Hyperspectral Optical, Thermal, and Microwave
L-Band Observations For Soil Moisture Retrieval
at Very High Spatial Resolution
Nilda Sánchez, Maria Piles, José Martínez-Fernández, Mercè Vall-llossera, Luca Pipia, Adriano Camps,
Albert Aguasca, Fernando Pérez-Aragüés, and Carlos M. Herrero-Jiménez
Abstract
The results of an experiment conducted in Spain over the
Soil Moisture Measurement Stations Network of the Uni-
versity of Salamanca (
REMEDHUS
) are presented. The obser-
vations included airborne observations from hyperspectral
optical, thermal, and microwave sensors coinciding with
intensive field measurements. The hyperspectral optical
and thermal datasets were first analyzed and processed to
select the best hyperspectral features to be included in the
soil moisture retrieval procedure. A linear model linking the
selected hyperspectral features to the microwave observa-
tions and the in situ soil moisture is proposed. The appli-
cation of this model resulted in soil moisture estimates that
agree with in situ measurements (correlation coefficient: R
>0.76, root mean squared differences:
RMSD
<0.07 m
3
m
-3
).
The hyperspectral dataset strengthened the link between
optical, thermal and microwave L-band observations with
soil moisture, and provided a spatial framework to disag-
gregate soil moisture at very high spatial resolution (3.5 m),
useful in hydrological modeling and precision agriculture.
Introduction
The first two space missions dedicated specifically to soil
moisture retrieval from passive L-band observations are lead-
ing to intense scientific activity. The European Spatial Agency
(
ESA
) Soil Moisture and Ocean Salinity (
SMOS
) mission (Kerr
et al
., 2010) has been providing soil moisture maps since
November 2009. The US National Aeronautics and Space
Administration (
NASA
) plans to launch the
SMAP
(Soil Mois-
ture Active Passive) mission in 2014; this satellite will carry
a radiometer and a synthetic aperture radar on-board (En-
tekhabi
et al
., 2010). Due to practical constraints on antenna
size and the altitude of low Earth orbits, the spatial resolu-
tion of
SMOS
and
SMAP
radiometers is limited to 40 to 50 km.
This resolution is adequate for many global applications but
restricts the use of the resulting data in regional studies over
land, where a resolution of 1 to10 km is needed (Crow
et al
.,
2000; Entekhabi
et al
., 1999; Piles
et al
., 2011). Multi-sensor
disaggregation techniques are emerging as a new technique
for refining broad resolution observations using a variety
of optical sensors. Combining optical and microwave data,
some studies have shown that soil moisture estimates can be
obtained at intermediate spatial resolutions that compare well
with
in situ
data. Visible/infrared/thermal sensors are used to
provide indirect measurements of soil moisture at high resolu-
tion; these measurements are combined with accurate passive
microwave observations to construct soil moisture maps with
resolutions ranging from several tens of meters with 15 day re-
visit (Merlin
et al
., 2013) to 1 km daily (Kim and Hogue, 2012;
Merlin
et al
., 2005; Piles
et al
., 2011). Microwave/optical data
merging methods for estimating high resolution soil moisture
are generally based on the intrinsic relationship between
vegetation indices (
VI
) and land surface temperature (
LST
)
with soil moisture using empirical approaches (Carlson
et al
.,
1994; Price, 1990; Sobrino
et al
., 2012). This relationship can
be graphically represented by a triangular-trapezoidal shape,
which is formed by the scatter plot of surface temperature ver-
sus
.
vegetation indices under a full range of vegetation covers
and soil moisture availability (Tang
et al
., 2010). The so-called
surface temperature-vegetation index (
LST-VI
) triangle method
was first introduced during the 1990s by Price (1990) and later
elaborated upon by Carlson
et al
. (1994 and 1995). It has been
frequently used to estimate evapotranspiration or evaporative
fraction (Batra
et al
., 2006; Jiang and Islam, 2001; Moran
et al
.,
1994; Venturini
et al
., 2004), determine soil moisture (Gillies
et al
., 1997; Sandholt
et al
., 2002) and downscale coarse-scale
microwave soil moisture estimates (Chauhan
et al
., 2003;
Kim and Hogue, 2012; Piles
et al
., 2011). In Piles
et al
. (2011),
a regression analysis was performed relating soil moisture
reference data (
SMOS
L2 product) to
MODIS LST
,
MODIS
Normal-
ized Difference Vegetation Index (
NDVI
) and
SMOS
L1 bright-
ness temperatures (
BT
). The objective was to develop a robust
model to link 40-km
SMOS
and 1-km
MODIS
observations to
soil moisture and to generate a
SMOS
-L4 soil moisture product
as an optimal blend of microwave/visible data at 1 km reso-
lution. In the present study, the objective is to improve upon
this research line by using simultaneous hyperspectral and
L-band airborne imagery.
Hyperspectral sensors have been used primarily to indi-
rectly estimate soil moisture, but the best spectral range for
detecting soil moisture has yet to be determined. Aside from
the microwave range, visible near-infrared (
VNIR
) (0.4–1.4
μm), near-infrared (
NIR
) (0.75–1.4 μm), short-wavelength
infrared (
SWIR
) (1.4–3.0 μm), and thermal infrared (
TIR
)
Nilda Sánchez, José Martínez-Fernández, and Carlos M.
Herrero-Jiménez are with the Universidad de Salamanca,
CIALE, Duero 12, 37185 Salamanca, Spain (
.
Maria Piles, Mercè Vall-llossera, Adriano Camps, and Albert
Aguasca are with the Universitat Politècnica de Catalunya,
Jordi Girona 1-3, E-08034 Barcelona, Spain.
Luca Pipia and Fernando Pérez-Aragüés are with the Programa
Català d’Observació de la Terra-Institut Cartogràfic de Catalun-
ya (
PCOT
/
ICC
), Parc de Montjuïc, Barcelona, 08038, Spain.
Photogrammetric Engineering & Remote Sensing
Vol. 80, No. 8, August 2014, pp. 745–755.
0099-1112/14/8008–745
© 2014 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.80.8.745
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
August 2014
745
