A new satellite-mapping framework provides 10-meter maps of wetlands across 43 global cities, distinguishing 18 wetland classes and tracking changes from 2016 to 2024. The dataset supports evaluation of conservation, restoration, and urban development, and could strengthen Ramsar accreditation reviews.
A research team compared 64 satellite-derived estimates with flux-tower observations, identifying reliable approaches and mapping global patterns from 2001 to 2015. Carbon-water coupled products, such as PMLv2 and BESSv2, reproduced flux-tower observations more accurately than independent products, providing a more consistent basis for...
A new study using satellite mapping identifies the Amazon forest restoration priorities and provides a framework to prioritize protection, natural regeneration, and restoration. The framework distinguishes between vulnerable, isolated regrowth and mature patches suitable for long-term protection.
A team developed a new active-passive remote sensing fusion approach, APEX, to generate continuous regional aerosol profiles at high resolution. The algorithm preserves fine atmospheric structure and supports planetary boundary layer height estimation. Future work will address coverage gaps and improve continuity.
A new framework combines satellite imagery, environmental DNA, and machine learning to predict fish biodiversity. The approach outperforms conventional methods, offering a scalable and noninvasive pathway for biodiversity mapping, conservation planning, and freshwater surveys.
The article outlines a roadmap for more accurate and interpretable Earth observation, highlighting limitations in existing systems. It proposes several technical pathways to address these challenges, including virtual constellations, AI agents, and multimodal real-time processing.
Researchers used LuTan-1 to test hybrid-polarimetric radar's effectiveness for various land covers, finding it performs well for many natural and agricultural surfaces but struggles with sloped or directionally organized targets. The study provides practical guidance for future Earth-observation missions.
The new PROSPECT-S model accurately estimates chlorophyll, water, and dry matter content in wheat spikes, allowing for non-destructive monitoring of physiological status. The model demonstrates greater stability across developmental stages than existing models, enabling improved light interception and photosynthetic capacity estimations.
Researchers developed a coarse-to-fine saliency-driven maritime ship detection network to improve detection accuracy in complex scenes. The framework achieves high F1 scores and mean average precision, outperforming strong baseline detectors.
A new AI model, PLGMamba, improves hyperspectral image super-resolution by combining local spectral similarity with global feature modeling. The approach achieves stronger reconstruction accuracy than several leading super-resolution methods and overcomes limitations of existing models.
Researchers have developed a deep-learning framework to reconstruct a global, high-resolution nighttime light dataset from 1992 to 2024. The new product improves upon existing datasets by reducing saturation-related bias and better capturing temporal changes in urbanization, economic activity, and human development.
A new framework called CH4Vision uses hyperspectral satellite imagery to estimate methane flux directly, improving accuracy and robustness. The method incorporates plume morphology and machine learning to infer emission rates, increasing reliability of satellite-based quantification.
A new study uses drone-based hyperspectral imaging to monitor changes in grassland ecosystems under different grazing pressures. The researchers found that biomass declines with increasing grazing intensity, while certain nutrient-related traits increase, suggesting a shift toward more stress-tolerant plant strategies.
A new long-term land cover dataset for China covers 1985 to 2022 and provides annual 30-meter maps. The dataset outperforms existing products, achieving high accuracy nationwide and in regional tests. It reveals major shifts in impervious surfaces, water bodies, forests, cropland, and grassland across China.
Researchers found that directional emissivity changes with viewing angle, canopy density, and tree arrangement in forests. Forests with higher leaf area index and more homogeneous tree distribution show relatively high mean emissivity values.
A new SAR interpretation method links bright scattering patterns to specific 3D structures, improving target analysis and physically grounded SAR simulation. The approach outperformed existing methods in image quality evaluation, processing time, and physical interpretability.
Researchers found a more reliable way to track seasonal productivity and drought stress in dryland ecosystems by combining multiple optical signals. This approach could improve drought monitoring and carbon-cycle modeling, supporting climate forecasting and ecosystem management.
A new study found that ignoring woody parts and using one representative leaf spectrum have minor effects on simulated waveforms, while assuming a uniform foliage area volume density causes larger errors. The research provides guidance for designing faster analytical models that remain physically reliable.
Researchers reconstructed annual canopy height maps for southern China, showing a 61% rise in average canopy height over three decades. Plantation forests grew faster than secondary forests, but ultimately reached greater heights. The study demonstrates the potential of satellite-based monitoring for tracking forest growth and informin...
A new machine learning framework for satellite-based CO2 retrievals provides highly accurate concentration estimates while quantifying uncertainty, outperforming traditional physics-based methods in speed and accuracy.
This study reveals that 29.2% of global drylands have significantly greened from 2001-2024, while only 4.9% experienced significant browning. Human activities, including cropland expansion and irrigation, have a more significant impact on greening than climate or CO₂ fertilization alone.
Researchers developed DeepForest AI to capture under-canopy vegetation using synthetic-aperture imaging and 3D neural networks. The technology improves deep-layer reflectance accuracy by 2–12 times and supports vegetation-index calculation for ecological monitoring.
A joint research team developed an AI-driven vision model that delivers sub-meter accuracy in estimating tree heights from RGB satellite images. The approach achieved near-lidar accuracy, enabling precise monitoring of forest biomass and carbon storage over large areas.
Researchers introduce ACA-SIM, a neural-network-based algorithm that reduces errors and striping artifacts in ocean color products. The method outperforms existing models in turbid water and complex-aerosol conditions, providing accurate retrievals of ocean color and water quality parameters.
Researchers developed an adaptive ensemble learning framework that combines hyperspectral and LiDAR data to precisely identify vegetation species in karst wetlands, achieving up to 92.77% accuracy. The approach uses local interpretable model-agnostic explanations to visualize how each feature contributes to the decision-making process.
Researchers have developed a new hybrid model that combines physical radiative transfer equations with neural networks to accurately retrieve soil moisture across China. The model achieved outstanding accuracy and generalization at a 1-km resolution, providing a powerful approach for global hydrological and climatic monitoring.
A recent study analyzed satellite data from 2000 to 2020 and found that while ecohealth began improving after 2012, degradation still dominates in Asia's drylands. The region's land ecosystems have changed significantly, with about 22% of the land remaining degraded and only 13% showing recovery.
Researchers develop a refined retrieval algorithm that leverages Diurnal Amplitude Variation in L-band brightness temperature to capture soil FT dynamics more precisely. The enhanced approach demonstrates stronger consistency with ERA5-Land and SMAP data.
Researchers developed an improved algorithm that reduces bias and error rates in satellite observations by integrating temperature-sensitive absorption of pure seawater. Temperature deviations of more than 10 °C can compromise accuracy, but the enhanced approach offers reliable monitoring across dynamic environments.
A new multimodal framework integrates satellite imagery, near-surface cameras, and AI to provide daily, high-resolution maps of global land changes. The approach improves accuracy and captures abrupt transitions such as snow cover and flooding.
A new deep learning framework, Themeda, achieves high accuracy in predicting annual land cover categories across Australia's vast savanna biome. By integrating satellite data with environmental predictors, the model delivers probabilistic outputs that reflect uncertainty and captures ecological shifts at multiple spatial scales.
Researchers developed an AI-powered framework to optimize urban space use in Beijing, proposing strategies for repurposing vacated areas into farmland, forests, or public facilities. The study demonstrated the framework's predictive power and adaptability, highlighting tensions between policy goals and economic incentives.
A new deep-learning model, AngleNet, significantly improves the retrieval accuracy of atmospheric relative humidity profiles from ground-based microwave radiometers. The model captures complex nonlinear relationships between brightness temperature and humidity profiles, leading to better understanding and forecasting of atmospheric the...
A new study has developed advanced, non-destructive models to measure the aboveground biomass and volume of Populus euphratica, a keystone tree in desert riparian forests. These models reduce errors by nearly half compared to traditional equations, providing new tools for forest conservation and carbon storage assessment.
The NISAR mission's multi-scale algorithm retrieves soil moisture at resolutions as fine as 100 meters, improving agricultural productivity and water resource management. Validation results show the algorithm met accuracy goals, with a root mean square error of less than 0.06 m³/m³.
Researchers developed satellite pixel-scale estimation models to quantify aquatic vegetation coverage and area, addressing challenges in ecological monitoring. The models offer a powerful tool for long-term analysis of lake ecosystem dynamics and carbon sink assessments.
Researchers analyzed data from CloudSat and CALIPSO satellites to understand how different atmospheric mechanisms shape cirrus cloud properties. The study found that convective and frontal systems dominate cirrus cloud formation globally, with convective clouds exhibiting higher ice water content and larger ice crystals.
The RAMI-V study evaluated various models for simulating radiation measurements over plant canopy surfaces, revealing agreement within 2% uncertainty thresholds. The findings will influence future model development and remote sensing methodologies, particularly for the Copernicus program.
A study using Landsat satellite data reveals significant increases in chlorophyll-a concentrations, indicating worsening eutrophication on the Qinghai-Tibet Plateau. The research forecasts future Chla levels until 2100, highlighting the need for targeted water management strategies to mitigate eutrophication and preserve lake health.
Researchers developed a multisensor approach combining optical and L-band InSAR data to improve snow water equivalent (SWE) retrievals. The study found that combining optical and radar data can enhance SWE estimations, particularly in mountain environments.
The EBD dataset leverages deep learning to speed up disaster recovery efforts, providing rapid and accurate building damage assessments. The semi-automated labeling system reduces manual workload by 80% and improves model accuracy, particularly in situations with limited labeled samples.
The new dataset provides a high-quality, readily available resource for researchers to streamline environmental monitoring and land use analysis in China. It addresses major challenges like cloud cover, sensor inconsistencies, and data gaps, offering a reliable, seamless dataset for large-scale analyses and informed decision-making.
A new method harnessing CYGNSS significantly improves soil moisture retrieval by eliminating vegetation interference and removing dependency on external datasets. The technique enables standalone, high-frequency monitoring of soil moisture, offering a transformative tool for climate research, agriculture, and disaster management.
A novel deep learning approach estimates TOA shortwave radiation using DSCOVR/EPIC satellite images, overcoming traditional challenges and improving accuracy. The method is validated against established CERES data products with a high correlation rate.
Researchers developed a new global flood dataset, FloodPlanet, which improves flood detection accuracy by up to 15.6% using high-resolution commercial satellite imagery and public sensor data. The dataset enables more reliable global inundation response systems and opens the door for AI-driven environmental monitoring tools.
A new algorithm, Geo-SETRA, reconstructs 3D models with denser and more accurate point clouds than traditional SAR techniques. It preserves fine architectural details while minimizing post-processing.
Researchers unveil a simple yet effective tool to track Spartina alterniflora, an aggressive invasive plant species threatening coastal wetlands. The new Spartina alterniflora Index (SAI) offers precise, large-scale mapping of the species using freely accessible Sentinel-2 imagery.
A recent study unveiled an innovative data-driven model that disentangles human-induced and natural water consumption in croplands, shedding light on the sustainability of arid lake ecosystems. The model provides actionable insights for sustainable resource management, quantifying the impact of agricultural expansion on water resources.
A novel Spectral Gaussian Mixture Model enhances classification accuracy by leveraging physiological traits and adapts to environmental variations. The model was tested across four regions, achieving an average accuracy of 87.5% to 90.7%, and has the potential to transform precision agriculture by providing real-time monitoring.
A new algorithm developed using data from China's Fengyun-3F satellite significantly improves the accuracy of key climate metrics, including TOA albedo, SWDR, and PAR. The innovation achieves remarkable correlation coefficients and rivals NASA's CERES products while reducing bias in SWDR estimation.