close

Вход

Забыли?

вход по аккаунту

?

Land surface dynamics monitoring using microwave passive satellite sensors

код для вставкиСкачать
NOTE TO USERS
This reproduction is the best copy available.
®
UMI
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
LAND SURFACE DYNAMICS MONITORING USING
MICROWAVE PASSIVE SATELLITE SENSORS
by
Lizbeth Noemi Guijarro
Bachelor of Sciences
University of Texas at El Paso, 2000
Submitted in Partial Fulfillment of the Requirements
for the Degree of Doctor of Philosophy in the
Department of Geological Sciences
University of South Carolina
2005
Major Professor
Chairman, Examining Committee
Committee Member
Committee Member
Committee M e m b e r
f
s
Dean of the Graduate School
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission
UMI Number: 3181952
Copyright 2005 by
Guijarro, Lizbeth Noemi
All rights reserved.
INFORMATION TO USERS
The quality of this reproduction is dependent upon the quality of the copy
submitted. Broken or indistinct print, colored or poor quality illustrations and
photographs, print bleed-through, substandard margins, and improper
alignm ent can adversely affect reproduction.
In the unlikely event that the author did not send a com plete manuscript
and there are missing pages, these will be noted. Also, if unauthorized
copyright material had to be removed, a note will indicate the deletion.
®
UMI
UMI Microform 3181952
Copyright 2005 by ProQuest Information and Learning Company.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
ProQuest Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor, Ml 48106-1346
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
© Copyright by Lizbeth Noemi Guijarro, 2005
All Rights Reserved.
ii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
I dedicate this dissertation to
Matthew, for his love and devotion
and
my parents, for their unwavering support.
iii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Acknowledgements
I would like to thank my dissertation director, Dr. Venkataraman Lakshmi for
giving me the opportunity to work with him on such interesting research projects for my
dissertation. His great enthusiasm motivated me to finish this dissertation. I appreciate
his positive attitude and support in difficult situations. I would like to thank him for the
knowledge he passed on to me over the past five years.
I would like to thank Dr. Yann Kerr for the opportunity to work with him at the
Centre d’Etudes Spatiales de la Biosphere (CESBIO) in Toulouse and for the awesome
experience I had there. I would also like to thank him for his input and for providing me
with the fundamental tools to carry out this research.
I would also like to thank Dr. Ghani Chehbouni from CESBIO for his input with
the scientific fundamentals of this research and for his help and support to attend
conferences in the EU. I also want to thank him for being such a great host in Toulouse
and for inviting me to a BBQ with his family.
I would like to thank my committee members, Dr. Robert Thunell and Dr. James
Kellogg for their comments, questions and suggestions that helped make this dissertation
better. I would like to thank them for being part of my committee and for their constant
support.
iv
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
I would also like to thank the CESBIO crew for teaching me programming
‘tricks’ in IDL and ENVI and troubleshooting, for helping me get settled in Toulouse, but
most of all for making my experience in Toulouse memorable. I want to especially thank
Mari Carmen, my friend and roommate in Toulouse, for being there for me all this time.
I would also like to thank my lab mates for their all help with the technical stuff
of this research and for making the soil moisture experiments a lot of fun.
It was great
working with you all.
I would like to thank Matthew Miles. I simply could have not done it without
you. Thanks for always being there when I needed you. Thank you for your dedication,
patience, support and words of encouragement. It all has brought me here.
Lastly but not least, I would like to thank my parents, Efren and Letty, my sister
Sandra and my brother David. Thank you for your support, prayers, love and
encouragement all this time because it gave the courage to finish this dissertation. Thank
you for always being there for me, but most of all thank you for believing in me.
v
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Abstract
LAND SURFACE DYNAMICS MONITORING USING
PASSIVE MICROWAVE SATELLITE SENSORS
Lizbeth Noemi Guijarro
Soil moisture, surface temperature and vegetation are variables that play an
important role in our environment. There is growing demand for accurate estimation of
these geophysical parameters for the research of global climate models (GCMs), weather,
hydrological and flooding models, and for the application to agricultural assessment, land
cover change, and a wide variety of other uses that meet the needs for the study of our
environment. The different studies covered in this dissertation evaluate the capabilities
and limitations of microwave passive sensors to monitor land surface dynamics.
The first study evaluates the 19 GHz channel of the SSM/I instrument with a
radiative transfer model and in situ datasets from the Illinois stations and the Oklahoma
Mesonet to retrieve land surface temperature and surface soil moisture.
The surface
temperatures were retrieved with an average error of 5 K and the soil moisture with an
average error of 6%.
The results show that the 19 GHz channel can be used to
qualitatively predict the spatial and temporal variability of surface soil moisture and
surface temperature at regional scales.
In the second study, in situ observations were compared with sensor observations
to evaluate aspects of low and high spatial resolution at multiple frequencies with data
collected from the Southern Great Plains Experiment (SGP99). The results showed that
the sensitivity to soil moisture at each frequency is a function of wavelength and amount
vi
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
of vegetation. The results confirmed that L-band is more optimal for soil moisture, but
each sensor can provide soil moisture information if the vegetation water content is low.
The spatial variability of the emissivities reveals that resolution suffers considerably at
higher frequencies.
The third study evaluates C- and X- bands of the AMSR-E instrument. In situ
datasets from the Soil Moisture Experiments (SMEX03) in South Central Georgia were
utilized to validate the AMSR-E soil moisture product and to derive surface soil moisture
with a radiative transfer model. The soil moisture was retrieved with an average error of
2.7 % at X-band and 6.7 % at C-band. The AMSR-E demonstrated its ability to
successfully infer soil moisture during the SMEX03 experiment.
Dr. Venkataraman Lakshmi, Disseration Director
vii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
TABLE OF CONTENTS
List of Tables...................................................................................................... xi
List of F igu res................................................................................................. xiii
Chapter 1...............................................................................................................1
1. G eneral In tr o d u c tio n ................................................................................................... 1
Chapter 2...............................................................................................................6
2. M icrowave T h e o r y ......................................................................................................... 6
2.1 Theory of Passive Microwave Remote Sensing o f L a n d ........................................6
2.2 Dielectric C onstant...................................................................................................... 9
2.3 Land Surface E ffects................................................................................................. 11
2.4 Atmospheric E ffects.................................................................................................. 13
Chapter 3.............................................................................................................18
3. Land Surface T emperature and S oil M oisture Retrieval U sing S S M /I
18
3.1 Introduction.................................................................................................................18
3.2 Data and M eth o d s......................................................................................................21
3.2.1 In Situ D atasets.............................................................................................21
3.2.2 Satellite Data ................................................................................................... 23
3.3 Calibration of the Radiative Transfer M o d el......................................................... 26
3.4 R esults......................................................................................................................... 28
3.4.1 Surface Temperature Retrieval ...................................................................... 28
3.4.2 Soil Moisture Retrieval ................................................................................... 29
3.5 Conclusions and Discussion .................................................................................... 34
viii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter 4 ........................................................................................................... 54
4. A MICROWAVE MULT SENSOR STUDY OF LAND SURFACE EMISSIVITY OVER THE
SOUTHERN GREAT PLAINS, OKLAHOMA .................................................................................54
4.1 Introduction................................................................................................................54
4.2 Data and M eth o d s..................................................................................................... 57
4.2.1 Study A r e a ........................................................................................................ 57
4.2.2 Airborne Sensors.......................................................................................... 58
4.2.2.1 Passive/Active L/S- Sensor (PA LS).................................................... 58
4.2.2.2 Polarimetric Scanning Radiometer at C-band (P S R /C )....................58
4.2.3 Space borne Sensors ....................................................................................... 59
4.2.3.1 Tropical Rainfall Measuring Mission (TRMM-TMI)........................59
4.2.3.1 Special Sensor Microwave Imager (SSM /I)....................................... 59
4.3 R esults.........................................................................................................................61
4.3.1 Comparisons o f Airborne and Space borne Observations ....................... 61
4.3.2 Observed and Simulated Emissivities ...........................................................64
4.3.3 Sensitivity ..........................................................................................................65
4.2.4 Simulation M o del......................................................................................... 66
4.4 Conclusions and Discussion .................................................................................... 71
Chapter 5 ............................................................................................................86
5. Soil M oisture R etrieval over S outh C entral G eorgia from A M S R -E
86
5.1 Introduction................................................................................................................ 86
5.2 Data and M eth o d s..................................................................................................... 89
5.2.1 SMEX03 Experiment Plan ...............................................................................89
5.2.2 Satellite D ata .................................................................................................... 92
5.2.2.1 AMSR-E ................................................................................................. 92
5.3 R esults......................................................................................................................... 94
5.3.1 AMSR Soil Moisture and In Situ Soil Moisure Validity...............................94
5.3.2 Predicted Soil Moisture with AMSR-E D ata ................................................ 97
5.3.3 Spatial Variability o f Soil Moisture During SMEXO3 ............................. 100
5.4 Conclusions and Discussion .................................................................................. 102
ix
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Chapter 6.......................................................................................................... 119
6. Summary and F ollow -up St u d ie s ........................................................................... 119
6.1 Summary.............................................................................................................. 119
6.2 Follow-up Studies and Applications.................................................................. 122
6.2.1 Surface Temperature.................................................................................. 122
6.2.2 Snow C o ver............................................................................................... 123
6.2.3 Atmospheric Models for Higher Frequency Channels .......................... 124
6.2.4 Vegetation Am ount..................................................................................... 125
6.2.5 Vegetation Optical Depth ........................................................................ 125
Bibliography ................................................................................................... 126
x
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
LIST OF TABLES
Table 3-1. Radiative transfer model inputs........................................................................ 37
Table 3-2. Mean, standard deviation, correlation coefficient, and root mean square of the
retrieved and in situ surface temperature for the Oklahoma Mesonet and the
Illinois stations...........................................................................................................38
Table 3-3. Mean, standard deviation, correlation coefficient, and standard error of the
monthly averaged retrieved and in situ soil moisture for the Oklahoma Mesonet
and the Illinois stations............................................................................................. 39
Table 3-4. Station ID, location, longitude, latitude, and soil texture for Stillwater,
Marena, Perkins, Acme, Goodwell, Seiling, Nowata, and Wister Stations in
Oklahoma................................................................................................................... 40
Table 3-5. Station ID, location, longitude, latitude, and soil texture for Peoria,
Springfield, Belleville, Carbondale, and Olney stations in Illinois....................... 41
Table 4-1. Sensor name, frequency, nadir angle, spatial resolution, and swath of the
sensors used in study..................................................................................................73
Table 4-2. Effects of the vegetation water content and surface temperature on the
polarization difference of brightness temperatures
at various
frequency
channels...................................................................................................................... 74
Table 4-3. Means and standard deviations of input parameters for the simulations at
three different scales..................................................................................................75
Table 4-4. Averaged brightness temperatures (v & h polarisation), standard deviation of
brightness temperatures (v & h), emissivities (v & h), standard deviation of
emissivities (v & h), coefficient of variation of brightness temperatures (v & h),
coefficient of variation of emissivities (v & h), and averaged polarization
difference of simulations for all sensors for Day 2 ................................................. 76
xi
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 5-1. Regional sampling sites and land use within each EASE-Grid cell box
104
Table 5-2. Correlation and standard deviation for the mv from all the regional sampling
sites and AMSR-E mv and the mean mv from all the regional sampling sites and
AMSR-E mv............................................................................................................. 105
Table 5-3. Input parameters of the radiative transfer model............................................ 106
Table 5-4. Statistics for the predicted soil moisture and in situ soil moisture at 0-1 cm
and 0-3 cm at 6.9 GHz and 10.7 GHz channel......................................................107
X ll
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
LIST OF FIGURES
Figure 2-1. Soil permittivity as a function of frequency....................................................15
Figure 2-2. Dielectric loss factor as a function of frequency............................................ 16
Figure 2-3. Dependence of the soil dielectric on soil texture........................................... 17
Figure 3-1. Time series of the estimated LAI for the Illinois stations............................. 42
Figure 3-2. Time series of the estimated LAI for the Oklahoma Mesonet...................... 43
Figure 3-3. Simulated brightness temperatures (19H) with the model and SSM/I
brightness temperatures (19H) for the Oklahoma Mesonet....................................44
Figure 3-4. Simulated brightness temperatures (19H) with the model and SSM/I
brightness temperatures (19H) for the Illinois stations........................................... 45
Figure 3-5. Retrieved surface temperatures versus in situ surface temperatures for the
Oklahoma Mesonet....................................................................................................46
Figure 3-6. Retrieved surface temperatures versus in situ surface temperatures for the
Illinois stations...........................................................................................................47
Figure 3-7. Retrieved soil moisture versus in situ soil moisture of the top layer for the
Illinois stations...........................................................................................................48
Figure 3-8. Retrieved soil moisture versus in situ soil moisture of the top layer for the
Oklahoma Mesonet ...................................................................................................49
Figure 3-9. Time series of the averaged observed and retrieved soil moisture for the top
layer for Stillwater, Marena, Perkins, Acme, Goodwell, Seiling, Nowata, and
Wister stations in Oklahoma..................................................................................... 50
Figure 3-10. Time series of the averaged observed and retrieved soil moisture for the top
layer for Peoria, Springfield, Belleville, Carbondale, and Olney stations
in Illinois.................................................................................................................... 51
Figure 3-11. Precipitation at the Oklahoma Mesonet, 5 April, 1999................................52
Figure 3-12. SSM/I brightness temperatures on 5 April, 1999........................................ 52
Figure 3-13. Retrieved soil moisture on 5 April, 1999.................................................... 52
Figure 3-14. SSM/I brightness temperatures on 6 April, 1999......................................... 52
Figure 3-15. Retrieved soil moisture on 6 April, 1999......................................................52
Figure 3-16. SSM/I brightness temperatures on 7 April, 1999........................................ 52
xiii
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 3-17. Retrieved soil moisture on 7 April, 1999..................................................... 52
Figure 3-18. SSM/I brightness temperatures on 8 April, 1999......................................... 52
Figure 3-19. Retrieved soil moisture on 8 April, 1999.................................................... 52
Figure 3-20. Precipitation, observed and simulated brightness temperatures, and
volumetric soil moisture for five consecutive days in 1998 for Lane Station,
Oklahoma................................................................................................................... 53
Figure 4-1. Time series of the observed horizontal polarization emissivities for all
sensors and the in situ soil moisture for a rangeland field......................................77
Figure 4-2. Time series of the observed horizontal polarization emissivities for all
sensors and the in situ soil moisture for a wheat field............................................ 78
Figure 4-3. Time series of the observed horizontal polarization emissivities for all
sensors and the in situ soil moisture for a com field............................................... 79
Figure 4-4. Time series of the observed horizontal polarization emissivities for all
sensors and the in situ soil moisture for an alfalfa field......................................... 80
Figure 4-5. Approach for the retrieval of emissivities from each sensor........................ 81
Figure 4-6. Observed and simulated emissivities in the horizontal polarization for three
field sites from 8 July through 14 July..................................................................... 82
Figure 4-7. Observed and simulated emissivities in the vertical polarization for three
field sites from 8 July through 14 July..................................................................... 83
Figure 4-8. Approach used in the simulation model of brightness temperatures for the
retrieval of emissivities at each frequency............................................................... 84
Figures 4-9a through 4-9j.
Spatial variability of soil moisture and the simulated
emissivity at the horizontal polarization for all frequencies for Day 2 .................. 85
Figure 5-1. Regional sampling sites, vegetation and vegetation-eddy flux sampling sites
at each EASE-Grid cell box.................................................................................... 108
Figure 5-2. AMSR-E mean Tb at C-band and in situ mv at 0-1 cm .............................. 109
Figure 5-3. AMSR-E mean Tb at X-band and in situ mv at 0-1 c m ............................. 110
Figure 5-4. Mean mv 0-1 cm and AMSR-E mv...............................................................I l l
Figure 5-5.
Mean mv 0-3 cm and AMSR-E mv...............................................................112
Figure 5-6.
Predicted and in situ soil moisture 0-1 cm from C-band T b ......................113
Figure 5-7.
Predicted and in situ soil moisture 0-3 cm from C-band T b ......................114
xiv
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 5-8. Predicted and in situ soil moisture 0-1 cm from X-band Tb......................... 115
Figure 5-9. Predicted and in situ soil moisture 0-3 cm from X-band Tb ........................116
Figure 5-10a through 5-10/. Spatial distribution of AMSR-E horizontal polarization
brightness temperatures at 10.7 GHz and mean predicted volumetric soil
moisture 0-1 cm........................................................................................................117
Figure 5-1 \a through 5-11/. Spatial distribution of AMSR-E horizontal polarization
brightness temperatures at 10.7 GHz and mean predicted volumetric soil
moisture 0-1 cm........................................................................................................118
xv
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
C h a p t e r 1 - G e n e r a l I n t r o d u c t io n
The use of satellites in the early sixties provided useful data to water resource
management agencies and research facilities.
During this time, visible and infrared
sensors were used to observe parameters such as snow cover, surface water areas, land
use, and surface temperature (Schmugge, 1985), but the signal of these sensors was
severely affected by atmospheric interference (i.e. clouds, water vapor).
Without the
means to counter the effect of these inhibiting factors, the hydrological parameters that
these sensors could observe were limited.
In the seventies, the potential for using microwave satellite sensors for land
surface monitoring was explored. Scientists observed that this new spectra of satellite
sensing, particularly at low frequencies had the capability to penetrate atmospheric
interference.
Microwave satellite sensors had the capability to provide their own
illumination regardless of local weather and thus, be accurately calibrated (Ulaby, 1985).
Studies in the late seventies demonstrated that microwave satellite sensors allowed for
disaggregation and diurnal variability studies, and for truly quantitative estimates of soil
moisture using physically based expressions such as radiative transfer models (Njoku
1
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
and Kong, 1977; Wilheit, 1975; Choudhury et al., 1979).
Today, the widespread
availability of microwave sensor observations makes it possible to conduct land surface
dynamics studies.
Soil moisture is a key variable in the hydrological cycle.
The exchange of
moisture fluxes in the form of evapotranspiration is a major link in the interaction
between hydrological and atmospheric processes (Delworth and Manabe 1989).
Soil
moisture also plays an important role in various land-atmosphere interactions. It controls
the partition of rainfall into runoff and infiltration (Shukla and Mintz, 1982). Thus, it is
an essential input for rainfall-runoff models. Soil moisture also controls the partition of
available energy at the surface into sensible and latent heat flux, thereby playing a crucial
role in boundary layer development. This exchange is a driving force for weather and
climate systems.
Numerous studies have shown that accurate estimation of soil moisture has
implications for the research of land-atmosphere interactions, particularly for the research
of General Circulation Models (GCMs) (Walker and Rowntree, 1977; Rowntree and
Bolton, 1983; Rind, 1982; Mintz, 1984). Understanding the role of soil moisture in the
hydrological cycle is also important for relating soil processes with water balance, plant
growth, watershed supply issues, and many other processes like irrigation scheduling,
precision farming, groundwater recharge, hydrological modeling, meteorological and
flood forecasting. Soil moisture also plays an important role in sub-surface contaminant
transport, which affects the nature of groundwater pollution. There is a distinct need for
a long-term temporal and spatially distributed database of land surface dynamics.
2
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Soil moisture has a high temporal and spatial variability due to a small-scale
component which is characterized by soil properties, vegetation, and topography, and a
large-scale component characterized by solar radiation, atmospheric forcing, and general
environmental conditions (Entin et al., 2000). The dearth in soil moisture data stems
from
the difficulty in monitoring this variability using conventional ground
measurements, apart from the high costs involved in monitoring soil moisture at larger
scales. Microwave satellite sensors offer a promising avenue of inferring regional and
global soil moisture.
Studies focused on the retrieval of soil moisture and other environmental
parameters using truck-mounted and aircraft radiometers have been conducted in the past
(O’Neill, 1985; Jackson et al., 1984). Despite excellent results, these past studies were
done at finer spatial scales with a more homogeneous surface and controlled
environmental conditions. Studies involving the use of space borne sensors have been
attempted with successful outcomes as well (Schmugge et al., 1977; Allison et al., 1979;
Blanchard et al., 1981; Wang, 1985; Schmugge et al., 1986; Choudhury et al., 1987;
Choudhury and Golus, 1988; Owe et al., 1988; Njoku and Li, 1999; Vinnikov et al.,
1999a and b). However, studies with space borne sensors have been limited to more or
less vegetated areas, specific seasons, certain cultivation characteristics, and smaller
areas. The estimation of soil moisture has been met with limited success for
representative large areas.
A mechanism for the estimation of soil moisture which
accounts for changing environmental conditions, vegetation cover types, spatial
resolutions among others, is needed.
3
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The work presented in this dissertation focused on the correlation between
models, satellites, and in situ observations to help improve models and measurement
procedures for characterizing land surface dynamics, with an emphasis on soil moisture.
This dissertation covers various studies that evaluated the capabilities and limitations of
multiple microwave passive sensors to retrieve soil moisture. The work aimed to assist
the ongoing efforts to understand the role of soil moisture in the land surface hydrological
cycle and help calibrate algorithms for the next generation of satellites that estimate soil
moisture such as the Soil Moisture and Ocean Salinity (SMOS) mission and the NASA
Hydrosphere State (HYDROS) mission scheduled for launch in 2007 and 2010,
respectively.
Soil moisture information and ocean salinity data together will improve
climate, weather and extreme-weather forecasting (http://www.esa.int/esaLP/smos.html).
The following chapters consist of three studies carried out over various regions in
the US that have typically provided good in situ datasets and sensor coverage. Chapter 2
describes the fundamentals of microwave theory and a physically based model namely
radiative transfer model. The methodology applied to all three studies for the various
parameter retrievals is based on the physics fundamentals presented in chapter 2. Chapter
3 gives a description of the data and methods for a study carried out over Illinois and
Oklahoma, followed by the results of the retrieval of land surface temperature and surface
soil moisture using in situ data and observations from the Special Sensor Microwave
Imager (SSM/I) at 19 GHz. Chapter 4 gives a description of the data and methods for a
study carried out over the southern Great Plains, followed by comparisons of observed
soil moisture with emissivities from aircraft and satellite microwave passive sensors.
Also, comparisons of observed and estimated emissivities for the various microwave
4
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
passive sensors are presented, followed by a sensitivity study and a simulation model of
land surface parameters to estimate emissivities at each frequency studied from the
various microwave passive sensors. Chapter 5 gives a description of the Soil Moisture
Experiments (SMEX03) in south central Georgia with a focus on ground truth protocols,
followed by validation of the Advanced Microwave Scanning Radiometer (AMSR-E) soil
moisture product with in situ soil moisture.
Also in this section, the results of the
retrieval of soil moisture using the Advanced Microwave Scanning Radiometer (AMSRE) at C- and X- bands are presented, followed by conclusions of the study. Chapter 6
provides a brief summary of the three studies and characterizes a few follow-up studies of
various land surface parameters along with their applications.
5
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
C h a p t e r 2 - M ic r o w a v e T h e o r y
2.1
THEORY OF PASSIVE MICROWAVE REMOTE SENSING OF LAND
Passive microwave remote sensing is based on the amount of incident radiation
measured by a radiometer expressed as a brightness temperature, TB. An approximation
of Plank’s equation for frequencies < 1 1 7 GHz is the Raleigh-Jeans expression for a
brightness temperature (Ulaby et al., 1986), in which a brightness temperature TB is the
product of the physical temperature Te of the radiating body and the emissivity e of the
soil surface (Njoku and Kong, 1977; Ulaby et al., 1986; Vinnikov et al., 1999a),
TB P =e-Te
[2.1]
where P refers to horizontal or vertical polarization of the observed brightness
temperature TB, Te is the physical or effective surface temperature of the emitting layer,
and e is the surface emissivity. The value of s ranges from 0 to 1, depending on the
composition and texture of the surface. For bare soil surfaces, Te is the weighted soil
temperature in the subsurface layer defined by the microwave penetration depth (Tsang et
al., 1975; Njoku and Kong, 1977).
6
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Kirchkoff s law states that the emissivity equals the absorbtivity A of a surface,
R +A =1
[2.2]
where R is the surface reflectivity. Thus, the emissivity e is given by
[2.3]
e =l-R
The reflectivity R can be calculated from Fresnel's equations (Equations 2.4 and
2.5), which describe the reflection and transmission of electromagnetic waves at an
interface. Fresnel's equations give the ratio of the reflected and transmitted electric field
amplitude to the initial electric field for electromagnetic radiation incident on a dielectric.
When a wave reaches a boundary between two different dielectric constants, part of the
wave is reflected and part is transmitted, with the total amount of the energies in these
two waves equal to that of the initial wave. Because electromagnetic waves are
transverse, there are separate coefficients in the directions perpendicular to and parallel to
the surface of the dielectric. The case for the horizontal polarization is that in which the
electric field is oriented parallel to the reflecting surface and perpendicular to the
direction of propagation. For the vertical polarization, the electric field has a component
perpendicular to the surface (Serway, 1994). These cases are represented by following
equations,
2
[2.4]
2
[2.5]
where e is the complex dielectric constant of a soil water mixture (e = e ’ + e ”i), 6 is the
incidence angle and H and V refer to the emitted radiation.
7
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
These two polarizations are equivalent at 6 = 0°; thus, these equations become,
[2.6]
■ \J £ +
1
It has been shown in previous studies that the soil emissivity in the horizontal
polarization is sensitive to changes in surface moisture.
The vertical emissivity will
increase until it reaches 1 as the nadir angle increases from 0° (Schmugge, 1980).
tanOB = ^
[2.7]
This means that the reflection coefficient equals zero and the indices of reflection are
equal to each other given the Brewster angle. At this angle from the normal, the reflected
light is completely polarized (Equation 2.7).
8
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2.2
DIELECTRIC CONSTANT
The dielectric constant, often called the relative permittivity, is a characteristic
quantity of a given dielectric substance. The dielectric constant is a complex number that
contains a real part £’ and an imaginary part e” (Equation 2.8).
The real part of a
dielectric gives the reflective surface properties (Fresnel reflection coefficients), and it
determines the propagation characteristics of the energy as it passes upward through the
soil. The imaginary part gives the radio absorption coefficient, and it determines the
energy loss (Schmugge et al., 1986).
The variations of e’ and e” as a function of
wavelength A are known as the Debye relaxation formulae (Equations 2.9 and 2.10).
The dielectric is defined as,
e = e '+ je "
[2.8]
where e’ is the real part, e” is the imaginary part, and j is V(-l).
The real and imaginary parts are described by Equations 2.9 and 2.10,
respectively,
/ / “I
,
(i
+0 0 )
[ 2 ' 9 ]
e " = (on)
(1 + ( c o z ) 2)
[2 .1 0 ]
CO
where £«, is the highfrequency limit of £, es is the permittivityof soil , aw is the ionic
conductivity of water, z is the molecular relaxation time, and co isgiven by Equation
2.11 w here/is the frequency in GHz.
co = 2 n f
[2.11]
In the 1 to 10 cm wavelength range, the value of £’ changes rapidly from es at low
frequencies to a much smaller value £„, at higher frequencies.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The dielectric constant is a difficult quantity to measure in the field. In 1980,
Wang and Schmugge produced a model to allow for theoretical calculations of this
constant.
Subsequently, Dobson et al. (1985) produced another dielectric model, and
Peplinski (1995) produced a modified version of Dobson’s model. The Peplinski model
is best suited for frequencies less than 18 GHz. The Wang and Schmugge model is used
for frequencies 19 GHz and higher.
Because a soil medium is heterogeneous, the dielectric constant results in a
combination of its components which are the various dielectric constants of air, water,
rock, etc. The dielectric constants are influenced by a series of factors (i.e. the soil
moisture content, temperature, salinity, soil texture, frequency) (Schmugge, 1985). The
influence of the both frequency and soil moisture content on the soil dielectric constant
can be seen in Figures 2-1 and 2-2. The relationship between the frequency and the soil
dielectric is linear, except when there is almost no moisture content (-volumetric soil
moisture, mv < 0.02). The higher the moisture content, the higher the dielectric constant
of soil is. The soil dielectric constant decreases with frequency.
The dependence of the soil dielectric on the soil texture is represented with three
common soil types (sand, loam, clay) in Figure 2-3. This relationship is generally linear
with an exception where there is low soil moisture content. This is due to the strong
bonds between the soil particles and the thin films of water which surround them. Thus,
each different type of soil texture exhibits a unique relationship with the soil dielectric
constant. A clayey soil (80 % clay) has a relatively low dielectric constant compared to
sandy soil (90% sand) at the same soil moisture content as seen in Figure 2.3.
10
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2.3
LAND SURFACE EFFECTS
The basis of soil surface roughness is represented by its effective temperature Te
and emissivity e in Equation 2.1. It has been demonstrated that surface roughness causes
the emissivity of natural surfaces to be higher than a smooth surface (Schmugge, 1985;
Choudhury et al. 1979). A semi-empirical relationship is utilized to estimate the surface
reflectivity rp given by Equation 2.12 (Wang and Choudhury, 1981),
rP ={Qroq + {l - Q ) roPy * p { - hc°s2 ev)
[2.12]
The surface reflectivity rp is given in terms of a polarization coupling factor Q, a viewing
angle dv, a roughness effective height h and the specular reflectivities of a corresponding
smooth soil surface with the same moisture and texture as the rough surface rop in
Equation 2.12. The h parameter is estimated using Equation 2.13 in which A: is a wave
number and cris the rms of the surface height.
h =(2 k a f
[2.13]
The parameters Q and h in Equation 2.12 account for the dependence of the surface
reflectivity rp of a randomly rough surface on both the rms surface height and a
horizontal correlation length I (Hallikainen et al., 1985; Kerr and Njoku, 1990).
The vegetation is modeled as a single homogenous layer above the soil following
the approach of Njoku and Entekhabi (1996). The brightness temperature equation is
given by,
Tbp
= Ts { l - rp )
exP M
+ Tc 0 - m) [ l - exP (~r )] [l + r p exP ( - T)]
[2-14]
where Ts is the soil temperature, rp are the soil reflectivities, t is the vegetation opacity,
Tc is the vegetation temperature, and co is the single scattering albedo.
11
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The opacity ris characterized as in Jackson and Schmugge (1991). It follows,
coseq,
[2.15]
where b is a function of canopy type, polarization and wavelength, Wc is the vegetation
water content (k g -n f), o.v is the incidence angle where cos a.v accounts for the non­
vertical path through the vegetation. The parameter b is frequency and vegetation cover
type dependent.
Due to the heterogeneity of the pixels observed by a space borne sensor, Equation
2.16 is used to determine the proportion of radiation sensed by the satellite from bare and
vegetation covered areas.
TBp = CTBPcan + (1 - C ) TBPbare
[2.16]
where C is the canopy fraction representing the percentage of surface covered by
vegetation, Tppbare is the brightness temperature of bare soil, and Tbpcan is the brightness
temperature of the surface at polarization P when completely covered by vegetation.
Following the approach by Kerr et al. (1992), the canopy fraction is given as a function of
the vegetation index NDVI expressed as,
c
( N D V l-N D V lml„)
(NDVImax-N D V Imi„)
where NDVImax is the highest NDVI value for a fully vegetated pixel, and NDV Imjn is the
minimum value of NDVI for bare soil. The N DVImax and NDVImin were extracted for
every station in Illinois and Oklahoma for each year, given that the vegetation has a full
cycle in one year.
The NDVImax and NDVImjn values correspond to the maximum and
minimum NDVI values of a station for a specific year.
12
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
2.4
ATMOSPHERIC EFFECTS
The influence of the atmosphere on the signal measured by a radiometer increases
with frequency. The brightness temperature equation becomes
Tbp
= T«+ T
b p
exP (-T ,) + exP (-T, ) \ Td + T
sky
exP {-Ta) ) rP
[2.14]
where Tu is the upwelling atmospheric emission, Td is the downwelling atmospheric
emission scattered at the surface and attenuated along the upward path by the atmosphere,
T sky
is the apparent temperature of the sky,
rp
is the surface reflectivity, and ra is the
opacity of air. The parameters in Equation 2.14 are functions of frequency. They were
computed for the 19 GHz and 37 GHz channel separately.
For small values of the opacity of air ra, the upwelling and downwelling
atmospheric emissions Tu and Td can be approximated as,
Ta ” Td = Ta\}-^ v {r^ a )\
t2’15]
where Ta is the weighted-mean temperature of the microwave-absorbing region of the
troposphere. The opacity of air za depends on gaseous and liquid-droplet attenuating
constituents (i.e. oxygen, water vapor, and clouds). The opacity of air za is given by,
=
[216]
cos#
where r0 is the oxygen nadir opacity, tv is the product of the water vapor coefficient kv
and the water vapor content qv, and t/ is the product of the cloud nadir opacity coefficient
ki and the cloud liquid content qi. The atmospheric temperatures and pressures needed for
the computation of the opacity of air ra were calculated with a density scale height of 9.5
km (Ulaby et al., 1986a). The oxygen absorption coefficient was computed using the
Van Vleck formulation for frequencies below 45 GHz (1947).
The water vapor
13
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
absorption coefficient was computed using the formulation for frequencies below 100
GHz and using the absorption coefficient for the 22 GHz line (Ulaby et al., 1986a). The
expression by Benoit (1968) was used to calculate the liquid extinction coefficient for
clouds containing negligible amount of ice particles and fog. Details concerning the
computation of these parameters can be found in Ulaby et al. (1986a).
14
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
40
mv=0.02
mv=0.08
mv=0.16
mv=0.23
mv=0.27
mv=0.32
mv=0.38
35
Soil Dielectric Constant
e'
30
5 --
0 -I - - - - - - - - - - - - - - - - - - - - - - 1- - - - - - - - - - - - - - - - - - - - - - 1- - - - - - - - - - - - - - - - - - - - - - 1- - - - - - - - - - - - - - - - - - - - - - 1- - - - - - - - - - - - - - - - - - - - - - 1- - - - - - - - - - - - - - - - - - - - - - 1- - - - - - - - - - - - - - - - - - - - - - 1- - - - - - - - - - - - - - - - - - - - - - 1- - - - - - - - - - - - - - - - - - - - - - 1—
1.4
3.4
5.4
7.4
9.4
11.4
13.4
15.4
17.4
Frequency (GHz)
Figure 2-1. Soil permittivity as a function of frequency.
15
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
19.4
40
mv=0.02
mv=0.08
mv=0.16
mv=0.23
mv=0.27
mv=0.32
mv=0.38
35
Soil Dielectric Constant c"
30
25 -
20
--
15 -
10
--
1.4
3.4
5.4
7.4
9.4
11.4
13.4
15.4
17.4
Frequency (GHz)
Figure 2-2. Dielectric loss factor as a function of frequency.
16
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
19.4
30
eps', sand'
eps", sand
eps', loam
25 --
eps", loam
eps', clay
eps", clay
Dielectric Constant t
20
--
15 --
10
5 --
Freq = 6.6 GHz
0.00
0.20
0.25
0.30
0.35
0.40
Volumetric Soil M oisture mv
Figure 2-3. Dependence of the soil dielectric on soil texture.
17
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0.45
0.50
C h a p t e r 3 - L a n d Su r f a c e T e m pe r a t u r e
and
S o il
M o is t u r e R e t r ie v a l U s in g S s m /i
3.1
INTRODUCTION
Soil moisture has a high temporal and spatial variability due to a small-scale
component which is characterized by soil properties, vegetation, and topography, and a
large-scale component characterized by solar radiation, atmospheric forcing, and general
environmental conditions (Entin et al., 2000). The dearth in soil moisture data stems
from the difficulty in monitoring this variability and the high costs involved in
monitoring soil moisture at larger scales. Microwave satellite sensors offer a promising
avenue of inferring regional and global soil moisture. They have proven to be effective
for soil moisture sensing because of the large contrast between the dielectric properties of
liquid water (« 80), dry soil (« 4) and the resulting soil-water mixes (4-40) (Schmugge,
1985).
Microwave satellite sensors have also proven to be effective because of the
dielectric properties’ effects on the natural emission from the soil (Schmugge et al.,
1986). Microwave sensors allow for the estimation of soil moisture using physically
based radiative transfer models (Njoku and Kong, 1977; Wilheit, 1978; Choudhury et al.,
1979; Kerr and Njoku, 1990).
18
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Jackson (1997) gives us an approximated error of 5.3% of the estimated soil
moisture in the grass-dominated sub-humid Oklahoma region. In his study, the Little
Washita watershed (-600 km ) was the focus of intensive field experiments where soil
moisture data were sampled at the 0-5 cm depth for a period of 10 days in June 1992 and
April 1994.
None of the works by Jackson (1997), England et al. (1992), Heymsfield and
Fulton (1992), Teng et al. (1993), compare SSM/I retrieved to observed soil moisture
specifically over a large area. A large area has been utilized with SMMR data by Owe
(1992) and Vinnikov (1999a), but they also do not carry out a direct comparison.
Lakshmi et al. (1997a) compare model output (0-1 cm) soil moisture to SSM/I retrieved
soil moisture. Hence, two issues remain: (a) lack of 0-1 cm estimates of soil moisture
that best correspond to the SSM/I sensing depth and (b) comparisons over an extended
time scale (seasonal cycle) and over large areas (several thousands of square kilometers).
In this study, we still fall short of our requirement by the use of 0-5 cm
(Oklahoma Mesonet) and 0-10 cm (Illinois stations) data for soil moisture retrievals.
However, we do accomplish soil moisture comparisons for 2 years over large areas in the
states of Oklahoma and Illinois.
We carry out an approach using data from the Special Sensor Microwave Imager
SSM/I and the Normalized Difference Vegetation Index derived from the Advanced Very
High Resolution Radiometer (AVHRR) to evaluate surface temperature and soil moisture
of the upper layer of soil in the states of Illinois and Oklahoma. This study investigates
the possibility of achieving a better characterization of these land surface variables over
large regions of heterogeneous vegetation character through the combination of passive
19
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
and optical satellite data using a radiative transfer model.
Chapter 2 describes the
radiative transfer model used for the simulation of brightness temperatures. Chapter 3.3
discusses the calibration and validation of the radiative transfer model.
Chapter 3.4
discusses the results of the retrieval of soil moisture and surface temperature and their
comparisons with in situ data. Chapter 3.5 continues with conclusions and discussion.
20
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.2
DATA AND METHODS
3.2.1
In Situ Datasets
In situ datasets of soil moisture and surface temperatures were obtained for the
states of Illinois and Oklahoma. The data spanned the years 1998 and 1999. For the state
of Illinois, the datasets obtained consist of soil moisture observations at 19 stations as
part of the Water and Atmospheric Resources Monitoring Program (WARM)
(http://climate.envsci.rutgers.edu/ soil_moisture/illinois.html). The 19 stations are spread
out across the state, mostly co-located with stations from the Illinois Climate Network.
The soil moisture is measured with a neutron depth probe, which is calibrated against
direct gravimetric measurements at each site (Vinnikov et al., 1999a). The RMSE error
of the surface probe calibration is approximately 4.5 % by volume (Vinnikov et al.,
1999a). The first measurement is taken from the upper 10 cm layer of soil, and then
every 20 cm thereafter down to a depth of 2 m. During the warmer months, the soil
moisture observations at each station are taken at the middle and at the end of the month.
During the colder months only one measurement is taken during the last week of the
month (Vinnikov et al., 1999a). The measurements are taken between 0700 and 1800
hours. The average Illinois soil moisture station spacing is about 93 km (Vinnikov et al.,
1999b). The Illinois stations are all grass-covered except for one station located on bare
soil. The stations are located primarily in loess soils of varying thickness, which are the
primary agricultural soils in Illinois. The stations in the southern one third of the state
have a dense layer of soil approximately 0.5 m below the surface, which limits root
development below 0.5 m. Stations in the northern two thirds of the state do not have
any root-restricting zones in the first 2 m. The precipitation datasets for the Illinois
21
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
stations were obtained from Earthlnfo’s National Climate Data Center (NCDC), National
Oceanic and Atmospheric Administration (NOAA) Summary of the Day Datasets. The
frequency of the precipitation observations for the Illinois soil moisture stations is hourly.
The climate in the state is sub-humid, and it receives an average rainfall of about 94 cm
per year.
The soil moisture and surface temperature datasets for the state of Oklahoma were
obtained from the Oklahoma Mesonet. The data delivery is the product of a combined
effort by the soil moisture research team and the Mesonet team at the Oklahoma
Climatological
Survey
(OCS).
Presently,
the
Oklahoma
Mesonet
comprises
approximately 114 stations where soil moisture, surface temperature, precipitation and
other land and meteorological parameters are monitored. However, during 1996-1997,
the network consisted of only 60 stations. In 1999, soil moisture sensors were installed at
approximately 40 additional sites. Thus, soil moisture data for these additional sites were
not available for the entire 1999 calendar year. For that reason, the data for this study is
from the 57 stations that have soil moisture data for the entire calendar year.
The soil-
water potential (used to derive soil-water content) is measured by a sensor which operates
on a heat dissipation principle. The sensor is identical to that used in the two Southern
Great Plains Networks (the ARM/SWATS and the ARS/SHWMS network).
The soil-
water potential measured and a soil-water retention curve are used to derive the soilwater content. The soil-water content is derived at depths of 5 cm, 25 cm, 60 cm, and 75
cm.
Soil moisture and surface temperature are measured at these depths daily on a 30-
minute interval. Only the observation values of soil moisture and surface temperatures
corresponding to the local time zone (LTZ) of the SSM/I overpasses were extracted for
22
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
this study. The Oklahoma stations are predominantly located on four soil types, which
are loams, sandy loams, silty loams, and silty clayey loams, and the average spacing
between the stations is about 32 km (Basara et al., 2001). Ancillary GCIP/EOP Surface
Oklahoma Mesonet precipitation datasets were derived by the Joint Office of Science
Support/University Corporation for Atmospheric Research. The datasets were retrieved
electronically from the JOSS Data Management Center via the CODIAC system. The
observation frequency of the precipitation data for the Oklahoma Mesonet is every 15
minutes. The climate of this state is sub-humid with an average rainfall of approximately
78 cm per year.
3.2.2
Satellite Data
Several studies using SSM/I data have been conducted for a wide array of land
and atmospheric applications. These studies include snow cover and precipitation
classification (Grody, 1991; Grody and Basist, 1996), surface wetness monitoring (Basist
et al., 2001), land surface temperature retrievals (McFarland et al., 1990), effects of
heterogeneity in vegetation and rainfall (Lakshmi et al., 1997b), and soil and plant
moisture content (Jackson and Schmugge, 1989; Owe et al., 1992; Lakshmi et al., 1997a).
SSM/I measures brightness temperatures at seven different channels over land and
ocean, but does not distinguish land from atmosphere or ocean. The SSM/I data is from
the Defense Meteorological Satellite Program (DSMP) F13 spacecraft, which operates in
a sun-synchronous orbit at an altitude of 833 km with an overpass at approximately 1815
local time (LTZ) at frequencies of 19.35 GHz, 22.235 GHz, 37.0 GHz, and 85.5 GHz.
All channels operate in both vertical and horizontal polarizations with the exception of
the 22.235 GHz channel which is only vertically polarized. The brightness temperatures
23
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
are measured across a 1400 km swath width at 53.1° of incidence angle. The spatial
resolution varies with frequency. The footprint size at 19 GHz is 69 km x 43 km, but the
data obtained for this study was reprocessed onto a global, equal-area (EASE) projection
to a 25-km cell resolution by the National Snow and Ice Data Center (NSIDC) in
Boulder, Colorado. The data used in this study are from the ascending orbit of the SSM/I
because during this period the conditions of the land surface appear relatively
homogeneous to the sensor (Lakshmi et al., 1997a). This allows for much easier
interpretation and analysis. A thorough description of the SSM/I instrument is found in
Hollinger et al. (1990) and Long and Daum (1998).
The vegetation data was derived from the Normalized Difference Vegetation
Index (NDVI) from the Advanced Very High Resolution Radiometer AVHRR onboard
NOAA polar orbiting satellites. The vegetation data used in this study are bi-weekly
composite observations, which minimize the amount of cloud contamination.
resolution is about 8 km x 8 km.
The
Sensor degradation and aerosol attenuation have been
corrected by the Global Inventory Monitoring and Modeling Studies (GIMMS). The data
were aggregated for retrievals and comparisons with the SSM/I and in situ datasets by
taking the average of 9 pixels. The latitudes and longitudes of each of the Oklahoma and
Illinois stations were used to center the pixel in the vegetation data. The vegetation
parameters that were derived from NDVI data are the Leaf Area Index (LAI) and the
Canopy Fraction (CF) (discussed in the next section). The vegetation parameter LAI can
be estimated from NDVI because it represents the relative seasonal changes in vegetation
rather than vegetation amount. In the literature, there are many relationships to derive
LAI as a function of NDVI, but these are mostly land cover dependent.
24
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Here, an
exponential formulation is used because this relationship better follows the canopy
geometry and is not canopy type specific, as are the others. It is important to note the
significance of the relationship between
and
NDVI
to derive this variable. The L A I estimates from
LAI
NDVI
as there is no “true” relationship
are highly dependent upon certain
factors such as canopy geometry, leaf and soil optical properties, sun position and cloud
coverage. These variations of N D V I and
1991).
LAI
best follow Beer’s Law (Baret and Guyot,
can be rewritten in terms of N D V I as
LAI
where
LAI
N D Vhs
is the
= — •In
-k
N D V I - N D V IX
N D V I bs- N D V I x
value for bare soil,
NDVI
NDVD
is the asymptotic value of N D V I
when L A I tends towards infinity; and k is the extinction coefficient that controls the slope
of the relationship (Asrar et al., 1985). The
N D Vhs
and
ND Vh
were extracted for each
station in Illinois and Oklahoma for 1998 and 1999 considering that the vegetation has a
full cycle in one year. The
N D V Im
The
N D Vhs
was set equal to the minimum
was set equal to the maximum
N D Vhs
NDVI
NDVI
value and the
value of the station and year considered.
values range from 0.13 to 0.19 for Illinois, and from 0.14 - 0.19 for
Oklahoma. The value of N D V L o for both datasets was reached with an L A I value of 7.5.
The extinction coefficient
kndvi
(Baret and Guyot, 1991).
monthly estimated
LAI
is equal to 0.93 for an average leaf inclination of 50°
Figures 3-1 and 3-2 show the temporal evolution of the
for each of the 19 stations in Illinois and the 57 stations in
Oklahoma for 1998 and 1999. During the colder months (December - February), there is
less variability than during the warmer months (July - August). In the winter, the
2
2
values range from 0 to 1 (m m ') whereas in the summer, the
9
LAI
LAI
values range from less
9
than 1 to slightly over 4 (m m‘ ).
25
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.3
CALIBRATION OF THE RADIATIVE TRANSFER MODEL
The calibration of the forward model discussed in Chapter 2 was performed using
daily data from 57 stations of the Oklahoma Mesonet and bi-weekly data from 6 stations
of Illinois for a period of three months (June to August 1998).
This physically based
radiative transfer model was used for the simulation of the SSM/I brightness
temperatures.
The brightness temperature simulations were carried out using in situ
surface temperature and soil moisture from the Oklahoma Mesonet and the Illinois
stations.
The ‘free’ parameters used in the calibration were the surface parameters
(roughness parameter h and polarization-mixing factor Q).
factor Q was set at zero.
The polarization-mixing
This assumption might not be appropriate at the SSM/I
frequency channels since the surface appears rougher than at lower frequencies (where Q
~ 0-0.3) (Jackson, 1993) but is the most re-assumable approximation under the present
conditions. There are no measured references of the polarization-mixing factor Q in the
literature. For the roughness parameter h, the values of 0.7 and 0.4 were assigned for
Oklahoma and Illinois, respectively. These calibrated values were used to simulate the
brightness temperatures in both vertical and horizontal polarizations for a period of 21
months. The vegetation parameter b was derived from b values as reported in Jackson
and Schmugge (1991), and yielded a value of 0.4 for both the Oklahoma Mesonet and the
Illinois stations.
The vegetation water content VWC was derived from the NDVI-LAI
relationship discussed in the previous section. Table 3-1 presents a summary of the input
parameters of the forward model. The left column lists the sensor and land parameters
with their respective units, and the right column lists the value or the type of data used.
26
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figures 3-3 and 3-4 show the SSM/I and simulated 19 GHz (H-pol) brightness
temperatures for the Oklahoma Mesonet and the Illinois stations. For the Illinois stations,
a correlation value R2 of 0.92 and a RMSE of 7.4 K were obtained with the vertically
2
,
polarized simulated brightness temperatures and a R of 0.96 and a RMSE of 2.9 K with
the horizontally polarized simulated brightness temperatures.
For the Oklahoma
Mesonet, a correlation value R of 0.95 and a RMSE of 6 K (number of observations=
10633) were obtained with the vertically polarized simulated brightness temperatures and
a R2 of 0.97 and a RMSE of 4.1 K with the horizontally polarized simulated brightness
temperatures. As expected, better correlation coefficients and minimum standard errors
were obtained with the horizontal polarization (Jackson, 1997). This demonstrates that
the 19 GHz horizontal polarized channel is more sensitive to soil moisture.
27
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.4
RESULTS
3.4.1
Surface Temperature Retrieval
The
surface temperature was estimated using the
observed brightness
temperatures from the SSM/I sensor following the approach by Hiltbrunner et al. (1994).
The vertical and horizontal polarizations of the SSM/I 19 GHz channel were used to
estimate the surface temperature.
The algorithm is given by the equation,
\}Tm 9 v - ( k - \ ) T m 9H ~]
As = ------------------------------------
L3 -AI
where & is a parameter to weight the vertical and horizontal polarization contributions
from the sensor, Tu is the brightness temperature at polarization V or H, and the ex
parameter is an equivalent emissivity. The ex parameter depends on the actual H and V
emissivities and the k factor, and it is set at 1 (Kerr et al., 1999). The k and ex parameters
were calibrated for 3 months (June to August 1998), and the best correlation was
obtained with a k value of 1.7 and an ex value of 0.99.
These calibrated values are
different than the k and ex values that Kerr et al. (1999) used (k= \ .95 and ex=l) because
the retrieved surface temperatures were over high latitude continental regions.
All available SSM/I brightness temperatures were used to estimate the surface
temperature for the Illinois stations and the Oklahoma Mesonet for 1998 and 1999. The
estimated surface temperatures were compared to the in situ top-layer soil temperatures
for Illinois and Oklahoma for the two years. The comparisons of the estimated and in
situ temperatures for Oklahoma and Illinois are shown in Figs. 3-5 and 3-6, respectively.
The results produced correlation coefficients of 0.82 for the Oklahoma Mesonet with a
28
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
RMSE of 3.1 K. For the Illinois stations, a correlation coefficient of 0.86 was obtained
with a standard error of 3.2 K. The statistics of these results are presented in Table 3-2.
The correlation of in situ and retrieved surface temperatures for the Oklahoma Mesonet
and the Illinois stations is within a similar range obtained in previous studies (McFarland
et al., 1990; Weng and Grody, 1998) on the retrieval of surface temperatures with the
SSM/I sensor.
3.4.2
Soil Moisture Retrieval
An iterative minimization algorithm is used to retrieve soil moisture using SSM/I
19 GHz and simulated averaged (// and V polarizations) brightness temperatures
normalized with the surface temperature (TBavg(ssM/i)-TBavg(simuiated)/Ts) where Ts is the
surface temperature in Kelvin.
Surface temperatures and surface soil moisture
observations from the Oklahoma Mesonet corresponding to the SSM/I local time zone
(LTZ) were used.
For the Illinois stations, this was not possible due to the time
variability of the in situ soil moisture measurements.
For Oklahoma, single
measurements from each station were used since the spacing between the stations (32
km) is similar to the data grid size of the SSM/I (25 km). For Illinois, the SSM/I data
were aggregated to compare to the approximate spacing of the stations (-93 km) by
taking the average of 4 pixels (-100 km). The latitudes and longitudes of the stations
were used to center the pixel.
A comparison of all in situ and retrieved soil moisture observations revealed that
for both networks, the retrieval model overestimated the soil moisture in the low moisture
range (< 0.15). The model also underestimated the moisture in the soil over one-third of
29
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
the time when it was highly vegetated (3>LAI>6).
9
The correlation coefficient (R )
between the retrieved and in situ soil moisture of the Illinois stations (number of
observations=642) is 0.38 with an RMSE of 0.11. The correlation coefficient for the
Oklahoma Mesonet (number of observations=10633) is 0.37 with a RMSE of 0.19. The
poor correlation coefficients and the large RMSE are likely due to several reasons. First,
the wavelength of the 19 GHz channel is only able to penetrate a few millimeters into the
soil in contrast to the in situ observations of the 10-cm layer. Furthermore, the presence
of vegetation strongly attenuates the signal received.
Second, there is inherently a
significant variation in the surface temperatures but not in soil moisture throughout the
year within the Oklahoma Mesonet (i.e. a value of soil moisture=0.20 might remain
constant while surface temperature values range from 20 to 30 °C over a given period).
Third, the retrieval error is likely related to when soil moisture measurements are
collected (time of day) for the Illinois stations, i.e. in the case where the soil moisture
profile may not be uniform. The retrieval error may also be related to errors in the probe
measurements for the Illinois case (i.e. calibration error of the sensor is 4.5 % by
volume). Lastly, the vegetation water content values used in the radiative transfer model
to retrieve soil moisture were empirically derived from bi-weekly NDVI data as opposed
to using in situ measurements, and seasonal values of the vegetation parameter b from
experimental data were used instead of in situ seasonal values.
These results may be daunting; however, one of the goals of climate and
hydrological models is to attain the ability to monitor soil moisture over large areas.
Averages of the data were performed to investigate the quality of the soil moisture
retrieval.
Better correlation results were obtained with monthly averages of in situ and
30
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
retrieved soil moisture than with daily values of soil moisture in both states. Figures 3-7
and 3-8 show the temporal variability of monthly averages of soil moisture for the two
year period. For the Illinois stations, a R2 of 0.69 with a RMSE of 0.04 was obtained for
the monthly averages of the retrieved and in situ soil moisture. This error is similar to the
error that Vinnikov et al. (1999a) obtained with averaged soil moisture data of the top 10cm soil layer using the SMMR instrument over the Illinois region. For Oklahoma, a R of
0.54 with a RMSE of 0.04 was obtained for the monthly averages of the retrieved and in
situ soil moisture. These correlation values are similar to those of a previous study (Teng
et al., 1993) using the SSM/I for the retrieval of soil moisture. Table 3-3 presents other
statistical parameters calculated from the retrieval.
Soil moisture observations at 0-1 cm or yet at 0-2.5 cm would be ideal to use to
carry a soil moisture experiment. However, soil moisture at these depths is not available
on a routine basis because of the difficulty in planting automated probes in the thin layer
close to the surface. It is our rationale to use 0 - 5 cm and 0 - 10 cm soil moisture for the
forward and inverse model because the vegetation for most of the stations is grass. In
addition, 0-1 cm soil moisture is seldom measured on a regular basis. Our assumption
that the soil moisture profile is “uniform” i.e. 0-1 cm soil moisture reasonably reflects the
0-5 cm and 0-10 cm soil moisture. As most of the sites are grass-covered sites, this
approximation also closely represents grasslands.
Also, in a study by Jackson, we find
that for a region where the vegetation cover is grass, the soil moisture profile in the top 5
or 10 cm is uniform and therefore, the top 1 cm is related to the 5 or 10 cm layer (1997).
Vinnikov (1999a) also used a soil moisture dataset from the upper 10-cm layer for a
correlation study of the SMMR instrument.
31
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The temporal evolution of in situ and retrieved soil moisture and precipitation for
5 stations from Illinois and 8 stations from the Oklahoma Mesonet is shown in Figs. 3-9
and 3-10.
The values on the graphs correspond to the average of daily in situ and
retrieved soil moisture and precipitation for 1998 and 1999. The plots for both networks
show a qualitative agreement between the in situ and retrieved soil moisture and
precipitation. Tables 3-4 and 3-5 show the station ID, location, longitude, latitude, and
soil texture of these stations for the Illinois stations and Oklahoma Mesonet.
The interpolation technique Kriging, which interpolates a regularly- or
irregularly-gridded set of points Z = f(x,y), was used to produce maps to show the spatial
distribution of the SSM/I 19 GHz brightness temperatures and the retrieved soil moisture
for 57 stations of the Oklahoma Mesonet. Spatial maps of four consecutive days of the
SSM/I Tbs and the retrieved soil moisture show a rain event on day 1, and three days after
rainfall to show the soil moisture dry drown (see Figs. 3-11 through 3-19).
The rain
event in Oklahoma occurred on 5 April, 1999 and distributed approximately 291 mm of
rain throughout the state.
The brightness temperatures after the rain event get gradually warmer with time
(Figs. 3-12, 3-14, 3-16, and 3-18). It is evident that the soil moisture content decreases
with time (Figs. 3-13, 3-15, 3-17, and 3-19). This demonstrates the ‘short memory’ of
soil moisture which is approximately 3 - 5 days. The retrieved soil moisture compares
well with the observed brightness temperatures.
The effect of rainfall on the brightness temperatures and soil moisture on a
smaller scale is illustrated in Fig. 3-20 with a station from the Oklahoma Mesonet. The
plot shows a precipitation event on day 1 along with the SSM/I and simulated brightness
32
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
temperatures, and the in situ and retrieved soil moisture for five consecutive days after
the precipitation event. The station is located at 34.30 °N and -95.99 °E near Lane Town,
Oklahoma. The effect of precipitation on the brightness temperatures and soil moisture is
what we expected, as the SSM/I and simulated brightness temperatures increase linearly
with time; conversely, the observed and the retrieved soil moisture decrease with time.
This shows a moisture content loss of approximately 4%.
33
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3.4
CONCLUSIONS AND DISCUSSION
This study utilized a physically based model (radiative transfer model) to derive
soil moisture and temperature of the top layer ( 0 - 5 cm) for the Oklahoma Mesonet and
( 0 - 1 0 cm) for the Illinois stations for 1998 and 1999.
The data from June to August
1998 were used to calibrate the values of the roughness parameters (Q and h). The
calibration was carried out by iteratively adjusting these values in the model while
comparing to the SSM/I data.
The surface temperature was retrieved using an empirical model and the vertical
and horizontal polarizations of the SSM/I 19 GHz channel to compare with in situ surface
temperatures from the states of Oklahoma and Illinois.
The estimated and in situ
temperatures resulted in a R2 of 0.82 for the Oklahoma Mesonet with a RMSE of 3.1 K,
and a R2 of 0.85 for the Illinois stations with a RMSE of 3.2 K. The soil moisture of the
top layer was retrieved using an iterative technique for minimization error. The average
of the vertical and horizontal polarizations of the SSM/I and simulated brightness
temperatures was used to derive soil moisture. The monthly averages of soil moisture
were retrieved with an error of 0.04 for Oklahoma and Illinois states. The predicted soil
moisture was poorly correlated with the in situ soil moisture. This is due to several
factors; first, the wavelength of the 19 GHz channel is only able to penetrate a few
millimeters into the soil in contrast to the in situ observations of the 10-cm layer.
Furthermore, the presence of vegetation strongly attenuates the signal received. Second,
there is inherently a significant variation in the surface temperatures but not in soil
moisture throughout the year within the Oklahoma Mesonet (i.e. a value of soil
moisture=0.20 might remain constant while surface temperature values range from 20 to
34
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
30 °C over a given period).
Third, the retrieval error is likely related to when soil
moisture measurements are collected (time of day) for the Illinois stations i.e. in the case
where the soil moisture profile may not be uniform. The retrieval error may also be
related to errors in the probe measurements for the Illinois case. Lastly, the water content
values used in the radiative transfer model to retrieve soil moisture were empirically
derived from by-weekly NDVI data as opposed to using in situ measurements, and
seasonal values of the vegetation parameter b from experimental data were used instead
of in situ seasonal values. This proves the need for large field experiments in which
vegetation and surface characteristics are measured.
Comparisons of satellite data which represent aerial averages to point data
collected on the ground by in-situ networks is at best an approximate technique. We
make these observations in hope that we may eventually trust the accuracy of the satellite
data in regions where these point observations are not available. In fact, comparing point
estimates to aerial estimates should be viewed as attempting to verify that the two are in
the same “ballpark” rather than match exactly. The spatial variability of land surface
properties such as soil moisture and surface temperature are large, and thus cannot be
captured by in situ measurements separated by great distances i.e. 100 kilometers or so
apart. Well calibrated satellite data, on the other hand, do capture some of this spatial
variability.
In the future, microwave missions will include SMOS (Soil Moisture and Ocean
Salinity) and HYDROS (Hydrosphere State Mission).
These are L-band sensors and
their retrievals will have a better accuracy than the 19 GHz channel from SSM/I. But in
this context, SSM/I data can still be used for initial estimates of soil moisture and surface
35
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
temperature which can form initial guesses to start the retrieval process with SMOS and
HYDROS sensors.
Furthermore, SMOS and HYDROS will not have complete daily
coverage. In this regard, interspersing SMOS, HYDROS retrievals of soil moisture and
surface temperatures with SSM/I estimates may go a long way to enhance data
continuity, which is useful in Global Climate Models. The SSM/I sensor can provide a
long 20-year span of data, which is very useful and can be harnessed in this regard.
36
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
SENSOR PARAMETERS
Frequency, GHz
Polarization
Viewing angle 6 (degrees)
SOIL
Sand %
Clay %
Wilting point, WP
Bulk density (g cm'3)
Roughness coefficient, h (cm)
Polarization mixing factor, Q
Soil temperature
Soil moisture %
VEGETATION
Vegetation parameter, b
Vegetation water content, VWC (kg m'2)
Canopy fraction
Canopy temperature
19
H, V
53.1
in situ
in situ
in situ (IL) and simulated (OK)
1.25
calibrated 0.7 (OK), 0.4 (IL)
0
in situ
in situ
0.4
derived from NDVI
derived from NDVI
in situ
Table 3-1. Radiative transfer model inputs.
37
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Surface Tem perature, Ts (0 - 5 cm)
Oklahoma
Observed Retrieved
289.9 K
285.6 K
h
9.2 K
10.8 K
a
0.82
R2
RMSE
3.1 K
4.3 K
Bias
Illinois
Observed
Retrieved
294.7 K
288.9 K
9.7 K
8.8 K
0.86
3.2 K
5.8 K
Table 3-2. Mean, standard deviation, correlation coefficient, and root mean square of
the retrieved and in situ surface temperatures for the Oklahoma Mesonet and the Illinois
stations.
38
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Top Layer of Volumetric Soil Moisture
E
a
R2
RMSE
Bias
Oklahoma
Observed
Retrieved
0.27
0.28
0.03
0.04
0.54
0.04
0.01
Illinois
Observed
Retrieved
0.19
0.22
0.05
0.06
0.69
0.04
0.03
Table 3-3. Mean, standard deviation, correlation coefficient, and standard error of the
monthly averaged retrieved and in situ soil moisture for the Oklahoma Mesonet and the
Illinois stations.
39
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
STID
LON
LAT
ELEV (m)
SOIL TYPE
SAND%
CLAY %
STIL
97°09’50”
36°12’11”
272
Silty clay loam
19
32
MARE
97°21’28”
36°06’44”
331
Sandy clay loam
54
20
PERK
97°04’81”
36°99’83”
292
Loam
52
12
ACME
98°00’56”
34°80’56”
397
Sandy loam
73
9
GOOD
101°60’14”
36°60’17”
996
Loam
41
23
SEIL
99°04’06”
36°19’03”
540
Loam
43
19
NOWA
95°60’78”
36°74’36”
206
Silty loam
27
13
WIST
94°68’81”
34°98’47”
143
Silty loam
21
17
Table 3-4. Station ID, location, longitude, latitude, and soil texture for Stillwater, Marena, Perkins,
Acme, Goodwell, Seiling, Nowata, and Wister stations in Oklahoma.
STID
ICC
LLC
OLN
FRM
SIU
LON
89°32’
89°37’
88°06’
89°53’
89°14’
LAT ELEV (m)
40°42’
207
39°31’
177
38°44’
144
133
38°31’
37°43’
137
SOIL TYPE
Silt loam
Silt loam
Silt loam
Silt loam
Silt loam
SAND%
CLAY%
-
-
38
23
28
28
18
12
12
12
Table 3-5. Station ID, location, longitude, latitude, and soil texture for Peoria,
Springfield, Belleville, Carbondale, and Olney stations in Illinois.
41
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Estimated LAI values for Illinois
soil moisture stations 1998-1999
4
+
+
*
+ f
<
_i
2
+
i; t I +
:: t t
+
0
JAN-9B
APR-98
JUL-9B
OCT-98
JAN-99
APR-99
JUL-99
OCT-99
Figure 3-1. Time series of the estimated LAI for the Illinois
stations.
42
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Estimated LAI values for Oklahoma
soil moisture stations 1998-1999
m m !
|i i ii 11 in | m ii i n i |n i ii [ i n [ ii i u n i 11 n 11 iL'r u'p TTTT
6
4
2
11.. il..
□
JAN-98
APR-98
JUL-98
OCT-98
JAN-99
APR-99
JUL-99
OCT-99
Figure 3-2. Time series of the estimated LAI for the
Oklahoma Mesonet.
43
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
S IM U LA T E D v s . SSM I T B„
320
300
I
05 280
2
260
D
E
cn
240
220
220
240
280
260
300
320
SSMI Tb 19H (K)
Figure 3-3. Simulated brightness temperatures (19H) with
the model and SSM/I brightness temperatures (19H) for the
Oklahoma Mesonet.
44
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
S IM U LA T E D vs. S S M I T B„
320
300
280
2
260
240
.xx'
220
220
240
260
280
320
SSMI Tb 19H (K)
Figure 3-4. Simulated brightness temperatures (19H) with the
model and SSM/I brightness temperatures (19H) for the
Illinois stations.
45
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
R ETRIEVED v s . IN SITU T s
^ =0.82
M=1063J
240
260
280
300
In S itu
320
T, (K)
Figure 3-5. Retrieved surface temperatures versus in situ
surface temperatures for the Oklahoma Mesonet.
46
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
340
RE T R IE V E D v s . IN SITU T s
340
320
XX X
.XX
300
I-"
X!
(>D
0)
280
X XX
xx
260
240
240
260
300
280
In S itu
320
Ts (K)
Figure 3-6. Retrieved surface temperatures versus in situ
surface temperatures for the Illinois stations.
47
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
340
TIME SERIES OF IN SITU AND RETRIEVED SOIL MOISTURE
1.40
Number of averaged points = 642
R2 = 0.69
Soil moisture,
i.30
i.20
0.10
in situ soil moisture
retrieved soil moisture
o.oo
Jan-98
Apr-98
Jul-98
Jan-99
Oct-98
Apr-99
Jul-99
Oct-99
Figure 3-7. Retrieved soil moisture versus in situ soil moisture of the top
layer for the Illinois stations.
48
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
TIME SERIES OF IN SITU AND RETRIEVED SOIL MOISTURE
0.40
Num ber of averaged points = 10633
R 2 = 0 .5 4
Soil moisture,
0.30
0.20
0.10
in situ soil moisture
retrieved soil moisture
o.oo
Jan-98
Apr-98
Jul-98
Oct-98
Jan-99
Apr-99
Jul-99
Oct-99
Figure 3-8. Retrieved soil moisture versus in situ soil moisture of the top
layer for the Oklahoma Mesonet.
49
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
volumetric soil moisture
SOIL MOISTURE OF THE 10 CM SOIL LAYER
0.3
0.2
0.1
0.0
Apr-98
Jul-98
□ precipitation (in)
Oct-98
Jan-99
Apr-99
A observed soil moisture
Jul-99
Oct-99
o retrieved soil moisture
Figure 3-9. Time series of the averaged observed and retrieved soil
moisture for the top layer for Stillwater, Marena, Perkins, Acme,
Goodwell, Seiling, Nowata, and Wister stations in Oklahoma.
50
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
SOIL MOISTURE OF THE 5 CM SOIL LAYER
o
o o
volumetric soil moisture
▲ A
0.3
O
O
▲
o
O
A A
O ▲
▲
O
o
0.2
7
o6
5
▲ p) 6
4
o
3
0.1
2
1
0.0
II
Apr-98
Jul-98
□ precipitation (in)
H
Oct-98
Jan-99
Apr-99
A observed soil moisture
Jul-99
Oct-99
o retrieved soil moisture
Figure 3-10. Time series of the averaged observed and retrieved soil
moisture for the top layer for Peoria, Springfield, Belleville, Carbondale,
and Olney stations in Illinois.
51
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0
precipitation (in)
0.4
P r e c ip ita t io n (m m )
1291
146
I.
Figure 3-11. Precipitation in the Oklahoma Mesonet, 5 April, 1999.
m
Figure 3-12. SSM/I TB, 5 April, 1999.
Figure 3-13. Retrieved soil moisture.
Figure 3-14. SSM/I TB, 6 April, 1999.
Figure 3-15. Retrieved soil moisture.
Figure 3-16. SSM/I TB, 7 April, 1999.
Figure 3-17. Retrieved soil moisture.
'JX*
Figure 3-19. Retrieved soil moisture.
Figure 3-18. SSM/I TB, 8 April, 1999.
0 .2 5
52
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
TB (H ) O b s e rv e d
and S im u la te d
Lane Station, 34.30° N -95.99° E
^
5/9/1998
5/10/1998
precipitation (in) X 2
—
observed soil moisture
•
—
5/11/1998
5/12/1998
5/13/1998
TB (H) observed
A TB (H) simulated
retrieved soil moisture
Figure 3-20. Precipitation, observed and simulated brightness
temperatures, and volumetric soil moisture for five consecutive
days in 1998 for Lane Station, Oklahoma.
53
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
C h a p t e r 4 - A M ic r o w a v e M u l t i -s e n s o r S t u d y
of
L a n d S u r f a c e E m i s s iv it y
over
The
S o u t h e r n G r e a t P l a in s , O k l a h o m a
4.1
INTRODUCTION
A long term goal for hydrologists is to be able to monitor soil moisture at a high
temporal repeat (i.e. daily), high spatial resolution (meters) and with wide coverage (i.e.
global) from remote sensing methods. It has been demonstrated that microwave satellite
sensors provide an effective tool for soil moisture estimation (Schmugge, 1985) with
accuracies within 5% (Jackson, 1997).
This is due to the sensitivity of the microwave
satellite sensors to the dielectric properties of liquid water (« 80) and dry soil (« 4)
(Schmugge, 1985). Microwave satellite sensors provide data to carry out studies over
long time periods with global coverage.
Despite the all-weather coverage and good accuracy retrievals of microwave
satellite sensors with certain types of low density and low water content vegetation (i.e.
wheat, com, etc), an important issue remains regarding the reliability of land surface
dynamics derived from microwave sensors.
Reliability is questioned when a sensor
54
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
signal is masked by densely vegetated areas (large biomass and water content) and when
land surface dynamics are retrieved over large regions, considering the poor spatial
resolution of microwave passive sensors.
In this study, we evaluate the aspects of low and high spatial resolution at
multiple frequencies by comparing in situ observations with sensor observations over the
Southern Great Plains, USA from the Passive/Active L- and S-band sensor (PALS), the
Polarimetric
Scanning
Radiometer at
C-band
(PSR/C),
the
Tropical Rainfall
Measurement Mission Microwave Imager (TMI) at 10 GHz (X-band) and 19.4 GHz, and
the Special Sensor Microwave Imager (SSM/I) at 19.4 GHz and 37 GHz.
We also
investigate the effects of sensitivities of brightness temperatures on land surface
parameters for the aforementioned sensors.
Lastly, we examine the effects of land
surface heterogeneity by using a simulation model based on passive microwave
measurements at all the frequencies used in this study.
Single-frequency studies have been carried out to derive soil moisture from active
and passive microwave radiometers over the SGP99 region (Njoku et al., 2002; Bolten et
al., 2003; Le Vine et al., 2001; Guha et al., 2003; Jackson et al., 1999; Jackson et al.,
2001; Jackson et al., 2002a; Jackson et al., 2002b).
Previous studies have examined the
sensitivity of high frequency brightness temperatures to land surface dynamics
(Choudhury et al., 1987; Lakshmi et al., 1998).
Studies have been carried out to
investigate the effect of land surface heterogeneity on land surface processes (Entekhabi
et al., 1989; Famiglietti and Wood, 1994) and on simulated brightness temperatures
(Lakshmi et al., 1998; Lakshmi et al., 1997).
55
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Whereas previous research over the SGP99 region has studied single and dual
frequencies for relating the emissivity to soil moisture, this study is the first of its kind to
use all SGP99 frequencies/sensors to compare the sensitivity of emissivity to soil
moisture.
56
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.2
DATA AND METHODS
4.2.1
Study Area
The SGP99 experiment took place in the Little Washita watershed near
Chickasha, Oklahoma from 8 July through 20 July, 1999. The region covers an area of
603 km2 and consists of different vegetation cover types which include bare, pasture, and
various agricultural crop surfaces. The climate of the state is sub-humid with an average
rainfall of approximately 78 cm per year. The topography of the watershed is very flat
(maximum relief of less than 200 m) (Jackson et al., 2002b). Land use is dominated by
agriculture, mainly winter wheat and pasture. Soil texture varies throughout the
watershed from fine to coarse, but the characteristic texture is 30% sand and 20% clay
(Jackson et al., 2002b).
In situ data was utilized for 18 field sites that vary between five different crop
types: rangeland (10 sites), wheat (4 sites), fallow (1 site), corn (2 sites), and alfalfa (1
site). The in situ data consists of observations of surface temperature and soil
temperatures at depths of 1 cm, 5 cm, and 10 cm, soil moisture observations at 0 - 2.5
cm, 2.5 - 5.0 cm, and at 5 cm, and a single biomass measurement taken at the beginning
of the experiment on 8 July. The field measurements were made between 0730 and 1200
CDT. The average spacing between the field sites is about 0.8 km.
vegetation water content of the fields ranged from 0 - 2.5 kg-m' .
57
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The average
4.2.2
Airborne Sensors
4.2.2.1 PALS
Aircraft data for SGP99 was retrieved from the Passive/Active L/S-band sensor
(PALS).
The radiometer provides data at L-band (1.41 GHz) and S-band (2.69 GHz)
with dual polarization. The PALS sensor has an incidence angle between 35° and 55°,
but the incidence angle was preset at 38° for the experiment. The instrument was flown
on a US C-130 aircraft over the SGP99 region from 8 July to 14 July (except 10 July),
1999 at a nominal altitude of 0.9 km with an approximately 300 x 400 footprint size in all
channels. The nominal daily time window for PALS coverage was 0830 to 1300 CDT
(1330 to 1800 UTC).
Additional details of the instrument design and engineering
performance are provided in Wilson et al. (2001).
4.2.2.2 PSR/C
The Polarimetric Scanning Radiometer (PSR) is an airborne microwave imaging
radiometer developed and operated by NOAA Environmental Technology Laboratory.
The PSR/C radiometer mounted on the NASA P3-B aircraft is capable of viewing
multiple angles within 70° due to a scan head drum rotated by a gimbal positioner
(Jackson et al. 2002b). In this study, a conical scan mode at 55° incidence from nadir is
used. The PSR/C is a multi-frequency radiometer (6.0, 6.5, 6.925, and 7.325 GHz), but
only vertical and horizontal polarization data from the 7.325 GHz channel is provided
here. The PSR/C flew at an altitude of ~ 8.2 km allowing an average footprint size of 2.3
km and a swath width of 25.5 km. The nominal daily time window for the aircraft coverage
was 0830 to 1130 CDT (1330 to 1630 UTC) on 8, 9, 11, 14, 15, 19 and 20 July, 1999.
Details on the PSR/C can be found in Jackson et al. (2002b).
58
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.2.3
Space borne Sensors
4.2.3.1 TRMM-TMI
The Tropical Rainfall Measurement Mission Microwave Imager (TMI) is one of
two satellite sensors used for this study. The TMI orbit extends from 35° N to 35° S at an
altitude of 402.5 km with a swath width of 897 km. There were a total of 25 orbits that
provided TMI data during the study period. Each orbit has a different start time of swath,
but the overpass times were generally made between 0600 and 1430 CDT (1100 to 1930
UTC). The TMI sensor has an incidence angle of 52.8° and is a nine-channel linearly
polarized system. It operates at frequencies of 10.7 GHz, 19.4 GHz, 21.3 GHz, 37 GHz,
and 85 GHz. In this study, only the 10.7 GHz and 19.4 GHz frequency channels are
used. The effective field of view (EFOV) at 10.7 GHz is 63 x 37 km, and at 19.4 GHz is
30 x 18 km. The TMI sensor takes approximately 91 minutes to complete an orbit, so it
makes approximately 15.8 orbits per day. The TMI provides one to three passes over the
SGP99 region per day. Further details concerning TMI can be found in Jackson et al.
(2001).
4.2.3.2 SSM/I
The Special Sensor Microwave Imager (SSM/I) mounted on the Defense
Meteorological Satellite Program (DMSP) provides data from the FI 1, F I3, and F14 US
spacecrafts, which operate in a sun-synchronous orbit at an altitude of 833 km. The start
time of swath in Local Standard Time (LST) for the F ll was predominantly between
0700 to 0900 and 1900, for the F13 between 0600 to 0700 and 1700 to 1800, and for the
F14 between 0900 and 1000. The ascending equatorial crossing time (UTC) for the FI 1,
F13, and F14 spacecrafts is 1925, 1743, and 2039, respectively. The SSM/I consists of
59
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
seven channels and a linearly polarized system that operates at frequencies of 19.4 GHz,
22.235 GHz, 37.0 GHz, and 85.5 GHz.
All channels operate in both vertical and
horizontal polarizations with the exception of the 22.235 GHz channel which is only
vertically polarized. Only the 19.4 GHz and 37.0 GHz channels are used in this study.
The SSM/I has a swath width of 1400 km at 53.1° of incidence angle. The footprint size
at 19.4 GHz is 69x43 km, and at 37.0 GHz is 37x28 km. A thorough description of
the SSM/I instrument is found in Hollinger et al. (1990) and Long and Daum (1998). The
key instrument characteristics of all the sensors used in this study are listed in Table 4-1.
60
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.3
RESULTS
4.3.1
Comparison o f Airborne and Space borne Observations
A comparison was made to evaluate the aspects of low and high spatial
resolutions at multiple frequencies between in situ observations and brightness
temperatures from the Passive/Active L- and S-band sensor (PALS), the Polarimetric
Scanning Radiometer at C-band (PSR/C), the Tropical Rainfall Measurement Mission
Microwave Imager (TMI) at 10 GHz (X- band) and 19.4 GHz, and the Special Sensor
Microwave Imager (SSM/I) at 19.4 GHz and 37 GHz. Brightness temperatures were
extracted from these sensors in all the aforementioned polarizations and frequencies. The
brightness temperature values from the closest footprint to the in situ field site were used
for the comparisons. The in situ data used for the comparisons are soil temperature
readings at 1 cm and soil moisture observations at 0 - 2.5 cm. The in situ measurements
were taken each day at 18 field sites for the entire length of the experiment, except for the
com, fallow and alfalfa fields.
This section presents the comparisons between emissivities of the sensors and the
in situ soil moisture for 4 different land cover field sites. The emissivities at a particular
frequency were estimated using the observed brightness temperature divided by the soil
temperature at 1 cm. Soil temperatures were observed in individual fields using handheld
infrared thermometers (IRTs).
Figures 4-1 through 4-4 show daily variations of the observed horizontal
polarization emissivity eH at different frequencies and the in situ soil moisture at 0 - 2.5
cm for the following field sites: rangeland, wheat, corn, and alfalfa, respectively. Note
that the emissivity at any frequency exhibits a dip corresponding to an increase in soil
61
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
moisture on 11 July following the rainfall on 10 July. This rainfall event is also reflected
by a soil moisture peak on all figures. For the wheat and alfalfa fields, there were no in
situ measurements of soil moisture and temperature on 10 July. Also, there were no in
situ measurements for the com and alfalfa fields beyond 14 July (Figs. 4-3 and 4-4).
The emissivity change, Ae was calculated for all sensors using the emissivity
values from 9 July and 11 July. The Ae is given in percentage (%) on all figures. The
results show that among the four fields, the largest decrease in emissivity is seen in the
wheat field (Fig. 4-2). In this field, the emissivity decreased 15.75% at L-band, 13.85%
at S-band, 7.37% at C-band, 6.30% at X-band, 3.60% at 19.4 GHz (TMI), 2.69% at 19.4
GHz (SSM/I), and 1.93% at 37 GHz. Conversely, the smallest decrease in emissivity is
seen in the com field (Fig. 4-3). Here, the emissivity decreased 6.92% at L-band, 5.36%
at S-band, 5.21% at C-band, 2.19% at X-band, 1.98% at 19.4 GHz (TMI), 1.75% at 19
GHz (SSM/I), and only 0.36% at 37 GHz. There is more than twice the difference in As
between the com field and wheat field (-8.83 %). This confirms that the emissivity has
higher sensitivity to soil moisture at low vegetation water content (wheat field: -0.35
2,
2
kg-nf ) at any frequency than at high vegetation water content (com field: -7.18 kg-m ).
It is evident that there is little variability in the emissivity during the soil moisture drydown period. This means that zle is very minimal from 14 July to 20 July (Figs. 4-1 and
4-2).
In general, all data show that the 1.41 GHz channel (L-band) has the widest
emissivity range and the emissivity range decreases with increasing frequency.
The
results also show that at frequencies higher than X-band (10.7 GHz), the range of the
emissivity does not change as abruptly as it does at L-band amongst the fields.
62
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The Ae and the rate at which it decreases can be related to the lower sensitivity to
soil moisture at higher frequencies (10 GHz - 37 GHz). This is further discussed in
Section 4.3.3.
Overall, qualitative agreement exists between the estimated emissivity
time series from all the frequencies and the observed soil moisture.
63
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.3.2
Observed and Simulated Emissivities
A forward microwave emission model (described in Chapter 2) is used to simulate
brightness temperatures at the polarizations of PALS (1.41 GHz and 2.69 GHz), PSR/C
(7.35 GHz), TMI (10 GHz and 19 GHz), and SSM/I (19 GHz and 37 GHz). Figure 4-5
shows the general approach of the model for retrieving the emissivities from each sensor.
The values for the vegetation parameter b and surface roughness h, at each field site were
extracted from Jackson et al. (1999). A semi-empirical relationship is utilized to estimate
the surface reflectivity rp as suggested in Wang and Choudhury (1981). The inputs to the
forward microwave emission model such as the soil parameters: sand, clay and bulk
density, and the soil temperature and soil moisture are from in situ data. The canopy
temperature was assumed to be equal to the surface temperature. In the case for high
frequencies such as 19.4 GHz and 37 GHz, an atmospheric correction model was used.
Here, the emissivities for all frequencies were estimated with the simulated brightness
temperatures divided by the soil temperatures at 1 cm.
Figures 4-6 and 4-7 are scatter plots of the observed and simulated emissivities in
the vertical and horizontal polarizations for three wheat field sites from 8 July to 14 July.
The scatter plots show that the emissivities at L- and S-bands have lower emissivity
values and larger emissivity ranges than the 19.4 and 37 GHz channels. The horizontal
1.41 GHz channel (L-band) resulted in a correlation coefficient of 0.92, and the vertical
polarization in a correlation coefficient of 0.85. The correlation coefficient decreases
with increasing frequency. The results also show that there is higher scattering in the
vertical polarization. The increased scattering can be related to the vertical structure of
the stems of wheat.
64
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.3.3
Sensitivity
A sensitivity analysis was conducted particularly to examine the effects of land
surface dynamics on polarization differences from simulated brightness temperatures at
all frequency channels in this study. The polarization difference is given by A T b
poI)-Tb(h-poI)-
= Tb(v-
The sensitivity analysis focused on the effect that variations of vegetation
water content VWC and surface temperature Ts had on the simulated
ATb
at each
frequency. The polarization mixing factor Q and the roughness height parameter h were
both set constant at 0 to observe solely the effects of the VWC on the sensitivity of A T b to
changes in soil moisture 8.
The VWC was varied from 0 to 6 (kg-m'2), the surface
temperature Ts from 20 °C to 33.5 °C, and the soil moisture 8 from 0 to 0.35. The
brightness temperatures were estimated using a constant soil textural type composed of
20% sand and 15% clay (loam soil), which is the common soil type in the SGP99 region.
The brightness temperatures were simulated at the frequencies, polarizations, and
incidence angle of each sensor using the forward microwave emission model discussed in
Chapter 2. The sensitivity of the polarization difference of brightness temperatures was
estimated using the following expression,
S(A Tb )
Ts
58
[4.1]
where S(ATB) is the polarization difference at 8= 34% and 0=14%, 88 is the volumetric
soil moisture difference taken to be equal to 20%, and Ts is the surface temperature. The
symbol 8 is used here to denote a difference. Table 4-2 shows the sensitivity values
obtained at 20 °C and 33.5 °C at each frequency with different values of vegetation water
content (0<VWC<6).
65
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The results show that the sensitivity of the polarization difference of brightness
temperatures is higher at any given frequency at lower surface temperature values (Ts=
20 °C) with low or no vegetation water content (VWC-0).
The sensitivity of the
polarization difference of brightness temperatures is a function of vegetation and
frequency. The sensitivity decreases as Wc and frequency increase, so at 37 GHz there is
little variability (Table 4-2). The results also show that the sensitivity values between Cband (7.35 GHz) and X-band (10.7 GHz) are comparable. Similarly, there are only slight
differences between the 19.4 GHz channel of TMI and SSM/I. The largest sensitivity
range is seen at L-band at Ts = 20 °C, showing that this frequency channel is optimal for
soil moisture sensing.
Nonetheless, this does not undermine the sensitivity to soil
moisture at other frequencies when the VWC is low. The sensitivity of the model is
limited beyond a VWC of 3 kg-m'2, which is seen at the higher frequency channels (19
GHz and 37 GHz) on Table 4-2.
4.3.4
Simulation Model
The effect of land surface heterogeneity on land surface processes has been
investigated in previous studies (Entekhabi et al., 1989; Famiglietti and Wood, 1994;
Guha and Lakshmi, 2002) and on simulated brightness temperatures (Lakshmi et al.,
1998; Lakshmi et al., 1997) with good results. In this study, we evaluate aspects of
sensor observations at low and high spatial resolutions and assess the impact of land
surface
heterogeneity
frequencies/sensors.
on
brightness
temperatures
derived
from
multiple
To achieve this, a numerical simulation model of brightness
temperatures was carried out based on polarizations and frequencies from PALS, PSR/C,
TMI, and SSM/I for five days. The simulation involved three parts: (1) A method for
66
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
generating 2-D grids of land surface parameters such as soil moisture, surface
temperature, vegetation water content, and surface roughness; (2) A forward microwave
emission model to simulate brightness temperatures from the aforementioned sensors;
and (3) An atmospheric model for high frequency channels (19.4 GHz and 37 GHz).
The simulation of brightness temperatures begins with the generation of land
parameters, which are inputs to the forward microwave emission model.
The land
surface parameters are soil moisture, surface temperature, vegetation water content, and
surface roughness.
The method generates three grids that are 10x10, 100x100 and
1000x1000 in resolution, comprised of the above parameters.
This is done with a
random number generator and a shuffle that removes low-order serial correlations. The
grids were generated with boundary limits defined by the land surface parameters
described below for a period of five days. The land surface parameters have the same
initial conditions (remain constant) in all grids for any given day, but they vary within
each grid due to its inherent resolution (standard deviation changes). This allows for the
simulation of brightness temperatures at the various spatial resolutions of the sensors (the
standard deviations increase with the resolution of the grid). The simulation also includes
different time conditions to account for the spatial heterogeneity of the land surface
parameters. The simulation begins on Day 1 with wet conditions (high soil moisture and
vegetation water content) and ends on Day 5 with relatively dry conditions.
The
boundary limits or maximum ranges that the land surface parameters could have during
the five days are as follows: the surface temperature varies between 293 K (Tmm) and 320
K (Tmax), soil moisture varies between 0.05 and 0.35, the VWC varies between zero and 3
kg-m' and the surface roughness varies between 0.15 and 0.32.
67
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The forward microwave emission model utilized to simulate the brightness
temperatures was described in Chapter 2. The brightness temperatures Tb were simulated
at the indicated frequencies, polarizations, and incidence angles of PALS, PSR/C, TMI,
and SSM/I using the three generated grids over five days. A particular grid was applied
to each sensor to generate a similar output produced by the sensor’s spatial resolution.
The 1000 x 1000 grid of land surface parameters was used at 1.41 GHz (L-band) and 2.69
GHz (S-band) for the PALS instrument and at 7.35 GHz (C-band) for the PSR/C sensor.
The 100x100 grid was used at 10 GHz (X-band) and 19.4 GHz for TMI.
10x10 grid was used at 19.4 GHz and 37 GHz for SSM/I.
Lastly, the
Figure 4-8 shows the
approach taken for the simulation of brightness temperatures for the retrieval of
emissivities at each frequency.
In each case, the resolution of the output grid of
brightness temperatures was the same as the resolution of the input grid. The parameters
used as inputs for the simulations used values which simulated conditions as realistic as
possible. The soil texture chosen is a silty loam with 20% sand, and 15% clay which is
the predominant soil type in the SGP99 region. The vegetation parameter varies between
0.085 and 0.095 which corresponds to various land cover types: alfalfa, legume, pasture,
and shrub. The bulk density was set constant at 1.2 g-cm' .
The emissivity, s, is defined as the ratio of a brightness temperature and its
effective surface temperature Teff-
Here, the emissivities were estimated for all the
frequencies and polarizations using the simulated brightness temperatures and the
generated surface temperatures (assuming Teff = Tsurf).
The results for Day 2 are
presented here. The initial conditions for Day 2 used in the simulations are shown in
Table 4-3.
The statistics of the simulations for Day 2: means, standard deviations,
68
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
coefficient of variations for the brightness temperatures and emissivities for all the
sensors are presented in Table 4-4. The coefficient of variation, Cy, is the ratio of the
standard deviation of the brightness temperature TB or emissivity s and its averaged
value.
The results show that at L-band the polarization difference of brightness
temperatures ATb is greater than at any other frequency or sensor (ATB(i.4 i gh:) >
ATB(freq>i.4 i gh:))■ The coefficient of variation for both the brightness temperatures and
emissivities decreases with increasing frequency. The coefficient of variation is always
greater at the horizontal polarization than at the vertical polarization.
Thereby, the
standard deviations for the brightness temperatures and emissivities were also greater in
the horizontal polarization. At L-band, the coefficient of variation for the horizontal
emissivity is 0.064, and at 37 GHz it decreases to 0.020. The simulations captured the
spatial effects on the emissivities discussed in Section 4.3.1. Here again, the L-band has
the largest difference in emissivity (z)e^(v-H)=0.11) between the two polarizations. As
the frequency increases, the zL^v-h) decreases. It is seen that at 37 GHz, the zle^v-H) is
only 0.03. This confirms that the sensitivity of the sensors to land surface parameters is
affected by their spatial resolution.
The results are comparable to those of a previous study (Guha and Lakshmi,
2002) where the spatial heterogeneity and scaling effects at C-band were investigated.
The authors used several grid sizes to show the effects of regridding on the polarization
difference of brightness temperatures. The results compare as follows, at a 1000x1000
grid the mean averaged brightness temperatures ATB obtained in this study is between
26.01 and 29.93, and the authors’ ATB is 21.37. At a 100x100 grid, ATg in this study
69
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
varies between 19.22 and 23.75, and the authors’ A T B is 20.37. Lastly, at a 10x10 grid,
A T Bm
this study varies between 11.26 and 15.93, and the authors’
ATB
is 20.35.
The spatial variability of the generated soil moisture grid (input) and the
simulated &// grid (output) at each frequency for Day 2 is depicted in Figures 4-9a through
4-9j. The three soil moisture grids (Figs. 4-9a through 4-9c) show the same pattern, but
the soil moisture at the 1000x1000 grid has a finer resolution. The emissivities at 1.41
GHz, 2.69 GHz, and 7.35 GHz, which were simulated using this grid, show the largest
difference in emissivity zte^v-H) (Table 4-4). The 1.41 GHz channel (L-band) has the
largest Eh range (Fig. 4-9d). The eh range decreases with increasing frequency. This is
because of the lower sensitivity to soil moisture at higher frequencies. The 7.35 GHz
channel (C-band) and the 10 GHz channel (X-band) share some similarities in the values
of the emissivities, brightness temperatures, and coefficient of variations due to the
closeness between these bands.
The 19.4 GHz and 37 GHz channels have lower
sensitivity to soil moisture as seen in Figs. 4-9i through 4-9j. Table 4-4 shows that there
is an eh change of only 0.019 at 37 GHz compared to 0.052 at 1.41 GHz. The
Cy
at 1.41
GHz is 0.064 and it decreases to 0.020 at 37 GHz. The spatial plots reveal that resolution
suffers considerably at higher frequencies (19.4 GHz and 37 GHz).
70
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
4.4
CONCLUSIONS AND DISCUSSION
This study evaluated the aspects of low and high spatial resolutions of emissivities
at multiple frequencies by comparing in situ observations with sensor observations over
the US Southern Great Plains from the Passive/Active L- and S-band sensor (PALS), the
Polarimetric
Scanning
Radiometer at
C-band
(PSR/C),
the
Tropical Rainfall
Measurement Mission Microwave Imager (TMI) at 10 GHz (X-band) and 19.4 GHz, and
the Special Sensor Microwave Imager (SSM/I) at 19.4 GHz and 37 GHz.
The results
show the 1.41 GHz channel (L-band) has the widest emissivity range of all the
frequencies studied.
For the four types of land cover fields used in this study, the
emissivity range decreases with increasing frequency. The results also show that the
emissivity at any frequency exhibits a dip corresponding to a sharp increase in soil
moisture, as seen on 11 July following the rainfall event on 10 July.
The range in the
emissivities also varied between the land cover types of the field sites. The emissivity
was found to be greatest at an agricultural cover type of wheat.
In general, the results
showed qualitative agreement between the estimated emissivity temporal series from all
the frequencies and the observed soil moisture.
This study also investigated the effects of land surface dynamics on polarization
differences from simulated brightness temperatures for the aforementioned sensors. The
results showed that (1) the sensitivity of the polarization difference of brightness
temperatures is a function of vegetation and frequency, (2) the sensitivity values between
C-band (7.35 GHz) and X-band (10.7 GHz) are comparable, (3) the 19.4 GHz channel of
TMI and SSM/I have only slight differences in sensitivity, (4) that each frequency can
provide soil moisture information if the vegetation water content is low, and (5) that
71
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
highly dense vegetation considerably masks the microwave signal, especially at low
spatial resolutions (19.4 GHz and 37 GHz).
Lastly, this study examined the effects of land surface heterogeneity by using a
simulation model based on passive microwave measurements at the aforementioned
frequencies. The results showed that at L-band the polarization difference is greater than
at any other frequency or sensor.
The polarization difference and the coefficient of
variations decrease with increasing frequency.
At high spatial resolution (L-band), the
pattern of the emissivities mimics the soil moisture pattern well. Also, at high spatial
resolution (L-band), there is a greater range in the emissivities than at any other spatial
resolution (S-, C-, X-bands, 19.4 GHz and 37.0 GHz).
The spatial plots reveal that
resolution suffers considerably at higher frequencies (19.4 GHz and 37.0 GHz).
72
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Sensor
PALS
Frequency
1.41 GHz
2.69 GHz
Nadir
angle
30° to 50°
7.35 GHz
TRMM -
10.65 GHz
TMI
SSM/I
19.4 GHz
37 GHz
2.3 km
53.1°
25.5
63 x 37 km
30 x 18 km
19 GHz
0.4
(at 1 km altitude)
55°
52.8°
Swath (km)
0.4 km
For SGP
38°
PSR-C
Spatial
resolution
69 x 43 km
37 x 28 km
790
1200
Table 4-1. Sensor name, frequency, nadir angle, spatial resolution, and swath
of the sensors used in study.
73
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
PALS
PSR/C
TMI
TMI
SSM/I
SSM/I
1.41 GHz
2.69 GHz
7.35 GHz
10.7 GHz
19.4 GHz
19.4 GHz
37 GHz
33.5°C
20°C
33.5°C
20°C
33.5°C
20°C
33.5°C
20°C
33.5°C
20°C
33.5°C
20°C
33.5°C
5.80
3.66
4.59
2.81
4.04
2.51
3.75
2.34
3.61
2.25
3.38
2.11
0.88
0.65
w c= 1
5.62
3.55
4.54
2.74
3.89
2.43
3.23
2.00
2.77
1.72
2.63
1.63
0.85
0.63
Wc= 2
5.45
3.35
4.24
2.60
3.51
2.17
2.22
1.37
1.64
1.02
1.56
0.97
0.83
0.61
Wc= 3
5.29
3.23
3.87
2.37
3.02
1.87
1.53
0.95
0.98
0.60
0.56
0.57
0.70
0.53
II
5.13
3.14
3.53
2.17
2.60
1.61
1.05
0.65
0.58
0.36
0.56
0.34
0.67
0.48
£
i/T
II
4.98
3.04
3.23
1.98
2.24
1.38
0.73
0.45
0.35
0.22
0.34
0.21
0.55
0.43
4.83
2.91
2.95
1.81
1.93
1.29
0.51
0.31
0.22
0.13
0.21
0.11
0.42
0.30
IIu
20°C
£
Ts
o
II
O
PALS
£
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
SENSOR
Table 4-2. Effects of the vegetation water content and surface temperature on the polarization difference of brightness
temperatures at various frequency channels.
10x10
Land Surface Parameters
100x100
1000x1000
a
V
a
M
a
0.26
0.005
0.26
0.009
0.26
0.094
306.65
0.28
306.65
2.34
306.65
13.37
2.72
0.009
2.72
0.086
2.72
0.77
0.24
0.084
0.24
0.041
0.24
0.022
Soil moisture
Surface temperature
(Kelvin)
Vegetation water content
(kg-m'2)
Surface roughness
(cm)
Table 4-3. Means and standard deviations of input parameters for the simulations at three
different scales.
75
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Sen so r
TBV
TbH
oTbV
aT„H
£y
£h
asv
oeH
CyTBy
V^BH
CySy
C£
290.23
261.65
4.67
5.30
0.92
0.81
0.029
0.052
0.016
0.020
0.031
0.064
29.93
294.96
265.03
3.55
4.74
0.94
0.85
0.036
0.053
0.012
0.018
0.038
0.063
28.58
297.71
271.70
3.05
3.48
0.95
0.86
0.023
0.040
0.010
0.013
0.024
0.046
26.01
296.86
273.10
3.17
3.42
0.96
0.88
0.024
0.039
0.011
0.013
0.025
0.045
23.75
299.95
280.73
2.43
2.36
0.97
0.91
0.018
0.032
0.008
0.009
0.019
0.035
19.22
301.46
285.53
1.65
2.05
0.97
0.93
0.015
0.026
0.005
0.007
0.015
0.028
15.93
304.43
293.17
1.36
0.61
0.98
0.95
0.009
0.019
0.002
0.005
0.009
0.020
11.26
ATb
pals
1.41 GHz
PALS
2.69 GHz
PSR
7.35 GHz
TMI
10 GHz
TMI
19 GHz
SSMI
19 GHz
SSMI
37 GHz
Table 4-4. Averaged brightness temperatures (v & h polarization), standard deviation of brightness temperatures (v &
h), emissivities (v & h), standard deviation of emissivities (v & h), coefficient of variation of brightness temperatures
(v & h), coefficient of variation of emissivities (v & h), and averaged polarization difference of simulations for all
sensors for Day 2.
0.40
1.00
Rangeland /V W C =0.16 kg-m'
i n te r p o la te d -
0.35
S en sor
A e(JU| 9 . Jum )
PALS 1.41 GHz
7.24%
PALS 2.69 GHz
6.78 %
PSR/C 7.35 GHz
6.88 %
TMI 10.7 GHz
3.87%
TMI 19.4 GHz
2.77%
SSM/I 19.4 GHz
1.80%
SSM/I 37.0 GHz
0.07 %
Soil moisture (0 - 2.5 cm)
0.85
0.25
0.20
0.15
0.80
0.10
so il m oisture
E m issiv ity , £ (H - p o l)
0.90
In situ volum etric
0.30
0 - 2.5 c m
0.95
0.75
0.05
0.00
0.70
Figure 4-1. Time series of the observed horizontal polarization emissivities for
all sensors and the in situ soil moisture for a rangeland field.
77
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Field Site: 23
W heat
VW C=0.35 kg-m'2
0.85
so il m oisture
Sensor
Ae(jUi9 . jUn i)
PALS 1.41 GHz 15.75 %
PALS 2.69 GHz 13.85 %
PSR/C 7.35 GHz
7.37 %
TMI 10.7 GHz
6.30 %
3.60 %
TMI 19.4 GHz
2.69 %
SSM/l 19.4 GHz
1.93 %
SSM/I 37.0 GHz
Soil moisture (0 - 2.5 cm)
0.15
0.80
0.10
0.75
0.05
0.70
Jul-08-09
Jul-11-09
Jul-15-09
Jul-13-09
Jul-17-09
Jul-19-09
Figure 4-2. Time series of the observed horizontal polarization emissivities
for all sensors and the in situ soil moisture for a wheat field.
78
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
In situ volum etric
E m issiv ity , i (H - p o l)
0.90
0-2.5 c m
0.95
1.00
Corn
0.98
Field Site: 25
VW C=7.18 kg-m"2
S'---------0.25
0.92
S en sor
0.90
»
a
_
#
A
'- H r —
i
A e(jU| 9 _jUMij
6.92 %
PALS 1.41 GHz
PALS 2.69 GHz
5.36 %
PSR/C 7.35 GHz
5.93 %
2.19 %
TMI 10 GHz
1.98 %
TMI 19 GHz
1.75 %
SSM/I 19 GHz
0.36 %
SSM/I 37 GHz
Soil moisture (0 -2 .5 cm)
0.15
0.10
0.84
0.05
0.82
0.00
0.80
Jul-08-99
Jul-09-99
Jul-10-99
Jul-11-99
Jul-12-99
Jul-13-99
Jul-14-99
Figure 4-3. Time series of the observed horizontal polarization emissivities
for all sensors and the in situ soil moisture for a com field.
79
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
so il m oisture
E m iss iv ity , c (H - pol)
0.20
In situ volum etric
0.94
0-2.5 c m
0.96
0.40
1.00
Field Site: 27
Alfalfa
VW C=1.0 kg-m'
0.35
0.25
0.20
0.85
0.15
0.80
0.10
0.75
R 7
0.05
0.00
0.70
Jul-08-99
Jul-09-99
Jul-11-99
Jul-12-99
Jul-13-99
Jul-14-99
Figure 4-4. Time series of the observed horizontal polarization emissivities
for all sensors and the in situ soil moisture for an alfalfa field.
80
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
s o il m oisture
E m issiv ity , e (H - p ol)
0.90
in situ volum etric
0.30
0-2.5 c m
0.95
LAND SURFACE PARAM ETERS
sand % , *clay % , *roughness param eters
bulk density (In situ data)
Sim ulated Brightness T em perature Ts mod
Sim ulated Emissivity £p mod
Observed Brightness T em perature TB ob»
Observed Emissivity £p
*Source: Jackson, T.J. et al. 1999
Figure 4-5. Approach for the retrieval of emissivities from each sensor.
81
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Observed vs. Estimated Emissivities
1.00
0.95
■
■
•
■
■
a
A
J^
S'
>
0.90
E
® 0.85
■-
Field sites: LW 21, LW 22, LW 23
July 8 th -J u ly 14th, 1999
PALS 1.41 GHz
PALS 2.69 GHz
PSRC 7.35 GHz
TMI 10.7 GHz
TMI 19.4 GHz
SSMI 19.4 GHz
SSMI 37.0 GHz
Linear (PALS 1.41 GHz)
Linear (PALS 2.69 GHz)
- Linear (PSRC 7.35 GHz)
- Linear (TMI 10.7 GHz)
- Linear (TMI 19.4 GHz)
Linear (SSMI 19.4 GHz)
Linear (SSMI 37.0 GHz)
A
''a A ■
0)
m
■o
3
E
w
PALS 1.41 GHz R = 0.92
0.80
PALS 2.69 GHz R2 = 0.88
PSR/C 7.35 GHz R2 = 0.85
TMI 10.7 GHz R2 = 0.83
TMI 19.4 GHz R2 = 0.67
0.75
SSM/I 19.4 GHz R2 = 0.60
SSM/I 37.0 GHz R2 = 0.21
0.70
0.70
0.75
0.80
0.85
O bserved em issivity,
0.90
0.95
1.00
eh
Figure 4-6. Observed and simulated emissivities in the horizontal
polarization for three field sites from 8 July through 14 July.
82
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Observed vs. Estimated Emissivities
1.00
Field sites: LW 21, LW 22, LW 23
July 8th - July 14th, 1999
a
■
•
■
a
A
PALS 1.41 GHz
PALS 2.69 GHz
PSRC 7.35 GHz
TMI 10.7 GHz
TMI 19.4 GHz
SSMI 19.4 GHz
A
SSMI 37.0 GHz
Linear (PALS 1.41 GHz)
Linear (PALS 2.69 GHz)
- - ■Linear (PSRC 7.35 GHz)
Linear (TMI 10.7 GHz)
— - - Linear (TMI 19.4 GHz)
Linear (SSMI 19.4 GHz)
Linear (SSMI 37.0 GHz)
0.95
&
I
TJ
0.90
3ra
3
E
PALS 1.41 GHz R = 0.89
i5)
PALS 2.69 GHz R2 = 0.82
0.85
PSR/C 7.35 GHz R2 = 0.80
-
TMI 10.7 GHz R2 = 0.71
TMI 19.4 GHz R2 = 0.52
SSM/I 19.4 GHz R2 = 0.45
SSM/I 37.0 GHz R2 = 0.12
0.80
0 .8 0
0.85
0.90
O b serv ed em issivity,
0.95
ev
Figure 4-7. Observed and simulated emissivities in the vertical
polarization for three field sites from 8 July through 14 July.
83
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
1.00
1000x1000
1000x1000
OUPUT
2-D GRIDS
INPUT
2-D GRIDS
10x10
10x10
*■
Figure 4-8. Approach used in the simulation model of brightness temperatures for the
retrieval of emissivities at each frequency.
84
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Soil moisture
0.40
0.96
0.96
0.96
0.25
0.85
0.85
0.85
o.io
4-7a. 1000 x 1000
00
Emissivity H-pol
Emissivity H-pol
Emissivity H-pol
4-7e. S-band (2.69 GHz)
4-7f. C-band (7.35 GHz)
0.40
0.96
0.96
0.25
0.85
0.85
0.74
0.74
0 .10
4-7b. 100x 100
4-7g. X-band (10.7 GHz)
0.74
0.74
0.74
4-7d. L-band (1.41 GHz)
4-7h. TMI 19 GHz
0.96
0.40
0.96
0.25
0.85
0.85
0 .10
0.74
10.74
I
4-7j. SSM/I 37 GHz
4-7i. SSM/I 19 GHz
4-7c. 10 x 10
Figures 4-9a through 4-9j. Spatial variability of soil moisture and the simulated emissivity at the horizontal polarization
for all frequencies for Day 2.
C h a p t e r 5 - S o il M o is t u r e R e t r ie v a l
S o il M o is t u r e E x p e r im e n t s
in
5.1
in
S o u t h e r n C e n t r a l G e o r g ia
fr o m the
2003 (S m e x 03)
from
A m s r -e
INTRODUCTION
The influence of soil moisture on land-atmosphere interactions and on weather
and climate systems is incontrovertible. Soil moisture plays a crucial role in the interplay
between hydrological and atmospheric processes by controlling the partition of available
energy at the surface into sensible and latent heat flux, by determining the partition of
rainfall into runoff and infiltration, and by the exchange of moisture fluxes in the form of
evapotranspiration (Shukla and Mintz, 1982; Delworth and Manabe, 1989; Brubaker and
Entekhabi, 1996). Hence, soil moisture information is important for continuing research
and for the application in geophysical processes that affect our environment (Leese et al.,
2001; Entekhabi et al., 1999). These processes, which include hydrology, meteorology,
agriculture, and climate, require accurate estimates of soil moisture for modeling and
monitoring. There is a need for soil moisture information at the global scale to better
understand and predict our climate system (WCRP-89, 1995) and to help improve models
to better relate soil processes to meet the needs of our environment.
86
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Soil moisture has a high temporal and spatial variability due to a small-scale
component which is characterized by soil properties, vegetation, and topography, and a
large-scale component characterized by solar radiation, atmospheric forcing, and general
environmental conditions (Entin et al., 2000). The dearth in soil moisture data stems
from the difficulty in monitoring this variability using
conventional ground
measurements, apart from the high costs involved in monitoring soil moisture at larger
scales. Microwave satellite sensors offer a promising avenue of inferring regional and
global soil moisture (Jackson and O’Neill, 1987; 1990; Jackson and Schmugge, 1991;
Laymon et al., 2001). Truly quantitative estimates of soil moisture are possible using
physically based radiative transfer models (Njoku and Kong, 1977; Wilheit, 1975;
Choudhury et al., 1979). Hydrological models that integrate passive microwave
observations and in situ observations can be then constructively harnessed in monitoring
soil moisture globally (Njoku et al, 2003).
Experiments have been conducted in the past to augment soil moisture datasets to
validate observations from current and future passive microwave space bomes for soil
moisture sensing (Njoku et al., 2002; Bolten et al., 2003; Le Vine et al., 2001; Guha et
al., 2003; Jackson et al., 1999; Jackson et al., 2001; Jackson et al., 2002a; Jackson et al.,
2002b).
These studies evaluated the capabilities of available airborne and space borne
instruments during the Southern Great Plain Experiment (SGP99). More recently, the
Soil Moisture Experiments in 2002 (SMEX02) have been conducted over areas with
more challenging vegetation conditions to monitor soil moisture using aircraft
observations and satellite observations particularly from the Polarimetric Scanning
Radiometer at C- and X-bands (PSR-C/X), Passive/Active L- and S- band sensor (PALS)
87
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
and the Advanced Microwave Scanning Radiometer (AMSR) (Ujjwal et al., 2005;
Bindlish et al., 2005; McCabe et al., 2004; McCabe et al., 2005). The experiments have
been conducted to support NASA’s Global Water and Energy Cycle Program and future
space borne missions such as the Soil Moisture and Ocean Salinity (SMOS) mission.
The Soil Moisture Experiments in 2003 (SMEX03) were conducted over areas with
broader ranges of vegetation cover types, which included grasslands and agricultural crop
surfaces in the states of Oklahoma, Georgia and Alabama in the United States to
rainforests in the Cerrado region, Brazil (http://hydrolab.arsusda.gov/smex03/) to provide
datasets to validate observations from the Advanced Microwave Scanning Radiometer
(AMSR-E).
The AMSR-E instrument is presently a newly available remote sensing
source for obtaining soil moisture information globally (Njoku et al., 2003).
The goal of this study is to determine how effectively we can utilize observations
from the AMSR-E instrument to infer soil moisture by the interpretation of data from the
Soil Moisture Experiments in 2003 (SMEX03). The in situ datasets utilized in this study
are from field measurements collected in southern central Georgia from 17 June to 21
July, 2003.
This paper is structured as follows: Section 5.2 describes the experiment
plan, study area, and the AMSR-E instrument. Section 5.3.1 presents an evaluation of the
AMSR-E soil moisture product with the in situ datasets from SMEX03. Section 5.3.2
presents the results from the soil moisture retrieval using a zero-order radiative transfer
model and brightness temperature observations from the AMSR-E instrument at 6.9 GHz
(C-band) and 10.7 GHz (X-band). Section 5.3.3 describes the spatial variability of the
AMSR-E brightness temperatures at X-band and the predicted soil moisture during
SMEX03. Section 5.4 presents the conclusions of this study.
88
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.2
DATA AND METHODS
5.2.1
SMEX03 Experiment Plan
The Soil Moisture Experiments (SMEX03) were conducted to provide datasets for
the validation of satellite and aircraft instruments for soil moisture estimation. The in situ
data used in this study was collected during the SMEX03 experiment in south central
Georgia near Tifton from 17 June to 21 July, 2003. The experiment consisted of
vegetation sampling and soil moisture and temperature measurements in various land
cover type fields. There were a total of fifty-two field sites for the data collection of field
parameters. The sampling area was within a minimum of 100 m from any field boundary
and consisted of three sample points within each field; within the plant row, lA of the way
between the rows, and ‘A of the way between the rows.
Vegetation sampling was conducted at twelve field sites for the entire study
period.
The primary objective of the sampling was to estimate green and dry biomass.
Vegetation samples were collected on 18 and 25 June and 1 and 21 July.
Other
vegetation measurements taken at each site included plant height, plant density,
percentage ground cover, phenology, surface roughness, and leaf area index (Bosch et al.,
2003). Eddy-flux evapotranspiration measurements were taken at 3 sites representative
of the three dominant land cover types: pasture, peanuts, and cotton.
Samples of soil moisture and temperature were collected at forty-nine field sites
from 23 June to 2 July. These are referred to as the regional sampling sites. The eastcentral part of the Little River Experimental Watershed (LREW) included a larger
number of sites which coincided with available network instrumentation. The soil
samples for the measurements of soil moisture were collected every day between 1030
89
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
and 1400 CDT to coincide with aircraft and satellite overpasses.
The samples were
collected at three depth intervals: 0-1 cm, 0-3 cm, and 3-6 cm and gravimetric
measurements were made from these samples at each depth interval. The bulk density
was estimated using fixed volume samples from the 0-3 cm and 3-6 cm depth intervals
(Bosch et al. 2003). The gravimetric soil moisture values and the bulk densities were
then used to estimate the volumetric soil moisture at each depth interval. For the soil
samples from the 0-1 cm interval, a value of 1.5 gm cm'3 was used to convert to
volumetric soil moisture.
In addition to the soil sampling, three theta probe measurements of soil moisture
within the row, 14 row, and 14 row were taken at each site. The soil sampling coincided
with the theta probe measurements at the % row location. A composite of volumetric soil
moisture from theta probe measurements was used to compare with the estimated
volumetric soil moisture from the 0-3 cm and 3-6 cm intervals.
Soil samples to
determine soil texture at 0-6 cm were also collected at each field site. The samples of soil
contained a high percentage of sand and a low percentage of clay. The soil samples
exhibited typical regional soil textures which are predominantly loamy sands and sandy
loams. Measurements of sub-surface and surface soil temperature at 1 cm, 5 cm, and 10
cm were collected at the Vi row position.
The sub-surface soil temperatures were taken
using digital thermometers and the surface soil temperatures using handheld infrared
thermometers (IRTs).
Climatic data were also collected in a network of 35 tipping bucket precipitation
gauges, four meteorological stations operated by the University of Georgia, and one
station
from
the
NRCS
Soil
Climate
Analysis
Network
90
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
(SCAN)
(http://www.wcc.nrcs.usda.gov/scan/).
The climatic data include measurements of air
temperature, barometric pressure, wind speed, precipitation, relative humidity, solar
radiation, soil temperature, and soil moisture.
The study area lies within the coastal plain region of the southeastern United States
and covers an area of 3,750 km . Geographically, it extends from latitude 31.20° N to
31.82° N and from longitude 83.43° W to 83.94°W. The area includes the 334 km2 Little
River Experimental Watershed (LREW) located in the headwaters of the Suwannee River
basin (Sheridan, 1997). The climate of the area is classified as humid with an average
rainfall of approximately 1,200 mm per year. The topography is relatively flat (slopes
vary from 1 to 5%). The study area consists of different vegetation cover types which
include forest (36%), pasture (18%), agricultural crop surfaces (40%), wetlands and
residential areas (6%).
Land use is dominated by agriculture, in which peanuts and
cotton are the main crop types. Other minor crop surfaces of the area include com and
cucumber. During the study period, the sampling fields exhibited various growth stages
that ranged from low initial biomass content to high biomass. The peanut flowers grew
from an average height of 9 cm to 51 cm allowing a range of vegetation water content
2
2
VWC from 0.1 kg-m' to 3.8 kg-m' . The cotton flowers grew from an average height of
28 cm to 110 cm with a VWC range from 0.11 kg-m'2 to 3 kg-m'2. The pasture fields
were cut and harvested at the beginning of the experiment and grew from an average
height of 24 cm to 45 cm with a VWC range from 0.38 kg-m'2 to 2.12 kg-m'2. These
conditions provided significant changes in biomass throughout the study period.
91
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.2.2
Satellite D ata
5.2.2.1 AMSR-E
Data from the Advanced Microwave Scanning Radiometer (AMSR-E) were used
in this study to evaluate the AMSR-E soil moisture product and to infer soil moisture
from AMSR-E brightness temperatures at the 6.9 GHz and the 10.7 GHz channels. The
AMSR-E instrument on the Earth Observation System (EOS) Aqua satellite was
launched in May 2002 by the National Aeronautics and Space Administration (NASA)
(Njoku et al., 2003).
The AMSR-E instrument is a modified version of the AMSR
instrument on the Japanese Advanced Earth Observing Satellite-II (ADEOS-II)
previously launched in December 2002 by the National Space Development Agency of
Japan (NASDA).
The AMSR-E instrument is a twelve channel linearly polarized system which has
equator crossings at 0130 and 1330 LST.
The radiometer operates at frequencies of
6.925 GHz, 10.65 GHz, 18.7 GHz, 23.8 GHz, 36.5 GHz, and 89.0 GHz, in vertical and
horizontal polarizations. The AMSR-E instrument operates in a sun-synchronous orbit at
an altitude of 705 km providing a swath width of 1,445 km. The footprint size produced
at a conical scanning mode and incidence angle of 55° is 76x44 km at 6.925 GHz,
49x28 km at 10.65 GHz, 28x16 at 18.7 GHz, 31x18 at 23.8 GHz, 14x8 at 36.5 GHz,
and 6 x 4 at 89.0 GHz. A thorough description of the AMSR-E instrument and the soil
moisture retrieval algorithm is found in Njoku et al. (2003).
The AMSR data used in this study is the Level-3 (L3) data product. The data was
reprocessed onto a global, Equal-Area Scalable Earth Grid (EASE-Grid) projection of a
25-km cell resolution by the National Snow and Ice Data Center (NSIDC) in Boulder,
92
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Colorado. The AMSR-E brightness temperatures at 6.9 GHz (C-band) and 10.7 GHz (Xband) and the soil moisture product at C-band resolution were extracted for this study.
During the SMEX03 experiment, the AMSR-E instrument provided data on June 23, 25,
27, 29, and 30 and July 2 in the ascending overpass.
93
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.3
RESULTS
5.3.1
AMSR-E Soil Moisture and In Situ Soil Moisture Validity
The goal of SMEX03 was to provide reliable estimates of soil moisture within the time
frame o f the ascending overpass of the AMSR-E instrument (1330 LST). The sampling
was conducted to include measurements of soil moisture and temperature within a single
satellite footprint and multiple measurements within EASE-Grid cells of 25 km (Bosch et
al., 2003). The study area comprises a total of six 25-km EASE-Grid cell boxes (Brodzik
and Knowles, 2002) covering forty-nine regional sampling sites with a nominal spacing
of 8-10 km. Figure 5-1 shows the location of the fifty-two sites, distinguishing between
the regional sampling sites (soil moisture and temperature), the vegetation sampling sites
and the vegetation-eddy flux sampling sites within each EASE-Grid cell box. Table 5-1
describes the distribution of the regional sampling sites and land use for each EASE-Grid
cell box.
This section discusses the comparison and validation of the AMSR-E soil
moisture product with the in situ volumetric soil moisture mv from the regional sampling
sites at the 25-km resolution for all common days between the ascending overpass of
AMSR-E and mv measurements from 23 June through 2 July.
The EASE-Grid point
vertices for the six cell boxes were provided by Brodzik and Knowles (2002) and were
used to extract the AMSR-E brightness temperatures Tb and the soil moisture product
from the AMSR-E L3 global datasets from 23 June through 2 July. The values of Tb and
soil moisture at any given four vertices were used to obtain a mean value of each
parameter to apply to this study. Figures 5-2 and 5-3 show the relationship between the
AMSR-E mean Tb estimated from the vertical and horizontal polarization Tg at C-band
94
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
and X-band and the in situ mv at 0-1 cm. The mean TB at X-band are slightly better
correlated with the in situ mv at 0-1 cm as seen in Figure 5-3. A correlation of 0.42 was
obtained with the TB at C-band and a R2 of 0.44 with the TB at X-band. The mean TB at Cand X-band are used in the retrieval of soil moisture which is discussed in Section 5.3.2.
A comparison of each value of mv from the regional sites and the AMSR-E mv
was conducted to investigate the relationship between the variability of soil moisture
within each EASE-Grid cell box. The comparison of the variance of mv was performed
between the mv values at 0-1 cm and 0-3 cm and the AMSR-E mv from the six EASEGrid cell boxes. Table 5-2 presents the statistics computed from the comparison at each
depth interval. The AMSR-E mv is found to be poorly correlated in both cases with the in
2
2
situ mv at 0-1 and 0-3 cm. A R value of 0.12 was obtained for the mv at 0-1 cm and a R
of 0.14 for the mv at 0-3 cm. The root mean square errors (RMSE) are also high in both
cases (18-21%) as seen in Table 5-2. The in situ mv has a higher deviation at both depth
intervals than the AMSR-E mv. In fact, the AMSR-E mv has almost no deviation. The
low correlation observed between the in situ mv and AMSR-E mv can be related to the
decrease in variability of the surface characteristics at the AMSR-E C-band scale (Njoku
et al., 2003). The AMSR-E C-band soil moisture values are lower than the in situ values
because both the soil moisture and vegetation water content products are averaged over a
60 km scale.
Therefore, the soil moisture values are a direct result of the spatial
smoothing of the AMSR-E C-band footprint (Njoku et al., 2003). For this reason and the
smaller deviation observed in the in situ mv at 0-3 cm, there is a slightly better correlation
between the mv from the deeper layer of soil and the AMSR mv.
95
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
An up-scaling method was employed to the regional sampling sites to keep the
same spatial domain as have the EASE-Grid cell boxes. It has been demonstrated that
key statistical characteristics are preserved from up-scaling (McCabe et al., 2005); with
this in mind, up-scaling was performed to compare the mean mv and the AMSR-E mv to
examine how much the AMSR mv was representative of the regional sampling sites. The
up-scaling method was also used to obtain the spatial mean mv and other in situ
parameters (i.e. soil temperature and vegetation water content) needed as inputs for a
radiative transfer model for the retrieval of soil moisture discussed in Section 5.3.2. The
spatial mean was obtained for the EASE-Grid cell boxes by taking the mean mv values
from the regional sites that fell within each cell box. The result was six values of in situ
mean mv. Figures 5-4 and 5-5 compare the mv at 0-1 cm and 0-3 cm values with the
AMSR-E mv. The correlation between the mean mv at both depth intervals (0-1 cm and
0-3 cm) and the AMSR-E mv improved from the previous relationship. The correlation
R2 obtained resulted in a value of 0.60 for the mv of the sub-surface and 0.65 for the 0-3
cm layer (Table 5-2).
The correlation between the AMSR-E mv and the mean mv at 0-3
cm is also found to be higher than at the sub-surface (0-1 cm). In the next section, the
observed and simulated AMSR-E Tg at C- and X- band are used in a physically based
model to retrieve soil moisture.
96
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5 . 3.2
Predicted Soil Moisture with AMSR-E Data
A zero-order radiative transfer model was utilized to simulate the Tb that would
be measured by the AMSR-E instrument at C-band and X-band. The contributions to the
microwave signal can be separated into three main components: a surface emission, a
vegetation emission, and an atmospheric emission. The effects from the surface and
vegetation are each parameterized in the radiative transfer model discussed in Chapter 2.
The effects from the atmosphere on the measured signal by a radiometer increase with
frequency. These effects are significant at frequencies higher than 18 GHz, but they can
be considered negligible at C- and X- bands.
The radiative transfer equation in a general form is given by,
TBP= f{0 }
(5-1)
where 0 are the input parameters of the radiative transfer (0=6, Ts, To b, VWC, pb, freq,
sand%, clay%, Q, h). Table 5-3 gives a summary and description of the input parameters
of the radiative transfer model used in this study. The radiative transfer model was
calibrated using 3 stations for the entire duration of the experiment.
The brightness
temperatures at C- and X-band were simulated using the in situ soil temperature at 1 cm
and the soil moisture at 0-1 cm. The bulk density, sand and clay mass fractions for each
of the regional sampling sites are from in situ data. The vegetation parameter b and the
surface roughness parameters were calibrated. A value of 0.04 was obtained for the
vegetation parameter b and a value of 1.21 for the roughness height h for C-band. The
value of b changed to 0.12 and the h to 0.81 for X-band. At both bands, a value of zero
was used for Q. The vegetation water content VWC was provided from in situ data.
97
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
As a first check before the soil moisture retrieval, a comparison was made
between the simulated Tb and the AMSR TB at C- and X-band. Overall, the AMSR Tb at
C-band had a higher range of Tg values than the simulated Tb. The correlation between
the simulated Tb at C-band in the vertical polarization resulted in a R2 of 0.87 and for the
horizontal polarization in a R2 of 0.90. The correlation at X-band, resulted in R2values of
0.95 and 0.97 in the vertical and horizontal polarizations, respectively. As expected, the
simulated horizontally polarized TB were better correlated with the AMSR horizontally
polarized 7g. The lower correlation obtained with the TB at C-band can be related to
residual effects from Radio Frequency Interference (RFI) in the AMSR-E Level-2 data
(Njoku et al., 2003; Li et al., 2004). The RFI effects were more notable on the Tb at Cband than at X-band. The AMSR-E TB at C-band are used in this study for the sake of
completeness.
The simulation of Tb at C- and X- bands allowed for adjusting the
radiative transfer model before being used in the inverse mode for the retrieval of soil
moisture.
The retrieval of soil moisture employed in this study uses an iterative least-square
minimization method. The method chooses the value of soil moisture that minimizes the
weighted-sum of squared differences between the observed and simulated averaged Tb
normalized with the surface temperature Ts to account for the surface temperature
contribution. The iteration uses soil moisture values ranging from 0% to 50% in 1%
increments. The retrieval of soil moisture was carried out using the mean in situ data at
the 0-1 cm and 0-3 cm soil layers for the six EASE-Grid cell boxes presented in Section
5.2. The retrieval was performed for C- and X- bands separately.
98
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Table 5-4 presents the statistics computed for the predicted mv and the in situ mv at
0-1 cm and 0-3 cm at C- and X- bands. The linear relationship between the predicted
volumetric soil moisture mv and the sub-surface in situ mv at 0-1 cm at C-band is shown
in Figure 5-6. The majority of the predicted mv overestimated the in situ mv values,
particularly under low moisture content (mv <0.12). The correlation resulted in a R2 of
0.57 with a root mean square error RMSE of 0.067. The deviation of observed in situ mv
is the same as the predicted mv, but the in situ mv have much smaller values. For the 0-3
cm soil layer, a better correlation was found between the predicted mv and the in situ mv
(Fig. 5-7). The correlation resulted in a R2 of 0.50 with a RMSE of 0.04. In this case, the
predicted mv overestimated the in situ mv less than half of the time compared to the sub­
surface layer under low moisture content. A possible reason for the better correlation is
that during most of the experiment (dry conditions) the mv was generally lower in the
deeper layers (i.e. 0-3 cm has lower values than mv at 0-1 cm). A similar deviation is
observed in the predicted mv values and the in situ mv at 0-3 cm.
The soil moisture
retrieval was carried out at X-band following the same approach. The predicted mv was
compared to the in situ mv at 0-1 cm and 0-3 cm. Figure 5-8 shows the linear relationship
between the predicted mv and the in situ mv at 0-1 cm. The predicted mv underestimates
the in situ mv under low moisture content.
0.027 was obtained.
A correlation value of 0.63 with a RMSE of
At X-band, the highest correlation was achieved as seen in Figure
5-9. This correlation resulted in a R2 of 0.72 with a RMSE of 0.022 for the 0-3 cm layer
of soil at X-band. In general, the predicted mv was found to be better correlated with the
in situ mv at the 0-3 cm layer.
99
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.3.3
Spatial Variability o f Soil Moisture During SMEX03
The spatial variability of the TB and the predicted mv during the study period is
depicted in Figures 5-10a through 5-11/.
The spatial distribution of the AMSR-E
horizontally polarized Tg at X-band is depicted in Figures 5-10a through 5-10/ Figures
5-10g through 5-10/ show the spatial distribution of the mean predicted mv at 0-1 cm.
The spatial distribution of the AMSR-E horizontally polarized Tg at X-band are shown
again in Figures 5-1 la through 5-1 1 /for comparison with the mean predicted mv at 0-3
cm (Figures 5-1 lg through 5-11/). Both the AMSR-E Tg and the mean predicted mv
changed over the course of the experiment with the greatest change occurring on 30 June
corresponding to a rainfall event and on 2 July after a rainfall event on 1 July.
The in situ mv across the regional sampling sites shows high variability on a daily
basis. This variability was preserved in the mean m v of the EASE-Grid cell boxes as seen
in the resultant maps of the predicted
mv
mv
The greatest variability is seen in the predicted
at 0-1 cm. The highest spatial variability in soil moisture is seen during dry conditions
(23 June through 28 June) (Warrick and Nielsen, 1980). A rain event on 30 June and 2
July affected the observed soil moisture and the AMSR-E Tg in the area. The greatest
increase in
mv
was seen in the west-central part and north-east part of the watershed and
the least change observed was in the east-central part of the watershed on 30 June
(Figures 5-10& and 5-11/:). The rainfall measured at the stations on 30 June was about
114 mm in the west-central part and about 77 mm in the north-east part.
The mean
predicted mv, particularly the soil layer at 0-3 cm, reflects the spatial variability of the
AMSR-E
Tb .
On July 2, the greatest increase in
mv
was seen in the east-central part and
south-west part of the watershed and the least change in the north-east part of the
100
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
watershed. The rainfall measured was the greatest at the end of the soil sampling which
related to a rainfall event on 1 July.
The 1 July rainfall was about 318 mm in the east-
central part of the watershed and about 293 mm in the south-west part. In general, the
predicted mv at 0-1 cm and 0-3 cm soil layers reflect the spatial and temporal variability
of the in situ mv and the AMSR-E observed 7g.
101
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
5.4
CONCLUSIONS AND DISCUSSION
The growing demand for the accurate estimation of soil moisture has been notable
in the past decade.
The application and research into soil moisture to the understanding
of the climate system, hydrological modeling, meteorological and flood forecasting and
many other disciplines is incontrovertible. In fact, numerous studies carried out for the
past twenty years have demonstrated that passive microwave instruments hold the most
promise for inferring regional and global soil moisture. The AMSR-E instrument was
developed in an attempt to collect soil moisture throughout the globe. This instrument
holds promise for the acquisition of soil moisture information on a long-term basis from
space. In particular, the AMSR-E at C-band has shown significant improvements over
the capabilities of previous sensors.
In this study, we utilized in situ datasets from SMEX03 to validate observations
from the ascending overpass of the AMSR-E instrument. The validation and comparison
results of the AMSR-E mean soil moisture product derived from C-band demonstrate that
the AMSR-E retrieval algorithm can infer soil moisture with certain limitations.
The
main contributors to errors in the retrieval of soil moisture are the spatial smoothing over
the footprint and the residual RFI effects. Comparisons of the AMSR-E soil moisture to
direct point measurements resulted in high errors of up to 20%. The results suggest that
single point measurements of mv may not provide a quantitative representation of the
AMSR-E footprint but mean mv data from point measurements over the footprint do.
The algorithm for the soil moisture product still needs validation and verification in other
areas. Improvements are expected from the soil moisture retrieval algorithm at X-band in
the near future.
102
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
The results obtained from using a physically based model (zero-order radiative
transfer model) and AMSR-E brightness temperature observations demonstrate that soil
moisture was retrieved with acceptable accuracy.
The results also demonstrated the
superiority of X-band over C-band in the retrieval of soil moisture.
The retrieval of soil
moisture was achieved with a 2.7 % RMSE at X-band and a 6.7 % RMSE at C-band. The
predicted soil moisture at C-band was better correlated with the sub-surface soil layer (01 cm).
On the other hand, the predicted soil moisture at X-band was found to be better
correlated with the 0-3 cm layer. In general, the comparisons and statistics of the results
show that the data is noticeably corrupted at C-band.
Overall, the AMSR-E soil
moisture product and the predicted soil moisture using AMSR-E brightness temperatures
were higher correlated within the deeper layer studied (0-3 cm).
The spatial distribution of the predicted soil moisture at 0-1 cm and 0-3 cm at Xband show similar spatial trends with the observed AMSR-E brightness temperatures at
X-band. The 0-1 cm soil layer was found to have the highest variability over the course
of the experiment, but the 0-3 cm soil layer had better agreement with the observed
AMSR-E brightness temperatures.
Anticipated changes in soil moisture were expected
after the rainfall events on 29 June and 1 July. These changes were observed in the spatial
maps of the predicted soil moisture.
Qualitative agreement amongst the AMSR-E
brightness temperatures, the predicted soil moisture and precipitation was observed
during the experiment. Overall, the AMSR-E demonstrated its ability to successfully
infer soil moisture during the SMEX03 experiment.
103
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
EASE-Grid Cell
Regional Sampling Sites
Row crop
Pasture
Forest
1 (North-west)
GA01-GA09
4
1
4
2 (North-east)
GA10-GA16
3
2
2
3 (West-central)
GA17-GA24
5
1
2
4 (East-central)
GA25-GA37
8
2
3
5 (South-west)
GA39-GA44
3
1
2
6 (South-east)
GA45-GA49
3
1
2
Table 5-1. Regional sampling sites and land cover within each EASE-Grid cell box.
104
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Statistical
Variance of mv
Mean of mv
parameter mv 0-1 cm mv 0-3 cm mv 0-1 cm mv 0-3 cm
R2
0.12
0.14
0.60
0.65
RMSE
0.21
0.18
0.04
0.05
C*insitu mv
0.14
0.09
0.06
0.04
^AMSR-Emv
0. 01
0.01
Table 5-2. Correlation and standard deviation for the mv from all the regional sampling
sites and AMSR-E mv and the mean mv from all the regional sampling sites and
AMSR-E mv.
105
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
SENSOR PARAMETERS
Frequency
Polarization
Viewing angle 0 (deg)
SURFACE PARAMETERS
Soil moisture, 6 (0-1 cm)
Soil temperature, Ts (1 cm)
Sand %
Clay %
Bulk density, pb (g cm'3)
Roughness coefficient, h (cm)
Polarization mixing factor, Q
VEGETATION PARAMETES
Vegetation parameter, b
Vegetation water content, VWC (kg- m'2)
Canopy temperature, Tc
Single scattering albedo, co
6.9 GHz
10.7 GHz
H, V
55°
in situ
in situ
in situ
in situ
in situ
calibrated 1.2
calibrated 0.8
calibrated 0.04
calibrated 0.12
in situ
in situ
0
Table 5-3. Input parameters of the radiative transfer model.
106
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
mv 0-1 cm
mv 0-3 cm
Soil moisture layer
R
6.9 GHz (C-band)
0.57
RMSE Opred Chn situ
0.067
0.06
R
RMSE
0.60
0.04
Opred ®In situ
0.05
0.06
10.7 GHz (X-band)
0.63
0.027
0.05
0.04
0.72
0.022
0.04
Table 5-4. Statistics for the predicted soil moisture and in situ soil moisture at 0-1 cm and
0-3 cm at 6.9 GHz and 10.7 GHz channels.
107
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
3530000
E A S E -G rid cell b o x 1
E A S E -G rid cell b ox 2
3510000
A
E A S E -G rid cell b o x 3
E A S E -G rid ce
box 4
3490000
E A S E -G rid cell b ox 5
E A S E -G rid cell b ox 6
3470000 r
3450000
215000
245000
230000
260000
• Regional sampling sites
♦ Vegetation &eddy flux sampling sites
■ EASE-Grid point vertices
a Vegetation sampling sites
Figure 5-1. Regional sampling sites, vegetation and vegetation-eddy flux
sampling sites within each EASE-Grid cell box.
108
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
275000
300
y = -39.98x + 289.12
R2 = 0.42
295
X
290
(!)
o>
<£>
H
c 285
<0
O
XX X
CD
E
O'
CO
<
280
275
270
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
In s itu mv 0-1 cm
Figure 5-2. AMSR mean TB at C-band and in situ mv at 0- cm.
109
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Figure 5-3. AMSR mean Tb at X-band and in situ mv at 0-1 cm.
110
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0.35
0.30
y= 0.18x + 0.14
R2 = 0.60
0.25
i 0.20
m
O'
w
f
0.15
0.10
0.05
0.00
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
In s itu m v 0-1 cm
Figure 5-4. Mean mv 0-1 cm and AMSR-E mv.
111
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0.35 r
0.30 -
0 .1 0
-
0.05
-
o.oo r ...................................................................,- x- x.^—...
0.00
0.05
0.10
0.15
0.20
0.25
0.30
In s itu m v 0-3 cm
Figure 5-5. Mean mv 0-3 cm and AMSR-E mv.
112
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0 .3 5
0.30
0.25
E
y = 0.73x + 0.06
o
©
R2 = 0.57
0.20
E
0>
■D
o
0.15
Q)
’B
Q.
0.10
0.05
0.00
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
In s itu m v 0-1 cm
Figure 5-6. Predicted and in situ soil moisture at 0-1 cm from C-band TB.
113
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0 .3 0
0.25
y = 0.96x + 0.05
.20
R2 = 0.50
0.15
0
♦♦
0.05
0.00
0.00
0.05
0.10
0.15
0.20
0.25
0.30
In situ mv 0-3 cm
Figure 5-7. Predicted and in situ soil moisture at 0-3 cm from C-band Tr.
114
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0 .3 5
0.30
0.05
.
0.00
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
In situ mv 0-1 cm
Figure 5-8. Predicted and in situ soil moisture at 0-1 cm from X-band Tb.
115
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
0 .3 0
0.25
y = 0.82x + 0.05
R2 = 0.72
♦
♦♦♦
0.05
0.00 — 1— 1— 1— 1— 1— '— '— '— 1— 1— 1— '— '— '— 1— 1— 1— 1—
0.00
0.05
0.10
0.15
'— 1— 1— 1— 1— 1— '— '— 1— ■— 1
0.20
0.25
0.30
In situ mv 0-3 cm
Figure 5-9. Predicted and in situ soil moisture 0-3 cm from X-band TB.
116
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
AMSR Brightness Temperatures, 7g H 10.7 GHz (X-band)
23 June, 2003
—83.90—63.80—
83.70—83.60—83.50
25 June, 2003
83.00 83.80 83.70 83.60- 83.50
27 June, 2003
-8
1- 83.50
29 June, 2003
30 June, 2003
—83.90—S3.80—83.70—83.60—83.50
- 83.90-S3.80- 83.70- 83.6Q-a3.50
2 July, 2003
- 83.90- 83.80- 83.7
T* (K)
285
278
I
270 a
70- 83.50- 83.50
3.80-83
50- 83.50 £ —
83
B3.50 /
-8
Predicted Volumetric Soil Moisture, 0 V 0-1 cm
- 83.90- 83.80- 83.70- 83.60- 83.50
6
- 83.90- 83.80- 83.70- 83.60- 83.50
- 83.90- 83.8Q- 83.70- 83.6Q- 83.50
.70- 83.60- 83.50
- 83.90- 83.8Q- 83.70- 83.6Q-83.50
- 83.90- 83.80- 83.70- 83.60- 83.50
0.25
I
1 0.18
0. 10
g
83.60- 83.50 1 -83
- 83.50 fc
83.90
Figures 5-10a through 5-10/. Spatial distribution of AMSR horizontal polarization brightness temperatures at 10.7 GHz and
mean predicted volumetric soil moisture 0-1 cm.
73
CD
■-5o
o
a.
c
o
CD
Q.
■o
CD
C
(f)/)
AMSR Brightness Temperatures, T r H 10.7 GHz (X-band)
23 June, 2003
25 June, 2003
—8 3 .9 Q —8 3 .8 0 —8 3 .7 0 —6 3 .6 Q —8 3 .5 0
-8 3 .9 0 -6 3 .8 0 -8 3 .7 0 -a 3 .6 Q -8 3 .5 Q
27 June, 2003
- 83,70- 63.60- 83.50
29 June, 2003
30 June, 2003
- 83.90- 83.80- 83.70- 83.60- 83.50
- 83.90- 63.80- 83.70- 63.60- 83.50
2 July, 2003
- 83.90- 83.80- 83.70- 83.60- 83.50
O
o
■O
ca
T„ (K)
285
c
3.
I 278
CD
CD
■-5o
o
I
270 a
-1
50
fc?
-S 3
Q.
C
ao
Predicted Volumetric Soil Moisture, 0vO-3 cm
■o
- 83.90- 63.80- 83,70- 83.60- 83.50
o
- 83.90- 83.80- 83.70- 83.60- 83.50
- 83.90- 83.80- 83.70- 83.6Q-83.5Q
- 83,90-e 3.80- 83.70-a 3.6Q-83.50
- 83.90- 83.8Q-83.70- 83.6Q-a3.50
CD
Q.
■o
CD
C
C/)/)
6
0.25
I
t 0.18
0.10 g - 83.90- 83.80- 83.70- 83.60- 53.50 h -B3.‘o-sA80- 83.
/
- 83.90- 83.8O-ai 'D-8^0
83:50j
8J.90 83.80 8.CTd 83.60 83.50
_g;
Figures 5-1 la through 5-11/. Spatial distribution of AMSR horizontal polarization brightness temperatures at 10.7 GHz and
mean predicted volumetric soil moisture 0-3 cm.
C h a p t e r 6 - Su m m a r y
6.1
and
F o l l o w -u p S t u d ie s
SUMMARY
The objective of the different studies covered in this dissertation was to evaluate
the capabilities and limitations of microwave passive sensors to monitor land surface
dynamics using a physically based radiative transfer model. The model can be applied to
any frequency and to any spatial-temporal domain but has certain limitations as
exemplified in the research of this dissertation. In the study carried out in Illinois and
Oklahoma, land surface temperature and surface soil moisture were retrieved using the
SSM/I sensor. The model by Hiltbrunner (1994) applied to the SSM/I at 19 GHz resulted
in land surface temperature retrievals with similar accuracies and errors to previous
studies that used SSM/I. A radiative transfer model was utilized to simulate brightness
temperatures at 19 GHz and was then used inversely with the SSM/I brightness
temperatures to retrieve soil moisture. Monthly averages of soil moisture over the length
of the study resulted in better correlations with the in situ soil moisture with errors within
4%. In the study over the US Southern Great Plains, the aspects of low and high spatial
resolution of emissivities at multiple frequencies were evaluated by comparing in situ
observations with observations from four different sensors.
The results from the
119
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
comparisons confirm that L-band (PALS sensor) is optimal for soil moisture retrieval.
The sensitivity to soil moisture observed at each frequency from the various microwave
passive sensors was clearly seen as a function of their wavelength and vegetation amount.
The results demonstrated that the microwave signal detected at higher frequencies is
considerably masked under highly dense vegetation. In the study carried out in southern
central Georgia, observations from the AMSR-E instrument were validated and compared
with data collected during the SMEX03 experiment. The results demonstrated that the
AMSR-E retrieval algorithm can infer soil moisture with certain limitations. The main
contributors to errors in the retrieval of soil moisture were the spatial smoothing over the
footprint and residual RFI effects, especially at C-band. The results obtained from using a
radiative transfer model and AMSR-E brightness temperatures demonstrated that soil
moisture may be retrieved with reasonable accuracy.
The results also demonstrated the
superiority of X-band over C-band in the retrieval of soil moisture, which can be related
to corrupted data at C-band caused by RFI. The retrieval of soil moisture using AMSR-E
Tb was achieved with a 2.7 % RMSE at X-band.
The results from each study in which aircraft or satellite passive sensors were
used demonstrate that surface soil moisture, surface temperature and vegetation can be
retrieved with acceptable accuracies. However, verification and validation of algorithms
for the retrieval of land surface dynamics and calibration and evaluation of future
missions still needs work. Furthermore, the limitations of a radiative transfer model fall
on the interaction between nonlinear retrieval physics and on the requirement of many
land and vegetation surface parameters as inputs in the model. The retrieved parameter
errors are partly due to data availability. There is a need to conduct more experiments to
120
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
collect datasets of land surface parameters to combine with hydrological models in order
to reduce the number of unknowns in the radiative transfer model and to minimize
retrieval uncertainties caused by local and regional variability such as topography,
vegetation and soil textures. In the next section, a number of follow-up studies and
applications is provided in consideration to the aforementioned issues.
121
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6.2.
FOLLOW-UP STUDIES AND APPLICATIONS
6.2.1
Surface Temperature
Further study in more geographic areas is needed to evaluate algorithms to
retrieve surface temperature and improve their accuracies. Improvements in the
estimation of surface temperature can yield more accurate retrievals of soil moisture and
vegetation optical depth (De Jeu et al., 2002). In this dissertation, SSM/I observations
were used in conjunction with an algorithm to retrieve surface temperatures in two
regions in the United States. The application of a surface temperature algorithm (i.e.
algorithm utilized in this dissertation) to new sensor platforms such as AMSR-E, SMOS
and HYDROS will significantly improve the accuracy at which surface temperature is
estimated.
Furthermore, global studies of surface temperature can be extended with SSM/I
observations and a surface temperature algorithm. The SSM/I can provide a 20-year span
of data which can be applied to long-term studies of surface temperature. The SSM/I as
well as any other radiometer can be used to retrieve land surface temperature
independently of weather conditions and without an a priori knowledge of the emissivity,
absorption, or scattering for known surface conditions (McFarland et al 1990).
The
limitations of land surface temperature retrievals are apparent when retrievals are
attempted over large areas with water bodies and areas with significant snow cover. An
algorithm that corrects for high water content is needed. The potential uses of the
retrieval of land surface temperatures include agricultural yield and crop condition
models, numerical weather models, atmospheric profile retrieval models,
and
hydrological models. Retrievals of land surface temperature may be used as potential
122
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
indicators of global warming by identifying temperature changes in sensitive regions of
the Earth where warming may affect the evolution of the climate.
6.2.2
Snow Cover
Studies concerning the detection and characterization of snow cover are
significant to understanding our climate and weather, primarily because of the energy
interplay that exists at the boundary between the land surface and the atmosphere (Kerr et
al., 1999) (i.e. regional energy and water balance). Snow cover studies can be applied to
detect global climate changes. One such way is by providing indicators of the onset of
snow melt and estimates of snow water equivalent, which is important for flood
prediction and for water resource applications such as reservoir management and other
agricultural activities. Awareness of seasonal snow cover fluctuations is also important
because it influences the water supply in many parts of the world (Kerr et al., 1999).
Passive microwave measurements can provide useful information on snow cover
(England, 1974; Chang et al., 1976) and freeze thaw cycles (Wismann et al., 1996)
because of the scattering properties of snow crystals, which reduce the upwelling
radiation measured by a radiometer (Kerr et al., 1999). It is well established that the
algorithm for snow cover (Chang et al., 1987) in which the 19 GHz and the 37 GHz
channels are utilized can provide a reasonable indication of snow depth (Chang et al.,
1976; Hallikainen and Jolma, 1986; Rott and Nagler 1994; Josberger et al., 1998). The
SSM/I sensor can be harnessed in this regard since it operates at these two channels and
can provide at least 20 years of data.
123
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6.2.3
Atmospheric Models for Higher Frequency Channels
The SSM/I sensor can provide a 20-year span of data which can be valuable for
long-term monitoring of land surface dynamics, especially soil moisture. However, the
frequencies at which the SSM/I sensor operate require atmospheric data for correction.
Algorithms that correct for these effects and are independent of water vapor and cloud
liquid content data are needed in order to utilize SSM/I data.
6.2.4
Vegetation Amount
Vegetation covers much of the earth and strongly influences the environment.
Passive microwave instruments are a tool for assessing characteristics and growth of
vegetation. They respond to vegetation properties differently than visible/near infrared
instruments, so studies that integrate observations from microwave passive sensors and
observations from these other instruments can provide better accuracy retrievals for
vegetation. Global coverage from passive microwave satellites and the AVHRR and
MODIS instruments provides the opportunity to derive estimates of global- and
continental-scale land cover. These instruments provide opportunities for studying and
monitoring vegetation conditions in various ecosystems.
Applications include
agricultural assessment, land cover mapping, producing image maps of large areas such
as countries or continents and tracking regional and continental snow cover. Vegetation
data from these remote sensing instruments can be used for many applications including
monitoring the increase of irrigated land, deforestation, the health of vegetation,
agricultural and timber productivity, global land cover change, and a wide variety of
other uses that meet the needs of our ever changing environment.
124
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
6.2.5
Vegetation Optical Depth
Further studies are needed to improve the algorithms that retrieve the vegetation
optical depth parameter and other related vegetation parameters that are used in the
radiative transfer model.
The physical formulation of the vegetation optical depth
illustrates the strong relationship the vegetation water content has with the vegetation
opacity. It is well known that the vegetation opacity is crop dependent, especially at low
frequencies (i.e. L-band), so values derived from empirical formulations affect the
accuracy of the vegetation optical depth. Furthermore, the use of visible/near infrared
instruments to estimate the vegetation water content when in situ data are not available
introduces errors into the retrieval of the vegetation optical depth because the signal that
these instruments detect is affected by atmospheric interference. One such way to make
these improvements is to develop an algorithm that accounts for the dependence of the
vegetation water content and vegetation optical depth.
125
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Allison, L., et al, 1974: Tropical cyclone rainfall as measured by the Nimbus-5
Electrically Scanning Microwave Radiometer. Bull. Amer. Met. Soc., 55, 10741089.
Asrar, G., E. T. Kanemasu and M. Yoshida, 1985: Estimates of leaf area index from
spectral reflectance of wheat under different cultural practices and solar angle.
Remote Sens. Environ., 17, 1 —11.
Baret, F., and G. Guyot, 1991: Potentials and Limits of vegetation indices for LAI and
APAR assessment. Remote Sens, o f Environ., 35, 161 - 173.
Basara, J.B., R. Elliot, K. Fisher, G. McManus, and K. Flumes, 2001: Estimation of soil
moisture for CYs 98 and 99 from the Oklahoma Mesonet Version 3.0. Oklahoma
Climatological Survey, Norman, Oklahoma.
Basist, A., C. Williams, N. Grody, T. R. Ross, S. Shen, A. T. C. Chang, R. Ferraro, and
M. J. Menne, 2001: Using the Special Sensor Microwave Imager to Monitor
Surface Wetness. J. Hydrometeor., 2, 291 - 308.
Benoit, A., 1968: Signal Attenuation Due to Neutral Oxygen and Water Vapor, Rain and
Clouds, Microwave J., 11, 73 - 80.
Bindlish, R., T. J. Jackson, A. J. Gasiewski, M. Klein, and E. G. Njoku, 2005: Soil
moisture mapping and AMSR-E validation using the PSR in SMEX02. Remote
Sens. Environ. In Press.
Bolten, J. D., 2003: Soil Moisture Retrieval Using the Passive/Active L-and S- Band
Radar/ Radiometer." IEEE Trans. Geosci. Remote Sens. 41(12): 2792-2801.
Bosch D. D., T. J. Jackson, V. Lakshmi, and J. Jacobs, 2004: Large Scale Measurements
of Soil Moisture for Validation of Remotely Sensed Data: Soil Moisture
Experiments of 2003. Submitted.
Brodzik, M. J., & Knowles, K., 2002: EASE-Grid a versatile set of equal-area projections
and grids. In M. Goodchild and A. J. Kimerling (Eds.), Discrete Global Grids.
Santa Barbara, CA, USA: National Center for Geographic Information &
Analysis.
Brubaker, K. L., and D. Entekhabi, 1996: Analysis of feedback mechanisms in landatmosphere interaction. Water Resour. Res., 32, 1343 - 1357.
Chang, A. T. C, P. Gloersen, T. Schmugge, T. T. Wilheit, and H. J. Zwally, 1976:
Microwave emission from snow and glacier ice. J. Glaciol. 16(74) 23-39.
126
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Choudhury, B. J., T. J. Schmugge, A. T. Chang, and R. W. Newton, 1979: Effect of
surface roughness on the microwave emission from soils. J. Geophys. Res., 84,
5699-5706.
Choudhury, B. J., C. J. Tucker, et al., 1987: Monitoring Vegetation Using Nimbus-7
Scanning Multichannel Microwave Radiometers Data. Int. J. Remote Sens. 8(3):
533-538.
Choudhury, B. J. and R. E. Golus, 1988: Estimating soil wetness using satellite data. Int.
J. Remote Sens., 9:1251-1257.
De Jeu, R. A. M. and M. Owe, 2002: Estimating spatially averaged surface soil
temperature for microwave surface soil moisture derivation. Remote Sens.
Environ. Submitted.
Delworth, T., and S. Manabe, 1989: The influence of soil wetness on near-surface
atmospheric variability. J. Climate, 2, 1447 - 1462.
Dirmeyer, P. A., F. J. Zeng, A. Duchame, J. C. Morrill, and R. D. Koster, 2000: The
sensitivity of surface fluxes to soil water content in three land surface schemes. J.
Hydrometeor., 1, 121 - 134.
Dobson, M. C., F. T. Ulaby, M. T. Hallikainen, and M. A. El-Reyes, 1985: Microwave
dielectric behavior of wet soil - Part 2: Dielectric mixing models. IEEE Trans.
Geosci. Remote Sens. GE-23, 35 - 46.
England, A. W., Galantowicz, J. F., and M. S. Schretter, 1992: The Radiobrightness
Thermal Inertia Measure of Soil Moisture. IEEE Trans. Geosci. Remote Sensing,
30, 1 3 2 - 139.
Entekhabi, D., and P. S. Eagleson, 1989: Land surface hydrology parameterization for
atmospheric and general circulation models including subgrid scale spatial
variability. J. Climate, 2, 816 - 831.
Entekhabi, D., G. Asrar, et al., 1999: An agenda for land-surface hydrology research and
a call for the second international hydrological decade. Bull. Amer. Meteorol.
Soc. 80(10): 2043-2058.
Entin, J. K., A. Robock, Y. V. Konstantin, S. E. Hollinger, S. Liu, and A. Namkai, 2000:
Temporal and spatial scales of observed soil moisture variations in the
extratropics. J. Geophys. Res., 105, 11865 - 11877.
Famiglietti, J. S., and E. F. Wood, 1994: Multiscale modeling of spatially variable water
and energy balance processes. Water Resour. Res., 30, 3060 - 3078.
127
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Fennessy, M. J., and J. Shukla, 1999: Impact of initial soil wetness on seasonal
atmospheric prediction. J. Climate, 12, 3167 - 3180.
Grody, N. C., 1991: Classification of snow cover and precipitation using the special
sensor microwave/imager. J.Geophys. Res., 94, 7423 - 7435.
Grody, N. C. and A. N. Basist, 1996: Global identification of snow cover using SSM/I
measurements. IEEE Trans. Geosci. Remote Sens., 34, 237 - 249.
Guha, A., J. M. Jacobs, et al., 2003: Soil moisture mapping using ESTAR under dry
conditions from the Southern Great Plains Experiment (SGP99). IEEE Trans.
Geosci. Remote Sens., 41(10), 2392-2397.
Guha, A. and V. Lakshmi, 2002: Sensitivity, Spatial Eleterogeneity, and Scaling of CBand Microwave Brightness Temperatures for Land Hydrology Studies. IEEE
Trans. Geosci. Remote Sens., 40(12), 2626-2635.
Guijarro, L. N., V. Lakshmi, and Y. H. Kerr, 2004: Land surface temperature and surface
soil moisture retrieval using the SSM/I instrument. J. Hydrometeo., In Press.
Guijarro, L. N. and V. Lakshmi, 2005: A microwave multi-sensor study of land surface
emissivity over the Southern Great Plains, Oklahoma, USA. Int. J. Remote Sens.
Submitted.
Guijarro, L. N., V. Lakshmi, and D. Bosch, 2005: Soil moisture retrieval from the Soil
Moisture Experiments (SMEX03) in South Central Georgia from AMSR-E.
Remote Sens. Environ. Submitted.
Hallikainen, M. T., F. T. Ulaby, M. C. Dobson, M. A. El-Rayes, and L. K. Wu, 1985:
Microwave dielectric behavior of wet soil - Part I: Empirical models and
experimental observations. IEEE Trans. Geosci. Remote Sensing, GE - 23, 25 34.
Hallikainen, M. T., and P. A. Jolma, 1986: Retrieval of the water equivalent of snow
cover in Finland by satellite microwave radiometry. IEEE Trans. Geosci. Remote
Sens. GE-24(6), 855-862.
Hiltbrunner, D., C. Matzler, and A. Wiesmann, 1994: Monitoring land surfaces with
combined DMSP-SSM/I and ERS-1 scatterometer data. Proc. IGARSS’94 Symp.,
pp. 1945-1947.
Heymsfield, G. M., and R. Fulton, 1991: Modulation of SSM/I microwave soil radiances
by rainfall. Remote Sens. Environ., 36, 203 - 212. Hollinger, J. P., J. L. Pierce,
and G. A. Poe, 1990: SSM/I Instrument Evaluation. IEEE Trans. Geosci. Remote
Sens., 39, 187-202.
128
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Jackson, T. J., T. J. Schmugge, and J. R. Wang, 1982: Passive microwave sensing of soil
moisture under vegetation canopies. Water Resour. Res., 18, 1137- 1142.
Jackson, T.J. and P. O’Neill, 1987: Temporal observations of surface soil moisture using
a passive microwave sensor. Remote Sens. Environ., 21:281-296.
Jackson, T. J., and T.J. Schmugge, 1989: Algorithm for the passive microwave remote
sensing of soil moisture in Microwave Radiometry and Remote Sens. Appl., P.
Pampaloni, Ed. Zeist, The Netherlands: VSP, 3 - 17.
Jackson, T.J. and P. O’Neill, 1990: Attenuation of soil microwave emission by com and
soybeans at 1.4 and 5 GHz. IEEE Trans. Geosci. Remote Sens., 28: 978-980.
Jackson, T. J., and T. J. Schmugge, 1991: Vegetation effects on the microwave emission
of soils. Remote Sens. Environ., 36, 203 - 212.
Jackson, T. J., 1993: Measuring surface soil moisture using passive microwave remote
sensing. Hydrol. Process., 7, 1 3 9 - 152.
Jackson, T. J., 1997: Soil moisture estimation using special satellite microwave/imager
satellite data over a grassland region. Water Resour. Res., 33(6), 1475 - 1484.
Jackson, T. J., D. M. Le Vine, et al., 1999: Soil moisture mapping at regional scales using
microwave radiometry: The Southern Great Plains Hydrology Experiment. IEEE
Trans. Geosci. Remote Sens. 37(5): 2136-2151.
Jackson, T. J. and A. Y. Hsu, 2001: Soil moisture and TRMM microwave imager
relationships in the Southern Great Plains 1999 (SGP99) Experiment. IEEE
Trans. Geosci. Remote Sens. 39(8): 1632-1642.
Jackson, T. J., A. Y. Hsu, and P.E. O’Neill, 2002a: Surface soil moisture retrieval and
mapping using high-frequency microwave satellite observations in the Southern
Great Plains. J. Hydrometeo., 3(6): 688-699.
Jackson, T. J., A. J. Gasiewski, et al., 2002b: Soil moisture retrieval using the C-band
polarimetric scanning radiometer during the Southern Great Plains 1999
Experiment. IEEE Trans. Geosci. Remote Sens. 40(10): 2151-2161.
Josberger, E. G., P. Gloersen, A. T. C. Chang, and A. Rango, 1996: The effects of snow
pack grain size on the passive microwave signatures from the upper Colorado
river basin snowpack. J. Geophys. Res. 101, 6679-6688.
Kerr, Y. H., and E. G. Njoku, 1990: A semiempirical model for interpreting microwave
emission from semiarid land surfaces as seen from space. IEEE Trans. Geosci.
Remote Sens., 28(3), 384 - 393.
129
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Kerr, Y. H., N. M. Mognard, S. Baillarin, J. C. Meunier, and S. Kalagdew, 1999:
Combined use of passive and active microwave data to monitor seasons over high
latitude continental regions. Nordic Hydrology, 1 - 40.
Kerr, Y. H., J. P. Lagouarde, and J. Imbernon, 1992: Accurate land surface temperature
retrieval from AVHRR data with the use of an improved split window algorithm.
Remote Sens. Environ., 41, 197 - 209.
Lakshmi, V., E. F. Wood, and B. J. Choudhury, 1997a: Evaluation of the Special Sensor
Microwave/Imager satellite data for regional soil moisture estimations over the
Red River Basin. J. Appl. Meteor., 36, 1309 - 1328.
Lakshmi, V., E. F. Wood, and B. J. Choudhury, 1997b: Investigation of effect of
heterogeneities in vegetation and rainfall on simulated SSM/I brightness
temperatures. Int. J. Remote Sens., 18, 2763 - 2784.
Laymon, C.A., W.L. Crosson, T.J. Jackson, A. Manu, A., and T.D Tsegaye, 2001:
Ground-based passive microwave remote sensing observations of soil moisture at
S-band and L-band with insight into measurement accuracy. IEEE Trans. Geosci.
Remote Sens., 39: 1844-1858.
Leese, J., T. J. Jackson, et al., 2001: Meeting summary: GEWEX/BAHC International
Workshop on Soil Moisture Monitoring, Analysis, and Prediction for
Hydrometeorological and Hydroclimatological Applications. Bull. Amer.
Meteorol. Soc., 82: 1423-1430.
Le Vine, D. M., T. J. Jackson, et al., 2001: ESTAR measurements during the Southern
Great Plains experiment (SGP99). IEEE Trans. Geosci. Remote Sens. 39(8):
1680-1685.
Li, L., E. G. Njoku, E. Im, P. S. Chang, and K. S. Germain, 2004: A preliminary surery of
radio-frequency interference over the US in Aqua AMSR-E data. IEEE Trans.
Geosci. Remote Sens., 422, 380-390.
Long, D. G., and D. L. Daum, 1998: Spatial Resolution Enhancement of SSM/I Data.
IEEE Trans. Geosci. Remote Sens., 36, 407 -4 1 7 .
McCabe, M. F., B. H. Gao, and E. F. Wood, 2004: An evaluation of AMSR-E derived
soil moisture retrievals using ground based, airborne and ancillary data during
SMEX02. J. Hydrometeo., In Press.
McCabe, M. F., E. F. Wood and B. H. Gao, 2005: Initial soil moisture retrievals from
AMSR-E: Multiscale comparison using in situ data and rainfall patterns over
Iowa. Geophys. Res. Letters, 32. In Press.
130
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
McFarland, M. J., R. L. Miller, and C. M. U Neale, 1990: Land surface temperature
derived from the SSM/I passive microwave brightness temperatures. IEEE Trans.
Geosci. Remote Sens., 28, 839 - 845.
Mo, T., B. J. Choudhury, T. J. Schmugge, J. R. Wang, and T. J. Jackson, 1982: A model
for the microwave emission from vegetation covered fields. J. Geophys. Res., 87,
11229- 11237.
Narayan, U., V. Lakshmi, and E. G. Njoku, 2004: Retrieval of soil moisture from passive
and active L/S band sensor (PALS) observations during the Soil Moisture
Experiment in 2002 (SMEX02). Remote Sens. Environ., 924, 483-496.
Njoku, E. G., and D. Entekhabi, 1996: Passive microwave remote sensing of soil
moisture, J. Hydrol., 184, 101 - 129.
Njoku, E. G., and J. A. Kong, 1977: Theory for passive microwave remote sensing of
near surface soil moisture. J. Geophys. Res. 87, 11229 - 11237.
Njoku, E. G., W. J. Wilson, et al. 2002: Observations of soil moisture using a passive and
active low-frequency microwave airborne sensor during SGP99. IEEE Trans.
Geosci. Remote Sens. 40(12): 2659-2673.
Njoku, E. G., T. J. Jackson, et al., 2003: Soil Moisture Retrieval from AMSR-E. IEEE
Trans, on Geosci. Remote Sensing 41(2): 215-229.
Owe, M., A. T. C. Chang, and R. E. Golus, 1988: Estimating Surface Soil-Moisture from
Satellite Microwave Measurements and a Satellite Derived Vegetation Index.
Remote Sens. Environ., 24(2): 331-345.
Owe, M., A. A. Van de Griend, and A. T. C. Chang, 1992: Surface soil moisture and
satellite passive microwave observations in semi-arid Southern Africa. Water
Resour. Res., 28(3), 829 - 839.
Peplinski, N. R., F. T. Ulaby, et al., 1995: Dielectric-Properties of Soils in the 0.3-1.3Ghz Range. IEEE Trans. Geosci. Remote Sens. 33(3): 803-807.
Rott, B. and A. E. Nagler, 1994: Intercomparison of snow retrieval algorithms by means
of spacebome microwave radiometry, in Passive Microwave Remote Sensing o f
Land-Atmosphere Interactions, Eds. Choudhury, B. J., Y. H. Kerr, E. G. Njoku
and P. Pampaloni, VSP, 227-244.
Rowntree, P. R., and J. R. Bolton, 1983: Simulation of atmospheric response to soil
moisture anomalies over Europe. Quart. J. Roy. Meteor. Soc., 109(641), 501 526.
131
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Schmugge, T. J., 1983: Remote sensing of soil moisture, IEEE Trans. Geosci. Remote
Sens., 21, 336 - 344.
Schmugge, T. J., 1985: Remote Sensing o f Soil Moisture, Hydrological Forecasting, John
Wiley & Sons Ltd., 100-125.
Schmugge, T. J., P. E. O’Neill, J. R. Wang, 1986: Passive microwave soil moisture
research. IEEE Trans. Geosci. Remote Sens., 24, 12 - 22.
Sheridan, J. M., 1997: Rainfall-streamflow relations for coastal plain watersheds. A.
Engineering in Agriculture 13(3): 333-344.
Shukla, J., and Y. Mintz, 1982: Influence of land-surface evapotranspiration on the
Earth’s climate. Science, 215, 1498 - 1500.
Teng, W. L., Wang, J. R., and P.C. Doraiswamy, 1993: Relationship between satellite
microwave radiometric data, antecedent precipitation index, and regional soil
moisture. Int. J. Rem. Sens., 14, no. 13, 2483 - 2500.
Tsang, L., and R. W. Newton, 1982: Microwave emission from soils with rough surfaces.
J. Geophys. Res., 87, 9017 - 9024.
Ulaby, F. T., R. K. Moore, and A. K. Fung, 1986a: Microwave Remote Sensing: active
and passive: From theory to applications. Vol.l. Dedham (MA), Artech, 262-263
pp.
Ulaby, F. T., R. K. Moore, and A. K. Fung, 1986b: Microwave Remote Sensing: active
and passive: From theory to applications. Vol.3. Dedham (MA), Artech, 1522 pp.
Van Vleck, J. El., 1947: The Absorption of Microwaves by Oxygen, Phys. Rev., 71, 413 424.
Vinnikov, K. Y., A. Robock, S. Qiu, J. K. Entin, M. Owe, B. J. Choudhury, S. E.
Hollinger, and E. G. Njoku, 1999a: Satelllite remote sensing of soil moisture in
Illinois, United States. J. Geophys. Res., 104(D4), 4145-4168.
Vinnikov, K. Y., A. Robock, S. Qiu, and J. K. Entin, 1999b: Optimal design of surface
networks for observation of soil moisture. J. Geophys. Res., 104(D16), 19743 19749.
Walker, J. M., and P. R. Rowntree, 1977: The effect of soil moisture on circulation and
rainfall in a tropical model. Quart. J. Roy. Meteor. Soc., 103, 29 - 46.
Wang, J. R. and T. J. Schmugge, 1980: An empirical model for the complex dielectric
permittivity of soils as a function of water content. IEEE Trans. Geosci. Remote
Sens., GE -18, 2 8 8 -2 9 5.
132
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Wang, J. R., and B. J. Choudhury, 1981: Remote sensing of soil moisture content over a
bare field at 1.4 GHz frequency. J.Geophys. Res., 86, 5277 - 5280.
Warrick, A.W. and D. R. Nielsen, 1980: Spatial variability of soil physical properties in
the field. In Applications in Soil Physics. D. Hillel (ed.). Academic Press, New
York, New York.
Weng, F., and N. C. Grody, 1998: Physical retrieval of land surface temperature using the
special sensor microwave imager. J. Geophys. Res., 103(D8), 8839 - 8848.
Wismann, V., A Cavanie, D. Hoekman, I. Woodhouse, K. Boehnke, C. Schmullius, 1996:
Land surface observations using ERS-1 windscatterometer, ifars, ESA Contract,
11103/94/NL/CN.
Wigneron, J. P., A. Chanzy, J. C. Calvet, A. Olioso, and Y. Kerr, 2002: Modeling
approaches to assimilating L band passive microwave observations over land
surfaces. J. Geophys. Res., 107(D14), 1 - 14.
Wilheit, T. T. J., 1978: Radiative transfer in a plane stratified dielectric. IEEE Trans.
Geosci. Remote Sens., 16, 138-143.
Wilson, W. J., S. H. Yueh, et al. 2001: Passive active L- and S-band (PALS) microwave
sensor for ocean salinity and soil moisture measurements. IEEE Trans. Geosci.
Remote Sens., 39(5): 1039-1048.
World Climate Research Programme, “CLIVAR a study of climate variability and
predictability,” WCRP-89, 1995.
133
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Документ
Категория
Без категории
Просмотров
0
Размер файла
2 861 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа