International journal of applied earth observation and geoinformation : ITC journal最新文献_第5页

Forecasting sea level maxima using Machine learning with explainability and extreme value analysis 使用机器学习预测海平面最大值，具有可解释性和极值分析

IF 8.6 Q1 REMOTE SENSING

International journal of applied earth observation and geoinformation : ITC journal

Pub Date : 2026-01-06 DOI: 10.1016/j.jag.2025.105064

Saeed Rajabi-Kiasari , Nicole Delpeche-Ellmann , Artu Ellmann , Tarmo Soomere

Accurately forecasting sea level maxima (SLM) and extremes, is vital for maritime management, engineering, and navigation. Most Machine learning (ML) models focus on moderate surges and often underestimate extremes. We use a two-fold forecasting framework: ML/DL for short-term daily SLM forecasting and extreme value theory (EVT) for long-term extremes (<100 years). ML models include Random Forest, Extreme gradient boosting, and Multilayer perceptron, CNN-LSTM and CNN-GRU. Long-term extremes are analysed via EVT using a block maximum method. The Baltic Sea, a semi-enclosed micro-tidal basin prone to extremes, serves as case study. The analysis used six tide gauge stations (Narva, Ristna, Oulu, Kungsholmsfort, Władysławowo, and Greifswald). Key features—wind speed, surface air pressure, Baltic Sea Index (BSI), and significant wave height (SWH), were selected using a mutual information and models’ hyperparameters tuned using Bayesian optimization.

Neural networks models, specifically the CNN-GRU and MLP, performed best (RMSE 7–15 cm) with strong generalization. Most models captured storm events, but underestimated extreme peaks (>150 cm), due to the rarity in the training, incomplete meteorological representation, and missing local physical processes. CNN-GRU excelled in RMSE, recall, and F1, while MLP led in

R^{2}

and precision. EVT analysis showed winter extremes have ∼ 5–7 years in the north-east (Narva and Oulu). Explainability analysis of CNN-GRU showed prefilling dominates SLM at all stations; BSI, pressure, and winds drive west, south, and north, while local pressure, wind, and SWH dominate in the east. The framework supports early warning and long-term risk assessment, though forecasting rare extremes remains challenging.

准确预测海平面最大值（SLM）和极值，对于海事管理、工程和航海至关重要。大多数机器学习（ML）模型关注的是适度的波动，往往低估了极端情况。我们使用双重预测框架：ML/DL用于短期每日SLM预测，极值理论（EVT）用于长期极值（100年）预测。ML模型包括随机森林、极端梯度增强和多层感知器、CNN-LSTM和CNN-GRU。利用块极值法对EVT进行长期极值分析。波罗的海是一个容易发生极端情况的半封闭微潮盆地，可以作为研究案例。分析使用了6个潮汐测量站（Narva、Ristna、Oulu、Kungsholmsfort、Władysławowo和Greifswald）。风速、地面气压、波罗的海指数（BSI）和有效波高（SWH）等关键特征通过互信息和贝叶斯优化调整的模型超参数进行选择。神经网络模型，特别是CNN-GRU和MLP，表现最好（RMSE 7-15 cm），泛化能力强。大多数模型捕获了风暴事件，但由于训练中的稀缺性、不完整的气象表示和缺少当地物理过程，低估了极端峰值（150厘米）。CNN-GRU在RMSE、召回率和F1方面表现优异，而MLP在R2和精度方面表现优异。EVT分析显示，东北部（纳尔瓦和奥卢）冬季极端事件约为5-7年。CNN-GRU的可解释性分析表明，各站点的SLM以预充为主；BSI、气压和风驱动西、南、北，而局地气压、风和西南偏南主导东部。该框架支持早期预警和长期风险评估，尽管预测罕见的极端情况仍然具有挑战性。

{"title":"Forecasting sea level maxima using Machine learning with explainability and extreme value analysis","authors":"Saeed Rajabi-Kiasari , Nicole Delpeche-Ellmann , Artu Ellmann , Tarmo Soomere","doi":"10.1016/j.jag.2025.105064","DOIUrl":"10.1016/j.jag.2025.105064","url":null,"abstract":"<div><div>Accurately forecasting sea level maxima (SLM) and extremes, is vital for maritime management, engineering, and navigation. Most Machine learning (ML) models focus on moderate surges and often underestimate extremes. We use a two-fold forecasting framework: ML/DL for short-term daily SLM forecasting and extreme value theory (EVT) for long-term extremes (<100 years). ML models include Random Forest, Extreme gradient boosting, and Multilayer perceptron, CNN-LSTM and CNN-GRU. Long-term extremes are analysed via EVT using a block maximum method. The Baltic Sea, a semi-enclosed micro-tidal basin prone to extremes, serves as case study. The analysis used six tide gauge stations (Narva, Ristna, Oulu, Kungsholmsfort, Władysławowo, and Greifswald). Key features—wind speed, surface air pressure, Baltic Sea Index (BSI), and significant wave height (SWH), were selected using a mutual information and models’ hyperparameters tuned using Bayesian optimization.</div><div>Neural networks models, specifically the CNN-GRU and MLP, performed best (RMSE 7–15 cm) with strong generalization. Most models captured storm events, but underestimated extreme peaks (>150 cm), due to the rarity in the training, incomplete meteorological representation, and missing local physical processes. CNN-GRU excelled in RMSE, recall, and F1, while MLP led in <span><math><mrow><msup><mrow><mi>R</mi></mrow><mn>2</mn></msup></mrow></math></span> and precision. EVT analysis showed winter extremes have ∼ 5–7 years in the north-east (Narva and Oulu). Explainability analysis of CNN-GRU showed prefilling dominates SLM at all stations; BSI, pressure, and winds drive west, south, and north, while local pressure, wind, and SWH dominate in the east. The framework supports early warning and long-term risk assessment, though forecasting rare extremes remains challenging.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"146 ","pages":"Article 105064"},"PeriodicalIF":8.6,"publicationDate":"2026-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145926329","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Advancing CyanoHAB monitoring with hyperspectral data from NASA PACE: First results and validation 利用NASA PACE的高光谱数据推进CyanoHAB监测：首次结果和验证

IF 8.6 Q1 REMOTE SENSING

International journal of applied earth observation and geoinformation : ITC journal

Pub Date : 2026-01-06 DOI: 10.1016/j.jag.2025.105032

Abhishek Kumar , Chintan B. Maniyar , Nathan Tesfayi , Brice K. Grunert , Isabella R. Fiorentino , Kendra Herweck , Emily Hyland , Bingqing Liu , Bradley Bartelme , Deepak R. Mishra

This study presents the first assessment of NASA’s Phytoplankton, Aerosol, Cloud, ocean Ecosystem (PACE) mission’s hyperspectral Ocean Color Imager (OCI) for cyanobacterial harmful algal blooms (cyanoHABs) monitoring. We conducted a direct comparison of PACE OCI with Sentinel-3’s multispectral Ocean and Land Color Instrument (OLCI), and its Cyanobacteria Assessment Network (CyAN) operational product using imagery from summer 2024 blooms in Lake Erie, Green Bay, and Clear Lake. We evaluated performance of both sensors using the established Cyanobacteria Index (CI_Cyano) and corresponding cyanobacterial cell density (CCD) to estimate bloom biomass. PACE OCI successfully captured bloom patterns comparable to Sentinel-3 OLCI. When benchmarked against the CyAN product, OCI-derived CCD showed strong agreement (

R^{2}

= 0.84, Normalized Root Mean Square Error (NRMSE) = 8.95%), though a negative bias (

β ≃ 11 %

) was observed for extreme bloom pixels. Validation with in situ measurements indicated that OCI significantly improved chlorophyll-a biomass retrievals compared to CyAN/OLCI (NRMSE = 21.57% for OCI vs 38.67% for CyAN/OLCI), emphasizing the value of hyperspectral observations for optically complex inland waters. Our results demonstrate PACE OCI’s capability to advance CyanoHAB monitoring, providing a critical first step in establishing continuity with existing operational products while offering new potential for improved biomass estimates and taxonomic discrimination.

这项研究首次评估了美国宇航局浮游植物、气溶胶、云、海洋生态系统（PACE）任务的高光谱海洋彩色成像仪（OCI）对蓝藻有害藻华（cyanoHABs）的监测。我们将PACE OCI与Sentinel-3的多光谱海洋和陆地颜色仪器（OLCI）及其蓝藻评估网络（CyAN）操作产品进行了直接比较，使用的是2024年夏季伊利湖、绿湾和清澈湖的水华图像。我们使用已建立的蓝藻指数（CICyano）和相应的蓝藻细胞密度（CCD）来评估这两种传感器的性能，以估计水华生物量。PACE OCI成功捕获了与Sentinel-3 OLCI相当的开花模式。当与青色产品进行基准比较时，oci衍生的CCD显示出很强的一致性（R2 = 0.84，归一化均方根误差(NRMSE) = 8.95%），尽管在极端开花像素上观察到负偏差（β≃11%）。原位测量验证表明，与CyAN/OLCI相比，OCI显著提高了叶绿素-a生物量的检索（OCI的NRMSE = 21.57%，而CyAN/OLCI的NRMSE为38.67%），强调了高光谱观测对光学复杂的内陆水域的价值。我们的研究结果表明，PACE OCI有能力推进蓝藻藻监测，为建立与现有业务产品的连续性提供了关键的第一步，同时为改进生物量估算和分类区分提供了新的潜力。

{"title":"Advancing CyanoHAB monitoring with hyperspectral data from NASA PACE: First results and validation","authors":"Abhishek Kumar , Chintan B. Maniyar , Nathan Tesfayi , Brice K. Grunert , Isabella R. Fiorentino , Kendra Herweck , Emily Hyland , Bingqing Liu , Bradley Bartelme , Deepak R. Mishra","doi":"10.1016/j.jag.2025.105032","DOIUrl":"10.1016/j.jag.2025.105032","url":null,"abstract":"<div><div>This study presents the first assessment of NASA’s Phytoplankton, Aerosol, Cloud, ocean Ecosystem (PACE) mission’s hyperspectral Ocean Color Imager (OCI) for cyanobacterial harmful algal blooms (cyanoHABs) monitoring. We conducted a direct comparison of PACE OCI with Sentinel-3’s multispectral Ocean and Land Color Instrument (OLCI), and its Cyanobacteria Assessment Network (CyAN) operational product using imagery from summer 2024 blooms in Lake Erie, Green Bay, and Clear Lake. We evaluated performance of both sensors using the established Cyanobacteria Index (CI<sub>Cyano</sub>) and corresponding cyanobacterial cell density (CCD) to estimate bloom biomass. PACE OCI successfully captured bloom patterns comparable to Sentinel-3 OLCI. When benchmarked against the CyAN product, OCI-derived CCD showed strong agreement (<span><math><msup><mrow><mi>R</mi></mrow><mrow><mn>2</mn></mrow></msup></math></span> = 0.84, Normalized Root Mean Square Error (NRMSE) = 8.95%), though a negative bias (<span><math><mrow><mi>β</mi><mo>≃</mo><mn>11</mn><mtext>%</mtext></mrow></math></span>) was observed for extreme bloom pixels. Validation with <em>in situ</em> measurements indicated that OCI significantly improved chlorophyll-<em>a</em> biomass retrievals compared to CyAN/OLCI (NRMSE = 21.57% for OCI vs 38.67% for CyAN/OLCI), emphasizing the value of hyperspectral observations for optically complex inland waters. Our results demonstrate PACE OCI’s capability to advance CyanoHAB monitoring, providing a critical first step in establishing continuity with existing operational products while offering new potential for improved biomass estimates and taxonomic discrimination.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"146 ","pages":"Article 105032"},"PeriodicalIF":8.6,"publicationDate":"2026-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145926446","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

MT-RoadNet: A heterogeneous network with local–global joint enhancement for road surface and centerline extraction MT-RoadNet：一个具有局部-全局联合增强的异构网络，用于路面和中心线提取

IF 8.6 Q1 REMOTE SENSING

International journal of applied earth observation and geoinformation : ITC journal

Pub Date : 2026-01-06 DOI: 10.1016/j.jag.2025.105080

Zejiao Wang, Longgang Xiang, Meng Wang, Xingjuan Wang, Fengwei Jiao

Road network information has extensive applications, such as urban planning and navigation. However, current road extraction methods mostly rely on single model architectures and single-source remote sensing images, neglecting the potential benefits of collaborative extraction with heterogeneous networks and multi-source data fusion. Moreover, existing methods often suffer from road fragmentation and poorly connected road graphs due to occlusions. To address these challenges, we propose MT-RoadNet, a local–global joint enhancement framework for extracting road surfaces and centerlines, which incorporates multi-network collaborative optimization. Specifically, MT-RoadNet adopts a dual-branch structure, which not only facilitates the collaborative extraction of local details and global semantics but also achieves the coupling of geometry and topology through a cross-task dynamic interaction mechanism. Besides, we propose the Local–Global Feature Fusion module (LGFF), which dynamically integrates local details and global semantics through multi-level feature interaction. Furthermore, to reduce interference features with high inter-class separability and low intra-class variation, we innovatively design the Visual State Space Module (VSSM) and the Spatial-Channel Mutual Attention (SCMA). The VSSM weighs features dynamically using multi-directional cross-scanning and global receptive fields, emphasizing prominent area information while improving computational efficiency. SCMA effectively guides the model to focus on semantically relevant regions. Finally, MT-RoadNet adopts a dual-path decoder to produce road surfaces and centerlines. Extensive experiments on three road datasets demonstrate that MT-RoadNet significantly outperforms existing state-of-the-art methods in terms of road completeness and recognition accuracy of topological structure. The code is available at https://github.com/508hz1207/MTNet.

路网信息在城市规划、导航等方面有着广泛的应用。然而，目前的道路提取方法大多依赖于单一模型架构和单源遥感图像，忽视了异构网络协同提取和多源数据融合的潜在优势。此外，现有的方法往往存在道路碎片化和由于遮挡导致的道路图连接不良的问题。为了应对这些挑战，我们提出了MT-RoadNet，这是一个局部-全局联合增强框架，用于提取路面和中心线，其中包含多网络协作优化。具体而言，MT-RoadNet采用双分支结构，既有利于局部细节和全局语义的协同提取，又通过跨任务动态交互机制实现了几何和拓扑的耦合。此外，我们还提出了local - global Feature Fusion模块（LGFF），该模块通过多层次的特征交互，动态集成了局部细节和全局语义。此外，为了降低类间可分离性高、类内变化小的干扰特征，我们创新性地设计了视觉状态空间模块（VSSM）和空间信道相互注意（SCMA）。VSSM利用多向交叉扫描和全局接受野动态加权特征，在强调突出区域信息的同时提高了计算效率。SCMA有效地引导模型关注语义相关区域。最后，MT-RoadNet采用双路解码器生成路面和中心线。在三个道路数据集上进行的大量实验表明，MT-RoadNet在道路完整性和拓扑结构识别精度方面明显优于现有的最先进方法。代码可在https://github.com/508hz1207/MTNet上获得。

{"title":"MT-RoadNet: A heterogeneous network with local–global joint enhancement for road surface and centerline extraction","authors":"Zejiao Wang, Longgang Xiang, Meng Wang, Xingjuan Wang, Fengwei Jiao","doi":"10.1016/j.jag.2025.105080","DOIUrl":"10.1016/j.jag.2025.105080","url":null,"abstract":"<div><div>Road network information has extensive applications, such as urban planning and navigation. However, current road extraction methods mostly rely on single model architectures and single-source remote sensing images, neglecting the potential benefits of collaborative extraction with heterogeneous networks and multi-source data fusion. Moreover, existing methods often suffer from road fragmentation and poorly connected road graphs due to occlusions. To address these challenges, we propose MT-RoadNet, a local–global joint enhancement framework for extracting road surfaces and centerlines, which incorporates multi-network collaborative optimization. Specifically, MT-RoadNet adopts a dual-branch structure, which not only facilitates the collaborative extraction of local details and global semantics but also achieves the coupling of geometry and topology through a cross-task dynamic interaction mechanism. Besides, we propose the Local–Global Feature Fusion module (LGFF), which dynamically integrates local details and global semantics through multi-level feature interaction. Furthermore, to reduce interference features with high inter-class separability and low intra-class variation, we innovatively design the Visual State Space Module (VSSM) and the Spatial-Channel Mutual Attention (SCMA). The VSSM weighs features dynamically using multi-directional cross-scanning and global receptive fields, emphasizing prominent area information while improving computational efficiency. SCMA effectively guides the model to focus on semantically relevant regions. Finally, MT-RoadNet adopts a dual-path decoder to produce road surfaces and centerlines. Extensive experiments on three road datasets demonstrate that MT-RoadNet significantly outperforms existing state-of-the-art methods in terms of road completeness and recognition accuracy of topological structure. The code is available at <span><span>https://github.com/508hz1207/MTNet</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"146 ","pages":"Article 105080"},"PeriodicalIF":8.6,"publicationDate":"2026-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145926449","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Large-scale high-resolution coastal subsidence mapping in eastern China with Sentinel-1 and Sentinel-2: Heterogeneous patterns and primary drivers 基于Sentinel-1和Sentinel-2的中国东部大尺度高分辨率沿海沉降制图：非均质模式和主要驱动因素

IF 8.6 Q1 REMOTE SENSING

International journal of applied earth observation and geoinformation : ITC journal

Pub Date : 2026-01-06 DOI: 10.1016/j.jag.2025.105047

Peng Li , Jianbo Bai , Lin Shen , Wei Tang , Cunren Liang , Bin Zhao , Zhenhong Li , Houjie Wang

A majorty of the low-lying coastal areas worldwide, where the population is densely concentrated, are confronted with high to extremetly high risks of land subsidence. However, a comprehensive detection and quantification of large-scale coastal subsidence patterns and their primary drivers in eastern China remain lacking. In this study, we employed Time Series Interferometric Synthetic Aperture Radar (TS-InSAR) technique and Sentinel-1 data to derive the vertical deformation pattern at a resolution of 90 m from 2017 to 2024. We developed a novel multi-frame mosaicking method, achieving spatially consistent InSAR observations over the land-sea transition areas. Our findings uncover extensive coastal subsidence in northeastern Shandong and eastern Jiangsu, highlighting several rapid-subsidence funnels with rates exceeding 50 mm/yr for the first time. By integrating Sentinel-2 multispectral imagery, subsidence time series, groundwater level measurements, and principal component analysis (PCA), we further analyzed the spatiotemporal distribution patterns and underlying drivers of these heterogeneous subsidence funnels. Our analysis demonstrates that anthropogenic factors are the dominant drivers of coastal subsidence. Four representative case studies reveal distinct subsidence mechanisms: (1) brine extraction for salt production and aquaculture, (2) excessive freshwater withdrawal for agricultural irrigation and industrial use, (3) groundwater depletion for intensive greenhouse aquaculture, and (4) land reclamation for industrial infrastructure development. In each region, subsidence patterns are predominantly controlled by a single dominant factor. This study is expected to provide valuable insights for monitoring and managing coastal subsidence, enhance our understanding of associated risks, and offer critical guidance for protecting communities in vulnerable coastal areas.

世界上大部分人口密集的沿海低洼地区都面临着高至极高的地面沉降风险。然而，中国东部沿海大尺度沉降模式及其主要驱动因素的综合检测和量化仍然缺乏。本研究利用时序干涉合成孔径雷达（TS-InSAR）技术和Sentinel-1数据，获得了2017 - 2024年90 m分辨率的垂直形变模式。我们开发了一种新的多帧拼接方法，实现了陆海过渡区InSAR观测的空间一致性。我们的研究结果揭示了山东东北部和江苏东部广泛的沿海沉降，并首次突出了几个速度超过50毫米/年的快速沉降漏斗。通过综合Sentinel-2多光谱影像、沉降时间序列、地下水位测量数据和主成分分析（PCA），进一步分析了这些非均匀沉降漏斗的时空分布格局及其驱动因素。分析表明，人为因素是沿海沉降的主要驱动因素。四个具有代表性的案例研究揭示了不同的沉降机制：(1)盐生产和水产养殖的盐水抽取，(2)农业灌溉和工业用水的过度淡水抽取，(3)集约化温室水产养殖的地下水枯竭，以及(4)工业基础设施建设的土地复垦。各区域沉降模式主要受单一主导因素控制。这项研究有望为监测和管理沿海下沉提供有价值的见解，增强我们对相关风险的理解，并为保护脆弱沿海地区的社区提供重要指导。

{"title":"Large-scale high-resolution coastal subsidence mapping in eastern China with Sentinel-1 and Sentinel-2: Heterogeneous patterns and primary drivers","authors":"Peng Li , Jianbo Bai , Lin Shen , Wei Tang , Cunren Liang , Bin Zhao , Zhenhong Li , Houjie Wang","doi":"10.1016/j.jag.2025.105047","DOIUrl":"10.1016/j.jag.2025.105047","url":null,"abstract":"<div><div>A majorty of the low-lying coastal areas worldwide, where the population is densely concentrated, are confronted with high to extremetly high risks of land subsidence. However, a comprehensive detection and quantification of large-scale coastal subsidence patterns and their primary drivers in eastern China remain lacking. In this study, we employed Time Series Interferometric Synthetic Aperture Radar (TS-InSAR) technique and Sentinel-1 data to derive the vertical deformation pattern at a resolution of 90 m from 2017 to 2024. We developed a novel multi-frame mosaicking method, achieving spatially consistent InSAR observations over the land-sea transition areas. Our findings uncover extensive coastal subsidence in northeastern Shandong and eastern Jiangsu, highlighting several rapid-subsidence funnels with rates exceeding 50 mm/yr for the first time. By integrating Sentinel-2 multispectral imagery, subsidence time series, groundwater level measurements, and principal component analysis (PCA), we further analyzed the spatiotemporal distribution patterns and underlying drivers of these heterogeneous subsidence funnels. Our analysis demonstrates that anthropogenic factors are the dominant drivers of coastal subsidence. Four representative case studies reveal distinct subsidence mechanisms: (1) brine extraction for salt production and aquaculture, (2) excessive freshwater withdrawal for agricultural irrigation and industrial use, (3) groundwater depletion for intensive greenhouse aquaculture, and (4) land reclamation for industrial infrastructure development. In each region, subsidence patterns are predominantly controlled by a single dominant factor. This study is expected to provide valuable insights for monitoring and managing coastal subsidence, enhance our understanding of associated risks, and offer critical guidance for protecting communities in vulnerable coastal areas.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"146 ","pages":"Article 105047"},"PeriodicalIF":8.6,"publicationDate":"2026-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145926090","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

City-scale building instance segmentation from LiDAR point clouds via structure-aware method 基于结构感知方法的LiDAR点云城市尺度建筑实例分割

IF 8.6 Q1 REMOTE SENSING

International journal of applied earth observation and geoinformation : ITC journal

Pub Date : 2026-01-06 DOI: 10.1016/j.jag.2026.105086

Jinpeng Li , Yuan Li , Yiping Chen , Hongchao Fan , Ruisheng Wang

Building instance segmentation from city-scale point cloud is of great significance to urban planning management, disaster response and recovery, and land resource management. However, due to the complexity of urban scenes and sparse nature of LiDAR data, existing methods are often limited by the problems of obscured building boundaries and incomplete building structures, particularly in densely populated urban areas with diverse architectural styles. To address these challenges, we propose a novel method that automatically extracts building instances from airborne LiDAR data and is especially aware of the building structures. The proposed method encompasses two main stages, building points semantic segmentation and individual building extraction. First, we design a point cloud semantic segmentation network, VPBE-Net, that innovatively utilizes voxel-point cloud fused features to efficiently extract building points from large-scale point cloud. Second, building instances are automatically and robustly extracted using a graph-based algorithm SI-DVDC, which comprehensively considers both object-level building structure property and point-level density accessibility. We evaluate the semantic segmentation performance on the DALES and Toronto datasets and the building instance segmentation performance on the UrbanBIS and City-BIS datasets. For the semantics, Overall Accuracy (OA) and mean Intersection over Union (mIoU) metrics reach 88.96 % and 70.28 % on DALES dataset, and 89.26% and 75.40% on Toronto dataset, which is 2.22 % and 3.25 % higher than the state-of-the-art methods. For the building instance extraction, the instance-level quality metric reach 88.65 % on UrbanBIS dataset and 76.97 % on City-BIS dataset, respectively. The experiments verify that the proposed method can extract individual buildings from complex urban and rural environments, while being aware of diverse building structures, thereby demonstrating the remarkable generalization ability. To facilitate future research, we make source code and dataset available at https://github.com/Lijp411/City-BIS.

基于城市尺度点云的建筑实例分割对城市规划管理、灾害响应与恢复、土地资源管理等具有重要意义。然而，由于城市场景的复杂性和LiDAR数据的稀疏性，现有的方法往往受到建筑物边界模糊和建筑物结构不完整的问题的限制，特别是在人口密集、建筑风格多样的城市地区。为了解决这些挑战，我们提出了一种新的方法，从机载激光雷达数据中自动提取建筑实例，并特别注意建筑结构。该方法主要包括两个阶段：建筑点语义分割和单个建筑提取。首先，设计了点云语义分割网络VPBE-Net，创新地利用体素-点云融合特征，从大规模点云中高效提取建筑点；其次，采用基于图的SI-DVDC算法，综合考虑对象级建筑结构属性和点级密度可达性，实现建筑实例的自动鲁棒提取；我们评估了DALES和Toronto数据集上的语义分割性能，以及UrbanBIS和City-BIS数据集上的建筑实例分割性能。在语义方面，DALES数据集的总体准确率（Overall Accuracy， OA）和平均交汇率（Intersection over Union, mIoU）指标分别达到88.96%和70.28%，Toronto数据集达到89.26%和75.40%，分别比目前最先进的方法高2.22%和3.25%。对于建筑实例提取，在UrbanBIS数据集和City-BIS数据集上，实例级质量指标分别达到了88.65%和76.97%。实验结果表明，该方法能够从复杂的城乡环境中提取出单个建筑，同时能够感知到建筑结构的多样性，具有显著的泛化能力。为了方便未来的研究，我们在https://github.com/Lijp411/City-BIS上提供了源代码和数据集。

{"title":"City-scale building instance segmentation from LiDAR point clouds via structure-aware method","authors":"Jinpeng Li , Yuan Li , Yiping Chen , Hongchao Fan , Ruisheng Wang","doi":"10.1016/j.jag.2026.105086","DOIUrl":"10.1016/j.jag.2026.105086","url":null,"abstract":"<div><div>Building instance segmentation from city-scale point cloud is of great significance to urban planning management, disaster response and recovery, and land resource management. However, due to the complexity of urban scenes and sparse nature of LiDAR data, existing methods are often limited by the problems of obscured building boundaries and incomplete building structures, particularly in densely populated urban areas with diverse architectural styles. To address these challenges, we propose a novel method that automatically extracts building instances from airborne LiDAR data and is especially aware of the building structures. The proposed method encompasses two main stages, building points semantic segmentation and individual building extraction. First, we design a point cloud semantic segmentation network, VPBE-Net, that innovatively utilizes voxel-point cloud fused features to efficiently extract building points from large-scale point cloud. Second, building instances are automatically and robustly extracted using a graph-based algorithm SI-DVDC, which comprehensively considers both object-level building structure property and point-level density accessibility. We evaluate the semantic segmentation performance on the DALES and Toronto datasets and the building instance segmentation performance on the UrbanBIS and City-BIS datasets. For the semantics, Overall Accuracy (OA) and mean Intersection over Union (mIoU) metrics reach 88.96 % and 70.28 % on DALES dataset, and 89.26% and 75.40% on Toronto dataset, which is 2.22 % and 3.25 % higher than the state-of-the-art methods. For the building instance extraction, the instance-level quality metric reach 88.65 % on UrbanBIS dataset and 76.97 % on City-BIS dataset, respectively. The experiments verify that the proposed method can extract individual buildings from complex urban and rural environments, while being aware of diverse building structures, thereby demonstrating the remarkable generalization ability. To facilitate future research, we make source code and dataset available at https://github.com/Lijp411/City-BIS.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"146 ","pages":"Article 105086"},"PeriodicalIF":8.6,"publicationDate":"2026-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145926450","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

WeavingUnet: Enhancing dense road extraction by integrating horizontal and vertical features WeavingUnet：通过整合水平和垂直特征，增强密集道路提取

IF 8.6 Q1 REMOTE SENSING

International journal of applied earth observation and geoinformation : ITC journal

Pub Date : 2026-01-06 DOI: 10.1016/j.jag.2025.105037

Xianzhi Ma, Jianhui Li, Mingyang Lv

Road extraction plays a crucial role in automatically identifying road networks from remote sensing images, which is essential for various applications. However, existing road extraction methods perform poorly when dealing with dense roads. Inspired by the act of ”weaving” in real life, this paper proposes a dense road extraction method named WeavingUnet, aimed at addressing the difficulties in effectively extracting dense roads using existing methods. This model explores a novel approach to road feature extraction by analyzing road features in various directions and simulating the geometric shapes of roads to more accurately perform dense road extraction. The model incorporates innovative components such as Snake Weaving Attention, Contextual Information Weaving Module, Global Information Extraction Module, and Multi-Scale Weaving Decoder, which collectively enhance the extraction of multi-level road information and optimize the road reconstruction process. The model is tested on the public datasets DeepGlobe and Massachusetts, achieving F1 scores of 82.73% and 72.52%, and IoUs of 71.44% and 61.01%, respectively. Compared with the latest SOTA models, the F1 scores increase by 1.53% and 1.22%, and the IoUs improve by 1.76% and 1.87%. This study also conducts a comprehensive analysis of dense road extraction and compared it with SOTA models. In the densest areas, the false negative rates (FNs) in the test images are reduced by 4.93‰ and 1.47‰, respectively, demonstrating the effectiveness of WeavingUnet in handling dense roads. The code and dataset are available at https://github.com/XianZhi-Ma/WeavingUnet.

道路提取在从遥感图像中自动识别道路网中起着至关重要的作用，在各种应用中都是必不可少的。然而，现有的道路提取方法在处理密集道路时表现不佳。本文受现实生活中“编织”行为的启发，提出了一种名为WeavingUnet的密集道路提取方法，旨在解决现有方法难以有效提取密集道路的问题。该模型探索了一种新的道路特征提取方法，通过分析不同方向的道路特征，模拟道路的几何形状，更准确地进行密集的道路提取。该模型融合了Snake编织关注、上下文信息编织模块、全局信息提取模块、多尺度编织解码器等创新组件，共同增强了多层次道路信息的提取，优化了道路重构流程。该模型在DeepGlobe和Massachusetts公共数据集上进行了测试，F1得分分别为82.73%和72.52%，IoUs得分分别为71.44%和61.01%。与最新款SOTA车型相比，F1分数分别提高了1.53%和1.22%，iou分别提高了1.76%和1.87%。本研究还对密集道路提取进行了综合分析，并与SOTA模型进行了比较。在最密集的区域，测试图像的假阴性率（FNs）分别降低了4.93‰和1.47‰，证明了织网在处理密集道路方面的有效性。代码和数据集可从https://github.com/XianZhi-Ma/WeavingUnet获得。

{"title":"WeavingUnet: Enhancing dense road extraction by integrating horizontal and vertical features","authors":"Xianzhi Ma, Jianhui Li, Mingyang Lv","doi":"10.1016/j.jag.2025.105037","DOIUrl":"10.1016/j.jag.2025.105037","url":null,"abstract":"<div><div>Road extraction plays a crucial role in automatically identifying road networks from remote sensing images, which is essential for various applications. However, existing road extraction methods perform poorly when dealing with dense roads. Inspired by the act of ”weaving” in real life, this paper proposes a dense road extraction method named WeavingUnet, aimed at addressing the difficulties in effectively extracting dense roads using existing methods. This model explores a novel approach to road feature extraction by analyzing road features in various directions and simulating the geometric shapes of roads to more accurately perform dense road extraction. The model incorporates innovative components such as Snake Weaving Attention, Contextual Information Weaving Module, Global Information Extraction Module, and Multi-Scale Weaving Decoder, which collectively enhance the extraction of multi-level road information and optimize the road reconstruction process. The model is tested on the public datasets DeepGlobe and Massachusetts, achieving F1 scores of 82.73% and 72.52%, and IoUs of 71.44% and 61.01%, respectively. Compared with the latest SOTA models, the F1 scores increase by 1.53% and 1.22%, and the IoUs improve by 1.76% and 1.87%. This study also conducts a comprehensive analysis of dense road extraction and compared it with SOTA models. In the densest areas, the false negative rates (FNs) in the test images are reduced by 4.93‰ and 1.47‰, respectively, demonstrating the effectiveness of WeavingUnet in handling dense roads. The code and dataset are available at <span><span>https://github.com/XianZhi-Ma/WeavingUnet</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"146 ","pages":"Article 105037"},"PeriodicalIF":8.6,"publicationDate":"2026-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145925516","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Decadal changes and influencing factors of global industrial heat sources 全球工业热源的年代际变化及其影响因素

IF 8.6 Q1 REMOTE SENSING

International journal of applied earth observation and geoinformation : ITC journal

Pub Date : 2026-01-06 DOI: 10.1016/j.jag.2025.105058

Yakun Xie , Ruifeng Xia , Jianbo Lai , Yaoji Zhao , Chaoda Song , Wen Song , Dejun Feng , Jun Zhu , Ya Hu

Industrial heat sources are major contributors to greenhouse gas emissions, environmental degradation, and energy transitions, yet systematic global monitoring has been hindered by fragmented and sector-specific datasets. Here, we present a deep learning framework applied to high-resolution satellite imagery to automatically classify six major categories of industrial facilities—cement plants, steel plants, chemical plants, oil and gas development platforms, coal chemical plants, and open-pit mines. Based on this approach, we generate the first annual global dataset of industrial heat sources from 2013 to 2023. Our method achieves an overall classification accuracy of 95.38 % and an average precision of 94.38 %, with efficiency gains exceeding 98 % compared to traditional manual surveys. The results reveal distinct decadal trends: oil and gas platforms steadily increased, coal chemical plants declined markedly, cement plants and open-pit mines remained stable, while steel and chemical plants exhibited moderate growth. Correlation analysis further shows that economic development, population, and development stage strongly shape the spatial distribution of industrial activity. In addition, the dataset captures short-term disruptions, such as reduced activity during the COVID-19 pandemic. This study provides a reproducible and scalable foundation for analyzing industrial transitions, advancing global carbon accounting, and informing environmental and energy policy.

工业热源是温室气体排放、环境恶化和能源转型的主要原因，但系统的全球监测一直受到零散和特定行业数据集的阻碍。在这里，我们提出了一个应用于高分辨率卫星图像的深度学习框架，以自动分类六大类工业设施——水泥厂、钢铁厂、化工厂、石油和天然气开发平台、煤化工工厂和露天矿。基于这种方法，我们生成了2013年至2023年全球第一个年度工业热源数据集。该方法总体分类精度为95.38%，平均分类精度为94.38%，与传统人工调查相比，分类效率提高98%以上。结果显示出明显的年代际变化趋势：石油和天然气平台稳步增长，煤化工企业明显下降，水泥厂和露天矿保持稳定，钢铁和化工企业呈现温和增长。相关分析进一步表明，经济发展、人口和发展阶段对产业活动的空间分布具有重要影响。此外，该数据集还记录了短期中断，例如COVID-19大流行期间活动减少。本研究为分析产业转型、推进全球碳核算以及为环境和能源政策提供信息提供了可复制和可扩展的基础。

{"title":"Decadal changes and influencing factors of global industrial heat sources","authors":"Yakun Xie , Ruifeng Xia , Jianbo Lai , Yaoji Zhao , Chaoda Song , Wen Song , Dejun Feng , Jun Zhu , Ya Hu","doi":"10.1016/j.jag.2025.105058","DOIUrl":"10.1016/j.jag.2025.105058","url":null,"abstract":"<div><div>Industrial heat sources are major contributors to greenhouse gas emissions, environmental degradation, and energy transitions, yet systematic global monitoring has been hindered by fragmented and sector-specific datasets. Here, we present a deep learning framework applied to high-resolution satellite imagery to automatically classify six major categories of industrial facilities—cement plants, steel plants, chemical plants, oil and gas development platforms, coal chemical plants, and open-pit mines. Based on this approach, we generate the first annual global dataset of industrial heat sources from 2013 to 2023. Our method achieves an overall classification accuracy of 95.38 % and an average precision of 94.38 %, with efficiency gains exceeding 98 % compared to traditional manual surveys. The results reveal distinct decadal trends: oil and gas platforms steadily increased, coal chemical plants declined markedly, cement plants and open-pit mines remained stable, while steel and chemical plants exhibited moderate growth. Correlation analysis further shows that economic development, population, and development stage strongly shape the spatial distribution of industrial activity. In addition, the dataset captures short-term disruptions, such as reduced activity during the COVID-19 pandemic. This study provides a reproducible and scalable foundation for analyzing industrial transitions, advancing global carbon accounting, and informing environmental and energy policy.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"146 ","pages":"Article 105058"},"PeriodicalIF":8.6,"publicationDate":"2026-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145925608","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A geometric consistency constrained hierarchical global SfM for large-scale UAV images 基于几何一致性约束的大尺度无人机图像分层全局SfM

IF 8.6 Q1 REMOTE SENSING

International journal of applied earth observation and geoinformation : ITC journal

Pub Date : 2026-01-06 DOI: 10.1016/j.jag.2025.105075

Yan Zhou , Xianwei Zheng , Jinding Gao , Qian Shi , Xiaoping Liu

Efficient and effective pose estimation and 3D point clouds reconstruction from low-altitude UAV images is crucial for digital twins and geospatial applications. However, conventional divide-and-conquer SfM pipelines remain constrained by the high computational cost of incremental reconstruction. While global SfM, though faster, often suffer from instability due to outlier sensitivity in translation averaging. To address these limitations, we propose a feature track enhanced global SfM within divide-and-conquer framework. To deal with outliers, a cluster consistency constrained outlier filtering method is proposed, which combines distribution-aware scene partition and cluster consistency to remove false matches. Furthermore, we propose a geometry-constrained alignment optimization strategy in sub-models merging process to eliminate misalignments and ghosting artifacts, obtaining complete and accurate 3D models. Extensive experiments on ETH3D and SYSU UAV large-scale datasets verify that the proposed method outperforms established baselines, with comparable accuracy to the incremental SfM but half of the computation time. The reconstructed 3D models demonstrate notable enhancements in both structural integrity and level of detail, proving the method’s high applicability in large-scale 3D reconstruction tasks. The code and data will be released: https://github.com/BunnyanChou/HieGSfM.

基于低空无人机图像的高效姿态估计和三维点云重建对于数字孪生和地理空间应用至关重要。然而，传统的分而治之的SfM管道仍然受到增量重建的高计算成本的限制。而全局SfM虽然速度更快，但由于平移平均的离群值敏感性，往往存在不稳定性。为了解决这些限制，我们提出了一个分而治之框架下的特征跟踪增强的全局SfM。针对异常点的处理，提出了一种聚类一致性约束的异常点滤波方法，将分布感知场景划分和聚类一致性相结合，去除虚假匹配。在子模型合并过程中，提出了一种几何约束的对齐优化策略，以消除子模型的不对齐和重影现象，获得完整、准确的三维模型。在ETH3D和SYSU无人机大规模数据集上的大量实验验证了所提出的方法优于已建立的基线，具有与增量SfM相当的精度，但计算时间只有一半。重建的三维模型在结构完整性和细节水平上都有显著提高，证明了该方法在大规模三维重建任务中的高适用性。代码和数据将发布：https://github.com/BunnyanChou/HieGSfM。

{"title":"A geometric consistency constrained hierarchical global SfM for large-scale UAV images","authors":"Yan Zhou , Xianwei Zheng , Jinding Gao , Qian Shi , Xiaoping Liu","doi":"10.1016/j.jag.2025.105075","DOIUrl":"10.1016/j.jag.2025.105075","url":null,"abstract":"<div><div>Efficient and effective pose estimation and 3D point clouds reconstruction from low-altitude UAV images is crucial for digital twins and geospatial applications. However, conventional divide-and-conquer SfM pipelines remain constrained by the high computational cost of incremental reconstruction. While global SfM, though faster, often suffer from instability due to outlier sensitivity in translation averaging. To address these limitations, we propose a feature track enhanced global SfM within divide-and-conquer framework. To deal with outliers, a cluster consistency constrained outlier filtering method is proposed, which combines distribution-aware scene partition and cluster consistency to remove false matches. Furthermore, we propose a geometry-constrained alignment optimization strategy in sub-models merging process to eliminate misalignments and ghosting artifacts, obtaining complete and accurate 3D models. Extensive experiments on ETH3D and SYSU UAV large-scale datasets verify that the proposed method outperforms established baselines, with comparable accuracy to the incremental SfM but half of the computation time. The reconstructed 3D models demonstrate notable enhancements in both structural integrity and level of detail, proving the method’s high applicability in large-scale 3D reconstruction tasks. The code and data will be released: <span><span><u>https://github.com/BunnyanChou/HieGSfM</u></span><svg><path></path></svg></span>.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"146 ","pages":"Article 105075"},"PeriodicalIF":8.6,"publicationDate":"2026-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145926331","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Canopy height mapping in the Western Himalayas, Pakistan: A deep learning approach using GEDI and Sentinel-2 fusion 巴基斯坦西喜马拉雅地区的冠层高度测绘：使用GEDI和Sentinel-2融合的深度学习方法

IF 8.6 Q1 REMOTE SENSING

International journal of applied earth observation and geoinformation : ITC journal

Pub Date : 2026-01-06 DOI: 10.1016/j.jag.2025.105030

Adeel Ahmad , Srikumar Sastry , Aayush Dhakal , Subash Khanal , Alex Levering , Hammad Gilani , Nathan Jacobs

The western Himalayas in Pakistan, characterized by a diverse range of conifer species at higher elevations, represent a critical biodiversity hotspot and habitat for numerous species. Accurate spatial assessments of canopy height are essential for improving estimates of aboveground biomass, carbon sequestration, and associated forest ecosystem services in this region. In this study, we estimated canopy heights in the western Himalayas in Pakistan using a U-Net CNN approach, fusing data from the Global Ecosystem Dynamics Investigation Mission (GEDI) with multi-band Sentinel-2 (S2) imagery. We produced a canopy height map at a 10 m resolution for 2020. To ensure accurate measurements across various canopy height groups, we implemented a stratified training approach that optimized the representation of GEDI data throughout the training, validation, and testing phases. We trained multiple models using varying thresholds and assigned different weights to taller trees to improve accuracy between different canopy height groups in the study region. Our best model achieved a Root Mean Square Error (RMSE) of 7.52 m and a Mean Absolute Error (MAE) of 5.71 m in the test set, significantly outperforming existing global canopy height models in this topographically complex region. We further validate our predictions against field inventory plots, achieving a coefficient of determination (R

^{2}

) of 0.49 for plots containing at least 15 trees. The resulting tree canopy height map, designated as the Western Himalaya Canopy Height Map (WHiCH Map), is publicly available.

巴基斯坦西喜马拉雅地区海拔较高，针叶树种类繁多，是一个重要的生物多样性热点和众多物种的栖息地。准确的冠层高度空间评估对于改善该地区地上生物量、碳固存和相关森林生态系统服务的估算至关重要。在这项研究中，我们使用U-Net CNN方法，融合了来自全球生态系统动力学调查任务（GEDI）的数据和多波段Sentinel-2 （S2）图像，估计了巴基斯坦西喜马拉雅地区的冠层高度。我们制作了一张2020年10米分辨率的冠层高度图。为了确保各种冠层高度组的准确测量，我们实施了一种分层训练方法，在整个训练、验证和测试阶段优化GEDI数据的表示。我们使用不同的阈值训练多个模型，并为较高的树木分配不同的权重，以提高研究区域不同冠层高度组之间的准确性。我们的最佳模型在测试集中实现了7.52 m的均方根误差（RMSE）和5.71 m的平均绝对误差（MAE），显著优于该地形复杂地区现有的全球冠层高度模型。我们进一步验证了我们的预测与实地调查样地的关系，对于至少包含15棵树的样地，其决定系数（R2）为0.49。生成的树冠高度图，被命名为西喜马拉雅树冠高度图（哪个地图），是公开的。

{"title":"Canopy height mapping in the Western Himalayas, Pakistan: A deep learning approach using GEDI and Sentinel-2 fusion","authors":"Adeel Ahmad , Srikumar Sastry , Aayush Dhakal , Subash Khanal , Alex Levering , Hammad Gilani , Nathan Jacobs","doi":"10.1016/j.jag.2025.105030","DOIUrl":"10.1016/j.jag.2025.105030","url":null,"abstract":"<div><div>The western Himalayas in Pakistan, characterized by a diverse range of conifer species at higher elevations, represent a critical biodiversity hotspot and habitat for numerous species. Accurate spatial assessments of canopy height are essential for improving estimates of aboveground biomass, carbon sequestration, and associated forest ecosystem services in this region. In this study, we estimated canopy heights in the western Himalayas in Pakistan using a U-Net CNN approach, fusing data from the Global Ecosystem Dynamics Investigation Mission (GEDI) with multi-band Sentinel-2 (S2) imagery. We produced a canopy height map at a 10 m resolution for 2020. To ensure accurate measurements across various canopy height groups, we implemented a stratified training approach that optimized the representation of GEDI data throughout the training, validation, and testing phases. We trained multiple models using varying thresholds and assigned different weights to taller trees to improve accuracy between different canopy height groups in the study region. Our best model achieved a Root Mean Square Error (RMSE) of 7.52 m and a Mean Absolute Error (MAE) of 5.71 m in the test set, significantly outperforming existing global canopy height models in this topographically complex region. We further validate our predictions against field inventory plots, achieving a coefficient of determination (R<span><math><msup><mrow></mrow><mrow><mn>2</mn></mrow></msup></math></span>) of 0.49 for plots containing at least 15 trees. The resulting tree canopy height map, designated as the Western Himalaya Canopy Height Map (WHiCH Map), is publicly available.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"146 ","pages":"Article 105030"},"PeriodicalIF":8.6,"publicationDate":"2026-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145925513","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

High spatial resolution GLASS FAPAR (version 2) product from Landsat imagery: Algorithm development using a knowledge transfer strategy 基于Landsat图像的高空间分辨率GLASS FAPAR（版本2）产品：使用知识转移策略的算法开发

IF 8.6 Q1 REMOTE SENSING

International journal of applied earth observation and geoinformation : ITC journal

Pub Date : 2026-01-06 DOI: 10.1016/j.jag.2025.105051

Yuting Qiao , Huaan Jin , Tao He , Shunlin Liang , Feng Tian , Wei Zhao , Zhouyang Liu

The fraction of absorbed photosynthetically active radiation (FAPAR) is a critical parameter for measuring the vegetation photosynthetic capacity. The Hi-resolution Global LAnd Surface Satellite (Hi-GLASS) FAPAR product (version 1, V1) from Landsat imagery has been successfully applied to the ecosystem productivity modeling; however, this product algorithm still exhibits some limitations, including the poor adaptability to heterogeneous surfaces and limited physical interpretability, due to the absence of real-world knowledge guidance. To address these issues, we integrated deep transfer learning and radiative transfer models to update the Hi-GLASS FAPAR algorithm and generate the corresponding product (i.e., version 2, V2). A long short-term memory (LSTM) model was pre-trained on Soil-Leaf-Canopy (SLC) simulations and then optimized using physical knowledge-guided transfer learning, which was used to generate the new FAPAR product from Landsat image series. Validation results demonstrated that the Hi-GLASS FAPAR V2 (R² = 0. 95, RMSE = 0.08) significantly outperformed V1 (R² = 0.94, RMSE = 0.11), with notable improvements in various vegetation categories and sensors. The greatest improvement of FAPAR was found over multiple forest types, where different forest categories showed substantial gains, with R² increasing by 2 % − 11 % and RMSE decreasing by 15 % − 55 %, confirming the improved adaptability of our proposed method to heterogeneous canopies. Moreover, the Hi-GLASS V2 product preserved better spatial details than MODIS、GLASS、GEOV2 products, and its temporal dynamics were more closely aligned with field measurements than the V1 product. These advancements highlight the potential of Hi-GLASS FAPAR V2 as valuable data for supporting terrestrial ecosystem studies.

光合有效辐射吸收分数（FAPAR）是衡量植被光合能力的一个重要参数。基于Landsat影像的高分辨率全球陆地表面卫星（Hi-GLASS） FAPAR产品（版本1，V1）已成功应用于生态系统生产力模型；然而，由于缺乏现实世界的知识指导，该乘积算法仍然存在一定的局限性，包括对异构表面的适应性差和物理可解释性有限。为了解决这些问题，我们整合了深度迁移学习和辐射迁移模型来更新Hi-GLASS FAPAR算法并生成相应的产品（即版本2，V2）。在土壤-叶-冠（Soil-Leaf-Canopy， SLC）模拟中对长短期记忆（LSTM）模型进行预训练，并利用物理知识引导迁移学习对其进行优化，利用Landsat影像序列生成新的FAPAR产品。验证结果表明，Hi-GLASS FAPAR V2 (R2 = 0。95, RMSE = 0.08)显著优于V1 (R2 = 0.94, RMSE = 0.11)，在各种植被类别和传感器上均有显著改善。FAPAR在多种森林类型上的改善最大，其中不同森林类型均有显著提高，R2增加2% ~ 11%，RMSE降低15% ~ 55%，证实了我们提出的方法对异质林冠的适应性提高。与MODIS、GLASS、GEOV2产品相比，Hi-GLASS V2产品保留了更好的空间细节，其时间动态比V1产品与现场测量更接近。这些进展突出了Hi-GLASS FAPAR V2作为支持陆地生态系统研究的宝贵数据的潜力。

{"title":"High spatial resolution GLASS FAPAR (version 2) product from Landsat imagery: Algorithm development using a knowledge transfer strategy","authors":"Yuting Qiao , Huaan Jin , Tao He , Shunlin Liang , Feng Tian , Wei Zhao , Zhouyang Liu","doi":"10.1016/j.jag.2025.105051","DOIUrl":"10.1016/j.jag.2025.105051","url":null,"abstract":"<div><div>The fraction of absorbed photosynthetically active radiation (FAPAR) is a critical parameter for measuring the vegetation photosynthetic capacity. The Hi-resolution Global LAnd Surface Satellite (Hi-GLASS) FAPAR product (version 1, V1) from Landsat imagery has been successfully applied to the ecosystem productivity modeling; however, this product algorithm still exhibits some limitations, including the poor adaptability to heterogeneous surfaces and limited physical interpretability, due to the absence of real-world knowledge guidance. To address these issues, we integrated deep transfer learning and radiative transfer models to update the Hi-GLASS FAPAR algorithm and generate the corresponding product (i.e., version 2, V2). A long short-term memory (LSTM) model was pre-trained on Soil-Leaf-Canopy (SLC) simulations and then optimized using physical knowledge-guided transfer learning, which was used to generate the new FAPAR product from Landsat image series. Validation results demonstrated that the Hi-GLASS FAPAR V2 (R<sup>2</sup> = 0. 95, RMSE = 0.08) significantly outperformed V1 (R<sup>2</sup> = 0.94, RMSE = 0.11), with notable improvements in various vegetation categories and sensors. The greatest improvement of FAPAR was found over multiple forest types, where different forest categories showed substantial gains, with R<sup>2</sup> increasing by 2 % − 11 % and RMSE decreasing by 15 % − 55 %, confirming the improved adaptability of our proposed method to heterogeneous canopies. Moreover, the Hi-GLASS V2 product preserved better spatial details than MODIS、GLASS、GEOV2 products, and its temporal dynamics were more closely aligned with field measurements than the V1 product. These advancements highlight the potential of Hi-GLASS FAPAR V2 as valuable data for supporting terrestrial ecosystem studies.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"146 ","pages":"Article 105051"},"PeriodicalIF":8.6,"publicationDate":"2026-01-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145925519","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0