Conference Agenda

Timely and reliable detection of forest disturbance is crucial for monitoring tropical ecosystems and informing conservation and policy actions. The European Space Agency’s BIOMASS mission, launched to provide global P-band Synthetic Aperture Radar (SAR) measurements, offers unprecedented potential for observing forest structure and biomass dynamics. Beyond its primary objective of estimating above-ground biomass, BIOMASS data may also enhance the capability to detect forest disturbances due to its sensitivity to forest vertical structure.

This study investigates the potential of Cal/Val BIOMASS multi-temporal observations for detecting tropical forest disturbances. BIOMASS assessments derived from multi-temporal, multi-baseline polarimetric and interferometric observations are compared with forest loss detection results obtained from C-band SAR Sentinel-1 (S1) and multispectral Sentinel-2 (S2) data, using a Bayesian Online Changepoint Detection (BOCD) framework that enables near real-time monitoring. In particular, BIOMASS results are compared against the detection performance of three BOCD configurations: S1 single-polarization (VH-BOCD), S1 dual-polarization (pol-BOCD), and the generalization of the approach to jointly handle asynchronous time series from S1 and S2 (ms-BOCD).

The evaluation is carried out over selected areas in the Brazilian Amazon using reference data from MapBiomas Alerta, an initiative devoted to land cover mapping in Brazil that provides spatially explicit records of deforestation events derived from visual interpretation of high-resolution optical imagery. Both spatial and temporal synergies between Sentinel-based detections and BIOMASS observations are investigated to assess the added value of integrating multi-source data for accurate forest disturbance monitoring.

The ERS-1 satellite was launched on 17th July 1991, ERS-2 on 21st April 1995, and ENVISAT on 1st March 2002. These three satellites all carried a Synthetic Aperture Radar (SAR) operating at C-band and acquired some data over two decades. All ERS-1/2 and ENVISAT SAR products, in accordance with ESA Earth Observation Data Policy, are freely available to users [1][2]. While some products have been bulk processed, other products are generated On-the-Fly using the same SAR processor. However, this processor is currently closed source, and no Algorithm Theoretical Baseline Document (ATBD) is available.

In 2023 an activity was started to develop from scratch a new SAR processor that is to be Open-Source, with an open ATBD, and capitalizing on GPU processing.

The objectives are:

- to step-up the possibilities of ESA service to its end users, both for On-The-Fly services, and to facilitate the execution of massive reprocessing campaigns by reducing both the processing times and associated costs;

- to progress in the migration away from proprietary code;

- to eliminate the dependencies from old legacy libraries;

- To facilitate the agility and the effectiveness in keeping up with IT security changes and evolutions.

These objectives shall be achieved while fulfilling all scientific requirements. This development represents a big step in the direction of Open-Science.

This paper will present the new SAR processor and a comparison between the legacy and the new processor in term of data quality and processing speed.

References

[1] https://earth.esa.int/eogateway/instruments/sar-ers

[2] https://earth.esa.int/eogateway/instruments/asar

The BioTech Quilombo project, part of the Amazônia +10 Initiative, is pioneering a community-led approach to biodiversity monitoring in the Amazon, placing Afro-Brazilian Quilombola communities at the heart of modern conservation science. Led by the University of Manchester and IPAM, in partnership with the co-leads UFRR, INPA, UNESP, and ETH Zürich, the project brings together over 40 scientists — most from Amazonian institutions — 10 Quilombola leaders, and seven funding agencies: UKRI, SNSF, FAPESPA, FAPEAM, FAPRR, FAPESP, and CNPq. By integrating traditional ecological knowledge with advanced technologies such as remote sensing, eDNA, DNA barcoding, and artificial intelligence, BioTech Quilombo is developing a flagship framework for long-term biodiversity and forest structure assessment across the Amazon. Working closely with communities in Pará, Amazonas, and Roraima, the initiative empowers local researchers through training and intercultural exchange, embedding equity, ethics, and co-production at every stage of scientific practice. The project’s study sites are l also contribute to the validation and calibration of the ESA Biomass SAR mission (PP0106231 - BioMassQuilombo-Amazon: Participatory Calibration and Validation of ESA BIOMASS SAR). We explored BIOMASS SAR datasets across a gradient of forest types and structures at three Amazonian sites, encompassing variations in terrain and vegetation characteristics across primary and degraded forests, river meanders, and floodplain mosaics. Although our analyses are preliminary, the results already demonstrate that the BIOMASS SAR data can effectively detect temporary river meanders and spatial differences in surface moisture among forests in the studied areas.

The ESA BIOMASS mission aims to retrieve above-ground biomass and forest height data to enhance our knowledge about the state of the Earth’s forests and the carbon cycle. Besides these primary mission parameters, its P-band (430 MHz) SAR sensor is also able to provide information on secondary geophysical variables, such as tree trunk water content (TWC) and upper root-zone soil moisture (RZSM). These parameters are essential for understanding the Earth’s climate and vegetation-hydrology interactions, and play an essential role in the soil-plant-atmosphere feedback processes [1,2]. However, they remain insufficiently validated in current global Earth Observation frameworks.

The TWIST-NZ project seeks to establish a framework to validate secondary mission products at three sites in New Zealand. New Zealand offers a unique environment with varying climatic zones, soil types, and geology with the same dominant tree species, Pinus radiata (D. Don) [3], making it an ideal location for testing satellite-derived estimates of TWC and RZSM based on P-band BIOMASS observations. Additionally, during the Forest Flows program (2019-2025) research sites were intensely measured with a series of integrated terrestrial and remote sensing measurements [4], including airborne P-Band SAR [5]. We will take advantage of the established sensor networks and previous measurements to evaluate the sensitivity of the P-band radar sensor to RZSM, assess the seasonality in TWC, and examine the variability in RZSM across three different forest with contrasting climate, geology and soil.

For the estimation of TWC and SM from P-band BIOMASS data, the approaches from [1,2,6] should be adapted and further refined. These approaches include polarimetric decomposition of SAR observations as well as signal modeling based on physically-constrained methods (such as the multi-layer soil scattering and radar vegetation interaction models). For the validation of the secondary products, in situ measurements from the Forest Flows monitoring network (https://www.forestflows.nz), acquired airborne P-band SAR observations from the SlimSAR system (aligning with BIOMASS satellite overpasses to ensure comparable environmental conditions), and satellite observations (e.g., NiSAR) should be employed and integrated in a multi-scale observation framework.

By validating these secondary products, TWIST-NZ aims to enhance the scientific reliability and extend the practical relevance of BIOMASS for ecosystem monitoring, hydrological modeling, and forest management. Further, all processed datasets, validation results, and analysis methods will be made publicly available, e.g., via the ESA Multi-Mission Algorithm and Analysis Platform (MAAP).

[1] Fluhrer, A., T. Jagdhuber, A. Tabatabaeenejad, H. Alemohammad, C. Montzka, P. Friedl, E. Forootan, and H. Kunstmann (2022): Remote sensing of complex permittivity and penetration depth of soils using P-band SAR polarimetry. Remote Sensing 14(12), 2755. DOI:10.3390/rs14122755

[2] Fluhrer, A., T. Jagdhuber, C. Montzka, M. Schumacher, H. Alemohammad, A. Tabatabaeenejad, H. Kunstmann, and D. Entekhabi (2024): Soil Moisture Profile Estimation by Combining P-band SAR Polarimetry with Hydrological and Multi-Layer Scattering Models. Remote Sensing of Environment 305, 114067. DOI:10.1016/j.rse.2024.114067

[3] Zhu, H., Meason, D.F., Salekin, S., Hu, W., Lad, P., Jing, Y. and J. Xue (2024). Time stability of soil volumetric water content and its optimal sampling design in contrasting forest catchments. Journal of Hydrology 131344; DOI:10.1016/j.jhydrol.2024.131344

[4] Meason, D.F., Matson, A., Baillie, B., Moller, D., Dudley, B., Srinivasan, M.S., Rajanayaka, C., Zammit, C., and D. White (2021): Forest Flows – Real time monitoring of water quantity and quality spatio-temporal dynamics in planted forests. IGARSS 2020 - 2020 IEEE International Geoscience and Remote Sensing Symposium, Waikoloa, HI, USA, 2020: 4626-4629; DOI:10.1109/IGARSS39084.2020.9324637

[5] Zhao, Y. -H., Moller, D., Meason, D., and M. Moghaddam (2024). Multifrequency Subsurface Soil Moisture Retrieval for Forest Flows: A Case Study in Te Hiku, New Zealand. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing; DOI:10.1109/JSTARS.2024.3493118

[6] Fluhrer, A., H. Alemohammad, and T. Jagdhuber (2025): Analyzing the dihedral scattering component of P-band SAR signals for trunk permittivity estimation–a concept study. Science of Remote Sensing 11, 100236. DOI:10.1016/j.srs.2025.100236

The knowledge of the health and distribution of forest biomass on our planet is one of the fundamental parameters for evaluating the many aspects that regulate life on the planet. In fact, important drivers of global warming, carbon cycle and biodiversity loss are recognized in deforestation, forest degradation and, more in general, forest management also in terms of bioeconomy [1] . Furthermore, there are many uncertainties regarding the total amount of carbon stored in the world's forests and how these reserves change over time due to temperatures, increasing CO2 concentrations in the atmosphere and not least human activities such as deforestation and agriculture.

In this context, satellite remote sensing is often proposed as a tool to overcome forest monitoring challenges, as it provides instantaneous and repetitive views of vast areas over time. However, while optical sensors can provide a view of the surface forest canopy, microwave sensors, and especially SAR, allow mapping of forest structural variability, deforestation and degradation. In this context, L-band or lower-frequency SAR data have demonstrated the best sensitivity to biomass in terms of Forest AboveGround Biomass (AGB) [2] - [3] . With the recent launch of ESA’s BIOMASS mission, carrying onboard a P band polarimetric SAR [4], the SAR potential for observing AGB is expected to further increase.

This study aims at assessing the contribution of different frequencies for improving the accuracy of AGB retrievals using SAR. Data from ESA’s BIOMASS obtained in the framework of ESA BRAVE project (PP0104553 - Biomass Radar for Assessing Vegetation and Ecosystems) have been combined with data derived from PALSAR-2, operating at L band, and Sentinel-1, operating at C band. It is well known indeed that, depending on the wavelength, each band has different penetration depth into the canopy and can therefore provide information on a different layer of the forest.

Two test areas have been identified for developing the proposed multifrequency approach, based on the availability of data from the three SAR sensors. The first area is located in Gabon, central Africa, and the second in the Amazon River Basin, Brazil, South America. Both areas are covered by equatorial forest with AGB ranging from 50 t/ha up to 350 t/ha and above. The latter area has further challenges related to its flooded nature.

The AGB retrieval has been based on two popular machine learning techniques, namely Neural Networks (ANN) and Random Forests (RF).

The ANNs have been largely applied to solve remote sensing problems (e.g. [5] - [7] ), thanks to their ability in approximating almost any kind of non-linear relationships ([8] [9] ). The ANN implementation proposed in this study is based on the feed-forward multi-layer perceptron neural networks (MLP-ANN), with iterative optimization of hyperparameters based on [10] .

Like the ANNs, RF are gaining increasing popularity for solving remote sensing problems (e.g., [11] - [14] ). The RFs belong to the ensemble learning methods ([15] ), which average the results coming from several weak predictors, also called decision trees, to establish the input/output relationship. An iterative optimization of the main hyperparameters, like the one adopted for ANN, has been carried out.

Algorithms have been trained considering the calibrated and coregistered backscattering from the three SAR sensors as input and the AGB as target. Given the lack of spatially distributed in-situ measurements suitable for the algorithms’ development, the target AGB has been derived from the last release of ESA CCI AGB map ([16] ).

Auxiliary data as the local incidence angle have been also considered as input while the not forested pixels and the areas in which the retrieval is not feasible have been discarded based on the land cover information derived from the ESA’s CCI LCC map [17] . All the data have been geocoded, coregistered and reprojected over a fixed grid at 100 m resolution.

To keep training and test as independent as possible, both algorithms have been trained over a subset of the available data and tested on the remaining. In order to assess the contribution of the higher frequencies (L and C bands), different input configurations have been exploited, by considering P band only, P + L band combination and P + L + C bands, at all the polarizations available. A predictor importance analysis has been also implemented to assess the contribution of each input to the overall results.

First tests showed good accuracy of the implementation using P band only, and some improvement when considering the multifrequency implementation, that is able to exploit the contribution of branches and leaves from different layers of the canopy.

References

[1] FAO and UNEP. 2020. The State of the World’s Forests 2020. Forests, biodiversity and people. Rome, https://doi.org/10.4060/ca8642en

[2] Woodhouse, I., Mitchard, E., Brolly, M. et al. Radar backscatter is not a 'direct measure' of forest biomass. Nature Clim Change 2, 556–557 (2012). https://doi.org/10.1038/nclimate1601

[3] Joshi, N., Mitchard, E.T.A., Brolly, M. et al. Understanding ‘saturation’ of radar signals over forests. Sci Rep 7, 3505 (2017). https://doi.org/10.1038/s41598-017-03469-3

[4] Quegan et al. (2019) "The European Space Agency BIOMASS mission: Measuring forest above-ground biomass from space," published in Remote Sensing of Environment.

[5] Dai X., Z. Huo, and H. Wang, “Simulation for response of crop yield to soil moisture and salinity with artificial neural network”, Field Crops Research, 121, 3, pp. 441-449, 2011. Doi: 10.1016/j.fcr.2011.01.016.

[6] Del Frate F., P. Ferrazzoli, and G. Schiavon, “Retrieving soil moisture and agricultural variables by microwave radiometry using neural networks”, Remote Sens. Environ., 84, 2, pp. 174–183, 2003. doi: 10.1016/S0034-4257(02)00105-0.

[7] Elshorbagy A., and K. Parasuraman, “On the relevance of using artificial neural networks for estimating soil moisture content”, Journal of Hydrology, 362, pp. 1–18, 2008.

[8] Hornik K., “Multilayer feed forward network are universal approximators”, Neural Networks, 2, 5, pp. 359-366, 1989.

[9] Linden A., and J. Kinderman, “Inversion of multi-layer nets”, Proc. Int. Joint Conf. Neural Networks, 2, pp. 425-43, 1989.

[10] Santi E., “Neural Networks applications for the remote sensing of hydrological parameters”, Artificial Neural Networks - Models and Applications book, InTechOpen, 2016. ISBN 978-953-51-2705-5.

[11] Camargo F.F., E.E Sano, C.M. Almeida, J.C. Mura, and T. Almeida, “A Comparative Assessment of Machine-Learning Techniques for Land Use and Land Cover Classification of the Brazilian Tropical Savanna Using ALOS-2/PALSAR-2 Polarimetric Images”, Remote Sens., 11, 1600, 2019.

[12] Marrs J., and W. Ni-Meister, “Machine Learning Techniques for Tree Species Classification Using Co-Registered LiDAR and Hyperspectral Data”, Remote Sens., 11, 819, 2019.

[13] Pal M., “Random Forest classifier for remote sensing classification”, Int. J. Remote Sens., 26, 217–222, 2005.

[14] Yu Y., M. Li, and Y. Fu, “Forest type identification by random forest classification combined with SPOT and multitemporal SAR data”, J. For. Res., 29, 1407–1414, 2018.

[15] Breiman L., “Random Forests”, Mach. Learn., 45, 5–32, 2001.

[16] Santoro, M.; Cartus, O. (2025): ESA Biomass Climate Change Initiative (Biomass_cci): Global datasets of forest above-ground biomass for the years 2007, 2010, 2015, 2016, 2017, 2018, 2019, 2020, 2021 and 2022, v6.0. NERC EDS Centre for Environmental Data Analysis, 17 April 2025.

[17] ESA. Land Cover CCI Product User Guide Version 2. Tech. Rep. (2017). Available at: maps.elie.ucl.ac.be/CCI/viewer/download/ESACCI-LC-Ph2-PUGv2_2.0.pdf

This study investigates the potential of radar polarimetric data for geological mapping in the Gobi Desert region of Mongolia, focusing on the period before the availability of Biomass P-band data. To address this gap, we analyzed historical L-band polarimetric datasets from ALOS and SAOCOM satellites, as well as C-band alt-pol data from the Sentinel-1 constellation. Geological information—including rock types, contact zones, and tectonic features—was extracted through various polarimetric decomposition techniques. Our analysis centered on the area of a Polish scientific expedition conducted in 2025, integrating both field, ground-truth data and hyperspectral imagery from archived PRISMA and newly acquired ENMap sources. Various classification approaches such as Spectral Angle Mapper (SAM), Spectral Feature Fitting (SFF), Spectral Feature Analysis (SFA), and Support Vector Machine (SVM) were employed to enhance interpretation. Data processing was performed in Python and Jupyter notebooks using SNAPISTA and SNAP software.

Preliminary results indicate that radar polarimetric data reveal previously undocumented geological features, offering significant insights into surface roughness, structure, and subsurface penetration compared to traditional optical sensors. These findings underscore the added value of radar data and highlight opportunities for further research. The anticipated integration of L-band SAR data from the Biomass satellite, which has not yet been explored from a geological perspective in this region, is expected to provide greater penetration capabilities and may represent a transformative advance for geological applications in arid environments.

Accurate estimation of above-ground biomass (AGB) is critical for carbon cycle science, yet existing methods, particularly in dense tropical forests, are often subject to significant and systematic biases. This project aims to characterize the sources of bias in global AGB datasets and develop a framework for its prediction and correction. We hypothesize that when models with different architectures and sources of features exhibit similar bias patterns, the cause is likely external, stemming from the reference data or environmental conditions rather than the model formulation itself. Our analysis first evaluated bias by comparing the European Space Agency (ESA) Climate Change Initiative (CCI) Biomass dataset and an independent AGB product for the Brazilian Legal Amazon against circa 900 overflights of airborne Light Detection and Ranging (LiDAR) observations calibrated with ground truth plots, totalizing circa 125.000 pixels. We found a consistent pattern across both datasets: models systematically overestimate AGB at the lower end of the biomass spectrum and underestimate it at areas with higher volumes of AGB. The signed raw and relative residuals between the datasets are significantly correlated (Pearson's R² up to 0.47, Spearman's S up to 0.68), indicating that bias occurs in similar geographical areas and have similar drivers. In the project's second phase, we employed machine learning techniques (XGBoost, Random Forest, Lasso) to model these residuals using a suite of geoenvironmental variables and the CCI AGB estimate itself. Predictors included vegetation indices (NDVI, EVI), precipitation history, and the CCI's own uncertainty metric. To address the ambiguity and systematic bias present in the reference samples, we implemented a post-processing bias correction technique. We focused exclusively on a high-confidence subset of LiDAR reference samples where the reported uncertainty was below the 10th percentile. Using this optimal subset of reference samples, we trained a model to predict the signed residuals of the original AGB estimates. The resulting model achieved a strong performance, with an R² of 0.87. By adding these predicted residuals to the original AGB estimations from the CCI dataset, we observed a substantial reduction in the overall prediction error for this high-confidence subset, reducing the Root Mean Square Error (RMSE) from 135.82 to 26.48. This post-processing approach represents a major validation of the bias-correction strategy and demonstrates its power to improve the AGB estimation accuracy when high-quality references are available. These findings suggest that a significant portion of the drivers for epistemic bias is concentrated in specific regions rather than being universally present. We propose that in the future a more effective, two-stage architecture for post-processing AGB estimates in areas prone to bias will be designed. The first stage would involve a classification module to classify areas where LiDAR is expected to exhibit lower uncertainty; followed by a targeted regression model to predict the magnitude of the bias specifically within those identified, high-confidence areas. By unrevealing dimensions of the tropical dense forest structure that are commonly unseen by current spaceborne operating bands (C and L bands), the forthcoming P-Band from the BIOMASS mission is anticipated to be a critical input variable in the feature set for both the classification and the regression stages. These results will be followed by an extrapolation to backcast post-processed AGB estimations, resulting in a dataset that takes advantage of the expected better performance of the BIOMASS product, and the spatial resolution and long-term time series of the CCI AGB project. By combining these results, this project contributes to the next generation of AGB datasets, achieving both improved accuracy and precision to meet not only critical climate science requirements, but also other applications that require a more granular precision, such as fire and deforestation monitoring.

Radar interferometry is a technique capable of measuring precise surface displacement by analyzing the interferometric phase between two images acquired at different times. However, an interferogram generated from Synthetic Aperture Radar(SAR) data contains various phase components unrelated to displacement, including those from topography, orbital errors, earth curvature, atmosphere, and noise. Time-series interferometry techniques effectively remove these non-displacement phase signals. A prominent time-series method is Persistent Scatterer Interferometry (PSI), which monitors displacement using stable Persistent Scatterers(PS) identified over the acquisition period. While SAR-based PSI is widely applied to monitor phenomena such as ground subsidence, earthquakes, and volcanic activity, its long revisit period, typically several days, limits its ability to capture rapid, short-term displacements. Furthermore, SAR's side-looking imaging geometry can introduce geometric distortions such as layover, foreshortening, and shadow which may result in observational blind areas.

In contrast, ground-based radar(GBR) offers flexible control over acquisition time, location, and antenna geometry to specific monitoring objectives. GBR can acquire high-precision time-series data, making it highly effective for observing rapid and localized displacements. Consequently, it is extensively utilized in diverse applications, including the monitoring of slopes, landslides, glaciers, and subsidence. An advantage of GBR is that its fixed antenna position and comparatively smaller observation area inherently exclude topographic and earth’s curvature phases from the interferogram. However, GBR is highly susceptible to atmospheric phase delay caused by changing weather conditions. Temporal and spatial variations in temperature, humidity, and pressure alter the atmospheric refractivity. This fluctuation modifies the propagation path and velocity of the radar waves, inducing an atmospheric phase unrelated to actual surface displacement. This atmospheric phase significantly limits high-precision measurements, and its effect is particularly pronounced at higher frequencies.

This study aims to evaluate the atmospheric interferometric phase in the application of PSI to GBR, and to propose a correction method using meteorological data. The GPRI-II (Gamma Portable Radar Interferometer-II) system operates in the Ku-band, with a frequency range of 17.1-17.3 GHz. Data were acquired continuously for 33 hours(from 17:00, Sept 18, to 02:00, Sept 20, 2025) at an levee in South Korea, with a 5-minute interval. A total of 397 Single-Look Complex(SLC) images were collected. Synchronous temperature and humidity data were collected at the same interval from hygrometers co-located with corner reflectors(CR) on the levee. Barometric pressure data was obtained from a meteorological station 6 km away. The 33-hour dataset was divided into 11 distinct 3-hour sections. Using the first image of each section as the reference, we calculated the interferometric phase and the corresponding changes in meteorological conditions. A correlation analysis revealed the highest coefficient of determination (R² = 0.9) in the section 20:00 to 23:00. Subsequently, we calculated the atmospheric refractivity from the meteorological data to generate an atmospheric phase model. This model was then applied to correct the atmospheric phase in the SLC stack. Finally, PSI was performed using the first acquired image as the reference. The atmospheric correction was evaluated using the interferometric phase at two installed corner reflectors(CR1, CR2), under the assumption of zero actual displacement. Before correction, the standard deviations of the phase were 0.26 radians(CR1) and 0.36 radians(CR2). After correction, these values were significantly reduced to 0.12 radians and 0.14 radians, respectively. When converted to Line-of-Sight(LOS) displacement rate, the displacement rates decreased from 0.5 cm/hour(CR1) and 0.7 cm/hour(CR2) before correction, to 0.1 cm/hour and 0.0 cm/hour after correction. This study successfully demonstrates the generation of an effective atmospheric phase model from meteorological data and its utility in correcting atmospheric phases observed in permanent scatterers. We conclude that this methodology can be effectively utilized to enhance the precision of displacement monitoring in ground-based radar PSI applications.

Validation of DTM Derived from P-band SAR Under the Amazon Rainforest: from Airborne Surveys to the Biomass Perspective

Laurent Polidori, Universidade Federal do Pará (UFPA), Belém, Brazil

Mhamad El Hage, Lebanese University (UL), Tripoli, Liban

Laurent Ferro-Famil, ISAE-SUPAERO / CESBIO, Toulouse, France

Ludovic Villard, CNRS / CESBIO, Toulouse, France

Carlos Caldeira, Universidade Federal Rural da Amazônia (UFRA), Belém, Brazil

Topics: BIOMASS calibration and validation, SAR Tomography (Tomo-SAR)

Keywords: Biomass, tomography, DTM, Amazon

Presentation preference : oral

Digital elevation models through forests is one of the most important products of the Biomass mission, although the case of tropical dense forests remains challenging. Previous tests carried out on the experimental sites of the TropiSAR (Paracou, French Guiana) and AfriSAR (Gabon) preparation campaigns have demonstrated the feasibility of extracting elevation from the P-band radar tomographic signal, but cannot fully conclude on the potential of Biomass to produce a DTM, for several reasons:

- The areas imaged are too small (we do not have access to the low frequencies of the terrain elevation or to a significant hydrographic catchment area enabling hydrographic coherence to be analysed).

- Slope is not assessed, even though it is a more relevant variable than elevation for most DTM applications.

- The quality indicators are incomplete, inspired by the practices of official mapping agencies which generally do not take into account the needs of the users of their products. Bias and standard deviation do not take into account the autocorrelation of the error and therefore do not allow conclusions to be drawn about the fidelity of the model to the landforms. A wider range of quality criteria must be considered, linked to the expectations of the users in the field of the geosciences.

- The Lidar elevation product is considered to be a ‘perfect’ reference, which is highly questionable.

In order to assess more comprehensively the potential of P-band radar for relief mapping under forest cover, we analysed InSAR products derived from aerial P-band radar acquisitions, available over an area of 160,000 km² in northern Brazil. Different quality criteria were considered, divided into two parts:

- External validation consists in comparing the DTM with reference data to characterise elevation and slope errors, their statistical distribution, their spatial distribution and the influence of landscape and acquisition parameters. The reference data are mainly Lidar surveys, which are more accurate but have a more limited spatial coverage.

- Internal validation consists in verifying the geomorphological realism and the coherence of the hydrographic network, based on hypotheses such as the fractal behaviour of terrestrial relief in geological environments shaped by water runoff and structured into watersheds.

Based on this work, a strategy for validating the MNT derived from Biomass is being implemented in order to better understand the potential and limitations of this product once it becomes available in the Amazon region, for a variety of purposes in geoscience and water management.

In the present paper we propose a two-component polarimetric SAR (PolSAR) decomposition method for estimating the depolarised and polarised components within the target and their corresponding coherency matrices. This idea stems from early works in the radar polarimetry field initiated by Huynen [1] and Cloude [2]. Later, several fundamental works dealt with the nonuniqueness issue of polarimetric decompositions [3,4]. Alternatively, Freeman and Durden moved to a different strategy based on accounting for the scattering physics describing the interaction of radar signals with the target according to some assumptions [5,6]. That work sparked a research line still active today in the field and focused on model-based decomposition techniques, being the final purpose the characterisation of the Earth surface through the retrieval of both quantitave and qualitative information from the decomposition outcomes (see for example [5, 6, 7, 8, 9, 10, 11, 12, 13, 14]). Nevertheless, the interpretation of polarimetric features is still subject to some ambiguity which can negatively affect both the derived land cover classification algorithms and the bio- and geophysical parameter retrieval techniques. Some recent attempts have dealt with this issue and tried to decouple different polarization channels [15] or improving the extraction of the dominant scattering mechanism through an eigendecomposition-based methodology [16]. The present paper is focused on this same issue, but a totally different methodology is employed instead. The method relies on the previous calculation of the backscattering powers given by the model-free three component (MF3C) decomposition [17], being the 3-D Barakat degree of polarisation [18] the key factor for separating polarised and depolarised backscattering components. Here, we propose to estimate the proportion of the polarised and depolarised contributions for all the elements of the observed coherency matrix under the reflection symmetry assumption. Basically, the proposed decomposition can be regarded as an extension of the MF3C method and, consequently, it enables the exploitation of both model-free and model-based approaches for parameter retrieval.

Indoor multi-frequency datasets acquired over three vegetation samples (i.e. cluster of small fir trees, maize, and rice) from the European Microwave Signature Laboratory (EMSL) have been employed for testing the proposed decomposition. Performance analysis has been supported by a quantitative analysis of the decomposed polarised and depolarised components of the elements of the observed coherency matrix and by the eigendecomposition of both resulting coherency matrices. Overall, it has been observed for the datasets employed that decomposition outcomes are consistent with the overall expected behaviour of polarimetric signatures. Limitations related to the decomposition performance specially regarding the T(1,2) element of the coherency matrix and implementation issues derived from the numerical procedure have been also investigated.

References:

[1] J. Huynen, “Phenomenological theory of radar targets,” Ph.D. dissertation, Dept. Elect. Eng., Math. Comput. Sci, Delft Univ. Technol., Delft, The Netherlands, 1970.

[2] S. R. Cloude, “Radar target decomposition theorems,” Electron. Lett., vol. 21, no. 1, pp. 22–24, 1985.

[3] W. A. Holm and R. M. Barnes, “On radar polarization mixed target state,” in Proc. USA Nat. Radar Conf., 1988, pp. 249–254.

[4] S. R. Cloude and E. Pottier, “A review of target decomposition theorems in radar polarimetry,” IEEE Trans. Geosci. Remote Sens., vol. 34, no. 2, pp. 498–518, Mar. 1996.

[5] Freeman, A. and Durden, S. L. A three component scattering model to describe polarimetric SAR data. In SPIE, Radar Polarimetry, volume 1748, pages 213–224, 1992.

[6] Freeman, A. and Durden, S. L. A three-component scattering model for polarimetric SAR data. IEEE Trans. Geosci. Remote Sens., 36(3):963–973, May 1998

[7] Yamaguchi, Y., Moriyama, T., Ishido, M., and Yamada, H. Four-Component scattering model for polarimetric SAR image decomposition. IEEE Trans. Geosci. Remote Sens., 43(8):1699–1706, 2005.

[8] Yamaguchi, Y., Sato, A., Boerner, W. M., Sato, R., and Yamada, H. Four-component scattering power decomposition with rotation of coherency matrix. IEEE Trans. Geosci. Remote Sens., 49(6):2251–2258, 2011.

[9] van Zyl, J.J., Arii, M., and Kim, Y. Model-based decomposition of polarimetric SAR covariance matrices constrained for nonnegative eigenvalues. IEEE Trans. Geosci. Remote Sens., 49(9):3452– 3459, 2011.

[10] Lee, J.-S., Ainsworth, T. L., and Wang, Y. Generalized polarimetric model-based decompositions using incoherent scattering models. IEEE Trans. Geosci. Remote Sens., 52(5):2474–2491, 2014.

[11] Chen, S. W., Wang, X. S., Xiao, S. P., and Sato, M. General polarimetric model-based decomposition for coherency matrix. IEEE Trans. Geosci. Remote Sens., 52(3):1843–1855, 2014.

[12] Jagdhuber, T., Hajnsek, I., and Papathanassiou, K. P. An iterative generalized hybrid decomposition for soil moisture retrieval under vegetation cover using fully polarimetric SAR. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 8(8):3911–3922, August 2015.

[13] Singh, G., Malik, R., Mohanty, S., Rathore, V. S., Yamada, K., Umemura, M., and Yamaguchi, Y. Seven-component scattering power decomposition of polsar coherency matrix. IEEE Trans. Geosci. Remote Sens., 57(11):8371–8382, Nov 2019.

[14] Ainsworth, T. L., Wang, Y., and Lee, J.-S. Model-based polarimetric SAR decomposition: An L1 regularization approach. IEEE Trans. Geosci. Remote Sens., 60, 2022.

[15] Han, W., Fu, H., Zhu, J., and Li, N. Decoupling between different polarization channels of polsar data. IEEE Geoscience and Remote Sensing Letters, 20, 2023.

[16] Hanis, D., Hadj-Rabah, K., Belhadj-Aissa, A., and Pallotta, L. Dominant scattering mechanism identification from quad-pol-sar data analysis. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 17:14408–14420, 2024.

[17] Dey, S., Ratha, D., and Frery, A. Target characterization and scattering power decomposition for full and compact polarimetric SAR data. IEEE Trans. Geosci. Remote Sens., 59:3981–3998, May 2021.

[18] Barakat, R. N-fold polarization measures and associated thermodynamic entropy of N partially coherent pencils of radiation. Optica Acta: International Journal of Optics, 30(8):1171–1182, 1983.

This study proposes a sub-canopy terrain extraction method based on the statistical characteristics of interferometric phase height histograms, using spaceborne L-band bistatic InSAR data acquired by China’s Lutan-1 system. The few-look InSAR phase height histogram reveals the vertical structure of volume scatterers, where the lower-height peaks correspond to ground scattering and the higher components reflect canopy scattering. By analyzing the statistical distribution of the histogram, a ground finding approach based on the histogram is developed to estimate the digital terrain model (DTM) without relying on external data for calibration. The method exploits both global and local statistical parameters of the phase height histogram to automatically separate sub-canopy and canopy scattering components. Experiments were conducted across multiple forest sites with varying canopy types (both temperate and tropical) and complex terrain conditions using Lutan-1 bistatic InSAR observations. The results demonstrate that the retrieved DTM shows strong agreement with spaceborne LiDAR-derived DTM (GEDI and ICESat-2/ATLAS), while the corresponding forest height estimates achieve the accuracy of a few meters. These findings validate the effectiveness of the proposed phase height histogram-based approach for automated and accurate forest height and DTM inversion without using external lidar data. This approach shows great potential for large-scale forest monitoring and surveying underlying terrain, serving as a complementary technique to other multi-polarization (PolInSAR) and/or multi-baseline (TomoSAR) methods for mapping forest vertical structure (e.g., ESA’s BIOMASS mission).

Accurate estimation of forest above-ground biomass (AGB) is essential for modeling global carbon cycle. Synthetic Aperture Radar (SAR) remote sensing is extensively used for forest AGB estimation. The role of polarimetric and time-series SAR data in improving estimates of AGB have been tested using limited airborne and spaceborne data. With the launch of NASA-ISRO SAR (NISAR) mission, we would soon have access to dual-polarimetric SAR data with global coverage at 12-day temporal resolution. In this work, we explore the utility of the dual-polarimetric scattering information from NISAR simulated data and its application for forest AGB modeling.

The NISAR simulated data is obtained from the UAVSAR AM-PM campaign with the raw SAR data reprocessed to NISAR-like 20 MHz bandwidth, added noise and a spatial resolution coarser than 6 m. The NISAR simulated data is acquired over Lenoir Landing site in Alabama, USA which is a temperate deciduous forest with mean and maximum field measured AGB of 125 Mg/ha and 373 Mg/ha respectively. Field validation is obtained from 41 plots of 1600 sq.m and 10 plots of 400 sq.m along with 1-m airborne lidar CHM. For Lenoir Landing site, six SAR acquisitions in 2019 and 2022 are used in this work. The NISAR simulated data is provided in HDF5 format and it is processed using the ISCE3 toolbox to obtain the dual-polarimetric (HH, HV) geocoded covariance matrix (C2).

The scattering information is obtained from dual-polarimetric C2 matrix using Scattering Power Factorization Framework (SPFF) to derive the dihedral-like (Pd), surface-like (Ps) and residual (Pr) scattering power components. These components are analyzed across the time-series as well as across the biomass range. The reference AGB is binned at 1 Mg/ha intervals and the behavior of scattering powers analyzed. We observe that the Ps increases initially till 20 Mg/ha and then declines and stabilizes around 60 Mg/ha. This would most likely be due to change in surface scattering from ground and smaller vegetated regions. The Pd and Pr both increase with AGB with the saturation characteristics showing time-dependence. These scattering powers are used for modeling AGB using semi-emperical models like the modified water cloud model (WCM) and the six-component model used in NISAR biomass ATBD. When the WCM is modeled using the HV-backscatter, a component due to dihedral-scattering reduces the sensitivity to AGB. This is removed when we use the Pd and Pr as inputs to WCM. We observe that the modified WCM model with the input of Pr component leads to a RMSE better than 25 Mg/ha across the six NISAR simulated acquisitions. The three components vary temporally as well as with changes in surface soil moisture and precipitation conditions.

The use of dual-pol scattering powers is likely to improve sensitivity to AGB changes as well as extend the inversion range due to higher saturation level. In our extended work, we will have an in-depth discussion on how the use of dual-polarimetric scattering power improves AGB retrieval accuracy, compare it with WCM and 6-component ATBD model, as well as try to understand the forest-type and temporal changes in SAR scattering over the temperate forest site.

The estimation of forest biomass in the Amazon is fundamental for supporting climate mitigation policies, both in Brazil and globally, as well as for understanding the significance of existing carbon stocks. This study compares data from the ESA's Biomass satellite, equipped with P-band SAR radar capable of penetrating clouds and dense canopies, with estimates derived from a local algorithm developed by the authors. Quantitative analyses employing RMSE and R² metrics assessed the concordance between the datasets, while discrepancy maps revealed systematic spatial differences.The qualitative assessment demonstrated the advantage of Biomass in cloud-covered regions and in capturing structural details of the forest. The results indicate that, although the local algorithm performs adequately in less dense areas, Biomass provides more reliable estimates in dense forests, such as those of the Brazilian Amazon. The study reinforces the value of integrating radar observations with algorithmic methods for comprehensive monitoring of Amazonian biomass.

Furthermore, the goal of the present analysis is to conduct comparative analyses to generate error metrics between the ESA data and those calculated by the 6BIOS project of the University of Aveiro and the Federal University of Paraíba, seeking to direct this discrepancy toward improving the calibration of biomass calculations acquired by the Biomass satellite. To perform this analysis, we conducted a study focused in a specific area of the Brazilian Amazon rainforest for which prior analyses have been performed, both by satellite and, in the future, enabling comparison with data obtained from regions that have been calculated in situ. The proposal is to demonstrate, in a highly restricted and focused manner, comparative data to advance the calibration and subsequent validation of data obtained from the Biomass satellite. In conclusion, we present suggestions for the need for more in-depth studies regarding coverage errors, such as occlusion, and the potential performance improvement that may be obtained from future measurements by the Biomass satellite.

Polarimetric Synthetic Aperture Radar (PolSAR) imaging is a powerful tool for studying how surfaces and objects interact with electromagnetic waves. It allows detailed observation of land cover, vegetation, and human-made structures. However, because SAR data are acquired through a coherent imaging process, they are affected by speckle noise a multiplicative effect that reduces image clarity and makes physical or statistical analysis less reliable.

Traditional methods for reducing speckle, such as multi-looking or model-based filters like Lee, Frost, and refined Gamma-MAP, rely on local statistics of pixel intensity or covariance matrices. While these approaches can effectively smooth noise and improve radiometric consistency, they often blur fine spatial details. Recent state-of-the-art despeckling methods have introduced several deep learning (DL) and data-driven approaches that have proven powerful in mitigating noise in SAR images while preserving structural details and avoiding blurring effects. However, most of these DL-based methods are designed for a specific number of polarization channels—either single, dual, or fully polarimetric data. Therefore, there remains a need for new methods capable of handling SAR image denoising across different polarimetric modalities.

To this end, we introduce a deep learning based framework for speckle suppression that works with any number of polarization channels. The key idea is a band-agnostic neural network architecture capable of handling single-, dual-, or quad-polarization data without any modification or retraining. The model uses a shared convolutional backbone to learn common spatial features across channels, while a cross-polarization attention mechanism captures the relationships between them preserving the physical information encoded in the scattering matrix.

Our method is trained entirely on real PolSAR observations, where the reference (noise-free) images are generated using a spatio-temporal averaging strategy applied to time-series data. Multiple co-registered SAR images of the same area are temporally averaged and subsequently processed with the MuLog spatial filtering technique to produce a noise-suppressed dataset. These filtered images serve as pseudo-clean references, enabling the network to learn directly from real data while preserving the physical characteristics and statistical authenticity of SAR observations.

During training, the loss function balances two goals: accurate spatial reconstruction and preservation of polarization coherence. This ensures that the denoised images stay true to the original scattering behavior, distinguishing random speckle variations from meaningful structural features.

In summary, this work presents a polarization-agnostic, data-driven deep learning approach to speckle reduction in PolSAR imagery. By combining multitemporal MuLog-filtered supervision with a flexible neural architecture, our method unites the strengths of traditional statistical filters and modern deep learning. The result is a practical, scalable, and physically consistent solution for improving the quality and interpretability of polarimetric SAR data across a wide range of remote sensing applications.

Forest height is one of the most important forest structure parameters due to its close relation with biomass, terrestrial carbon storage and ecosystem dynamics. Among the forest height observation methods such as passive optical, lidar and manual in-situ measurement, InSAR has unique advantages, including its height sensitivity as well as all-weather, day-and-night, and wall-to-wall mapping capability, which makes it a powerful tool for large-scale forest height observation. As the majority of spaceborne InSAR coverage, single-baseline single-polarization InSAR data was used in this research due to its data accessibility.

The accuracy of physical scattering model-based forest height inversion with single-baseline single-polarization InSAR is restricted by the limited number of solvable unknown parameters, especially in lower frequency like L-band where the ground-associated contributions become significant. In this paper, we propose a method that exploits a backscatter model with the help of GEDI and ATLAS/ICESat-2 spaceborne lidar height products as auxiliary information to reduce the number of unknowns.

In particular, the Random Volume over Ground (RVoG) model was derived by incorporating ground-related contributions (both ground-surface backscattering and double-bounce) into the the Random Volume (RV) model, which describes the relation between InSAR complex coherence and randomly oriented and distributed scatterers within a volume. This process introduces two additional unknowns rendering the forest height inversion problem underdetermined. To address this problem, the backscatter RVoG model is adopted in this work. The two above-mentioned parameters are estimated as global constants via backscatter regression, utilizing Lutan-1 backscatter and spaceborne lidar forest height at each lidar footprint. Global parameters estimated in this step are used in the pixelwise wall-to-wall mapping from Lutan-1 InSAR complex coherence to forest height through using the interferometric RVoG model.

Lutan-1 bistatic data with HH polarization at Hainan Tropical Rainforest National Park was selected to validate this method for complicated tropical forest. Standard SAR and InSAR processing were performed on the Lutan-1 data to obtain backscatter power, corrected coherence amplitude and interferometric phase. The underlying digital terrain model (DTM) was obtained by using the few-look InSAR phase height histogram method developed in our previous work. The physical scattering model-based forest height inversion method described above was then applied on a pixel-by-pixel basis to estimate forest height.

Validation against ALS lidar measurements demonstrates the proposed method's strength. It effectively corrects the significant global underestimation seen with the RV model, achieving a negligible bias (0.39 m) and a RMSE of 5.65 m in contrast to the -2.48 m bias and 5.76 m RMSE of RV model. Performance, however, varies by vegetation height, which was analyzed in three distinct height ranges. The model shows high linearity and accuracy for medium-height trees (15–30 m), as the global parameters are optimized around this average. However, it tends to overestimate short vegetation (<15 m). This is attributed to the model's increased sensitivity to decorrelation and potential contamination from non-forest targets like crops, where the applied forest SAR/InSAR scattering model is invalid. For taller trees (≥30 m), the method underestimates height. This is identified as an intrinsic property of the L-band's deep penetration capability when observing the sparse upper canopies, resulting in a loss of sensitivity of the complex coherence to height. As a result, tall trees are indistinguishable from medium ones.

Accurate retrieval of snow depth (SD) and snow water equivalent (SWE) is critical for understanding seasonal snow dynamics, meltwater contribution, and hydrological variability across mountainous terrain. While several studies have employed quad-polarized SAR systems for snow parameter retrieval, the application of free source dual-polarized C-band Sentinel-1 SAR for quantitative SD and SWE estimation remains limited. This study introduces a multi-level Polarimetric SAR-based framework that utilises Sentinel-1 dual-polarization observations for snow parameter retrieval over the Dhundi–Manali region of the western Himalaya, with cross-study using SNOTEL Colorado and Salt Lake regions, USA and SnowEx datasets.

The framework was developed in three progressive stages: (1) a SAR GRD-based model employing VH/VV backscatter ratios between snow and reference (non-snow) acquisitions to capture surface scattering variations; (2) a polarimetric C-matrix decomposition model, designed to quantify dielectric and structural anisotropy of the snowpack; and (3) a coherence-assisted PolInSAR extension, utilizing multi-temporal Sentinel-1 acquisitions to evaluate temporal coherence decay associated with snow metamorphism. The integration of coherence with polarimetric parameters enhances the sensitivity of Sentinel-1 to vertical snow structure and wetness evolution.

Results indicate that the cross-polarized (VH) coherence exhibits a strong negative correlation with in-situ snow depth (r ≈ 0.65) under dry to moderately wet snow conditions. The empirical models achieved an RMSE of approximately 10–15 cm, demonstrating good retrieval accuracy across varying topography and snow types. Application of the same empirical relationship over the other alpine sites gave better model transferability and comparable sensitivity trends across contrasting climatic conditions.

Overall, the study provides demonstrations of applying PolInSAR principles to dual-polarized Sentinel-1 SAR data for quantitative estimation of SD and SWE. The proposed methodology establishes a scalable, all-weather, and temporally consistent approach for snowpack characterization at 50–100 m spatial resolution, offering substantial potential for regional-to-continental cryosphere monitoring.

Precise quantification of the spatial and temporal variability of surface soil moisture (SSM) is crucial for crop monitoring and agricultural water resource management. Synthetic Aperture Radar (SAR) offers distinct advantages for surface soil moisture (SSM) retrieval, including high spatial resolution, strong backscatter signal strength, coherence properties, and greater penetration capability, etc. However, the development of stable and reliable soil moisture products using SAR remains in a nascent stage and has yet to reach full maturity. The challenges include the complexity of interpreting scattering mechanisms, the insufficient revisit frequency of current SAR systems for adequately capturing the spatiotemporal variability of SSM, and the ongoing development of retrieval algorithms for converting SAR observations into accurate soil moisture estimates. This study set out to develop a high accuracy soil moisture retrieval algorithm over crop vegetation areas utilizing polarimetric SAR data. In our work, a two-component polarimetric target decomposition model is used for the removal of vegetation scattering contribution. Subsequently, the soil dielectric constant is estimated using the alpha approximation model with the time-series soil surface scattering coefficients. Regarding the ill-posed issue of time-series soil moisture estimation method, we proposed to introduce the constraints of soil permittivity with the minimum and maximum coarse-resolution microwave surface soil moisture products. Time-series L-band UAVSAR datasets collected by SMAPVEX12 campaign, and the CCI combined soil moisture data product at a resolution of 0.25 degree in an agricultural region south of Winnipeg, Manitoba (Canada) are used for the application and accuracy assessment of the proposed soil moisture algorithm. Comparisons between SAR-derived SSM estimates and in-situ measurements over Manitoba showed retrieval accuracies with root mean square errors (RMSE) ranging from 0.03 to 0.08 cm^3/cm^3 and correlation coefficients (R) between 0.5 and 0.86 across canola, corn, soybean, wheat, and winter wheat fields. In addition, the proposed algorithm is applied to high-resolution SSM retrieval and mapping over the Huanghuaihai Plain agricultural area in China to validate the effectiveness of the algorithm. Sentinel-1 backscatter measurements and SMAP 9 km daily SSM data (SPL3SMP_E) from January 1 to December 31 in 2022 are collected for this study. Experimental results show that 0.04 ≤ RMSE ≤ 0.16 cm^3/cm^3, and 0.41 ≤ R^2 ≤ 0.84. The proposed SM retrieval algorithm that integrates the advantages of active and passive observation data not only enables high-resolution SSM retrieval at field-scale and region-scale but also improves the accuracy of SSM retrieval results. This study demonstrates the potential of the proposed technique to produce high-resolution global soil moisture products, particularly with future L-band NISAR and P-band BIOMASS satellite missions.

Accurate, high-resolution mapping of above-ground forest biomass density (AGBD) is essential for carbon accounting, sustainable forest management, and ecosystem monitoring. While missions such as GEDI, NISAR, BIOMASS, and the ESA CCI Biomass project have advanced global biomass estimation, producing consistent, wall-to-wall, and accurate products with both high spatial and temporal resolution across diverse ecosystems remains a major challenge.

A complementary approach exploits TanDEM-X digital elevation models (DEMs) together with coarse-resolution digital terrain models (DTMs), where the DEM–DTM difference serves as a proxy for canopy height and biomass. With the NASA–ISRO NISAR mission, the ESA BIOMASS mission (successfully launched on 29 April 2025 and set to deliver a global DTM), and the forthcoming ESA ROSE-L mission, synergistic use of P-, L-, C-, and X-band SAR data offers unprecedented potential for long-term, continental-scale AGBD monitoring at sub-hectare and monthly scales.

Within the FORBEAR project, we produced national AGBD maps for Spain, Poland and the Netherlands using the 12 m TanDEM-X DEM, national lidar DTMs, and simple one- and two-parameter models. Comparisons with national forest inventory data show that AGBD can be estimated accurately using simple models that leverage the strengths of individual datasets. We also analysed Sentinel-1 polarimetric backscatter and repeat-pass temporal coherence time series over forest dynamics plots, revealing distinct spatial and seasonal patterns linked to forest structure and moisture variability.

With the recent launches of BIOMASS and NISAR, and the upcoming ROSE-L mission, a clear roadmap is emerging toward accurate, frequent, and synergistic AGBD mapping from multi-frequency SAR systems. This presentation will show our recent results and outline the path toward global, operational biomass monitoring.

The novel ESA Biomass satellite utilizes a P-band synthetic aperture radar sensor, whose signals can penetrate forest canopies and also interact with understory woody elements such as trunks, branches, and stems. Together with its specific observation scenario, the Biomass sensor collects these signals multiple times over the same area, allowing the mapping of vegetation height and vertical structure, which is beyond the forest cover readily derived from optical and higher-frequency radar sensors. Despite these characteristics, biomass ground data are also required to calibrate those signals and build more reliable machine learning models of forest biomass and forest height. Here, we present Geo-Quest, our free citizen-science mobile app, which includes two modules, Tree-Quest and Plot-Quest, that enable open crowdsourced ground data collection of single-tree and forest-plot attributes, including tree height, stem diameter, and species identification, all of which are necessary for estimation of aboveground biomass. Our results highlight that our app can be used to estimate tree height and stem diameter with a relative accuracy of 11% compared to measurements obtained with traditional forest inventory tools. Our first citizen science campaign, launched at the Living Planet Symposium 2025 in Vienna, engaged over 200 participants and collected more than 1100 ground observations of tree height, diameter, and species. These results show that the app is suitable for global citizen science campaigns, which can be used to provide unique open crowdsourced datasets for calibration and validation of ESA Biomass satellite products.

Synthetic Aperture Radar Tomography (TomoSAR), by virtue of its capability in three dimensional resolution, can be used to study the three dimensional structure of semi-transparent targets like forests, icebergs and snowpacks. Currently, TomoSAR measurements, especially spaceborne TomoSAR, are mostly realized by repeat-pass observations, which brings out two major problems: the first one is temporal decorrelation, and the second one is signal delay caused by troposphere or ionosphere. Severe temporal decorrelation and signal delay can lead to defocused tomogram, which make it impossible to reconstruct the three dimensional structure of a target. Unlike repeat-pass TomoSAR system, multi-static TomoSAR system collects multibaseline images simultaneously, reducing temporal decorrelation to zero and cancelling out all kinds of signal delay, which make it an ideal tool for TomoSAR three dimensional reconstruction. Hongtu-1 X-band SAR constellation, launched in 2023 and operated by PIESAT Information Technology Limited, is the world’s first spaceborne multi-static SAR system. In this paper, we conduct spaceborne multi-static TomoSAR processing and forest height estimation experiment using Hongtu-1 multi-static images. By comparing the tomograms from tropical forest and temperate forest, it is found that the X-band signal of Hongtu-1 cannot reach the ground in dense tropical forest, but it can reach the ground in temperate forest due to much smaller tree and leaf density, which means Hongtu-1 is capable of forest height measurement in temperate forest. The forest height inversion experiment is carried out in Saihanba forest by high resolution airborne LiDAR, Hongtu-1 multi-static TomoSAR and GEDI. Using the airborne LiDAR CHM as reference, it shows that Hongtu-1 TomoSAR measurements can provide more accurate forest height inversion (with 55% RMSE improvement, from 5.6m to 2.5m), much more measurement points and higher resolution product than GEDI, which proves the capability and superiority of Hongtu-1 TomoSAR in forest height estimation.

Forest disturbances such as fires, windthrows, and bark beetle outbreaks have major impacts on forest condition and ecosystem functioning. Changes in canopy structure and dielectric properties resulting from these events can serve as valuable indicators for studying disturbance processes using radar remote sensing.

Radar interferometric coherence is particularly sensitive to temporal and geometric variations in scattering mechanisms. This study investigates how the temporal and perpendicular baselines of image pairs contribute to coherence changes, compared to those caused by actual forest disturbances. The analysis focuses on Central European forests affected by various disturbance types and is based on Sentinel-1 SLC data, from which coherence maps are derived for a range of baseline combinations. Coherence dynamics are evaluated over disturbed and reference forest stands to separate baseline-related effects from disturbance-induced changes.

Preliminary statistical analyses of coherence changes across different baseline configurations provide a quantitative assessment of the relative influence of geometric and temporal decorrelation. The results help identify optimal baseline ranges for reliable detection of forest disturbances and support improved interpretation of coherence changes in operational forest monitoring.

Understanding the global carbon budget, including its carbon sources and sinks, is scientifically important and economically relevant. Vegetation and soils are major dynamic carbon pools in the Earth System, and a substantial part of the terrestrial carbon budget is influenced by land use changes, vegetation dynamics, and soil processes.

Recent advances in Foundation Models (FMs) are transforming AI, enabling remarkable generalization and zero-shot learning capabilities. Within the Helmholtz Foundation Model Initiative, we are developing the 3D-ABC FM to target the accurate mapping of above- and below-ground carbon stocks in vegetation and soils at high spatial resolution. 3D-ABC aims to provide a comprehensive view of terrestrial carbon distribution by integrating multimodal datasets, including remote sensing, climate and elevation datasets, while addressing challenges such as varying spatial resolution and multi-dimensionality in FMs.

The 3D-ABC FM combines large-scale remote sensing data, including multispectral imagery from the Harmonized Landsat-Sentinel-2 (HLS) dataset, TanDEM-X InSAR coherence data, and 3D lidar data from space (GEDI, ICESat 1&2), aircraft, and ground-based platforms. The TanDEM-X data provides coherence and interferometric information on vegetation structure as well as forest and soil parameterization. We aim to include ERA-5 Land climate reanalysis information, GLO-30 digital elevation data, and local lidar and field measurements on vegetation, soils, and carbon fluxes. High-resolution forest models will be employed to benchmark and validate carbon fluxes.

To accommodate the diverse data modalities assembled for 3D-ABC and to support eight downstream tasks, the AI model uses an adaptive architecture. This consists of a multi-modal input processor, an FM encoder, an adaptive fusion neck, and task-specific prediction heads. The multi-modal input processor handles data with varying spectral dimensions, automatically mapping inputs into a unified feature space. The encoder extracts generalized deep features from the normalized inputs. These features are integrated into universal feature representations through the adaptive fusion neck, enhancing interactions across modalities, before the universal features are decoded into outputs tailored to the specific downstream tasks. Model training proceeds in two phases. In the first phase, a masked autoencoder is used to pretrain the input processor, encoder, and fusion neck in an unsupervised manner, allowing the model to develop robust feature representations. In the second phase, using the principles of transfer learning, the pretrained network is fine-tuned using labeled datasets from the downstream tasks.

3D-ABC primarily targets use of two high-performance computing (HPC) systems located at the Jülich Supercomputing Centre (JSC): the JUWELS Booster and JUPITER. The JUWELS Booster comprises 936 compute nodes, each equipped with four NVIDIA A100 GPUs. JUPITER, the first European exascale supercomputer, is currently being installed at the JSC. Its Booster module will consist of approximately ~6,000 compute nodes, each featuring four NVIDIA GH200 GPUs. To maximize efficient JUPITER utilization, 3D-ABC is leveraging the JUPITER Research and Early Access Program, which provides early access for code optimization and preparation.

Currently, activities are being carried out in the training of the foundation model using GEDI footprints, HLS data and TanDEM-X coherence to obtain a forest height map over the amazon forest area as a downstream task. Hence, TanDEM-X data for more than 900 HLS tiles have been prepared by mosaicking interferometric coherences calculated at 20 m multilook resolution from more than 20000 acquisitions since January 2018. First results will be presented at the workshop, and the related challenges and future perspectives will be addressed.

3D-ABC Team: Josh Hashemi(AWI), Lona van Delden(AWI), Tillmann Lübker(AWI), Ingmar Nitze(AWI), Jens Strauss(AWI), Stefan Kruse(AWI), Ulrike Herzschuh(AWI), Peter Steinbach(HZDR), Gunjan Joshi(HZDR), Weikang Yu(HZDR), Aldino Rizaldy(HZDR), Richard Gloaguen(HZDR), Ehsan Zandi(FZJ), Rocco Sedona(FZJ), Samy Hashim(FZJ), Sayan Mandal(FZJ), Qian Song(GFZ), Simon Besnard(GFZ), Mikhail Urbazaev(GFZ), Mike Sips(GFZ), Leonard Schulz(UFZ), Matteo Pardini(DLR), and Kostas Papathanassiou(DLR)

The snow water equivalent (SWE) describes the amount of water stored in a snow pack. The retrieval of SWE with SAR methods has mostly been explored for snow over land. This study seeks to explore SAR-based SWE accumulation retrieval on glaciers.

Differential Interferometric SAR (D-InSAR) combines temporally separated acquisitions that ideally have no spatial baseline. A common D-InSAR assumption is that scatterers stay constant in between acquisitions. This assumption is already debatable for snow over land, but requires even more consideration on glaciers. When it holds, the D-InSAR phase is solely a propagation effect from SWE change and an established inversion method for snow over land exists. The objective of this study is to better understand what, besides the SWE-induced propagation effect, needs to be accounted for in SWE change retrieval with D-InSAR over glaciers.

For this, polarimetric SAR (PolSAR) plays an important role. The copolar phase difference (CPD), which is the phase difference between HH and VV polarizations, can be related to the snow height. The polarimetric scattering angle α describes scattering mechanisms independent of line-of-sight rotation. Possible values are 0° ≤ α ≤ 90° where angles close to zero tend to relate to surface scattering, while angles closer to 90° correspond to dihedral scattering events. If scatterers do not change in between acquisitions, even if there was snowfall, α should remain constant, while the new snow should give a change in CPD. Further, changes in scattering mechanisms and snow accumulation should manifest differently across frequencies in α and CPD, where in an ideal scenario, the CPD should scale linearly with wavelength, while scattering mechanisms are generally frequency dependent.

The data used in this study stems from two fully polarimetric, airborne SAR campaigns of the Aletsch Glacier in Switzerland in 2022 and 2024 conducted by DLR and partners. Both campaigns entail short time series with DLR’s F-SAR sensor in X-, C-, and L-band over the course of multiple weeks. Snow accumulation and density were measured at several locations during the campaigns and corner reflectors were placed as reference targets. In 2024, snow anisotropy measurements were carried out by SLF. In the campaigns, the weather conditions differed greatly, with warm temperatures in 2022 and cooler, more stable conditions with more fresh snow in 2024. As a first assessment of the constant-scatterer assumption, the scattering angle α was analyzed for the 2022 and 2024 Aletsch campaigns.

In the polarimetric analysis of the 2022 data changes in α are observed, fitting a melting event that took place between two acquisitions, resulting in more surface scattering. After a snowfall event, α displays more volume scattering. In contrast, the 2024 acquisitions barely differ in scattering mechanisms. Constant cold temperatures and new snow relate to a very stable α. The contrast of these two campaigns shows that while the assumption of constant scatterers does hold for 2024, it would be incorrect for 2022. Therefore, α is an important parameter in deciding where D-InSAR SWE retrieval is reasonable.

Further research should help understand what assumptions common to snow over land hold up for snow over glaciers, and how polarimetric SAR can help to better understand influences on the D-InSAR phase. This will facilitate the model development towards D-InSAR SWE change retrieval over glaciers.

The resilience of terrestrial ecosystems to droughts and stress is key for the future of the terrestrial carbon balance. Satellite observations of sub-daily variations in vegetation water content (VWC) could provide us information on health, stress and resilience of key ecosystems across the globe. Water dynamics in vegetation are central in these ecosystems, as they are closely coupled to carbon assimilation at the plant stomata. Understanding diurnal variations in VWC provides insight into the water status, stress and health of plants. However, sub-daily water dynamics in ecosystems are still poorly understood and weakly represented in terrestrial biosphere models. Furthermore, there are no existing or planned satellite missions capable of resolving fluctuations in VWC on sub-daily scales.

To address this critical knowledge and observation gap, SLAINTE was being developed as one of ESA’s New Earth Observation Mission Ideas with a first mission concept submitted in response to ESA’s 12th Call for Earth Explorers (Steele-Dunne et al., 2024, Matar et al., 2024). It comprised a constellation of identical, decametric, monostatic SARs to capture sub-daily variations in vegetation water storage (e.g. via vegetation optical depth (VOD), vegetation water content (VWC), plant water potential (PWP) and surface soil moisture (SSM). One of the key challenges during the development of the SLAINTE mission concept was the limited availability of sub-daily radar backscatter data, both real and synthetic. Real observations are restricted to a few tower sites. Generating synthetic data, on the other hand, requires sub-daily vegetation water content information that is difficult to obtain through destructive sampling, as it is both expensive and logistically demanding.

Here, this challenge is addressed by using continuous, non-destructive ground-based measurements of vegetation water storage dynamics at various forested sites throughout Europe. These observations provide the high-temporal-resolution input needed to drive a physically-based, bistatic and polarimetric radiative transfer (RT) model. The simulations produce synthetic time series of radar backscattering coefficient at sub-daily resolution, allowing us to quantify and characterize the influence of sub-daily variations in plant water dynamics, vegetation structural parameters and biogeophysical properties on the backscattering coefficient. Disentangling and isolating pertinent signals (such as plant internal and surface water content) from confounding factors (e.g. forest geometry, vertical water redistribution effects), will allow us to better understand the sensitivity of radar backscatter to key variables. In this presentation, we will show initial results from simulations of Loobos, a forested site in the Netherlands. We will use these results to demonstrate the value of RT simulations to further strengthen the science case, consolidate observation and measurement requirements and develop retrieval approaches for future sub-daily SAR missions.

References:

Matar, J., Sanjuan-Ferrer, M. J., Rodriguez-Cassola M., Steele-Dunne, S. & De Zan, F. (2024). A Concept for an Interferometric SAR Mission with Sub-daily Revisit. EUSAR 2024; 15th European Conference on Synthetic Aperture Radar, pp. 18-22. IEEE, 2024.

Steele-Dunne, S., Basto, A., De Zan, F., Dorigo, W., Lhermitte, S., Massari, C., Matar J. et al. (2024). SLAINTE: A SAR mission concept for sub-daily microwave remote sensing of vegetation. EUSAR 2024; 15th European Conference on Synthetic Aperture Radar, pp. 870-872. VDE, 2024.

This study investigates the advantage of combining structural information obtained from the newly launched ESA BIOMASS (P-band) Synthetic Aperture Radar (SAR) with spectral information derived from PRISMA hyperspectral data to assess forest diversity in the Amazon region of Brazil. The study area was selected based on the overlap of available PRISMA and BIOMASS acquisitions.

Structural diversity was estimated from BIOMASS data (6 June 2025) using polarisation variability, Polarimetric SAR (PolSAR) metrics, and texture of above-ground biomass to represent differences in forest vertical structure and stand complexity. The cross-polarised backscattering coefficient (HV/VH) is particularly sensitive to forest height variations due to volume scattering, providing key information on canopy structure and scattering mechanisms.

Surface reflectance from PRISMA Level-2D (29 July 2025, 234 bands, 406–2497 nm) was used to derive spectral diversity indicators, including Rao’s Q, spectral variance, and clustering-based “spectral species.” These metrics describe variability in canopy composition and biochemical traits that reflect species and functional diversity across heterogeneous forest areas.

The analysis evaluates relationships between spectral diversity from PRISMA and structural diversity from BIOMASS, examining how compositional and structural heterogeneity correspond across forest environments. Available LiDAR canopy height data and, where possible, field observations of species composition and functional traits will serve as supporting reference information. Correlation and multivariate analyses will assess the consistency and complementarity of spectral and radar-derived indicators.

This multi-modal EO approach aims to advance satellite-based monitoring of forest biodiversity across tropical ecosystems, highlighting the potential of the ESA BIOMASS mission for forest applications.

Water scarcity is a growing problem in the southern Mediterranean region, identified as a climate change hotspot. The agricultural sector is most threatened by water scarcity, as it is the largest consumer of freshwater, mainly extracted from groundwater. Morocco is an illustrative example of such a problem where groundwater resources are under anthropogenic and climatic pressures: overexploitation due to water pumping and lack of precipitation and long periods of drought.

The subsurface is a critical zone where water exchanges between the surface and underground layers take place. This is why considerable efforts have been devoted to monitoring underground soil moisture using in situ measurements. Despite their high accuracy, in situ measurements cannot cover the high spatial and temporal variability of soil moisture. Remote sensing data are used as an alternative to estimate underground moisture on a large spatial scale, as in the case of GRACE (Gravity Recovery and Climate Experiment) data or synthetic aperture radar (SAR) data in the C or L bands.

Due to the limited penetration of C-band or L-band signals into scatters, the P-band is used for estimating subsurface soil moisture. In this context, this project aims to explore the potential of P-band radar data for estimating subsurface water content using the first and exclusive BIOMASS satellite acquisitions in P-band in two of Morocco's most important watersheds (Oum Er-Rbia and Tensift). These watersheds are equipped with time-domain reflectivity sensors and piezometers that provide a vast dataset for monitoring soil moisture at different depths and water levels, respectively. The in situ measurements will be collected at the same time as BIOMASS acquisitions and will be used to analyze tomographic (TomSAR) and interferometric (InSAR) data sensitivity to the dynamics of subsurface soil water content. The sensitivity of P-band data will be compared to that of C-band radar data. The detection and mapping of groundwater content are among the objectives pursued using TomSAR and InSAR data, through different models. The results of this project will help stakeholders make informed decisions based on subsurface soil water content monitoring in order to effectively govern the risk of water resource depletion and, consequently, ensure sustainable water management.

Desert regions such as Bir Safsaf in Egypt conceal a complex palaeohydrological history, with ancient riverbeds (palaeochannels) buried beneath arid surfaces. Mapping these features is essential for reconstructing past hydrological networks, understanding palaeoclimatic evolution, and supporting resource management in hyper-arid environments. However, their detection is challenging due to sediment cover, surface roughness, and the limited effectiveness of optical remote sensing.

This study investigates the use of polarimetric Synthetic Aperture Radar (SAR) data and differential entropy analysis to identify and map palaeochannels in the Bir Safsaf region. The methodology leverages the complementary capabilities of multiple SAR sensors: Sentinel-1 (C-band, VV and VH polarizations), ALOS PALSAR (L-band, HH and HV polarizations), and, prospectively, BIOMASS (P-band, pending availability of calibrated data). Each sensor offers distinct penetration depths and sensitivities to surface and subsurface features, while their polarimetric channels provide diverse information about scattering mechanisms.

Central to the approach is the computation of the differential entropy — a statistical measure of the complexity or randomness of radar backscatter distributions — applied to the polarimetric channels. For Sentinel-1, entropy is calculated using both VV and VH channels; for ALOS, HH and HV are used. Importantly, the methodology allows for the computation of differential entropy not only on individual channels but also on combinations of polarizations across different sensors (e.g., VV from Sentinel-1 with HH from ALOS). This cross-sensor, cross-polarization analysis enhances the sensitivity to subtle textural and structural variations that may indicate the presence of buried or relict fluvial features.

By exploiting the diversity of polarimetric information, the approach aims to improve the discrimination of palaeochannel signatures from the surrounding desert matrix. The fusion of entropy maps derived from various polarization combinations is expected to highlight features that may remain undetected in single-polarization or single-sensor analyses. This is particularly valuable in desert contexts, where surface and subsurface contrasts are often faint and spatially heterogeneous.

Ancient riverbeds in desert environments are often characterized by subtle textural and structural variations that are barely distinguishable from sensor noise in conventional SAR imagery. The features of interest frequently lie at or below the noise floor, making their detection particularly challenging. By leveraging differential entropy analysis across multiple polarimetric channels and sensor combinations, the proposed methodology enhances the signal-to-noise ratio (SNR), increasing the likelihood of revealing these faint palaeochannel signatures within the noisy radar background.

The workflow involves preprocessing and calibration of SAR data, computation of local differential entropy for each polarization and selected combinations, and the integration of these entropy maps to generate enhanced indicators of palaeochannel presence. The methodology is designed to be robust to surface cover variability and atmospheric conditions, making it suitable for application in hyper-arid regions where optical methods are limited.

The study will apply this framework to the Bir Safsaf region, with the integration of BIOMASS data considered if calibrated products become available by the time of the conference. As the research is ongoing, the abstract focuses on the methodological framework and the potential of polarimetric differential entropy analysis for desert geomorphology. Preliminary results and case studies will be presented if available.

Monitoring flood dynamics within forested and densely vegetated regions remains a major challenge due to the limited penetration of conventional radar frequencies through canopy layers. This study investigates the comparative potential of P-band synthetic aperture radar (SAR) observations from the ESA BIOMASS mission and C-band data from Sentinel-1 for detecting flooded areas beneath vegetation cover. The analysis is conducted within the BIOMASS Calibration and Validation (Cal/Val) framework, aiming to assess how multi-frequency SAR observations can enhance hydrological and ecohydrological monitoring in complex environments.

P-band radar, operating at longer wavelengths, provides deeper penetration through vegetation canopies and upper soil layers compared to C-band sensors. This characteristic enables BIOMASS to capture hydrological features that are often obscured in Sentinel-1 imagery, such as subsurface and under-canopy inundation. By comparing backscatter and polarimetric responses from BIOMASS and Sentinel-1, this study quantifies differences in sensitivity to vegetation water content, soil moisture, and surface water extent. Temporal and spatial variations in SAR backscatter are analyzed to evaluate the complementary roles of both frequencies in characterizing flood dynamics, particularly in forested floodplains and wetland ecosystems. The comparison is extended to SWOT High-Rate products (with Ka-band Interferometry) providing water surface elevation and extent.

Flood detection methods, including change detection, coherence analysis, and polarimetric decomposition, are applied to both datasets. The integration of these techniques enables improved discrimination between open water, flooded vegetation, and dry land surfaces. In particular, change detection using time-series P-band data enhances the identification of inundated zones beneath vegetation, while Sentinel-1 contributes high temporal resolution for near-real-time monitoring. The combination of multi-frequency SAR data, supported by ancillary datasets such as optical imagery, topography, and in-situ hydrological measurements, allows for a comprehensive assessment of surface and subsurface water processes.

Automatic mapping of land cover types on flooded Riparian landscapes is one of the most challenging problems in remote sensing. Although Object based approach has been embraced in Land Use Land cover studies few studies have applied it on riparian vegetation discrimination. In order to improve on the accuracy and efficacy of any classification, new approach to data collection and extraction has become increasingly necessary. The study investigated an Object based classification based on Sentinel 1 Single Look Complex (Synthetic Aperture Radar) data using three classifiers namely Naıve Bayes, Decision Tree and Random Forest. Four land cover types i.e. Acacia Forest, Built-up, Grasslands, water and others were successfully retrieved. Change detection was noted on the North western strands having reduced from an area of 212.7 hectares in 2008 before the floods to an area of 64.64 hectares in 2019 after floods. ALOS-1 Level 1.1 was used as reference image captured in 2008 before floods and Sentinel 1 SLC data captured in 2015, 2017 and 2019 after the floods, Acacia Forest strands were captured to have been degraded especially on the North western part of Lake Nakuru.Classification Results based on Machine learning and Polarimetric Matrix generated bands C11, C22 and their ratio were used to obtain the results as follows; Naive Bayes 91.1% , Decision Tree 94.1% while Random Forest took the lead by 94.4%. One-way Analysis of Variance (ANOVA) was used to compare variations among the group algorithms. However, there was no significant difference between and amongst the Classifier performance since F,2,15=0.529 with a p-value of 0.60 was achieved

Index Terms—Synthetic Aperture Radar, Object Detection, Machine Learning, Change Detection, flooding

Gabon's tropical forests, estimated to cover 89% of the country’s territory [1], play a major role in the global carbon cycle and climate regulation, particularly in terrestrial carbon storage. The ESA BIOMASS satellite, operating in the P-band, offers a unique opportunity to estimate the aboveground biomass and monitor forest dynamics on a global scale [2]. Long wavelength and 3-days repeat-pass of BIOMASS, allow to acquire the high-quality interferograms.

Interferometric coherence is intrinsically sensitive to the volumetric distribution of scatterers inside the resolution cell, and, therefore, also sensitive to the forest height and stored biomass. However, the interferometric coherence, is subject to temporal decorrelation, that can be either wind-induced or caused by dielectric changes in the scattering volume. The propagation through the ionosphere requires additional calibration processing steps, that can complicate the interpretation of temporal changes in the reflectivity volume.

In this context, the new AfriSAR airborne campaign is conducted in Gabon in November 2025, at the end of BIOMASS in-orbit commissioning. The campaign data allow to assess the temporal stability of P-band radar coherence during the BIOMASS acquisition, matching the temporal and interferometric baselines of the satellite.

This study therefore aims to quantify, model, and interpret the temporal decorrelation of the P-band radar signal using fully polarimetric and interferometric/tomographic airborne datasets and compare it to the one of BIOMASS. In addition, the long-term temporal coherence of several months and years will be assessed based on the previous AfriSAR campaigns in 2016 and 2023.

References

1. Cartographie de l’occupation du sol du Gabon en 2015 - changements entre 2010 et 2015 - Scientific Figure on ResearchGate. Available from: https://www.researchgate.net/figure/Identification-des-forets-inondees-et-des-prairies-aquatiques-par-seuillage-de-la-bande_fig15_355287478 [accessed 21 Feb 2025]

2. Quegan et.al., The European Space Agency BIOMASS mission: Measuring forest above-ground biomass from space. Remote Sensing of Environment, 227, 44-60, 2019

3. I. Hajnsek et al., "Technical assistance for the development of airborne SAR and geophysical measurements during the AfriSAR campaign, Final technical report, 2016

Seasonal settlement of permafrost terrain reflects the thickness and ice content of the active layer. Such changes can be quantified through differential interferometric SAR (InSAR). Polarimetric SAR (PolSAR), in turn, is sensitive to scattering mechanisms related to vegetation structure, soil moisture, and dielectric contrast at the thaw front. Integrating these complementary observables provides a means to link subsurface processes with surface scattering behavior and to improve the physical interpretation of SAR backscatter in permafrost environments.

In this work, we exploit long-wavelength airborne SAR data (P- and L-band) acquired during the NASA ABoVE campaign over the North Slope of Alaska. Three P-band acquisitions spanning the 2017 thaw season are used to estimate cumulative surface subsidence via InSAR, isolating the net seasonal deformation associated with active layer thaw. To complement the sparse temporal sampling of the airborne data, dense Sentinel-1 C-band interferograms (12-day repeat) are employed to track the temporal evolution of subsidence and to constrain the seasonal trajectory. In parallel, polarimetric parameters and decompositions derived from the L- and P-band datasets are used to map scattering mechanisms and land-cover classes, yielding proxies for soil and vegetation conditions that govern thaw settlement and drainage.

By jointly analyzing the interferometric deformation observations and polarimetric indicators, we examine how scattering behavior varies with the magnitude and spatial pattern of subsidence across contrasting tundra types, including ice-rich polygonal terrain. The results demonstrate the potential of combined InSAR–PolSAR observations to link geophysical and electromagnetic properties of permafrost landscapes. This work anticipates the capabilities of the recently launched spaceborne missions such as NISAR (L-band) and BIOMASS (P-band), which will enable global, long-wavelength monitoring of permafrost thaw dynamics and related carbon–climate feedbacks.

Aboveground biomass density (AGBD) is a key indicator of crop productivity and carbon storage. However, accurate regional estimation remains challenging because optical remote sensing suffers from cloud contamination, while process-based crop growth models, such as WOFOST, require extensive calibration and often fail to capture spatial heterogeneity. To improve AGBD estimation performance and operational flexibility, this study integrates the complementary strengths of cloud-insensitive microwave remote sensing and crop growth models through a SAR-driven, simple-calibration assimilation framework. AGBD is estimated from dual-polarized Sentinel-1 radar backscatter calibrated with field measurements. A modified and simplified version of the WOFOST potential production (WOFOST-PP) model was proposed, which reduces data requirements and calibration complexity. Pixel-wise WOFOST parameters are updated via data assimilation by minimizing the least-squares difference between SAR-derived AGBD and model outputs, improving their spatial heterogeneity representation and accuracy.

Validation with multi-year cotton observations from two contrasting farms in Georgia, USA (rainfed ACF; irrigated TCF) shows that assimilation configuration achieves the best performance compared with both WOFOST-PP and SAR-derived estimates. In WOFOST-PP, assimilated carbon is partitioned among leaves, stems, and storage organs; the dry weight of storage organs, a subset of AGBD, is defined as yield and enables computation of the harvest index (HI). Pixel-level maps of AGBD, yield, and HI capture spatial heterogeneity and maintain operational coverage under frequent cloudiness, where optical data are sparse. Spatial fields of YRF reveal persistent low-efficiency zones suitable for targeted management, enabling earlier stress detection than end-of-season diagnostics. The SAR-driven, simple-calibration assimilation offers a practical pathway to regional AGBD and yield mapping with limited field data. It remains competitive with optical-based systems, maintains accuracy while providing temporal robustness in cloudy conditions.

Climate change presents one of the greatest challenges to humankind, with the rapid depletion of forest ecosystems worldwide intensifying the crisis. Sanjay Gandhi National Park (SGNP), located in Borivali, Mumbai, India, stands as a rare example of a thriving green ecosystem nestled within a dense urban environment. To assess the changing dynamics between forest and non-forest areas in such critical landscapes, this study develops a hybrid Variational Autoencoder–Self Organizing Feature Map (VAE–SOFM) speckle filtering framework integrated into a polarimetric SAR application for urban forest mapping.

Dual-polarization Sentinel-1 (C-band) and ALOS-2 (L-band) datasets are utilized to exploit complementary scattering characteristics: C-band enabling high-temporal, cloud-independent monitoring, and L-band offering deeper canopy penetration for biomass sensitivity. Preliminary results demonstrate that the proposed VAE–SOFM hybrid filter achieves superior speckle suppression while retaining essential polarimetric and textural features, thereby enhancing the delineation of forest and non-forest areas using the Wishart Supervised Classifier. The integration of forthcoming NISAR mission data is expected to further enhance biomass sensitivity and improve temporal continuity in forest monitoring.

The resulting high-resolution and non-destructive assessment of forest cover dynamics in SGNP provides critical insights into deforestation patterns and vegetation changes. Such analysis will significantly contribute to policy frameworks on carbon accounting, sustainable forest management, and climate change mitigation strategies.

The increasing availability of global Sentinel-1 observations offers new opportunities for large-scale forest structural mapping. Although temporal decorrelation in repeat-pass Sentinel-1 acquisitions limits the applicability of Polarimetric-Interferometric (PolInSAR) techniques for forest height estimation, recent studies [1] have demonstrated the potential of Sentinel-1 coherence for biophysical parameter retrieval. Such approaches rely on dense or seasonal time-series data, where the ground and volume scattering components are inferred indirectly from backscatter statistics under simplified, regionally constant assumptions.

In this study, we investigate an alternative data-driven approach for forest height estimation from Sentinel-1 interferometric observables, leveraging deep learning models trained on globally distributed SAR–LiDAR data. We assembled an extensive dataset of over 900 Sentinel-1 interferometric pairs acquired between 2019 and 2024, covering diverse forest biomes worldwide. Each pair corresponds to a single-baseline configuration with a 12-day temporal baseline and dual-polarization (VV, VH) channels. Reference forest heights were derived from GEDI L2A and ICESat-2 ATL08 products, filtered by confidence and terrain criteria, and projected into SAR geometry (azimuth and slant range). This dataset represents a uniquely comprehensive resource, encompassing a wide range of forest types, forest heights, and ecological regions across multiple continents. The Sentinel-1 interferometric pairs span diverse spatial baselines as well as varying heights of ambiguity (HoA), acquired under different InSAR geometry. This diversity of the data enables the development and evaluation of machine learning models that are robust and generalizable across biomes, acquisition geometries, and environmental conditions. To achieve this objective, a deep convolutional network was developed to map elements of the polarimetric interferometric covariance matrix to forest height. Preliminary results indicate that the model successfully learns meaningful relationships between Sentinel-1 observables and LiDAR-derived forest heights, demonstrating the feasibility of data-driven forest height mapping at the global scale. Despite the effects of temporal decorrelation, Sentinel-1 interferometric observables retain measurable sensitivity to vegetation structure: coherence generally decreases with increasing forest height, although the relationship remains nonlinear and spatially variable. Unlike TanDEM-X data, where simplified RVoG formulations can directly relate coherence to height, Sentinel-1 coherence requires alternative modeling strategies that can capture its more complex behavior.

This work contributes to ongoing efforts to integrate Sentinel-1, GEDI, and forthcoming L-band missions within the ESA’s SUPSAR framework. The research is funded by the ESA under the Sentinel2Height initiative. This initiative aims to develop data-driven approaches for global forest height mapping. Detailed results, model architecture, and assessments will be presented at the conference.

Reference

[1] Lavalle, Marco, C. Telli, Nazzareno Pierdicca, Unmesh Khati, Oliver Cartus, and Josef Kellndorfer. "Model-based retrieval of forest parameters from Sentinel-1 coherence and backscatter time series." IEEE Geoscience and Remote Sensing Letters 20 (2023): 1-5.

Airborne Synthetic Aperture Radar (SAR) systems have greatly advanced remote sensing by providing high-resolution, all-weather, day-and-night imaging capabilities, enabling detailed analysis of Earth’s surface features. Among emerging SAR techniques, Polarimetric SAR Tomography (PolTomSAR) has proven highly effective for reconstructing three-dimensional representations of forests, urban landscapes, and natural terrains by exploiting multi-polarization and multi-baseline acquisitions. This study presents the application of PolTomSAR for forest vertical structure estimation using L-band UAVSAR datasets acquired during NASA’s AfriSAR airborne campaign. The selected data correspond to the Rabi Forest region in Gabon, collected on June 15, 2015, with an azimuth resolution of 1.5 m and a range resolution of 12 m. Advanced tomographic inversion algorithms, including classical beamforming and adaptive Capon spectral estimation methods, were applied to resolve vertical scattering mechanisms within the forest canopy. The reconstructed tomographic profiles indicate an average forest height of approximately 31 m, consistent with the characteristics of dense tropical vegetation. Validation was performed using a Digital Surface Model (DSM) derived from LiDAR point cloud data acquired by the Land, Vegetation, and Ice Sensor (LVIS), showing strong agreement with PolTomSAR-based estimates. The results demonstrate the robustness of SAR tomography for detailed forest height mapping and three-dimensional canopy characterization, highlighting its potential for large-scale forest monitoring, biomass estimation, and climate-related ecosystem studies.

Accurate estimation of above-ground biomass (AGB) is essential for understanding the global carbon cycle and monitoring the impacts of deforestation, forest degradation, and land-use change, which significantly contribute to greenhouse gas emissions and influence climate regulation at regional and global scales.

In this context, Synthetic Aperture Radar (SAR) data represents a powerful and reliable source of Earth observation, offering cloud- and weather-independent, day-and-night imaging capabilities that make them particularly well-suited for continuous and large-scale forest monitoring.

While the Biomass mission provides unprecedented P-band SAR observations for forest structure and biomass retrieval, its spatial coverage is limited, excluding regions such as North America and parts of Europe. In contrast, Sentinel-1 offers global C-band SAR coverage, which can serve as a complementary data source to transfer algorithmic knowledge beyond the Biomass acquisition zones.

This study explores the potential of context learning, a paradigm in which an embedding model enables downstream algorithms to adapt to new, unseen domains by leveraging environmental or contextual similarities between regions. It involves mapping satellite observations to ground truth data through multimodal learning, allowing the retrieval and transfer of relevant settings information to regions where in situ measurements are sparse or unavailable, while simultaneously enabling the handover of relevant satellite-derived information to areas where specific radar observations are not accessible.

A first embedder will be initially trained on Sentinel-1 data taking advantage of its global coverage to assess the ability of the model to retrieve relevant satellite information from environmental variables. Subsequently, Biomass P-band data will be used to train a second embedder with the final goal to retrieve contextually informed Biomass products for regions lacking direct P-band coverage, using bioclimatic affinity rather than geographic proximity. This approach would enable the application of P-band–derived biomass and structural estimates in areas such as North America and Europe, by transferring insights from ecologically comparable regions. The technique applies the concept of knowledge transfer through contextual correspondence, enabling the model to generalize biophysical relationships beyond direct measurement zones.

This proof-of-concept will assess the feasibility, robustness, and transferability of context learning in SAR applications, paving the way toward cross-mission AI models capable of integrating data from different radar frequencies and acquisition geometries.

Finally, by leveraging advanced data-driven methodologies and context-aware modeling approaches, this work aims to support the development of next-generation frameworks capable of integrating heterogeneous datasets, improving decision-making processes, and enabling scalable environmental monitoring at regional and global levels.

Introduction: Tropical forests play a central role in the global carbon cycle, biodiversity conservation, and climate regulation. However, monitoring their structure and carbon dynamics frequently across large and heterogeneous landscapes remains a major challenge. In Brazil, the Atlantic Forest exemplifies this complexity: a densely populated and highly fragmented biome encompassing a wide gradient of successional stages, from young regenerating stands to mature forests. Despite significant restoration efforts, the lack of consistent structural data limits our ability to assess forest quality and functionality for decision-making. Current large scale monitoring frameworks rely primarily on optical remote sensing, which maps canopy extent but poorly captures vertical structure and above-ground biomass (AGB). Optical indices tend to saturate in dense canopies, obscuring processes such as biomass stagnation among other key indicators of ecological integrity.

ESA's BIOMASS mission, equipped with a fully polarimetric P-band Synthetic Aperture Radar (SAR), represents a breakthrough in tropical forest monitoring. Its ability to penetrate dense canopies and its sensitivity to woody components are promising capabilities for enhancing and including structural indicators essential for improving forest quality monitoring at large scales.

Objectives: This study aims to evaluate the performance and complementarity of ESA’s BIOMASS mission products for monitoring forest structure and carbon stocks across diverse tropical formations of the Atlantic Forest. Specifically, it seeks to: (1) Validate BIOMASS AGB and canopy height products across successional gradients; (2) Compare BIOMASS and NASA GEDI LiDAR derived products to explore complementarities and hybrid modeling potential; (3) Explore “not intend” potential applications of BIOMASS products to enhance forest quality assessment (such as the detection of stagnant growth); and (4) Develop an applied framework for integrating multi-sensor products to monitor forest structural quality at regional scales.

Methods: We will use an extensive dataset of 700–800 georeferenced forest plots distributed across ombrophilous and seasonal Atlantic Forest formations, encompassing early to old-growth stages. A subset of these plots includes airborne LiDAR data, providing accurate benchmarks for canopy height, density, and vertical complexity.

BIOMASS Level-2 AGB and canopy height products will be validated against these references following ESA’s Cal/Val criteria (≈20% RMSE for AGB, ≈30% for height). The analysis will quantify accuracy, bias, and consistency across successional gradients and forest types. Additionally, full-polarimetric P-band data will be explored to identify parameters most sensitive to variations in structure, layering, and biomass distribution.

Expected Results: (1) A rigorous validation of BIOMASS AGB and height estimates across structurally diverse tropical forests; (2) A comparative assessment of BIOMASS and GEDI derived products performance, emphasizing complementarity and potential for integrated modeling; (3) Insights into P-band sensitivity to structural patterns beyond traditional biomass metrics; and (4) An applied multi-sensor decision-making-oriented framework for large-scale monitoring of forest structural quality.

Beyond quantitative validation, this study advances the conceptual integration of large-scale structural indicators into restoration and conservation planning. By linking remote sensing–derived metrics with ecological attributes such as maturity, successional stage, and potential growth stagnation, it supports a transition from assessing how much forest exists to evaluating how well these forests function. The results will inform ongoing initiatives under the Brazilian Atlantic Forest Restoration Pact, global restoration programs, and ESA’s Cal/Val activities, contributing to improved tropical forest monitoring worldwide.

Understanding above-ground biomass (AGB) in forests is essential for unravelling the complexities of biogeochemical cycles and climate change. Remote sensing, particularly through NASA's GEDI (Global Ecosystem Dynamics Investigation) mission, significantly boosts our ability to monitor AGB on a large scale, thanks to advanced technologies like LiDAR (Light Detection and Ranging). As for the PolInSAR (Polarimetric Interferometric Synthetic Aperture Radar) technique, it is particularly effective in measuring forest height, which in turn enhances AGB estimation across different forest types by utilising radar data and models such as the Random Volume over Ground model for better accuracy because biomass is related to forest height and her volume. Recent studies emphasise the significance of diverse data sources and sophisticated machine learning (ML) techniques in achieving precise AGB estimates, showcasing a variety of successful models and data integrations from global research. Additionally, recent research findings indicate that ML algorithms significantly enhance AGB estimation by utilising multisensor data. Our study specifically focuses on estimating AGB density in Gabon's Pongara National Park, where we fused PolInSAR, Sentinel-2, Dynamic World V1, and GEDI (specifically GEDI Level 4B Gridded AGB density) data with ML regression algorithms such as Gradient Tree Boosting (GTB), Random Forest (RF), and Classification and Regression Trees (CART). Our findings achieved the best estimation for biomass mapping using the GTB model, and alongside all models, they demonstrated a high correlation (up to 0.99) with GEDI, emphasising the effectiveness and influence of PolInSAR features specifically, topographic and structural metrics (e.g., DEM heights and incidence angle), followed by forest height derived from the RVoG inversion model and percent canopy cover, while Sentinel-2 indices primarily played a supporting role.

Monitoring vegetation water dynamics at sub-daily scales can reveal early hydraulic stress in forests and improve drought early-warning systems. Existing radar satellites such as Sentinel-1 provide frequent, all-weather observations of vegetation structure and moisture but lack the temporal resolution to resolve diurnal fluctuations that reflect plant water transport processes. Upcoming missions like ESA’s BIOMASS and ROSE-L will advance forest parameter retrievals, yet complementary UAV-scale measurements are needed to capture short-term canopy moisture changes and support calibration and validation.

This contribution outlines a measurement concept for a UAV-based multi-band SAR system (C/L/P) to monitor diurnal changes in vegetation water dynamics at the forest-stand scale. The approach builds on recent studies linking radar backscatter to canopy hydraulic properties (1–4) and overnight rehydration patterns, which signal early drought stress (5).

As a precursor, we analyse Sentinel-1 backscatter in combination with Sentinel-2 vegetation indices to characterize radar–optical relationships and moisture regime variability across selected German forest sites (six beech-dominated, three pine-dominated, and five pre-forest stands). This analysis will guide site selection for UAV-SAR campaigns by identifying locations with contrasting canopy phenologies and moisture responses.

Subsequent UAV campaigns will acquire multi-band SAR observations three times daily, accompanied by in-situ measurements of soil moisture, sap flow, and stem radius variations. Integrating these data streams will allow us to model root–canopy water exchange and validate radar-derived vegetation water metrics at unprecedented temporal resolution.

By bridging the temporal gap between satellite and ground observations, this work contributes to algorithm development for vegetation moisture retrieval and supports the integration of UAV-SAR in calibration and validation frameworks for future ESA missions.

1. Chaparro D, Piles M, Vall-llossera M, Camps A, Konings AG, Entekhabi D. L-band vegetation optical depth seasonal metrics for crop yield assessment. Remote Sensing of Environment. 2018;212:249-259. doi:10.1016/j.rse.2018.04.049

2. Konings AG, Yu Y, Xu L, Yang Y, Schimel DS, Saatchi SS. Active microwave observations of diurnal and seasonal variations of canopy water content across the humid African tropical forests. Geophysical Research Letters. 2017;44(5):2290-2299. doi:10.1002/2016GL072388

3. Steele-Dunne S, Friesen J, van de Giesen N. Using Diurnal Variation in Backscatter to Detect Vegetation Water Stress. IEEE Transactions on Geoscience and Remote Sensing. 2012;50(7):2618-2629. doi:10.1109/TGRS.2012.2194156

4. Steele-Dunne S, Basto A, de Zan F, et al. SLAINTE: A SAR mission concept for sub-daily microwave remote sensing of vegetation. In: EUSAR 2024; 15th European Conference on Synthetic Aperture Radar. 2024:870-872. Accessed June 25, 2025. https://ieeexplore.ieee.org/abstract/document/10659518

5. Ziegler Y, Grote R, Alongi F, Knüver T, Ruehr NK. Capturing drought stress signals: the potential of dendrometers for monitoring tree water status. Tree Physiol. 2024;44(12). doi:10.1093/treephys/tpae140

This study investigates the potential of multi-polarimetric features extracted from Sentinel-1 Synthetic Aperture Radar (SAR) data for the detection and mapping of burnt areas (BA) caused by a severe wildfire that affected Stromboli Island, located in the Aeolian Archipelago, Italy, in May 2022. A novel change detection technique based on Dual-Polarimetric (DP) SAR data is proposed to improve the sensitivity to surface alterations induced by fire-related effects. The method relies on the computation of the normalized difference between covariance matrices derived from pre- and post-fire SAR acquisitions, effectively capturing modifications in backscattering mechanisms associated with vegetation loss and soil exposure. The resulting change detection map highlights areas exhibiting significant radiometric and structural variations, which are subsequently analyzed to delineate the fire-affected regions. To automatically extract the burnt area, an unsupervised clustering approach based on the k-means algorithm is applied, classifying image pixels according to their polarimetric and radiometric properties and separating burnt from unburnt zones. The reliability of the proposed methodology is assessed through a comparative analysis with conventional Single-Polarimetric (SP) change detection techniques, employing Sentinel-2 optical imagery as reference data. Results demonstrate that the DP-based approach significantly enhances the discrimination of fire-impacted surfaces, providing a more accurate and spatially coherent mapping of burnt areas compared to SP methods. Overall, the study highlights the effectiveness and operational potential of integrating SAR-based multi-polarimetric information with unsupervised classification techniques for near-real-time wildfire monitoring and post-event damage assessment.

This paper explores the use of multiple correlation coefficient R and complex correlation coefficient for reflection symmetry testing in multi-look polarimetric synthetic aperture radar (PolSAR) data. The recently proposed block-diagonality test statistic is actually related to the former, which arises from multiple linear regression. Moreover, the exact distribution of the squared multiple correlation R^2 for a 3x3 polarimetric covariance matrix is a beta distribution, which depends only on the number of looks L. As the number of looks tends to infinity, the limiting distribution of LR^2 is a gamma distribution with unit scale parameter and a shape parameter of two. For the complex correlation coefficient, its magnitude squared is also beta distributed. In this study, both the multiple correlation coefficient and complex correlation coefficient were applied to the NASA/JPL AIRSAR data acquired over Camp Roberts, California. The rugged terrain showed its characteristic of non-reflection symmetry. In addition, the reflection asymmetric target detection was carried out using multi-look ALOS-2 PALSAR-2 quad-polarisation data over selected test sites in Southeast Asia. The experimental results confirmed the usefulness of reflection symmetry property of geophysical media, particularly in detecting man-made objects, such as bridge, slightly oriented buildings, vessels, aquaculture farms, etc. However, some buildings, vessels, and part of bridges belonging to dihedrals were categorised as reflection symmetric targets.

Detecting objects lying beneath a forest cover using SAR measurements represents a major challenge, due to the response of the overlying vegetation volume, wave attenuation caused by propagation through the forest canopy, and high-intensity scattering mechanisms occurring at the ground level. 3D polarimetric SAR imaging, through Polarimetric SAR Tomography (PolTomoSAR), represents an appealing solution to address these limitations, as it enables discriminating objects from their background by leveraging both polarimetric and spatial diversities. This work investigates two approaches based on PolTomoSAR processing, and adapted to different tomographic acquisition configurations, i.e. different vertical resolution and ambiguity compromises.

The first method relies on PolTomoSAR data that feature high vertical resolution and a wide unambiguous elevation range. Full-Rank polarimetric SAR tomographic focusing techniques are employed to isolate, with a high resolution, scattering sources located a few meters above the ground, and estimate their full-rank polarimetric responses. The concealed object detection is then conducted, based on polarimetric parameters provided by classical decomposition techniques. A simple detector, combining a few source descriptors, such as the polarimetric entropy and indicators of double-bounce scattering, as well as the elevation information, proves effective in identifying artificial objects embedded in dense, masking vegetation.

The second approach considers an extreme, but far less complex configuration, consisting solely of a two-image PolinSAR acquisition. Ground-notched InSAR processing is applied to suppress ground-scattering contributions, whose polarimetric and radiometric features may prevent the detection of objects, providing a filtered image representing a possibly ambiguous sampling of the scene reflectivity in the vertical direction. In the context of concealed object detection, the choice of the interferometric baseline separating the acquisition trajectories is crucial, as it balances the suppression of the forest canopy contribution and the preservation of responses from above-ground objects. Further discrimination is then carried out through a polarimetric analysis.

Both methods are applied to a 21-image fully polarimetric L-band data set, acquired by the DLR F-SAR sensor over Dornstetten, Germany. The study site consists of a mixed forest area containing several man-made objects, such as vehicles, containers, and corner reflectors, that are deployed both inside and outside the forest. Results show that combining spatial and polarimetric diversity modes allows both methods to successfully detect the different artificial objects in the scene, outside and below the forested areas.

Tropical forests play a pivotal role in the global carbon cycle, sequestering vast amounts of atmospheric CO₂ and storing it as aboveground biomass (AGB). Accurate AGB estimation is fundamental for monitoring forest health and understanding carbon dynamics under changing climatic conditions. Satellite-based observations, such as those provided by the European Space Agency’s (ESA) BIOMASS mission, offer powerful means for large-scale forest carbon assessment; however, their reliability depends on robust ground-based calibration and validation. This project aims to calibrate and validate the BIOMASS mission’s AGB and canopy height (H100) products across tropical forests in São Paulo State, Brazil. The mission’s novel P-band synthetic aperture radar (SAR) provides unprecedented sensitivity to forest structural properties, but its performance must be evaluated in the highly heterogeneous environments of tropical forests. We will leverage four 10-hectare permanent plots from the GEO-TREES network, encompassing key Atlantic Forest types: cerradão, semideciduous forest, dense ombrophilous forest, and restinga. The methodology integrates detailed forest inventories (diameter, height, species composition) with terrestrial laser scanning (TLS) for high-resolution 3D structural modeling and airborne LiDAR (ALS) for landscape-scale mapping. These datasets will inform advanced allometric models and support BIOMASS product calibration at 4-hectare (200 × 200 m) resolution. Additionally, polarimetric P-band SAR data will be analyzed to characterize radar–forest structure interactions. Expected outcomes include: (1) forest type–specific allometric models, (2) quantitative validation of BIOMASS AGB products with target accuracy (RMSE ≤ 20% for AGB > 50 t ha⁻¹), and (3) improved understanding of radar signal responses in tropical forest canopies. The results will contribute to global carbon monitoring, REDD+ implementation, and the advancement of radar-based remote sensing, with outputs disseminated through peer-reviewed publications and collaboration with ESA.

Accurate estimation of forest height is essential to improve the understanding and quantification of forest structure, aboveground biomass (AGB) and carbon dynamics. The integration between spaceborne and airborne remote sensing technologies such as synthetic aperture radar (SAR), airborne laser scanning (ALS) and field data might be the key to reliably assess the contribution of these environments to the carbon cycle and the impacts of human-induced disturbances. Several satellites provide forest height estimates at fine resolution, however, until now, these height estimates have not been systematically validated, especially in the carbon-dense tropical forests and using novel satellite missions.

The ESA BIOMASS mission — the first spaceborne satellite to operate a fully polarimetric SAR P-band — plays a fundamental role in advancing forest height estimation, since its wavelength (~70 cm) allows the radar signal and its backscattering to deeply penetrate the forest canopy. Light detection and ranging (LiDAR) technology in combination with in situ measurements can support overall calibration and validation (Cal/Val) BIOMASS activities related to forest height estimations.

Our goal is to gather and process airborne LiDAR and field data collected over the Brazilian Amazon, creating a reference dataset to validate the BIOMASS forest height product. The LiDAR data was collected by the Brazilian Forest Service over forests under sustainable forest management, enabling pre- (undisturbed) and post-disturbance (logged) assessment. The field data was collected by companies operating in the same areas, including estimates such as geolocation, diameter at breast height, height and stem volume of each tree, related to its corresponding logging annual site.

ALS point cloud data can be processed to generate canopy height models (CHM). The currently automated pipeline performs LiDAR point cloud cleaning, normalization, ground classification and rasterization, producing standardized DTM, DSM, and CHM outputs. Quality assurance products, such as pulse density, scan angle, and laser penetration maps, are also generated to ensure traceability. Field data processing uses the variables from logged trees to measure gaps created after logging and to compare them with ALS tree height estimates. The reference dataset is rasterized.

The validation strategy uses the reference dataset to perform a multi-level spatial comparison by interpolation methods with the BIOMASS height product. Height layers and field data can be co-registered and compared through pixel-based and aggregated statistical analyses. The expected results include the computation of coefficient of determination (R²), root mean square error, and bias, complemented by a regression analysis to identify systematic deviations between datasets. The analysis is an independent validation framework that captures both canopy-level and tree-level variations.

Our results aim to contribute to the ongoing Cal/Val activities of the ESA BIOMASS mission. We strongly believe that ALS data and in situ measurements can provide benchmark information for calibrating and validating overall remote sensing products. We will thus help reduce uncertainties in global carbon stock and flux estimations, particularly those associated with land-use change, forest degradation, and regrowth.

ESA’s BIOMASS mission is the first P-band SAR satellite. The penetration capability of P-band microwave and the SAR tomography (TomoSAR) configuration of BIOMASS provide an unprecedented opportunity to improve the quantification of carbon stock in tropical forests. Robust in-situ and airborne data are necessary for the calibration and validation (Cal/Val) of BIOMASS products, including above-ground biomass densities (AGB), forest height (FH) and forest disturbances (FD). To support the BIOMASS Cal/Val, this project will collect forest inventory and unmanned aerial vehicle (UAV) lidar data in the Peruvian rainforest.

Along the Tambopata River in tropical Peru, we will establish further reference sites to specifically validate BIOMASS estimates along gradients from intact to degraded tropical forests and by quantifying changes in biomass from 2024 to 2026/2027 by using UAVs with LiDAR and aerial imaging. Some areas were already surveyed in 2021 and 2024. Forest inventories on smaller plots (max. 20 × 50 m) were also conducted, recording species, stem diameter, and tree height. To meet the standards set out in the BIOMASS Products Verification and Validation Plan, larger inventories and further UAV surveys are planned to first represent a spatial gradient from degraded to undisturbed rainforest and second to quantify changes in biomass from 2024 to 2026/2027. The inventory and UAV data will be published in an open data repository.

The objectives of this project are to: (1) validate the BIOMASS products using field inventory and UAV lidar data; (2) estimate changes in forest height and AGB using BIOMASS and comparing them with the changes detected from multitemporal UAV lidar, and (3) investigate the sensitivity of the TomoSAR reflectivity profiles to the vertical forest structure by comparing the BIOMASS P-band TomoSAR reflectivity profiles with the lidar-derived forest vertical structure profiles (e.g., canopy cover profile).

The Arctic is warming more than twice as fast as the global average, making climate change’s polar effects more intense than anywhere else in the world. The Arctic accounts for half of the organic carbon stored in soils. There is high confidence that the thaw of terrestrial permafrost will lead to carbon release, but only low confidence regarding timing, magnitude and relative role of CO2 versus CH4. There is general consensus that these issues can be tackled through support by satellite observations, but this has not been fully exploited to date. The recently inaugurated NASA-ESA Arctic Methane and Permafrost Challenge (AMPAC) strives to address these questions inter alia through making use of synergistic measurements, activities to improve satellite retrievals with a clear focus on high latitudes, and promoting new dedicated satellite sensors as well as improving validation of existing and upcoming satellite missions.

The objectives are to evaluate the capability of P-band SAR to support AMPAC, specifically for wetland and freeze/thaw dynamics monitoring. Results will support the discussion on gaps of current and future missions for the purpose of AMPAC, continuing Bartsch et al. (2025). Requirements towards recent and future missions will be presented and preliminary results will be discussed depending on BIOMASS mission data availability.

Bartsch, A., Gay, B.A., Schüttemeyer, D., Malina, E., Miner, K. Grosse, G. , Fix, A., Tamminen, J., Bösch, H., Parker, R. J., Rautiainen, K., Hashemi, J., Miller, C. E: "Advancing the Arctic Methane Permafrost Challenge (AMPAC) With Future Satellite Missions," in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 18, pp. 6279-6298, 2025a, doi: 10.1109/JSTARS.2025.3538897

Topics:

• Differential interferometric SAR (DInSAR)

• Permafrost, retrogressive thaw slumps, and coastal erosion

ABSTRACT (Abstract + references maximum 1000 words)

Periglacial coasts are degrading due to thermo-abrasive and thermo-denudative processes acting on permafrost [1]. This not only reshapes coastal geomorphology but also enhances nearshore nutrient input and microbial activity, which in turn mobilises organic carbon, part of which is released into the atmosphere as greenhouse gases [2]. With an estimated 1,035 Pg of carbon stored in permafrost worldwide [3] and roughly 34 % of the world’s coastlines underlain by permafrost [4], such degradation has far-reaching implications for the global climate. Yet, the mechanisms that drive coastal erosion, particularly abrupt thaw processes such as retrogressive thaw slumps (RTS), remain poorly understood.

This study aims to improve the understanding of permafrost coastal erosion dynamics, and in particular addresses the question: How can interferometric SAR time-series be used to explain spatial and temporal patterns of surface deformation associated with coastal permafrost degradation and retrogressive thaw slump activity?

The research question is approached by combining differential interferometric SAR (DInSAR) time-series analysis, with geomorphological and thermal covariates. The study focuses on Herschel Island (Qikiqtaruk), Yukon Coast, Canada – a key site for monitoring coastal permafrost degradation in the southern Beaufort Sea [5]. Using multi-temporal TerraSAR-X data (2020–2024), differential InSAR (DInSAR), small baseline subset (SBAS), and 2D inversion techniques were applied to derive millimetre- to centimetre-scale vertical and horizontal surface displacements. Multipolarimetric signatures from the X-band acquisitions were further utilised to infer surface scattering differences.

These deformation patterns were analysed in relation to coastal erosion and retrogressive thaw slump attributes derived from TanDEM-X differential digital elevation models (dDEMs) [6]. The dDEMs were further used to estimate eroded sediment volumes and soil organic carbon (SOC) losses following the approach of [7]. The deformation patterns and dDEMs were additionally compared with environmental covariates such as air temperature, land surface temperature (LST), geomorphology, and vegetation indices to assess how these factors influence RTS behaviour and coastal stability.

By resolving multi-year deformation trends and linking them to geomorphological features, this work contributes to a quantitative understanding of permafrost coastal instability. It highlights the potential of combining radar interferometry with environmental covariates to characterise the spatio-temporal evolution of coastal permafrost degradation – an essential step towards predicting future Arctic coastal responses under continued climate warming.

REFERENCES

[1] Günther, F., Overduin, P. P., Sandakov, A. V., et al.: Thermo-erosion along the Yedoma Coast of the Buor Khaya Peninsula, Laptev Sea, East Siberia, in Proceedings of the Tenth International Conference on Permafrost, Volume 1: International Contributions, Salekhard, Russia, 25–29 June 2012, 137–142, 2012.

[2] Couture, N.: Fluxes of soil organic carbon from eroding permafrost coasts, in Canadian Beaufort Sea, McGill University, 2010.

[3] Hugelius, G., Strauss, J., Zubrzycki, S., et al.: Estimated stocks of circumpolar permafrost carbon with quantified uncertainty ranges and identified data gaps, in Biogeosciences, 11, 6573–6593, https://doi.org/10.5194/bg-11-6573-2014, 2014.

[4] Lantuit, H., Overduin, P.P., Couture, N. et al.: The Arctic Coastal Dynamics Database: A New Classification Scheme and Statistics on Arctic Permafrost Coastlines, in Estuaries and Coasts 35, 383–400, https://doi.org/10.1007/s12237-010-9362-6, 2012.

[5] Obu, J., Lantuit, H., Myers-Smith, I., et al.: Effect of terrain characteristics on soil organic carbon and Total nitrogen stocks in soils of Herschel Island, Western Canadian Arctic, in Permafrost and Periglacial Processes, 28 (1), 92–107, https://doi.org/10.1002/ppp.1881, 2017.

[6] Maier, K., Bernhard, P., Ly, S., et al.: Detecting mass wasting of Retrogressive Thaw Slumps in spaceborne elevation models using deep learning, International Journal of Applied Earth Observation and Geoinformation, 137, 1569-8432, https://doi.org/10.1016/j.jag.2025, 2025.

[7] Ramage, J. L., Irrgang, A. M., Morgenstern, A., et al..: Increasing coastal slump activity impacts the release of sediment and organic carbon into the Arctic Ocean, in Biogeosciences, 15, 1483–1495, https://doi.org/10.5194/bg-15-1483-2018, 2018.

SAR Histomography, also known as Phase Histogram technique, enables the reconstruction of tomographic-like vertical forest profiles from a single-polarization, single-interferometric pair with a non-zero spatial baseline. This approach drastically reduces the acquisition requirements and costs typically associated with tomographic missions. Although the method provides only an approximation of the true vertical distribution of scatterers within the canopy, it has demonstrated promising capabilities for digital elevation model (DEM) estimation and for improving the understanding of radar–forest interactions [1,2,3].

This work investigates the applicability and limitations of SAR histomography for forest-structure retrieval and DEMs estimation. The analysis relies on airborne interferometric and polarimetric L- and P-band SAR data acquired by NASA-JPL’s UAVSAR system during the AfriSAR-1 and AfriSAR-2 campaigns over tropical forests in Gabon, complemented by LVIS LiDAR waveforms used as reference data. Two representative study sites, namely the Rabi Forest and Pongara National Park, were selected for their contrasting soil conditions and vegetation types, enabling the influence of both system and scene properties on the derived radar profiles to be examined.

To quantify the spatial variability of the generated radar profiles at L- and P-band, a set of structural metrics was defined. These metrics describe the vertical distribution of the profiles, providing insights into canopy stratification and scattering complexity. Furthermore, a method for estimating digital elevation models from radar profiles while accounting for signal attenuation through the canopy, is proposed. The obtained results are compared and discussed to evaluate the performance and limitations of the approach.

[1] Shiroma H.X. G., et al. Digital terrain, surface, and canopy height models from

insar backscatter-height histograms. IEEE Transactions on Geoscience and Remote Sensing,

58(6):3754–3777, 2020.

[2] Lavalle, M., et al. "Histogram SAR Tomography: Model and SAR Observations." EUSAR 2024; 15th European Conference on Synthetic Aperture Radar. VDE, 2024.

[3] Wu, Chuanjun, et al. "Evaluating Phase Histograms for Remote Sensing of Forested Areas Using L-Band SAR: Theoretical Modeling and Experimental Results." IEEE Transactions on Geoscience and Remote Sensing (2024).

As the world’s second largest rainforest, the Congo Basin rainforest plays a crucial role in the global carbon cycle. Furthermore, recent data suggests that it is more carbon-dense and more resistant to climate change than the Amazon (White et al, 2021). It is also a vital resource for local livelihoods and regional climate regulation. Increasing human disturbance to this rainforest due to demographic growth is generating large uncertainties in the regional carbon balance, mainly due to a lack of understanding of forest regrowth trajectories. The Afrocards consortium (U. Gent, U. Liege and U. catholique de Louvain) works to better understand regional regrowth trajectories following slash-and-burn agriculture, which is the dominant cultivation system in the region. In particular, we aim to shed light on the role of land use history and environmental variables in determining forest regrowth. To that end, we work to develop a regional land surface model calibrated on field, airborne, and satellite remote sensing data.

Here we present results related to the calculation of regrowth curves based on satellite remote sensing data using a space-for-time approach, where forest patches of different age are coupled with their above ground biomass. Building on a methodology initially established at the Laboratoire des Sciences du Climat et de l’Environnement (P. Ciais, Y. Xu), we use the time since last disturbance as a proxy for forest age, derived from the Tropical Moist Forest dataset, paired with height and biomass estimates derived from the GLAD and CCI-Biomass products as our input data. The coupled age/biomass data is grouped by land use history classes and used as input to fit local sigmoidal (Richard-Chapman) regrowth curves using a Bayesian approach at the 1-degree grid cell level, across the Congo Basin. By using a Bayesian modeling approach we can better account for uncertainties on the input data and output model parameter estimates. We use the posterior distributions of the fit parameters for all 312 grid cells and 4 land use history classes together together with gridded bioclimactic variable datasets to carry out an exploratory analysis of variable importance and interaction by means of machine learning techniques, including random forest and clustering methods. Ultimately, we aim to use such local regrowth curves to calibrate the Ecosystem Demography Biosphere model (version 2) to carry out mechanistic modeling of forest regrowth in the Congo Basin under different climate change and demographic growth scenarios.

References:

Congo Basin rainforest –invest US$150 million in science, White et al., Nature, 2021

In recent years an increase in development and publishing of large scale (pan-troical to global) remote sensing based aboveground biomass (AGB) maps is observed. These datasets differ in spatial resolution, input data and modelling approaches. Given these differences it is important to compare the datasets with a common global reference data and understand implications for different use cases.

Here we present and overview and preliminary results from an ongoing validation of selected pan-tropical and global biomass products using AGBref, “A global forest biomass reference dataset”, and pan-tropical Airborne Laser Scanning (ALS) data collected in Brazilian Amazon, Democratic Republic of Congo and in Borneo, Indonesia. To do this, we compiled 23 AGB maps, matched each mapped year with AGBref, harmonized the forest definition by applying a consistent forest mask, disaggregated to 0.1 degree pixel spacing using mean values and calculated statistical metrics (i.e., bias, RMSE) at different biomass intervals. Furthermore, we used ALS-based AGB estimates to assess how well the global maps represent local spatial patterns. For this, we used high resolution AGB maps (i.e., at least 100 m pixel spacing) and calculated a local heterogeneity ratio index, defined as the ratio of standard deviation (SD) from a map within a moving window to the SD from the ALS-AGB within the same moving window.

Preliminary results indicate that the maps exhibit systematic differences, either as a uniform bias (e.g. overestimation across all biomass bins) or as over- and under-estimation in the lower and upper biomass ranges. An important trend is observed is that these biases tend to decrease in more recently produced maps. Furthermore, product versioning (e.g., the first vs. the last version) leads to a lower biases at lower and upper biomass levels. Finally, using ALS-AGB as reference, we show how much spatial details are captured, highlighting that recent maps based on high-resolution satellite imagery (<30m) and novel modelling approaches (e.g. deep learning) do not always preserve fine-scale spatial details.

Forest water availability is a major driver of forest growth, transpiration, productivity, and carbon turnover at the global scale1. Monitoring spatial and temporal dynamics of forest water content is therefore essential. Satellite remote sensing provides large-scale observations, and several water-related indicators already exist, such as multispectral indices, live fuel moisture content (LFMC) from optical and microwave data2,3 or vegetation optical depth (VOD) from passive microwave instruments4,5. However, current products are limited either by insufficient canopy penetration (optical domain) or coarse spatial resolution (passive microwave domain), which restricts their applicability in forest ecosystems.

High-resolution active microwave systems can overcome these limitations. Yet, current methods have not achieved the accuracy required to retrieve forest water content, and applications remain focused mainly on sparse vegetation6,7. Here, we introduce a novel approach using corner reflectors as stable reference targets to improve active microwave retrievals of forest water content. The strong and well-defined backscatter response of corner reflectors reduces modelling uncertainties and simplifies theoretical models such as the Water Cloud Model.

This enables direct assessment of vegetation attenuation and scattering effects from the canopy and allows the derivation of physically interpretable parameters related to forest VOD. Previous microwave tower experiments have demonstrated the feasibility of retrieving vegetation status using reflector-based configurations8,9. We will present the theoretical basis, system design, and first experimental results demonstrating the potential of corner-reflector-based SAR observations as a reference measurement within the framework of PolInSAR techniques or the BIOMASS mission.

1. Konings, A. G. et al. Detecting forest response to droughts with global observations of vegetation water content. Global Change Biology 27, 6005–6024 (2021).

2. Yebra, M. et al. A global review of remote sensing of live fuel moisture content for fire danger assessment: Moving towards operational products. Remote Sensing of Environment 136, 455–468 (2013).

3. Rao, K., Williams, A. P., Flefil, J. F. & Konings, A. G. SAR-enhanced mapping of live fuel moisture content. Remote Sensing of Environment 245, 111797 (2020).

4. Schmidt, L. et al. Assessing the sensitivity of multi-frequency passive microwave vegetation optical depth to vegetation properties. Biogeosciences 20, 1027–1046 (2023).

5. Bueso, D. et al. Soil and vegetation water content identify the main terrestrial ecosystem changes. National Science Review 10, nwad026 (2023).

6. El Hajj, M. et al. First Vegetation Optical Depth Mapping from Sentinel-1 C-band SAR Data over Crop Fields. Remote Sensing 11, 2769 (2019).

7. Vreugdenhil, M. et al. Sentinel-1 Cross Ratio and Vegetation Optical Depth: A Comparison over Europe. Remote Sensing 12, 3404 (2020).

8. Lemmetyinen, J. et al. Attenuation of Radar Signal by a Boreal Forest Canopy in Winter. IEEE Geosci. Remote Sensing Lett. 19, 1–5 (2022).

9. Fleischman, J. G., Ayasli, S., Adams, E. M. & Gosselin, D. R. Foliage attenuation and backscatter analysis of SAR imagery. IEEE Trans. Aerosp. Electron. Syst. 32, 135–144 (1996).

Forest height and biomass are critical indicators for evaluating forest structure and quantifying ecosystem carbon storage. They play an essential role in forest resource monitoring and carbon cycle assessments. As most forests are located in mountainous regions, topographic effects exert a substantial influence on radar signal propagation and Polarimetric Interferometric Synthetic Aperture Radar (PolInSAR) measurements. In such regions, the conventional Random Volume over Ground (RVoG) model often yields significant systematic biases in forest height inversion results, primarily due to its omission of the coupling relationship between terrain slope and satellite observation geometry.To address this limitation, this study proposes an enhanced Sloped Random Volume over Ground (S-RVoG) model that incorporates ascending/descending orbit fusion and Digital Elevation Model (DEM) constraints to improve the accuracy and stability of forest height estimation using L-band SAOCOM PolInSAR data.Specifically, the proposed approach employs 12.5 m DEM data to accurately derive the slope and aspect of each pixel. By integrating the satellite incidence angle with terrain aspect information, slope-facing and back-slope regions are identified to correct both the local incidence angle and the vertical wavenumber (k_z). Furthermore, the DEM is utilized to directly compute and remove the terrain phase (φ₀), effectively eliminating phase distortions caused by complex topography and preventing systematic errors in the S-RVoG inversion process arising from terrain-induced effects.To further mitigate the influence of terrain-induced systematic bias, an error-constrained fusion mechanism is introduced within the improved S-RVoG framework. Forest height inversion is independently performed using both ascending and descending SAOCOM datasets. The reliability weights of each inversion result are determined according to their respective root mean square error (RMSE) values obtained through comparison with ground measurements or UAV LiDAR data. The final forest height is then derived via RMSE-inverse weighted fusion, enabling adaptive correction of systematic errors under varying observation geometries. This RMSE-constrained multi-orbit fusion strategy effectively reduces slope-induced biases and substantially enhances the accuracy and robustness of forest height estimation in mountainous areas.To validate the proposed model, a comprehensive field campaign was conducted in the western Qinling Mountains, a region characterized by highly variable topography with slopes exceeding 50° and an average slope of approximately 20°. Field measurements included tree height, species type, diameter at breast height (DBH), slope, and other relevant attributes. Additionally, UAV LiDAR data acquired over the same sampling areas were employed for validation and accuracy assessment.Experimental results demonstrate that the proposed method effectively suppresses slope-related systematic errors, improving forest height estimation accuracy by approximately 5–10% in steep terrain. Overall, the DEM-constrained, multi-orbit SAOCOM-integrated, and UAV LiDAR-validated S-RVoG model provides a physically consistent and empirically robust framework for L-band PolInSAR-based forest parameter inversion in mountainous regions with complex topography.

Understanding the spatial heterogeneity of Amazonian ecosystems requires radar observations that capture both canopy structure, forest understory, and hydrological characteristics. This study investigates the complementary sensitivities of P-band data from the ESA BIOMASS mission and C-band data from Sentinel-1 Synthetic Aperture Radar (SAR) across representative ecosystems of the Amazon Basin.

We employ BIOMASS SAR data in DGM (Detected Ground Multi-look) format, acquired on 2025-06-09, and two Sentinel-1 SAR datasets in GRD (Ground Range Detected) format, acquired on 2025-06-03 and 2025-06-15, which were co-registered and radiometrically calibrated, over an Amazonian region in Brazil. The selection of the study area was based on the availability of BIOMASS data. MAPBIOMAS land-cover data serve as the primary reference classification, providing detailed delineations of forest and non-forest areas. Within the area of interest, land-cover types include Forest Formation, Floodable Forest, Wetland, Grassland Formation, River, among others. Additionally, a peatland map of Brazil was utilised as complementary information to assess radar sensitivity to moisture and hydrological variations.

Backscatter intensity, polarisation ratios (e.g. VV/VH), and texture metrics are extracted from both datasets to characterise wavelength-dependent scattering properties for each land-cover type. Initial visual inspection indicates that BIOMASS P-band data clearly delineate river corridors and floodable areas that are only partially visible in Sentinel-1 C-band imagery, conforming to deeper canopy penetration and reduced attenuation at longer wavelengths. While C-band can provide information on forest canopy structure, P-band could be used for better differentiation of land cover types within the study area. These complementary sensitivities suggest that combined C- and P-band observations can better capture structural and hydrological gradients across tropical landscapes.

The study provides a quantitative assessment of multi-frequency radar responses across Amazonian ecosystems, highlighting wavelength-specific advantages for identifying and delineating various land-cover types, including forest boundaries, as well as for hydrological mapping. It advances the understanding of radar backscatter mechanisms in humid tropical environments and demonstrates how multi-frequency SAR synergy can enhance the characterisation of structural and environmental properties across the Amazon Basin.

The timely and accurate monitoring of inland water bodies, particularly dynamic wetlands and catastrophic flood events, is critical for environmental management, hydrological modeling, and disaster response. Synthetic Aperture Radar (SAR) provides a powerful tool for this purpose, given its all-weather and day-night imaging capabilities. While quad-polarization data (i.e., exploiting the full covariance or coherency matrices) offers the possibility to extract information on different scattering mechanisms, many operational systems are limited to dual-polarization (dual-pol) acquisitions. This research proposes a novel framework that maximizes the information content of dual-pol SAR data for robust inland water detection, primarily aiming at distinguishing flooding in complex scenes such as floodplains, wetlands, and vegetated areas.

The proposed approach aims to identify inundation events through the application of different polarimetric techniques. First, a comparison will be carried out between the magnitude of the eigenvalues computed from the 2x2 dual-polarimetric covariance matrix (C2) for both pre-event (baseline) and post-event SAR acquisitions. The underlying assumption is that the significant change in the scattering properties caused by water presence will manifest as changes in the derived eigenvalues. The performance of the eigenvalue-based approach in detecting subtle inundation, including flooded vegetation, will be compared against traditional amplitude-based techniques. In addition, two polarimetric decompositions suitable for dual-polarization SAR data will be investigated. The first is the H/ (Entropy/Alpha) decomposition, which enables the extraction of information on the scattering behaviors within the scene through the generation of the H/α scattering plane. The second is the m- decomposition based on the definition of the Stokes parameters, which in turn are related to the degree of polarization, the ellipticity and the orientation angle of the polarization ellipse associated with the polarized component of the electromagnetic wave.

By synergistically combining the temporal change information from the eigenvalue analysis with the physical scattering characterization from the H/ and m- decomposition, this framework aims to significantly reduce classification ambiguity.

Our hypothesis is that this dual-method approach will demonstrate enhanced performance over conventional techniques, particularly in complex wetland environments where mixtures of open water, submerged vegetation, and saturated soils co-exist. This investigation validates the proposed method using dual-pol C-band Sentinel-1 data and BIOMASS P-band data. The results will be discussed at PolinSAR to foster a discussion on advanced physically-based methods to improve inland water mapping, providing a new valuable tool for hydrological applications.

Synthetic Aperture Radar (SAR) has emerged as a critical data source for continuous and reliable forest monitoring, particularly in tropical regions where persistent cloud cover limits the applicability of optical remote sensing. Owing to its sensitivity to structural and dielectric properties of vegetation, SAR enables the study of forest dynamics through a variety of polarimetric and interferometric derivatives. This study investigates the potential of polarimetric and interferometric SAR variables derived from C-band Sentinel-1 data for deforestation mapping across tropical and sub-tropical forest regions. Specifically, the research focuses on the Haldwani Forest Range in India, which is characterized by complex forest management activities and seasonal soil moisture variations that pose challenges to accurate change detection.

A statistical change detection algorithm, Cumulative Sums of Change (CuSUM) was employed to analyze backscatter time-series data for identifying abrupt and persistent changes indicative of deforestation events. The backscatter-driven change detection algorithm was applied using VH-polarized Sentinel-1 data. However, while this approach effectively captured major deforestation events, it also resulted in a considerable number of false positives, primarily arising from grazing activity and short-term soil moisture variations that caused fluctuations in backscatter intensity unrelated to forest loss. To address this limitation, a coherence-based compensation mechanism was developed to refine the backscatter-derived change maps.

In this proposed framework, temporal coherence values were used as a weighting factor to assess the reliability of detected changes. Coherence served as an indicator to distinguish between genuine deforestation and temporary disturbances. A rule-based decision-making approach was then implemented to retain or discard changes flagged by the backscatter-based method based on the corresponding coherence values. Specifically, changes associated with low coherence were examined to confirm whether they represented consistent structural alteration (e.g., tree removal) or temporary effects (e.g., grazing or moisture fluctuations). This coherence-based decision mechanism significantly reduced the number of false positives, thereby improving the accuracy and reliability of deforestation detection.

The results demonstrated a marked improvement in classification accuracy through the integration of coherence information. Using a backscatter-only approach, the overall accuracy and kappa coefficient achieved were 0.58 and 0.23, respectively, for VH polarization. After applying the coherence-based compensation approach, these values improved by approximately 39.6% (overall accuracy = 0.81) and 120.8% (κ = 0.53), indicating a substantial reduction in commission errors. The study highlights the importance of incorporating temporal coherence as an auxiliary variable in SAR-based change detection to mitigate false alarms caused by dynamic environmental factors such as soil moisture variations, isolated rainfall events, and inconsistencies in land cover classification.

Low-frequency synthetic aperture radars are monitored with a special SAR transponder. Passive radar targets are less convenient, because of structural aspects and dimensionality.

Large reflector antenna structures operating with their antenna function at higher frequency bands can provide a potential mono-static backscattering, provided the antenna is pointed towards the SAR satellite. That is conveninet for low frequency SAR.

Biomass Earth Explorer mission measures data, with which tomographic assessment is intended for various scenarios. An array of radio telescope antennas, which for their function operating at higher frequency, can provide supporting mono-static backscattering, the level of which is systematic though not a priori known at the frequency of the SAR. Initial analysis serves as a prediction.

Assessment of responses for a SAR instrument implies pointing to the satellite. It implies a possibility to derive out-of-band scattering. Certain radio telescope antennas make use of three positioning axes, which can be convenient also to assist polarisation aspects

Vegetation optical depth (VOD) derived from microwave satellites provides critical information on vegetation water content as well as biomass and has emerged as a valuable remote sensing metric for quantifying water stress impacts on vegetation carbon uptake. However, as ground validation of VOD remains challenging, terrestrial biosphere models (TBMs) have been developed as essential tools for comprehensive validation and broader application of VOD. By incorporating plant hydraulic modeling schemes, TBMs can simulate key vegetation water status variables (such as leaf water potential), thereby enabling the simulation of VOD. In this study, we developed a novel plant hydraulics module within the adjoint-based Nanjing University Carbon Assimilation System (NUCAS) to facilitate VOD simulation. Parameter sensitivity analysis was conducted to identify key parameters influencing VOD. Simulated VOD was validated against microwave remote sensing-derived VOD observations and biomass data. Subsequently, VOD observations were assimilated into the NUCAS integrated with the plant hydraulic module to improve model representation and to enhance understanding of ecosystem processes. Results showed that the model accurately reproduces the seasonal dynamics observed in satellite-based VOD, and the simulated VOD is strongly sensitive to plant hydraulic parameters, including saturated hydraulic conductivity and minimum leaf water potential. Moreover, the assimilation of VOD effectively constrained vegetation hydraulic processes, leading to improved model representation of ecosystem carbon and water fluxes/storage. In summary, our study provides valuable insights for model development, validation, and application related to microwave satellite-derived VOD, and opens new perspectives for advancing the understanding of carbon-water processes within terrestrial ecosystems.

To explore the lunar surface several studies were carried out with many lunar missions. Continuous efforts were made by many scientists to confirm the presence of water on the lunar surface. Using remote sensing techniques much of the information was obtained from the lunar surface. Synthetic Aperture Radar (SAR) an active sensor, utilizes microwaves frequencies such as L, X, S-bands for analyzing Lunar surface and subsurface. Since, SAR is sensitive to water and has more penetration capability into the lunar surface, chandrayaan-2 mission with Dual Frequency Synthetic Aperture Radar (DFSAR) payload was manifested. DFSAR, a microwave imaging instrument was the first of its kind to utilizes dual frequencies (L and S - Bands) to operate in full polarimetric and hybrid polarimetric modes. In hybrid polarimetry, signals are circularly transmitted and linearly received. DFSAR is the first payload to operate in full polarimetric capability with L-band. In this study, several sites were used Such as Chandrayaan-3 landing site and south pole to carry out the polarimetric analysis. Yamaguchi, H-A- Alpha, and Raney M-Delta and M-Chi decomposition were applied on the datasets and several child parameters are computed.

In this study L-Band Full pol and Hybrid pol data were analyzed on the same site for a better analysis. Yamaguchi four component decomposition was performed on full pol data on lunar south pol. We compared our analysis of full pol data with hybrid pol for estimating the response of circular polarized component interaction with the lunar surface. Polarimetric analysis on Ch-3 landing site, post and pre-landing were contradicting with each other, and this may be due to the lander impact on the Lunar surface. The landing site has almost a flat terrain with few small craters. In this study, we observed that hybrid pol data has almost similar information when compared to full pol data.

The use of Tomographic Synthetic Aperture Radar (TomoSAR) has opened new perspectives for three-dimensional forest structure retrieval, yet its operational exploitation in tropical environments remains challenged by volume decorrelation, sidelobe leakage, and signal noise. These effects often obscure the vertical separation between ground and canopy scatterers and hinder the derivation of accurate height products. Moreover, the limited availability of fully polarimetric datasets restricts the capacity to distinguish scattering mechanisms across forest layers. In this study, we present a geometry-driven methodology for the extraction of vertical forest layers directly from single-polarized TomoSAR reconstructions. The proposed framework employs a concave-hull-based volume delimitation and adaptive envelope refinement to delineate canopy and terrain surfaces from the reconstructed reflectivity profiles, while maintaining the spatial coherence of the tomographic signal. The approach is independently applied to each polarization channel of the fully polarimetric dataset acquired by ESA’s TropiSAR campaign over a dense tropical forest located in Paracou, French Guiana. This polarization-wise processing avoids decomposition artifacts and allows for a direct evaluation of single-channel structural sensitivity. Validation against LiDAR-derived reference models demonstrates that the proposed method recovers Digital Terrain Models and Canopy Height Models consistent with the vertical scattering behavior of the forest. The results highlight the robustness of the proposed strategy for noise-affected TomoSAR data and its potential for large-scale forest structure monitoring.

The structure of forests directly indicates the amount of carbon stored in the ecosystem. This total mass of living vegetation above the soil, Above-Ground Biomass (AGB), impacts how the ecosystem cycles carbon, nutrients, and water. With the launch of BIOMASS (P-band) and NISAR (L-band and S-band) missions this year, we are now better equipped to make comprehensive, accurate measurements of Earth's surface features, owing to the longer wavelength of P-band, which can penetrate deeper into the forest canopy, and S-band, which is sensitive to light vegetation.

This study focuses on modeling the forest AGB using full polarimetric L-band SAR data from the NISAR Simulated UAVSAR campaign. These datasets simulate NISAR satellite acquisitions and include geocoded covariance products in HDF5 format with associated metadata. The CX products used in this study provide HH and HV backscatter, which are further modelled according to the NISAR's Algorithm Theoretical Basis Document. The validation is carried out using the CHM-AGB relationship, which provides the biomass in Mg/ha based on LiDAR data. For parameter optimization, the differential evolution method is employed. As the NISAR ATBD is a nonlinear implicit function that cannot be inverted mathematically, multiple inversion techniques were tested, including both parametric and non-parametric methods. The preliminary results are produced using individual and cumulative acquisitions from multiple sites in North America and Africa for polarisations HH and HV. The model has been tested on binned and unbinned data. We see a decrease in the RMSE for binned data when cumulative acquisitions are used. It is observed that the highest RMSE is 8.61±0.05 Mg/Ha, and the lowest is 7.91±0.03 Mg/Ha, for the LENO site with HV polarization. The temporal invariance of D, 𝛂, and γ is also tested over multiple sites. The soil moisture content and seasonality exhibit significant changes in the backscatter values; however, the S parameter should account for this variation, thereby improving model performance.

Semi-empirical models for AGB estimation have already demonstrated robust performance across boreal, subtropical, and tropical forest types. Nonetheless, not much is studied in low AGB regions. This study re-visited the relationship between AGB and C-band backscatter over low AGB regions in a managed forest site in India, since forest plantations are homogeneous and are closer to true representation.

In this study, we chose four models, Rise-to-max exponential model, Rational model, Water Cloud model and Michaelis-Menten model to explore the estimation of low AGB using the Global Seasonal Sentinel-1 interferometric coherence and backscatter dataset for both VV and VH polarisations,

This study uses SAR Backscatter data and GEDI Lidar Point Cloud metrics to estimate AGB.

The process involves these main steps:

1. Data Collection and Pre-processing: Backscatter is calibrated radiometrically, and Field Above-Ground Biomass measurements were collected through ground surveys. Additionally, Lidar-Derived Canopy height metrics (RH95) are acquired over the study area. 2. Lidar-Derived AGB Derivation: Using the simple power-law relationship to derive Aboveground Biomass (AGB) from RH95 and Field AGB measurements, Lidar-Derived AGB is calculated. 3. Apply Methods: In this study, backscatter data and Lidar-Derived Aboveground Biomass (AGB) are the inputs into the Rise-to-max exponential model, Water Cloud Model , Rational Model, and Michaelis-Menten Model to estimate AGB . Initially, the data is divided into a 60:40 ratio, with 60% of the data for model training and the remaining 40% reserved for testing. This training-testing split was used to ensure that the model parameters are derived from a sufficiently large dataset while allowing for independent validation of model predictions. The model is applied to the training data to estimate the model parameters. This involves running the models through 50 iterations, to optimize the fit between observed and predicted AGB values. The Lidar-Derived AGB is used as a reference for this study, and the model was iteratively adjusted to minimize errors in AGB prediction. After completing the 50 iterations, the mean values of the model parameters are computed to represent the optimal values. 4. AGB Maps: These averaged parameter values are then applied to the respective model to the backscatter data from the testing set, allowing for the estimation of AGB at each spatial location across the study area. The final output is a spatial map that displays the predicted AGB values for each pixel, providing a detailed view of biomass distribution across the landscape. The spatial AGB map is generated using the derived model parameters, which allows for a comprehensive understanding of AGB across the study area.

Seasonal Influence on Polarisation: Both VH and VV backscatter doesn’t seem to have an influence on performance of the models. However, the distribution of data is different in polarisation. In the case of VH Polarisation the change in backscatter trend is captured i.e., at high AGB range, firstly in Dec-Jan-Feb dataset, the backscatter is stable demonstrating almost no sensitivity. With the onset of leaf-fall and/or dry season during Mar-Apr-May, being highly sensitive than the rest. Moving to monsoon season, Jun-Jul-Aug, not only the high soil moisture effects resulting in higher backscatter values but also the leaf-on during this period. And during Sep-Oct-Nov, though the precipitation is the least, because of high canopy coverage, the attenuation is the highest and the backscatter trend is seen decreasing with AGB. This shows the phenology of the low AGB areas can be studied using VH backscatter because of its interaction with leaf and small branches. In the case of VV Polarisation, though the backscatter increase or decrease with AGB is same as VH Polarisation for corresponding datasets, the Mar-Apr-May dataset stands out due to the high sensitivity due to trunk-ground interaction and canopy openness due to leaf-fall.

Seasonal Influence on other parameters: The Mar-Apr-May dataset demonstrated the most favorable conditions for estimating Above-Ground Biomass (AGB). This improved performance is primarily attributed to two key factors. First, the availability of multiple Sentinel-1 SAR acquisitions during this period helps mitigate the impact of non-vegetation related noise in the backscatter signal, aligning with previous findings. Second, the seasonal conditions of Mar-Apr-May which is characterized by dry season, leaf-fall, and minimal understory vegetation lead to a reduction in canopy density. This results in lower Leaf Area Index (LAI) and potentially higher canopy openness, both of which enhance the radar signal’s sensitivity to forest structure. This effect is particularly pronounced in C-band VV polarization, which, as observed, maintains a higher saturation threshold compared to VH polarization (Musthafa, M. et al (2022), Nizalapur, V., Jha (2010)). Collectively, these conditions contribute to improved model sensitivity and result in the lowest Root Mean Square Deviation (RMSD) values across all four models. In contrast, the Sep–Oct–Nov and Dec–Jan–Feb seasons exhibited higher RMSD values,reflecting reduced model performance. During the Sep-Oct-Nov period, residual foliage from late-season growth leads to a denser canopy, while in the Dec-Jan-Feb season, evergreen species retain their leaves. These conditions correspond to elevated LAI values, which increase canopy saturation and diminish radar sensitivity to structural variations in forest biomass. The denser canopy limits the contribution of trunk and ground elements to the backscatter signal, resulting in greater uncertainty in AGB estimation. The lower RMSD observed in the Mar-Apr-May dataset may reflect a period of ecological stability within the study area. This transitional phase, situated between winter dormancy and monsoon-driven growth, is likely associated with relatively stable biomass conditions in natural vegetation and also attributed to dry climatic conditions and no understory. This period coincides with lower soil moisture and precipitation levels, leading to drier climatic conditions. These conditions reduce canopy density and biomass. Similar observations have been reported in Musthafa, M. et al (2022) over other site. The spatial and temporal uniformity contributes to improved model predictability and reinforces the suitability of Mar-Apr-May acquisitions for reliable large-scale biomass mapping.

Methodological Considerations: We do not include any machine learning models, since our objective is to see how SAR backscatter provides a foundational input for AGB modeling, with VV and VH polarizations capturing different aspects of vegetation and soil conditions. Additionally, the availability of multiple SAR acquisitions is expected to minimize the influence of short-term anomalies. While it offers theoretical robustness, its performance is sensitive to seasonal changes in vegetation water content, soil moisture, and canopy structure. The main challenge is the model parameters are difficult to calibrate and estimate accurately. Moreover, at higher biomass levels, saturation effects limit the model’s sensitivity, leading to underestimation. Overall, Michaelis-Menten model followed by rational model provides valuable physical insight, highlights the interaction with the empirical models.

Limitations : The low AGB region where AGB <20 Mg/ha, seems to have noise and ambiguous trends in backscatter. This ambiguity may stem from factors such as low vegetation density and high soil influence. These factors introduce instability in the model outputs, particularly in the lower biomass ranges. Another limitation is the geographical bias introduced by training the model on a spatially limited area. Furthermore, the field plots are located in homogeneous managed forests, which do not fully represent the heterogeneity of the surrounding region. This mismatch will reduce the generalizability of the model to broader forest landscapes. Allometric variability, geolocation errors, and limited representation of forest structure can further compromise AGB retrieval accuracy, which will be explored in this paper. Given that forest allometries and phenological behavior vary across regions, models calibrated in one locale will not perform well elsewhere. The results highlighted that the Mar-Apr-May dataset provided the most favorable conditions for AGB estimation. During this period, the leafless canopy structure allowed for greater wave penetration and enhanced surface-trunk interactions, improving backscatter sensitivity to biomass. In contrast, the Jun-Jul-Aug (monsoon) period showed a limited backscatter dynamic range due to high soil moisture, while the dense canopy in Sep-Oct-Nov and Dec-Jan-Feb led to saturation effects and poor performance, reducing accuracy. Mar-Apr-May dataset appears to be a strong candidate for operational monitoring, offering better estimates with reduced computational burden for low AGB estimations compared to other three datasets, which is a dry seasonal period over the study area. Focusing on this optimal dataset could also significantly reduce the storage burden on annual biomass estimations and assessments, supporting climate modeling, carbon accounting, and forest management strategies. Furthermore, expanding the analysis to phenology and soil characteristics across diverse ecoregions and forest structures will help to track the biomass dynamics.

Forest parameters inversion is essential for quantifying carbon cycle dynamics, monitoring climate change, and sustaining global ecological balance. Recently, long-wavelength Synthetic Aperture Radar (SAR) Tomography (TomoSAR) has gained increasing interest for forest parameter inversion due to its excellent three-dimensional (3D) imaging and penetration capabilities. This paper presents an airborne TomoSAR processing framework, starting from raw radar echo data, proceeding to focused SAR image stacks, followed by 3D tomographic reconstruction, and finally forest parameter inversion. Compared with other SAR imaging algorithms, the Time Domain Back Projection (TDBP) algorithm does not rely on the straight flight path assumption and therefore does not require additional motion compensation operations, leading to higher robustness. To improve its computational efficiency, the algorithm is first accelerated using GPU parallel processing. Furthermore, the two-dimensional antenna pattern and range attenuation effects of the radar signal are taken into account by utilizing high-precision airborne trajectory and attitude data to ensure the radiometric accuracy of the SAR images. Even with high-precision navigation data, residual centimeter-level trajectory errors can lead to non-negligible misregistration and phase errors among SAR image stacks, which in turn degrade the quality of 3D imaging. A two-stage airborne trajectory error correction algorithm is developed for compensating multi-baseline trajectory errors to ensure high-quality TomoSAR imaging. The first stage corrects the trajectory errors of each baseline using strong scatterer echoes based on the Phase Gradient Autofocus (PGA) concept, while the second stage further refines the relative trajectory errors among baselines by exploiting interferometric information from distributed scatterers following the Multisquint approach.

The method is applied to an area of approximately 200 km² in the Saihanba Mechanical Forest Farm, the largest man-made forest and a national forest park in China. Located in the temperate zone of northern China, this forest is dominated by Pinus sylvestris, Larix gmelinii, Betula platyphylla, and Picea asperata. An airborne P-band TomoSAR campaign was conducted over the study area in October, 2023. Twelve tracks of P-band quad-polarization SAR data were acquired, providing azimuth, range, and height resolutions of 0.9 m, 0.75 m, and 5 m, respectively. Based on the TomoSAR results of HH and HV polarizations, the underlying topography and forest height of the Saihanba Forest were retrieved, showing good consistency with a 10 m-resolution airborne LiDAR topography product and a 1 m-resolution forest height product from Meta and the World Resources Institute.

We have been approved to carry out the Biomass Cal/Val project (ID: PP0106173), and plan to perform spaceborne TomoSAR imaging and forest parameter retrieval based on the Biomass Level-1C data acquired over the Saihanba Forest during the Commissioning phase of the Biomass satellite. Based on the previous airborne P-band TomoSAR forest parameters retrieval results, the accuracy of spaceborne P-band TomoSAR forest parameters retrieval will be evaluated, and the effects of temporal decorrelation on spaceborne TomoSAR will be analyzed.