Speaker
Description
The Soil Moisture Ensemble Method (SMEM) is a project funded by the European Space Agency (ESA) aimed at enhancing both the accuracy and spatial detail of soil moisture (SM) and vegetation water content (VWC) retrievals derived from multi-frequency Synthetic Aperture Radar (SAR) observations. The initiative addresses some of the major limitations affecting current retrieval approaches, particularly in areas characterized by complex topography, dense vegetation cover, or arid environmental conditions.
The SMEM framework relies on an ensemble methodology that integrates information from multiple SAR frequencies and sensors together with optical multispectral observations. This integration is achieved through advanced Artificial Intelligence (AI) techniques coupled with physically based models, enabling a more robust exploitation of the complementary information provided by different remote sensing systems.
A major component of the research focuses on the development of machine learning methods for estimating soil moisture (m³/m³) and vegetation biomass, expressed as Plant Water Content (PWC, kg/m²), in agricultural environments. The approach exploits the combined use of microwave observations collected by both satellite radiometers and SAR systems. The dataset includes radiometric measurements from NASA’s SMAP mission and JAXA’s AMSR2/3 sensors, together with L-, C-, and X-band SAR acquisitions provided by ALOS-2, SAOCOM, Sentinel-1, and COSMO-SkyMed.
Multi-temporal observations have been gathered over agricultural sites located in Tuscany, Central Italy, where the dominant land covers include wheat, maize, sorghum, sunflower, vineyards, and pasturelands. Additional information has been obtained from Sentinel-2 multispectral imagery. To support satellite data analysis and validation, dedicated field campaigns have been conducted during satellite overpasses, collecting key soil and vegetation parameters such as soil moisture, surface roughness, crop structural characteristics, and fresh and dry biomass.
The initial phase of the study investigated the sensitivity of microwave signals to soil and vegetation properties across different frequencies. These analyses were aimed at quantifying the relative contributions of soil and vegetation to the overall radar backscatter response. Building upon these results, crop and vegetation classes were identified through Deep Learning techniques applied to COSMO-SkyMed and Sentinel-2 data [1].
Subsequently, high-resolution soil moisture maps were generated through the disaggregation of low-resolution soil moisture products obtained from radiometric observations using the HydroAlgo algorithm [2]. Artificial Neural Networks (ANNs), following the methodology proposed in [3], were employed by combining high-resolution Sentinel-1 C-band data with L-band observations from ALOS-2 and SAOCOM. The inclusion of L-band measurements was intended to exploit their higher sensitivity to soil moisture variations.
Finally, high-resolution PWC estimation was achieved through the integration of COSMO-SkyMed X-band SAR data, crop classification outputs, and the derived high-resolution soil moisture products.
The outcomes obtained to date demonstrate the potential of the proposed methodology. Future activities will extend the analysis in both space and time, with the objective of assessing the robustness and transferability of the approach and performing a more comprehensive validation of the generated products
References
[1] Fontanelli, G., Lapini, A., Santurri, L., Pettinato, S., Santi, E., Ramat, G., Pilia, S., Baroni, F., Tapete, D., Cigna, F., and Paloscia, S. (2022). Early-season crop mapping on an agricultural area in italy using X-band dual-polarization SAR satellite data and convolutional neural networks. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 15, 6789-6803. doi:10.1109/JSTARS.2022.3198475.
[2] Santi E., S. Pettinato, S. Paloscia, P. Pampaloni, G. Macelloni, and M. Brogioni, 2012, “An algorithm for generating soil moisture and snow depth maps from microwave spaceborne radiometers: HydroAlgo”, Hydrol. Earth Syst. Sci., 16, pp. 3659-3676, doi:10.5194/hess-16-3659-2012.
[3] Santi E., Baroni F., Fontanelli G., Palchetti E., Paloscia S., Pettinato S., Pilia S., Ramat G., Santurri L. High-Resolution Mapping of Soil Moisture by AMSR2 Data Disaggregation Based on Sentinel-1 and Machine Learning (2024) IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 17, pp. 15077 - 15088, DOI: 10.1109/JSTARS.2024.3445111.