IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL.

60, 2022 4210114

Status of the Dielectric Constant of Sea Water at

L-Band for Remote Sensing of Salinity
David M. Le Vine , Life Fellow, IEEE, Roger H. Lang, Life Fellow, IEEE, Yiwen Zhou , Member, IEEE,
Emmanuel P. Dinnat , Senior Member, IEEE, and Thomas Meissner , Senior Member, IEEE

Abstract— The model expressing the dielectric constant of sea the importance of a good model. A major milestone was the
water at microwave frequencies as a function of salinity and tem- development of a model for saltwater by Klein and Swift [3]
perature is an important element in remote sensing of sea surface based on laboratory measurements at L- and S-band and
salinity. It is also important independently as a description of
the physical properties of salt water. A major milestone was the employing a functional dependence on frequency based on the
development in the late 1970s by Klein and Swift of a model based response of polar molecules [4], [5]. Most of the work on the
on laboratory measurements at L- and S-band and a functional dielectric constant of sea water afterward was done to extend
form supported by theory for polar molecules and previous work the model to higher frequencies [6], [7], [8], [9], [10], [11] for
on freshwater. Much of the subsequent work has focused on applications, such as remote sensing of sea surface temperature
measurements at higher frequency and determining model para-
meters tuned to apply for applications, such as remote sensing of (SST), but the development of sensors to measure sea surface
sea surface temperature (SST). Interest in the dielectric constant salinity (SSS) from space, such as soil moisture ocean salinity
at 1.4 GHz (L-band) increased again with the development of soil (SMOS) [12], [13] and Aquarius [14], motivated new work
moisture ocean salinity (SMOS) and Aquarius to measure salinity at L-band, including new measurements [15], [16] and new
from space, but there have been few new measurements at L-band models [15], [17], [18], [19].
and often confusion regarding the applicability of new models
at 1.4 GHz. The objective of this article is to compare available Research on the microwave response of the water mole-
models in the context of how well they represent the dielectric cule [4] established the functional form for the dielectric
constant of sea water at 1.4 GHz. Among the criteria applied will constant, and as a result, all the models are similar in their
be the recent measurements at the George Washington University functional dependence on frequency. Mathematically, they can
of the dielectric constant at 1.4 GHz. all be evaluated at any frequency and they are remarkably
Index Terms— Dielectric constant, L-band, microwave remote similar from 1 to 100 GHz. The result is a plethora of models
sensing, ocean salinity, sea water. with similar features but fitted to measurements covering a
varying range of frequency, temperature, and salinity. To apply
I. I NTRODUCTION the models without regard to the limitation set by the data used
in fitting the model can lead to misrepresentation and error in
A MODEL for the dielectric constant of water as a function
of salinity and temperature is an essential element in
remote sensing of parameters of the ocean surface, such as
the retrieval of science products. This is especially the case
in remote sensing of salinity, where high accuracy is required
salinity and temperature. The beginning of passive microwave on both the radiometer (e.g., 0.1 K) and the model for the
remote sensing of the oceans in the 1970s [1], [2] increased dielectric constant (0.25%) [20].
The objective of this article is to take a close look at the
Manuscript received 29 April 2022; revised 16 August 2022; accepted models for the dielectric constant of sea water in the context
12 September 2022. Date of publication 19 September 2022; date of current
version 7 October 2022. The work of Yiwen Zhou and Roger H. Lang was of remote sensing of SSS and how well they represent the
supported by the National Aeronautics and Space Administration (NASA) dielectric constant of sea water at 1.4 GHz. These are not nec-
through the Physical Oceanography Program under Grant NN17AK01G essarily identical criteria as some models are tuned to optimize
and Grant NNG05GO48G. The work of Emmanuel P. Dinnat was sup-
ported by NASA under Grant 80NSSC22K0215. (Corresponding author: remote sensing, such as [10], [11], and [19], and others are
David M. Le Vine.) tuned to fit laboratory measurements of the dielectric constant
David M. Le Vine is with the Cryospheric Sciences Laboratory, [3], [7], [17]. The success of a model in remote sensing is
NASA Goddard Space Flight Center, Greenbelt, MD 20771 USA (e-mail:
[email protected]). not necessarily a measure of its accuracy representing the
Roger H. Lang is with the Electrical and Computer Engineering Department, dielectric constant of sea water. This is so because empirical
The George Washington University, Washington, DC 20052 USA (e-mail: adjustments in the retrieval algorithm can hide errors in the
[email protected]).
Yiwen Zhou is with the Swiss Federal Institute for Forest, Snow and model for emission from the surface. In the following sections,
Landscape Research WSL, CH-8903 Birmensdorf, Switzerland (e-mail: nine models reported in the literature for the dielectric constant
[email protected]). of sea water will be compared at 1.4 GHz. It will be clear
Emmanuel P. Dinnat is with the Cryospheric Sciences Laboratory, NASA
Goddard Space Flight Center, Greenbelt, MD 20771 USA, and also with the that some of the models should not be used at L-band. The
Center of Excellence in Earth Systems Modeling and Observations, Chapman remainder will be compared with the laboratory measurements
University, Orange, CA 92866 USA (e-mail: [email protected]). at GWU [15], [17] to assess their representation as a function
Thomas Meissner is with Remote Sensing Systems, Santa Rosa, CA 94501
USA (e-mail: [email protected]). of temperature and salinity of the dielectric constant of sea
Digital Object Identifier 10.1109/TGRS.2022.3207944 water. In Section II, a brief introduction will be given to the
1558-0644 © 2022 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: Nanjing University of Information Science and Technology. Downloaded on April 17,2024 at 12:33:35 UTC from IEEE Xplore. Restrictions apply.
 4210114 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 60, 2022

functional form of the models of the dielectric constant of The last term in (1) follows from the definition of current:
sea water. The models are described in Sections III and IV that is, a macroscopic model with the assumption that the
and they are compared as a function of frequency and com- motion of charge is proportional to the electric field with a
pared at 1.4 GHz as a function of salinity and temperature. proportionality constant called conductivity, σ . The definition
A comparison by model component (e.g., conductivity, of “practical salinity” in terms of conductivity established
relaxation time, and static limit) is given in Appendix I. a relationship between salinity, temperature, and conductiv-
In Section V, the models are compared against the laboratory ity [25]. Stogryn et al. [6] inverted this definition to establish
measurements at 1.4 GHz. The implications with respect to the model, σ (S, T ), needed in (1). This inversion has been
remote sensing and the representation of the dielectric constant revisited and is available as public code [47]. The models by
of sea water are discussed in Section VI. Meissner and Wentz [10], [11] and Boutin et al. [19] adopted
this definition, but there is little variation and σ (S, T ) is very
II. BACKGROUND : T HE M ODEL F UNCTION similar for most models (Appendix I).
The parameter ε∞ is mathematically the limiting value at
The dielectric constant of sea water consists of two parts, a f → ∞, but the value in this limit is not established and ε∞
contribution due to orientation and/or distortion of the water is treated as a parameter to be determined. It is perhaps the
molecule (called “polarization” [5], [21]) and a contribution parameter with most variation among the models discussed
due to motion of charge (current). Theory for the polarization here. Most models have values ranging from 3 to 6 but with
of an ideal polar molecule in a viscous (i.e., damping) medium choices ranging from a constant to a function of S and T
was developed by Debye [4] and Somaraju and Trumpf [5], (details are in Appendix I).
and experimental evidence supports this solution for water In the following sections, the models will be compared.
[22], [23]. All the models discussed here employ this form for In Section III, they are compared as a function of frequency
the frequency dependence. About half the models examined and then at 1.4 GHz compared as a function of salinity and
here employ one Debye “resonance” and half employ a second temperature. In Section IV, the models will be compared to
resonance as suggested by Stogryn et al. [6] to get a better laboratory measurements of the dielectric constant at 1.4 GHz
fit at higher frequency. Adding salt to water makes the liquid [15], [17]. The component parts of each model (i.e., relaxation
conducting, necessitating an additional term to account for the time, static term, conductivity, and high frequency limit) are
current. The model with two resonances has the following compared as functions of S and T in Appendix I.
form:
εs − ε1 ε1 − ε∞ σ III. M ODELS TO B E C OMPARED
ε = ε∞ + + −j (1) The models to be compared are listed in the following in two
1 + j ωτ1 1 + j ωτ2 ωεo
categories, those composed of a single Debye resonance and
where f is the frequency, εo is the permittivity of vacuum, those using two resonant terms. Generally, those composed
and the other parameters, such as conductivity, σ , relaxation of a single resonance are designed for use near 1.4 GHz and
times, τ1 and τ2 , and static limit, εs , are to be determined from those with two resonances were intended for use at higher
measurements. In those models with a single resonance, ε1 is microwave frequencies. However, there are exceptions, such
replaced by ε∞ . as FASTEM [18] and the model of Meissner and Wentz
Measurements suggest a relaxation time for the first res- [10], [11], which use two resonances and are applicable
onance near τ1 = 0.05 ns (near 20 GHz) depending on at 1.4 GHz, and the model by Guillou et al. [9], which
temperature [22], [24]. Models employing a second resonance includes a single resonance and was intended for use at higher
have resonant frequencies above 100 GHz and for applications frequencies.
at the low end of the microwave spectrum, for example
at 1.4 GHz for remote sensing of SSS, the first resonance A. Models With a Single Resonance
is dominant. Most models intended for use specifically near KS: Klein and Swift [3]: Based on laboratory measurements
1.4 GHz only include the first Debye resonance term. at 1.43 and 2.65 GHz using a reflection cavity technique by
Since the model for the dielectric constant should reduce to Ho et al. [26] and Ho and Hall [27]. The data cover a range
fresh water when S = 0, it can be argued that the static term, of approximately 5 < T < 30 ◦ C and 4 < S < 35 psu.
εs , can be written in the following form: BA: Blanch and Aguasca [16]: Based on laboratory mea-
εs = A(T ){1 − B(T )S} (2) surements at 1.4 GHz over the range of 0.5 < f < 2.5 GHz
using a transmission line method and with salinity and temper-
where A(T ) and B(T ) are the functions to be determined. ature covering the range 0 < S < 40 psu and 0 < T < 38 ◦ C,
This form was used by Guillou et al. [9] and adopted by the later in steps of 7 ◦ C.
Boutin et al. [19] as suggested by Somaraju and Trumpf [5]. GW: Zhou et al. [17]: Based on laboratory measurements at
The form was also adopted in the recent model of 1.4 GHz using a resonant cavity technique at 1.413 GHz and
Zhou et al. [17] but with B dependent on salinity covering the range 0 < S < 38 psu and −1.5 < T < 35 ◦ C.
[i.e., B = B(S, T )]. Zhou et al. [17] found that using this This is an extension of earlier work, which fit a polynomial
constraint reduced the number of unknowns and resulted in to the measurements [28].
a better fit than an earlier model, which attempted a third- BV: Boutin et al. [19]: Uses the functional form in (2)
order polynomial in S and T for the static term [15]. for the static term with A(T ) the freshwater value and

Authorized licensed use limited to: Nanjing University of Information Science and Technology. Downloaded on April 17,2024 at 12:33:35 UTC from IEEE Xplore. Restrictions apply.
 LE VINE et al.: STATUS OF THE DIELECTRIC CONSTANT OF SEA WATER AT L-BAND 4210114

B(T ) tuned to optimize the retrieval from SMOS measured

brightness temperature. The conductivity is as proposed by
Stogryn et al. [6] and McDougall et al. [47], and the relaxation
time, A(T ) and the high frequency limit are the S = 0 values
as given by Meissner and Wentz [10], [11]. Tuned to apply at
1.4 GHz and 0 < T < 30 ◦ C and 32 < SSS < 38 psu.
EL: Guillou et al. [9] and Ellison et al. [7]: Based on
measurements from 3 to 20 GHz in 0.85-GHz steps and
covering −2 < T < 30 ◦ C in 1◦ steps for each of six salinities
23.2, 28.0, 30.024, 35.0, 38.024, and 38.893 psu also with
selected measurements in this range at 23.8, 36.5, and 89 GHz.
Nominal range for model fit: 3 < f < 20 GHz, −2 < T <
30 ◦ C, and 20 < S < 40 psu.

B. Models With Two Resonances

MW: Meissner and Wentz [10], [11]: Uses the conductivity
given by Stogryn et al. [6] and the measurements of Ho et al.
[26] and Ho and Hall [27] at L- and S-band and Guillou et
al. [9] at 85 GHz; and the dielectric constant of fresh water
is fitted to laboratory measurements in the frequency range
1.7–410 GHz [29], [30], [31], [32]. The relaxation time and
other parameters have been tuned to improve the fit to special
sensor microwave/imager (SSM/I) and Windsat measured TB
especially near 37 GHz [33], [34]. The range of application
is 0 < S < 40 psu, 2 < T < 29 ◦ C, and frequencies up
to 90 GHz.
FM: FASTEM-4: Liu et al. [18]: Input data cover the range
−2 < T < 30 ◦ C and 0 < S < 35 psu and 1.4 < f <
410 GHz. The model parameters are fitted using measurements
of the dielectric constant as follows: 1) the measurements at
23.8, 36.5, and 89 GHz for a constant salinity of 38.89 psu
and water temperatures of −2 ◦ C, 12 ◦ C, 20 ◦ C, and 30 ◦ C as Fig. 1. Brightness temperature predicted by each model versus frequency at
nadir with S = 35 psu and T = 20 ◦ C. (Top) Full scale and (Bottom) with
reported by Ellison et al. [7]; 2) the measurements at 1.43 and expanded scale to show detail below 5 GHz.
2.65 GHz reported by Ho et al. [26] and Ho and Hall [27] used
in the KS model; 3) measurements at 35 psu at frequencies
of 30–105 GHz and at temperatures of −2 ◦ C and 5 < T < the dielectric constant and brightness temperature predicted
30 ◦ C in steps of 5 ◦ C as reported by Lamkaouchi et al. [35]; by models are very similar as a function of frequency. This
and 4) the dielectric constant of fresh water over the frequency is illustrated in Figs. 1 and 2. Fig. 1 shows the brightness
range of 1.7–410 GHz reported by Kaatze and Uhlendorf [29], temperature predicted by the models as a function of frequency
Hasted et al. [32], and Ellison et al. [36]. at S = 35 psu and T = 20 ◦ C and nadir incidence angle. The
EL2: Ellison [8]: Proposed for range: 0 < T < 30 ◦ C, frequency range is 0 < f < 100 GHz on the top and with
0 < SSS < 40 psu, and 0 < freq < 500 GHz for sea water. better resolution for f < 5 GHz on the bottom. Except for
This article provides a good review of past work, but the two model BA above 10 GHz, there is not a great difference at
models provided (“best fit” and “practical”) are out of bounds this resolution among the models, although the models with
compared to the other models above and it was decided not two Debye resonant terms (ST, MW, and FM) tend to predict
to include them here. Uses the Stogryn et al. [6] model for lower TB for f > 20 GHz. Fig. 2 shows the real and imaginary
conductivity. parts of the dielectric constant predicted by the eight models
ST: Stogryn et al. [6]: Based on measurements from 7 to 14 as a function of frequency for S = 35 psu and T = 20 ◦ C
GHz in 1-GHz steps and salinities of 0, 2.09, 3.92, 7.17, 11.2, again with full scale in the top panel and with an expanded
15.46, 20.14, 22.47, 31.68, and 35.96 psu and temperatures 0 scale for f < 5 GHz in the bottom panel. On the bottom, the
< T < 30 ◦ C in 5◦ steps. Inverts the definition of salinity [25] BA model (solid green) stands out from the others. At lower
to obtain conductivity as a function of S and T . frequencies (bottom panel), there are noticeable differences in
IV. C OMPARISON the real part of the models.
A. Dependence on Frequency
The models for the dielectric constant of sea water con- B. Behavior at 1.4 GHz
sidered here all have the functional form given by (1) with When the focus is on 1.4 GHz, significant differences are
one or two resonances. Consequently, it is not surprising that evident in both the brightness temperature predicted by the

Authorized licensed use limited to: Nanjing University of Information Science and Technology. Downloaded on April 17,2024 at 12:33:35 UTC from IEEE Xplore. Restrictions apply.
 4210114 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 60, 2022

Fig. 2. Each panel shows the real part (top) and imaginary part (bottom) of
the models versus frequency. (Top) Full scale and (Bottom) with expanded Fig. 3. Brightness temperature at 1.4 GHz for nadir. (Top) As a function of
frequency scale to show detail below 5 GHz. temperature for S = 35 psu. (Bottom) As a function of salinity for T = 20 ◦ C.

models and in the dielectric constant itself. This is illustrated the models (the major exceptions are models EL and BA:
in Figs. 3–5. see Appendix I).
Fig. 3 shows the brightness temperature, TB, at nadir pre- For remote sensing of salinity, an accuracy of TB of
dicted by each model at 1.4 GHz as a function of temperature about 0.1 K is required to obtain an accuracy of SSS of
(top) for S = 35 psu and as a function of salinity (bottom) 0.2 psu depending on temperature and incidence angle (see
for T = 20 ◦ C. The models differ much more as a function [37, Fig. 1]). It is clear from the top panel in Fig. 3 that
of temperature than as a function of salinity. The salinity several of the models (e.g., BA, ST, and EL) stand apart from
dependence is primarily in the conductivity, σ , and static the others at this level of accuracy. These three models also
terms, εs , which are compared in Appendix I. The conductivity appear as outliers in Fig. 4 (top panel) for their dependence
(Appendix I, Fig. 8) is almost identical among the models, and on temperature (i.e., real part of ST and EL and imaginary
the dependence of εs on salinity is similar and approximately part of BA) and in Fig. 4 (bottom panel) for the dependence
linear for most models (Appendix I, Fig. 9, bottom). The on salinity of the real part [i.e., ST (red) and EL (green dash)
relative dependence on S and T shown in Fig. 3 at nadir models]. The ST and EL models are based on measurements
is independent of polarization and incidence angle. This is at higher frequencies and L-band is out of the range of the
illustrated in Appendix II, where the brightness temperature is data used in the model fit. It is not reasonable to expect them
shown as a function of salinity and temperature at an incidence to fit at 1.4 GHz. Consequently, these two models will be
angle of 40◦ . dropped from further analysis, since the focus is on models
Fig. 4 shows the same information for the dielectric con- representing the dielectric constant at 1.4 GHz. On the other
stant. On the top is shown the real and imaginary parts of the hand, the BA model (green solid) was based on measurements
dielectric constant at S = 35 psu as a function of T and on at 1.4 GHz and intended for use in remote sensing of salinity,
the bottom is the dielectric constant for the eight models as but Figs. 1–4 suggest something is amiss. This is especially
a function of salinity for T = 20 ◦ C. The greatest difference obvious in Fig. 3 (top), which shows the brightness tempera-
among the models is in the real part of the dielectric constant. ture predicted by this model as a function of temperature. It is
The imaginary parts are very similar because σ (S, T ) is almost also clear from the examination of the component parts of the
identical and the salinity dependence in εs is similar among models (Appendix I) that the model for conductivity and the

Authorized licensed use limited to: Nanjing University of Information Science and Technology. Downloaded on April 17,2024 at 12:33:35 UTC from IEEE Xplore. Restrictions apply.
 LE VINE et al.: STATUS OF THE DIELECTRIC CONSTANT OF SEA WATER AT L-BAND 4210114

Fig. 4. Comparison at L-band of (Top) real part and (Bottom) imaginary Fig. 5. Comparison at L-band of the brightness temperature predicted at
part of the dielectric constant for all models. (Top) Comparison as a function nadir and 1.4 GHz by selected models. (Top) Comparison as a function of
of temperature for S = 35 psu. (Bottom) Comparison as a function of salinity temperature for S = 35 psu. (Bottom) Comparison as a function of salinity
for T = 20 ◦ C. for T = 20 ◦ C.

relaxation time associated with the Debye resonance term in To get a better quantitative assessment of how well these
this model are unusual. Consequently, this model will also be models represent the dielectric constant of sea water at L-band,
dropped from further analysis. they will be compared in Section V below with laboratory
Fig. 5 shows the brightness temperature as a function of measurements of the dielectric constant.
temperature (top) and salinity (bottom) for the remaining
models. Each of these models advertises applicability at V. C OMPARISON W ITH M EASUREMENTS AT 1.4 GHz
L-band. Differences in the dependence on temperature (top) While there has been much work on the dielectric constant
exist that are well in excess of the 0.1-K figure of merit for of pure water and several relatively recent reports of measure-
remote sensing of salinity (see Fig. 16 in Appendix III). The ments of the dielectric constant of sea water at frequencies
differences are greatest for T < 20 ◦ C and smallest near 26 ◦ C. associated with remote sensing of SST [6], [7], [9], the
There is much more uniformity in the variation with salinity measurements at 1.4 GHz are very limited (see [8], [15]
(Fig. 5, bottom) and the agreement is good for all values of for a review). The measurements at 1.4 GHz include the
salinity. A figure with expanded resolution for salinity typical measurements by Ho and Hall [27] used by Klein and Swift [3]
of the open ocean, 32 < S < 38 psu, is shown in Appendix III, with measurements at 2.85 GHz [26] to develop the KS model.
where it can be seen that the differences in the dependence on More recently, measurements were made at the Universitat
S are on the order 0.1 K. Politècnica de Catalunya (UPC) and used to develop the BA
Brightness temperature is important because it is the para- model [16]. Unfortunately, the only public record of this work
meter measured by satellite remote sensing systems, but it is is the article by Blanch and Agusca [16]. The data were not
not necessarily a good indicator of how well the model actually published and the model appears to have issues as discussed
represents the dielectric constant of sea water. This is because above.
the relationship between the dielectric constant and TB is not The development of radiometers at 1.4 GHz for remote
unique and given TB one cannot uniquely determine both sensing of salinity from satellite missions, such as Aquarius/
the real and imaginary parts of the dielectric constant [38]. SAC-D and SMOS, raised the visibility of the need for

Authorized licensed use limited to: Nanjing University of Information Science and Technology. Downloaded on April 17,2024 at 12:33:35 UTC from IEEE Xplore. Restrictions apply.
 4210114 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 60, 2022

visualization of the comparison is to compare with an interpo-

lation function that represents this data. This introduces addi-
tional error and the accuracy of the fit, but it permits smooth
curves that allow one to see patterns in how the models com-
pare with the data as a function of temperature and salinity.
The obvious interpolation function for this data is the model
developed by the GWU team [17], which is a best fit to the
data. This is the model GW listed in Section III and discussed
above. It is a fit to the data using the functional form of (1)
with a polynomial in S and T for the unknowns, εs and σ . The
relaxation time was assumed to be the same as for freshwater
and fitted to the measurements (red circles in Fig. 6). The
high-frequency limit was set at 4.9 as recommended by Klein
and Swift [3]. Fortunately, a careful analysis of the accuracy of
the measurements in 2016 was documented [15]. It is assumed
that this is representative although improvements were made
leading to the measurements reported in [17]. In the case
of the real part, the standard deviation (STD) of the total
measurement error is 0.23, and for the imaginary part, the
STD of the total measurement error is 0.26. The STD of
the difference between the GW model and the data at each
point is 0.11 for the real part and 0.29 for the imaginary
part. Assuming that measurement and representation errors are
independent, the net error for representation (square root of the
sum of the squares) is 0.25 for the real part and 0.36 for the
imaginary part. Differences between a particular model under
consideration here and the GW interpolation that are within
one STD (i.e., ±0.25 for the real part) will be used here to
suggest good agreement with the data. The differences are
reported in Fig. 7.
Fig. 6. Measurement of the dielectric constant of sea water as a function Fig. 7 shows the difference between GW and the mod-
of temperature (Top) and salinity (Bottom). The data are from [Lang et al.,
2015; Table V] and [17, Tables I–III]. Freshwater (S = 0) in red.
els, BV, KS, MW, and FM as a function of temperature
for SSS = 35 psu (top) and as a function of salinity with
SST = 20 ◦ C (bottom). The green dashed horizontal lines
accurate measurements of the dielectric constant of sea water indicate the STD of the combined error of the interpolation
at 1.4 GHz [20], [39]. In response, research at The George (fit of GW to the data) and the data itself. In general, the fits are
Washington University was started to make a comprehensive better for the imaginary parts than for the real parts and better
set of measurements at 1.4 GHz [15], [17], [28]. These mea- as a function of salinity than as a function of temperature.
surements comprise a wide range of salinity (0 < S < 38 psu) As mentioned above, this is a consequence of a great deal of
and temperature (−1.5 < T < 30 ◦ C) and with an emphasis similarity in how the salinity dependence is included among
on values likely to be encountered in the open ocean. This the models. In particular, the BV, MW, and FM model all
is a contemporary set of measurements that used a proven use the Stogryn [6] inversion of the definition of salinity
technique (resonant cavity) and takes advantage of modern or equivalent (see [47]). The GW and KS versions, which
microwave instrumentation to achieve high accuracy [15]. were determined independently using polynomial fits to their
Measurements were made at fixed salinity as a function of respective data, are almost identical (Appendix I). The largest
temperature at discrete temperatures and salinity. The data variation among the models is in the behavior of the real part
are shown in Fig. 6 as a function of temperature (top) and as a function of temperature. This reflects the differences in
as a function of salinity (bottom). The spread in the data in how the models included T in the parameterization of the
the top panel represents the dependence of the dielectric on static term, εs .
salinity at each of the temperatures at which measurements The best fit of all appears with KS, which also happens
were made and in the bottom panel represents the dependence to be the other model based on laboratory measurements at
on temperature for each salinity used in the measurements. The L-band. The KS model is consistently within the error bounds
measured values of dielectric constant are in [15, Table V] in all four panels except for the real part at temperatures
and [17, Tables I–III]. The complete dataset is available below 5 ◦ C and above 30 ◦ C (top panel), which are tempera-
online at [40]. tures out of the range of the measurements used in developing
Rather than comparing the models for the dielectric constant the KS model. At temperatures and salinities in the range of
point by point with the data, an alternative that yields a better 20 < T < 30 ◦ C and 30 < S < 40 psu, the FM model is

Authorized licensed use limited to: Nanjing University of Information Science and Technology. Downloaded on April 17,2024 at 12:33:35 UTC from IEEE Xplore. Restrictions apply.
 LE VINE et al.: STATUS OF THE DIELECTRIC CONSTANT OF SEA WATER AT L-BAND 4210114

temperature dependence of the static term [term A(T ) in (2)],

which is strongly coupled to the real part. The MW model
is a double Debye resonance model intended to cover a large
range of higher frequencies and the tuning was done to fit the
satellite observations from Windsat and SSM/I in this higher
range and particularly near 37 GHz. The FM model, which
is also a double resonance model intended for use over an
extended range of frequency, uses data at several frequencies
from several sources. The fit at 1.4 GHz is mixed. It fits the
real part well as a function of temperature at 35 psu (Fig. 7,
top panel) but not as well for S < 30 (Fig. 7, bottom panel).
The imaginary part agrees with the measurements well for
20 ◦ C and above (Fig. 7, bottom panel) but not as well at
lower temperatures.
Three of the models in Fig. 7 have been used in remote
sensing of salinity: KS and BV have been used in SMOS
retrievals and MW was used for Aquarius and is now used for
retrieving salinity from soil moisture active passive (SMAP).
In addition, a combination of the MW and KS model has
been used in the combined active passive (CAP) algorithm for
retrievals from Aquarius and SMAP [41]. These models have
been used quite successfully to retrieve salinity. However, this
success does not guarantee that they are a good representation
of the dielectric constant of sea water. For example, the MW
model was used quite successfully to retrieve salinity from
both Aquarius and SMAP [42], but this does not mean it is
the best representation of the dielectric constant of sea water.
On the contrary, except for cold temperatures, KS appears to
be a better fit to the dielectric constant at 1.4 GHz.
Fig. 7. Comparison of models for the dielectric constant with the GW model
The contradiction lies in the realities of the retrieval
as a function of (Top) temperature and (Bottom) salinity. algorithm. There are many inputs to the retrieval algorithm
each with associated uncertainty that require adjustments. For
example, the algorithm starts with the best available models
also consistently within the error bounds. This model used for the sensor antenna and for propagation from the surface
in its development the same measurements at 1.4 GHz by to the sensor and makes adjustment to improve the retrieval
Ho et al. [26] as used for developing the KS model. of the science product via iteration with surface truth. This
process can hide deficiencies in models adopted as “known”
VI. D ISCUSSION input. For example, an SST-dependent bias was noticed in one
Perhaps, it is not surprising that the two models based on of the early versions (Version 3.0) of the retrieval of SSS from
laboratory measurements of the dielectric constant, KS and Aquarius [43]. This was corrected with an adjustment to the
GW, should be close. The two sets of measurements are surface emissivity used in the retrieval algorithm (see [42]).
completely independent and so are the model fits, although the The result was a retrieval in excellent agreement with surface
two models use the same functional form [(1)] and the same truth and little SST dependent bias, but the source of the error
high-frequency limit, ε∞ = 4.9 (Appendix I). The agreement was not identified. It was later shown that switching models
of these two models is a positive indication for both the quality for the dielectric constant to KS reduced this bias [43]. The
of the measurements and the quality of the data fits (but choice of models may well be the source of this error, but it
see [15] for more information about the Ho and Hall [27] was not necessary to make a change because of adjustments
data). made in the retrieval algorithm.
In the two models, BV and MW, parameters were tuned to The example above is a reminder that the success of
optimize the science retrieval algorithm by choosing the model Aquarius, SMAP, and SMOS in retrieving accurate values of
parameters to minimize the differences between radiative SSS does not automatically assure that the model used in
transfer simulations of the brightness temperature and the TB the algorithm for the dielectric constant is accurate. On the
actually observed by the satellite. These two models are close other hand, the more accurate the model, the more likely the
to the measurements but diverge in some significant respects. retrieval is to be successful and the more potential there is for
This is particularly evident in the temperature dependence improvement of the retrieved SSS. For example, suppose one
of the real part (Fig. 7, top). The BV model was tuned to had a perfect model for the dielectric constant and encountered
optimize the retrieval of salinity from SMOS observations of a temperature-dependent bias in the retrieval as occurred
brightness temperature at 1.4 GHz. The tuning was done on the with Aquarius. In this case, having the correct model would

Authorized licensed use limited to: Nanjing University of Information Science and Technology. Downloaded on April 17,2024 at 12:33:35 UTC from IEEE Xplore. Restrictions apply.
 4210114 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 60, 2022

Fig. 9. Comparison of GW and KS conductivity, σ (S, T ) with the Stogryn,

Fig. 8. Conductivity as a function of temperature and salinity. (Top) Conduc- ST, model. (Top): Difference. (Middle): Comparison. (Bottom): Slope dS/dσ
tivity at S = 20 psu as a function of temperature. (Bottom) Conductivity at in PSU/(S/m) for several values of SST.
T = 25 ◦ C as a function of salinity. GW—black dash; ST and MW—red;
KS—black dashed dotted; EL—green dash; BV—black; BA—green; and
FM—red dashed dotted.
suggesting that a wideband radiometer operating in the range
0.5–2.0 GHz could enhance the accuracy of the retrieval of
eliminate this part of the algorithm as a potential source SSS especially in cold water by including frequencies closer
of error and would direct the research for the root cause to the peak [44], [45]. Simply applying existing models to
elsewhere. The result likely would be a better understanding these frequencies does not guarantee their validity, and new
of errors in the retrieval algorithm itself. measurements will be needed to determine the coefficients of
The search for a model for the dielectric constant of sea models suitable for this range of frequencies with the accuracy
water at 1.4 GHz accurate enough to promote improvements needed to retrieve SSS. Work to address this problem has
in the retrieval of is not yet complete. There are at least already begun [46].
two challenges. One is that making measurements that are In the meantime, a problem with many models is the use
consistent with an accuracy of the salinity product of better of the models outside of their range of validity. There are no
than 0.2 psu is very hard. For example, at nadir, an accuracy physical restrictions, which prevent using any of these models
of 0.2 psu corresponds to radiometric accuracy of 0. 1 K at any frequency, salinity, or temperature, but all the models
for a measurement at 1.4 GHz (see [37, Fig. 4]). Assum- discussed here are based on measurements of a finite range in
ing equal error, , in the real and imaginary parts of the S and T and use mathematical functions (usually polynomials)
dielectric constant, an accuracy of about  = 0.25% in the to fit the unknown parameters to the data in this range. The
measurement of the dielectric constant is required (at 35 psu fits are unconstrained outside of the range of the data. This
and 25 ◦ C) to have an error of less than 0.1 K in TB. The is evident in the case of the KS model. As can be seen in
current measurement accuracy of the GW measurements at this Fig. 7 (top panel), the real part diverges strongly from the
temperature and salinity is about 0.35% [15]. So, there is yet a measurements for high (T > 30 ◦ C) and low temperatures
need for improvement, and if the goal is eventually to achieve (T < 5 ◦ C). Using this model for retrievals at low temperature
0.1 psu, even more progress is needed. The second issue is results in large errors, which have been well documented [43].
the possibility that remote sensing of salinity in the future Another obvious example is the EL and ST models, which
will become a multispectral process. The peak sensitivity to were built on measurements at frequency above L-band
SSS lies below 1.4 GHz [37], and research has been reported (minimum frequency of 3 and 7 GHz, respectively).

Authorized licensed use limited to: Nanjing University of Information Science and Technology. Downloaded on April 17,2024 at 12:33:35 UTC from IEEE Xplore. Restrictions apply.
 LE VINE et al.: STATUS OF THE DIELECTRIC CONSTANT OF SEA WATER AT L-BAND 4210114

Fig. 10. Static term as a function of (Top) temperature for S = 35 psu

(Bottom) and as a function of salinity T = 20 ◦ C. Fig. 11. First relaxation time as a function of temperature for (Top) freshwater
and (Bottom) for S = 35 psu.

Mathematically, there is no problem applying them at 1.4 GHz.

However, they are not representative of the brightness temper- on top as a function of T for S = 35 psu and on the bottom as
ature of the ocean surface at 1.4 GHz as the examples in a function of S for T = 20 ◦ C for each of the models. With
Figs. 3 and 4 show. the exception of EL (green dash) and BA (green), there is
very good agreement among the models. The models ST, MW,
A PPENDIX I and BV use the inversion from the definition for conductivity.
All the models discussed in this article have the form of (1), The KS and GW models use a polynomial in S and T
which has several common parameters that are determined to represent σ (S, T ) with coefficients that are determined
by fitting the model to observation of the real world. In this from their respective measurements of the dielectric constant
Appendix, the parameters, conductivity, static dielectric con- (i.e., see [26], [27] in the case of KS and [15], [17] in
stant, the relaxation time of the first resonance, and the high- the case of GW). The FM model uses the KS expression
frequency limit will be compared. for conductivity. The two polynomial fits are in very good
agreement with the definition. This is illustrated in Fig. 9,
which shows the KS and GW expressions for conductivity
A. Conductivity, σ (S, T) together with the expression derived by Stogryn et al. [6] and
With the establishment of the practical salinity scale [25], used in ST as a function of salinity in the top two panels.
the definition of salinity was tied to conductivity. Salinity was On the top, the differences ST–KS and ST–GW are shown
defined as a ratio of measured conductivity to that of a standard to make the difference more visible. To put these differences
solution at a 15 ◦ C and experiments provided conversions in context, the change in salinity with conductivity, dS/dσ ,
for temperature and pressure [25, eq. (9)]. Stogryn et al. [6] is plotted in the bottom panel for T = 25 ◦ C. A difference of
reported an inversion of this definition to produce an expres- about 0.03 S/M corresponds to a salinity difference of about
sion for σ (S, T ), which is equivalent to a more recent ver- 0.2 psu at 35 psu and this temperature (25 ◦ C) and less with
sion [47]. Most models adhere closely to this definition. This a decrease in salinity. Using the goal of contemporary remote
is illustrated in Fig. 8, which shows the conductivity, σ (S, T ), sensing of SSS from space of an accuracy of 0.2 psu as

Authorized licensed use limited to: Nanjing University of Information Science and Technology. Downloaded on April 17,2024 at 12:33:35 UTC from IEEE Xplore. Restrictions apply.
 4210114 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 60, 2022

Fig. 12. High-frequency limit. As a function of temperature for (Top) S = 0 Fig. 13. Brightness temperature at 1.4 GHz and 40◦ incidence angle as a
and (Bottom) S = 35. Only MW depends on salinity. GW, BA, and KS all function of temperature with (Top) S = 35 psu and as a function of salinity
use the constant value = 4.9. with (Bottom) T = 20 ◦ C. In each panel, horizontal polarization is on the
top and vertical polarization is on the bottom.
a metric, the models for conductivity determined by model fits
to measurements are in good agreement with the expression associated with the second resonance is around 100 GHz and
obtained by inverting the definition of salinity (which is a this term is not important for application near L-band. The
reassuring result for both measurements and inversion of the first relaxation time, τ1 , for all the models is close to the
definition). freshwater value with a weak or no dependence on salinity.
B. Static Term, εs This is illustrated in Fig. 11, in which τ1 is plotted as a function
of temperature for freshwater (S = 0) on top and for SSS =
The static term, εs , is shown in Fig. 10 as a function of T 35 psu on the bottom. In the case of fresh water, all the models
for S = 35 psu (top) and as a function of S for T = 20 ◦ C except for BA (green dash) are close together and have a
on the bottom. The model, EL, is clearly an outlier and does
similar dependence on temperature. The resonant frequency at
not converge to the same value for freshwater (S = 0, bottom 20 ◦ C is about 17 GHz and varies with temperature increasing
panel) as all the other models. Its temperature dependence to about 30 GHz at 40 ◦ C and decreasing to about 10 GHz
at S = 0 (not shown here) is also much different than that at 0 ◦ C. All the models assume a weak dependence on salinity
of the other models. The variation with both salinity and except for BV and GW, in which τ is independent of S. This
temperature is approximately linear over the range shown in
can be seen by comparing the curves in the panel on the top
Fig. 10, although the BA model (green solid) has a noticeable (S = 0) with those in the bottom panel of Fig. 11, which
curvature in its dependence on salinity (bottom panel). The KS shows τ1 as a function of temperature for S = 35 psu.
model is an outlier at very cold and very warm temperatures
The dependence on salinity in those models that include a
(dashed-dotted curve, top panel), but the KS model was based dependence on SSS is very weak for T > 20 ◦ C and increases
on measurements for 5 ◦ C < T < 30 ◦ C, and there is no reason at lower temperatures as can be seen in the spread of the curves
to expect the polynomials fitted to the data to extrapolate well in the bottom panel of Fig. 11.
to temperatures out of this range.

C. Relaxation Time, τ D. High-Frequency Limit, ε∞

Comparison is made here only for the relaxation time, τ1 , The high-frequency limit, ε∞ , is shown in Fig. 12 as a
of the first Debye resonance. The resonant frequency, τ2 , function of SST for freshwater (S = 0) on the top and for sea

Authorized licensed use limited to: Nanjing University of Information Science and Technology. Downloaded on April 17,2024 at 12:33:35 UTC from IEEE Xplore. Restrictions apply.
 LE VINE et al.: STATUS OF THE DIELECTRIC CONSTANT OF SEA WATER AT L-BAND 4210114

Fig. 15. Dependence of brightness temperature, TB, at 1.4 GHz and nadir
Fig. 14. Same as Fig. 13 but with models ST, EL, and BA removed to permit incidence as a function of salinity with expanded scale to show differences.
enhanced resolution. Brightness temperature at 1.4 GHz and 40◦ incidence The differences among the models depend on the value of salinity: (Top) for
angle as a function of temperature (Top) with S = 35 psu and as a function 32 < S < 38 and (Bottom) for 12 < S < 18. The vertical and horizontal scales
of salinity with (Bottom) T = 20 ◦ C. In each panel, horizontal polarization have the same resolution in each panel to facilitate comparison.
is on the top and vertical polarization is on the bottom.

shown in Fig. 12 (top) with the exception of EL (green dash),

which is an outlier.
water (SSS = 35 psu) on the bottom. Only the MW model When frequency f  1.4 GHz, the infinite frequency limit,
(red dash) includes a dependence on salinity. In the models ε∞ , has negligible effect on the dielectric constant. In this
using a single Debye term (KS, GW, BA, BV, and EL), case, ωτ2 < ωτ1  1, and both ε∞ and ε1 cancel themselves
ε∞ ≡ ε1 . Three models, KS, GW, and BA, assume a constant in (1). The effect of ε∞ is also small at 1.4 GHz, although not
value, ε∞ = 4.9, which was suggested by Klein and Swift [3]. necessarily negligible. For example, using the KS model as an
With the exception of EL (green dash), all of other models example and varying ε∞ over the range 3.5 < ε∞ < 6, which
have values close to this but with a small dependence on covers the range of values appearing in the models discussed
temperature. here (except for EL), then the change in the imaginary part
There appears to be no consensus as to what the dielectric (where the change is greatest) is about 0.3%. This is small, but
constant should be at very high frequency, and this parameter this is comparable to the accuracy needed to retrieve salinity
plays the role of a free parameter to be adjusted to improve at an accuracy of 0.2 psu (see Section VI).
the fit. The choice made by model builders appears to depend
on the functional form (one or two Debye terms) chosen for A PPENDIX II
the dielectric constant. For example, the red curves correspond Fig. 13 shows the brightness temperature predicted at
to models (ST, MW, and FM) that employ two Debye reso- 1.4 GHz by each of the models for the dielectric constant at
nance. On the other hand, the constant, 4.9, and black curve 40◦ incidence angle. The brightness temperature at S = 35 psu
(BV) correspond to models using a single Debye resonance. is shown as a function of temperature on the top and TB is
A similar partition depending on the number of Debye terms shown at T = 20◦ C as a function of salinity on the bottom.
is reported by Liebe et al [24] in the case of freshwater. The Horizontal polarization is on the top and vertical polarization
values for ε∞ reported by Liebe are consistent with those on the bottom in both panels. Fig. 14 shows the same as

Authorized licensed use limited to: Nanjing University of Information Science and Technology. Downloaded on April 17,2024 at 12:33:35 UTC from IEEE Xplore. Restrictions apply.
 4210114 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 60, 2022

[6] A. P. Stogryn, H. T. Bull, K. Rubayi, and S. Iravanchy, “The microwave

permittivity of sea and fresh water,” Gen. Corp Aerojet, Azusa, CA,
USA, Tech. Rep., 1995.
[7] W. Ellison et al., “New permittivity measurements of seawater,”
Radio Sci., vol. 33, no. 3, pp. 639–648, May/Jun. 1998, doi:
10.1029/97RS02223.
[8] W. Ellison, “Fresh and sea water, section 5.2,” in Thermal Microwave
Radiation: Applications for Remote Sensing (IET Electromagnetic Wave
Series), vol. 52, Matzler, Ed. London, U.K.: Institution of Engineering
and Technology, 2006.
[9] C. Guillou et al., “Impact of new permittivity measurements on sea
surface emissivity modeling in microwaves,” Radio Sci., vol. 33, no. 3,
pp. 649–667, May 1998, doi: 10.1029/97rs02744.
[10] T. Meissner and F. J. Wentz, “The complex dielectric constant of pure
and sea water from microwave satellite observations,” IEEE Trans.
Geosci. Remote Sens., vol. 42, no. 9, pp. 1836–1849, Sep. 2004.
[11] T. Meissner and F. J. Wentz, “The emissivity of the ocean surface
between 6 and 90 GHz over a large range of wind speeds and earth
Fig. 16. Difference in brightness temperature at 1.4 GHz between each model incidence angles,” IEEE Trans. Geosci. Remote Sens., vol. 50, no. 8,
and the GW model (TBmodel − TBGW ) as a function of temperature for nadir pp. 3004–3026, Aug. 2012.
incidence and SSS = 35 psu. [12] S. Mecklenburg et al., “ESA’s soil moisture and oceansalinity mis-
sion: Mission performance and operations,” IEEE Trans. Geosci.
Remote Sens., vol. 50, no. 5, pp. 1354–1366, May 2012, doi:
10.1109/TGRS.2012.2187666.
Fig. 13 but with models ST, EL, and BA removed to permit [13] Y. H. Kerr et al., “The SMOS mission: New tool for monitoring
enhanced resolution. key elements of the global water cycle,” Proc. IEEE, vol. 98, no. 5,
pp. 666–687, May 2010.
[14] D. M. L. Vine, G. S. E. Lagerloef, F. R. Colomb, S. H. Yueh, and
F. A. Pellerano, “Aquarius: An instrument to monitor sea surface
A PPENDIX III salinity from space,” IEEE Trans. Geosci. Remote Sens., vol. 45, no. 7,
Figs. 15 and 16 in this Appendix provide additional detail pp. 2040–2050, Jul. 2007.
[15] R. Lang, Y. Zhou, C. Utku, and D. L. Vine, “Accurate measurements of
for the comparison of the models shown in Figs. 3 and 5. The the dielectric constant of seawater at L band,” Radio Sci., vol. 51, no. 1,
two panels in Fig. 15 are the same as Fig. 5 (bottom) in the pp. 2–24, Jan. 2016, doi: 10.1002/2015RS005776.
main text but with an expanded scale to show the differences [16] S. Blanch and A. Aguasca, “Seawater dielectric permittivity model
from measurements at L band,” in Proc. IEEE Int. Geosci. Remote
among the models in more detail. Each figure shows the Sens. Symp. (IGARSS), vol. 2, Sep. 2004, pp. 1362–1365, doi:
brightness temperature predicted by the selected models at 10.1109/IGARSS.2004.1368671.
1.4 GHz and T = 20 ◦ C as a function of salinity. The panel on [17] Y. Zhou, R. H. Lang, E. P. Dinnat, and D. M. L. Vine, “Seawater
Debye model function at L-band and its impact on salinity retrieval
top is the same as Fig. 5 but limited to 32 < S < 38 psu and from aquarius satellite data,” IEEE Trans. Geosci. Remote Sens.,
the panel on the bottom is the same as Fig. 5 but limited to vol. 59, no. 10, pp. 8103–8116, Oct. 2021, doi: 10.1109/TGRS.2020.
12 < S <18 psu. The vertical scale resolution is the same in 3045771.
each panel in Fig. 15 to facilitate comparison. Fig. 16 provides [18] Q. Liu, F. Weng, and S. J. English, “An improved fast microwave water
emissivity model,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 4,
additional information on the dependence of the models on pp. 1238–1250, Apr. 2011, doi: 10.1109/TGRS.2010.2064779.
SST (see top panel of Fig. 5). Fig. 16 shows the difference [19] J. Boutin et al., “Correcting sea surface temperature spurious effects in
in brightness temperature between each model and the GW salinity retrieved from spaceborne L-band radiometer measurements,”
IEEE Trans. Geosci. Remote Sens., vol. 59, no. 9, pp. 7256–7269, Sep.
model. The ST and EL models shown in Fig. 3 have been 2021 doi: 10.1109/TGRS.2020.3030488.
included in this comparison for completeness. To put the [20] E. P. Dinnat, J. Boutin, G. Caudal, J. Etcheto, and P. Waldteufel,
difference in TB in context, the sensitivity of TB to a change “Influence of sea surface emissivity model parameters at L-band for
in SSS at nadir is about 0.5 K/psu at SST = 20 ◦ C and SSS = the estimation of salinity,” Int. J. Remote Sens., vol. 23, no. 23,
pp. 5117–5122, Jan. 2002, doi: 10.1080/01431160210163119.
35 psu [37], [44]. Thus, a difference of 0.1 K corresponds to [21] C. J. F. Böttcher and P. Bordewijk, “Theory of electric polarization,”
a change of about 0.2 psu, which is the goal of contemporary in Dielectrics in Time-Dependent Fields. Amsterdam, The Netherlands:
remote sensing from space [12], [14]. Elsevier, 1978.
[22] P. W. Rosenkranz, “A model for the complex dielectric constant of super-
cooled liquid water at microwave frequencies,” IEEE Trans. Geosci.
Remote Sens., vol. 53, no. 3, pp. 1387–1393, Mar. 2015.
R EFERENCES
[23] A. Stogryn, “Equations for calculating the dielectric constant of
[1] K. R. Carver, C. Elachi, and F. T. Ulaby, “Microwave remote saline water (Correspondence),” IEEE Trans. Microw. Theory Techn.,
sensing from space,” Proc. IEEE, vol. 73, no. 6, pp. 970–996, vol. MTT-19, no. 8, pp. 733–736, Aug. 1971.
Jun. 1985. [24] H. J. Liebe, G. A. Hufford, and T. Manabe, “A model for the complex
[2] C. T. Swift and R. E. Mcintosh, “Considerations for microwave remote permittivity of water at frequencies below 1 THz,” Int. J. Infr. Millim.
sensing of ocean-surface salinity,” IEEE Trans. Geosci. Remote Sens., Waves, vol. 12, no. 7, pp. 659–675, Jul. 1991.
vol. GRS-21, no. 4, pp. 480–491, Oct. 1983. [25] E. Lewis, “The practical salinity scale 1978 and its antecedents,” IEEE
[3] L. Klein and C. Swift, “An improved model for the dielectric constant of J. Ocean. Eng., vol. OE-5, no. 1, pp. 3–8, Jan. 1980.
sea water at microwave frequencies,” IEEE J. Ocean. Eng., vol. OE-2, [26] W. W. Ho, A. W. Love, and M. J. VanMelle, “Measurements of the
no. 1, pp. 104–111, Jan. 1977. dielectric properties of sea water at 1.43 GHz,” NASA, Washington,
[4] P. J. W. Debye, Polar Molecules. Dover, IL, USA: Dover, 1929. DC, USA, Tech. Rep. CR-2458, 1974.
[5] R. Somaraju and J. Trumpf, “Frequency, temperature and salinity [27] W. Ho and W. F. Hall, “Measurements of the dielectric properties of
variation of the permittivity of seawater,” IEEE Trans. Antennas Propag., seawater and NaCl solutions at 2.65 GHz,” J. Geophys. Res., vol. 78,
vol. 54, no. 11, pp. 3441–3448, Nov. 2006. no. 27, pp. 6301–6315, Sep. 1973, doi: 10.1029/JC078i027p06301.

Authorized licensed use limited to: Nanjing University of Information Science and Technology. Downloaded on April 17,2024 at 12:33:35 UTC from IEEE Xplore. Restrictions apply.
 LE VINE et al.: STATUS OF THE DIELECTRIC CONSTANT OF SEA WATER AT L-BAND 4210114

[28] R. Lang, Y. Zhou, E. Dinnat, and D. L. Vine, “The dielectric constant David M. Le Vine (Life Fellow, IEEE) received
model function and implications for remote sensing of salinity,” in the Ph.D. degree in electrical engineering from the
Proc. IEEE Int. Geosci. Remote Sens. Symp. (IGARSS), Jul. 2017, University of Michigan, Ann Arbor, MI, USA.
pp. 3572–3574. His background is electrical engineering with spe-
[29] U. Kaatze and V. Uhlendorf, “The dielectric properties of water at cialization in electromagnetic theory and physics.
microwave frequencies,” Zeitschrift Für Physikalische Chem., vol. 126, He does his research at the Earth Sciences Division,
no. 2, pp. 151–165, Feb. 1981. NASA Goddard Space Flight Center, Greenbelt,
[30] D. Bertolini, M. Cassettari, and G. Salvetti, “The dielectric relax- MD, USA, where he works to develop techniques for
ation time of supercooled water,” J. Chem. Phys., vol. 76, no. 6, microwave remote sensing of the environment from
pp. 3285–3290, Mar. 1982, doi: 10.1063/1.443323. space. His research has focused on passive remote
[31] J. Barthel, K. Bachhuber, R. Buchner, H. Hetzenauer, and sensing at the long wavelength end of the microwave
M. Kleebauer, “A computer-controlled system of transmission lines spectrum (e.g., L-band) with applications to remote sensing of soil moisture
for the determination of the complex permittivity of lossy liquids and sea surface salinity (SSS). Examples of this work are the development of
between 8.5 and 90 GHz,” Berichte der Bunsengesellschaft Für the synthetic aperture radiometer, ESTAR, and the launch of AQUARIUS,
Physikalische Chem., vol. 95, no. 8, pp. 853–859, Aug. 1991, doi: a National Aeronautics and Space Administration (NASA) Earth System
10.1002/bbpc.19910950802. Science Pathfinder (ESSP) mission to measure SSS. He was Deputy Principal
[32] J. B. Hasted, S. K. Husain, F. A. M. Frescura, and J. R. Birch, Investigator for AQUARIUS. He is a member with the Ocean Salinity Science
“The temperature variation of the near millimetre wavelength optical Team, which continues research on remote sensing of SSS. He is also a
constants of water,” Infr. Phys., vol. 27, no. 1, pp. 11–15, Jan. 1987. member of the Science Team for NASA’s Soil Moisture Active Passive
(SMAP) mission and the Quality Working Group supporting the European
[33] F. J. Wentz, “A well calibrated ocean algorithm for special sen-
Space Agency (ESA) Soil Moisture and Ocean Salinity (SMOS) mission.
sor microwave/imager,” J. Geophys. Res., vol. 102, pp. 8703–8718,
His teaching experience includes the Department of Electrical Engineering,
Apr. 1997.
University of Maryland, College Park, MD, and an Adjunct Faculty with The
[34] F. J. Wentz and T. Meissner. (2000). AMSR Ocean Algorithm, George Washington University, Washington, DC, USA.
Version 2, Report Number 121599A-1, Remote Sensing Systems. Dr. Le Vine is a member of the GRSS, Antennas and Propagation Society,
Santa Rosa, CA, USA, [Online]. Available: https://images.remss.com/ the International Union of Radio Science (URSI), and American Geophysical
papers/rsstech/2000_121599A-1_Wentz_AMSR_Ocean_Algorithm_ Union. He was a recipient of the IEEE/GRSS Distinguished Achievement
ATBD_Version2.pdf Award (2016) and the Golden Florin Award (2014), for contributions to
[35] K. Lamkaouchi, A. Balana, G. Delbos, and W. J. Ellison, “Permittivity microwave radiometry. He has served on the Geoscience and Remote Sensing
measurements of lossy liquids in the frequency range 20–110 GHz,” Society (GRSS) AdCom and several IEEE committees focused on engineering
Meas. Sci. Technol., vol. 14, no. 4, pp. 1–7, Apr. 2003. accreditation.
[36] W. J. Ellison et al., “A comparison of ocean emissivity models using
the advanced microwave sounding unit, the special sensor microwave
imager, the TRMM microwave imager, and airborne radiometer obser-
vations,” J. Geophys. Res., Atmos., vol. 108, no. 21, p. 4663, Nov. 2003, Roger H. Lang (Life Fellow, IEEE) received the
doi: 10.1029/2002JD003213. B.S. and M.S. degrees in electrical engineering and
[37] D. M. L. Vine and E. P. Dinnat, “The multifrequency future for remote the Ph.D. degree in electrophysics from the Poly-
sensing of sea surface salinity from space,” Remote Sens., vol. 12, no. 9, technic Institute of Brooklyn, New York, NY, USA,
p. 1381, Apr. 2020, doi: 10.3390/rs12091381. in 1962, 1964, and 1968, respectively.
[38] P. Waldteufel, J. L. Vergely, and C. Cot, “A modified cardioid model for He did his post-doctoral research in random media
processing multiangular radiometric observations,” IEEE Trans. Geosci. under Joe Keller at the Courant Institute of Mathe-
Remote Sens., vol. 42, no. 5, pp. 1059–1063, May 2004. matical Sciences, New York University, New York,
[39] E. P. Dinnat, J. Boutin, G. Caudal, and J. Etcheto, “Issues concerning the NY. He is currently the L. Stanley Crane Professor of
sea emissivity modeling at L band for retrieving surface salinity,” Radio engineering and applied science with George Wash-
Sci., vol. 38, no. 4, pp. 1–25, Aug. 2003, doi: 10.1029/2002RS002637. ington University, Washington, DC, USA. He is
[40] GWU. (2021). The Following Website is Located on the NASA/JPL known for the early development of the discrete scattering model for veg-
PODAAC; See NASA Salinity Dielectric Constant Data and Docu- etation. More recently, he has been involved in remote sensing of seawater
mentation. [Online]. Available: https://podaac.jpl.nasa.gov/Aquarius? salinity and soil moisture under vegetation. His research interests include
sections=data, and https://podaac-tools.jpl.nasa.gov/drive/files/allData/ microwave remote sensing, electromagnetic wave propagation, and dielectric
aquarius/docs/Dielectric_Constant/GW2019_dielectric_constant_ measurements.
measurements.h5 Dr. Lang received the Distinguished Achievement Award from the IEEE
[41] S. Yueh, W. Tang, A. Fore, A. Hayashi, Y. T. Song, and G. Lagerloef, Geoscience and Remote Sensing Society. He is an Active Participant in the
“Aquarius geophysical model function and combined active passive IEEE Geoscience and Remote Sensing Society. He was an Associate Editor
algorithm for ocean surface salinity and wind retrieval,” J. Geo- for Microwave Scattering and Propagation, and the co-chair of the Technical
phys. Res., Oceans, vol. 119, no. 8, pp. 5360–5379, Aug. 2014, doi: Program Committee for the IGARSS’90 meeting held at College Park, MD,
10.1002/2014JC009939. USA, in 1990. He was the Chair of the International URSI Commission F
[42] T. Meissner, F. J. Wentz, and L. Vine, “The salinity retrieval algorithms and is a member of the Editorial Board of Waves in Random and Complex
for the NASA aquarius version 5 and SMAP version 3 releases,” Remote Media.
Sens., vol. 10, p. 1121, Jul. 2018, doi: 10.3390/rs10071121.
[43] E. Dinnat, D. L. Vine, J. Boutin, T. Meissner, and G. Lagerloef, “Remote
sensing of sea surface salinity: Comparison of satellite and in situ
observations and impact of retrieval parameters,” Remote Sens., vol. 11, Yiwen Zhou (Member, IEEE) received the B.S.
no. 7, p. 750, Mar. 2019, doi: 10.3390/rs11070750. degree in electrical engineering from Southeast Uni-
versity, Nanjing, China, in 2010, and the M.S. and
[44] D. M. L. Vine and E. Dinnat, “Sensitivity of wide bandwidth radiometer
Ph.D. degrees in electrical engineering from The
for remote sensing of ocean salinity,” IEEE Trans. Geosci. Remote
George Washington University (GWU), Washington,
Sens., vol. 60, 2022, Art. no. 5301517, doi: 10.1109/TGRS.2021.
DC, USA, in 2012 and 2017, respectively.
3101962.
He then did his post-doctoral research at GWU,
[45] N. Vinogradova et al., “Satellite salinity observing system: Recent for modeling the radar backscatter from vegetation
discoveries and the way forward,” Frontiers Mar. Sci., vol. 6, p. 243, canopy and the impact of seawater dielectric model
May 2019, doi: 10.3389/fmars.2019.00243. on ocean salinity retrieval. He is currently a Research
[46] R. Lang, Y. D. Zhou, and L. Vine, “Seawater dielectric measurements Scientist with Lincoln Agritech Ltd., Lincoln Uni-
at 700 MHz,” in Proc. Electromagn. Res. Symp., vol. 2, Hangzhou, versity, Christchurch, New Zealand. His research interests include remote
China, 2022, pp. 25–29. sensing of ocean salinity and soil moisture, scattering model development,
[47] T. McDougall, P. Barker, and R. Pawlowicz. (Jan. 27, 2015). Version accurate dielectric measurements, metamaterials, and microwave medical
3.05. [Online]. Available: https://[email protected] imaging.

Authorized licensed use limited to: Nanjing University of Information Science and Technology. Downloaded on April 17,2024 at 12:33:35 UTC from IEEE Xplore. Restrictions apply.
 4210114 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 60, 2022

Emmanuel P. Dinnat (Senior Member, IEEE) Thomas Meissner (Senior Member, IEEE) received
received an advanced studies degree in instrumental the B.S. degree in physics from the University of
methods in astrophysics and spatial applications and Erlangen-Nürnberg, Erlangen, Germany, in 1983, the
the Ph.D. degree in computer science, telecommu- M.S. (Diploma) degree in physics from the Univer-
nications, and electronics from the University Pierre sity of Bonn, Bonn, Germany, in 1987, and the Ph.D.
and Marie Curie, Paris, France, in 1999 and 2003, degree in theoretical physics from the University of
respectively. Bochum, Bochum, Germany, in 1991.
He is currently a Senior Research Scientist with From 1992 to 1998, he conducted his postdoc-
the Center of Excellence in Earth Systems Modeling toral research at the University of Washington,
and Observations (CEESMO), Chapman University, Seattle, WA, USA, the University of South Carolina,
Orange, CA, USA, and the Cryospheric Sciences Columbia, SC, USA, and Carnegie Mellon Uni-
Laboratory, NASA Goddard Space Flight Center (GSFC), Greenbelt, MD, versity, Pittsburgh, PA, USA, in Theoretical Nuclear and Particle Physics.
USA. He is working on the calibration and validation and algorithm In 1998, he joined Remote Sensing Systems (RSS), Santa Rosa, CA, USA.
improvements for the Soil Moisture and Ocean Salinity (SMOS), Aquarius/ Since then, he has been working on the development and refinement of
SAC-D, and Soil Moisture Active Passive (SMAP) missions. His latest radiative transfer models, calibration, validation, and ocean retrieval algo-
research focuses on high latitude oceanography and the interactions between rithms for various microwave instruments [special sensor microwave/imager
the cryosphere and oceans. His research interests include active and passive (SSM/I), Tropical Rainfall Measuring Mission (TRMM) Microwave Imager
microwave remote sensing, sea surface salinity, scattering from rough surfaces, (TMI), Advanced Microwave Scanning Radiometer for Earth Observing
atmospheric radiative transfer, and numerical simulations. System (AMSR-E), WindSat, Special Sensor Microwave Imager Sounder
(SSMIS), Global Precipitation Measurement (GPM) Microwave Imager
(GMI), AQUARIUS, Soil Moisture Active Passive (SMAP), Weather System
Follow on – Microwave Imager (WSF-MWI), and Copernicus Imaging
Microwave Radiometer (CIMR)].
Dr. Meissner has been serving on the review panel for the National
Academies’ Committee on Radio Frequencies (CORF). As a member of
the AQUARIUS Launch, Early Orbit Operations, and Commissioning Team,
he was recognized with the NASA Group Achievement Award in 2012.
In 2013, he and Frank Wentz received the IEEE Transactions on Geoscience
and Remote Sensing Prized Paper Award for the paper describing the RSS
ocean radiative transfer model.

Authorized licensed use limited to: Nanjing University of Information Science and Technology. Downloaded on April 17,2024 at 12:33:35 UTC from IEEE Xplore. Restrictions apply.