A Novelty Data Fusion Approach for Integrating Multi-Band/Multi-Sensor Persistent Scatterers

The growing accessibility of multi-sensor Persistent Scatterer (PS) data in the advent of the European Ground Motion Service offers a well-established methodology for detecting and monitoring ground displacement over extended areas with sub-centimetric precision. The detection of ground deformation phenomena relies on the available PS density, which is influenced by the sensor resolution and specific site characteristics, such as the presence of stable natural and artificial reflectors. This study proposes a novel Data Fusion (DF) approach that integrates the displacement along the line of sight of PS products to unleash the full potential of multi-sensor combinations by synthesizing multi-band displacement information. The DF approach, developed by NHAZCA S.r.l. and the Research Center “CERI - Centro di Ricerca Previsione e Prevenzione dei Rischi Geologici” of the Sapienza University of Rome in the frame of the “MUSAR” project funded by ASI (Italian Space Agency), overcomes the limitations associated with individual sensor data, allowing for improved information content and data coverage.

The method based on the strain tensor approach combines data with different orbital geometries (i.e., ascending and descending) to obtain a comprehensive deformation map by extracting synthetic measurement points called Ground Deformation Markers. In our analysis, we applied, tested, and validated the fusion method in the Basilicata region of southern Italy, combining data extracted from the C-band Sentinel-1 Copernicus initiative and the COSMO-SkyMed constellation in X-band. We evaluated the DF performance within a test site characterized by homogeneous spatial and velocity PS data distribution. The method accuracy was assessed by comparing its interpolation capabilities to estimate the velocity of deformation at a specific location with those estimated by widely used traditional (i.e., linear interpolation, cubic spline interpolation, and inverse distance weighting) and advanced techniques (i.e., universal kriging and k-nearest neighbors). The predictions of interpolators were compared with randomly extracted ground truth datasets given by the observed PS velocities. The evaluation took into consideration several metrics, such as Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and the coefficient of determination (R-squared). After validating the DF application, we compared multi-sensor results with single-sensor PS to assess the capability of the method to improve spatial coverage and information content, enabling a more comprehensive understanding of ground displacements. The results verified the capabilities and robustness of the DF approach and underscored its efficacy in enhancing the accuracy and spatial coverage of ground deformation monitoring. The proposed study highlighted the DF approach as a valuable tool in geospatial analysis and satellite monitoring applications.