LiDAR, SAR, Optical
Optical
SAR, Optical
LAGIRS 2025
LAGIRS 2025
GFW
Estimating forest biomass through remote sensing is crucial in the context of carbon sequestration, as it is the only way to systematically and periodically quantify the amount of carbon stored in forest ecosystems. Forests act as carbon sinks by absorbing carbon dioxide (CO₂) from the atmosphere during photosynthesis and storing it in their biomass. Among the most widely used satellite systems, ranked in terms of error in aboveground biomass estimation (AGBD), are lidar, SAR, and optical sensors (each with its own challenges). The development of accurate estimation algorithms is an open issue. However, estimates from these instruments are regularly used to monitor deforestation, forest degradation, and natural regeneration processes, as input for decision-making in conservation, reforestation, and sustainable management policies. These data are essential for meeting international climate-related commitments, such as the Paris Agreement and REDD+ (Reducing Emissions from Deforestation and Forest Degradation) mechanisms.
Deforestation poses a significant threat to natural ecosystems, particularly in Argentina’s Chaco region—one of the world’s most rapidly changing forest areas. This study focuses on the detection of sudden deforestation events, where forest cover is rapidly removed within a few months. Monitoring such changes across vast areas requires the use of satellite-based vegetation indices, such as EVI and NDVI from MODIS. However, accurately identifying deforestation events is challenging due to seasonal variability, sensor noise, data gaps, and algorithmic inconsistencies. These factors can obscure true deforestation signals or generate false positives. To address these issues, a robust detection approach must explicitly model time-series dynamics—capturing trends, seasonality, and uncertainty—to reliably distinguish genuine deforestation breakpoints from natural variation and noise. In this paper, three models for the detection of breakpoints in EVI time series were proposed: a simple z-score anomaly detector, and two fully Bayesian models; one temporally uncorrelated and one fully correlated. Results indicate that the Bayesian schemes significantly improve over the naive approach (zscore: AUC=0.921, F1-score=0.870, Bayes: AUC=0.959, F1-score=0.925), for a reasonable cost in computing time (x1000).
Following the PROSAT II call from the National Commission for Space Activities (CONAE), a system was developed using active microwave polarimetric satellite information from the Argentine SAOCOM Mission, together with other optical and radar satellite data, for monitoring wetlands in Argentina (in line with the use of satellite information for wetland inventory programs). Work was carried out in pilot wetland landscape units: the Ramsar Site in the Upper Delta (Entre Ríos Province), Concepción del Uruguay (Entre Ríos Province), and Bajo de los Saladillos (Santa Fe Province). The first objective was to develop a catalog of backscatter coefficients (active microwave, C and L bands) and spectral responses (visible and reflective infrared) associated with the wetland types in the study areas. The catalog is an interactive dashboard that displays graphs for each wetland type, allowing users to view spectral signatures or characteristic backscatter (mean values and deviation or range of variation) for a selected period, as well as the temporal dynamics of backscatter or synthetic indices derived from optical satellite data. The dashboard was designed with two modalities: a) exploratory, in which each user can obtain graphs by freely filtering by scene type, date, and wetland type; b) guided, in which pre-built graphs are accessed, accompanied by an interpretation made by the consultants (for example, differences between dry and wet periods, or analysis of differences between wetland types). It is useful for managers involved in wetland mapping, as it provides a summary and interpretation of SAR and optical information.
Deforestation poses a significant threat to natural ecosystems, particularly in Argentina’s Chaco region—one of the world’s most rapidly changing forest areas. This study focuses on the detection of sudden deforestation events, where forest cover is rapidly removed within a few months. Monitoring such changes across vast areas requires the use of satellite-based vegetation indices, such as EVI and NDVI from MODIS. However, accurately identifying deforestation events is challenging due to seasonal variability, sensor noise, data gaps, and algorithmic inconsistencies. These factors can obscure true deforestation signals or generate false positives. To address these issues, a robust detection approach must explicitly model time-series dynamics—capturing trends, seasonality, and uncertainty—to reliably distinguish genuine deforestation breakpoints from natural variation and noise. In this paper, three models for the detection of breakpoints in EVI time series were proposed: a simple z-score anomaly detector, and two fully Bayesian models; one temporally uncorrelated and one fully correlated. Results indicate that the Bayesian schemes significantly improve over the naive approach (zscore: AUC=0.921, F1-score=0.870, Bayes: AUC=0.959, F1-score=0.925), for a reasonable cost in computing time (x1000).
Deforestation poses a significant threat to natural ecosystems, particularly in Argentina’s Chaco region—one of the world’s most rapidly changing forest areas. This study focuses on the detection of sudden deforestation events, where forest cover is rapidly removed within a few months. Monitoring such changes across vast areas requires the use of satellite-based vegetation indices, such as EVI and NDVI from MODIS. However, accurately identifying deforestation events is challenging due to seasonal variability, sensor noise, data gaps, and algorithmic inconsistencies. These factors can obscure true deforestation signals or generate false positives. To address these issues, a robust detection approach must explicitly model time-series dynamics—capturing trends, seasonality, and uncertainty—to reliably distinguish genuine deforestation breakpoints from natural variation and noise. In this paper, three models for the detection of breakpoints in EVI time series were proposed: a simple z-score anomaly detector, and two fully Bayesian models; one temporally uncorrelated and one fully correlated. Results indicate that the Bayesian schemes significantly improve over the naive approach (zscore: AUC=0.921, F1-score=0.870, Bayes: AUC=0.959, F1-score=0.925), for a reasonable cost in computing time (x1000).
The Pilar Municipal Nature Reserve (Buenos Aires) protects a sector of the Luján River floodplain with diverse plant communities. Our objective was to evaluate the feasibility of estimating vegetation height from information obtained with an Unmanned Aerial Vehicle (UAV). In May 2025, we conducted automated grid flights over cleared forest (12 ha) and grassland (13 ha) environments using the DJI Phantom 4 Pro UAV, equipped with a 20-megapixel RGB camera. The photographs were processed using digital photogrammetric techniques that allow the generation of dense point clouds. We obtained georeferenced orthomosaics, surface elevation models, and terrain elevation models (DTMs). In the field, we geolocated plant individuals with Differential GNSS (dGNSS) and measured terrain altitude. We also used a laser rangefinder to estimate the height of individual plants of *Celtis tala*, *Phytolacca dioica*, *Solanum granulosum-leprosum*, *Salix humboldtiana*, *Morus* sp., *Baccharis* sp., and *Cortaderia selloana*. The consistency between the DTMs obtained from dGNSS measurements and those obtained from UAV data was low in areas of high tree density. Vegetation height estimates showed a very good fit with field measurements, regardless of the DTM used (Pearson correlation, r = 0.98 to 0.95). However, biases in individual height were greater when using the UAV-derived DTM. The results highlight the potential of UAV data acquisition and photogrammetry to characterize vegetation structure.
La deforestación representa una amenaza significativa para los ecosistemas naturales, en particular en la región del Chaco argentino, una de las áreas forestales con mayor cambio del mundo. Este estudio se centra en la detección de eventos repentinos de deforestación, donde la cubierta forestal se elimina rápidamente en pocos meses. El monitoreo de estos cambios en vastas áreas requiere el uso de índices de vegetación satelitales, como el EVI y el NDVI de MODIS. Sin embargo, la identificación precisa de los eventos de deforestación es un desafío debido a la variabilidad estacional, el ruido de los sensores, la falta de datos y las inconsistencias algorítmicas. Estos factores pueden ocultar las verdaderas señales de deforestación o generar falsos positivos. Para abordar estos problemas, un enfoque de detección robusto debe modelar explícitamente la dinámica de las series temporales (capturando tendencias, estacionalidad e incertidumbre) para distinguir de manera confiable los puntos de quiebre de la deforestación genuina de la variación natural y el ruido. En este artículo, se propusieron tres modelos para la detección de puntos de quiebre en series temporales de EVI: un detector simple de anomalías de z-score y dos modelos completamente bayesianos; uno no correlacionado temporalmente y otro completamente correlacionado. Los resultados indican que los esquemas bayesianos mejoran significativamente con respecto al enfoque ingenuo (z-score: AUC=0,921, F1-score=0,870, Bayes: AUC=0,959, F1-score=0,925), por un coste razonable en tiempo de cálculo (x1000).
La deforestación representa una amenaza significativa para los ecosistemas naturales, en particular en la región del Chaco argentino, una de las áreas forestales con mayor cambio del mundo. Este estudio se centra en la detección de eventos repentinos de deforestación, donde la cubierta forestal se elimina rápidamente en pocos meses. El monitoreo de estos cambios en vastas áreas requiere el uso de índices de vegetación satelitales, como el EVI y el NDVI de MODIS. Sin embargo, la identificación precisa de los eventos de deforestación es un desafío debido a la variabilidad estacional, el ruido de los sensores, la falta de datos y las inconsistencias algorítmicas. Estos factores pueden ocultar las verdaderas señales de deforestación o generar falsos positivos. Para abordar estos problemas, un enfoque de detección robusto debe modelar explícitamente la dinámica de las series temporales (capturando tendencias, estacionalidad e incertidumbre) para distinguir de manera confiable los puntos de quiebre de la deforestación genuina de la variación natural y el ruido. En este artículo, se propusieron tres modelos para la detección de puntos de quiebre en series temporales de EVI: un detector simple de anomalías de z-score y dos modelos completamente bayesianos; uno no correlacionado temporalmente y otro completamente correlacionado. Los resultados indican que los esquemas bayesianos mejoran significativamente con respecto al enfoque ingenuo (z-score: AUC=0,921, F1-score=0,870, Bayes: AUC=0,959, F1-score=0,925), por un coste razonable en tiempo de cálculo (x1000).
La Reserva Natural Municipal del Pilar (Buenos Aires) protege un sector de la planicie de inundación del río Luján, con diversas comunidades vegetales. Nuestro objetivo fue evaluar la factibilidad de estimar la altura de la vegetación a partir de información obtenida con un Vehículo Aéreo No Tripulado (VANT). En mayo de 2025 realizamos vuelos automatizados en grillas sobre ambientes de talar (12 ha) y pastizal (13 ha) con el VANT DJI Phantom 4 Pro, equipado con una cámara RGB de 20 megapíxeles. Las fotografías fueron procesadas mediante técnicas fotogramétricas digitales que permiten generar nubes de puntos densas. Obtuvimos ortomosaicos georreferenciados, modelos de elevación de superficie y de elevación del terreno (DTM). A campo, geolocalizamos individuos vegetales con un GNSS Diferencial (dGNSS) y medimos la altitud del terreno. Asimismo, con un telémetro láser estimamos la altura de individuos vegetales de *Celtis tala*, *Phytolacca dioica*, *Solanum granulosum-leprosum*, *Salix humboldtiana*, *Morus* sp., *Baccharis* sp. y *Cortaderia selloana*. La consistencia entre los DTM obtenidos con las mediciones dGNSS y con información VANT fue baja en zonas de alta densidad arbórea. Las estimaciones de altura de la vegetación mostraron muy buen ajuste con las mediciones a campo, independientemente del DTM utilizado (correlación de Pearson, r = 0.98 a 0.95). Sin embargo, los sesgos en la altura de los individuos fueron mayores al usar el DTM elaborado con VANT. Los resultados destacan el potencial de la adquisición de datos con VANT y de la fotogrametría para caracterizar la estructura de la vegetación.
In this work we present an innovative and low-cost way to estimate the dielectric constant of real soils -which is relatedto soil moisture- using GPS satellite signal. The trajectory of the satellites was followed with an automated device,and with an antenna type patch was measured the interference pattern generated by the direct GPS signal and thatreflected by the ground. To analyze the acquired signal, an electromagnetic dispersion model based on the methodof small perturbation was used, which may include stratifications in the underground and from which the dielectricconstant that best fits the measurements was obtained. These values were compared with the obtained using a sensorthat measures the soil dielectric constant textit in-situ. The comparisons show that the method based on the interferencepattern provides satisfactory results to estimate the dielectric constant of real bare soils. Finally, proposals are outlinedto improve the results obtained in this work.
En este trabajo presentamos una forma innovadora y de bajo costo para estimar la constante dieléctrica de suelos reales - relacionada con la humedad del mismo - utilizando señal satelital GPS. Con un dispositivo automatizado se siguió la trayectoria de los satélites y con una antena tipopatchse realizaron las mediciones del patrón de interferencia generado por la señal directa del GPS y la reflejada por el suelo. Para analizar la señal adquirida se utilizó un modelode dispersión electromagnética basado en el método de pequeñas perturbaciones, que puede incluir estratificaciones en el medio subterráneo y a partir del cual se obtuvo la constante dieléctrica que mejor ajusta a las mediciones. Estos valores fueron comparados contra los obtenidos con un sensor de humedad que mide la constante dieléctrica in-situ. Las comparaciones muestran que el método basado en el patrón de interferencia brinda resultados satisfactorios para estimar la constante dieléctrica del suelo. Finalmente, se delinean propuestas para mejorar los resultados obtenidos eneste trabajo.
Soil moisture is a key geophysical variable that can be estimated using remote-sensing techniques by making use of the known relation between soil backscattering and the dielectric constant in the microwave regime. However, since SAR system observations depend on geometrical and dielectric surface parameters (besides instrument parameters like operation frequency, incidence angle, and received/transmitted polarization), the uncertainties associated with a given retrieval scheme are difficult to evaluate. In this letter, these uncertainties associated with the estimation of soil dielectric constant from a single quad-pol SAR image are studied using a physically based interaction model (i.e., a two-layer version of the small perturbation method (SPM) model at second order) coupled with a Bayesian approach. The overall scheme was validated using SAOCOM quad-pol data and in situ soil dielectric constant measurements in experimental agricultural plots in Argentina. Both theoretical end-to-end experiments and actual retrieval from real SAR data were implemented. From the simulations, the intrinsic ambiguities in the estimations of soil dielectric constant from a single image were studied, and the benefits of using two images with different incidence angles were discussed. Finally, by analyzing SAOCOM data using the proposed retrieval scheme, soil dielectric constants were estimated and compared with in situ measurements, with a root-mean-square error (RMSE) of ≤2.
Soil moisture is a key geophysical variable that can be estimated using remote-sensing techniques by making use of the known relation between soil backscattering and the dielectric constant in the microwave regime. However, since SAR system observations depend on geometrical and dielectric surface parameters (besides instrument parameters like operation frequency, incidence angle, and received/transmitted polarization), the uncertainties associated with a given retrieval scheme are difficult to evaluate. In this letter, these uncertainties associated with the estimation of soil dielectric constant from a single quad-pol SAR image are studied using a physically based interaction model (i.e., a two-layer version of the small perturbation method (SPM) model at second order) coupled with a Bayesian approach. The overall scheme was validated using SAOCOM quad-pol data and in situ soil dielectric constant measurements in experimental agricultural plots in Argentina. Both theoretical end-to-end experiments and actual retrieval from real SAR data were implemented. From the simulations, the intrinsic ambiguities in the estimations of soil dielectric constant from a single image were studied, and the benefits of using two images with different incidence angles were discussed. Finally, by analyzing SAOCOM data using the proposed retrieval scheme, soil dielectric constants were estimated and compared with in situ measurements, with a root-mean-square error (RMSE) of ≤2.
In this study, the estimation of soil moisture from a single quad-pol Synthetic Aperture Radar (SAR) image from SAOCOM mission along with a physically based interaction model (two-layer version of the small perturbation method model at the second order) was validated with in-situ measured data in the central region of Argentina. The relationship between microwave backscattering and dielectric constant (ϵ) allows for precise estimation of ϵ effectively disambiguating the influence of soil roughness in bare soil conditions. Measured dielectric constants from the study area are used for validation.
In this study, the estimation of soil moisture from a single quad-pol Synthetic Aperture Radar (SAR) image from SAOCOM mission along with a physically based interaction model (two-layer version of the small perturbation method model at the second order) was validated with in-situ measured data in the central region of Argentina. The relationship between microwave backscattering and dielectric constant (ϵ) allows for precise estimation of ϵ effectively disambiguating the influence of soil roughness in bare soil conditions. Measured dielectric constants from the study area are used for validation.
Soil moisture (SM) retrieval using signals of opportunity based on specularly reflected signals has gained significant attention in the past two decades. Specifically, with the Cyclone Global Navigation Satellite System (CyGNSS), the reflected signal is often modeled in its simplest form, utilizing the Fresnel reflection coefficients for a semi-infinite dielectric medium corrected with an effective roughness parameter. Within this framework, for bare soils condition, only two parameters need to be inferred: the dielectric permittivity ε (related to SM) and the effective roughness σ. Although this approach is relatively simple, our results show that both the estimated dielectric constant and the modeled reflectivity consistently overestimate CyGNSS observations. To address these overestimations, we propose a model where the reflected signal is scattered by a medium comprising two layers: one with a finite thickness d and permittivity ε1 and the other semi-infinite with permittivity ε2. We observe that both the in-situ measurements of ε1 and the reflectivity reported by CyGNSS align with the optimal values obtained from the fit, resulting in a 73% reduction in root mean square error when compared to the traditional approach. To further enhance SM retrieval, we propose incorporating full polarimetric images from SAOCOM. This will allow us to combine the low revisit time of CyGNSS with the high spatial resolution offered by SAOCOM.
Soil moisture (SM) retrieval using signals of opportunity based on specularly reflected signals has gained significant attention in the past two decades. Specifically, with the Cyclone Global Navigation Satellite System (CyGNSS), the reflected signal is often modeled in its simplest form, utilizing the Fresnel reflection coefficients for a semi-infinite dielectric medium corrected with an effective roughness parameter. Within this framework, for bare soils condition, only two parameters need to be inferred: the dielectric permittivity ε (related to SM) and the effective roughness σ. Although this approach is relatively simple, our results show that both the estimated dielectric constant and the modeled reflectivity consistently overestimate CyGNSS observations. To address these overestimations, we propose a model where the reflected signal is scattered by a medium comprising two layers: one with a finite thickness d and permittivity ε1 and the other semi-infinite with permittivity ε2. We observe that both the in-situ measurements of ε1 and the reflectivity reported by CyGNSS align with the optimal values obtained from the fit, resulting in a 73% reduction in root mean square error when compared to the traditional approach. To further enhance SM retrieval, we propose incorporating full polarimetric images from SAOCOM. This will allow us to combine the low revisit time of CyGNSS with the high spatial resolution offered by SAOCOM.
In recent years, in addition to the traditional uses of forest plantations, there has been an interest in their capacity to capture and fix carbon in their structures, acting as greenhouse gas sinks. The Global Ecosystem Dynamics Investigation (GEDI) project features a light detection and ranging sensor aimed at quantifying the distribution of biomass and carbon in woody vegetation worldwide. Although the GEDI level 4A product (L4A) generates aboveground biomass density (AGBD) estimates on a global scale using semi-empirical relationships between the energy reflected at a given height (rh metric) and AGBD, its forest segmentation is coarse and presents limitations at a local scale, which can result in less accurate estimates. We evaluated the accuracy of GEDI L4A AGBD estimates and in situ AGBD data obtained from forest inventories. The analysis of the relationship between in situ AGBD and GEDI L4A product AGBD estimates shows that, although the correlation is generally good, some plots exhibit high variability in GEDI AGBD values despite having intermediate values of in situ AGBD [root mean squared error (RMSE) = 89.1 Mg/h]. Moreover, we found that these underestimations are associated with low values of rh50. In particular, these low values of rh50 (related to large AGBD RMSE errors) were not related to thinned/non-thinned forests but were related to understory conditions (less error for the “very dense” understories). Then, we explored landcover misassignment; as the Pinus sp. plantations correspond to implanted North American evergreen needleleaf forests (ENFs), we reconstructed the L4A product for this region and South America, using the corresponding expressions to estimate AGBD. The results show an improvement in the overall correlation and a reduction in errors, with RMSE reducing to 68.4 Mg/h for South American ENF and 64.2 Mg/h for North American ENF. Finally, we discuss ways to improve AGBD estimations (or flag wrong ones) using all available values of the rh curve.
In recent years, in addition to the traditional uses of forest plantations, there has been an interest in their capacity to capture and fix carbon in their structures, acting as greenhouse gas sinks. The Global Ecosystem Dynamics Investigation (GEDI) project features a light detection and ranging sensor aimed at quantifying the distribution of biomass and carbon in woody vegetation worldwide. Although the GEDI level 4A product (L4A) generates aboveground biomass density (AGBD) estimates on a global scale using semi-empirical relationships between the energy reflected at a given height (rh metric) and AGBD, its forest segmentation is coarse and presents limitations at a local scale, which can result in less accurate estimates. We evaluated the accuracy of GEDI L4A AGBD estimates and in situ AGBD data obtained from forest inventories. The analysis of the relationship between in situ AGBD and GEDI L4A product AGBD estimates shows that, although the correlation is generally good, some plots exhibit high variability in GEDI AGBD values despite having intermediate values of in situ AGBD [root mean squared error (RMSE) = 89.1 Mg/h]. Moreover, we found that these underestimations are associated with low values of rh50. In particular, these low values of rh50 (related to large AGBD RMSE errors) were not related to thinned/non-thinned forests but were related to understory conditions (less error for the “very dense” understories). Then, we explored landcover misassignment; as the Pinus sp. plantations correspond to implanted North American evergreen needleleaf forests (ENFs), we reconstructed the L4A product for this region and South America, using the corresponding expressions to estimate AGBD. The results show an improvement in the overall correlation and a reduction in errors, with RMSE reducing to 68.4 Mg/h for South American ENF and 64.2 Mg/h for North American ENF. Finally, we discuss ways to improve AGBD estimations (or flag wrong ones) using all available values of the rh curve.