Pub Date : 2026-01-10DOI: 10.1016/j.rse.2026.115237
Hao Xu , Nan Xu , Wenyu Li , Kai Tan , Chunpeng Chen , Huan Li , Lucheng Zhan , Pei Xin , Jiaqi Yao , Peng Li , Zhen Zhang , Haipeng Zhao , Bolin Fu , Yifei Zhao , Yufeng Li , Qi Wang , Fan Zhao , Xiaojuan Liu , Zhongwen Hu , Guofeng Wu , Qingquan Li
Tidal flat topography is a fundamental attribute affecting inundation dynamics, sediment transport, and ecosystem functioning, yet accurate and spatially consistent large-scale monitoring remains challenging. Here, we leveraged satellite altimetry from the Surface Water and Ocean Topography (SWOT) mission to develop a novel, large-scale framework for deriving tidal flat topography from SWOT data, and demonstrated its capability by generating a high-accuracy, national-scale elevation dataset for China. By combining a percentile-based aggregation of multi-temporal water-surface elevation observations with a tide-constrained, adaptive best-quantile (best-q) reconstruction strategy, followed by linear interpolation for gap filling, we improved both vertical accuracy and spatial completeness. Validation against airborne LiDAR, GNSS-RTK surveys, and ICESat-2 photon data demonstrates robust performance across diverse coastal settings, achieving RMSE = 0.34–0.47 m and R2 = 0.81–0.88 at a horizontal resolution of 100 m. Compared with existing large-scale digital elevation models (DEMs), the SWOT-derived topography not only improves vertical accuracy by over 80% but also providing substantially more complete spatial coverage of tidal flat elevations. Spatial analyses reveal pronounced latitudinal gradients, with higher tidal flats concentrated in low-latitude regions and extensive low-lying flats dominating northern estuarine and deltaic systems. This study establishes a scalable framework for tidal-flat elevation retrieval and provides a foundational dataset to support coastal monitoring and sustainable management.
潮滩地形是影响淹没动态、泥沙输运和生态系统功能的基本属性，但精确和空间一致的大规模监测仍然具有挑战性。在这里，我们利用来自地表水和海洋地形（SWOT）任务的卫星测高数据，开发了一个新的、大规模的框架，用于从SWOT数据中获取潮滩地形，并通过生成中国高精度的国家尺度高程数据集来证明其能力。通过将基于百分位的多时间点水面高程观测集合与潮汐约束的自适应最佳分位数（best-q）重建策略相结合，然后采用线性插值进行间隙填充，我们提高了垂直精度和空间完整性。针对机载LiDAR、GNSS-RTK调查和ICESat-2光子数据的验证表明，在不同的沿海环境下，该方法具有强大的性能，在100米的水平分辨率下，RMSE = 0.34-0.47 m, R2 = 0.81-0.88。与现有的大尺度数字高程模型（dem）相比，swot衍生的地形不仅垂直精度提高了80%以上，而且提供了更完整的潮滩高程空间覆盖。空间分析显示了明显的纬度梯度，高纬度潮滩集中在低纬度地区，北部河口和三角洲系统主要是广泛的低洼滩。本研究建立了一个可扩展的潮坪高程检索框架，为支持沿海监测和可持续管理提供了基础数据集。
{"title":"A large-scale framework for deriving tidal flat topography from SWOT data","authors":"Hao Xu , Nan Xu , Wenyu Li , Kai Tan , Chunpeng Chen , Huan Li , Lucheng Zhan , Pei Xin , Jiaqi Yao , Peng Li , Zhen Zhang , Haipeng Zhao , Bolin Fu , Yifei Zhao , Yufeng Li , Qi Wang , Fan Zhao , Xiaojuan Liu , Zhongwen Hu , Guofeng Wu , Qingquan Li","doi":"10.1016/j.rse.2026.115237","DOIUrl":"10.1016/j.rse.2026.115237","url":null,"abstract":"<div><div>Tidal flat topography is a fundamental attribute affecting inundation dynamics, sediment transport, and ecosystem functioning, yet accurate and spatially consistent large-scale monitoring remains challenging. Here, we leveraged satellite altimetry from the Surface Water and Ocean Topography (SWOT) mission to develop a novel, large-scale framework for deriving tidal flat topography from SWOT data, and demonstrated its capability by generating a high-accuracy, national-scale elevation dataset for China. By combining a percentile-based aggregation of multi-temporal water-surface elevation observations with a tide-constrained, adaptive best-quantile (best-q) reconstruction strategy, followed by linear interpolation for gap filling, we improved both vertical accuracy and spatial completeness. Validation against airborne LiDAR, GNSS-RTK surveys, and ICESat-2 photon data demonstrates robust performance across diverse coastal settings, achieving RMSE = 0.34–0.47 m and R<sup>2</sup> = 0.81–0.88 at a horizontal resolution of 100 m. Compared with existing large-scale digital elevation models (DEMs), the SWOT-derived topography not only improves vertical accuracy by over 80% but also providing substantially more complete spatial coverage of tidal flat elevations. Spatial analyses reveal pronounced latitudinal gradients, with higher tidal flats concentrated in low-latitude regions and extensive low-lying flats dominating northern estuarine and deltaic systems. This study establishes a scalable framework for tidal-flat elevation retrieval and provides a foundational dataset to support coastal monitoring and sustainable management.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"334 ","pages":"Article 115237"},"PeriodicalIF":11.4,"publicationDate":"2026-01-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145947722","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}
Pub Date : 2026-01-09DOI: 10.1016/j.rse.2025.115206
Shulin Pang , Zhanqing Li , Lin Sun , Biao Cao , Zhihui Wang , Xinyuan Xi , Xiaohang Shi , Jing Xu , Jing Wei
Cloud detection is crucial in many applications of satellite remote sensing data. Traditional cloud detection methods typically operate at the pixel level, relying on empirically tuned thresholds or, more recently, machine learning classification schemes based on training datasets. Motivated by the success of the Transformer with its self-attention mechanism and convolutional neural networks for enhanced feature extraction, we propose a new encoder-decoder method that captures global and regional contexts with multi-scale features. This new model takes advantage of two advanced deep-learning techniques, the Swin Transformer and UPerNet (named STUPmask), demonstrating improved cloud detection accuracy and strong adaptability to diverse imagery types, spanning spectral bands from visible to thermal infrared and spatial resolutions from meters to kilometers, across a wide range of surface types, including bright scenes such as ice and desert, globally. Training and validation of the STUPmask model are conducted using data obtained from the Landsat 8 and Sentinel-2 Manually Cloud Validation Mask datasets on a global scale. STUPmask accurately estimates cloud amount with a marginal difference against reference masks (0.27 % for Landsat 8 and −0.81 % for Sentinel-2). Additionally, the model captures cloud distribution with a high overall classification accuracy (97.51 % for Landsat 8 and 96.27 % for Sentinel-2). Notably, it excels in detecting broken, thin, and semi-transparent clouds across diverse surfaces, including bright surfaces like urban and barren lands, especially with acceptable accuracy over snow and ice. These encompass the majority of challenging scenes encountered by cloud identification methods. It also adapts to cross-sensor satellite data with varying spatial resolutions (4 m–2 km) from both Low-Earth-Orbit (LEO) and Geostationary-Earth-Orbit (GEO) platforms (including GaoFen-2, MODIS, and Himawari-8), with an overall accuracy of 94.21–97.11 %. The demonstrated successes in the automatic identification of clouds with a variety of satellite imagery of different spectral channels and spatial resolutions render the method versatile for a wide range of remote sensing studies.
{"title":"Enhancing cloud detection across multiple satellite sensors using a combined Swin Transformer and UPerNet deep learning model","authors":"Shulin Pang , Zhanqing Li , Lin Sun , Biao Cao , Zhihui Wang , Xinyuan Xi , Xiaohang Shi , Jing Xu , Jing Wei","doi":"10.1016/j.rse.2025.115206","DOIUrl":"10.1016/j.rse.2025.115206","url":null,"abstract":"<div><div>Cloud detection is crucial in many applications of satellite remote sensing data. Traditional cloud detection methods typically operate at the pixel level, relying on empirically tuned thresholds or, more recently, machine learning classification schemes based on training datasets. Motivated by the success of the Transformer with its self-attention mechanism and convolutional neural networks for enhanced feature extraction, we propose a new encoder-decoder method that captures global and regional contexts with multi-scale features. This new model takes advantage of two advanced deep-learning techniques, the Swin Transformer and UPerNet (named STUPmask), demonstrating improved cloud detection accuracy and strong adaptability to diverse imagery types, spanning spectral bands from visible to thermal infrared and spatial resolutions from meters to kilometers, across a wide range of surface types, including bright scenes such as ice and desert, globally. Training and validation of the STUPmask model are conducted using data obtained from the Landsat 8 and Sentinel-2 Manually Cloud Validation Mask datasets on a global scale. STUPmask accurately estimates cloud amount with a marginal difference against reference masks (0.27 % for Landsat 8 and −0.81 % for Sentinel-2). Additionally, the model captures cloud distribution with a high overall classification accuracy (97.51 % for Landsat 8 and 96.27 % for Sentinel-2). Notably, it excels in detecting broken, thin, and semi-transparent clouds across diverse surfaces, including bright surfaces like urban and barren lands, especially with acceptable accuracy over snow and ice. These encompass the majority of challenging scenes encountered by cloud identification methods. It also adapts to cross-sensor satellite data with varying spatial resolutions (4 m–2 km) from both Low-Earth-Orbit (LEO) and Geostationary-Earth-Orbit (GEO) platforms (including GaoFen-2, MODIS, and Himawari-8), with an overall accuracy of 94.21–97.11 %. The demonstrated successes in the automatic identification of clouds with a variety of satellite imagery of different spectral channels and spatial resolutions render the method versatile for a wide range of remote sensing studies.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"334 ","pages":"Article 115206"},"PeriodicalIF":11.4,"publicationDate":"2026-01-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145938803","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}
Pub Date : 2026-01-09DOI: 10.1016/j.rse.2026.115234
Wenyuan Li , Shunlin Liang , Keyan Chen , Yongzhe Chen , Han Ma , Jianglei Xu , Yichuan Ma , Yuxiang Zhang , Shikang Guan , Husheng Fang , Zhenwei Shi
Climate change and population growth intensify the demand for precise agriculture mapping to enhance food security. Such mapping tasks require robust modeling of multi-scale spatiotemporal patterns from fine field textures to landscape context, and from short-term phenology to full growing-season dynamics. Existing methods often process spatial and temporal features separately, limiting their ability to capture essential agricultural dynamics. While transformer-based remote sensing foundation models (RSFMs) offer unified spatiotemporal modeling ability, most of them remain suboptimal: they either use fixed windows that ignore multi-scale crop characteristics or neglect temporal information entirely. To address these gaps, we propose AgriFM, a multi-source, multi-temporal foundation model for agriculture mapping. AgriFM introduces a synchronized spatiotemporal downsampling strategy within a Video Swin Transformer backbone, enabling efficient handling of long and variable-length satellite time series while preserving multi-scale spatial and phenological information. It is pre-trained on a globally representative dataset comprising over 25 million samples from MODIS, Landsat-8/9, and Sentinel-2 with land cover fractions as pre-training supervision. AgriFM further integrates a versatile decoder specifically designed to dynamically fuse multi-source features from different stages of backbone and accommodate varying temporal lengths, thereby supporting consistent and scalable agriculture mapping across diverse satellite sources and task requirements. It supports diverse tasks including agricultural land mapping, field boundary delineation, agricultural land use/land cover mapping, and specific crop mapping (e.g., winter wheat and paddy rice) with different data sources. Comprehensive evaluations show that AgriFM consistently outperforms existing deep learning models and general-purpose RSFMs across multiple agriculture mapping tasks. Codes and models are available at https://github.com/flyakon/AgriFM and https://glass.hku.hk
{"title":"AgriFM: A multi-source temporal remote sensing foundation model for Agriculture mapping","authors":"Wenyuan Li , Shunlin Liang , Keyan Chen , Yongzhe Chen , Han Ma , Jianglei Xu , Yichuan Ma , Yuxiang Zhang , Shikang Guan , Husheng Fang , Zhenwei Shi","doi":"10.1016/j.rse.2026.115234","DOIUrl":"10.1016/j.rse.2026.115234","url":null,"abstract":"<div><div>Climate change and population growth intensify the demand for precise agriculture mapping to enhance food security. Such mapping tasks require robust modeling of multi-scale spatiotemporal patterns from fine field textures to landscape context, and from short-term phenology to full growing-season dynamics. Existing methods often process spatial and temporal features separately, limiting their ability to capture essential agricultural dynamics. While transformer-based remote sensing foundation models (RSFMs) offer unified spatiotemporal modeling ability, most of them remain suboptimal: they either use fixed windows that ignore multi-scale crop characteristics or neglect temporal information entirely. To address these gaps, we propose AgriFM, a multi-source, multi-temporal foundation model for agriculture mapping. AgriFM introduces a synchronized spatiotemporal downsampling strategy within a Video Swin Transformer backbone, enabling efficient handling of long and variable-length satellite time series while preserving multi-scale spatial and phenological information. It is pre-trained on a globally representative dataset comprising over 25 million samples from MODIS, Landsat-8/9, and Sentinel-2 with land cover fractions as pre-training supervision. AgriFM further integrates a versatile decoder specifically designed to dynamically fuse multi-source features from different stages of backbone and accommodate varying temporal lengths, thereby supporting consistent and scalable agriculture mapping across diverse satellite sources and task requirements. It supports diverse tasks including agricultural land mapping, field boundary delineation, agricultural land use/land cover mapping, and specific crop mapping (e.g., winter wheat and paddy rice) with different data sources. Comprehensive evaluations show that AgriFM consistently outperforms existing deep learning models and general-purpose RSFMs across multiple agriculture mapping tasks. Codes and models are available at <span><span>https://github.com/flyakon/AgriFM</span><svg><path></path></svg></span> and <span><span>https://glass.hku.hk</span><svg><path></path></svg></span></div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"334 ","pages":"Article 115234"},"PeriodicalIF":11.4,"publicationDate":"2026-01-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145938802","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}
Pub Date : 2026-01-09DOI: 10.1016/j.rse.2025.115197
Baptiste Vandecrux , Ghislain Picard , Pierre Zeiger , Marion Leduc-Leballeur , Andreas Colliander , Alamgir Hossan , Andreas Ahlstrøm
As the Arctic warms, surface melt extends into the Greenland Ice Sheet's accumulation zone, where much of the water infiltrates into the snowpack. This makes monitoring the subsurface water depth and spatial extent important for accurate ice sheet runoff estimations. Subsurface water can be detected using remotely sensed microwave brightness temperatures (TB). We use vertically polarized TB at 1.4 GHz from Soil Moisture and Ocean Salinity satellite (SMOS) and at 6.9, 10.7, and 18.7 GHz from the Advanced Microwave Scanning Radiometers (AMSR-E/2) to estimate the upper depth of liquid water (UDLW) on the ice sheet accumulation area. We build a catalogue of simulated UDLW and TB: realistic UDLW are modeled by the Geological Survey of Denmark and Greenland (GEUS) snow model, forced by the Copernicus Arctic Regional Reanalysis (CARRA), and the corresponding TB are calculated by the Snow Microwave Radiative Transfer (SMRT) model at 19 sites. We train on this catalogue an ensemble of cross-validated Random Forest (RF) models to predict UDLW and its uncertainty from TB at four frequencies. On hold-out modeled data and for water within 5 m of the surface, the RF ensemble achieves a median RMSE of 0.68 m and mean error of −0.09 m. Our retrieval, when applied to observed TB, matches within 2 m UDLW inferred from subsurface temperature profiles down to 4–6 m depth. Performances decrease beyond 5 m depth and for low liquid water amounts. Our retrieval produces daily UDLW maps over the ice sheet's accumulation area during 2010–2023 which reveal the seasonal evolution of UDLW, deliver the first quantitative estimates of subsurface liquid water depth on the ice sheet and offer new insights into meltwater infiltration and storage processes.
{"title":"Estimating the upper depth of subsurface water on the Greenland Ice Sheet using multi-frequency passive microwave remote sensing, radiative transfer modeling, and machine learning","authors":"Baptiste Vandecrux , Ghislain Picard , Pierre Zeiger , Marion Leduc-Leballeur , Andreas Colliander , Alamgir Hossan , Andreas Ahlstrøm","doi":"10.1016/j.rse.2025.115197","DOIUrl":"10.1016/j.rse.2025.115197","url":null,"abstract":"<div><div>As the Arctic warms, surface melt extends into the Greenland Ice Sheet's accumulation zone, where much of the water infiltrates into the snowpack. This makes monitoring the subsurface water depth and spatial extent important for accurate ice sheet runoff estimations. Subsurface water can be detected using remotely sensed microwave brightness temperatures (T<sub>B</sub>). We use vertically polarized T<sub>B</sub> at 1.4 GHz from Soil Moisture and Ocean Salinity satellite (SMOS) and at 6.9, 10.7, and 18.7 GHz from the Advanced Microwave Scanning Radiometers (AMSR-E/2) to estimate the upper depth of liquid water (UDLW) on the ice sheet accumulation area. We build a catalogue of simulated UDLW and T<sub>B</sub>: realistic UDLW are modeled by the Geological Survey of Denmark and Greenland (GEUS) snow model, forced by the Copernicus Arctic Regional Reanalysis (CARRA), and the corresponding T<sub>B</sub> are calculated by the Snow Microwave Radiative Transfer (SMRT) model at 19 sites. We train on this catalogue an ensemble of cross-validated Random Forest (RF) models to predict UDLW and its uncertainty from T<sub>B</sub> at four frequencies. On hold-out modeled data and for water within 5 m of the surface, the RF ensemble achieves a median RMSE of 0.68 m and mean error of −0.09 m. Our retrieval, when applied to observed T<sub>B</sub>, matches within 2 m UDLW inferred from subsurface temperature profiles down to 4–6 m depth. Performances decrease beyond 5 m depth and for low liquid water amounts. Our retrieval produces daily UDLW maps over the ice sheet's accumulation area during 2010–2023 which reveal the seasonal evolution of UDLW, deliver the first quantitative estimates of subsurface liquid water depth on the ice sheet and offer new insights into meltwater infiltration and storage processes.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"334 ","pages":"Article 115197"},"PeriodicalIF":11.4,"publicationDate":"2026-01-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145938861","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}
Pub Date : 2026-01-08DOI: 10.1016/j.rse.2025.115220
Guoqing Zhang, Hongjie Xie, Alfonso Fernandez, Christophe Kinnard, Stef Lhermitte
Driven by rapid technological advances in cryospheric science and the emergence of new generations of remote sensing observations, this special issue of Remote Sensing of Environment, entitled “Remote sensing of the global cryosphere: status, processes, and trends”, brings together 23 studies published between 2023 and 2025. Collectively, these papers showcase how multi-sensor satellite observations, high-resolution digital elevation models (DEMs), and cutting-edge deep learning techniques are revolutionizing the monitoring of glaciers, snow, glacial lakes, permafrost, sea ice, and ice shelves across the Earth's three poles: the Arctic (including Greenland), Antarctica, and High Mountain Asia (the Third Pole). These studies integrate diverse datasets – including multisource DEMs, optical, thermal, and passive microwave imageries, as well as RADAR, LiDAR, and GRACE observations - to quantify glacier mass balance, map glacial lakes, assess permafrost thermal conditions, classify sea-ice types, and detect icebergs. We organize the publications by major cryospheric themes and their distribution across polar regions and summarize the dominant remote sensing datasets and methodologies employed. Finally, we outline future directions, emphasizing multi-sensor data fusion, physics-informed modeling, and AI-driven approaches to improve predictions of cryospheric change under a warming climate.
{"title":"Remote sensing of the global cryosphere: Status, processes, and trends","authors":"Guoqing Zhang, Hongjie Xie, Alfonso Fernandez, Christophe Kinnard, Stef Lhermitte","doi":"10.1016/j.rse.2025.115220","DOIUrl":"https://doi.org/10.1016/j.rse.2025.115220","url":null,"abstract":"Driven by rapid technological advances in cryospheric science and the emergence of new generations of remote sensing observations, this special issue of <em>Remote Sensing of Environment</em>, entitled “Remote sensing of the global cryosphere: status, processes, and trends”, brings together 23 studies published between 2023 and 2025. Collectively, these papers showcase how multi-sensor satellite observations, high-resolution digital elevation models (DEMs), and cutting-edge deep learning techniques are revolutionizing the monitoring of glaciers, snow, glacial lakes, permafrost, sea ice, and ice shelves across the Earth's three poles: the Arctic (including Greenland), Antarctica, and High Mountain Asia (the Third Pole). These studies integrate diverse datasets – including multisource DEMs, optical, thermal, and passive microwave imageries, as well as RADAR, LiDAR, and GRACE observations - to quantify glacier mass balance, map glacial lakes, assess permafrost thermal conditions, classify sea-ice types, and detect icebergs. We organize the publications by major cryospheric themes and their distribution across polar regions and summarize the dominant remote sensing datasets and methodologies employed. Finally, we outline future directions, emphasizing multi-sensor data fusion, physics-informed modeling, and AI-driven approaches to improve predictions of cryospheric change under a warming climate.","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"254 1","pages":""},"PeriodicalIF":13.5,"publicationDate":"2026-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145947720","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}
Pub Date : 2026-01-08DOI: 10.1016/j.rse.2025.115213
João M.B. Carreiras , Thomas Higginbottom , John L. Godlee , Sam Harrison , Lorena Benitez , Penelope J. Mograbi , Aurora Levesley , Karina Melgaço , David Milodowski , Georgia Pickavance , Geoff Wells , Edmar Almeida de Oliveira , Luzmila Arroyo , Sam Bowers , Roel J.W. Brienen , Domingos Cardoso , António Alberto Jorge Farias Castro , Ezequiel Chavez , Ítalo A.C. Coutinho , Tomás F. Domingues , Casey M. Ryan
<div><div>Accurate characterization of the role of the dry tropics in the global carbon cycle requires precise estimation of woody biomass changes due to ecological and anthropogenic change, including deforestation, forest degradation, regrowth, mortality and enhanced tree growth due to climate change. L-band Synthetic Aperture Radar (SAR) backscatter observations offer a reliable option to consistently map these processes as they are (i) available globally since 2007 (JAXA ALOS-1, ALOS-2 and ALOS-4), and (ii) sensitive to woody structure, such as aboveground biomass density (<span><math><mi>AGBD</mi></math></span>) up to ∼100 t ha<sup>−1</sup>. However, we lack multi-site empirical understanding of the scattering processes that determine the relationship between L-band SAR and woody vegetation structure in the dry tropics, and how this is mediated by soil properties.</div><div>This study used observations from ground plots in Africa (<em>n</em> = 171), Australia (<em>n</em> = 6), and South America (<em>n</em> = 44) to understand the impact of vegetation structure and soil properties on spatially and temporally coincident fully-polarimetric L-band SAR data. Fully-polarimetric L-band SAR single-look complex data were converted to scattering mechanisms/parameters using van Zyl, Cloude-Pottier, and Freeman-Durden polarimetric decompositions to elucidate the physical mechanisms involved. Multivariate SAR-vegetation-soil relationships were analysed using a theory-informed structural equation modelling approach. The strongest positive effects on volume scattering come from stem density (stems ha<sup>−1</sup>) and mean stem biomass of trees, and soil water and sand content (standardized regression coefficients of 0.3, 0.1, 0.2 and 0.1, respectively). The only significant effect on surface scattering is from stem density (0.1). Significant effects on double bounce scattering are from stem density (0.3) and soil sand content (−0.2). Since <span><math><mi>AGBD</mi></math></span> is the product of stem density and mean stem biomass, this modelling framework points to a stronger effect from the number of trees rather than their size/biomass. Therefore, <span><math><mi>AGBD</mi></math></span> maps relying solely on radar intensity may not reflect significant changes when <span><math><mi>AGBD</mi></math></span> is increasing due to the growth of existing stems. Additionally, such maps might overestimate changes in <span><math><mi>AGBD</mi></math></span> when driven by the recruitment of new stems or loss of existing stems. Full-polarimetric observations allow the decomposition of the radar signal into volume scattering, surface scattering, and double bounce, enabling the inversion of structural equation models to retrieve both stem density and mean stem biomass. This provides a more comprehensive description of forest structure compared to retrieving only <span><math><mi>AGBD</mi></math></span>. As this approach depends on full-polarimetric data, its effective
要准确描述干燥热带地区在全球碳循环中的作用，就需要精确估计由于生态和人为变化造成的木质生物量变化，包括森林砍伐、森林退化、再生、死亡和气候变化导致的树木生长增强。l波段合成孔径雷达（SAR）后向散射观测提供了一种可靠的选择，可以一致地绘制这些过程，因为它们(i)自2007年以来在全球范围内可用（JAXA ALOS-1， ALOS-2和ALOS-4），并且（ii）对木质结构敏感，例如地上生物量密度（AGBD）高达~ 100 t ha -1。然而，我们缺乏对l波段SAR与干旱热带木本植被结构之间关系的散射过程的多站点经验理解，以及土壤性质如何介导这种关系。本研究利用非洲（n = 171）、澳大利亚（n = 6）和南美洲（n = 44）的地面样地观测资料，了解植被结构和土壤性质对时空重合全极化l波段SAR数据的影响。利用van Zyl、cloud - pottier和Freeman-Durden极化分解方法，将全极化l波段SAR单目复杂数据转换为散射机制/参数，以阐明所涉及的物理机制。利用结构方程建模方法分析了多变量sar -植被-土壤关系。对体积散射的正向影响最大的是树木的茎密度（茎ha−1）和平均茎生物量，以及土壤含水量和含沙量（标准化回归系数分别为0.3、0.1、0.2和0.1）。唯一对表面散射有显著影响的是茎密度（0.1）。茎密度（0.3）和土壤含沙量（−0.2）对双弹跳散射有显著影响。由于AGBD是茎密度和平均茎生物量的产物，该模型框架指出，树木数量的影响比它们的大小/生物量更强。因此，单纯依靠雷达强度的AGBD地图可能无法反映出由于现有系统的生长而增加的AGBD的显著变化。此外，这样的图谱可能会高估AGBD的变化，因为它是由新茎的吸收或现有茎的丧失所驱动的。全极化观测允许将雷达信号分解为体散射、表面散射和双反弹，从而实现结构方程模型的反演，从而获得茎密度和平均茎生物量。与仅检索AGBD相比，这提供了更全面的森林结构描述。由于这种方法依赖于全极化数据，其有效性与这种观测的可用性密切相关。我们的研究结果强调了ALOS-4、PALSAR-3、BIOMASS和ROSE-L等近期和即将开展的任务的价值，并强调了优先获取四极SAR数据以支持未来大规模植被结构属性检索的必要性。
{"title":"Determinants of L-band backscatter in dry tropical ecosystems: Implications for biomass mapping","authors":"João M.B. Carreiras , Thomas Higginbottom , John L. Godlee , Sam Harrison , Lorena Benitez , Penelope J. Mograbi , Aurora Levesley , Karina Melgaço , David Milodowski , Georgia Pickavance , Geoff Wells , Edmar Almeida de Oliveira , Luzmila Arroyo , Sam Bowers , Roel J.W. Brienen , Domingos Cardoso , António Alberto Jorge Farias Castro , Ezequiel Chavez , Ítalo A.C. Coutinho , Tomás F. Domingues , Casey M. Ryan","doi":"10.1016/j.rse.2025.115213","DOIUrl":"10.1016/j.rse.2025.115213","url":null,"abstract":"<div><div>Accurate characterization of the role of the dry tropics in the global carbon cycle requires precise estimation of woody biomass changes due to ecological and anthropogenic change, including deforestation, forest degradation, regrowth, mortality and enhanced tree growth due to climate change. L-band Synthetic Aperture Radar (SAR) backscatter observations offer a reliable option to consistently map these processes as they are (i) available globally since 2007 (JAXA ALOS-1, ALOS-2 and ALOS-4), and (ii) sensitive to woody structure, such as aboveground biomass density (<span><math><mi>AGBD</mi></math></span>) up to ∼100 t ha<sup>−1</sup>. However, we lack multi-site empirical understanding of the scattering processes that determine the relationship between L-band SAR and woody vegetation structure in the dry tropics, and how this is mediated by soil properties.</div><div>This study used observations from ground plots in Africa (<em>n</em> = 171), Australia (<em>n</em> = 6), and South America (<em>n</em> = 44) to understand the impact of vegetation structure and soil properties on spatially and temporally coincident fully-polarimetric L-band SAR data. Fully-polarimetric L-band SAR single-look complex data were converted to scattering mechanisms/parameters using van Zyl, Cloude-Pottier, and Freeman-Durden polarimetric decompositions to elucidate the physical mechanisms involved. Multivariate SAR-vegetation-soil relationships were analysed using a theory-informed structural equation modelling approach. The strongest positive effects on volume scattering come from stem density (stems ha<sup>−1</sup>) and mean stem biomass of trees, and soil water and sand content (standardized regression coefficients of 0.3, 0.1, 0.2 and 0.1, respectively). The only significant effect on surface scattering is from stem density (0.1). Significant effects on double bounce scattering are from stem density (0.3) and soil sand content (−0.2). Since <span><math><mi>AGBD</mi></math></span> is the product of stem density and mean stem biomass, this modelling framework points to a stronger effect from the number of trees rather than their size/biomass. Therefore, <span><math><mi>AGBD</mi></math></span> maps relying solely on radar intensity may not reflect significant changes when <span><math><mi>AGBD</mi></math></span> is increasing due to the growth of existing stems. Additionally, such maps might overestimate changes in <span><math><mi>AGBD</mi></math></span> when driven by the recruitment of new stems or loss of existing stems. Full-polarimetric observations allow the decomposition of the radar signal into volume scattering, surface scattering, and double bounce, enabling the inversion of structural equation models to retrieve both stem density and mean stem biomass. This provides a more comprehensive description of forest structure compared to retrieving only <span><math><mi>AGBD</mi></math></span>. As this approach depends on full-polarimetric data, its effective","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"334 ","pages":"Article 115213"},"PeriodicalIF":11.4,"publicationDate":"2026-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145939014","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}
Pub Date : 2026-01-08DOI: 10.1016/j.rse.2026.115233
Rui Chen , Yiru Zhang , Yanxi Li , Marta Yebra , Chunquan Fan , Hongguo Zhang , Binbin He
High-intensity forest fires have significant destructive impacts on ecosystems and society, and are an increasing concern worldwide. Accurate probabilistic risk assessment of these fires can effectively enhance the ability to guide wildfire management, particularly for large and extreme fires. However, forecasting large-scale fire behavior characteristics remains challenging, limiting the effectiveness of spatial estimations of high-intensity forest fire potential (HIFFP). This study aims to integrate fire spread simulations and machine learning (ML) algorithms to enhance HIFFP estimations through multi-step time-series forecasting on fire rate of spread and fireline intensity at regional scales. We first established a high-intensity forest fire dataset based on remote sensing-informed fire spread simulations from the Weather Research and Forecasting coupled fire-spread model (WRF-SFIRE), incorporating explanatory variables on fuel, weather, climate, and topography. Then, the knowledge-guided framework (multi-step time series-based ML, MTS-ML) was designed to estimate HIFFP within different hours after fires occur, integrating with Bayesian Network (BN), Random Forest (RF), and copula models. Results indicate that MTS-ML improved HIFFP modeling compared with ML-based methods, achieving AUC (the area under the receiver operating characteristic curve) > 0.95 (with ∼0.04 increments), F1 score > 0.85 (with ∼0.08 increments), and MAE < 0.15. Topographic index, foliage fuel load, and wind speed are identified as primary contributors to HIFFP. Probabilistic mapping of HIFFP represents wildfire danger, which is closely linked to burn severity and fire-induced carbon emissions. This study presents a novel framework for enhancing regional risk assessment of high-intensity forest fires, providing valuable guidance in wildfire control and management.
{"title":"Probabilistic mapping of high-intensity forest fire potential via time series machine learning and remote sensing-informed fire spread simulations","authors":"Rui Chen , Yiru Zhang , Yanxi Li , Marta Yebra , Chunquan Fan , Hongguo Zhang , Binbin He","doi":"10.1016/j.rse.2026.115233","DOIUrl":"10.1016/j.rse.2026.115233","url":null,"abstract":"<div><div>High-intensity forest fires have significant destructive impacts on ecosystems and society, and are an increasing concern worldwide. Accurate probabilistic risk assessment of these fires can effectively enhance the ability to guide wildfire management, particularly for large and extreme fires. However, forecasting large-scale fire behavior characteristics remains challenging, limiting the effectiveness of spatial estimations of high-intensity forest fire potential (HIFFP). This study aims to integrate fire spread simulations and machine learning (ML) algorithms to enhance HIFFP estimations through multi-step time-series forecasting on fire rate of spread and fireline intensity at regional scales. We first established a high-intensity forest fire dataset based on remote sensing-informed fire spread simulations from the Weather Research and Forecasting coupled fire-spread model (WRF-SFIRE), incorporating explanatory variables on fuel, weather, climate, and topography. Then, the knowledge-guided framework (multi-step time series-based ML, MTS-ML) was designed to estimate HIFFP within different hours after fires occur, integrating with Bayesian Network (BN), Random Forest (RF), and copula models. Results indicate that MTS-ML improved HIFFP modeling compared with ML-based methods, achieving AUC (the area under the receiver operating characteristic curve) > 0.95 (with ∼0.04 increments), F1 score > 0.85 (with ∼0.08 increments), and MAE < 0.15. Topographic index, foliage fuel load, and wind speed are identified as primary contributors to HIFFP. Probabilistic mapping of HIFFP represents wildfire danger, which is closely linked to burn severity and fire-induced carbon emissions. This study presents a novel framework for enhancing regional risk assessment of high-intensity forest fires, providing valuable guidance in wildfire control and management.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"334 ","pages":"Article 115233"},"PeriodicalIF":11.4,"publicationDate":"2026-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145938801","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}
Pub Date : 2026-01-08DOI: 10.1016/j.rse.2025.115230
Yaqing Dou , Huaiqing Zhang , Hua Sun , Hui Lin , Yang Liu , Meng Zhang
Desertification is a global ecological and environmental problem, dynamic monitoring and accurate assessment of desertification are essential for restoring regional ecology and achieving sustainable development. Current desertification monitoring methods face dual challenges including unclear remote sensing mechanisms, non-robust extraction methods and the absence of high-resolution large-scale desertification products. This study constructs a comprehensive desertification index (CDI) by integrating multisource remote sensing data (Sentinel-1/2), incorporating three features —phenomenal indices (vegetation cover), cause indices (soil moisture) and essence indices (soil roughness)—on the basis of the multidimensional driving mechanisms of desertification physical processes. The Gaussian mixture model (GMM) was then applied to the CDI to automate desertification mapping for yielding the first 10 m-resolution annual desertification dataset in northern China (NCDMD, 2016–2024). The results demonstrated that the CDI-GMM based method achieves superior performance in mapping desertification scope across northern China in 2019, with an overall accuracy of 93.5 % and an overall accuracy of 86.4 % for desertification degree according to field survey data. In comparison, traditional approaches showed significantly lower accuracy, with the pixel dichotomy model (FVC-based) achieving 82.2 % in scope extraction and 50.3 % in degree classification, while the DDI feature space method reached 86.1 % and 64.2 %, respectively. Comparative experiments with five unsupervised classification methods (GMM, K-Means, MiniBatch K-Means, Jenks natural breaks, and Weka LVQ algorithms) indicated that the CDI combined with the GMM clustering algorithm can optimize the extraction of desertification and maintain stable performance, with an overall classification accuracy of over 93 %. Moreover, the NCDMD achieved consistent desertification mapping accuracies above 83 % throughout the 2016–2024 period, further demonstrating the robust spatiotemporal reliability of the proposed product. In summary, as a nationally significant high-resolution base dataset, the NCDMD not only fills the gap in high-precision desertification monitoring in China but also provides scientific support for ecological restoration assessment and land management policy-making.
{"title":"High-resolution annual desertification mapping in northern China using a novel comprehensive desertification index and unsupervised algorithm","authors":"Yaqing Dou , Huaiqing Zhang , Hua Sun , Hui Lin , Yang Liu , Meng Zhang","doi":"10.1016/j.rse.2025.115230","DOIUrl":"10.1016/j.rse.2025.115230","url":null,"abstract":"<div><div>Desertification is a global ecological and environmental problem, dynamic monitoring and accurate assessment of desertification are essential for restoring regional ecology and achieving sustainable development. Current desertification monitoring methods face dual challenges including unclear remote sensing mechanisms, non-robust extraction methods and the absence of high-resolution large-scale desertification products. This study constructs a comprehensive desertification index (CDI) by integrating multisource remote sensing data (Sentinel-1/2), incorporating three features —phenomenal indices (vegetation cover), cause indices (soil moisture) and essence indices (soil roughness)—on the basis of the multidimensional driving mechanisms of desertification physical processes. The Gaussian mixture model (GMM) was then applied to the CDI to automate desertification mapping for yielding the first 10 m-resolution annual desertification dataset in northern China (NCDMD, 2016–2024). The results demonstrated that the CDI-GMM based method achieves superior performance in mapping desertification scope across northern China in 2019, with an overall accuracy of 93.5 % and an overall accuracy of 86.4 % for desertification degree according to field survey data. In comparison, traditional approaches showed significantly lower accuracy, with the pixel dichotomy model (FVC-based) achieving 82.2 % in scope extraction and 50.3 % in degree classification, while the DDI feature space method reached 86.1 % and 64.2 %, respectively. Comparative experiments with five unsupervised classification methods (GMM, K-Means, MiniBatch K-Means, Jenks natural breaks, and Weka LVQ algorithms) indicated that the CDI combined with the GMM clustering algorithm can optimize the extraction of desertification and maintain stable performance, with an overall classification accuracy of over 93 %. Moreover, the NCDMD achieved consistent desertification mapping accuracies above 83 % throughout the 2016–2024 period, further demonstrating the robust spatiotemporal reliability of the proposed product. In summary, as a nationally significant high-resolution base dataset, the NCDMD not only fills the gap in high-precision desertification monitoring in China but also provides scientific support for ecological restoration assessment and land management policy-making.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"334 ","pages":"Article 115230"},"PeriodicalIF":11.4,"publicationDate":"2026-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145938860","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}
Pub Date : 2026-01-08DOI: 10.1016/j.rse.2026.115232
Wen Zhou , Claudio Persello , Dongping Ming , Shaowen Wang , Alfred Stein
Livable cities enhance urban economic development, improve physical and mental health, foster well-being, and foster urban sustainability. Evaluating urban livability is therefore important for policymakers to develop urban planning and development strategies aimed at improving livability. Mainstream methods of evaluating urban livability assign different weights to diverse indicators extracted from survey data, statistical data, and geospatial data. To relieve such time-consuming and labor-intensive data collection, this study proposes a transformer-based multi-task multimodal regression (TMTMR) model for the simultaneous evaluation of urban livability focusing on five domain-specific scores. Pretrained state-of-the-art computer vision and natural language processing models serve as backbones to extract features from high spatial resolution remote sensing (RS) images, digital surface models (DSM), night light remote sensing (NLRS) images and point of interest (POI) data. An attention mechanism helps the TMTMR model to assign varying significance levels to features from different modalities, thus capturing both intrinsic information and interrelationships among modalities for livability evaluation. Focusing on 13 Dutch areas, our research demonstrates that the TMTMR model efficiently evaluates urban livability with correlation coefficients ranging from 0.605 to 0.779, and root mean square error values between 0.070 and 0.112 in four unseen test areas. Furthermore, we analyze the synergy between different modalities. We found that modalities of urban livability can be effectively evaluated by aligning, in a descending order, contributions from RS images, NLRS images, DSM, and POI data. We demonstrated that the proposed TMTMR model is capable of effectively evaluating urban livability directly from multimodal geospatial data.
{"title":"A transformer based multi-task deep learning model for urban livability evaluation by fusing remote sensing and textual geospatial data","authors":"Wen Zhou , Claudio Persello , Dongping Ming , Shaowen Wang , Alfred Stein","doi":"10.1016/j.rse.2026.115232","DOIUrl":"10.1016/j.rse.2026.115232","url":null,"abstract":"<div><div>Livable cities enhance urban economic development, improve physical and mental health, foster well-being, and foster urban sustainability. Evaluating urban livability is therefore important for policymakers to develop urban planning and development strategies aimed at improving livability. Mainstream methods of evaluating urban livability assign different weights to diverse indicators extracted from survey data, statistical data, and geospatial data. To relieve such time-consuming and labor-intensive data collection, this study proposes a transformer-based multi-task multimodal regression (TMTMR) model for the simultaneous evaluation of urban livability focusing on five domain-specific scores. Pretrained state-of-the-art computer vision and natural language processing models serve as backbones to extract features from high spatial resolution remote sensing (RS) images, digital surface models (DSM), night light remote sensing (NLRS) images and point of interest (POI) data. An attention mechanism helps the TMTMR model to assign varying significance levels to features from different modalities, thus capturing both intrinsic information and interrelationships among modalities for livability evaluation. Focusing on 13 Dutch areas, our research demonstrates that the TMTMR model efficiently evaluates urban livability with correlation coefficients ranging from 0.605 to 0.779, and root mean square error values between 0.070 and 0.112 in four unseen test areas. Furthermore, we analyze the synergy between different modalities. We found that modalities of urban livability can be effectively evaluated by aligning, in a descending order, contributions from RS images, NLRS images, DSM, and POI data. We demonstrated that the proposed TMTMR model is capable of effectively evaluating urban livability directly from multimodal geospatial data.</div></div>","PeriodicalId":417,"journal":{"name":"Remote Sensing of Environment","volume":"334 ","pages":"Article 115232"},"PeriodicalIF":11.4,"publicationDate":"2026-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145938772","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}
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.
{"title":"Multi-sensor (since 1997) global soil moisture mapping with enhanced Spatio-temporal coverage through machine learning framework fusion","authors":"Haojie Zhang , Tianjie Zhao , Zhiqing Peng , Jingyao Zheng , Yu Bai , Nemesio Rodriguez-Fernadez , Donghai Zheng , Huazhu Xue , Zhanliang Yuan , Qian Cui , Peng Guo , Zushuai Wei , Peilin Song , Lixin Dong , Panpan Yao , Qiangqiang Yuan , Lingkui Meng , 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 m<sup>3</sup>/m<sup>3</sup>, and the ubRMSE is 0.059 m<sup>3</sup>/m<sup>3</sup>, 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":"2026-01-07","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}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号