Bare Surface Soil Moisture and Surface Roughness Estimation Using Multi-band Multi-Polarization NISAR-like SAR Data

Sami Najem^1*Nicolas Baghdadi¹Henri Hassan Bazzi²Mehrez Zribi³Dharmendra Kumar Pandey⁴Sekhar Muddu⁵

• ¹Institut National de recherche pour l’agriculture, l’alimentation et l’environnement (INRAE), Paris, France
• ²AgroParisTech Institut des Sciences et Industries du Vivant et de L'environnement, Paris, Ile-de-France, France
• ³Institut de Recherche Pour le Développement (IRD), Marseille, Provence-Alpes-Côte d'Azur, France
• ⁴Indian Space Research Organisation, Bengaluru, Karnataka, India
• ⁵Indian Institute of Science (IISc), Bangalore, Karnataka, India

The upcoming NISAR Earth-observation satellite will utilize dual frequencies simultaneously, providing synthetic aperture radar (SAR) remote sensing data in both the L-band and S-band. With its ability to operate single-polarization, dual-polarization, and quad-polarization modes, NISAR will offer significant capabilities for land surface observation applications, particularly for estimating surface soil moisture (SSM) and surface roughness (Hrms). This study aims to demonstrate NISAR's future potential in SSM and Hrms estimation by evaluating the single (SP), double (DP) and quad (QP) polarization configurations. Noisy synthetic NISAR-like data was generated using the Dubois-B model for both S-and L-bands. The use of a priori information on the soil moisture was also examined for SSM and Hrms estimations. Various neural networks (NNs) were trained using the noisy synthetic dataset. Validation was performed on noisy synthetic data, as real NISAR data is not yet available. Out of the NISAR configurations tested, the QP configuration was shown to be the most performant, with RMSE on SSM estimation of 4.2 vol.%, for QP configuration compared to 5.1 and 8.2 vol.% for SP and DP configurations when not using a priori knowledge of soil moisture conditions. RMSE on Hrms was 0.3 cm for QP configuration, compared to 0.7 and 0.6 cm for SP and DP configurations. The QP was also shown to be more capable of mitigating the effect of the incidence angle on the estimation of SSM and Hrms compared to the two other configurations. Moreover, simultaneous use of S-and Lbands enhances SSM and Hrms estimation compared to using either of these frequency bands alone in single, dual, or quad-polarization configurations. Furthermore, using a priori knowledge of soil moisture conditions was successful in improving the estimation precision for SSM for all NISAR configurations. Notably, for QP configuration, RMSE on SSM estimation was 3.9 vol.% and 3.2 vol.% when a priori information on SSM was considered respectively in dry to slightly wet and very wet conditions. These findings demonstrate the high potential of the future NISAR sensor for estimating SSM and Hrms.