International journal of applied earth observation and geoinformation : ITC journal最新文献

{"title":"Lightweight segmentation of rock-mass cracks based on anisotropic feature modeling for geological hazard monitoring","authors":"Qin Bao,&nbsp;Yong Yang,&nbsp;Yongyi Xiang,&nbsp;Heng Zhang","doi":"10.1016/j.jag.2026.105540","DOIUrl":"10.1016/j.jag.2026.105540","url":null,"abstract":"<div><div>Accurate segmentation of rock-mass cracks is essential for geological hazard investigation, yet remains challenging due to the slender morphology, strong anisotropy, and discontinuous distribution of cracks in complex field environments. To address these challenges, we propose a lightweight anisotropic feature modeling network for rock-mass crack segmentation, termed AFMNet. The proposed network enhances orientation-dependent crack representations through an anisotropic feature aggregation module and a multi-orientation perception module in the encoder to effectively capture directional and geometric characteristics of slender cracks. In the decoder, cross-scale attention and axis-gated attention are employed to progressively fuse multi-level features and reinforce structural continuity during feature reconstruction. Furthermore, a dual-task complementary output head is introduced to jointly optimize crack region segmentation and edge detail prediction, enabling boundary information to guide the final segmentation results. Experimental results on the CrackSeg9k dataset and a self-constructed geological hazard crack dataset (GHCrack) show that AFMNet achieves mIoU scores of 82.64% and 82.38%, respectively, while maintaining a compact model size of only 0.59 million parameters, demonstrating a favorable balance between segmentation accuracy and computational efficiency.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"153 ","pages":"Article 105540"},"PeriodicalIF":8.2,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148933342","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"DecAttNet: A decoder-centric multi-task attention network for rapid, large-scale land use change detection and 2 M resolution products generation","authors":"Jun Sun,&nbsp;Weiguo Jiang,&nbsp;Haijiang Luo,&nbsp;Ziyan Ling,&nbsp;Shuhui Lai,&nbsp;Xiang Long","doi":"10.1016/j.jag.2026.105568","DOIUrl":"10.1016/j.jag.2026.105568","url":null,"abstract":"<div><div>To address the issues of insufficient resolution and a lack of fine-grained categories in traditional land use and land cover (LULC) products, this study integrates lightweight deep learning, a decoder-centric multi-task learning architecture, and a significant change attention mechanism to propose a deep learning framework (Decoder-Centric Attention Network, DecAttNet) for large-scale, high-precision land use change mapping. By mitigating cross‑task feature conflicts within the multitask‑learning framework via its significant‑change attention module, DecAttNet can adaptively focus on key change areas while sharing features and improving computational efficiency. Additionally, this study utilized DecAttNet to generate a 2023 land‑use product for North China (ten provinces, municipalities and autonomous regions) (LULC_NorthChina_2023) comprising 28 fine‑grained land‑cover classes based on 2‑meter high‑resolution imagery. The results of this study demonstrate that (1) DecAttNet significantly outperforms other deep learning models in terms of both accuracy and computational efficiency, with an average overall accuracy of 90.28 %, which is higher than the figures achieved by the FCNN model (83.12 %), the ChangeMask model (87.73 %), the BGSNet model (86.63 %), and the TripleS model (88.24 %), while improving computational efficiency by approximately 44 % compared to the FCNN model. (2) Based on an independent field validation dataset, the DecAttNet model achieved high extraction precision (precision = 94.86 %) for changed land cover features. (3) The LULC_NorthChina_2023 dataset constructed in this study, which is based on the DecAttNet model, combines high precision (precision = 91.85 %) with high spatial detail. (4) The DecAttNet model demonstrates significant temporal generalization capabilities, achieving average classification accuracies of 89.23 % and 88.78 % in selected sample test regions in 2024 and 2025, respectively, with interannual variability of less than 1.5 %, and exhibits high stability across different geographic re. The DecAttNet framework provides an efficient and accurate solution for large-scale, high-resolution land use dynamic monitoring, and the resulting dataset serves as a valuable benchmark for fine-scale environmental research.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"153 ","pages":"Article 105568"},"PeriodicalIF":8.2,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148936099","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Corrigendum to “A partially supervised joint segmentation method for farmland and rural roads from single-class annotated remote sensing datasets” [Int. J. Appl. Earth Obs. Geoinf. 152 (2026) 105509]","authors":"Zhaoxiang Cao,&nbsp;Yuchun Huang,&nbsp;Yibo Zhou,&nbsp;Peng Zhou,&nbsp;Zhiqiang Qin,&nbsp;Shengyuan Zhang","doi":"10.1016/j.jag.2026.105538","DOIUrl":"10.1016/j.jag.2026.105538","url":null,"abstract":"","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"153 ","pages":"Article 105538"},"PeriodicalIF":8.2,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148884191","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"SegRail: A Framework for semantic segmentation of long-distance railway MLS data via global priors and local geometry integration","authors":"Yue Zhang,&nbsp;Tao Xu,&nbsp;Wen Xiao,&nbsp;Liang Fei,&nbsp;Jiawei Wang,&nbsp;Yuanwei Yang,&nbsp;Lunche Wang","doi":"10.1016/j.jag.2026.105560","DOIUrl":"10.1016/j.jag.2026.105560","url":null,"abstract":"<div><div>Mobile Laser Scanning (MLS) enables accurate and efficient data acquisition in long-distance railway environments, yet directly processing the resulting large-scale point clouds imposes a prohibitive computational burden. Conventional rigid discretization methods induce artificial boundary artifacts and disrupt the intrinsic topology of continuous assets, compromising the structural integrity required for reliable engineering monitoring. To address these challenges, we propose SegRail, a dedicated framework for semantic segmentation of long-distance railway MLS data that explicitly integrates global priors and local geometry. The framework first introduces a sectioning strategy that converts continuous railway scenes into structured segments, reducing computational complexity while preserving geometric coherence. Based on these segments, we design a semantic segmentation network that jointly models global semantic priors and local geometric structures. Specifically, a global prior extraction module is introduced to capture long-range contextual dependencies and category distribution priors, while a specialized neighborhood operator encodes fine-grained local geometric relationships to preserve structural boundaries. Extensive experiments on five railway datasets demonstrate the effectiveness of SegRail, which achieves the highest mIoU than the compared baseline methods under multiple runs. Specifically, it achieves mIoUs of 84.37%, 77.36%, 75.35%, 83.55%, and 92.60% on WhuRailway3D-Urban, WhuRailway3D-Rural, WhuRailway3D-Plateau, SNCF, and ExpressRail, respectively.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"153 ","pages":"Article 105560"},"PeriodicalIF":8.2,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148933345","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Monitoring forest growth dynamics from harmonized multi-source time series of canopy height","authors":"Hugo de Lame,&nbsp;Thibauld Collet,&nbsp;Pauline Depoortere,&nbsp;Alexia Favaro,&nbsp;Philippe Lejeune,&nbsp;Adrien Michez,&nbsp;Antoine Plumacker,&nbsp;Jérôme Perin,&nbsp;Thomas Pollet,&nbsp;Quentin Ponette,&nbsp;Zoé Rousseau,&nbsp;Arthur Vander Linden,&nbsp;Marie-Pierre Tasseroul,&nbsp;Christian Messier,&nbsp;Jean-François Bastin","doi":"10.1016/j.jag.2026.105570","DOIUrl":"10.1016/j.jag.2026.105570","url":null,"abstract":"<div><div>Most remote sensing assessments of forest change focus on canopy cover, whereas the dynamics of vertical growth remain less investigated despite their sensitivity to external forcings. Yet, the increasing availability of canopy height products from multiple platforms facilitates spatially explicit characterization of vertical growth dynamics, provided that systematic measurement biases between sources are properly addressed. Here, we present a transferable workflow to map vertical growth dynamics from mixed-source canopy height model (CHM) time series, harmonized against forest inventory plots at ∼ 1000 m² resolution. We estimated vertical growth to model reference trajectories as a function of initial canopy height. Deviations from these trajectories were quantified, providing contextualized growth information. This approach was illustrated in the Belgian Ardenne using multi-source CHMs derived from recent aerial imagery and lidar data (2006–2021). Across all acquisitions, harmonization reduced systematic bias in dominant height estimates from 2.53 m to 0.01 m (RMSE = 1.79 m, R2 = 0.92), with canopy structure rather than acquisition parameters as the primary driver of bias. Subsequent growth decreased as initial heights increased, with earlier declines in broadleaf than coniferous stands. Spatially clustered growth deviations (Moran’s I = 0.31–0.46, p &lt; 0.001) suggested systematic variation in performance among otherwise similar forests. By leveraging heterogeneous sources of canopy height, this workflow enables spatially exhaustive characterization of forest growth, adding structural understanding into multi-scale forest monitoring and resilience assessments.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"153 ","pages":"Article 105570"},"PeriodicalIF":8.2,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148933347","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A multi-scale spatial-conditioned spectral routing network for hyperspectral and multispectral image fusion","authors":"Bo Zhou,&nbsp;Xianfeng Zhang,&nbsp;Ziyuan Feng,&nbsp;Miao Ren,&nbsp;Xiaobo Zhi","doi":"10.1016/j.jag.2026.105547","DOIUrl":"10.1016/j.jag.2026.105547","url":null,"abstract":"<div><div>Hyperspectral and multispectral image fusion (HMIF) aims to reconstruct high spatial resolution hyperspectral images by jointly leveraging the complementary spectral information from hyperspectral imagery and the spatial details from high spatial resolution multispectral imagery. Existing deep learning-based methods typically improve fusion performance by enlarging the receptive field to capture richer spatial–spectral contextual information, which inevitably increase computational complexity and may introduce redundant spatial–spectral interactions. To address these issues, this study proposes a new multiscale Spatial-Conditioned Spectral Routing Network (SCSRNet). Specifically, a Dynamic Fusion Gate (DFG) module is first developed to adaptively regulate the contributions of hyperspectral and multispectral features based on local spatial characteristics. Subsequently, a Large-Kernel Spatial–Spectral Residual Block (LK-SSRB) is designed, in which a Multi-Scale Spatial Perception Module (MSPM) captures spatial contextual information with large receptive fields, while a Spatial-Conditioned Spectral Routing Module (SCSRM) is innovatively designed to dynamically establish spectral dependencies guided by spatial priors. Additionally, a Residual Scaling Connection (RSC) is introduced to improve optimization stability and feature propagation efficiency. Extensive experiments conducted on three public datasets (i.e. Pavia Center, Chikusei, and MDAS) and a self-collected UAV-based Daxing dataset have demonstrated that the proposed method consistently outperforms seven state-of-the-art approaches in both quantitative metrics and visual quality. The performance gains of the proposed SCSRNet observed on the MDAS dataset achieves improvements of 6.0%, 9.8%, 0.7%, 9.7%, and 7.2% over the second-best comparative method. Furthermore, SCSRNet achieves superior reconstruction performance with relatively low computational complexity, underscoring its effectiveness and practicality for HMIF.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"153 ","pages":"Article 105547"},"PeriodicalIF":8.2,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148884168","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Optimal soil dielectric model map for global soil moisture retrieval from SMAP","authors":"Yulin Shangguan,&nbsp;Jian Peng,&nbsp;Cheng Tong,&nbsp;Xingyu Xue,&nbsp;Xiaojun Xu,&nbsp;Xiaodong Deng,&nbsp;Xiaoxiao Min,&nbsp;Zhou Shi,&nbsp;Jean-Pierre Wigneron","doi":"10.1016/j.jag.2026.105543","DOIUrl":"10.1016/j.jag.2026.105543","url":null,"abstract":"<div><div>Accurate characterization of soil complex dielectric permittivity is essential for passive microwave soil moisture (SM) estimation, but the uncertainty introduced by dielectric model choice remains insufficiently characterized. Moreover, the impact of soil organic matter (SOM) on SM retrieval has not been systematically assessed at the global scale. Therefore, this study retrieved SM using five mineral-soil and two organic-soil dielectric models, evaluated their performances and quantified how their structural differences affect SM retrieval uncertainties based on L-band Soil Moisture Active Passive (SMAP) data. We revealed a substantial inter-model divergence of retrieved SM among dielectric models with a global mean RMSE of 0.035 m³/m³ and this discrepancy was most pronounced in forests and northern high-latitude regions. We further demonstrated that soil dielectric model choice accounted for 24% of the overall SM retrieval error, and this contribution greatly increased under high SOM conditions. Validation using in-site measurements showed that organic soil-based models generally outperformed mineral soil-based models with significantly higher R values. The SOM-aware Mironov 2019 model had the highest correlation (mean R = 0.67), but also showed the largest absolute error primarily due to its overestimation of SOM effect. Evaluation using spatial representative sites yielded similar results regarding the relative performances of dielectric models. Using the triple collocation analysis, we demonstrated that organic soil-based models were the optimal choice over 36.51% of global regions, showing both higher correlation and lower random error. We then derived a global map of optimal soil dielectric models, which could further improve SM retrieval performance relative to SMAP SCA SM product across 76.73% of global areas. Our results underscore the critical role of soil dielectric model selection in L-band SM retrieval and provide new insights for developing SM retrieval algorithms and improving global SM products.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"153 ","pages":"Article 105543"},"PeriodicalIF":8.2,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148852860","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Global, Gap-Free daily Ocean-Colour dataset (2000–2024) for robust marine ecological monitoring","authors":"Yuan Zhang,&nbsp;Hui Yang,&nbsp;Fang Shen,&nbsp;Tingwei Cui,&nbsp;Liu Cui,&nbsp;Renhu Li,&nbsp;Qingzhou Lv,&nbsp;Xuerong Sun","doi":"10.1016/j.jag.2026.105532","DOIUrl":"10.1016/j.jag.2026.105532","url":null,"abstract":"<div><div>Satellite ocean-colour data are crucial for monitoring marine ecosystems, yet data gaps caused by limitations such as cloud cover compromise the reliability of derived ecological metrics, particularly phytoplankton phenology indicators. To enable robust marine ecological monitoring, we reconstruct a gap-free global daily ocean-colour data cube at 9 km resolution for 2000–2024 using a GPU-accelerated Discrete Cosine Transform–Penalized Least Squares (DCT–PLS) framework applied to OC-CCI (Ocean Colour-Climate Change Initiative) dataset. The product provides continuous fields of chlorophyll-a (Chl-a), diffuse<!--> <!-->attenuation coefficients <em>K_d</em>(490)<!--> <!-->, and six-band remote-sensing reflectance (<em>R_rs</em>). Validation against in situ Chl-a data shows high consistency (<em>R</em>2 = 0.75), and comparison with NOAA’s DINEOF Level-4 product demonstrates good agreement. The missing-rate sensitivity experiments indicate that the proposed reconstruction framework maintains generally reliable performance when the final total missing rate is below approximately 80–85 %. Case studies of the Hunga Tonga–Hunga Haʻapai eruption and Hurricane Lorenzo confirm that the reconstruction preserves the spatiotemporal structure of bio-optical anomalies while substantially enhancing data continuity. Using the gap-free Chl-a record, we generated a global climatology of phytoplankton phenology metrics, and quantified the impact of realistic satellite sampling patterns (simulated using MODIS-Aqua masks) on these phenology metrics. Monte Carlo reference–gap analysis reveals that data gaps cause small errors in mean bloom Chl-a (typically &lt; 5 %), but substantially larger uncertainties in phenological timing: root-mean-square errors for bloom initiation, peak date, and duration often reach 10–30 days or more. Uncertainties peak in frequently clouded and dynamic regions, such as the eastern boundary upwelling systems. These findings establish gap-filling as essential for robust phenology analysis and long-term trend detection from satellite ocean-colour data. The gap-free dataset and associated uncertainty benchmarks provide a foundation for advancing studies of marine ecosystem variability and climate-driven change. The reconstructed dataset and processing code are publicly available on <span><span>https://doi.org/10.5281/zenodo.18190723</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"153 ","pages":"Article 105532"},"PeriodicalIF":8.2,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148852862","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Satellite tracking of long-term spatiotemporal dynamics in China’s largest seagrass meadow","authors":"Jun Tang,&nbsp;Yingcheng Lu,&nbsp;Zhiyuan Ouyang,&nbsp;Yarong Zou,&nbsp;Futao Wang,&nbsp;Yuntao Wang,&nbsp;Chao Zhou,&nbsp;Yanlong Chen","doi":"10.1016/j.jag.2026.105536","DOIUrl":"10.1016/j.jag.2026.105536","url":null,"abstract":"<div><div>Seagrass meadows are vital coastal ecosystems that provide essential ecological functions but are highly vulnerable to human disturbances and environmental change. Effective monitoring is therefore critical for understanding their long-term dynamics and supporting conservation efforts. Satellite remote sensing provides an efficient approach for large-scale and long-term observation of seagrass ecosystems; however, accurate monitoring remains challenging due to tidal variability, seasonal growth dynamics, and complex underwater environments. This study focuses on the Caofeidian (CFD) seagrass meadow, the largest continuous seagrass habitat in China. Using Sentinel-2 Multispectral Instrument (MSI) imagery, we quantified its spatiotemporal dynamics from 2016 to 2024. A standardized workflow integrating pixel-level classification, image composition, and trend analysis was developed to characterize long-term variation patterns and identify areas associated with degradation, stability, and recovery. The results reveal a persistent net decline in seagrass extent over the past decade, despite localized recovery observed during the period of restoration activities after 2019. Five variation classes were identified: Significant Degradation (SD), Slight Degradation (SID), Stable (ST), Slight Recovery (SIR), and Significant Recovery (SR), accounting for 5.2 %, 15.5 %, 55.8 %, 22.9 %, and 0.6 % of the meadow area, respectively. Severe losses in several localized areas showed strong spatial and temporal associations with sustained human activities, including land reclamation, clam harvesting, and waterway development, whereas recovery patterns were mainly observed in regions with targeted restoration backgrounds. This study provides the first region-wide assessment of long-term seagrass spatiotemporal dynamics in CFD and demonstrates the potential of multi-temporal satellite observations for monitoring coastal seagrass ecosystems and supporting adaptive conservation management.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"153 ","pages":"Article 105536"},"PeriodicalIF":8.2,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148852883","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A spatiotemporal framework for identifying urban flood priority areas in megacities using social media data","authors":"Xi Huang,&nbsp;Guicheng Yang,&nbsp;Yinyin Chen,&nbsp;Luo Liu,&nbsp;Jiufeng Li,&nbsp;Xianzhe Tang","doi":"10.1016/j.jag.2026.105562","DOIUrl":"10.1016/j.jag.2026.105562","url":null,"abstract":"<div><div>Urban flood (UF) poses major challenges to sustainable development and public safety. Geographic information-based studies commonly apply machine learning (ML) with UF samples and multi-domain factors to assess urban flood susceptibility (UFS) and identify priority regions for UF management. However, most existing approaches rely solely on the latest UFS map, overlooking the dynamic nature of UF and thereby reducing the reliability of the identified priority areas. To address this limitation, we propose an ML-based framework for spatiotemporal UFS estimation to delineate more accurate and practical UF priority regions. The framework integrates: <strong>(1)</strong> Natural Language Processing to extract flood locations from social media (<strong>Weibo)</strong>, and <strong>(2)</strong> a Spatiotemporal Susceptibility Extraction (SSE) method that combines the spatial and temporal characteristics of UFS. Guangzhou, a megacity in China with frequent UF events and extensive Weibo usage, is employed as the case study. We found: <strong>(1)</strong> incorporating UF locations extracted from social media effectively complements official records, as evidenced by both the increased information content of the positive class across multiple flood-conditioning factors and the improved classifier performance and robustness; and <strong>(2)</strong> compared with priority areas directly delineated from the latest-year UFS map, the SSE-derived UF priority areas show higher accuracy when evaluated against ground-truth UF data, further supported by high-resolution remote sensing imagery, while also exhibiting more spatially concentrated and continuous patterns. The proposed framework enables rapid and reliable identification of UF priority areas and provides practical decision support for UF prevention and management in rapidly urbanizing megacities under resource constraints.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"153 ","pages":"Article 105562"},"PeriodicalIF":8.2,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148895349","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}