A MATRIC-POTENTIAL-BASED DIELECTRIC MODEL FOR MOIST SOILS: LABORATORY VALIDATION AND APPLICATION TO AMSR2 RETRIEVALS

K. Tsujimoto¹, T. Ohta², I. Kaihotsu³

¹Okayama University, ²Assistance Unit for Research and Engineering Development (U-PRIMO), ³Hiroshima University

Dielectric models of moist soils are a critical component of microwave radiative transfer modeling and satellite soil moisture retrieval algorithms. Conventional models, such as those proposed by Wang and Schmugge (1978), Dobson et al. (1985), and Mironov et al. (2004), have advanced the field substantially but remain limited across diverse soil textures. Moreover, these models rely on empirical formulations that are not directly connected to the parameters used in soil water flow simulations. This disconnect constrains the integration of satellite observations with land surface models and data assimilation frameworks, which are needed to fill observational gaps and interpolate soil moisture estimates to finer temporal and spatial resolutions.

In this study, we propose a refinement of soil dielectric models by explicitly linking dielectric properties to soil hydraulic characteristics. Specifically, we explore the use of van Genuchten (1980) parameters, which describe the soil water retention curve in Richards’ equation (1931), to represent the relationship between volumetric water content and matric potential. These parameters can be estimated from pedotransfer functions using sand and clay content together with bulk density, and such soil information is available from global databases such as those provided by the FAO. This framework therefore allows the proposed dielectric model to be applied consistently in global-scale remote sensing, while ensuring physical compatibility with hydrological modeling.

We conducted a comparative analysis of the Dobson, Mironov, and Wang–Schmugge models together with our newly developed model under a wide range of hypothetical soil conditions, focusing on how each represents bound water content and dielectric properties. These results were further validated against complex permittivity measurements obtained using a vector network analyzer (VNA). In the proposed model, bound water content was represented by the air-dry water content at a relative humidity of 50%, and the dielectric constant of bound water was derived from Hilhorst et al. (2001) at a matric potential of –500 MPa. Above this threshold, the dielectric properties of free water were incorporated using the Birchak mixing model (Birchak et al., 1974), and the frequency dependence was described using Debye’s (1929) formula with an estimated relaxation frequency. The matric-potential-based model reproduced measured complex permittivity across a wide frequency range (25 MHz–18 GHz) and diverse soil types, including Toyoura sand, bentonite, Japanese soils (Andosol, Shirasu, Masado, paddy soils), and Cambodian paddy soils. Comparisons revealed that although the Mironov and Wang–Schmugge models yield similar overall dielectric estimates despite their different theoretical foundations, they diverge markedly in their treatment of bound water. This confirms that the characterization of bound water content and its dielectric properties remains a key source of uncertainty and an essential target for refinement in dielectric modeling.

Furthermore, when the refined dielectric model was implemented within the AMSR2 soil moisture retrieval algorithm—replacing the Dobson model currently used in the JAXA standard product—it reduced the regional dependence of retrieval accuracy. The original AMSR2-JAXA algorithm exhibited a pronounced dry bias across most global validation sites, showing relatively higher accuracy only at the Mongolian site. By contrast, the application of the proposed model alleviated this dry bias and improved consistency across diverse regions, demonstrating the critical importance of adequately modeling the soil-type dependence of the soil moisture–dielectric relationship.

In conclusion, integrating matric potential into dielectric modeling provides a promising framework for reconciling soil physics with microwave remote sensing. By enabling parameter sharing between hydrological modeling and radiative transfer simulations, and by leveraging globally available soil datasets, the proposed approach strengthens the physical basis of dielectric models. This refinement not only enhances the reliability of satellite soil moisture retrievals, but also paves the way for next-generation retrieval algorithms and data assimilation applications using AMSR2, AMSR3, and beyond.