Vegetation Effects Modeling in
Soil Moisture Retrieval Using MSVI
Mina Moradizadeh and Mohammad R. Saradjian
Abstract
Brightness temperature (
BT
) measured by passive mi-
crowave sensors is usually affected by soil moisture,
vegetation cover, and soil roughness. Soil moisture esti-
mates have been limited to regions that had either bare
soil or low to moderate amounts of vegetation cover.
In this study, Simultaneous Land Parameters Retrieval
Model (
SLPRM
) as an iterative least-squares minimization
method has been used. This algorithm retrieves surface soil
moisture, land surface temperature, and canopy temperature
simultaneously using brightness temperature data in bare
soil, low to moderate and higher amounts of vegetation cover.
Furthermore, a new index called
MSVI
(Multi Sensor Vegeta-
tion Index) has been introduced to approximate vegetation
effects on properly observed brightness temperatures. The
algorithm includes model construction, calibration, and
validation using observations carried out for the
SMEX03
(Soil
Moisture Experiment 2003) region in the South and North of
Oklahoma. The results indicated about 0.9 percent improve-
ment on soil moisture estimation accuracy using the
MSVI
.
Introduction
Previous studies have shown that there is a large number of
factors such as soil moisture, vegetation characteristics, sur-
face roughness, temperature of the soil along with soil texture
affecting the brightness temperature of the surface as observed
from space especially in the soil-vegetation medium (Njoku
and Chan, 2006; Njoku and Entekhabi, 1996).
These perturbing factors introduce varying amounts of
uncertainty into the relationship between brightness temper-
ature and soil moisture. These factors are typically required
to model radiative transfer in all regions except for regions of
snow cover and water (Njoku and Chan, 2006; Colliander
et
al.
, 2012; Tsang and Newton, 1982; Mo
et al.
, 1987; Jackson
et al.
, 1982; Ulaby
et al.
, 1983; Pampaloni and Paloscia, 1986;
Jackson and Schmugge, 1991; Wang and Schmugge, 1980;
Dobson
et al.
, 1985; Njoku, 1995, and Barnes
et al.
, 2003).
This is a challenging task to separate these variables from
the signal for those who interpret passive microwave data
over land by using a minimum of ancillary data. In order to
do this, different soil moisture retrieval algorithms have been
developed. Each of them accounting in their own way for
the various parameters which will contribute to the bright-
ness temperature (Jackson
et al.
, 1993; Wigneron
et al.
, 1995
and 2003; Njoku and Li, 1999,and Owe
et al.
, 2001; Santi
et
al.
2012; Pan
et al.
2014; Zheng
et al.
2015). In some meth-
ods, extra parameters are retrieved simultaneously from the
microwave observations of multiple polarizations, frequency,
or view angle (Li
et al.
, 2011; Njoku and Li, 1999). In most of
these algorithms, vegetation effects are generally parameter-
ized by using the radiative transfer model of Mo
et al.
(1982).
The characteristics of microwave remote sensing have
long been recognized and various methodologies have been
described to obtain surface soil characteristics accordingly
(Jackson, 1993; van de Griend and Owe, 1994; Wigneron
et al.
,
1995; Owe
et al.
, 2001; Njoku and Li, 1999; and Chanzy and
Wigneron, 2000) however, in most of the models, vegetation
effect has not been modeled properly. In such models, soil
and canopy temperature have been considered equal or have
been solely developed for vegetation areas (Owe
et al.
, 2001;
Li
et al.
, 2011; Njoku and Chan, 2006). On the other hand, due
to the lack of consideration of roughness parameter in some
models, there exist serious problems in estimation of different
soil parameters properly (Owe
et al.
, 2001; Li
et al.
, 2011).
The microwave emissivity of land surfaces is determined
primarily by characteristics of the soil and vegetation while
the biophysical and biochemical parameters of vegetation
change with soil moisture and may cause some changes
in the vegetation reflectance (Gao
et al.
, 2011; Njoku and
Chan, 2006). However, there are rare models such as the one
discussed in Wigneron
et al.
(2007) in which both rough-
ness and vegetation effects have been considered. In this
study, in order to estimate the surface parameters including
surface volumetric soil moisture (
VSM
), land surface tempera-
ture (
LST
), and canopy temperature (
CT
) simultaneously, an
iterative least-squares minimization algorithm was used. The
solution is based on a multi-parameter inversion algorithm
(
SLPRM
) which is based on considering surface roughness pa-
rameter and vegetation. Furthermore,
SLPRM
’s optimization is
performed jointly using the three lower frequencies of
AMSR-E
instead of optimizing for each band separately.
The
SLPRM
retrieval method has been applied on
AMSR-E
(Advance Microwave Scanning Radiometer-
EOS
) observations
data. The observations have been carried out for the
SMEX03
(Soil Moisture Experiment 2003) in the North of Oklahoma
(
ON
) and the South of Oklahoma (
OS
) regions.
In this study, the potential capability of the
SLPRM
algo-
rithm has been investigated using
SMEX03
. Also, a new index
called
MSVI
has been developed to estimate vegetation trans-
missivity using three different sensors.
Here, soil moisture is obtained by the mentioned model in
three levels of vegetation density (i.e., bare soil, low to moder-
ate amounts of vegetation cover, fully vegetated cover) by
means of relevant ground truth data such as soil roughness,
soil dielectric constant, soil temperature and canopy tempera-
ture. The paper is organized as follows: in the next section
the study region and the data used are described followed by
the
SLPRM
algorithm and proposed method to modeling the
vegetation effect is presented. Next, the results and discussion
of soil moisture estimates obtained using the model. The final
section presents the summary and conclusion of the study.
Remote Sensing Division, School of Surveying and Geospatial
Engineering, College of Engineering, University of Tehran,
Iran (
).
Photogrammetric Engineering & Remote Sensing
Vol. 82, No. 10, October 2016, pp. 803–685.
0099-1112/16/803–685
© 2016 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.82.10.677
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
October 2016
803
