2686 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 49, NO. 7, JULY 2011
Effect of Radiative Transfer Uncertainty on
L-Band Radiometric Soil Moisture Retrieval
Alexandra G. Konings, Dara Entekhabi, Senior Member, IEEE,
Steven K. Chan, Senior Member, IEEE, and Eni G. Njoku, Fellow, IEEE
Abstract—Microwave radiometry soil moisture retrieval meth-
ods suffer from uncertainties about the representation of sev-
eral effects, including dielectric mixing, surface roughness, and
vegetation opacity. These uncertainties lead to two major types
of error: systematic bias and random errors. The effect of the
uncertainties is studied using the Soil Moisture Active Passive
Algorithm Testbed, a simulation environment for evaluating error
propagation in retrieval algorithms, and two different common
retrieval algorithms (single and dual polarizations). The two types
of errors are simulated by using different representations for
each factor in the forward and retrieval parts. For both algo-
rithms, this approach introduces a spatially variable bias, which
is particularly large when using a single-polarization retrieval
algorithm. This paper illustrates the emergence of both this bias
and the random error due to uncertainty in the representation
of vegetation and soil texture effects in retrieval algorithms. The
dependence of these two types of error on vegetation and soil
texture properties is shown through mapping them over the sim-
ulation region. The relative contribution of these errors to the
total error is strongly dependent on the simulation conditions and
is not necessarily indicative of what may be experienced during
actual observations. Uncertainty due to roughness representation
causes a lower error than uncertainty in vegetation opacity and
dielectric mixing parameterizations in the simulated soil mois-
ture retrieval. Summation and compensation of multiple errors
can cause the estimate error to increase with improved radiative
transfer knowledge, even after bias removal. The retrieval of soil
moisture from microwave measurements depends on several other
parameterizations that are also uncertain. This paper is limited to
only three parameterizations that are considered to be among the
larger contributors to bias.
Index Terms—Microwave radiometry, radiative transfer, soil
moisture.
I. I NTRODUCTION
T
HE HIGH spatial and temporal variability of soil mois-
ture limits the use of in situ measurements for global
mapping. Remote sensing provides a method for retrieving
large-scale soil moisture fields. Microwave radiometry has
Manuscript received July 9, 2010; revised November 10, 2010; accepted
December 15, 2010. Date of publication March 3, 2011; date of current version
June 24, 2011.
A. G. Konings is with the Nicholas School of the Environment, Duke
University, Durham, NC 27705 USA (e-mail: konings@alum.mit.edu).
D. Entekhabi is with the Department of Civil and Environmental Engi-
neering, Massachusetts Institute of Technology, Cambridge, MA 02139 USA
(e-mail: darae@mit.edu).
S. K. Chan and E. G. Njoku are with the Jet Propulsion Laboratory,
California Institute of Technology, Pasadena, CA 91109 USA (e-mail: steven.k.
chan@jpl.nasa.gov; eni.g.njoku@jpl.nasa.gov).
Digital Object Identifier 10.1109/TGRS.2011.2105495
been well studied for satellite soil moisture retrieval [1]–[3].
L-band (0.95–1.45 GHz) is particularly suited to soil moisture
retrieval because it has greater vegetation penetration and re-
duced atmospheric attenuation compared to higher frequencies
[1]. Nevertheless, retrieval capabilities are limited under dense
vegetation cover [4]. In addition, limits on antenna technology
force radiometer spatial resolution to be generally low (on the
order of 40–50 km, depending on frequency and antenna).
Radar measurements may be made at higher spatial resolution
(on the order of a few kilometers) but suffer from increased
sensitivity to the confounding effects of soil roughness and
vegetation [5]. The Soil Moisture Active Passive (SMAP) ob-
servatory will combine the two technologies to provide global
soil moisture measurements. It is expected to launch around
2014. SMAP will provide data products at several resolutions
based on measurements from the two individual instruments
and on merging their information (e.g., [6]). The focus of this
paper is the 40-km radiometer-only product.
Most microwave radiometry soil moisture retrieval algo-
rithms are based on a zero-order single-scattering characteriza-
tion that accounts for variations due to soil type and vegetation.
Nevertheless, uncertainty remains about the optimal parametric
representation of several terms in this model. Some data to con-
strain and improve existing parameterizations may be obtained
by field experiments and measurements from dedicated flights
(e.g., [7] and [8]), but the expense and resource needs of such
campaigns limit their frequency and coverage. Furthermore,
they are hampered by the difficulty of designing a network
to obtain validation data at satellite-based measurement scales
[9]. Alternatively, retrieval algorithms can be partially tested
using an observing system simulation environment [10]. In the
SMAP Algorithm Testbed, geophysical fields taken to represent
a synthetic “truth” are used to simulate emission (the forward
component of the testbed). After coupling this emission model
to an orbital sampling module, the resulting brightness temper-
atures are passed through the proposed retrieval algorithms to
study the retrieval error. Although the dependence of these ex-
periments on an explicitly modeled forward method eliminates
their applicability for constraining the models themselves, they
may still be used to characterize the error structures associated
with soil moisture retrieval, including the contributions of land
surface heterogeneity [11], parameter uncertainty, and measure-
ment error.
Crow et al. [10] carried out a simulation experiment for
The Hydrosphere State Mission (Hydros) (the cancelled pre-
decessor to SMAP) radiometer products in preparation for
that project. The Hydros simulation experiment covered the
0196-2892/$26.00 © 2011 IEEE