Zhenshi Li , Dilxat Muhtar , Feng Gu , Yanglangxing He , Xueliang Zhang , Pengfeng Xiao , Guangjun He , Xiaoxiang Zhu
{"title":"LHRS-Bot-Nova: Improved multimodal large language model for remote sensing vision-language interpretation","authors":"Zhenshi Li , Dilxat Muhtar , Feng Gu , Yanglangxing He , Xueliang Zhang , Pengfeng Xiao , Guangjun He , Xiaoxiang Zhu","doi":"10.1016/j.isprsjprs.2025.06.003","DOIUrl":"10.1016/j.isprsjprs.2025.06.003","url":null,"abstract":"<div><div>Automatically and rapidly understanding Earth’s surface is fundamental to our grasp of the living environment and informed decision-making. This underscores the need for a unified system with comprehensive capabilities in analyzing Earth’s surface to address a wide range of human needs. The emergence of multimodal large language models (MLLMs) has great potential in boosting the efficiency and convenience of intelligent Earth observation. These models can engage in human-like conversations, serve as unified platforms for understanding images, follow diverse instructions, and provide insightful feedbacks. In this study, we introduce LHRS-Bot-Nova, an MLLM specialized in understanding remote sensing (RS) images, designed to expertly perform a wide range of RS understanding tasks aligned with human instructions. LHRS-Bot-Nova features an enhanced vision encoder and a novel bridge layer, enabling efficient visual compression and better language-vision alignment. To further enhance RS-oriented vision-language alignment, we propose a large-scale RS image-caption dataset, generated through feature-guided image recaptioning. Additionally, we introduce an instruction dataset specifically designed to improve spatial recognition abilities. Extensive experiments demonstrate superior performance of LHRS-Bot-Nova across various RS image understanding tasks. We also evaluate different MLLM performances in complex RS perception and instruction following using a complicated multi-choice question evaluation benchmark, providing a reliable guide for future model selection and improvement. Data, code, and models will be available at <span><span>https://github.com/NJU-LHRS/LHRS-Bot</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50269,"journal":{"name":"ISPRS Journal of Photogrammetry and Remote Sensing","volume":"227 ","pages":"Pages 539-550"},"PeriodicalIF":10.6,"publicationDate":"2025-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144518809","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"A cross-spatiotemporal weakly supervised framework for land cover classification: Generating temporally and spatially consistent land cover maps","authors":"Junqi Zhao , Zhanliang Yuan , Xiaofei Mi , Jian Yang , Xueke Chen , Xianhong Meng , Hongbo Zhu , Yuke Meng , Zhenzhao Jiang , Zhouwei Zhang","doi":"10.1016/j.isprsjprs.2025.06.005","DOIUrl":"10.1016/j.isprsjprs.2025.06.005","url":null,"abstract":"<div><div>High-resolution land cover mapping tasks guided by publicly available decameter-level land cover products often suffer from label inaccuracies caused by land cover changes and scale discrepancies resulting from spatiotemporal resolution inconsistencies. To address this issue, this study proposes a cross-spatiotemporal weakly supervised dual-stage classification framework (CTS-WS) that implements temporal and spatial correction strategies to rectify erroneous labels and scale differences, achieving spatiotemporal consistent high-resolution land cover mapping. In the cross-temporal stage, we establish an NDVI screening and uncertainty noise correction mechanism by leveraging the spectral characteristics of high-resolution imagery and the feature fitting capability of convolutional neural networks, effectively eliminating pixels with spectral feature mismatches. The cross-spatial stage proposes a dual-branch parallel network integrating spatial and spectral features, which combines a periodic label screening module with boundary metric loss to learn fine-grained spatial features and refine boundaries. To validate the effectiveness of the proposed method, this study constructed the GF1-CTS dataset by integrating Gaofen-1 satellite imagery with ESA-GLC10 product, and conducted parallel experiments on both the GF1-CTS dataset and a large-scale Chesapeake Bay watershed dataset. Experimental results demonstrate that CTS-WS successfully achieves cross-spatiotemporal resolution land cover mapping from 10 m to 2 m and from 30 m to 1 m, outperforming various mainstream methods and state-of-the-art technologies. This study provides a novel solution for high-resolution remote sensing image land cover mapping across spatiotemporal resolutions.</div></div>","PeriodicalId":50269,"journal":{"name":"ISPRS Journal of Photogrammetry and Remote Sensing","volume":"227 ","pages":"Pages 519-538"},"PeriodicalIF":10.6,"publicationDate":"2025-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144518808","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Zhan Zhang, Daoyu Shu, Cunyi Liao, Chengzhi Liu, Yuanxin Zhao, Ru Wang, Xiao Huang, Mi Zhang, Jianya Gong
{"title":"FlexiSAM: A flexible SAM-based semantic segmentation model for land cover classification using high-resolution multimodal remote sensing imagery","authors":"Zhan Zhang, Daoyu Shu, Cunyi Liao, Chengzhi Liu, Yuanxin Zhao, Ru Wang, Xiao Huang, Mi Zhang, Jianya Gong","doi":"10.1016/j.isprsjprs.2025.05.028","DOIUrl":"https://doi.org/10.1016/j.isprsjprs.2025.05.028","url":null,"abstract":"Fine-grained land use and land cover (LULC) classification using high-resolution remote sensing (RS) imagery is fundamental to scientific research. Recently, the Segment Anything Model (SAM) has emerged as a major advance in deep learning-based LULC classification due to its robust segmentation and generalization capabilities. However, existing SAM-based models predominantly rely on single-modal inputs (e.g., optical RGB or SAR), limiting their ability to fully capture the complex spatial and spectral characteristics of RS imagery. Although multimodal RS data can provide complementary information to enhance classification accuracy, integrating multiple modalities into SAM presents significant challenges, including modality adaptation, semantic interference, and domain gaps. Building on this, we propose FlexiSAM, a SAM-based multimodal semantic segmentation model designed to overcome these challenges. FlexiSAM uses RGB as the primary modality while seamlessly integrating auxiliary RS modalities through a modular pipeline. Key innovations include the Dynamic Multimodal Feature Fusion Unit (DMMFU) and Dynamic Attention and the Context Aggregation Mixer (DACAM) for robust cross-modal feature fusion and refinement, and the Semantic Cross-Modal Integration Module (SCMII) for mitigating modality-induced feature misalignments and ensuring coherent multimodal integration. These are then processed by the adapted SAM encoder, enhanced with a lightweight adapter tailored for RS data, and followed by a dedicated decoder that produces precise classification outputs. Extensive experiments on the Korea, Houston2018, and Mini-FLAIR datasets, conducted using LuoJiaNET for core evaluations and PyTorch for cross-method comparisons, demonstrate FlexiSAM’s effectiveness and superiority, surpassing state-of-the-art models by at least 1.58% on Korea, 0.77% on Houston2018, and 1.14% in mIoU. Importantly, the LuoJiaNET framework delivers higher accuracy and efficiency compared to PyTorch. FlexiSAM also demonstrates strong adaptability and robustness across diverse RS modalities, establishing it as a versatile solution for fine-grained LULC classification.","PeriodicalId":50269,"journal":{"name":"ISPRS Journal of Photogrammetry and Remote Sensing","volume":"24 1","pages":""},"PeriodicalIF":12.7,"publicationDate":"2025-06-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144515941","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Luis Lizcano-Sandoval , Marcus W. Beck , Sheila Scolaro , Edward T. Sherwood , Frank Muller-Karger
{"title":"Cloud-based satellite remote sensing for enhancing seagrass monitoring and ecosystem management","authors":"Luis Lizcano-Sandoval , Marcus W. Beck , Sheila Scolaro , Edward T. Sherwood , Frank Muller-Karger","doi":"10.1016/j.isprsjprs.2025.06.034","DOIUrl":"10.1016/j.isprsjprs.2025.06.034","url":null,"abstract":"<div><div>We examined the feasibility of monitoring interannual and intra-annual changes in seagrass extent in Tampa Bay, Florida between 1987 and 2023 using remote sensing with the Landsat 5, Landsat 7, Landsat 8, and Sentinel-2 satellite sensor series. This study filled gaps and extended the time series developed for the period 1990–2021 by Lizcano-Sandoval et al. (2022). Seagrass extent was evaluated for six Tampa Bay segments: Hillsborough Bay (HB), Old Tampa Bay (OTB), Middle Tampa Bay (MTB), Lower Tampa Bay (LTB), Boca Ciega Bay (BCB), Manatee River + Terra Ceia Bay (MRTC). Results were compared with reference data from the biennial mapping program conducted by the Southwest Florida Water Management District (SWFWMD). Overall, seagrass showed long-term increases in extent in Tampa Bay. The highest increasing rates over 1987–2023 were observed in HB (+5.7 % yr<sup>−1</sup>) and OTB (+3.5 % yr<sup>−1</sup>). Smaller increases were observed in BCB (0.9 % yr<sup>−1</sup>). However, in HB and MRTC seagrass extent showed decreases in particular years between 2015 and 2023 (34 % and 11 %, respectively), and in OTB between 2017 and 2023 (41 %). Intra-annual changes in seagrass extent were observed during 2021–2023. Intra-annual coefficient of variation was estimated to be as low as 3 % in BCB and as high as 60 % in HB. Seagrass extent estimated by remote sensing was highly correlated with the reference data (r > 0.74), except in the MRTC segment (r = -0.29). A Google Earth Engine app was developed to allow public access to the temporal and spatial changes of seagrass extent and distribution in Tampa Bay. The results showed that seagrass extent assessments to complement existing field and airborne seagrass monitoring programs are feasible at low cost with public satellite imagery in the cloud.</div></div>","PeriodicalId":50269,"journal":{"name":"ISPRS Journal of Photogrammetry and Remote Sensing","volume":"227 ","pages":"Pages 508-518"},"PeriodicalIF":10.6,"publicationDate":"2025-06-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144502531","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Yichuan Ma , Shunlin Liang , Wanshan Peng , Tao He , Han Ma , Yongzhe Chen , Wenyuan Li , Jianglei Xu , Shikang Guan
{"title":"A universal physically-based topographic correction framework for high-resolution optical satellite data","authors":"Yichuan Ma , Shunlin Liang , Wanshan Peng , Tao He , Han Ma , Yongzhe Chen , Wenyuan Li , Jianglei Xu , Shikang Guan","doi":"10.1016/j.isprsjprs.2025.05.027","DOIUrl":"10.1016/j.isprsjprs.2025.05.027","url":null,"abstract":"<div><div>Surface reflectance, retrieved via atmospheric correction from top-of-atmosphere (TOA) observations, characterizes the intrinsic properties of the Earth’s surface. Despite its importance, topographic effects are often neglected in surface reflectance retrievals, introducing significant uncertainties, particularly over mountainous regions. Existing topographic correction methods, while numerous, commonly face challenges: physically-based approaches require accurate atmospheric parameters (often unavailable in complex terrains) and are computationally intensive, whereas semi-empirical methods depend on empirical parameters that can lead to overcorrection or inconsistent performance across diverse conditions. Additionally, prior studies have predominantly utilized data from a single satellite platform, limiting the applicability and transferability of topographic correction algorithms across diverse datasets. To overcome these limitations, we introduce a Universal Topographic Correction (UTC) framework, a physically-based approach designed for seamless integration with multiple high-resolution satellite and airborne datasets. The UTC integrates spectral information from extensive radiative transfer simulations with image-derived spatial information to optimize spectral direct irradiance ratios, a key component of physically-based correction, while accounting for shadow effects and digital elevation model (DEM)-induced errors through targeted processing along shadow boundaries. We evaluated UTC’s performance against established methods, including C-correction, SCS + C, and Statistical-Empirical (SE), using a 3D radiative transfer model as a reference across varied topographic and illumination conditions. Results show that UTC consistently outperforms these methods, particularly in shadowed areas, with mean absolute deviations in the near-infrared band of 0.0103 for UTC compared to 0.0179 (C), 0.0362 (SCS + C), and 0.0311 (SE). Testing across Landsat 9 (30 m), Sentinel-2 (20 m), SPOT 4/5 (10–20 m), PlanetScope (3 m), and AVIRIS-3 (∼2.9 m) datasets further demonstrates UTC’s robustness, effectively reducing overcorrection in complex terrains and improving reflectance accuracy in shadowed regions. UTC’s advantages lie in (i) requiring no external atmospheric inputs, and (ii) its physically-informed design based on spectral and spatial information for broad applicability. This study underscores critical limitations in existing topographic correction methods and proposes a robust solution for addressing them. Future research could enhance the UTC framework by integrating atmospheric effects, thereby achieving combined atmospheric and topographic correction from top-of-atmosphere observations.</div></div>","PeriodicalId":50269,"journal":{"name":"ISPRS Journal of Photogrammetry and Remote Sensing","volume":"227 ","pages":"Pages 459-480"},"PeriodicalIF":10.6,"publicationDate":"2025-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144490363","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Kim Calders , Martin Herold , Jennifer Adams , John Armston , Benjamin Brede , Wout Cherlet , Zane T. Cooper , Karun Dayal , Pieter De Frenne , Shaun R. Levick , Patrick Meir , Niall Origo , Cornelius Senf , Luna Soenens , Louise Terryn , Wouter A.J. Van den Broeck , Mikko Vastaranta , Hans Verbeeck , Ludovic Villard , Mathias Disney
{"title":"Realistic virtual forests for understanding forest disturbances and recovery from space","authors":"Kim Calders , Martin Herold , Jennifer Adams , John Armston , Benjamin Brede , Wout Cherlet , Zane T. Cooper , Karun Dayal , Pieter De Frenne , Shaun R. Levick , Patrick Meir , Niall Origo , Cornelius Senf , Luna Soenens , Louise Terryn , Wouter A.J. Van den Broeck , Mikko Vastaranta , Hans Verbeeck , Ludovic Villard , Mathias Disney","doi":"10.1016/j.isprsjprs.2025.06.031","DOIUrl":"10.1016/j.isprsjprs.2025.06.031","url":null,"abstract":"<div><div>Forests worldwide are undergoing large-scale and unprecedented changes in terms of structure and composition due to land use change and natural disturbances. We have some understanding of how disturbances impact forest structure. Still, we lack knowledge of the structural impact at fine spatial and temporal resolution, as well as across large spatial extents. Here, we provide a perspective on new approaches to observe, quantify and understand forest disturbances and recovery from space by using time series of the most detailed 3D virtual forest models that aim to digitise real-life forests fully. These virtual forests are important for enhancing our fundamental understanding of how we observe forest disturbance and recovery monitoring from space. We define virtual forests in the context of this paper as explicit 3D reconstructed models that are parameterised so they can be used and manipulated for radiative transfer modelling. Realistic virtual forests can be created through empirical reconstruction of explicit forest structure measured by terrestrial laser scanning, coupled with radiometric parameterisation. We argue that these realistic virtual forests, capturing the temporal dimension of forest disturbances, combined with physically-based radiative transfer modelling, provide a critical link between detailed in situ observations and large spatial coverage from satellite observations.</div></div>","PeriodicalId":50269,"journal":{"name":"ISPRS Journal of Photogrammetry and Remote Sensing","volume":"227 ","pages":"Pages 501-507"},"PeriodicalIF":10.6,"publicationDate":"2025-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144490821","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Piecewise-ICP: Efficient and robust registration for 4D point clouds in permanent laser scanning","authors":"Yihui Yang, Christoph Holst","doi":"10.1016/j.isprsjprs.2025.06.026","DOIUrl":"10.1016/j.isprsjprs.2025.06.026","url":null,"abstract":"<div><div>The permanent terrestrial laser scanning (PLS) system has significantly improved the temporal and spatial resolution of surface capture in geomonitoring tasks. Accurate registration of the four-dimensional (3D space + time) point clouds (4DPC) generated by PLS is the prerequisite for subsequent deformation analysis. However, due to the massive data volume and potential changes between scans, achieving automatic, efficient, and robust registration of 4DPC remains challenging, especially in scenarios lacking signalized and reliable targets. To address the challenges in target-free registration of 4DPC from PLS, we propose Piecewise-ICP, a robust and efficient fine registration method. Assuming the stable areas on monitored surfaces are locally planar, we employ supervoxel-based segmentation to generate planar patches from 4DPC. These patches are then refined and classified by comparing defined correspondence distances to a monotonically decreasing distance threshold, thus progressively eliminating unstable areas in an iterative process and preventing convergence to local minima. Subsequently, an improved point-to-plane ICP (Iterative Closest Point) is applied to the centroids of identified stable patches. We introduce the Level of Detection to determine the minimum distance threshold, mitigating the influence of outliers and surface changes on registration accuracy. Based on derived transformation uncertainties, we further smooth the transformation sequence using a Kalman filter, yielding more accurate registration parameters. We demonstrate our registration approach on two datasets: (1) Synthetic point cloud time series with predefined changes and transformation parameters, and (2) a real 4DPC dataset from a PLS system installed in the Alpine region for rockfall monitoring. Experimental results show that Piecewise-ICP improves the average registration accuracy by more than 50% compared to the target-based method and existing robust ICP variants such as Trimmed-ICP and Generalized-ICP.</div></div>","PeriodicalId":50269,"journal":{"name":"ISPRS Journal of Photogrammetry and Remote Sensing","volume":"227 ","pages":"Pages 481-500"},"PeriodicalIF":10.6,"publicationDate":"2025-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144490820","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Zhenbang Hao , Lili Lin , Christopher J. Post , Elena A. Mikhailova , Jeffery Allen
{"title":"Mapping dynamics of large-scale high-precision pond datasets using a semi-automated method based on deep learning","authors":"Zhenbang Hao , Lili Lin , Christopher J. Post , Elena A. Mikhailova , Jeffery Allen","doi":"10.1016/j.isprsjprs.2025.06.018","DOIUrl":"10.1016/j.isprsjprs.2025.06.018","url":null,"abstract":"<div><div>With their large numbers and widespread distribution, ponds are crucial in stormwater interception, biodiversity, and freshwater resource conservation. However, due to their small size and shallow depth, ponds are highly susceptible to anthropogenic activities and climate variability, making it necessary to map their numbers, distribution, and change dynamics. Relying only on deep learning (DL) techniques is insufficient to create a pond identification dataset that does not contain errors. This study is the first of its kind that proposes a workflow to identify small ponds (<5 ha) with minimal errors from the National Agricultural Imagery Program (NAIP) high-resolution aerial imagery using the combination of DL and a manual cross-correction approach. Ponds in South Carolina, United States, were detected and delineated for 2017 and 2019 using U-Net models. Next, the detection results from both years were used as reference data for cross-correction, removing false detections and adding omissions to obtain the refined high-precision pond datasets. The pond datasets were compared to the existing public datasets (JRC and NWI) to evaluate the performance of the proposed method. Finally, changes in ponds between two years and the predominant land cover around each pond in 2019 were analyzed in our study. The results showed that the refined high-precision pond dataset containing 70,449 ponds in 2017 and 71,858 ponds in 2019, with an average size of 0.5 ha, fills an important gap in existing pond data. The existing public datasets (JRC and NWI) do not identify 61.72% and 41.03% of the new high-precision pond dataset developed as part of our study in 2019. Based on land cover data, the largest number of ponds were located in forested areas (23,188 ponds, 0.76 ponds/km<sup>2</sup>), followed by wetlands (15,782 ponds, 0.76 ponds/km<sup>2</sup>). In contrast, barren land and hay/pasture had the highest pond density, reaching 2.67 ponds/km<sup>2</sup> and 1.93 ponds/km<sup>2</sup>. A total of 2,979 ponds experienced changes between 2017 and 2019, and 69,664 ponds remained unchanged. The types of pond changes can be categorized as new pond construction, water level changes, and pond disappearance. Our study significantly advances a workflow and method for pond detection that leverages deep learning over large areas in diverse ecological regions and can provide high-precision pond datasets with minimal errors for pond evaluation and management.</div></div>","PeriodicalId":50269,"journal":{"name":"ISPRS Journal of Photogrammetry and Remote Sensing","volume":"227 ","pages":"Pages 438-458"},"PeriodicalIF":10.6,"publicationDate":"2025-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144490097","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Conor McGlinchey , Jesus Torres Palenzuela , Luis Gonzalez-Vilas , Mortimer Werther , Dalin Jiang , Andrew Tyler , Yolanda Pazos , Evangelos Spyrakos
{"title":"Optical properties of a toxin-producing dinoflagellate and its detection from Sentinel-2 MSI in nearshore waters","authors":"Conor McGlinchey , Jesus Torres Palenzuela , Luis Gonzalez-Vilas , Mortimer Werther , Dalin Jiang , Andrew Tyler , Yolanda Pazos , Evangelos Spyrakos","doi":"10.1016/j.isprsjprs.2025.06.017","DOIUrl":"10.1016/j.isprsjprs.2025.06.017","url":null,"abstract":"<div><div>Harmful algal blooms (HABs) caused by the dinoflagellate <em>Alexandrium minutum</em> can pose risks to human and ecosystem health. HABs of different species can coexist in coastal waters and accumulate near the shoreline, challenging their detection through Earth observation (EO). In this study, we use <em>in situ</em> geo-bio-optical and taxonomical data from the <em>Rías Baixas</em> (NW Spain) to develop a new method for identifying high-concentration blooms of <em>A. minutum</em> and its application to Sentinel-2 Multispectral Instrument (S2 MSI). Our approach named <em>A. minutum index</em> (AMI) was developed to capture the low absorption and high backscattering properties of <em>A. minutum</em> cells between 560 and 570 nm. We tested and validated the performance of three atmospheric correction algorithms (AC) (C2RCC, POLYMER and ACOLITE) using matchups between <em>in situ</em> and satellite-derived R<sub>rs</sub>. Results show that C2RCC had the lowest error across most wavelengths. Applying AMI to S2 MSI indicates that our approach can accurately identify high-concentration blooms of <em>A. minutum</em> (F1 score: 70 %, Kappa: 68.3 %, balanced accuracy: 87.7 %, MCC: 68.3 %) and discriminate blooms of <em>A. minutum</em> from other phytoplankton species. We compared AMI with three existing indices for detecting HABs in coastal waters and found that our approach achieved a better performance, with the NDTI, RGCI and NDCI yielding F1 scores of 21.28, 21.74, and 0.0 % and MCC values of 15.0, 15.0 and 0.0 %, respectively. We also investigated the spatial resolution of S2 MSI to Sentinel-3 Ocean and Land Colour Instrument (S3 OLCI) for mapping fine-scale variations in <em>A. minutum</em> blooms. We found that the higher spatial resolution data from S2 MSI were highly useful for detecting small-scale variations in <em>A. minutum</em> in nearshore waters, enabling their detection in the mid-inner part of the <em>Rías</em>, where aquaculture activities are more prominent. This study also showcases the significance of accurate AC in near-shore waters, where high-concentration blooms can be more prevalent. Our findings show that greater errors in AC are observed in near-shore pixels, where the socio-economic and environmental impact of HABs are typically more severe.</div></div>","PeriodicalId":50269,"journal":{"name":"ISPRS Journal of Photogrammetry and Remote Sensing","volume":"227 ","pages":"Pages 415-437"},"PeriodicalIF":10.6,"publicationDate":"2025-06-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144480395","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Wouter A.J. Van den Broeck, Louise Terryn, Shilin Chen, Wout Cherlet, Zane T. Cooper, Kim Calders
{"title":"Pointwise deep learning for leaf-wood segmentation of tropical tree point clouds from terrestrial laser scanning","authors":"Wouter A.J. Van den Broeck, Louise Terryn, Shilin Chen, Wout Cherlet, Zane T. Cooper, Kim Calders","doi":"10.1016/j.isprsjprs.2025.06.023","DOIUrl":"10.1016/j.isprsjprs.2025.06.023","url":null,"abstract":"<div><div>Terrestrial laser scanning (TLS) is increasingly used in forest monitoring, providing detailed 3D measurements of vegetation structure. However, the semantic understanding of tropical tree point clouds, particularly the segmentation of leaf and wood components, remains a challenge. Deep learning (DL) on point clouds has been gaining traction as a valuable tool for automated leaf-wood segmentation, but its widespread adoption is impeded by data availability, a lack of open-source trained models, and knowledge on its influence on subsequent woody volume reconstruction. To address these issues, this paper makes three key contributions. First, it introduces a new dataset consisting of 148 tropical tree TLS point clouds from north-eastern Australia with manual leaf-wood annotations. Second, it uses this dataset to compare several state-of-the-art point-wise DL networks and benchmark these against traditional approaches, using a common training and inference pipeline to allow for a fair model comparison. We conduct an ablation study to examine the effects of various hyperparameters and modelling choices, focusing solely on point coordinates as input to develop a model adaptable to different forest types, platforms, and point cloud qualities. Third, we assess the impact of point-wise segmentation quality on tropical tree volume estimation using quantitative structure model (QSM) reconstruction on the extracted woody component. Results show that our newly trained DL models significantly outperform traditional benchmarks for leaf-wood segmentation of tropical tree point clouds from TLS, with PointTransformer achieving the highest performance (mIoU = 92.2 %). Quantitative and qualitative analyses reveal that DL methods excel in distinguishing woody points, crucial for woody volume estimation via QSMs, but may suffer from connectivity issues due to lack of physical constraints. Volumes of trees segmented using PointTransformer closely match those of manually segmented trees (MAE = 7.1 %), highlighting its suitability for automated woody volume estimation. Although this study demonstrates the effectiveness of state-of-the-art neural architectures for tropical tree point cloud processing, advocating for their integration into forest structure analysis pipelines, future work should focus on enhancing quantity, quality and variety of training data, to increase model robustness and generalisability. We make the dataset, code and trained DL models publicly available.<span><span><sup>1</sup></span></span></div></div>","PeriodicalId":50269,"journal":{"name":"ISPRS Journal of Photogrammetry and Remote Sensing","volume":"227 ","pages":"Pages 366-382"},"PeriodicalIF":10.6,"publicationDate":"2025-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144472286","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}