Global retrieval of soil moisture using neural networks trained with synthetic radiometric data

Nemesio Rodriguez‐fernandez (1) , Philippe Richaume (1) , Yann H. Kerr (1) , Filipe Aires (2) , Catherine Prigent (2) , Jean-Pierre Wigneron (3)

1 CESBIO - Centre d'études spatiales de la biosphère
2 Observatoire de Paris
3 UMR ISPA - Interactions Sol Plante Atmosphère

This paper discusses a methodology to construct a synthetic dataset using realistic geophysical data and the L-MEB model to compute synthetic brightness temperatures (Tb's) and to train a Neural Network (NN) for global retrievals of soil moisture (SM). The trained NNs are applied to real Tb's measured by the Soil Moisture and Ocean Salinity (SMOS) satellite (L-MEB NN). The objective is twofold. First, to compare and provide feedback to the operational algorithm. Second, to evaluate this approach in the context of pre-launch algorithm development. The performance of the L-MEB NN dataset was evaluated by comparing with time series of in situ measurements in North America. The correlation, standard deviation and bias of NN SM and in situ SM are similar to those obtained with the SMOS L3 SM product and with ECMWF models. The L-MEB NN dataset was also compared globally to the SMOS Level 3 SM and SM from ECMWF models. The L-MEB NN dataset is in general wetter than SMOS L3 SM, closer to ECMWF models. Some possible reasons are briefly discussed.

Domaines: Sciences du Vivant [q-bio]; Sciences de l'environnement

Nemesio Rodriguez‐fernandez, Philippe Richaume, Yann H. Kerr, Filipe Aires, Catherine Prigent, et al.. Global retrieval of soil moisture using neural networks trained with synthetic radiometric data. IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Jul 2017, Fort Worth, United States. ⟨10.1109/IGARSS.2017.8127273⟩. ⟨hal-02736324⟩