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

{"title":"Intelligent detection and 3D visualization for cracks in concrete lining panels of long-distance water conveyance channels: A case study on crack detection in deep-cut sections of the middle route of China ’s South-to-North water diversion project","authors":"Qingfeng Hu ,&nbsp;Weiqiang Lu ,&nbsp;Wenkai Liu ,&nbsp;Ximin Cui ,&nbsp;Ruimin Feng ,&nbsp;Shoukai Chen ,&nbsp;Weibo Yin ,&nbsp;Xidong Chen ,&nbsp;Zirui Zhang ,&nbsp;Zilin Liu ,&nbsp;Peipei He ,&nbsp;Kaifeng Ma ,&nbsp;Dantong Zhu ,&nbsp;Peng Wang ,&nbsp;Tangjing Ma ,&nbsp;Shiming Li ,&nbsp;Jinping Liu ,&nbsp;Qifan Wu ,&nbsp;Hui Zhang","doi":"10.1016/j.jag.2026.105308","DOIUrl":"10.1016/j.jag.2026.105308","url":null,"abstract":"<div><div>This study addresses the challenge of intelligent detection and spatial localization of massive cracks in deep-cut canal sections of the Middle Route of China’s South-to-North Water Diversion Project. An integrated framework for crack detection, localization, and three-dimensional visualization of concrete canal linings is proposed. UAV-based imitation ground photogrammetry was first employed to acquire 114,220 high-resolution images, from which a photogrammetric textured mesh model with a surface resolution of 0.47 <span><math><mrow><mi>c</mi><mi>m</mi><mo>/</mo><mi>p</mi><mi>i</mi><mi>x</mi><mi>e</mi><mi>l</mi></mrow></math></span> was constructed. Based on 2886 representative images, a dedicated training dataset for intelligent crack detection was established. By integrating photogrammetric collinearity equations with real-time position and orientation system (POS) data from Unmanned aerial vehicle (UAV) imitation ground flights, a single-image crack coordinate calculation model was developed and embedded into the YOLOv7 object detection framework. This integration enables direct computation of crack spatial coordinates from a single image without reliance on stereo image pairs, allowing crack identification and spatial localization to be synchronously achieved within a unified deep learning framework. Experimental results show that YOLOv7 achieves an <span><math><mrow><mi>m</mi><mi>A</mi><mi>P</mi><mi>@</mi><mn>0</mn><mo>.</mo><mn>5</mn></mrow></math></span> of 84.3% at an <span><math><mrow><mi>I</mi><mi>o</mi><mi>U</mi></mrow></math></span> threshold of 0.5, and the planar localization accuracy is better than 0.1 <span><math><mi>m</mi></math></span>. Finally, the detected and localized cracks are mapped onto the millimeter-level photogrammetric textured mesh model, enabling intuitive visualization of crack spatial distribution and providing technical support for structural condition assessment and intelligent operation and maintenance of long-distance water conveyance channels.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"149 ","pages":"Article 105308"},"PeriodicalIF":8.6,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147798184","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}

{"title":"Dynamic Arctic wetland mapping: A multi-mission satellite time series approach","authors":"Masoud Mahdianpari ,&nbsp;Jean Granger ,&nbsp;Fariba Mohammadimanesh ,&nbsp;Oliver Sonnentag ,&nbsp;Mohammad Marjani","doi":"10.1016/j.jag.2026.105315","DOIUrl":"10.1016/j.jag.2026.105315","url":null,"abstract":"<div><div>Arctic-boreal wetlands play a central role in global carbon cycles, biodiversity, and hydrological regulation, yet their extent and dynamics remain poorly characterized. Static wetland maps often fail to capture short-term fluctuations in inundation caused by snowmelt, rainfall, river flooding, and permafrost degradation, contributing to an underestimation of wetland area and associated greenhouse gas emissions. This study evaluates a multi-sensor framework under two scenarios: (I) a typical static scenario based on multi-sensor median summer composites from 2021 to 2025, and (II) a dynamic scenario that integrates Sentinel-1 time-series backscatter with Sentinel-2 vegetation indices to represent seasonal and interannual inundation behavior. Results for two 500 × 500 km2 study areas indicate substantial improvements under the dynamic scenario. Overall accuracy reached 95.2% in the Northwest Territories, Canada, and 96.1% in northern Sweden, compared with 91% and 92% for the static products. Gains are most evident for seasonally flooded and vegetated classes, with User's and Producer's accuracies increasing by 3–7% relative to the static scenario. The dynamic scenario also identified that 14–18% of wetland area exhibits seasonal or episodic inundation, information that static methods fail to reveal. In addition to higher accuracy, the dynamic framework provides ecologically meaningful outputs such as maximum inundation extent, inundation frequency, and seasonal hydroperiod maps, offering direct value for methane modelling, biodiversity assessment, and water resource management.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"149 ","pages":"Article 105315"},"PeriodicalIF":8.6,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147798273","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}

{"title":"Operational change detection for geographical information: Overview and challenges","authors":"Nicolas Gonthier","doi":"10.1016/j.jag.2026.105278","DOIUrl":"10.1016/j.jag.2026.105278","url":null,"abstract":"<div><div>The rapid evolution of territories due to climate change and human impact necessitates prompt and effective updates to geospatial databases maintained by National Mapping Agencies. That is why change detection is a rapidly evolving and increasingly important field with crucial applications in environmental monitoring, urban expansion, and disaster management. In this review, we present a comprehensive overview of change detection methods and propose future research directions to guide the community towards robust, interpretable, and operational systems, extending beyond the alert stage. This paper first outlines the fundamental definition of change, emphasizing its multifaceted nature, from temporal to semantic characterization. It categorizes automatic change detection methods into four main families: rule-based, statistical, machine learning, and simulation methods. The strengths, limitations, and applicability of every family are discussed in the context of various input data. Then, key applications for National Mapping Agencies are identified, particularly the optimization of geospatial database updating, change-based phenomena, and dynamics monitoring. Finally, the paper highlights the current challenges for leveraging change detection such as the variability of change definition, the missing of relevant large-scale datasets (for both training and evaluation), the diversity of input data, the unstudied no-change detection and the human in the loop integration.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"149 ","pages":"Article 105278"},"PeriodicalIF":8.6,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147798281","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}

{"title":"Rapid decline of ice thickness across Finnish lakes","authors":"Mohammad Milad Salamattalab ,&nbsp;Roohollah Noori ,&nbsp;Mohsen Shahmohammad ,&nbsp;Mehran Mahdian ,&nbsp;Majid Hosseinzadeh ,&nbsp;Peiman Kianmehr ,&nbsp;Bentolhoda Asl-Rousta ,&nbsp;Senlin Zhu ,&nbsp;Xihui Gu ,&nbsp;Markus Saari ,&nbsp;Mikko Kolehmainen ,&nbsp;Yunlin Zhang ,&nbsp;Sapna Sharma ,&nbsp;Qiuhong Tang ,&nbsp;R. Iestyn Woolway","doi":"10.1016/j.jag.2026.105293","DOIUrl":"10.1016/j.jag.2026.105293","url":null,"abstract":"<div><div>The complex relationship between lake ice thickness (LIT) and its environmental drivers, along with the high cost of <em>in-situ</em> measurements and the limited historic resolution of satellite observations—which also carries considerable uncertainties—makes precise regional-scale LIT modeling particularly challenging. To address this, we propose a black-box modeling approach that incorporates a novel parameter: freezing-days (FDs), derived from widely available air temperature data. Using this method, we project future LIT trends across 40 Finnish lakes. We also determine the historical LIT range to which aquatic species have adapted. Compared to the well-established Stefan model, our modeling approach enhances LIT estimation accuracy, achieving a 35% increase in the coefficient of determination and a 38% reduction in the mean absolute error in the validation phase. Our results indicate rapid declines in LIT, with projected reductions of 32% under SSP4.5 and 44% under SSP8.5 by 2100. Under SSP4.5, LIT variability is expected to exceed historical ranges in northern, central, and southern Finland by 2058, 2076, and 2098, respectively. In both emission scenarios, northern lakes are projected to cross this threshold by the 2050s, suggesting an ecological tipping point, regardless of future warming levels. These findings highlight the vulnerability of lake ice ecosystems to climate change and offer critical insights for developing adaptation strategies in a warming Arctic-boreal region. Our proposed model suggests a more accurate, widely applicable alternative for the study of LIT changes across spatial and temporal scales.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"149 ","pages":"Article 105293"},"PeriodicalIF":8.6,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147850595","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}

{"title":"Aligning legacy NLCD land cover maps based on Landsat Collection 1 to Collection 2","authors":"Congcong Li ,&nbsp;Suming Jin","doi":"10.1016/j.jag.2026.105314","DOIUrl":"10.1016/j.jag.2026.105314","url":null,"abstract":"<div><div>The transition from Landsat Collection 1 to Collection 2 introduced significant improvements in radiometric and geometric accuracy. However, the improvements cause location misalignment between the existing Landsat-derived land cover products and the new collection. The legacy National Land Cover Database (NLCD) has been used as a cornerstone land cover source for a variety of research. Therefore, a method aligning the legacy NLCD product to Collection 2 is required to ensure its continuity and consistency of service. We developed a strategy to not only align legacy NLCD to match new Collection 2 geometric locations but also improve land cover labeling in the region that was affected by the geometric shifts. The method identifies boundary pixels of homogeneous land cover patches as potential problem areas that are likely impacted by geometric shifts and generates candidate labels from 3 × 3 window with the target pixel at the center and segmentation-derived majority label. Standard phenology patterns of each candidate land cover type are established based on the random samples except boundary pixels within a 1000-pixels × 1000-pixels processing window region. The phenological distance to each standard land cover type pattern is calculated through a penalty dynamic time warping (DTW) method for each target pixel in the boundary region. Finally, the method determines the most suitable label based on the phenological distance from the candidate labels. Both visual and accuracy assessment results demonstrate that the alignment preserves the overall land cover patterns in the original legacy NLCD product while reducing the spatial discrepancies between the Landsat Collection 2 and land cover. In addition, it enhances the accuracy of land cover labeling of boundary pixels. The overall accuracy (OA) was increased by 7% in the land cover boundary regions after alignment. The quality and confusion matrix comparison between the alignment results and the original legacy NLCD confirm the reliability of the method. Our alignment method has the potential to serve as a framework for aligning other Landsat-derived land cover products to future collections.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"149 ","pages":"Article 105314"},"PeriodicalIF":8.6,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147798185","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}

{"title":"SDGSAT-1 MSI for monitoring water quality parameters in rivers: A comprehensive evaluation and demonstration","authors":"Zichen Mao ,&nbsp;Shenglei Wang ,&nbsp;Dalin Jiang ,&nbsp;Junsheng Li ,&nbsp;Changyong Dou ,&nbsp;Fangfang Zhang ,&nbsp;Bing Zhang","doi":"10.1016/j.jag.2026.105318","DOIUrl":"10.1016/j.jag.2026.105318","url":null,"abstract":"<div><div>River water quality remains insufficiently monitored at large scales despite its ecological and societal importance. SDGSAT-1, launched in 2021 with a 10-m, seven-band Multispectral Imager (MSI), enables observation of narrow river systems. This study evaluated the capability of SDGSAT-1 MSI for river water quality monitoring and developed a turbidity-adaptive retrieval framework for estimating total suspended matter (TSM) and chlorophyll-a (Chl-a) in optically diverse river environments. Using synchronized in-situ measurements from rivers across China, atmospheric correction schemes were evaluated and optimized. After classifying waters into relatively clean and extremely turbid types using a green-to-red band ratio, Green Line Height and Deep-Blue Line Height model based on visible bands were developed for TSM and Chl-a retrieval, respectively. The TSM retrieval achieved an average unbiased relative error (AURE) of 39.87% across turbidity levels, with an RMSE of 45.06 mg/L. The Chl-a model achieved an AURE of 35.04% in clean waters, with an RMSE of 10.70 mg/m³, but its performance declined in extremely turbid conditions due to sediment masking. This framework was then applied to generate TSM and Chl-a products for seven major rivers in China (Songhua, Hai, Liao, Yellow, Huai, Yangtze and Pearl) using SDGSAT-1 MSI imagery acquired during 2022–2023. Despite limited data coverage, our results revealed reasonable spatial–temporal patterns of water quality and potential links between river turbidity and basin land cover. Overall, this study demonstrates the usability of SDGSAT-1 MSI for dynamic river water quality monitoring and offers guidance for future research and applications.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"149 ","pages":"Article 105318"},"PeriodicalIF":8.6,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147798190","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}

{"title":"Phased, Seasonal, and regional asymmetries in China’s temperature change during 1948–2022: Insights and uncertainties from a Multi-Dataset analysis","authors":"Yao Shen","doi":"10.1016/j.jag.2026.105306","DOIUrl":"10.1016/j.jag.2026.105306","url":null,"abstract":"<div><div>Global climate change manifests heterogeneously across regions, yet reliably characterizing these patterns is contingent upon the datasets employed. As a climatically diverse and populous nation, China serves as a critical testbed for evaluating the consistency of temperature products. This study systematically compares six reanalysis datasets (JRA-55, MERRA-2, NCEP/NCAR, NCEP/DOE, CRA/Land, ERA5-Land) with in-situ observations to assess air temperature (Ta) and land surface temperature (LST) changes across China from 1948 to 2022. While all datasets confirm a mid-20th century cooling to late-century warming transition, the magnitudes show notable dataset-dependent variations. For example, the maximum pre-1980 s cooling rates range from −3.1 to −1.8 K/decade. Seasonal contrast is consistently identified: historical cooling was strongest in hot seasons, while recent warming is most rapid in cold seasons—a pattern aligned with aerosol forcing and snow-albedo feedbacks, though its intensity varies across products. A key finding is the significant inter-dataset discrepancy in regional trends, which directly impacts climate vulnerability interpretation. Estimated hot-season warming in northern China varies from 0.8 to 1.2 K/decade across datasets, compared to more modest rates in southern regions. These differences are attributable to variations in spatial resolution, assimilation schemes, and temperature variable definitions (e.g., ground vs. skin temperature). Crucially, systematic differences emerge between in-situ and reanalysis products: in-situ observations capture a larger regional temperature range (approximately 30 K) compared to reanalysis datasets (approximately 19 K), and exhibit faster cooling and warming rates—approximately 0.5 K/decade higher than the reanalysis average during recent decades—highlighting the tendency of gridded products to underestimate both spatial heterogeneity and the pace of change. Our analysis underscores that conclusions on regional temperature changes are contingent upon dataset choice. The substantial spread among state-of-the-art products highlights a critical uncertainty for impact assessments, underscoring the necessity of multi-dataset ensembles to robustly constrain trends and inform climate adaptation strategies over China.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"149 ","pages":"Article 105306"},"PeriodicalIF":8.6,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147798191","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}

{"title":"Improving SPM retrieval across diverse transitional waters through satellite-derived OWT classification and per-class algorithm optimization","authors":"Junfang Lin ,&nbsp;Elizabeth C. Atwood ,&nbsp;Xiaohan Liu ,&nbsp;Emmanuel Nwokocha ,&nbsp;Thomas Jackson ,&nbsp;Steve Groom","doi":"10.1016/j.jag.2026.105302","DOIUrl":"10.1016/j.jag.2026.105302","url":null,"abstract":"<div><div>Satellite remote sensing provides synoptic coverage and frequent revisit capabilities, making it an essential tool for estimating suspended particulate matter (SPM) in optically complex aquatic environments. Recent advances in satellite-derived optical water type (OWT) classification offer new opportunities for improving SPM retrieval across diverse water bodies, yet a systematic per-OWT recalibration and comparative evaluation of existing SPM algorithms has not been fully assessed. In this study, we employ a comprehensive framework that recalibrates 17 widely used SPM algorithms for each of 11 satellite-derived OWT classes and uses a Round-Robin evaluation to identify the optimal algorithm for each class while minimizing cross-type performance variability. Using Sentinel-2 MultiSpectral Instrument (MSI) reflectance, the resulting OWT-based blending scheme achieves substantially improved accuracy across optically diverse transitional waters, with a Root Mean Square Error (<em>Ψ</em>) of ∼4.6–8.2 g/m³ and a coefficient of determination (<em>r²</em>) of 0.87–0.91 when validated against in-situ observations, including measurements from regions not used for calibration. The framework is further demonstrated through spatial mapping and multi-year trend analysis of SPM in two transitional coastal systems. The results highlight the value of satellite-derived OWTs for enhancing the transferability and robustness of SPM retrievals and provide a scalable approach for water-quality monitoring and ecosystem-assessment applications.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"149 ","pages":"Article 105302"},"PeriodicalIF":8.6,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147798278","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}

{"title":"Integrating aerial optical and bathymetric LiDAR data for coral reef mapping","authors":"Jiangying Qin ,&nbsp;Ming Li ,&nbsp;Armin Gruen ,&nbsp;Deren Li ,&nbsp;Xuan Liao","doi":"10.1016/j.jag.2026.105311","DOIUrl":"10.1016/j.jag.2026.105311","url":null,"abstract":"<div><div>Coral reef ecosystems face severe threats from global warming and human activities in coastal regions, necessitating advanced mapping approaches to monitor the geographic distribution dynamics and support conservation efforts. This study proposes an innovative framework leveraging multi-modal aerial remote sensing data to enhance coral reef recognizing and sustainable management for Tetiaroa Atoll, Society Islands, French Polynesia. By integrating aerial optical imagery with bathymetric LiDAR (Light Detection and Ranging) data, we present a novel method for accurately mapping coral reefs. A depth-informed coral multi-modal semantic segmentation neural network is developed, achieving a notable mean Intersection over Union (mIoU) of 80.4% for coral segmentation by incorporating domain expertise and coral-specific features. Additionally, we explore the synergistic impact of combining aerial optical imagery with bathymetric LiDAR data on segmentation accuracy and evaluate the influence of varying depth information sources. The proposed approach offers a rapid, flexible, and scalable solution for coral reef aerial remote sensing mapping. It enables provide a cost-effective tool for addressing key challenges in precision coastal ecological surveys. This work contributes to advancing coral reef recovery efforts and underscores the value of remote sensing technologies in the sustainable management of marine ecosystems. The relevant code can be found in <span><span>https://github.com/jyqinnn/Depth-informed-coral-segmentation</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"149 ","pages":"Article 105311"},"PeriodicalIF":8.6,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147798280","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}

{"title":"A transferable deep learning framework for flood mapping: Spatial generalization across hydro-climatic regimes using satellite imagery","authors":"Claudie Ratté-Fortin ,&nbsp;Karem Chokmani ,&nbsp;Richard Turcotte","doi":"10.1016/j.jag.2026.105321","DOIUrl":"10.1016/j.jag.2026.105321","url":null,"abstract":"<div><div>Flood hazard assessment has traditionally relied on hydraulic models that require extensive boundary conditions, detailed topographic data, and substantial computational resources. Their reliability, however, is often undermined by the scarcity of in-situ observations during extreme events, which increases uncertainty in flood extent delineation.<!--> <!-->Remote sensing has therefore emerged as a valuable alternative for flood mapping. Yet, most recent deep learning studies have focused on maximizing in-sample accuracy while providing limited evaluation of spatial transferability and operational readiness when applied outside the training domain.</div><div>To address these constraints and move beyond the usual focus on in-sample accuracy in flood mapping, this study introduces an operationally oriented deep learning framework designed for spatial transferability. The approach combines two modality-specific U-Net models, one for Sentinel-1 (SAR) and one for Sentinel-2 (optical), trained on a globally sourced dataset (C2S-MS) of flood events. Beyond conventional model training, we incorporate key steps to support generalization: (i) a<!--> <!-->hydro-climatic and geomorphological characterization of the training dataset, combined with a multivariate similarity analysis<!--> <!-->to assess the environmental representativeness of the target region, (ii) an assessment of structural biases to orient model design, and (iii) a<!--> <!-->deployment-driven validation strategy<!--> <!-->to rigorously evaluate spatial transferability under operational constraints.</div><div>The S1 model achieved an explanatory performance (within the training dataset) of 85 % IoU and 92 % F1-score, while the S2 model reached 89 % and 94 %, respectively. Predictive performance, evaluated on an independent flood event in the target region, remained consistent with these results, with only a limited degradation observed (−2 % in IoU and − 1 % in F1-score). These findings provide empirical support for the role of environmental representativeness in supporting spatial transferability.</div><div>By unifying environmental representativeness assessment, modality-specific deep learning architectures, and rigorous spatial validation, this work advances a transferable and operationally oriented framework for satellite-based flood mapping. More broadly, the results highlight the importance of aligning training data coverage, validation strategy, and deployment context to achieve reliable generalization under domain shift. Beyond this specific application, the proposed methodology provides insights into model design and evaluation for Earth observation studies confronted with spatial extrapolation.</div></div>","PeriodicalId":73423,"journal":{"name":"International journal of applied earth observation and geoinformation : ITC journal","volume":"149 ","pages":"Article 105321"},"PeriodicalIF":8.6,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147850596","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}