IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 33, NO.

3, MAY 1995 803

Dielectric Properties of Soils in the 0.3-1.3-GHz Range

20 , .., ~. , * . . .,. ,
I , . .L
Neil R. Peplinski, Fawwaz T. Ulaby, and Myron C. Dobson Old Model C.

Abstract-In 1985, we reported the development of a semiempirical

Fieldl: 0.3
I5J 0.3 GHz //i New M&l

dielectric model for soils, covering the frequency range between 1.4 and
18 GHz. The model provides expressions for the real and imaginary parts 10-
of the relative dielectric constant of a soil medium in terms of the soil’s
textural composition (sand, silt, and clay fractions), the bulk density and __--.
volumetric moisture content of the soil, and the dielectric constant of 5- __---_ _ _ _ _ _---- -
water at the specified microwave frequency and physical temperature.
This communication provides similar expressions for the 0.3-1.3-GHz
range. Upon comparing experimental results measured in this study with 0 ~ ~ ~ “ . . . ~ 1 ’ ’ ~ ’ 1 ’ ~ ’ ’ 1 ’ ’ ~ *
predictions based on the semiempiricalmodel, it was found that the model 0 5 10 15 20 25
underpredicts the real part of the dielectric constant for high-moisture
cases and underestimates the imaginary part for all soils and moisture Volumetric Moisture m, (%)
conditions. A small linear adjustment has been introduced to correct the
expression for the real part and a new equation was generated for the
effective conductivity to correct the expression for the imaginary part. In
addition, dielectric measurements were made to evaluate the dependence - --- - - - Old Model
of the dielectric constant on clay type. The results show significant New Modsl
variations for the real part and large variations for the imaginary part
among soils with the same clay fractions but with clays of different specific Fieldl: 1.3 GHz
surface areas.
10-
I. INTRODUCTION
The purpose of the study presented in this paper is to develop 5-
expressions for characterizing the real and imaginary parts of the
relative dielectric constant of soils in the 0.3-1.3-GHz frequency
range. The development was motivated by a need to use such 0
expressions at 440 MHz in the analysis of radar observations made 0 5 10 15 20 25
by the P-band channel of the JPL AIRSAR system [ 2 ] . The study
Volumetric Moisture m, (%)
consisted of three steps: 1) dielectric measurements were conducted
using a dielectric probe for various soil types at various moisture Fig. 1. Comparison of values of E’ and 8’measured at 0.3 and 1.3 GHz
with calculations based on the old model given by (1H8) and the new model
contents, 2) the measured data were compared with predictions based with the corrections given by (9) and (10).
on the semiempirical expressions that were developed in [ 11 for the
1.4-18-GHz frequency range, and 3) then those expressions were
modified to better fit the observed behavior in the 0.3-1.3-GHz range. TABLE I
These steps are described in the following sections. WEIGHT PERCENTAGES
OF SAND,
SILTAND
CLAYFOR THE VARIOUS
SAMPLES
MEASURED
Soil Mixtures Sand (%) Silt (%) Clay (%)
11. MEASUREMENT
PROCEDURE Field 1 50 35 15
The dielectric data measured in support of this study consisted Field 2 40 55 5
of 19 sets of distinctive soil conditions, distributed among four soil Field 3 30 60 10
types and several moisture conditions. For each soil condition, the Field 4 15 65 20
real and imaginary parts of the relative complex dielectric constant
E were measured at each of 21 equally spaced frequency points
covering the range between 0.3 and 1.3 GHz. This corresponds to
used to solve for E’ and E” of the material under test. The probe tip is
399 data points for the real part e’ and the same number for the
modeled by an equivalent circuit that consists of a capacitance due to
imaginary part E “ . The measurements were made using a coaxial
the fringing field between the inner and outer conductor, a fringing
probe technique [3] that had been used previously to measure the
capacitance due to the field in the dielectric medium around the probe
dielectric properties of several types of materials. With the coaxial-
tip, and a resistance that accounts for losses due to radiation into the
probe tip inserted in the material under test (in this case, a soil
material [3]. The exact value of the elements in the equivalent circuit
sample), the magnitude and phase of the reflection coefficient are
measured by a network analyzer, and then the measurements are are determined through a calibration procedure that uses materials
with known dielectric properties; namely, a perfect conductor (short
circuit), air (open circuit), distilled water, and methanol.
Manuscript received September 22, 1994. This work was supported by Using the dielectric probe technique, the dielectric constant E
NASA under Grant NAGW-2151. was measured for four soil types, each of which was prepared by
The authors are with The Radiation Laboratory, Department of Electrical
Engineering, and Computer Science, The University of Michigan, Ann Arbor, mixing together prescribed amounts of sand, silt, and Georgia Kaolin
MI 48109-2122 USA. clay (Table I). Sand includes particles with diameters in the range
IEEE Log Number 9409911. between 0.05 and 2.0 mm, silt includes particles with diameters in

019&2892/95$04.00 0 1995 IEEE

 IEEE TRANSACTIONS ON GEOSCIENCEAND REMOTE SENSING, VOL. 33, NO. 3, MAY 1995

4- Field 1: m, = 5%

0.3 0.5 0.7 0.9 1.1 1.3

Calculated Permittivity E‘
Frequency (GHz)
(a)

5 I** **..-

0 2 4 6 6 10 12 14 16 18
0.3 0.5 0.7 0.9 1.1 1.3 Calculated Permittivity E‘
Frequency (GHz) (b)
Fig. 2. Comparison of measured spectra of e’ and err at low and high Fig. 3. Comparision of measured values of E‘ with (a) calculated values
moisture content with calculations based on the new and old models. based on the original model (2), and (b) calculated values after applying the
correction given by (9).
the 0.0024.05-mm range, and clay includes particles with diameters
smaller than 0.002 mm. The soil mixtures listed in Table I, which dielectric mixing model of the following form:
were chosen to approximate the type of soils measured previously
in [l], were each measured at volumetric moisture contents of 5 , 10,
15, 20, and 25% (except for the soil denoted “Field 4” for which no
measurements were conducted at the highest moisture state because
the network analyzer was no longer available).
After drying the sand, silt and clay in an oven at 110°C for 24
h, the soil samples were mixed according to the weight fractions where em is the relative complex dielectric constant of the soil-water
listed in Table I. Once the mixtures were ground with a mortar and mixture, mu is the water volume fraction (or volumetric moisture
pestle to remove any clods that may have existed, 20-mL samples content) of the mixture, P b is the bulk density in grams per cubic
of the dry soils were added to the test beakers and weighed. A centimeter, pa = 2.66 g/cm3 is the specific density of the solid soil
predetermined amount of distilled water was then added to each particles, CY =0.65 is an empirically determined constant, and p’ and
container to achieve the desired volumetric moisture content. After p” are empirically determined soil-type dependent constants given by
weighing the wet soils, the samples were then sealed and allowed
to cure for another 24 h. From the weight of the dry soil and wd 0’ = 1.2748 - 0.519s - 0.152C (4)
the weight of the water added Ww,the gravimetric moisture content
is given by mg = ww/wd.Assuming that the soil does not shrink and
or swell as a function of moisture content, the volumetric moisture
p” = 1.33797 - 0.603s - 0.166C (5)
content mu is then determined by pbm,, where ,ob is the bulk density
of the soil sample (given by wd/v,where v
is the volume of the where S and C represent the mass fractions of sand and clay,
dry sample). respectively (Le., 0 \leqS,C 5 1). The quantities and cyw are
the real and imaginary parts of the relative dielectric constant of
free water, given by a Debye-type dispersion equation, with the latter
III. SEMIEMPIRICAL
MODEL modified to include a term that accounts for the effective conductivity
of the soil mixture
Using an extensive dielectric database covering five soil types,
a wide range of moisture conditions, and a broad frequency range
extending between 1.4 and 18 GHz, [l] developed a semiempirical
~;w = Ewm + 1+ (2TfTW)*
- EwO Ewm (6)
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 33, NO. 3, MAY 1995 805

and
I/ 2Xf7w(EwO - E w w ) geff (Ps -pb)
Efw =
+
1 (27rf7,)Z + 5GJ (7)

where EO is the permittivity of free space, T~ is the relaxation time

for water, f is the frequency in hertz, E ~ Ois the static dielectric
constant for water, and E, = 4.9 is the high-frequency limit of
Expressions for T~ and E,O are given as a function of temperature
by Ulaby et al. [4, Appendix E-21. At room temperature (20°C),
27r7, = 0.58 x s and E,O = 80.1.
The database used in [l] consisted of dielectric measurements at
1.4 GHz and additional measurements at multiple frequences in the
4-18-GHz range. Because of the l/f dependence of the conductivity
0 1 2 3 4 5 6 7 8
term of E'Jw (second term in (7)), the relative contribution of that
Calculated Loss Factor E"
term was found to be insignificant at 4 GHz and higher frequencies.
Hence, only the data at 1.4 GHz were used to develop the following (a)
empirical expression for the effective conductivity, geffrin terms of
the textural properties of the soil:

Ueff = -1.645 + 1.939 p b - 2.25622s + 1.594C. (8)

As will be discussed in the next section, this expression does not lead
to good agreement between calculated and measured values of E" at
frequencies below 1.4 GHz.

IV. DIELECTRIC MODELFOR THE 0.3-1.3-GHz RANGE

As was stated earlier in Section 11, a total of 399 measurements of
E', and the same for E", were made, covering the range from 0.3 to 1.3
GHz for four soil mixtures under various moisture conditions. The 0 0.5 1 1.5 2 2.5 3 3.5
first step in the analysis of the data was to compare the measured Calculated Loss Factor E"
values with those calculated on the basis of the model outlined in (b)
the previous section. Fig. 1 shows measured values of E' and E"
Fig. 4. Comparision of measured values of e'' with (a) calculated values
plotted as a function of volumetric moisture content m , . The data
based on the original model, and (b) calculated values after replacement of
shown correspond to the measurements made at the two ends of the o,fi with the new expression given by (IO).
0.3-1.3-GHz range. Also shown are two sets of curves, one based on
the old model given by (1)-(8), and the other based on the modified
model using (9) and (10). Another perspective is presented in Fig. 2 a new empirically derived expression is used for the effective
where E' and E" are plotted as a function of frequency, for specific conductivity, namely
values of m,. The two figures basically show that the old model
slightly underestimates E' for high values of m, at frequencies in the
= 0.0467 + 0.2204pb - 0.4111s + 0.6614C. (10)
lower part of the 0.3-1.3-GHz range, and significantly overestimates Use of (10) instead of (8) in (7) leads to much better agreement
E" for all moisture conditions. The curves based on the modified between E" and E:, as shown in Fig. 4(b).
model discussed below are clearly in much better agreement with the Fig. 5 provides typical comparisons between measured spectra of
measured data. E' and E" and those calculated on the basis of the corrections indicated
For the data set as a whole, comparison between measured and (old by (9) and (10). All four soil mixtures listed in Table I are represented
model) calculated values of E' is shown in Fig. 3(a). Although the in Fig. 5, each at a different moisture content.
measured and calculated values exhibit a high degree of correlation
(T = 0.974), the model underestimates E' at high moisture conditions. v. DIELECTRIC
DEPENDENCE
ON SELECTED CLAY TPE
This can be corrected by a simple linear adjustment
The dependence of E' and E" on soil textural composition (sand,
E' = 1.156; - 0.68 (9) silt, and clay fractions) is a consequence of the role played by bound
water. In the model advanced in [l], the soil mixture is comprised of
where EL is the model-predicted value given by (2). Such an four components: air, solid soil particles, free water, and bound water.
adjustment leads to the comparison shown in Fig. 3(b) and to a linear The apportionment of the total water volume fraction m, between
correlation coefficient of 0.985. free and bound water is governed by the specific surface area A,
For the imaginary part (dielectric-loss factor), the model given (in m2 /g) of the soil particles because the bound water is defined as
by (3) overpredicts the measured value of E", with the error being adsorbed cations that are tightly held by negatively charged particle
relatively small at 1.3 GHz, but increases rapidly with decreasing surfaces composed predominantly of clay. Whereas for pure sand, A,
frequency. At 0.3 GHz, some of the high-moisture values predicted is on the order of lop3 to 10-1 m'/g, depending on the particle size,
by the model were more than double the values measured by the A, is on the order of 10-1 to 1 m2/g for silt particles, and anywhere
dielectric probe. This deviation leads to the poor correlation ( T = between 1 and lo3 m2/g for clay. Thus it is the clay fraction, and more
0.423) exhibited in Fig. 4(a). Analysis of the measured behavior of specifically the mineralogy of the clay particles, that determines A, of
E" relative to the behavior of E; revealed that the basic structure of the soil mixture, and hence the portion of m, in the form of bound
the model remains applicable over the 0.3-1.3-GHz range, provided water. The dielectric constant of free water is given by the Debye
806 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 33, NO. 3, MAY 1995

0.9 0.6 0.7 0.9 1.1 18 0.9 0.6 0.7 0.9 1.1 1.9
Frequency (GHz) Frequency (GHz)

Field 1: 9= 20.0 %
..* .- ,
Field 3: m., = 16.0%

9j.
.................
*:-
;
-i-
;
9 = .-
.................
j mj. . .j:
................. i ....
20
18 *__._._
................. ................. ....
.......................................................... i......................
................. .................
1

................. L... ................. ................. ................ ................. .................

............... ................. c.. ............. .................
.................i...
................................ ................ ............... ...................... ................. ................. .................
.................i................. i................. ;................. i.................i.... ................. ................. ................. ................. ....
..............,.................j .................;... ................................... ................ ....
............. .................:.................:....
4
u 0.” v 1 :
O l 1 I I I I 0 I I I I I
08 0.6 0.7 0.9 1.1 1.9 0.S 0.6 0.7 0.9 1.1 1.9

Frequency (GHz) Frequency (GHz)

Fig. 5. Comparision between measured and predicted values for various soils.

Effects of Clay Type: f=700MHz Effects of Clay Type:f=700MHz

-w

0 6 1 0 1 6 2 0 2 6 9 0 0 5 10 16 20 25 30
Volumetric MoisRve m, (5) Volumetric Mobture mv(%)
Fig. 6. Measured variations of e/ and e‘/ with volumetric moisture content at 0.7 GHz for three soils with the same sand, silt, and clay fraction (0.5,
0.35, and 0.15, respectively), but with different clay type component.

dispersion equation, but that of bound water is not well understood. TABLE II
It has been estimated that the dielectric constant of bound water is TYPICAL. VALUES FOR THE SPECJPIC SURFACE
AREASOF DIFFEMNT ’IkpEs OF CLAYS
on the order of Ebw = 35 - j l 5 . At 1 GHz, the dielectric constant of
pure free water at room temperature is efw = 79.8 - j 4 . 3 (excluding Clay S p e Specific Surface Area
the conductivity term in (7)). Because < and ~b:,> eYw, (m2/g)
as we progress from primarily sandy soils towards primarily clayey Georgia Kaolin 5-20
soils, and therefore increase the relative proportion of bound water to
total water in the mixture, we should expect E‘ of the soil mixture to
Kentucky Ball
Westem Bentonite
20.9
700 -
decrease and e” to increase. This is precisely the behavior exhibited
by experimental data at 1.4 GHz [ll, [51. of clay and their estimated specific surface areas. The Georgia Kaolin
By way of exploring this behavior further, we conducted an clay was the clay used in the study reported in the preceding sections.
experiment to compare the dielectric properties of three soil mixtures, Fig. 6 displays the variations of e’ and E” as a function of volumetric
all having the same sand, silt, and clay fractions (0.5, 0.35, and 0.15, moisture content at 0.7 GHz. Not only do the results confirm the
respectively), but different types of clay. Table II lists the three types expected dependence on A,, they also point out how strong the
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 33, NO. 3, MAY 1995 807

dependence is, particularly for E ” . These results suggest the need Microwave Backscatter and Attenuation Dependence
for conducting a more exhaustive study to determine and model the on Leaf Area Index for Flooded Rice Fields
dependence of E on A, at frequencies below 1 GHz.
Stephen L. Durden, Leslie A. Momssey, and Gerald P. Livingston
VI. CONCLUDING
REMARKS
Comparison of experimental measurements of the dielectric con- Abstract- Wetlands are important for their role in global climate
stant performed over the frequency range between 0.3 and 1.3 GHz as a source of methane and other reduced trace gases. As part of an
for four soil mixtures at various moisture contents with values cal- effort to determine whether radar is suitable for wetland vegetation
monitoring, we have studied the dependence of microwave backscatter
culated according to the semiempirical model developed previously and attenuation on leaf area index (LAI) for flooded rice fields. We find
in [l] on the basis of higher frequency data led to the following that the radar return from a flooded rice field does show dependence on
conclusions: LAI. In particular, the C-band W cross section per unit area decreases
with increasing LAI. A simple model for scattering from rice fields is
1) The model remains equally applicable over the 0.3-1.3 GHz- derived and fit to the observed HH and VV data. The model fit provides
range if a linear correction is applied to E ’ , as indicated by insight into the relation of backscatter to LA1 and is also used to calculate
(9). the canopy path attenuation as a function of LAI.
2) The model also is applicable for E” if the expression for the
effective conductivity is replaced with the new expression given I. INTRODUCTION
by (10).
Wetlands are important for their role in global climate as a source
The above results are based on soils whose clay component is
of methane and other reduced trace gases. Primary controlling factors
Georgia Kaolin clay. Experimental measurements conducted for other
for methane emission include vegetation characteristics and presence
clays with much higher specific surface areas A , reveal a strong
of surface water [l]. With the availability of ERS-1 and JERS-1
dependence, particularly for e”, on A,.
and the planned flight of RADARSAT, imaging radar is a candidate
for remote sensing of these controlling factors. It has been shown
REFERENCES that radar is sensitive to the biomass of woody vegetation (e.g.,
M. C. Dobson, F. T. Ulaby, M. T. Hallikainen, and M. A. El-Rayes, trees) for low to moderate biomass [2]. It has also been shown
“Microwave dielectric behavior of wet soil, Part II: Dielectric mixing that radar is sensitive to surface water under trees [3], at least at
models,” IEEE Trans. Geosci. Remote Sensing, vol. GRS-23, pp. 3546, longer wavelengths (e.g., L-band). Most measurements of herbaceous
Jan. 1985. vegetation (e.g., grasses) by radar have focused on agricultural crops
J. J. van Zyl, R. Carande, Y. Lou, T. Miller, and K. Wheeler,
“The NASNJPL three frequency polarimetric AIRSAR system,” in without surface water present. These studies have demonstrated
IGARSS’92 Proc., pp. 649-650. sensitivity to leaf area index (LAI) [4], biomass, and canopy height
M. A. El-Rayes and F. T. Ulaby, “Microwave dielectric spectrum of [5].Radar’s response to herbaceous vegetation in wetlands is less well
vegetation, Part I: Experimental observations,” IEEE Trans. Geosci. documented. Ott et al. [6] presented evidence that multipolarization
Remote Sensing, vol. GRS-25, pp. 541-549, Sept. 1987. SAR can distinguish wetlands vegetation in a mid-latitude setting,
F. T. Ulaby, R. K. Moore, and A. K. Fung, Microwave Remote Sensing,
vol. 3. Dedham, MA: Artech House, 1986, Appendix E. while [7] found similar results for a tropical setting. Reference [8]
M. T. Hallikainen, F. T. Ulaby, M. C. Dobson, M. A. El-Rayes, and demonstrated that ERS-1 SAR backscatter is sensitive to water table
L. Wu, “Microwave dielectric behavior of wet soil, Part I: Empirical position and LA1 for herbaceous Arctic tundra. Here, we examine
models and experimental observations,” IEEE Trans. Geosci. Remore the radar response to LA1 in flooded rice fields as a step toward
Sensing, vol. GRS-23, pp. 25-34, Jan. 1985.
understanding this response in natural herbaceous wetlands.

11. MEASUREMENTS
The data examined here were acquired in 19 flooded rice fields at
mid growing season in central California near the town of Colusa in
July 1991. At each field in situ measurements were made at one or two
sample plots located approximately 10 m from the edge of the field.
Each sample plot had an area of roughly 500 cm’. LA1 was measured
by harvesting the vegetation, measuring the total green leaf area, and
normalizing by the area of the sample plot. Stem number density
was found by counting the total number of stems in the plot and
normalizing by the plot area. Mean plant height was also computed.
Manuscript received May 3, 1994.; revised December 20, 1994. This
work was performed by the Jet Propulsion Laboratory, California Institute
of Technology, under contract with the National Aeronautics and Space
Administration (NASA), and by NASA Ames Research Center. Funding
was provided by the NASA Polar and Ecological Processes and Modeling
Programs.
S. L. Durden is with the Jet Propulsion Laboratory, California Institute of
Technology, Pasadena, CA 91109 USA.
L. A. Momssey and G. P. Livingston are with the Johnson Controls World
Services, NASA Ames Research Center, Moffett Field, CA 94035 USA.
IEEE Log Number 9409909.

019&2892/95$04.00 0 1995 IEEE