SOIL MOISTURE AND SURFACE ROUGHNESS RETRIEVAL FROM SAR DATA

1
SOIL MOISTURE AND SURFACE ROUGHNESS RETRIEVAL
FROM SAR DATA
Dr. Jakob J. van Zyl RADAR SCIENCE AND
ENGINEERING SECTION JET PROPULSION
LABORATORY CALIFORNIA INSTITUTE OF
TECHNOLOGY 4800 OAK GROVE DRIVE PASADENA, CA 91109
2
OUTLINE

• INTRODUCTION
• ALGORITHM 1 Oh et al.
• ALGORITHM 2 Dubois et al.
• ALGORITHM 3 Shi et al.
• In Situ MEASUREMENTS OF SOIL MOISTURE

3
Introduction Hydrologic Cycle
4
Introduction Biosphere-Atmosphere Models
5
Introduction Dielectric Constant vs. Soil
Moisture
6
Introduction Radar Response

• Models of scattering from rough surfaces predict
  both the absolute radar cross-section and the
  ratio of the co-polarized returns ( HH and VV) to
  be functions of the surface roughness and
  dielectric constant.
• The surface dielectric constant is a function of
  the surface soil moisture the wetter the
  surface, the higher the dielectric constant. Dry
  soils have dielectric constants of 2-3, while
  water has a dielectric constant of 80.
• Unfortunately, most analytical and numerical
  models are difficult to invert
• Using measured data, several empirical models
  have been developed to invert the observed radar
  cross-sections for surface roughness and soil
  moisture. All these use different combinations
  of radar cross-sections measured at HH, VV and/or
  HV

7
Introduction Theoretical Basis for Inversions
8
Introduction Definition of Errors
9
Algorithm 1 Oh et al. Reference
10
Oh et al. Description

• Using data measured by the University of
  Michigans POLARSCAT truck mounted scatterometer
  operating at L-, C- and X-Band, Oh et al. derived
  the following empirical expressions for
  scattering from bare soil surfaces
• In these expressions,
  and is the real part of the soil
  dielectric constant, and is the surface
  r.m.s. height.

11
Oh et al. Data Fit
12
Oh et al. Graphical Representation
13
Oh et al. Results
14
Algorithm 2 Dubois et al. Reference
15
Dubois et al. Description

• Using data measured with truck-mounted
  scatterometers from the University of Michigan
  and the University of Berne at frequencies
  ranging from L-band to X-band, Dubois et al.
  derived expressions for the cross-sections at HH
  and VV
• In these expressions, is the incidence angle,
  is the real part of the dielectric constant,
  is the radar wavenumber, and
  is the r.m.s. surface height.

16
Dubois et al. Data Fit
17
SOIL MOISTURE AND SURFACE ROUGHNESS Dubois et
al. Graphical Representation
INCREASING MOISTURE
INCREASING ROUGHNESS
18
Dubois et al. Inversion

• The Dubois et al. equations can be directly
  inverted to yield
• with

19
Dubois et al. Results

• The algorithm has been tested with SAR data from
  AIRSAR as well as SIR-C
• Soil moisture accuracies compared to in situ
  measurements of the top 5 cm were found to be
  about 4.2
• Surface roughness accuracies found were on the
  order of 0.34 cm
• Algorithm should only be applied for incidence
  angles between 30 and 60 degrees
• Surfaces with roughness exceeding about 3 cm at
  L-band yield less accurate results
• Results are less accurate for very wet (gt30)
  surfaces

20
Algorithm 3 Shi et al. Reference
21
Shi et al. Description

• The model functions proposed by Shi et al. are
  based on a parameterization of the results of the
  Integral Equation Model published by Fung et al.
  The results are

22
Shi et al. Model Fit
23
Shi et al. Results
24
In Situ Measurements of Soil Moisture

• When deciding the usefulness of remotely sensed
  soil moisture estimates, one has to consider the
  natural spatial variability of the soil moisture
• Remotely sensed estimates of soil moisture
  represents areal averages of soil moisture
• This is typically compared to in situ
  measurements which represent point measurements
• Typical techniques for measuring soil moisture in
  situ include neutron probes, Time Domain
  Reflectometry (TDR), and gravimetric soil
  sampling
• In situ measurements generally provide good
  definition with depth, but due to the labor
  intensive nature of the measurements, usually
  extensive spatial sampling is not done
• Field scale variability was shown to be between
  2-3 soil moisture for fields of 16 hectares
• Therefore, the natural spatial variability of
  soil moisture is of the same order of magnitude
  as the demonstrated accuracies fro radar remote
  sensing.