Conference Agenda

Global surface soil moisture (SSM) mapping at a 9-km intermediate resolution from microwave remote sensing could play a pivotal role in advancing detailed global hydrological investigations. Despite the wide recognition of the Soil Moisture Active Passive (SMAP) mission’s 9-km SSM products based on oversampling of the L-band radiometry, concerns persist regarding its ability to capture higher SSM heterogeneity at finer resolutions. For addressing this concern, a novel methodological framework was proposed in this study. This framework advocates the downscaling of the SMAP 36-km dataset through a fusion with high-frequency (Ka-band) passive microwave observations. The resultant 9-km all-weather SSM product, derived from this novel approach, is evaluated using ground-based measurements worldwide. The findings reveal a significantly enhanced accuracy compared to the SMAP conventional oversampling-based one, especially in areas exhibiting pronounced local variations in SSM patterns. The study therefore represents a substantial step forward, providing new insights into the design and use of a multi-frequency satellite radiometer for global SSM mapping.

Glaciers in the Pamir Mountains play a crucial role in the regional water supply and the seasonal buffering capacity especially in warm and dry conditions; therefore, it is highly important for understanding the response of glaciers to climate change in High Mountain Asia. However, due to the climatic and glaciological complexity of the region, more precise spatiotemporal analysis is required. Challenges include sparse in situ data, highly variable climate conditions with an East-West gradient of temperature and precipitation, and complex glacier processes including debris cover, surging behavior, and collapse features. Previous geodetic mass balance estimates, including global assessments derived by ASTER datasets, have suggested increasing mass loss since the early 2000s; however, the results are inconsistent due to high uncertainty in measuring the behavior of smaller glaciers and mass balance with shorter time resolution. In this study, we present a time-series analysis of geodetic data from 2003 to 2024 using high-resolution (<5m) stereo satellite imagery independent of ASTER data to determine a tendency towards mass balances in the 21st century. This study focuses on a set of 41 glaciers in the western Pamir as a case study and provides mass balance trends with improved accuracy at a time resolution of up to 5 years before 2019 and up to 1 year after that.

Using ALOS, SPOT5, SPOT6, and Pléiades stereo satellite images acquired since the 2000s, we adopted a stereo image processing workflow from the Ames Stereo Pipeline, and generated DEMs from these images using the More-Global-Matching stereo processing method. Elevation changes were derived by subtracting the coregistered and bias-corrected DEMs and converted to elevation change rates. The bias correction includes removing erroneous artifacts such as jitter-induced undulations using Fourier transforms and direct subtraction of the averaged error in the along and cross-track direction. We evaluate the elevation change uncertainty based on the modeling of the heteroscedasticity and the spatial correlation in elevation errors and propagate the errors to each glacier scale and regionally. Finally, we quantified the glacier mass balance in the study area over a 20-year period, and the results over the region were generally consistent with the ASTER data set from 2003 to 2019, and showed an accelerating decrease from 2019 to 2024. Our results demonstrate the potential of very high resolution satellite imagery for evaluating the spatial consistency of temporal variations in the mass balance of individual glaciers and entire regions, despite the challenge of short-interval observations, and highlight the value of multiple independent high-precision geodetic mass balance estimates for resolving glacial changes at shorter time resolutions.

Mountain glaciers are sensitive indicators of climate change and can respond rapidly to climatic variations. In contrast to most glaciers worldwide, glaciers in the Pamir region are a notable exception, showing limited mass loss during the early 21st century. The past decade, however, has been the warmest on record, with 2024 marking the first year that global average temperatures exceeded 1.5 °C above pre-industrial levels, surpassing the Paris Agreement target. As a consequence, Pamir region has been fast warm and dry since 2017. The local meteorology relevant to the glacier domain is not only crucial to capture glaciological processes, but also plays a key role in buffering local climate change. Yet, due to the inaccessibility of high mountain regions and the challenges of maintaining instrumentation, in-situ observations remain sparse, limiting our understanding of glaciers' response to intensified warming in this unique region.

Fortunately, a recently developed inversion approach, which couples physical glacier modeling with remote sensing data, now offers a promising approach to retrieve meteorological information relevant to the glacier scale. This approach can reconstruct glacier meteorology and mass balance without any in-situ data. In this study, we leverage satellite-derived glacier albedo and surface temperature from 2017 to 2024 to assess the feasibility of the inversion method for obtaining three key meteorological variables (air temperature, precipitation, and incoming shortwave radiation) at annual scale. This assessment is conducted on six benchmark glaciers with available in-situ measurements (Abramov, Sangvor, Zulmart, G457, Yakarcha, and GGP Glaciers). We analyze annual anomalies in these variables and in glacier mass balance by physical modelling driving by this improved meteorology. Our results show that the inversion approach reliably captures annual meteorology and mass balance variations. Notably, glacier air temperature anomalies are lower than the global average, while precipitation increases and incoming shortwave radiation decreases. Despite this, glacier mass loss accelerates significantly under a 1.5 °C warming scenario. These findings offer critical insights for climate policy and underscore the utility of the inversion method for assessing glacier responses to extreme warming, potentially at global scale.

Crop classification in cloudy and rainy regions faces the dual challenges of insufficient high-quality remote sensing images and limited samples. To address this, we propose a method that integrates optical and synthetic aperture radar (SAR) data to generate spatiotemporal samples. This approach enables the estimation of crop distribution in other years based on field survey samples from a single year. The method reconstructs vegetation index time series through pixel-level optical-SAR fusion and identifies crop types of generated spatiotemporal samples using time-weighted dynamic time warping and sample refinement. These samples support rapid and large-scale crop classification, achieving an accuracy of 67%-82% with random forest. Compared to methods relying solely on optical satellite data, the proposed approach demonstrates the potential of SAR in addressing data gaps. The results highlight the potential of training machine learning with spatiotemporal samples generated from optical-SAR fusion to enhance crop classification accuracy in similar climates.

Water resources in semi-arid areas are limited, and socioeconomic developments mainly depend on groundwater withdrawals for irrigation agricultural activities. Increased groundwater extraction leads to aquifer depletion and subsidence occurrences, causing a gap in driving the mechanisms, relationships, and correlations between the groundwater extraction across the hydrological Units, and land subsidence remains undistinguishable. The study aims at bridging this gap by estimating the groundwater depth and its relationships towards the spatial-temporal subsidence pattern characteristics in the Limpopo watershed using the river basin scale model developed to quantify the impact of land management practices in large, complex watersheds (SWAT) and Small Baseline Subset Interferometric Synthetic Aperture Radar (SBAS-InSAR) techniques algorithms. Multi-source datasets were used for the modeling, including climate data, soil, land use land cover, Shuttle Radar Topography Mission Digital Elevation Model, and Sentinel-1 data from 2022 to 2024. The results show that land subsidence occurs in the areas with intensive agricultural pumping activities (4m to 27m as the average depths) and around the settlement areas where the subsidence ranges from -10 cm to +10cm per year. This is because the water demand is too high. The correlation and extent between the substance and groundwater extraction were computed by the regression coefficients that vary significantly, influenced by different transitions in hydrological units (R² = 0.85), revealing that groundwater decline is the primary major in the subsidence.

In conclusion, the findings highlight the insufficient water resources to meet agricultural demands, resulting in continuing extraction from stagnant groundwater supplies and increased surface subsidence. This situation needs immediate and more effective groundwater management techniques to prevent additional ground subsidence.

Keywords: SBAS-InSAR, Agricultural Irrigation Landscapes, Groundwater Depth, and Hydrological Units

Accurate characterization of soil moisture (SM) and evapotranspiration (ET) is crucial for water resources assessment and climate impact studies. While land surface models (LSMs) provide a powerful tool for estimating these variables, their accuracy is often limited by uncertainties in model structure, parameterization, forcing and input data. Data assimilation (DA) has been shown to effectively reduce these uncertainties by integrating observational data into the modelling framework. In this work, we apply the Ensemble Smoother with Multiple Data Assimilation (ES-MDA) method to estimate sensitive vegetation and soil parameters within the European Community Land Model (eCLM). By assimilating Soil Moisture Active Passive (SMAP) SM data and ET measurements from Integrated Carbon Observation System (ICOS) sites, we refine pedotransfer function parameters linking soil texture to hydraulic properties, as well as key plant physiological parameters governing processes like photosynthesis and stomatal conductance. The estimated plant physiological parameters are upscaled in a regionalization approach. The optimized parameters are then used to generate SM and ET datasets at 12 km spatial resolution and daily temporal resolution over the EURO-CORDEX domain for the period 2000–2022.

Validation against the International Soil Moisture Network (ISMN) across 300 sites indicates notable improvements: the ubRMSE decreases from 0.058 to 0.055 cm³/cm³, median bias decreases from 0.057 to -0.012 cm³/cm³. The bias reduction indicates that the updated parameters help mitigate systematic model errors, such as the overestimation of SM that is reported in many studies. Additionally, the SM and ET are compared with ERA-Land reanalysis data, ESA Climate Change Initiative (CCI) SM products, and GLEAM ET products. The comparison also indicates consistent improvements in capturing spatial patterns. Further assessments include comparisons of simulated runoff and total water storage (TWS) against E-RUN and GRACE satellite observations, respectively. However, these evaluations indicate that DA does not consistently translate to better performance for all hydrological states and fluxes.

Satellite altimetry has been the major data source for marine geoid determination and gravity recovery in recent decades. In general, altimetry-derived geoid and gravity anomaly models are typically released with a 1'×1' gridding interval. However, their actual spatial resolution is far lower than the nominal ∼2 km level. Therefore, analyzing the marine geoid resolution capability from satellite altimetry observations is crucial for marine gravity recovery studies. The Surface Water and Ocean Topography (SWOT) Mission is a newly launched satellite using advanced radar technology to make headway in observing the variability of water surface elevations, providing new information through along-track and across-track two-dimensional swath observations. Here, we present the analysis results of marine geoid resolution capability for both typical conventional nadir altimeters (Ka&Ku&C band, LRM&SAR mode) and the SWOT Ka-band radar interferometer (KaRIn) in 2°×2° bins worldwide between 60°N and 60°S. We demonstrate the potential of SWOT KaRIn to capture along-track short-wavelength signals below 10 km and analyze the bin-based statistics of key marine geophysical factors correlated with this marine geoid resolution capability. Generally, SWOT KaRIn exhibits better marine geoid resolution capability over bins with large-scale seamounts or trenches.

Scatterometers are vital remote sensing instruments for observing ocean surface wind fields. In China, the majority of operational scatterometers adopt one of two systems: the rotating pencil-beam system (e.g., HY-2 HSCAT) and the rotating fan-beam system (e.g., CFOSAT CSCAT and FY-3 WindRAD). In this study, we provide a summary of the operating characteristics of the two systems with reference to the design parameters of the HSCAT scatterometer and the CSCAT system. Thereafter, we conduct a detailed evaluation of the observational performance of the scatterometer for varying wind speeds and assess the observational capability of the two scatterometer systems for tropical cyclones (TC) based on reanalysis data.

The study commences with a statistical analysis of the number of measurement samples N_sample, and the normalized signal-to-noise ratio SNR’, for both type of systems, utilizing the orbital data obtained from the simulation. The N_sample and the SNR’ are dependent solely on the parameters of the scatterometer system and have a substantial impact on the accuracy of the scatterometer's backscattering coefficient measurements. The transfer error Kpc of backscattering coefficient measurement can be used as a direct reflection of the impact of the scatterometer system design on measurement accuracy. The performances of the scatterometer are then discussed with this general metrics. In order to evaluate the accuracy of the backscattering coefficient measurement and the performance of wind field inversion for the two different systems, simulated wind fields at 4 m/s, 18 m/s, and 30 m/s.

A preliminary evaluation of the TC observation performance of the two scatterometer systems, based on the reanalyzed data, indicates that the rotating fan-beam system is more effective for observing the eyewall of TCs and the surrounding high wind regions, whereas the rotating pencil-beam system is more suitable for monitoring low wind structures, such as eyewalls. Moreover, the rotating fan-beam system demonstrates superior consistency in the observation of high wind speeds above 20 m/s, while the rotating pencil-beam system exhibits enhanced consistency in the observation of lower wind speeds below 14 m/s. Further discussion on fixed fan-beams are next steps.

Lidars provide high-precision vertical profiles and clearly defined cloud top height (CTH), as the upper geometrical boundary of the highest atmospheric layer containing cloud particles. In contrast, passive remote sensors such as satellite imagers estimate CTH indirectly by measuring reflected or emitted radiation. Common methods include the Infrared Brightness Temperature Method, CO₂ Slicing Technique, and Water Vapor Channel Method. The problem is that CTh retrievals from passive sensors often become ambiguous in complex atmospheric conditions, such as optically thin cirrus, multi-layered clouds, or low clouds under polar or nighttime conditions. Challenges include difficulty in detecting semi-transparent clouds, separating overlapping layers, and distinguishing low clouds from cold surface emissions, which causes the retrieved CTH to deviate from its supposed definition. Aerosol layers, both near the surface and in the free atmosphere, add further complexity and can introduce significant errors.

Spaceborne lidars, with their precise vertical profiling and layer detection capabilities, offer opportunities to assess the accuracy of passive CTH retrievals. By comparing lidar-based and radiometer-based CTH measurements, we can better understand the limitations of passive methods and their impact on cloud characterization, i.e., what we have actually been obtaining and how it influences our knowledge of CTH.

This study analyzes CTH data from CALIPSO/CALIOP, DQ-1/ACDL, and EarthCARE/ATLID lidars, along with passive retrievals from geostationary imager FY-4B/AGRI. The collaborative observations are categorized into different scenarios to analyse the major error sources in passive CTH retrievals, and implications are discussed in terms of global cloud datasets. In addition, with comparison to EarthCARE combined lidar and radar products, limitations on observations by lidar itself will be discussed.

Aerosols contribute to one of the greatest uncertainties in global radiative forcing of the changing climate. They reflect, scatter and absorb radiation, and can act as cloud condensation nuclei to form clouds and modulate cloud properties which consequently affect radiation balance. Therefore, it is important to obtain a thorough understanding of the global aerosol distribution and population to better predict climate change. The sun-synchronous EarthCARE (Earth Cloud, Aerosol and Radiation Explorer) satellite has been operating since May 2024, providing extensive aerosol and cloud measurements. The ATmospheric LIDar (ATLID) lidar onboard EarthCARE efficiently measures aerosol profiles at a 355 nm wavelength, while the Target Classification (TC) product provides valuable information about cloud and aerosol (sub-)types per layer. Their combination may potentially reveal how aerosol of different compositions vary with altitude globally. However, both products are at their preliminary stages and require validation and calibration.

In our study, we will validate the observations (e.g. extinction coefficient and lidar ratio) from ATLID and the TC products with the EARLINET measurements (A European Aerosol Research Lidar Network to Establish an Aerosol Climatology). We will additionally derive aerosol optical depth using the profiles of extinction coefficient and compare it with the CAMS (Copernicus Atmosphere Monitoring Service) forecast. CAMS forecast provides high-quality data by incorporating observations, which enhances the accuracy of atmospheric composition estimates. Co-located comparisons from all three datasets will improve our understanding of satellite retrieval quality and vertical distributions worldwide. They will also provide insights into lidar measurements and support improvements in global models for more accurate climate change predictions.

The geological conditions in the alpine gorge regions of western China are complex, with widespread and frequent disasters that result in hundreds to thousands of deaths and billions of yuan in direct economic losses annually. Currently, nearly 300,000 potential geological hazard sites have been engagement identified in China, yet over 70% of the geological hazards that lead to disastrous consequences occur outside these known potential hazard areas. Therefore, conducting early and precise identification of large-scale landslide hazards is of great significance for enhancing China's geological disaster prevention and control capabilities. Interferometric Synthetic Aperture Radar (InSAR), due to its characteristics of large coverage, all-weather measurement, and non-contact measurement, has been widely used and valued in the early identification and monitoring of landslide hazards. However, in engineering applications of early landslide hazard identification using InSAR technology, there are issues such as geometric distortions caused by SAR satellite oblique viewing that are unclear in terms of how to accurately identify them on a large scale and their impact on InSAR monitoring, as well as the quantitative relationship between detected Line of Sight (LOS) deformation and true deformation, and the unclear removal methods for atmosphere-related and spatially heterogeneous atmospheric effects caused by unique topographies in alpine gorge regions when external data are not available.

This study reviews the current application status of InSAR technology in the early identification and monitoring of landslide hazards, and clearly and innovatively addresses the key issues in its engineering applications, such as limitations in spatial detection capabilities, LOS detection sensitivity, and atmospheric correction methods in mountainous areas. It also summarizes the application characteristics and future prospects of InSAR technology, which is of great significance for effectively conducting engineering applications of InSAR technology in geological disaster prevention and control.

Sinkhole is a depression and a pervasive geohazard in karst terrains, where the progressive dissolution of carbonate bedrock leads to subsurface void formation and eventual ground collapse. These phenomena pose substantial risks to urban infrastructure, as their concealed development often precedes sudden surface subsidence, resulting in structural and settlement damage. The Gharbanyiat area located in the western part of Borg Al Arab region exemplifies such challenges, with documented cases of ongoing cavity propagation beneath populated areas. In this study, we employ Multi-Temporal Interferometric Synthetic Aperture Radar (MT-InSAR) techniques to monitor and analyze surface deformation associated with sinkhole development. Furthermore, we integrate field data to frame InSAR-derived deformation within the region’s hydrogeological framework. Results demonstrate that areas exhibiting the highest subsidence rates correlate with zones of unconsolidated sediments overlying carbonate rocks, a configuration prone to exposure-collapse sinkholes. Using Sentinel-1 SAR imagery, we detect and quantify subtle ground displacements, revealing a direct correlation between subsidence patterns and active sinkhole formation. Time-series analysis reveals localized subsidence patterns aligned with known sinkhole occurrences, validating the technique's sensitivity to incipient karst-induced instability. Our findings underscore MT-InSAR’s utility as a proactive monitoring tool for sinkhole risk assessment in urbanized karst environments. By quantifying deformation dynamics and spatially constraining vulnerable areas, this approach supports infrastructure resilience planning and early-warning strategies.

Multi-temporal synthetic aperture radar interferometry (SAR, MT-InSAR) has been widely recognized as an effective technique for monitoring surface deformation and marking a significant advancement in satellite geodesy to millimeter-level precision. As one of the most representative MT-InSAR methods, persistent scatters interferometry (PS, PSI) focuses on the most elite pixels over the temporal and spatial scales of SAR images. The selection of PS candidates is the cornerstone of the excellent performance of PSI, directly influencing the accuracy and density of surface deformation products. Most traditional methods employ particular metrics and thresholds to discriminate between PS and non-PS pixels. Their results are driven by the employed metric, like the mean coherence for distributed targets or the dispersion of amplitude for deterministic ones, but there can be stable pixels not detectable by any of the metrics. Benefiting from the development of deep learning, data-driven methods have been widely proposed in recent years and exhibit superior efficiency. However, existing approaches do not fully exploit the contextual relationships between phase, amplitude, time, and spatial dimensions. This will result in a set of selected PS points that are representative only in certain dimensions.

Therefore, a novel deep learning method for persistent scatters selection that leverages the temporal-spatial context features of amplitude images and interferometric phase has been proposed. Specifically, a temporal-spatial vision transformer (TS-ViT) architecture is employed to process input amplitude and phase time-series stacks simultaneously. A positional embedding method based on InSAR temporal-spatial baselines is proposed, which enables TS-ViT to learn the PS distribution patterns of different interferogram combinations. In the backbone, the positional information is incorporated into the image tokens via the temporal and spatial embedding layers, and the local features in the context of the time series images are derived by the temporal and spatial encoder. Then, inspired by the idea of temporal phase coherence estimation, a phase consistency attention mechanism (PCAM) model is used in transformers to improve the model's perception of stable phase noise in the complex domain. Furthermore, an amplitude and phase interaction (API) module is proposed to achieve multi-scale fusion of amplitude and phase features. Finally, the feature map is output to the decoding head, and the high-quality PS points are predicted pixel by pixel through a multilayer perceptron.

The proposed model was trained on a dataset containing time-series SAR amplitude images and interferometric phase stacks of Barcelona, acquired by TerraSAR-X between 2009 and 2011. The dataset includes 8,689 samples for training and 965 samples for validation, with data pre-processing and PS annotation performed using the SUBSIDENCE software from the Universitat Politècnica de Catalunya. To address the class imbalance between PS and non-PS points, the focal loss function was employed. The proposed model was evaluated using metrics like intersection over union (IoU), accuracy, Dice coefficient, F1-score, precision, and recall.

The results shows that our method achieves good results in urban areas with the presence of tall buildings. Compared to other deep learning methods, the mean IoU and mean Dice coefficient improved up to 9.1% and 8.43% respectively. Compared to the most widely used method, the new method extracts 67.78% more points with high spatial coherence. Additionally, the proposed method performs robustly across diverse land types and is extendable to distributed scatterer (DS) pixel selection. All model codes and training configurations will be available at https://dagshub.com/zhangyfcsu/pssformer.

BIOMASS is ESA’s 7^th Earth Explorer mission, selected for implementation in 2012 and is foreseen for launch in April 2025. BIOMASS is implemented as a single platform repeat pass interferometric P-band SAR mission. The instrument operates in a simple Stripmap mode, acquiring global data in a predefined orbit scenario.

The BIOMASS operation scenario foresees to fly the satellite in a drifting repeat pass orbit of 3 days. Spatial baselines are defined to enable the interferometric processing of data. Besides supporting the generation of primary mission products (i.e., above-ground biomass, forest height), the interferometric acquisitions can also be used to estimate the ground topography. While today several global surface models (DSM) such as SRTM, GTOPO, AW3D30, and most prominent, the Tandem-X and Copernicus DSM exist, the use of P-band will allow for the first time to estimate the true ground topography below the vegetation and specifically below the forest cover at a spatial resolution in the 50 – 200 m range. The systematic acquisition with a repeated global coverage on both the ascending and the descending passes and the baseline variety of the system will enable a flexible selection of geometries and baselines to optimize the topographic processing of data.

The generation of BIOMASS products, including DTM and forest maps, requires starting from coregistered and calibrated stacks so that main disturbances affecting data have been removed. These include orbital inaccuracies, ionosphere, and troposphere. This work reviews the interferometric processor required to estimate and correct the abovementioned effects. The main topics addressed are:

• Split-spectrum for low-varying ionosphere removal
• Multi-squint interferometry for fast-varying ionosphere and orbital error removal
• The sum of Kronecker Products (SKP) decomposition of ground and volume contributions

Apart from presenting the functionalities of each block, we stress the importance of adjusting the implementation to the chain configuration so that each step operates correctly even after previous blocks were applied. The correctness of the processor is then demonstrated by a stack of simulated data injected by the different disturbances as well as a realistic DTM error.

Tropical forests are vital ecosystems that play a crucial role in climate regulation, biodiversity preservation, and carbon sequestration. However, these ecosystems are increasingly threatened by human activities such as selective logging. Although more sustainable than clear-cutting, selective logging can still cause significant environmental impacts if not carefully managed. Techniques like Reduced Impact Logging (RIL), under the framework of Sustainable Forest Management (SFM), aim to mitigate these effects. Nonetheless, efficient and scalable monitoring tools are essential to evaluate logging outcomes and inform forest management strategies.

This study presents a cost-effective and scalable methodology for monitoring selective logging impacts in tropical forests using Sentinel-1 Synthetic Aperture Radar (SAR) data. In contrast to expensive technologies like airborne LiDAR, Sentinel-1 SAR is freely available and operates independently of weather and lighting conditions, making it especially suitable for vast and remote areas such as the Amazon.

The approach was tested in Brazil’s Jamari National Forest across three areas: two that were selectively logged and one undisturbed control site. Logging occurred in 2020 and 2017, affecting areas of 432.52 ha and 448.89 ha, respectively, while the 614.20 ha control area remained untouched. Field data collection involved GNSS-based mapping of logging infrastructure and road width sampling. Sentinel-1 SAR data (VV polarization) were obtained for both pre- and post-logging periods. After standard preprocessing, including radiometric calibration, speckle filtering, and terrain correction, backscatter change detection was performed to assess canopy alterations.

Results indicated canopy losses of 6.90% and 4.11% in the logged areas, while the control area experienced only 0.16% change. Spatial overlap between detected changes and mapped infrastructure reached 60.94% and 51.65%, confirming the accuracy of the method.

These findings demonstrate the potential of Sentinel-1 SAR as a practical alternative to high-cost remote sensing technologies for monitoring selective logging in tropical forests. While effective in detecting moderate disturbances, the method's sensitivity to low-intensity impacts remains limited. Future work will focus on integrating artificial intelligence to enhance detection accuracy and on extending the methodology to other forest regions, contributing to global forest conservation and sustainable management efforts.

Salt mining has long been a key socioeconomic activity in Maceió, Brazil, where extensive extraction from underground salt deposits has caused increasing geological instability. Since the 1970s, over 35 wells have been developed in three urban neighborhoods, leading to widespread structural damage—including fissures in streets, cracks in buildings, and ground depressions. These risks were amplified following a 2018 seismic event and culminated in the collapse of Mine 18 on 10 December 2023, highlighting the urgent need for robust monitoring tools.

This study applies the Quasi-Persistent Scatterer Interferometric Synthetic Aperture Radar (QPS-InSAR) technique to analyze ground deformation in Maceió using 145 Sentinel-1A images acquired between June 2019 and April 2024. Initial analysis employed Persistent Scatterer Interferometry (PSI) to estimate linear deformation velocity and target height. The workflow included: (1) image import and selection of reference scenes, (2) preliminary assessment using reflectivity maps and Amplitude Stability Index (ASI), (3) interferometric processing for coherence and interferogram generation, and (4) multi-temporal InSAR processing with Atmospheric Phase Screen (APS) compensation. A stable point network was first selected using ASI > 0.7, followed by APS estimation from residuals and a threshold reduction to ASI > 0.6 to enhance point density.

While PSI revealed major deformation zones around the mines, its effectiveness was limited near critical areas due to low coherence and steep displacement gradients. To overcome these challenges, QPS-InSAR with “Small Area Processing” was applied, selecting scatterers based on reflectivity (>0.5) and applying coherence-weighted estimation. This resulted in 23,460 scatterer points at a density of 1303 points/km²—a 422% increase over PSI-derived results. Detected cumulative displacements ranged from −1750 mm to +10 mm, with significant subsidence (−1200 mm) near Mine 18.

In addition to phase-based measurements, amplitude data were analyzed to assess changes in surface reflectivity as a proxy for instability. A marked reduction in radar return was observed in late 2023, particularly around Mine 18. Notably, 17 scatterers lost amplitude stability in 2023—11 of them in December—coinciding with the timing of the collapse. Amplitude time series revealed that this decline began approximately two weeks prior to the event, suggesting that QPS-InSAR can provide early warning signals of catastrophic ground failure.

The findings confirm the effectiveness of QPS-InSAR in detecting and characterizing complex deformation patterns in low-coherence environments. This approach offers a powerful tool for disaster risk mitigation and urban planning in geologically vulnerable regions. Future efforts should integrate satellite-based InSAR with in-situ data to improve the accuracy of predictive models and support timely interventions in at-risk communities.

This study utilizes the radio occultation observations from the Macau Science Satellite-1 (MSS-1) to investigate the ionospheric response to the May 2024 G5 geomagnetic storm within the South Atlantic Anomaly (SAA) region. The distinctive data from MSS-1, complemented by ground-based Global Navigation Satellite System (GNSS) and the Constellation Observing System for Meteorology, Ionosphere, and Climate follow-on mission (COSMIC-2) observations, reveals a super plasma fountain effect during the main phase of the storm. This effect is marked by peaks in the equatorial ionization anomaly extending beyond their typical latitude range. The MSS-1 observations, particularly in the northern hemisphere of the SAA region, confirm the role of prompt penetration electric fields in driving ionospheric disturbances and amplifying scintillation at higher altitudes. The study also identifies a decrease in total electron content and a reduction in scintillation occurrence during the recovery phase of the storm. The results demonstrate the pivotal role of MSS-1 combined with ground-based and COSMIC-2 observations, providing a comprehensive understanding of the ionospheric response to severe geomagnetic storms.

The main goal of this work is to map ancient water management features in archaeological landscapes of southern Iraq to generate a robust satellite-based validation layer to test the novel artificial intelligence and deep learning methods that are being developed within the Dragon-6 STAI4CH project and will be applied to the waterscapes of the UNESCO World heritage site of the Ahwar of southern Iraq: refuge of biodiversity and the relict landscape of the Mesopotamian cities. The work stems from the bilateral CNR/Royal Society of the UK research collaboration project “Vanishing archaeological landscapes under anthropogenic and climate change threats”, that the European project team of STAI4CH successfully delivered in 2023-2024, with further collaboration from the University of Al-Qadisiyah of Iraq for field surveying, and UNOSAT for high-resolution image interpretation.

Satellite remote sensing data are increasingly being used to identify and record archaeological sites and traditional landscapes. Given the wealth of high-resolution imagery that is now available, from declassified historical imagery to commercial multispectral data, it is possible to analyse how these locations have evolved. This study focuses on the development of irrigation from ancient times to the present day in an area of southern Iraq bounded between the western desert and the modern course of the Euphrates, and the remaining areas of the marshes to the south. A particular emphasis is put on the spring of Ayn Sayed and nearby irrigation systems, including features pre-dating the 1st millennium AD, and gravity-flow canal systems originating in the Sasanian (224-651) AD period drawing from both river and spring abstraction points. Some of these systems were in use until very recently and a few areas still intermittently are cultivated using them. Using HEXAGON imagery (1974-1981), the layout of the irrigation systems was recorded prior to more recent intensification of agriculture. The HEXAGON data were orthorectified using ERDAS with RMSEs of 1.4-2.9 m. With a resolution of up to 0.6 m, fine details of features could be mapped. A standardised image interpretation approach was developed to separate the systems into ancient canals, historical canals, 20th century canals, and modern canals; the mapping of these canals provides a proxy for field system locations. These systems are now at risk due to climate change and modern development. While traditional canals may offer a sustainable alternative to modern pumping methods in the face of climate change, socio-environmental issues pose a threat to them. An analysis of Landsat (from 1985) and Sentinel-2 (from 2018) obtained via Google Earth Engine allowed the impact of the recent changes on these systems to be mapped and measured. Using the 6 bands of Pléiades Neo imagery purchased from Airbus DS and processed in ArcGIS Pro to pansharpen them to the 0.3 m resolution of the panchromatic band, along with the sequences of imagery available on Google Earth, it was also possible to identify the land use activities which are involved, for example the production of bricks.

The assimilation of wave data from swath of scatterometters missions and directional wave spectra from Sentinel-1 and CFOSAT missions has significantly improved sea state forecasting and our understanding of wave climatology over global ocean, particularly in critical regions such as marginal ice zones. The neural network algorithms for retrieving wave data have been upgraded through the use of reprocessed data from HY2 and wave spectra from CFOSAT (IPF-7.1), which detects waves with wavelengths exceeding 500 meters.

The objective of this work is to evaluate the impact of new directional wave spectra provided by Sentinel-1C in the frame of the assimilation of multi-missions observations. We have analyzed and assessed the impact of assimilating multi-mission significant wave height data from HY2B, 2C, 2D, and directional wave spectra from SWIM, S1A, and S1C. The model simulations are performed for a recent time period from April to June 2025. Validation of the simulations, both with and without assimilation, is conducted using independent altimeters wave data and observations from drifting buoys collected during field campaigns at sea.

We have also examined the impact of data assimilation on extreme waves and the forecasting of maximum wave heights in regions affected by strong ship navigation route. These preliminary results have been developed by starting junior scientists in both chinese and french teams.

With the continuous development of society, there is an increasing emphasis on the quality of wheat grains. Wheat grain protein content (GPC) is one of the key indicators used to assess wheat grain quality, and remote sensing-based predictions of GPC have significant practical relevance. However, the large-scale cultivation of wheat, coupled with China's diverse topography and complex climatic conditions, creates challenges such as time lags, abrupt changes, and overlaps in the relationship between wheat quality and spatial data, resulting in severe spatiotemporal biases in large-scale predictions of wheat grain quality. Consequently, research on nationwide wheat grain quality prediction remains relatively scarce, and studies addressing model uncertainty are even rarer. To address these challenges, this study develops a phenology-guided time series prediction model (PBA-LSTM) for wheat grain protein content (GPC) using nationwide data collected from 2008 to 2019. This model integrates phenological (P) information, Bayesian inference (B), self-attention mechanisms (A), and the advantages of Long Short-Term Memory (LSTM) for mining temporal information, alleviating the spatiotemporal bias issues in large-scale wheat GPC predictions. Specifically, first, to address the monitoring omissions in the corrected shape model during large-scale applications, a multi-calibration point integrated corrected shape model is proposed, which provides seamless and complete phenological information for large-scale wheat GPC predictions. Second, based on the phenological information, time-series features from multi-source remote sensing, meteorological, and soil data are aggregated to construct the PBA-LSTM model, which integrates Bayesian inference, self-attention mechanisms, and LSTM. Finally, the PBA-LSTM model is used to predict winter wheat GPC nationwide and perform uncertainty analysis. The results show that the proposed phenology monitoring method significantly improved monitoring accuracy (mean R² = 0.95, RMSE = 8.29). Compared to other traditional machine learning models, the PBA-LSTM quality prediction model achieved the best performance, with R² of 0.66 and RMSE of 0.86%. In the PBA-LSTM model, meteorological data played a more important role than remote sensing data, with the phenological stages from stem elongation to heading being particularly influential. Regarding uncertainty assessment, stochastic uncertainty was found to be the primary source of model uncertainty, with greater uncertainty introduced during the earlier phenological stages. As the quantity of training data across different years increased, the model's uncertainty decreased. The PBA-LSTM model developed in this study enhances the accuracy of wheat GPC predictions across a wide range of phenological differences and complex environmental conditions. Moreover, it provides a relevant uncertainty assessment, offering important technical support and quality evaluation references for agricultural management, food production planning, and policy-making decisions.

As an international natural biological disaster, locust disaster is also one of the three natural disasters in China, which will cause huge losses to agriculture, forestry and animal husbandry production, and then cause serious ecological and environmental problems. It is of great significance to carry out research and prevention work related to grassland locust for improving the ability to resist grassland locust disasters and safeguard the grassland ecological environment. However, a wide area of grassland is often distributed with a variety of locusts, and it is difficult to obtain accurate distribution data of all kinds of locusts by relying on limited field investigation, which brings great troubles to the prevention and control of locusts. To this end, this paper takes Hulunbuir area, where there are many kinds of grassland locusts distributed, as the research area, and uses the existing distribution data of grassland locusts in Hulunbuir area to carry out research. First, according to the development mechanism of the grassland locust, four life cycles of the grassland locust were divided according to the empirical model, and 22 ecological environment factors were selected from the four factors of meteorology, vegetation, soil and terrain. Then, factor detector and interactive detector in the geographical detector method are used to screen the initially selected ecological factors. Then, based on the occurrence data of grassland locusts in Hulunbuir area from 2008 to 2020 and the selected ecological environment factors, the Maxent model was used to predict the suitable areas of grassland locusts in Hulunbuir area, so as to obtain the potential distribution areas of grassland locusts that had not been reached in field investigation.The ecological environment factors screened in this study included DEM, vegetation type, soil type, above-ground biomass at eclosion stage, above-ground biomass at spawning stage, soil moisture at incubation stage, precipitation at incubation stage, NDVI at eclosion stage, NDVI at spawning stage, mean surface temperature at overwinter stage, mean surface temperature at eclosion stage and minimum surface temperature at overwinter stage.The AUC value of the model running every year is above 0.9, which proves that the accuracy of the model results is good, and the prediction results of the suitable areas are in good agreement with the existing locust distribution data. The suitable area is mainly concentrated in the west of Hulunbuir, Chen Balhu Banner, New Balhu Left Banner, Ewenki Autonomous Banner and New Balhu Right Banner. The prediction results of suitable areas showed that Maxent model could obtain high prediction accuracy without spatial autocorrelation and overfitting in the case of limited actual pest distribution data. The research results can provide strong support for the scientific control of locusts, and have important significance for the stable development of agriculture and animal husbandry and ecological protection.

Forest type plays a crucial role in characterizing and understanding the diversity and distribution of forests. In this study, we focus on the potential of applying the model of the dominant leaf type (DLT) developed for forest type mapping in Switzerland to the study area north of Genhe, China. The model used leaf-off and leaf-on Sentinel-1 backscatter data of both polarizations, VV and VH, cloud-free leaf-on Sentinel-2 imagery, an Airborne Laser Scanning (ALS)-based Digital Terrain Model (DTM), and a forest mask, resulting in a map with the thematic classes of broadleaf and coniferous forest types Secondly, the original S2 bands were used with initial spatial resolutions of 10 and 20 meters, respectively. All datasets were resampled to 10 m ground sampling distance (GSD). The model calibration was repeated five times, each time re-selecting the training, validation and test datasets. In addition, the predicted final maps were compared to independent Swiss National Forest Inventory (NFI) plot data. In addition, three vegetation indices (VIs) were used, namely the Normalized Vegetation Index (NDVI), the Modified Soil Adjusted Vegetation Index (MSAVI2) and the Global Environmental Monitoring Index (GEMI). Based on the trained model, DLT mapping is produced for the study area of Genhe in China using S1 & S2 composites, DTM, slope and aspect from the open Shuttle Radar Topography Mission (SRTM) data.

This study seeks to develop a deep learning-based method for Urban Functional Zone (UFZ) change detection through the integration of multi-temporal Street View (SV) and Remote Sensing (RS) imagery. The objective is to overcome the limitations of traditional approaches, such as manual surveys and RS imagery, which often fail to capture fine-grained or ground-level details, by introducing the Multi-Source Temporal Fusion (MSTF) model. The MSTF aims to dynamically fuse SV and RS features via a Multi-Source Fusion (MSF) module and to capture temporal dynamics through a Multi-Temporal Fusion (MTF) module. An additional goal is to improve detection accuracy by aggregating SV classifications at the UFZ level. Ultimately, the study intends to establish a robust framework for multi-source data integration, thereby enhancing UFZ analysis to support improved land use optimization and urban management.

Human activities have profoundly reshaped fundamental ecological processes, pushing ecosystems toward unsustainable trajectories, particularly in the ecologically fragile regions such as Hohhot-Baotou-Ordos-Yulin urban agglomeration. To address these challenges, a quantitative framework was developed to evaluate anthropogenic sensitivity of ecological patterns. Functional connectivity was modeled using an omnidirectional circuit model to represent regional ecological patterns, with landscape elements extracted through morphological analysis. The mechanisms, intensity and pathways of anthropogenic sensitivity were explored using Geodetector and structural equation modeling, identifying habitat quality as a key mediating factor. The findings indicate that improving habitat quality greatly enhances omnidirectional connectivity. Among landscape elements, islets exhibit lower connectivity than the edges of core areas, despite higher resistance being assigned to the edges. In the Mu Us Desert, strip-like corridors serve as connectors but remain fragile, whereas in the Kubuqi Desert, patch-like corridors primarily function as barriers. Habitat quality and cultivated land emerge as dominant drivers of omnidirectional connectivity, while population and bare land contribute negatively through interactive effects that exceed the impacts of individual factors. Habitat quality directly enhances omnidirectional connectivity, with a path coefficient of 0.67. Bare land negatively impacts habitat quality, with a coefficient of -0.65, while cultivated land has a negative effect on grassland, with a coefficient of -0.82, indirectly shape regional ecological patterns. This study provides a quantitative understanding of the mechanisms driving anthropogenic sensitivity in ecological patterns, offering valuable insights to guide and optimize ecological spatial planning in arid urban agglomerations.

Rapid urbanization and climate change are exacerbating ecological vulnerability in dryland urban agglomerations, posing risks to sustainability and ecosystem stability. However, persistent technological gaps in large-scale, fine-grained and long-term monitoring hinder a comprehensive understanding of vulnerability patterns in these fragile regions. To address this, a novel Dryland Ecological Vulnerability Index (DEVI) is proposed by integrating six key indicators, combining remote sensing and machine learning to simplify the complex vulnerability scoping diagram (VSD). Using the Hohhot-Baotou-Ordos-Yulin urban agglomeration, a typical fragile region in China's drylands, as a case study, we analyzed spatiotemporal dynamics and sustainability of ecological vulnerability from 1986 to 2024. SHapley Additive exPlanations (SHAP) coupled with XGBoost revealed the impact mechanisms and nonlinear interactions of natural and anthropogenic drivers. Results showed that DEVI effectively captured the surface ecological features, such as quicksand, with high correlations (>0.79) to indicators. Over nearly 40 years, DEVI initially increased, then decreased, declining by 14% overall. Vulnerability notably reduced in Loess Hills but remained high in Ordos Plateau and northern Inner Mongolia, with degradation areas still exceeding improvements by 4.5%, reflecting imbalance in current ecological governance. Fortunately, improved areas have been increasing, with spatial sustainability reaching 82.8%, largely driven by land cover restoration (46.6%) and socioeconomic factors (27.0%). The study also identified the thresholds and interaction effects of key drivers, delineated new ecological management zones, and proposed targeted improvement suggestions. This study provides a novel index for ecological vulnerability monitoring and further offers practical guidance for sustainable development in dryland urban agglomerations.

Due to differences in data sources and processing methods, the accuracy of Digital Elevation Models (DEMs) varies greatly in different regions, which poses challenges for users in data selection. To address this issue, this study proposes the Optimal Terrain Retrieval (OTR) method, which integrates the optimal segments from public DEMs by using spaceborne Light Detection And Ranging (LiDAR). The OTR method involves selecting LiDAR photons, assigning weights, ranking DEMs by error statistics relative to LiDAR benchmarks, extracting and merging the most accurate segments of each DEM. OTR ensures integrating the highest quality data from each DEM while preserving the original data’s integrity. The experiments were conducted in the Loess Plateau, and results show that the OTR-derived DEM (OTRDEM) has a 25.71% lower Mean Absolute Error (MAE) than the leading COP30 and a 25.40% lower RMSE than the FABDEM. Additionally, OTRDEM demonstrates advantages in rough terrain and densely vegetated areas. The proposed method provides a scalable and adaptable approach for DEM optimization for various regions and datasets. It allows for the incorporation of new DEMs without requiring degradation assumptions or extensive training processes, enhancing terrain representation as more LiDAR observation data becomes available.

Urbanization is a key primary driver of global change, significantly impacting societal, economic, and environmental dynamics. Monitoring urban development is essential for sustainable urban planning and management. Traditionally, multispectral imaging has been used for urban studies; however, its application is hindered by environmental factors, such as cloud cover, that restrict high-frequency monitoring of urban developments.

Synthetic Aperture Radar (SAR) sensors, on the other hand, provide round-the-clock observation, enabling high-frequency urban change detection, including building reconstruction, demolition, and disaster risk management. Despite this, directly comparing and detecting changes using heterogeneous SAR data remains a challenge due to the unique backscattering characteristics of each sensor.

This research proposes a novel framework for urban change detection by leveraging multi-temporal SAR images acquired from different sensor modalities. At the core of the approach is a deep learning-based image translation method that aligns data across different SAR modalities. Deep learning models, particularly for image-to-image translation, are harnessed to learn high-level semantic features and facilitate cross-modal analysis. The main objective is to extract accurate urban activity changes by aligning SAR data from different sensors, improving urban monitoring capabilities.

By leveraging a Cycle-GAN framework to convert co-occurring image pairs with minimal time gaps from one SAR sensor domain to another, this methodology converts images from one SAR sensor domain to another, thereby enhancing the spatio-temporal resolution of the image sequences and improving the extraction of change-related information.

Preliminary evaluations over Shanghai, using SAR data from Sentinel-1 and COSMO-SkyMed, demonstrate the efficacy of this approach. Performance metrics, including Mean Squared Error (MSE), Peak Signal-to-Noise Ratio (PSNR), and Cosine Similarity, confirm that the dual-generator Cycle-GAN effectively translates images from one sensor modality to emulate another, achieving significantly better results compared to direct comparisons of heterogeneous data sources.

Looking ahead, two main avenues of research will be explored. First, the algorithmic aspect of the methodology will be enhanced by investigating the use of more recent and cutting-edge diffusion models applied to the SAR-to-SAR image translation task. Diffusion models, which have shown promise in high-quality image generation, may offer improvements over traditional Generative Adversarial Networks (GANs) in terms of both translation accuracy and image fidelity. Second, future work will focus on developing application-specific scenarios to demonstrate the practical utility of the proposed framework, including detailed urban construction monitoring with early-stage detection and progress tracking.

By expanding on these directions, this research aims to refine the methodology and explore more advanced techniques, ultimately providing a robust and versatile tool for urban change detection using heterogeneous SAR data. This advancement has significant implications for urban planning, disaster response, and environmental monitoring.