Conference Agenda

.

Recent years have seen a rapid expansion of commercial Earth Observation capabilities, particularly in Synthetic Aperture Radar (SAR). New private missions are significantly increasing data availability, revisit frequency, and imaging flexibility, enabling new operational uses forgeospatial intelligence. In this context, a new commercial X-band SAR satellite developed in Singapore introduces imaging capabilities specifically suited for high-frequency monitoring and analysis of dynamic environments.

This presentation provides an overview of the mission characteristics and the first insights derived from its data products. The system operates in the X-band and supports full polarimetric acquisitions, enabling detailed characterisation of surface scattering mechanisms and man-made structures. A repeat-acquisition strategy with a revisit time of approximately 3 days enables consistent monitoring of areas of interest and facilitates time-series analysis. This short revisit time is possible because the satellite operates on an equatorial orbit, enabling frequent, geometrically consistent observations over equatorial regions.

The combination of high spatial resolution, short revisit intervals, and full polarimetric capability makes this dataset particularly relevant for geospatial intelligence. The repeat observation strategy also provides opportunities for interferometric and time-series analyses, supporting the extraction of stable scatterers and the monitoring of subtle surface changes.

To highlight the capabilities of this dataset, we investigate a change detection framework based on the Frozen Background Reference (FBR) concept [1]. The FBR approach estimates a temporally stable background scene from a SAR time series, enabling the detection of ephemeral objects that appear between acquisitions. While previous studies mainly considered single-polarisation or dual-polarisation configurations, the availability of full polarimetric data enables the extension of this concept toward a polarimetric covariance-based FBR framework.

Through selected examples, this presentation highlights the potential of this emerging commercial SAR dataset and discusses its complementarity with existing Earth Observation missions. The objective is to explore how new-generation commercial SAR systems can contribute to operational monitoring frameworks and support the growing demand for timely, reliable geospatial intelligence.

The growing exposure of critical infrastructure to hydrologically driven geomorphic hazards necessitates monitoring approaches that are both reliable and non-intrusive. Interferometric Synthetic Aperture Radar (InSAR), particularly Persistent Scatterer InSAR (PS-InSAR), has proven effective in capturing millimetric ground deformation over large areas, making it well suited for detecting early signs of instability. In this study, a multi-frequency SAR framework is developed by integrating PS-InSAR deformation analysis with SAR-based soil moisture retrieval to enhance infrastructure monitoring within a Geospatial Intelligence (GEOINT) context. The approach is applied to the Petacciato landslide in Italy, a slow-moving, rainfall-sensitive slope affecting nearby buildings, roads, and transport infrastructure.

Soil moisture is retrieved using the first-order radiative transfer model (RT1) applied to SAR datasets from SAOCOM (L-band), Sentinel-1 (C-band), and COSMO-SkyMed (X-band). Long-term L-band retrievals (2021–2023) show strong agreement with reference products such as ASCAT, with correlations exceeding r ≥ 0.67. Further validation using in situ measurements collected between March and September 2025 confirms the superior performance of L-band (r = 0.76), while Bayesian fusion of L- and C-band data improves the robustness of the estimates and enables uncertainty quantification. These results highlight the enhanced sensitivity of L-band SAR to soil moisture dynamics, particularly under vegetated and heterogeneous surface conditions.

PS-InSAR time series spanning 2011–2025 reveal spatially variable deformation, with line-of-sight velocities ranging from −10 to −40 mm/year in the most active sectors. L-band observations increase the density of persistent scatterers in vegetated areas, whereas X-band demonstrates higher sensitivity in built-up environments. The application of Sequential Turning Point Detection (STPD) identifies key phases of deformation change, which exhibit a clear temporal correspondence with antecedent wetness conditions derived from the Antecedent Precipitation Index (API). Periods of accelerated movement are consistently preceded by elevated API values, reinforcing the role of hydrological forcing in landslide reactivation.

These components are integrated within the PS–SMaRT framework, which combines deformation and soil moisture indicators to detect unstable zones and derive hazard indices. By incorporating slope-projected kinematics, density-based clustering, and statistical evaluation, the framework produces spatially explicit hazard maps.

From a GEOINT perspective, this approach enables continuous, wide-area monitoring of infrastructure without reliance on ground access. Linking deformation signals with moisture conditions improves the interpretation of emerging instabilities and provides a robust basis for hazard assessment and risk-informed decision-making in sensitive or inaccessible regions.

KEYWORDS: Interferometric Synthetic Aperture Radar (InSAR), Persistent Scatterer InSAR (PS-InSAR), Multi-Frequency SAR, Soil Moisture Retrieval, Radiative Transfer Modelling (RT1), Ground Deformation Monitoring, Critical Infrastructure Monitoring, Landslide Hazard Assessment, Mass Movement, Geospatial Intelligence (GEOINT)

The Nile Valley and Delta are among the most densely populated regions, home to more than 95% of Egypt's population and hosting critical infrastructure, the majority of big cities, agricultural systems, and coastal economic hubs. Monitoring land subsidence in this region and understanding the driving factors is therefore essential to reduce risks to infrastructure fracturing, flooding, water abstraction , and excessive loading due to urban development.

This research introduces, for the first time, comprehensive land subsidence mapping of the Nile Valley in Egypt and its surroundings. Using 14 Sentinel-1 ascending frames covering nearly 1,000 km along the Nile corridor and an area of approximately 375,000 km², we generated detailed InSAR surface deformation time series covering November 2017 and July 2024 from about 15,000 interferograms processed with GMTSAR. The results show that while most of the Nile Valley is stable, several governorates face both widespread and localized subsidence, with rates sometimes exceeding –12 mm/year, indicating a clear need for mitigation.

To investigate the causes, mean and 5th percentile velocities were analyzed and combined with 10 key factors: clay content, precipitation, faults, daytime and nighttime temperature, geology and soil outcrops, Land Use and Land Cover (LULC), urban load, Normalized Difference Vegetation Index (NDVI), and water content. Deep learning modeling over five targeted regions reliably captured the magnitude and spatial patterns of deformation, with PSNR rising up to 36 dB and SSIM reaching 1.0, especially in urbanized zones like Port Said and Alexandria.

SHAP analysis shows the main drivers of subsidence vary from region to region: water soil content and urban load dominate in the Delta region, while geology, faulting, and soils are found to be more critical further south. The innovative integration of InSAR, deep learning, and SHAP analysis provides decision-makers with valuable insights to identify risk zones and understand regional subsidence processes, offering essential support for targeted early warning and mitigation actions across the Nile Valley and Delta.

Persistent Scatterer Interferometry (PSI) and other multi-temporal InSAR techniques have become key tools for monitoring surface deformation and assessing the stability of critical infrastructures. However, transforming large volumes of SAR-derived deformation measurements into actionable information for infrastructure monitoring remains challenging, particularly when linking geoscientific observations with asset-level information.

The GIGASAR project addresses this challenge by integrating advanced multi-SAR deformation processing with infrastructure-oriented analytics and digital building models. The approach combines deformation time series derived from multi-temporal SAR analysis with Building Information Modelling (BIM) data to enable object-level interpretation of ground motion affecting infrastructure assets. SAR-derived deformation points are automatically associated with BIM elements through spatial mapping and clustering techniques, enabling the derivation of infrastructure-specific indicators and structural monitoring metrics related to ground deformation.

A key aspect of the project is the integration of multi-source SAR datasets to improve spatial detail, temporal continuity, and the robustness of deformation analysis. This supports a more reliable interpretation of infrastructure-related motion in complex urban and peri-urban environments. Beyond the analytical workflow, GIGASAR implements a web-based geospatial platform for the interactive exploration and analysis of SAR deformation products together with BIM models and auxiliary geospatial datasets. This integration improves situational awareness for infrastructure operators and supports maintenance planning and risk assessment.

The presentation will describe the overall system architecture and processing workflow, including the integration of multi-source SAR datasets, the PSI–BIM linkage methodology, and the platform-based visualisation and analytics environment. It will demonstrate how advanced InSAR processing combined with digital infrastructure models can enhance the operational exploitation of SAR data for monitoring critical infrastructures.

Accurate crop type mapping is critical for agricultural monitoring and food security assessment at regional scales. This study presents a multi-sensor deep learning framework for crop classification using Sentinel-1 SAR and Sentinel-2 optical time-series data over Perambalur District, Tamil Nadu, India. An agricultural mask derived from land-use information and manual digitization was applied to restrict analysis to cultivated areas. Ground truth crop samples were collected as field polygons across multiple seasons to capture phenological variability. Multi-temporal Sentinel-1 data from July 2024 to February 2025 were processed to derive polarization-based features, backscatter combinations, and radar vegetation indices, while cloud-free Sentinel-2 observations were used to compute spectral bands and vegetation indices. All features were temporally stacked to construct a unified multi-sensor dataset. Several deep learning models were evaluated, including 3D convolutional neural networks, recurrent neural networks with Bidirectional Long Short-Term Memory, ConvLSTM, a fully connected Dense Neural Network, and a transformer-based Temporal Vision Transformer. Results show that models effectively leveraging phenological information achieve superior performance. Notably, the Dense Neural Network yielded the highest classification accuracy, demonstrating that carefully engineered multi-sensor temporal features can outperform more complex spatio-temporal architectures under cloud-constrained conditions.