Semi-empirical regressions at L-band applied to surface soil moisture retrievals over grass Kauzar Saleh a, ⁎ , Jean-Pierre Wigneron a , Patricia de Rosnay b , Jean-Christophe Calvet c , Yann Kerr b a INRA-Unité d'Ecologie Fonctionnelle et Physique de l'Environnement (EPHYSE), B.P. 81, Villenave d'Ornon Cedex, 33883, France b Centre d'Etudes Spatiales de la Biosphère (CESBIO), Toulouse, France c Centre National des Recherches Météorologiques (CNRM), Météo France, Toulouse, France Received 26 October 2005; received in revised form 12 January 2006; accepted 14 January 2006 Abstract The L-band brightness temperature of natural grass fields is strongly influenced by rainfall interception. In wet conditions, the contribution of the soil, mulch, and vegetation to the overall microwave emission is difficult to decouple, thus rendering the retrieval of surface soil moisture from a direct emission model difficult. This paper investigates the development and assesses the performances of statistical regressions linking passive microwave measurements to surface soil moisture in order to assess the potential of soil moisture retrievals over natural grass. First, statistical regressions were analytically derived from the L-Band Emission of the Biosphere model (L-MEB). Single configuration (1 angle, 1 polarisation), and multi-configuration regressions (2 angles, or 2 polarisations) were developed. Second, the performance of statistical regressions was evaluated under different rainfall interception conditions. For that purpose, a modified polarisation ratio at L-band was used to build three data sets with different interception levels. In the presence of interception, a regression based on one observation angle (50°) and two polarisations was able to reduce the effects of vegetation and soil roughness on the soil moisture retrievals. The methodology presented in this study is also able to provide estimates of the vegetation and soil roughness contribution to the brightness temperature. © 2006 Elsevier Inc. All rights reserved. Keywords: Surface soil moisture; L-band; Microwave; Statistical methods 1. Introduction Soil moisture content (w s )near the surface has a strong impact on the intensity of the longwave microwave radiation naturally emitted by the Earth. Two space missions, SMOS (Soil Moisture and Ocean Salinity, Kerr et al., 2001), and HYDROS (Hydrosphere State, Entekhabi et al., 2004) will soon exploit this physical feature in order to obtain global maps of the surface soil moisture from the Earth emission at L-band (≈ 21 cm wavelength). Several approaches exist to estimate surface soil moisture from radiometric data at L-band (Wigneron et al., 2003).The principal methods published in the literature are the forward model inversion where the forward model is a brightness temperature (TB) physical model dependent on soil moisture — and statistical regressions relating TB to w s . The strength of the forward model inversion approach relies on the small depen- dence of physical models on local conditions as opposed to statistical approaches. However, the difficulty lies with deter- mining appropriate model parameters depending on the soil and vegetation types. Nevertheless, very good results can be achieved once the model parameters are established as reported in a vast number of studies, particularly for crops, for which the emission behaviour at L-band is fairly well understood (Pardé et al., 2004; Wigneron et al., 2004a; Van de Griend, in press). Currently this is the preferred approach for processing the data from the SMOS mission in preparation (Wigneron et al., 2000). Conversely, statistical regressions include local conditions in the statistical relationship, and therefore they are generally site- specific. However, statistical approaches may be of interest to investigate the potential for soil moisture retrievals over surfaces for which no accurate emission model exists, and may eventually be applied to areas in which the forward inversion of a physical model fails. Remote Sensing of Environment 101 (2006) 415 – 426 www.elsevier.com/locate/rse ⁎ Corresponding author. Tel.: +33 5 57 12 24 19; fax: +33 5 57 12 24 20. E-mail address: ksalemco@bordeaux.inra.fr (K. Saleh). 0034-4257/$ - see front matter © 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.rse.2006.01.008