Remote Sensing of Environment最新文献

{"title":"A hybrid physics-informed and data-driven model for estimating ocean internal wave phase speeds from remote sensing imagery","authors":"Guangxi Cui ,&nbsp;Zhongya Cai ,&nbsp;Zhiqiang Liu","doi":"10.1016/j.rse.2026.115247","DOIUrl":"10.1016/j.rse.2026.115247","url":null,"abstract":"<div><div>The propagation speed of internal waves is a fundamental parameter for understanding their physical mechanisms, dynamic behavior, and environmental impact. However, traditional estimation methods are typically based on numerical simulations or sparse <em>in-situ</em> observations, which limit their accuracy and scalability, and results in a significant scarcity of available phase speed datasets. To overcome these challenges, we propose a physics-informed and data-driven model for estimating internal wave phase speed from satellite imagery. The proposed model incorporates three key innovations: (1) the integration of theoretical equations (KdV, BO, and eKdV equations) as physical constraints to ensure consistency with real-world ocean dynamics; (2) the adoption of an adaptive ensemble learning framework that fuses data-driven and physical-informed features to improve model robustness and prediction accuracy; and (3) the introduction of a transfer learning strategy to mitigate discrepancies between theoretical predictions and observational real-world internal wave results. Experimental results demonstrate that the model achieves superior performance across varying water depths, with an average RMSE of 0.04 m/s, MRE of 2.5%, and R² of 98.8% on the testing set. Additionally, the model was applied to the South China Sea, revealing a distinct propagation pattern: average phase speed initially increased (from 2.427 m/s to 2.53 m/s), then decreased (to 1.464 m/s), and subsequently increased again (to 1.703 m/s) as internal waves propagated westward across the Dongsha Islands and Hainan Island. The model was further validated on a global scale, achieving an average percentage error of 4.95%, confirming its scalability and generalization capability. This study presents an efficient and automated approach for accurately retrieving internal wave phase speed.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"335 ","pages":"Article 115247"},"PeriodicalIF":11.4,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146072688","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Generating an annual 30 m rice cover product for monsoon Asia (2018–2023) using harmonized Landsat and Sentinel-2 data and the NASA-IBM geospatial foundation model","authors":"Husheng Fang ,&nbsp;Shunlin Liang ,&nbsp;Wenyuan Li ,&nbsp;Yongzhe Chen ,&nbsp;Han Ma ,&nbsp;Jianglei Xu ,&nbsp;Yichuan Ma ,&nbsp;Tao He ,&nbsp;Feng Tian ,&nbsp;Fengjiao Zhang ,&nbsp;Hui Liang","doi":"10.1016/j.rse.2026.115256","DOIUrl":"10.1016/j.rse.2026.115256","url":null,"abstract":"<div><div>Timely and accurate information on rice distribution is crucial for food security and agricultural decision-making. Monsoon Asia hosts the world's largest rice cultivation area and is dominated by smallholder farming systems with fragmented farmlands, making it challenging for accurate rice field mapping based on coarse-resolution satellite imagery. Machine learning methods have been widely used in remote sensing classification tasks due to their high efficiency and accuracy. However, their performance can be limited in scenarios where sufficient training samples are unavailable. Recently, the advent of geospatial foundation models has offered a promising solution with a pre-training strategy on massive unlabeled satellite data, which achieves higher accuracy than traditional deep learning models in downstream tasks. This study proposes a conceptual framework for adapting geospatial foundation models (GFMs) to large-scale agricultural mapping, representing the first application of the NASA–IBM Prithvi model for continental-scale rice monitoring across Monsoon Asia. We first automatically generate rice labels from multiple existing regional rice products, then fine-tune Prithvi using the Harmonized Landsat and Sentinel-2 satellite data with high spatiotemporal resolution. Finally, the 30 m rice distribution product for Monsoon Asia from 2018 to 2023 is generated. Validation shows that the overall accuracy is 84.14% for the entire Monsoon Asia, with accuracies ranging from 83.07% to 90.06% across four climatic zones. Our product exhibits strong consistency with existing scattered local-scale high-resolution products and shows improved accuracy compared to the MODIS-based and SAR-based products covering the Monsoon Asia. This study indicates that geospatial foundation models can play a key role in remote sensing. Combining geospatial foundation models with massive satellite data has enormous potential for future applications. The rice cover product and code for fine tuning are available at <span><span>www.glass.hku.hk</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"335 ","pages":"Article 115256"},"PeriodicalIF":11.4,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146015022","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Daily 4D landslide movements monitoring via InSAR: A fusion framework integrating physics-based and data-driven models","authors":"Wanji Zheng ,&nbsp;Min Zhao ,&nbsp;Bo Huang ,&nbsp;Aoqing Guo ,&nbsp;Jun Hu","doi":"10.1016/j.rse.2026.115263","DOIUrl":"10.1016/j.rse.2026.115263","url":null,"abstract":"<div><div>Continuous monitoring of deep-seated slow-moving landslides is a critical measure to mitigate risks posed by catastrophic events, especially under increasing extreme weather conditions. As a non-contact, high-resolution measurement technique, space-borne interferometric synthetic aperture radar (InSAR) has been widely applied in landslide monitoring. However, when used for continuous monitoring, challenges remain, including incomplete monitoring dimensions, slow update efficiency, and insufficient temporal resolution. To address these, this paper proposes a fusion framework of physics-based and data-driven models, built on a Kalman Filter, to rapidly obtain daily 4D (spatial 3D with time) landslide movements based on movement characteristics and dependence on hydrometeorological factors. The method's performance was evaluated using both synthetic and real datasets. On synthetic data, RMSEs in the east, north, and vertical directions were 9.6 mm, 3.8 mm, and 1.4 mm, respectively. For real data, the daily 4D movements were projected onto the Line-of-Sight (LOS) directions of Sentinel-1 Track 11 and ALOS2 PALSAR2 for comparison, achieving sub-centimeter Root Mean Square Errors (RMSEs). These results confirm the accuracy of the estimated movements and demonstrate enhanced update efficiency enabled by the Kalman Filter, which allows rapid assimilation of new data without reprocessing the full historical archive. Additionally, by incorporating geophysical and geodynamic methods, we leveraged daily 4D movements to derive various landslide parameters to analyze the kinematics of the Chuwangjing landslide during 2016–2024. The findings indicate that daily 4D movements not only enhance InSAR's performance in continuous landslide monitoring but also provide additional derivative products, deepening the understanding of deep-seated landslide kinematics.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"335 ","pages":"Article 115263"},"PeriodicalIF":11.4,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146036820","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"TIDA-SR: A time-conditioned deformable attention network for DEM super-resolution in cloud-covered mountainous regions","authors":"Kai Chen ,&nbsp;Fayuan Li ,&nbsp;Sijin Li ,&nbsp;Haoyu Cao ,&nbsp;Wen Dai ,&nbsp;Guoan Tang","doi":"10.1016/j.rse.2026.115254","DOIUrl":"10.1016/j.rse.2026.115254","url":null,"abstract":"<div><div>The complex terrain and variable climatic conditions in mountainous regions often cause cloud and fog occlusions, which hinder the generation of high-resolution Digital Elevation Models (DEMs) (below 10 m) using optical satellite photogrammetry. To address the issues of insufficient accuracy and data gaps in cloud-affected DEM areas, this study proposes a Time-Conditioned Deformable Convolution Attention Super-Resolution Network (TIDA-SR). The network performs super-resolution reconstruction using open-source DEMs and integrates multiple DEM sources through a feathered blending strategy to achieve high-accuracy results. Its architecture incorporates a diffusion process, deformable convolutions, and a Convolutional Block Attention Module (CBAM), combined with a composite loss function, to enhance the recovery of complex terrain details. TIDA-SR network, combined with the cloud-affected 5 m DEMs generated from GF-7 stereo imagery and open-source 30 m DEMs, is employed to reconstruct and fuse cloud-affected regions with 5 m resolution. The experimental results in the Loess Plateau and the Rocky Mountains demonstrate that, compared with traditional interpolation methods and existing deep learning approaches, TIDA-SR reduces RMSE and MAE by approximately 5%–78% on the validation dataset and by 3%–25% on the open-source DEM dataset. Slope accuracy improvements of approximately 3%–42% on the validation dataset and 3%–8% on the open-source DEM dataset are observed. The feathered blending strategy effectively mitigates stitching artifacts between cloud and noncloud areas, enhancing overall spatial continuity. TIDA-SR exhibits superior performance in high-resolution DEM reconstruction for cloud-affected mountainous regions and shows strong potential for practical applications, including surface process simulations, mountain hydrological modeling, geomorphological analysis, and other terrain-driven geoscience tasks.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"335 ","pages":"Article 115254"},"PeriodicalIF":11.4,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146072687","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Characterizing diurnal variability in power plant carbon emissions in Asia: A top-down estimation approach constrained by geostationary NO2 and OCO-3 CO2 observations","authors":"Tianyi Xu ,&nbsp;Chengxin Zhang ,&nbsp;Cheng Liu","doi":"10.1016/j.rse.2026.115261","DOIUrl":"10.1016/j.rse.2026.115261","url":null,"abstract":"<div><div>Accurate quantification of carbon dioxide emissions is crucial for addressing climate change. However, traditional top-down CO₂ estimates are limited by sparse satellite observations and coarse temporal resolution. Although NO₂ data from polar-orbiting satellites can help constrain CO₂ emissions, temporal mismatch with OCO-3 measurements introduce additional uncertainty. To address this, we estimate hourly CO₂ emissions by combining OCO-3 XCO₂ observations with high temporal resolution NO₂ data from the GEMS instrument onboard the GEO-KOMPSAT-2B satellite. We developed an algorithm based on a Gaussian plume model and wind rotation techniques to estimate CO₂ emissions and NO_x/CO₂ emission ratios from near-synchronous NO₂/CO₂ observations. Hourly CO₂ emissions were further derived using GEMS-based NO_x emissions estimated via the flux divergence method. A total of 59 power plant cases across six Asian countries were identified. For these cases, the estimated CO₂ emissions exhibit distinct diurnal, seasonal, and interannual emission variability, primarily driven by heating demand, decarbonization measures, and pandemic-related industrial slowdowns. These top-down estimates, constrained by GEMS NO₂ data, show strong consistency with bottom-up inventories (<em>R</em> = 0.89), supporting the validity of our optimization approach. Furthermore, comparisons with daytime mean estimates suggest that CO₂ emission estimates constrained by polar-orbiting satellite observations can exhibit biases of approximately 60% relative to GEO-based approaches, underscoring the importance for high-temporal-resolution measurements. This study highlights the value of integrating geostationary NO₂ and CO₂ observations to capture the diurnal dynamics of power plant emissions and improve the accuracy of top-down CO₂ emission monitoring.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"335 ","pages":"Article 115261"},"PeriodicalIF":11.4,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146071830","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Centimeter-resolution 4D dynamics of retrogressive thaw slumps from repeat UAV photogrammetry on the Tibetan Plateau","authors":"Siru Gao ,&nbsp;Yushuo Liu ,&nbsp;Xinyu Men ,&nbsp;Hongting Zhao ,&nbsp;Luyang Wang ,&nbsp;Ziteng Fu ,&nbsp;Zhongqiong Zhang ,&nbsp;Yuzhong Yang ,&nbsp;Guanli Jiang ,&nbsp;Qingbai Wu","doi":"10.1016/j.rse.2026.115262","DOIUrl":"10.1016/j.rse.2026.115262","url":null,"abstract":"<div><div>Retrogressive thaw slumps (RTSs) are critical indicators of permafrost degradation, with significant implications for ecosystems, infrastructure, and carbon cycling. However, their evolutionary processes remain poorly understood due to limited high-resolution observations. Here, centimeter-scale UAV surveys were conducted from 2019 to 2024 to track eight RTSs on the Tibetan Plateau, with site-specific monitoring frequencies ranging from one to six surveys per year. These RTSs displayed tongue-, trumpet-, or funnel-shaped elongated morphologies with an average aspect ratio of 3.0 and were predominantly north-facing. Their areas, volumes, and headwall heights ranged from 1688.4 to 38,991.9 m², 594.1 to 51,961.1 m³, and 0.9 to 4.6 m, respectively. Most RTSs featured a three-part structure—rear-edge slumping, mid-slope sliding, and front-edge accumulation—each with distinct deformation characteristics. Two evolutionary pathways of RTS expansion, immediate-triggered and threshold-delayed, were identified. RTS activity was observed from June to October, with the most intense activity occurred between September and late October. Annual and monthly expansion averaged 1182.9 m² yr⁻¹ and 505.7 m² mo⁻¹, while headwall retreat reached 14.5 m yr⁻¹ and 6.2 m mo⁻¹. Surface elevation exhibited rear-edge subsidence (mean 2.4 m), central uplift (mean 1.4 m), and stable front-edge, with net volume loss mainly due to subsidence (mean 81%). Surface movement decreased from the rear-edge to the front-edge, with maximum values at the rear-edge slumping zone. This study quantified the evolution of small-scale RTSs—including changes in area, headwall retreat, surface movement, elevation, and volume—particularly focusing on monthly dynamics, thereby enhancing understanding and impact assessment of RTS.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"335 ","pages":"Article 115262"},"PeriodicalIF":11.4,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146033599","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"InSAR analysis using both co- and cross-polarized data at Death Valley, California from 2017–2025","authors":"Olivia Paschall,&nbsp;Rowena Benfer Lohman","doi":"10.1016/j.rse.2026.115265","DOIUrl":"10.1016/j.rse.2026.115265","url":null,"abstract":"<div><div>The Sentinel-1 satellite mission has been key to the achievement of interferometric synthetic aperture radar (InSAR)-based displacement rates that approach mm/yr precision, particularly in regions without significant vegetation and where long time series of observations exist. However, for more subtle displacement signals, separating the effects of surface processes from deformation due to deeper sources is still challenging. Here, we present a new method based on combinations of co-polarized (VV) and cross-polarized (VH) InSAR data. Cross-polarized data is typically noisier than the co-polarized data and are not widely used for InSAR. However, comparisons of co- and cross-polarized phase data can allow separation of the contributions from different processes. Signals due to deeper sources, such as slip along faults, should appear the same in both data types, while differences can be due to changes in surface characteristics. We examine full-resolution, unfiltered, VV and VH Sentinel-1 data covering Death Valley, California between January 2017 and March 2025. We find that displacement rates derived from VV and VH data differ by several mm/yr in some areas. We also show that rates based only on the VV imagery differ by a few mm/yr between subsets of pixels where the VV-VH differences are large or small, suggesting that VV-VH combinations can help researchers reliably identify pixels that are the least impacted by surface processes. While our work focuses on Death Valley, similar mm/yr-scale biases could impact endorheic basins around the world and influence analyses of interseismic motion, hazard estimates, and groundwater studies.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"335 ","pages":"Article 115265"},"PeriodicalIF":11.4,"publicationDate":"2026-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146072689","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Why do classification models go wrong? The importance of adaptations and acclimations in driving landscape-level spectral variation in Fremont cottonwood","authors":"M.M. Seeley ,&nbsp;B.C. Wiebe ,&nbsp;G.P. Asner ,&nbsp;A.J. Abraham ,&nbsp;H.F. Cooper ,&nbsp;C.A. Gehring ,&nbsp;K.R. Hultine ,&nbsp;G.J. Allan ,&nbsp;T.G. Whitham ,&nbsp;T. Goulden ,&nbsp;C.E. Doughty","doi":"10.1016/j.rse.2026.115240","DOIUrl":"10.1016/j.rse.2026.115240","url":null,"abstract":"<div><div>Spatially explicit predictions of species distributions can inform ecosystem processes and conservation, particularly under global change. While imaging spectroscopy could enable accurate species classifications, accuracy generally declines outside training regions, limiting its utility for regional-scale mapping. To investigate mechanisms constraining classification generalizability (e.g., spatial autocorrelation, local adaptation), we used National Ecological Observatory Network Airborne Observation Platform imaging spectroscopy data collected across riparian systems in Arizona, Colorado, and Utah. We extracted canopy spectral data of <em>Populus fremontii</em> (Fremont cottonwood), a foundation riparian tree known to form locally adapted ecotypes across its range, and spatially co-occurring species from seventeen 6 × 6 km sites. Combining this library with site-level environmental data, and support vector machine (SVM) models, we observed that environmental, not geographic, distance between training and test sites limited classification generalizability. Specifically, differences in mean annual temperature, winter precipitation, and spring precipitation, key drivers of local adaptation of <em>P. fremontii</em>, were associated with lower classification accuracy (∼50% lower). We then evaluated specific wavelength regions for improved generalizability. Classification models using only near-infrared (750–1400 nm) and shortwave infrared (1400–2500 nm) outperformed those using full-spectrum models in regions not represented in the training data, consistent with lower heritability in visible and red-edge wavelengths. In conclusion, spatially structured spectral phenotypes of <em>P. fremontii</em>, shaped by local adaptation and acclimation to environmental conditions, reduced species classification generalizability. By integrating ecology into remote sensing workflows, such as spectral band selection, we can improve species classification accuracy, thereby advancing scalable biodiversity monitoring and conservation efforts.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"334 ","pages":"Article 115240"},"PeriodicalIF":11.4,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145962390","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"RPI-GMM: A novel structure-based and phenology-independent algorithm for mapping latest 10-m resolution national-level rubber plantations","authors":"Chiwei Xiao ,&nbsp;Zilong Yue ,&nbsp;Zhiming Feng ,&nbsp;Jinwei Dong ,&nbsp;Juliet Lu ,&nbsp;Khin Htet Htet Pyone ,&nbsp;Khampheng Boudmyxay","doi":"10.1016/j.rse.2026.115241","DOIUrl":"10.1016/j.rse.2026.115241","url":null,"abstract":"<div><div>Accurate and updated maps of rubber plantations are beneficial to eco-environmental and socio-economic impact assessment and sustainable agroforestry management. However, existing remotely-sensed approaches to identifying rubber plantations primarily rely on phenological signals from time-series optical data, which are limited by persistent cloud cover, regional phenological variability or inconsistency, and high data demands. To address these challenges, here, we propose an innovative phenology-independent framework that integrates a rubber plantation index (RPI) with an unsupervised Gaussian Mixture Model (GMM) classifier. The RPI is a structure sensitive index derived from dual-polarized Sentinel-1 SAR backscatter (VV/VH) and Sentinel-2 SWIR reflectance (Band 11), capturing plantation regularity and canopy moisture characteristics. We evaluated the RPI-GMM framework across six diverse sample areas of rubber plots in tropics representing variations in phenology, topography, and plantation structure. Results demonstrated high classification accuracy, with F1 scores over 0.87 under both phenologically strong and weak conditions, as well as across mountainous and fragmented landscapes. Our RPI-GMM method achieved an overall accuracy of 87.0% in Laos, and estimated 234,206 ha of rubber plots in 2024. Spatial analysis revealed that approximately 70% of rubber plantations are located in Laotian border areas near China and Vietnam, 90% are situated at elevations below 1000 m, and 80% are found on slopes with gradients ranging from 3^° to 16^°. Notably, our simple and integrated method of RPI-GMM requires no temporal or labeled data, ensuring robustness, cost-efficiency, and transferability. The results highlight valuable insights of structure-based SAR-optical fusion for future global or tropical monitoring of tree-plantation dynamics and support broader applications in agroforestry management.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"334 ","pages":"Article 115241"},"PeriodicalIF":11.4,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145962389","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-sensor (since 1997) global soil moisture mapping with enhanced Spatio-temporal coverage through machine learning framework fusion","authors":"Haojie Zhang ,&nbsp;Tianjie Zhao ,&nbsp;Zhiqing Peng ,&nbsp;Jingyao Zheng ,&nbsp;Yu Bai ,&nbsp;Nemesio Rodriguez-Fernadez ,&nbsp;Donghai Zheng ,&nbsp;Huazhu Xue ,&nbsp;Zhanliang Yuan ,&nbsp;Qian Cui ,&nbsp;Peng Guo ,&nbsp;Zushuai Wei ,&nbsp;Peilin Song ,&nbsp;Lixin Dong ,&nbsp;Panpan Yao ,&nbsp;Qiangqiang Yuan ,&nbsp;Lingkui Meng ,&nbsp;Jiancheng Shi","doi":"10.1016/j.rse.2025.115221","DOIUrl":"10.1016/j.rse.2025.115221","url":null,"abstract":"<div><div>The successful deployment of multiple satellites equipped with passive microwave sensors has been pivotal for monitoring global soil moisture. Despite their importance, limitations including varying service durations, orbital scanning gaps, and differences in retrieval algorithms result in poor spatio-temporal consistency and coverage. This study introduces a two-stage paradigm to overcome the inconsistency of multi-sensors: Firstly, high-precision soil moisture is generated from SMAP L-band observations through the multi-channel collaborative algorithm (MCCA) as the physically anchored training target. Then, a long short-term memory (LSTM) network specifically designed for global gridded soil moisture dynamics is trained based on cross-calibrated brightness temperature observations (inclined orbit satellite sensors (TMI and GMI) and polar orbit satellite sensors (AMSR-E and AMSR2)) to obtain the high-quality retrieval accuracy of MCCA SMAP. Finally, the daily global soil moisture product (25 km resolution, 1997–2023) is provided by fusing the instantaneous soil moisture data of the four sensors from the model output. The study performed extensive validation with ground measurements and cross-validation with other datasets for both temporal and spatial consistency. The results indicate that the spatial distribution and seasonal variation patterns of MCCA-ML closely match those of MCCA SMAP, reflecting global climatic and geographic features. Verified by 24 dense global observation networks, the global r value of MCCA-ML SM is 0.76, the RMSE is 0.068 m³/m³, and the ubRMSE is 0.059 m³/m³, which well inherits the excellent performance of SMAP. During the service period of two or more satellites, the daily global land coverage of MCCA-ML SM usually exceeds 80 %, and it has a good ability to detect soil moisture.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"334 ","pages":"Article 115221"},"PeriodicalIF":11.4,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145939013","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}