Elsevier

Remote Sensing of Environment

Volume 163, 15 June 2015, Pages 70-79
Remote Sensing of Environment

SMOS soil moisture retrievals using the land parameter retrieval model: Evaluation over the Murrumbidgee Catchment, southeast Australia

https://doi.org/10.1016/j.rse.2015.03.006Get rights and content

Highlights

  • Land parameter retrieval model is applied to low frequency microwave observations.

  • 1.4 GHz measurements from the Soil Moisture and Ocean Salinity mission were used.

  • LPRM uses single instead of multiple incidence angle observations.

  • The model results were evaluated with an in situ dataset of southeast Australia.

  • The results fall close to the operational SMOS soil moisture retrievals.

Abstract

The land parameter retrieval model (LPRM) is a methodology that retrieves soil moisture from low frequency dual polarized microwave measurements and has been extensively tested on C-, X- and Ku-band frequencies. Its performance on L-band is tested here by using observations from the Soil Moisture and Ocean Salinity (SMOS) satellite. These observations have potential advantages compared to higher frequencies: a low sensitivity to cloud and vegetation contamination, an increased thermal sampling depth and a greater sensitivity to soil moisture fluctuations. These features make it desirable to add SMOS-derived soil moisture retrievals to the existing European Space Agency (ESA) long-term climatological soil moisture data record, to be harmonized with other passive microwave soil moisture estimates from the LPRM. For multi-channel observations, LPRM infers the effective soil temperature (Teff) from higher frequency channels. This is not possible for a single channel mission like SMOS and therefore two alternative sources for Teff were tested: (1) MERRA-Land and (2) ECMWF numerical weather prediction systems, respectively. SMOS measures brightness temperature at a range of incidence angles, different incidence angle bins (45°, 52.5° and 60°) were tested for both ascending and descending swaths. Three LPRM algorithm parameters were optimized to match remotely sensed soil moisture with ground based observations: the single scattering albedo, roughness and polarization mixing factor. The soil moisture retrievals were optimized and evaluated against ground-based data from the Murrumbidgee Soil Moisture Monitoring Network (OzNet) in southeast Australia. The agreement with single-angle SMOS LPRM retrievals was close to the official SMOS L3 product, provided the three parameters were optimized for the OzNet dataset, with linear correlation of 0.70–0.75 (0.75–0.77 for SMOS L3), root-mean-square error of 0.069–0.085 m3 m 3 (0.084–0.106 m3 m 3 for SMOS L3) and small bias of − 0.02–0.01 m3 m 3 (0.03–0.06 m3 m 3 for SMOS L3). These results suggest that the LPRM can be applied successfully to single-angle SMOS L-band observations, but further testing is required to determine if the same set of parameters can be used in other geographic areas.

Introduction

A better understanding of the dynamics of near-surface soil moisture (θ, m3 m 3 for a top soil layer of defined thickness) with increased spatial and temporal details can be expected to improve the knowledge of energy and water fluxes between the Earth surface and the atmosphere. Evidence suggests that several important practical applications can benefit from satellite-derived θ estimates, including flood forecasting, drought monitoring and weather and climate modeling (Bisselink et al., 2011, Bolten et al., 2010, Brocca et al., 2010). Space-borne microwave observations at low frequencies (i.e. L-band, C-band, X-band) have the potential to fulfill this need. Over the years several algorithms have been developed to derive θ from passive microwave observations, resulting in numerous data products developed from 1978 onwards (Owe, De Jeu, & Holmes, 2008, and references therein). The datasets have proven their value in research applications (e.g., Jung et al., 2010, Liu et al., 2007, Taylor et al., 2012). They become even more valuable once estimates from subsequent satellite missions are combined into one consistent multi-decadal data record (De Jeu et al., 2012). This was addressed by the European Space Agency (ESA) through the Water Cycle MultiMission Observation Strategy (WACMOS) project and the Climate Change Initiative Program (CCI), in which a single consistent 32 year data record was produced by harmonizing soil moisture estimates from historical passive- and active microwave and observations (Liu et al., 2012a). This data record makes use of the land parameter retrieval model (LPRM) (Owe, De Jeu, & Walker, 2001) to derive soil moisture from the passive microwave sensors and the change detection algorithm to derive θ from the active microwave observations (Wagner, Lemoine, & Rott, 1999) as baseline algorithms to develop the long-term soil moisture record.

The LPRM is one of several methods for inferring θ from passive microwave observations. This method has been applied to observations from multiple passive microwave sensors, such as the Scanning Multi-channel Microwave Radiometer (SMMR), the Special Sensor Microwave Imager (SSM/I), the Tropical Rainfall Measuring Mission's Microwave Imager (TRMM-TMI), the Advanced Microwave Scanning Radiometer-Earth Observing System (AMSR-E) and WindSat (Owe et al., 2008, Parinussa et al., 2011a), and has been demonstrated to generate good quality θ estimates (Gruhier et al., 2010, Rossato et al., 2011, Rudiger et al., 2009, Su et al., 2013, Wagner et al., 2007). Unlike other θ retrieval methods, the LPRM simultaneously retrieves both θ and vegetation optical depth (τv, dimensionless) from microwave brightness temperatures (Tb in K) via inversion of the radiative transfer model. It therefore does not require prior external information on vegetation (Huilin et al., 2004, Kerr et al., 2012, Meesters et al., 2005).

In November 2009, ESA launched the Soil Moisture and Oceans Salinity (SMOS) satellite (Kerr et al., 2010); the first mission dedicated to soil moisture. It observes at the 1.4 GHz (L-band) frequency which is considered to be optimal for θ retrievals because of the low sensitivity to cloud and vegetation contamination, a thermal sampling depth of several centimeters, and a high sensitivity to soil moisture fluctuations (Njoku & Entekhabi, 1996). SMOS is the first of several satellite missions measuring at L-band; in 2011 Aquarius (Le Vine, Lagerloef, Colomb, Yueh, & Pellerano, 2007b) was launched, and the Soil Moisture Active Passive (SMAP) mission (Entekhabi et al., 2010) is launched in 2015. The spatial resolution of the current SMOS Level 3 soil moisture product (SMOS L3) is 43 km. The unique capabilities of SMOS make it desirable to include retrievals from the sensor in ESAs long-term soil moisture climate record. To maintain consistency in the CCI data record, arguably application of the LPRM algorithm to derive θ from SMOS may be more preferable than to use the SMOS L3 product produced by alternative algorithms. In particular, (a) the other passive microwave θ retrievals in the long-term record are derived by the LPRM, (b) the SMOS L3 produces soil moisture for the dominant land type rather than an area-averaged soil moisture estimate for the entire footprint, which introduces conceptual differences and (c) LPRM uses as little as possible ancillary data, which is highly desired for the CCI θ dataset (De Jeu et al., 2014).

LPRM has not yet been thoroughly tested in combination with L-band measurements. De Jeu, Holmes, Panciera, and Walker (2009) showed promising results applying LPRM to L-band observations and ground data from the National Airborne Field Experiment 2005 (NAFE05) over southeast Australia, but stressed that verification with satellite observations was needed, especially because of the lesser incidence angles (up to 40°) and the higher radiometric accuracy (< 0.7 K) of the airborne data, when compared to SMOS observations (up to 65°and 2.5–3 K, respectively). LPRM has typically been applied to incidence angles between 50–55° and the applicability of LPRM for a wider range of incidence angles, such as those available from SMOS, has not yet been tested. Like most θ retrieval methods, LPRM requires an estimate of the effective soil temperature (Teff in K) as input to the retrieval scheme. For multi-channel observations, Teff may be inferred from higher frequency channels (e.g. AMSR-E 37 GHz vertical polarized brightness temperature; Holmes, De Jeu, Owe, & Dolman, 2009). However the SMOS (and SMAP) sensors only have a single frequency radiometer at 1.4 GHz, and therefore ancillary temperature data are needed for θ estimation. To address this, two methods to estimate Teff from model simulated land surface temperature have been proposed: (1) by applying a phase-shift and amplitude reduction to a temperature dataset (Holmes et al., 2012, Parinussa et al., 2011b) and (2) as a function of the surface skin temperature (Tsurf), deep soil temperature (Tdeep) and θ (De Rosnay et al., 2006, Wigneron et al., 2001), which is in line with the SMOS L3 product. In this study, two objectives are addressed:

  • 1.

    Establish the quality of LPRM θ retrievals from SMOS L-band observations over the Murrumbidgee catchment and compare this with SMOS L3 θ retrievals;

  • 2.

    Understand the dependence of retrieval quality on incidence angles of 45° to 60° and on the time of overpass.

To test the parameterization of the LPRM and to evaluate its retrieval outputs ground-based data of the Murrumbidgee Soil Moisture Monitoring Network (OzNet) in southeast Australia was used, because of its dense ground observation network, the variety in land cover types and its applicability to remote sensing studies (Smith et al., 2012).

Section snippets

SMOS

The SMOS satellite carries the Microwave Imaging Radiometer with Aperture Synthesis (MIRAS); a two-dimensional interferometric radiometer that measures the passive radiation emitted by the Earth's surface at the L-band frequency (1.4 GHz). The satellite is in a polar sun-synchronous orbit with a distance of around 758 km from the Earth. The measurements are made for incidence angles between 0° and 65° (Kerr et al., 2010), have an average ground resolution of 43 km and a swath width of 1000 km in

The land parameter retrieval model

The LPRM (De Jeu et al., 2014, Owe et al., 2001, Owe et al., 2008) is developed to retrieve land surface parameters from passive microwave observations. It is based on a forward model that uses horizontally and vertically polarized microwave brightness temperature (Tb(P) in K, where P is H for horizontal or V for vertical polarization) and Teff to simultaneously solve for θ and τv. The basis of the model is the radiative transfer theory of Mo, Cloudhury, Schmugge, Wang, and Jackson (1982),

Adjusting LPRM for L-band, parameterization

The optimized parameters of h, Q and ω are listed in Table 2. Optimal values of ω ranged between 0.15 and 0.18, decreasing with incidence angle. The optimal Q was 0 in all cases, which is in line with findings of Wigneron et al. (2001). Given the h parameterization in Eq. (8), h1 values were optimal from 1.0 to 1.8, with increasing incidence angle, and similarly for h2 that valued from 3.5 to 6.3 (3.5 times h1). All three parameters where independent of acquisition time and apply to both the

Conclusion and outlook

This work demonstrated that the LPRM is capable of retrieving soil moisture estimates over OzNet using single-angle L-band observations by the SMOS satellite of equivalent quality when compared to alternative methods and products. We focused our comparison on the official SMOS L3, but existing literature (Su et al., 2013) also suggests a comparable performance against other remotely sensed soil moisture products (ASCAT and AMSR-E LPRM). Optimization and evaluation against OzNet in situ

Acknowledgments

This work was partially supported under the Water Information Research and Development Alliance between the Australian Bureau of Meteorology and CSIRO Water for a Healthy Country Flagship. I would like to thank CSIRO for their hospitality and for giving me the opportunity to conduct this research with their support. We would also like to thank the staff at the University of Melbourne and Jeff Walker and his colleagues at Monash University who have led the OzNet program. This project is partly

References (62)

  • A. Camps et al.

    Improved image reconstruction algorithms for aperture synthesis radiometers

    IEEE Transactions on Geoscience and Remote Sensing

    (2008)
  • B. Choudhury et al.

    A parameterization of effective soil temperature for microwave emission

    Journal of Geophysical Research

    (1982)
  • W. Crow et al.

    An improved approach for estimating observation and model error parameters for soil moisture data assimilation

    Water Resources Research

    (2010)
  • De Jeu

    Retrieval of land surface parameters using passive microwave remote sensing

    (2003)
  • R.A.M. De Jeu et al.

    [Global climate] soil moisture

  • R.A.M. De Jeu et al.

    Parameterization of the land parameter retrieval model for L-band observations using the NAFE'05 data set

    IEEE Geoscience and Remote Sensing Letters

    (2009)
  • R.A.M. De Jeu et al.

    Soil moisture retrieval from passive microwave sensors: Algorithm theoretical baseline document, version 2.0

    ESA CCI Soil Moisture

    (2014)
  • G.J.M. De Lannoy et al.

    Global calibration of the GEOS-5 L-band microwave radiative transfer model over nonfrozen land using SMOS observations

    Journal of Hydrometeorology

    (2013)
  • P. De Rosnay et al.

    Parameterizations of the effective temperature for L-band radiometry. Intercomparison and long-term validation with SMOSREX field experiment

  • P. De Rosnay et al.

    Parameterizations of the effective temperature for L-band radiometry. Inter-comparison and long term validation with SMOSREX field experiment

  • W.A. Dorigo et al.

    Error characterisation of global active and passive microwave soil moisture datasets

    Hydrology and Earth System Sciences

    (2010)
  • G. Dumedah et al.

    Can SMOS data be used directly on the 15-km discrete global grid?

    IEEE Transactions on Geoscience and Remote Sensing

    (2014)
  • D. Entekhabi et al.

    The soil moisture active passive (SMAP) mission

    Proceedings of the IEEE

    (2010)
  • M.J. Escorihuela et al.

    Simple model of the bare soil microwave emission at L-band

    IEEE Transactions on Geoscience and Remote Sensing

    (2007)
  • C. Gruhier et al.

    Soil moisture active and passive microwave products: Intercomparison and evaluation over a Sahelian site

    Hydrology and Earth System Sciences

    (2010)
  • T.R.H. Holmes et al.

    Land surface temperature from Ka band (37 GHz) passive microwave observations

    Journal of Geophysical Research-Atmospheres

    (2009)
  • T.R.H. Holmes et al.

    An assessment of surface soil temperature products from numerical weather prediction models using ground-based measurements

    Water Resources Research

    (2012)
  • G. Huilin et al.

    Using a microwave emission model to estimate soil moisture from ESTAR observations during SGP99

    Journal of Hydrometeorology

    (2004)
  • M. Jung et al.

    Recent decline in the global land evapotranspiration trend due to limited moisture supply

    Nature

    (2010)
  • Y.H. Kerr et al.

    The SMOS soil moisture retrieval algorithm

    IEEE Transactions on Geoscience and Remote Sensing

    (2012)
  • Y.H. Kerr et al.

    The SMOS mission: New tool for monitoring key elements of the global water cycle

    Proceedings of the IEEE

    (2010)
  • Cited by (39)

    • The first global soil moisture and vegetation optical depth product retrieved from fused SMOS and SMAP L-band observations

      2022, Remote Sensing of Environment
      Citation Excerpt :

      Another typical case is the ESA Climate Change Initiative (CCI) project, which is devoted to constructing long-term satellite-based SM records by combining various single-sensor passive and active microwave SM products (Liu et al., 2012; Dorigo et al., 2017). Several approaches/studies have explored the integration of SMOS retrievals in the CCI SM data record, including neural networks (Rodríguez-Fernández et al., 2016), multi-linear regression methods (Al-Yaari et al., 2016, 2017) and merging retrieval algorithm (Van der Schalie et al., 2015). In the latter case, for instance, by improving the Land Parameter Retrieval Model (LPRM), the driving inversion algorithm of the CCI passive products, the SM retrievals from LPRM with different Hr and ω values based on SMOS and AMSR-E (the Advanced Microwave Scanning Radiometer for EOS) were integrated into a consistent dataset using a simple linear regression (Van der Schalie et al., 2017).

    View all citing articles on Scopus
    View full text