1973; Zhang et al.
1989; Rott et al. 1993; Surdyk and Fily, 1993; Steffen et al. 1999), and layer effects were also
found in snow-covered sea ice (e.g. Mätzler et al. 1984). Microwave emission of layered media,
such as firn, was computed in the past by several authors (e.g. Gurvich et al. 1973; Djermakoye
and Kong 1979; West et al. 1996). Their models are applicable at frequencies below about 20
GHz where volume scattering by the granular snow structure is not dominant.
When both volume scattering by snow grains and by the stratification are relevant, the models
become more complex. An example is the strong fluctuation theory (Stogryn 1986). In the
implementation by Surdyk (1992), this model was tested by the experimental snow data of
3
Weise [1996] to be used also in the present work. The model validation clearly failed, probably
because multiple scattering is ignored. Indeed, West et al. (1993) showed on the basis of the
dense-medium radiative transfer theory applied to backscattering from snow that multiple
scattering by snow grains is important.
Driven by the need for a realistic snowpack emission model, we started with the development
of MEMLS, including multiple scattering both by stratification and by the granular snow
structure, and a tuned combination of coherent and incoherent superpositions of different scat-
tering contributions. First, the detailed behavior of single layers was investigated experimentally
by Weise [1996] and later by Wiesmann et al. [1998]. The work lead to a physical model of the
emissivity of homogeneous snow slabs in the frequency range from 10 to 100 GHz and for
correlation lengths from 0.05 to 0.3 mm. With the exception of the empirical determination of
the scattering coefficient, the parameters are physically based. In MEMLS this information is
applied to describe the behavior of individual layers. Here, the snowcover is considered as a
stack of n horizontal, planar layers, characterized by thickness, correlation length, density,
liquid-water content and temperature. The sandwich model of Wiesmann et al. [1998] is
extended to couple internal scattering and reflections at the interfaces of all layers. Internal
volume scattering is accounted for by a six-flux model (streams in all space directions). Homo-
geneity in the horizontal directions reduces to the well-known two-stream model (up and down
welling streams) where the two-stream absorption and scattering coefficients are functions of
the six-flux parameters. The absorption coefficient depends on density, frequency and tem-
perature, and the scattering coefficient depends on the correlation length, density and frequency
[Wiesmann et al. 1998]. The purpose of this paper is to describe and illustrate the model and to
pave the way for further improvements. Details on the realization are given by Wiesmann and
Mätzler [1998].
2. Microwave Emission Model of Layered Snowpacks
The snowcover is assumed to be a stack of n horizontal snow layers (j = 1,...n) with planar
boundaries between air-snow and snow-snow (Figure 1), characterized at frequency f, polari-
zation p, and incidence angle θ = θn by
− the snowpack brightness temperature Tb,
− the ground-snow interface reflectivity s0 and the ground temperature T0,
4
− the interface reflectivity sj on top of each layer j,
− the thickness dj, temperature Tj, internal reflectivity rj, emissivity ej and transmissivity tj of
each layer,
− and the downwelling (sky) radiation, expressed by the brightness temperature Tsky.
The snow-ground interface is allowed to be rough; we only specify its reflectivity s0, based on
observation or theory.
2.1 Radiation in multi-layered media
Considering layer j in Figure 2, we can derive the following equations relating the up- and
downwelling brightness temperatures at the layer boundaries:
A r B t C e Tj j j j j j j= + + (1)
( )B s A s Dj j j j j= + −− − −1 1 11 (2)
( )C s A s Dj j j j j= − ++1 1 (3)
D t B r C e Tj j j j j j j= + + (4)
Note that for j = 1 on the left-hand side of (2) we have a term with D0 on the right-hand side
which is identified as the ground temperature
D T0 0= (5)
and s0 is the ground-snow reflectivity. Similarly for j = n we have in (3) a term with An+1,
namely the sky brightness
A Tn sky+ =1 (6)
Solving the system of equations (1) to (4) for j = 1 to n, and using (5) and (6) leads to Dn and
thus to the brightness temperature Bn+1 = Tb of the snowpack according to
( )T s D s Tb n n n sky= − +1 (7)
Sometimes the emissivity esp or the transmissivity tsp of the whole snowpack is of more interest
than the brightness temperature Tb. The computation of these parameters is not obtained by
simply dividing Tb by a physical temperature because the snowpack is not regarded as isother-
mal. Nevertheless by computing Tb under various illumination conditions the required quantities
can be found. As an example we can get the total emissivity (snowpack and background) from
5
one minus the reflected fraction of the downwelling sky radiation by choosing two choices of
Tsky (0K and 100K),
( ) ( )e
T T T Tsp
b sky b sky= −
= − =1
100 0
100
K K
K(8)
Likewise tsp can be obtained from
( ) ( )( )t
T T T T
ssp
b b== − =0 0273 173
100
K K
K 1- 0(9)
If only one layer is present, Tb follows from (7) where Dn=D1 is given by
( ) ( ) ( ) ( )
01
2110
11
111100101
111100101
1
11
111
11
sr
tsssr
TeTsrTstsr
TeTstTsrst
Dsky
sky
−−−
+−+−+−
+−+−
= (10)
2.2 Radiative transfer in a refractive medium - the six flux model
In order to compute realistic values of rj, ej, and tj, volume scattering and absorption within each
layer j are to be described by a radiative transfer model. An important distinction between
radiative transfer in the atmosphere and in snow is that refraction plays a role in snowpacks.
Even more important is the fact that total internal reflection occurs for certain rays. Our model
takes these processes into account. We use a simplified radiative transfer model, reducing the
radiation at a given polarization to six fluxes streaming along and opposed to the three principal
axes of the slab (Figure 3). The horizontal fluxes represent trapped radiation, i.e. radiation
whose internal incidence angle θ is larger than the critical angle θc for total reflection:
( )'/1arcsin nc =θ>θ (11)
where n' is the real part of the refractive index of the slab. Only the vertical fluxes (θ < θc) can
escape from the snowpack. Scattering by angles near 90° causes a coupling between the vertical
and the horizontal fluxes. The trapped horizontal fluxes are represented by T03 to T06. Radiation
at θ < θc is represented by T01 for downwelling radiation and by T02 for upwelling radiation. We
assume that the radiation of the vertical fluxes can be represented by an effective propagation
direction represented by the angle θ with respect to the z axis. Then the transfer equations for
the six fluxes, expressed by brightness temperatures, are of the type
6
( ) ( )− = − − − −dT
dzT T T Ta b
0101 01 02cosθ γ γ ( )− − − − −γc T T T T T4 01 03 04 05 06 (12)
( ) ( )01020202 cos TTTT
dz
dTba −γ−−γ−=θ+ ( )− − − − −γc T T T T T4 02 03 04 05 06 (13)
and similar equations for T03 to T06, where γa is the absorption coefficient, γb the scattering
coefficient in the backward direction and γc the coefficient for coupling between the vertical and
horizontal fluxes. Since the medium is plane parallel and isotropic in the (x, y) plane, the fluxes
T03 to T06 are the same with vanishing derivatives in x and y directions, leading to
T T T T03 04 05 06 = = = ( )( )( )= γ γ γ γa c a cT T T+ + + −01 02
12 (14)
Inserting (14) in (12) and (13) leads to two-flux equations with modified coefficients:
( ) ( )− = − ′ − − ′ −dT
dzT T T Ta b
0101 01 02cosθ γ γ (15)
+ ( ) ( )dT
dzT T T Ta b
0202 02 01cosθ γ γ= − ′ − − ′ − (16)
The two-flux absorption coefficient aγ′ and scattering coefficient sγ′ can be written in terms of
the six-flux parameters:
( )( )′ = + + −γ γ γ γ γa a c a c1 4 21
(17)
( )′ = + + −γ γ γ γ γb b c a c4 22 1(18)
The typical difference between γ'a and γa is about a factor 2, and γa agrees well with the theo-retical absorption coefficient (Wiesmann et al. 1998). For constant T and constant coefficientswithin the given layer, T01 and T02 can be written as
( ) ( )T T A d B d01 = + ′ + − ′exp expγ γ (19)
( ) ( )dBrdArTT ′γ−+′γ+= − expexp 10002 (20)
where
d' = d/|cos θ|. (21)
The reflectivity r0 at infinite layer thickness is given by
( )r b a b01= ′ ′ + ′ + −γ γ γ γ (22)
and the damping coefficient γ (eigenvalue) is)2( baa γ′+γ′γ′=γ (23)
For a snow layer of thickness d the internal r and t, considering all multiple reflections, are
( )( )r = r - t - r t0 1 102
02
02 1−
(24)
7
( )( )t t r r t= − −−
0 02
02
02 1
1 1 (25)
where
( )t d0 = − ′exp γ (26)
is the one-way transmissivity through the slab. The layer emissivity is obtained from e=1-r-t.
Note that e, r, t are to be indexed by the layer number j.
The way γb and γc are related is determined by the internal scattering phase function and by the
critical angle θc given by (11). The solid angle of the beams for T01 and T02 is
√√
↵
−−π=Ω'
1'12
2
1n
n(27)
and the solid angle Ω2 of trapped radiation is the complement 2π - Ω1; that is,
'
1'2
2
2n
n −π=Ω (28)
The scattering coefficient γs is the sum over all fluxes
cbs γ+γ=γ 42 (29)
and using (27) and (28) we get for isotropy of scattering
1
22
ΩΩ=
γγ
b
c(30)
With Equations (27) - (30) we can determine the six-flux coefficients γa, γb, γc and γs from n' and
the two-flux coefficients aγ′ and bγ′ , using (17) and (18).
2.3 Reflectivity at layer interfaces
The interface reflectivity sj between two regular layers is given by sj = Fj 2 where Fj is the
Fresnel reflection coefficient of the interface between layers j and j+1. By regular we mean that
a simple transition occurs in the permittivities εj to εj+1, and that coherent superpositions due to
reflections at other interfaces can be ignored because if the layers are deep enough, the natural
variability in thickness, local incidence angle and also the different frequencies allowed within
the radiometer bandwidth tend to cancel the interference effects. This is the usual situation.
However, for layers thinner than about half a wavelength the variability is not large enough to
cancel the coherent effects, and thus they cannot be ignored; in other words, the quarter-wave-
length resonance must be taken into account. The one-way phase through layer j is given by
8
λθπ
= −1cos'2 jjjj
ndP . (31)
which is a function of the thickness dj, of layer j, the real part jn' of the refractive index, the
propagation angle θj-1 and vacuum wavelength λ. Defining
jiPj eE 2… , (32)
the reflection coefficient R2 of a coherent double layer (Figure 4) can be written as
RF F E
F F E22 1 2
2 1 21=
++ (33)
and the reflectivity s is given by
s RF F E
F F E= =
++2
2 2 1 2
1 2 2
2
1 . (34)
For a loss-less medium
( )( )s
F F F F P
F F F F P=
+ ++ +2
21
21 2 2
12
22
1 2 2
2 2
1 2 2
cos
cos(35)
The coherent effect is represented by the terms containing the phase P2. If they were neglected,
the equation would correspond to the incoherent treatment of the reflections. A continuous
transition from the coherent to the incoherent behavior is possible at every zero of cos(2P2).
Here, we choose the second zero at
2P2 = 3π/2 ≅ 4.712. (36)
In MEMLS the layer is considered to be coherent for phases below the limit (36), i.e. for P <
2.36 and incoherent if P is larger. If it is coherent, the effect a layer j is taken into account by
using the coherent reflectivity s of this thin layer as the interface reflectivity sj between the two
adjacent (incoherent) layers, j+1 and j-1, separated by the thin layer. Afterwards the coherent
layer is removed from the input table of the layered snowpack emission model, i.e. the coherent
layer is represented exclusively by s. This is a valid approximation because volume scattering
and absorption are negligible in layers thinner than about half a wavelength. This procedure has
to be done at each frequency and incidence angle. Care must be taken when two ore more
adjacent layers are thin.
9
2.4 Effective propagation angle and polarization mixing
So far it has been assumed that at any position within the snowpack the effective propagation
angle θj (in layer j+1, see Figure 1) with respect to the surface normal is given by Snell's law for
both vertically and horizontally polarized (up- and downwelling) radiation:
1'
sinsin
+
θ=θj
njs
n(37)
Here the index s in θjs refers to Snell. In addition to refraction, deviations in propagation
direction occur due to volume scattering. The proposed change leads to a first-order correction
in effective propagation path and polarization mixing for radiation originating at the radiometer
and propagating into the snowpack (i.e. we describe the reciprocal and equivalent path of
photons as if they were emitted by the radiometer antenna).
Effective propagation angle
For a radiometer observing at the incidence angle θn the effective propagation angle is well
approximated by (37) as long as volume scattering is not too important. This is true just below
the smooth snow surface. However with increasing snow depth, more and more of the forward
streaming flux consists of forward-scattered radiation, leading to an effective propagation angle
which may differ from (37). Assuming that, in the limit of strong scattering, the forward
(downward) and backward (upward) fluxes are diffuse then the effective propagation angle θjd
can be represented by
coscos
θθ
jdjc=
+1
2 (38)
where θjc is the critical angle for total reflection, i.e. the maximum propagation angle of radia-
tion belonging to the vertical fluxes; it is given by
21 )'(1cos −
+−=θ jjc n (39)
For a dilute medium ( 1' +jn = 1) we get cosθjd = 0.5; this is the usual value of the two-stream
model Ishimaru [1978]. In our case of six streams or "fluxes", the streams have more narrow
beams, therefore we have to use the modified form given by Equation (38).
10
The actual value of the effective propagation angle θj in layer j+1 is somewhere between θjs and
θjd. We assume that the two cosines are weighted according to the transmissivity of tz for non-
scattered radiation from the snow surface (at height zn) to height z < zn, i.e.
cos cos ( ) cosθ θ θj z js z jdt t= + −1 (40)
and tz is taken to be
tz = θγ
− nz
z js
s dzz
)'cos2
)'(exp( (41)
where γs(z') is the scattering coefficient at height z'. For negligible scattering tz ⇒ 1, leading to θj
= θjs, whereas for strong scattering we get tz = 0 and thus θj = θjd.
Polarization mixing
After diffuse scattering the photons loose their memory about where they came from. As a con-
sequence they cannot remember their plane of incidence which is needed to define the state of h
or v polarization. Therefore the v- and h-polarized fluxes are mixed in a similar way as θjs and
θjd. This mixing can be modeled by modifying the polarization-dependent parts of the
snowpack model, i.e. the reflectivities sjv and sjh at the interface between Layer j and j+1. The
computation of these reflectivities without polarization mixing is described in Section 2.3. Here
we describe the modification due to the mixing effect.
Let us denote the difference ∆sj by
∆sj = sjh - sjv (42)
This difference is reduced due to scattering, leading to effective interface reflectivities sjveff and
sjheff between Layer j and j+1 at position z = zj:
sjheff = [ ]0 5. s s t sjh jv z j+ + ∆ (43)
sjveff = [ ]0 5. s s t sjh jv z j+ − ∆ (44)
In (43)and (44) the values of sjv and sjh are the ones computed for the angles θjs. In a further
improvement, these coefficients could be averaged over the range of incidence angles. However,
except near the Brewster angle, the effects are small because they tend to cancel.
11
2.5 Primary model parameters (input data)
At a given frequency f, polarization p and incidence angle θ the following snow physical
parameters are needed for each layer j (j = 1,… n):
− density ρj
− temperature Tj
− liquid water content Wj
− correlation length pcj
− vertical extent zj
Furthermore the ground temperature T0 and the snow-ground reflectivity s0 are needed. These
primary parameters define the derived secondary model parameters to be described below.
2.6 Secondary model parameters
Dielectric properties of dry snow
For dry snow the real part of the microwave permittivity can be well expressed according to
Mätzler [1996] in terms of the density ρ in g/cm3:
′ = + +ε ρ ρd 1 15995 1861 3. . ; 0 ≤ ρ ≤ 0.4 g/cm3 (45)
( )( )′ = − +ε ν ε νεd h s13
; ρ > 0.4 g/cm3 (46)
where εh = 1.0, εs = 3.215 and ν = ρ/0.917 g/cm3.
The imaginary part of the dielectric constant of dry snow is given by:
′′ = ′ ′′ε ε νεd d iK2(47)
where K2 is the squared field factor (Figure 1 in Paper 2); for dry snow K2≅ 0.44 for ν≤0.33,
then increasing to K2=0.54 at ν=0.5. A similar expression, optionally used by MEMLS, is the
one proposed by Tiuri et al. [1984].
( )′′ = ′′ ? + ?ε ε ρ ρd i 0 52 0 62 2. . (48)
For the imaginary part εi" of the dielectric permittivity of ice, the formula of Mätzler (1998b)
(see also Equations (61) to (64) of Wiesmann et al. 1998) is used:
12
′′ = + ?εα
βi ff (49)
θ = −300
1K
T(50)
( ) ( )α θ θ= + ? ? − ?0 00504 0 0062 221. . exp . (51)
( )( )( )
β = ?−
+B
T
b T
b TB f1
2 22
1
exp
exp( )( )+ − + ? −exp . .10 02 0 0364 273T K (52)
where f is the frequency in GHz, B1=0.0207 K⋅GHz-1, b=335 K and B2=1.16⋅10-11 GHz-3.
Dielectric properties of wet snow
Wet snow exhibits a distinct Debye relaxation spectrum in the microwave range. According to
physical mixing models (e.g. Mätzler [1987]), for small values of the volumetric liquid water
content W, the effective permittivity ε = ε' + iε" is linear in W and given by 4 terms
ε = εa + εb + εc + εd (53)
where εd is the permittivity of dry snow, and εa, εb, εc are Debye terms expressed as
ε εε ε
k ksk k
kif f= +
−−×
×1 0/ ; k = a,b,c (54)
The Debye parameters εsk, ε∞k, f0k are given by the following three relations
√√↵
−
εε?+
ε−ε=ε11
3
d
swk
dswsk
A
W ; k = a,b,c (55)
√√↵
−
εε?+
ε−ε=ε×
××
113
d
wk
dwk
A
W ; k = a,b,c (56)
( )
ε−ε?+ε
ε−ε+=×
×
dwkd
wswkwk
A
Aff
)(100 ; k = a,b,c (57)
where the static εsw and infinite permittivity ε∞w and the relaxation frequency f0w of liquid water
at T = 0°C are: εsw = 88, ε∞w = 4.9, f0w = 9 GHz (Ulaby et al. 1986). The depolarization factors
are taken from experimental data: Aa = 0.005, Ab = Ac = 0.4975 (Mätzler,1987).
13
Absorption coefficient
The absorption coefficient γa follows from the imaginary part n" of the refractive index, and
thus from the complex permittivity, ε = ε'+iε" = (n'+in")2, according to
λπ=γ "4 n
a (58)
where λ is the vacuum wavelength.
Scattering coefficient
According to Wiesmann et. al. [1998], the six-flux scattering coefficient can be obtained from
snow physical properties. Several possible fits were found among which the following two
were of equal quality with regard to the available snow-sample data:
5.25.2
3 GHz5054.0
cm/g123.1
mm12.9 √√
↵
√√↵
+ρ−=γ fpec
s (59)
5.25.2
GHz50mm1295
mm116.3 √√
↵
?√
√
↵
+=γ fpp ecec
s (60)
where pec is the exponential correlation length, i.e. a length obtained by fitting the measured
correlation function to exp(-x/pec). The range of validity includes snow densities from 0.1 to 0.4
g/cm3, pec from 0.05 to 0.3 mm and frequencies from 5 to 100 GHz. An extension to larger
correlation lengths is given in Paper 2 (Mätzler and Wiesmann, 1998). For pec<0.05mm Equa-
tion (60) should be used. In the following examples we will use γs according to (59).
3. Examples
3.1 Emissivity of a snowpack
MEMLS was written in Matlab 5 [MathWorks, 1996]. All subroutines are commented, and a
help text is available with help subroutine. Many simulations have been made so far, and the
data seem in general to be realistic as long as the limited parameter ranges are respected. For
illustration purposes we present an example of an emissivity simulation. Table 1 shows the
input data, i.e. the snow profile of Dec. 21, 1995 found at Weissfluhjoch-Davos, Switzerland. It
14
is a winter snow situation with a snow height of 60.3 cm. The top layer (j=4) was 20cm of new
snow, below was a thin crust, and the bottom layer consisted of coarse-grained snow.
The measured and simulated emissivities as functions of frequency and incidence angle are
shown in Figure 5. The lines indicate computed data while the symbols indicate measurements.
Note the significant polarization difference at frequencies below 35 GHz where Layer 3 acts as
a coherent reflector in agreement with the measurements. The increase of the measured emis-
sivity at small incidence angles is due to the shadowing of the snowcover by the radiometers.
Whereas the emissivity at vertical polarization is slightly underestimated, the agreement is
excellent at horizontal polarization. The broad minimum in the h-polarized spectrum around 25
GHz is the result of the 3 mm crust of Layer 3 acting there as a quarter-wave resonance. The
transition from coherent to incoherent interaction takes place at 35 GHz where a kink is recog-
nized. Table 2 shows the intermediate model data of the simulation at 25 GHz. The indices h
and v stand for horizontal and vertical polarization, respectively. Layer 3 disappeared here
because it is considered a coherent layer. Its effect is expressed by the large value of sh for j=2.
3.2 Simulation of spatial snowpack emissivity variations
Sturm and Holmgren [1994] describe a vertical cross section through a snowpack covering a
distance of several meters at Imnavit Creek, Alaska, for November 18, 1989. Five layers were
observed; they consisted of depth hoar and wind crusts in alternation. The situation was
characterized by type, thickness of each layer and soil temperature. The corresponding layer
temperatures were interpolated from the soil and air temperatures. The correlation lengths and
densities were specified using typical values from our snow sample catalog (Wiesmann, 1998).
The spatial variation of the emissivity at 37 GHz for an incidence angle of 50° and the snow
height are shown in Figure 6. The expected anticorrelation between these two parameters is
apparent. Due to the strong scattering in the depth hoar layers, the emissivity is below 0.7. At a
displacement of 2m, the depth hoar is very shallow, leading to an increase of the emissivity up
to 0.9 and thus to a pronounced spatial variation in phase with the microtopography. The mean
emissivity agrees quite well with SSM/I data (assuming Tsky=30K), as shown in the figure.
4. Discussion and outlook to further development
The purpose of the present paper was mainly to describe MEMLS and all its elements. First
applications were used to show that reasonable results can be obtained. In order to further test
15
the model and in order to assess its parameter range and errors, MEMLS should be applied to
many kinds of snowpacks and firn with well-known physical and structural properties. Since
the number of parameters can be large, it is difficult at present to give a perspective on the fea-
sibility to retrieve special snow parameters from microwave measurements. However, this
potential can now be studied by simulations. First sensitivity studies of MEMLS were done by
Pulliainen et al. (1998). Some possible steps for future considerations are proposed in the fol-
lowing:
• Weighting functions, defined by wij = ∂ep/∂xij, where xij is the ith physical parameter of layer
j, can be computed in order to assess the sensitivity of the model to any parameter.
• The assumption of an isotropic and exponential correlation function is not valid in general,
but it is a reasonable simplification. Furthermore it leads to a representative definition of the
structure parameter, i.e. the correlation length pec, which is obtained from fitting the
observed correlation function to exp(-x/pec) as described by Weise (1996). The index e in pec
is meant to indicate this exponential fit. Wiesmann et al. (1998) used the slope of the cor-
relation function at zero displacement x=0 to define a slightly different correlation length pc.
This lead to noisier and slightly different values, especially in case of fine-grained snow.
• By including the occasional snow-structure anisotropy (e.g. observed in new snow depos-
ited under calm wind condition, surface hoar, depth hoar) the model can be improved,
especially its polarization behavior at frequencies above 30 GHz.
• Occasionally, we observed clear deviations of the spatial correlation function from simple
exponentials. Improvements can be expected by choosing more suitable functions. Using the
improved Born Approximation and measured correlation functions, it can be tested how
these functions are related to snow types and how they influence the microwave properties.
• Further improvements of the polarization behavior of the model is possible by using a fully
polarimetric radiative transfer code. However, the expected gain from this more elaborate
approach is limited, because volume scattering in snow tends to be insensitive to polariza-
tion (Wiesmann et al. 1998). The most obvious polarization effects arise from reflections at
layer interfaces which are treated by the present model.
16
• A limitation of MEMLS at very high frequencies results from the assumption of smooth
interfaces between layers and of a smooth snow surface. Improvements can be expected by
including roughness effects.
• A limitation of the present MEMLS version is given by the limited range of correlation
lengths. The range covers the snow structure observed at Weissfluhjoch during two winter
seasons. Larger grains are observed for old wet snow; when it refreezes the scattering leads
to extreme microwave signatures which cannot be accounted for by the present model.
Paper 2 will be dedicated to this situation.
The complexity of MEMLS can be large. In fact, arguments were raised for preferring single-
layer snow emission models as they may be more practical and thus more feasible. This state-
ment can also be tested by simulations using MEMLS for different values of n where the single-
layer situation (n =1) is met as well. Nevertheless it should be pointed out that layering is
inherent to most snowpacks as described in the introduction and found in the examples.
5. References
Armstrong R.L., “Continuous monitoring of metamorphic changes of the internal snow struc-
ture as a tool in avalanche studies”, J. Glaciol., Vol. 19, pp. 325-334 (1977).
Armstrong R.L., “Metamorphism in subfreezing, seasonal snow cover: the role of thermal and
vapor pressure conditions”, PhD Thesis, University of Colorado, Fac. Graduate School,
Dept. Geography (1985).
Arons E.M and S.C. Colbeck, “Geometry of heat and mass transfer in dry snow: a review of
theory and experiment”, Reviews of Geophys. Vol. 33, pp. 463-493 (1995).
Djermakoye B., and J.A. Kong, “Radiative-transfer theory for the remote sensing of layered
random media”, J. Appl. Phys. Vol. 50, pp. 6600-6604 (1979).
Fierz C. “Field observations and modelling of weak-layer evolution”, Annals Glaciol. 26, pp. 7-
13 (1998).
Föhn P.M., C. Camponovo and G. Krüsi, “Mechanical and structural properties of weak snow
layers measured in situ”, Annals Glaciol. Vol. 26, pp. 1-6 (1998).
17
Gurvich A.S., V.I. Kalinin, and D.T. Matveyev, “Influence of the internal structure of glaciers
and their thermal radio emission”, ATM. Oceanic Phys. Vol. 12, pp. 713-717 (1973).
Ishimaru A., "Wave propagation and scattering in random media", Vols. 1 and 2, Academic
Press, Orlando (1978).
Looyenga H., "Dielectric constant of heterogeneous mixtures", Physica, Vol. 21, pp. 401 - 406
(1965).
MathWorks The, Inc., "Matlab, the language of technical computing", 24 Prime Park Way,
Natick, MA 01760-1500 (1996).
Mätzler C., R.O. Ramseier, and E. Svendsen, "Polarization effects in sea-ice signatures", IEEE
J. Oceanic Engin., Vol. OE-9, pp. 333-338 (1984).
Mätzler C., "Applications of the Interaction of Microwaves with the Seasonal Snow Cover",
Remote Sensing Reviews, Vol. 2, pp. 259 - 387 (1987).
Mätzler, C. "Passive microwave signatures of landscapes in winter", Meteorology and Atmos-
pheric Physics, Vol. 54, pp. 241-260 (1994).
Mätzler C.: "Microwave permittivity of dry snow", IEEE Transactions Geosci. and Rem.
Sens., Vol. 34, No. 2, pp. 573 - 581 (1996).
Mätzler C., "Improved Born Approximation for scattering of radiation in a granular media", J.
Appl. Phys., Vol. 83, pp. 6111-6117 (1998a).
Mätzler C., "Microwave properties of ice and snow", in B. Schmitt et al. (eds.), Solar System
Ices, pp. 241 - 257, Kluwer Academic Publ. Dordrecht (1998b).
Mätzler C. and A. Wiesmann, "Extension of the Microwave Emission Model of Layered
Snowpacks to Coarse-Grained Snow", to be published in Remote Sensing of Environment,
Paper (1999).
Mishima O., D. Klug, and E. Whalley, "The far-infrared spectrum of ice in the range 8-25cm-1:
Sound waves and difference bands, with application to Saturn's rings", J. Chem. Phys., 78,
6399 - 6404 (1983).
18
Polder D., and J.H. van Santen, "The effective permeability of mixtures of solids", Physica, Vol.
12, pp. 257 - 271 (1946).
Pulliainen J., K. Tigerstedt, W. Huining, M. Hallikainen, C. Mätzler, A. Wiesmann, and C.
Wegmüller, "Retrieval of geophysical parameters with integrated modeling of land surfaces
and atmosphere (models/inversion algorithms)", Final Report, ESA/ESTEC, Noordwijk,
Netherlands, Contract No. 11706/95/NL/NB(SC), Oct. (1998).
Reber B., C. Mätzler, E. Schanda, "Microwave Signatures of Snow Crusts: Modelling and
Measurements", Internat. J. of Remote Sensing, Vol. 8, pp. 1649 - 1665 (1987).
Rott H., K. Sturm and H. Miller, “Active and passive microwave signatures of Antarctic firn by
means of field measurements and satellite data”, Annals Glaciol. Vol. 17, pp. 337-343
(1993).
Steffen K. and W. Abdalati, “Faceted crystal formation in the NE-Greenland low accumulation
region”, J. Glaciol. (in press 1999).
Stogryn A. “A study of the microwave brightness temperature of snow from the point of view
of strong fluctuation theory”, IEEE Trans. Geosci. Remote Sens., GE-24, pp. 220-231
(1986).
Sturm M., and J. Holmgren, "Effects of microtopography on texture, temperature and heat flow
in Arctic and sub-Arctic snow", Annals of Glaciol., Vol. 19, pp. 63 - 68, 1994.
Surdyk S. “Etudes des signatures spectrales micro-ondes obtenues par télédétection sur la
calotte polaire Antarctique”, Thèse de doctorat de l’institut national polytechnique de
Grenoble (1992).
Surdyk S. and M. Fily, “Comparison of the passive microwave spectral signature of the
Antarctic ice sheet with ground traverse data”, Annals Glaciol., Vol. 17, pp. 161-166 (1993).
Tiuri, M.E., A. Sihvola, E. Nyfors, and M. Hallikainen, "The complex dielectric constant of
snow at microwave frequencies", IEEE J. Oceanic Eng., OE-9, pp. 377 - 382 (1984).
Ulaby F., R. Moore, A. Fung, "Microwave Remote Sensing, Vol. III, Artech House, pp. 2022 -
2025 (1986).
19
West R., L. Tsang and D.P. Winebrenner, “Dense medium radiative transfer theory for two
scattering layers with a Rayleigh distribution of particle sizes”, IEEE Trans. Geosci. Remote
Sensing, Vol. 31, pp. 426-436 (1993).
West R.D, D.P. Winebrenner, L. Tsang and H. Rott, “Microwave emission from density-strati-
fied Antarctic firn at 6 cm wavelength”, J. Glaciol., Vol. 42, pp. 63-76 (1996).
Weise T., "Radiometric and Structural Measurements of Snow", Ph.D. Thesis, Institute of
Applied Physics, Phil. nat. Fakultät, Universität Bern, Bern, Switzerland, February (1996).
Wiesmann A., T. Strozzi, and T. Weise, "Passive Microwave Signature Catalogue of Snow-
covers at 11, 21, 35, 48 and 94 GHz", IAP Research Report 96-8, University of Bern, Swit-
zerland (1996).
Wiesmann A., C. Mätzler and T. Weise, "Radiometric and structural measurements of snow
samples", Radio Sci. Vol. 33, No. 2 , pp. 273-289 (1998).
Wiesmann A. and C. Mätzler, "Documentation for MEMLS 98.2, Microwave Emission Model
of Layered Snowpacks", Research Report No. 98-2, Microwave Dept., Institute of Applied
Physics, University of Bern (1998).
Zhang H., L.Toudal-Pedersen, and P. Gudmandsen, “Microwave brightness temperatures of the
Greenland ice sheet”, Adv. Space Res. Vol. 9, pp. (1)277-(1)287 (1989).
20
Tables
Table 1: Snow profile of data from Dec. 21, 1995 where Tsnow is the snow temperature, W
the liquid water content, ρ the snow density, d the layer thickness, and pec the expo-
nential correlation length. The layer number increases with height above ground (from
Wiesmann et al. 1996).
Date/layer No.
Tsnow
[K]W
[%]ρ
[kg/m3]d
[cm]pec
[mm]21.12.95
1 273.0 0.00 259.0 25.0 0.16162 272.0 0.00 177.0 15.0 0.07203 271.8 0.00 400.0 0.3 0.06004 271.4 0.00 70.0 20.0 0.0550
Table 2: Intermediate model data of the simulation at 25 GHz and 50° incidence angle.The upwelling brightness temperatures in each layer are given by Dh and Dv.
j θ[°]
d'[cm]
sh sv r t Dh
[K]Dv
[K]0 39.5 0.096
50.0535
1 42.3 32.4 0.0024
0.0001
0.0696
0.8358
252.9 255.9
2 46.6 20.3 0.0972
0.0041
0.0110
0.9582
251.8 254.8
4 50.0 29.1 0.0036
0.0000
0.0180
0.9615
235.9 251.7
Air 235.4 251.7
21
Figure CaptionsFigure 1: A multi-layer system with a wave incident from above at an angleθn.
Figure 2: Vertical fluxes at the boundaries of layer j.
Figure 3: The six fluxes within a selected layer j.
Figure 4: Thin layer j between two thicker, scattering layers j-1 and j+1, here for j=2.
Figure 5: Measured (circles, triangles) and simulated (curves) data of the snowpack of
Dec. 21, 1995. Left: Emissivity versus frequency at θn = 50°. Right: Emissivity versus
incidence angle at f = 21 GHz.
Figure 6: Emissivity (upper lines) and snowheight (lower line) versus distance. Observa-
tion frequency is 37 GHz and incidence angle θn=50°. Input data are taken from Sturm
and Holmgren [1994]. The horizontal dashed line indicates the emissivity of the nearest
SSM/I pixel at 37 GHz ( v-pol.) assuming Tsky=30K.
22
Figures
s0
s1
s2
s3
s4
sn-1
sn
e1 r1
θn
θn-1
T0
t1 T d 1 1
e2 r2 t2 T d 2 2
e3 r3 t3 T d 3 3
e4 r4 t4 T d 4 4
en-1 rn-1 tn-1 T d n-1 n-1
en rn tn T d n n
b T sky T
θ0
θ1
θ2
Figure 1: A multi-layer system with a wave incident from above at an angleθn.
sj
sj-1
e , r , t , T
A
C
A
C
j+1
j
j
j-1
B
D
B
Dj-1
j
j
j+1
z
zj-1
zj
j j j j
Figure 2: Vertical fluxes at the boundaries of a selected layer j.
23
Figure 3: The six fluxes within a selected layer j.
n2
n3
n1
F1
F2
θ0
θ1
θ2
d2j
j-1
j+1
Figure 4: Thin layer j between two thicker, scattering layers j-1 and j+1, here for j=2.
10 100F [GH ]
0.5
0.6
0.7
0.8
0.9
1.0
emis
sivit
y e
v - pol
h - pol
v - pol
h - pol
10 20 30 40 50 60 70 8incidence angle [degrees]
0.5
0.6
0.7
0.8
0.9
1.0
emis
sivi
t y e
v - pol
h - pol
v - pol
h - pol
24
Figure 5: Measured (circles, triangles) and simulated (curves) data of the snowpack ofDec. 21, 1995. Left: Emissivity versus frequency at θn= 50°. Right: Emissivity versusincidence angle at f=21 GHz.
0.00 1.00 2.00 3.00 4.00 5.00distance [m]
0.40
0.50
0.60
0.70
0.80
0.90
1.00
emis
sivi
t y
0 .00
0.20
0.40
0.60
0.80
sno
whei
ght [cm
]
h-pol
v-pol
snow height
Figure 6: Emissivity (upper lines) and snowheight (lower line) versus distance. Observa-tion frequency is 37 GHz and incidence angle θn=50°. Input data are taken from Sturmand Holmgren [1994]. The horizontal dashed line indicates the emissivity of the nearestSSM/I pixel at 37 GHz ( v-pol.) assuming Tsky=30K.