Science of Remote Sensing最新文献_第7页

A spectral-preserving resampling for spatial upscaling of hyperspectral imagery 高光谱图像空间升级的保谱重采样方法

IF 5.2 Q1 ENVIRONMENTAL SCIENCES

Science of Remote Sensing

Pub Date : 2025-11-06 DOI: 10.1016/j.srs.2025.100330

Yuxin Tian, Zhenghai Wang

This paper proposes a spectral-preserving hyperspectral image resampling method (Spectral-Preserving Resampling, SpePR) designed to effectively retain the diagnostic spectral features of minerals during spatial upscaling. In this method, the band correlation structure of hyperspectral imagery is utilized as an intrinsic representation of spectral features, and Tikhonov regularized pseudo-inversion is introduced to mitigate spectral distortion induced by resampling. Within the proposed framework, spectral structural information is initially characterized by band correlation matrices. Subsequently, during the spatial resampling stage, the spectral preservation and spatial resampling terms are jointly optimized to ensure coordinated preservation of spectral and spatial information. The performance of the method was validated by analyzing multi-scale hyperspectral imagery data, with flight altitudes ranging from 30m to 150m, acquired using unmanned aerial vehicles. The results indicate that as spatial resolution decreases, mineral spectral features exhibit a corresponding decrease in absorption depth and absorption area, while maintaining stable absorption positions. Compared with seven conventional interpolation algorithms, SpePR reduces errors by 8.2 %–15.7 % in spectral angular mapping (SAM) and by 23.6 %–27.9 % in spectral correlation relative to the best conventional method. The proposed method also demonstrated significant advantages in metrics such as spectral gradient angle (SGA) and spectral correlation, while also more accurately preserving key mineral absorption features. Concurrently, SpePR demonstrated superior spatial information retention compared to conventional methods, as its resulting spatial features more closely approximated the actual observational imagery. The research findings confirm that the SpePR approach effectively preserves diagnostic spectral features of minerals, thereby providing reliable technical support for multi-scale hyperspectral mineral mapping.

本文提出了一种保留光谱的高光谱图像重采样方法（spectral-preserving resampling, SpePR），旨在有效地保留矿物在空间上尺度的诊断光谱特征。该方法利用高光谱图像的波段相关结构作为光谱特征的内在表征，并引入Tikhonov正则化伪反演来缓解重采样引起的光谱畸变。在该框架内，光谱结构信息首先由波段相关矩阵表征。随后，在空间重采样阶段，对光谱保存项和空间重采样项进行联合优化，保证光谱信息和空间信息的协调保存。通过分析无人机获取的飞行高度在30m ~ 150m之间的多尺度高光谱图像数据，验证了该方法的性能。结果表明：随着空间分辨率的降低，矿物光谱特征的吸收深度和吸收面积相应减小，但吸收位置保持稳定；与传统插值方法相比，SpePR在光谱角映射（SAM）上的误差降低了8.2% ~ 15.7%，在光谱相关上的误差降低了23.6% ~ 27.9%。该方法在光谱梯度角（SGA）和光谱相关性等指标上也具有显著的优势，同时也能更准确地保留关键的矿物吸收特征。与传统方法相比，SpePR显示出更好的空间信息保留能力，因为其得到的空间特征更接近实际观测图像。研究结果证实，SpePR方法有效地保留了矿物的诊断光谱特征，为多尺度高光谱矿物填图提供了可靠的技术支持。

{"title":"A spectral-preserving resampling for spatial upscaling of hyperspectral imagery","authors":"Yuxin Tian, Zhenghai Wang","doi":"10.1016/j.srs.2025.100330","DOIUrl":"10.1016/j.srs.2025.100330","url":null,"abstract":"<div><div>This paper proposes a spectral-preserving hyperspectral image resampling method (Spectral-Preserving Resampling, SpePR) designed to effectively retain the diagnostic spectral features of minerals during spatial upscaling. In this method, the band correlation structure of hyperspectral imagery is utilized as an intrinsic representation of spectral features, and Tikhonov regularized pseudo-inversion is introduced to mitigate spectral distortion induced by resampling. Within the proposed framework, spectral structural information is initially characterized by band correlation matrices. Subsequently, during the spatial resampling stage, the spectral preservation and spatial resampling terms are jointly optimized to ensure coordinated preservation of spectral and spatial information. The performance of the method was validated by analyzing multi-scale hyperspectral imagery data, with flight altitudes ranging from 30m to 150m, acquired using unmanned aerial vehicles. The results indicate that as spatial resolution decreases, mineral spectral features exhibit a corresponding decrease in absorption depth and absorption area, while maintaining stable absorption positions. Compared with seven conventional interpolation algorithms, SpePR reduces errors by 8.2 %–15.7 % in spectral angular mapping (SAM) and by 23.6 %–27.9 % in spectral correlation relative to the best conventional method. The proposed method also demonstrated significant advantages in metrics such as spectral gradient angle (SGA) and spectral correlation, while also more accurately preserving key mineral absorption features. Concurrently, SpePR demonstrated superior spatial information retention compared to conventional methods, as its resulting spatial features more closely approximated the actual observational imagery. The research findings confirm that the SpePR approach effectively preserves diagnostic spectral features of minerals, thereby providing reliable technical support for multi-scale hyperspectral mineral mapping.</div></div>","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"12 ","pages":"Article 100330"},"PeriodicalIF":5.2,"publicationDate":"2025-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145528146","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

Estimating tree diameter at breast height (DBH) from UAV data: A comparison of oblique–Vertical imagery fusion and allometric modeling 利用无人机数据估算树胸径：斜垂直影像融合与异速生长建模的比较

IF 5.2 Q1 ENVIRONMENTAL SCIENCES

Science of Remote Sensing

Pub Date : 2025-11-06 DOI: 10.1016/j.srs.2025.100331

Yousef Erfanifard , Ali Hosingholizade , Verena C. Griess , Virginia Elena Garcia Millan , Saied Pirasteh

Accurate estimation of tree diameter at breast height (DBH) is essential for forest inventory, biomass assessment, and ecological monitoring. Unmanned aerial vehicles (UAVs) have emerged as powerful tools for remote DBH estimation, yet challenges remain in accurately capturing stem dimensions from aerial perspectives. This study compares two approaches for estimating the DBH of pine trees: (I) a data fusion method that combines vertical and oblique UAV imagery to generate high-density point clouds, and (II) a Random Forest-based allometric modeling approach based on metrics derived from vertical UAV data. The first approach evaluated three fusion configurations combining vertical (90°) and oblique (30°, 60°) imagery: S1 (30°/90°), S2 (60°/90°), and S3 (30°/60°/90°). The second approach (S4) employed a Random Forest regression model using features derived from vertical UAV data, including tree height and crown size metrics to estimate DBH. Results demonstrate that the geometric approach using all three viewing angles (S3) achieved the highest accuracy (R² = 0.985, RMSE = 2.47 cm), followed closely by S2 (R² = 0.949), indicating the effectiveness of multi-angle image integration in improving DBH prediction. S4, while less accurate (R² = 0.824), provided moderately reliable estimates, offering a simpler and more scalable alternative in operational settings. In contrast, S1 significantly overestimated DBH, especially for larger trees, with a high positive bias and the largest RMSE. When evaluated across DBH size classes, S3 consistently outperformed other methods, with strong agreement across small, medium, and large trees. These findings highlight the value of oblique–vertical image fusion in enhancing DBH estimation accuracy, particularly when multiple viewing angles are used. While this study focuses on open-canopy pine stands, future research should assess these methods in denser forests and explore deep learning algorithms for DBH estimation from complex point clouds.

准确估算胸径对森林资源清查、生物量评估和生态监测具有重要意义。无人驾驶飞行器（uav）已经成为远程DBH估计的强大工具，但从空中角度准确捕获茎尺寸仍然存在挑战。本研究比较了两种估算松树胸径的方法：(I)结合垂直和倾斜无人机图像生成高密度点云的数据融合方法，以及（II）基于垂直无人机数据导出的度量的基于随机森林的异速建模方法。第一种方法评估了三种融合配置，结合垂直（90°）和倾斜（30°，60°）图像：S1（30°/90°），S2（60°/90°）和S3（30°/60°/90°）。第二种方法（S4）采用随机森林回归模型，利用垂直无人机数据的特征（包括树高和树冠大小指标）来估计胸径。结果表明，3个视角（S3）下的几何方法预测精度最高（R2 = 0.985, RMSE = 2.47 cm）， S2方法次之（R2 = 0.949），说明多角度图像集成在提高胸径预测中的有效性。S4虽然不太准确（R2 = 0.824），但提供了中等可靠的估计，在操作设置中提供了更简单和更可扩展的替代方案。相比之下，S1显著高估了胸径，特别是对于较大的树木，具有较高的正偏差和最大的RMSE。当跨DBH大小类进行评估时，S3始终优于其他方法，并且在小型、中型和大型树中都具有很强的一致性。这些发现突出了斜垂直图像融合在提高DBH估计精度方面的价值，特别是当使用多个视角时。虽然本研究侧重于开放冠层松林，但未来的研究应在更密集的森林中评估这些方法，并探索从复杂点云中估计DBH的深度学习算法。

{"title":"Estimating tree diameter at breast height (DBH) from UAV data: A comparison of oblique–Vertical imagery fusion and allometric modeling","authors":"Yousef Erfanifard , Ali Hosingholizade , Verena C. Griess , Virginia Elena Garcia Millan , Saied Pirasteh","doi":"10.1016/j.srs.2025.100331","DOIUrl":"10.1016/j.srs.2025.100331","url":null,"abstract":"<div><div>Accurate estimation of tree diameter at breast height (DBH) is essential for forest inventory, biomass assessment, and ecological monitoring. Unmanned aerial vehicles (UAVs) have emerged as powerful tools for remote DBH estimation, yet challenges remain in accurately capturing stem dimensions from aerial perspectives. This study compares two approaches for estimating the DBH of pine trees: (I) a data fusion method that combines vertical and oblique UAV imagery to generate high-density point clouds, and (II) a Random Forest-based allometric modeling approach based on metrics derived from vertical UAV data. The first approach evaluated three fusion configurations combining vertical (90°) and oblique (30°, 60°) imagery: S1 (30°/90°), S2 (60°/90°), and S3 (30°/60°/90°). The second approach (S4) employed a Random Forest regression model using features derived from vertical UAV data, including tree height and crown size metrics to estimate DBH. Results demonstrate that the geometric approach using all three viewing angles (S3) achieved the highest accuracy (R<sup>2</sup> = 0.985, RMSE = 2.47 cm), followed closely by S2 (R<sup>2</sup> = 0.949), indicating the effectiveness of multi-angle image integration in improving DBH prediction. S4, while less accurate (R<sup>2</sup> = 0.824), provided moderately reliable estimates, offering a simpler and more scalable alternative in operational settings. In contrast, S1 significantly overestimated DBH, especially for larger trees, with a high positive bias and the largest RMSE. When evaluated across DBH size classes, S3 consistently outperformed other methods, with strong agreement across small, medium, and large trees. These findings highlight the value of oblique–vertical image fusion in enhancing DBH estimation accuracy, particularly when multiple viewing angles are used. While this study focuses on open-canopy pine stands, future research should assess these methods in denser forests and explore deep learning algorithms for DBH estimation from complex point clouds.</div></div>","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"12 ","pages":"Article 100331"},"PeriodicalIF":5.2,"publicationDate":"2025-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145528105","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 systematic review on soil moisture estimation using remote sensing data for agricultural applications 农业应用遥感数据估算土壤水分的系统综述

IF 5.2 Q1 ENVIRONMENTAL SCIENCES

Science of Remote Sensing

Pub Date : 2025-11-04 DOI: 10.1016/j.srs.2025.100328

Ana C. Teixeira , Matus Bakon , Domingos Lopes , António Cunha , Joaquim J. Sousa

Soil moisture plays a central role in agricultural sustainability and water-resource management under climate change and increasing water scarcity. Remote-sensing technologies have transformed soil-moisture estimation by enabling large-scale, high-resolution, and continuous monitoring. Following the PRISMA framework, this systematic review analyzes 64 studies published between 2016 and 2024, selected from 379 screened articles, focusing on agricultural applications. Remote-sensing data span optical, thermal, and microwave observations from satellites and unmanned aerial vehicles (UAVs), with estimation approaches classified as empirical, semi-empirical, physical, or learning-based. Satellite observations dominate the literature (73% of studies), while UAVs are increasingly used for high-resolution, site-specific assessments. Multi-sensor fusion, combining optical, thermal, and microwave data, is a growing strategy to overcome the limitations of individual sensors. Active SAR systems provide weather-independent measurements with high spatial resolution, whereas optical and thermal sensors offer valuable spectral indices but are limited by cloud cover and shallow penetration depth. Learning-based methods are the most frequent approach (54% of studies), using machine and deep learning to model complex relationships between soil moisture and remote-sensing variables. Principal challenges include vegetation interference, surface roughness, and limited in-situ calibration data. Mitigation strategies involve longer-wavelength SAR (L- and P-bands), multi-sensor fusion, downscaling, and integration of auxiliary datasets (soil texture, elevation, meteorology). By synthesizing recent advances and emerging trends, this review provides practical guidance for accurate, scalable, and operational soil-moisture monitoring in precision agriculture and environmental management.

在气候变化和水资源日益短缺的情况下，土壤湿度在农业可持续性和水资源管理中发挥着核心作用。遥感技术通过实现大规模、高分辨率和连续监测，改变了土壤湿度估算。遵循PRISMA框架，本系统综述分析了2016年至2024年间发表的64项研究，从379篇筛选文章中选出，重点关注农业应用。遥感数据包括来自卫星和无人机的光学、热和微波观测，估计方法分为经验、半经验、物理或基于学习。卫星观测在文献中占主导地位（73%的研究），而无人机越来越多地用于高分辨率、特定地点的评估。多传感器融合，结合光学、热和微波数据，是克服单个传感器局限性的一种日益发展的策略。主动SAR系统提供与天气无关的高空间分辨率测量，而光学和热传感器提供有价值的光谱指数，但受云层覆盖和浅穿透深度的限制。基于学习的方法是最常见的方法（54%的研究），使用机器和深度学习来模拟土壤湿度和遥感变量之间的复杂关系。主要挑战包括植被干扰、表面粗糙度和有限的原位校准数据。缓解策略包括较长波长的SAR （L波段和p波段）、多传感器融合、降尺度和辅助数据集（土壤质地、海拔、气象）的整合。通过综合最新进展和新兴趋势，本文综述为精准农业和环境管理中精确、可扩展和可操作的土壤湿度监测提供了实用指导。

{"title":"A systematic review on soil moisture estimation using remote sensing data for agricultural applications","authors":"Ana C. Teixeira , Matus Bakon , Domingos Lopes , António Cunha , Joaquim J. Sousa","doi":"10.1016/j.srs.2025.100328","DOIUrl":"10.1016/j.srs.2025.100328","url":null,"abstract":"<div><div>Soil moisture plays a central role in agricultural sustainability and water-resource management under climate change and increasing water scarcity. Remote-sensing technologies have transformed soil-moisture estimation by enabling large-scale, high-resolution, and continuous monitoring. Following the PRISMA framework, this systematic review analyzes 64 studies published between 2016 and 2024, selected from 379 screened articles, focusing on agricultural applications. Remote-sensing data span optical, thermal, and microwave observations from satellites and unmanned aerial vehicles (UAVs), with estimation approaches classified as empirical, semi-empirical, physical, or learning-based. Satellite observations dominate the literature (73% of studies), while UAVs are increasingly used for high-resolution, site-specific assessments. Multi-sensor fusion, combining optical, thermal, and microwave data, is a growing strategy to overcome the limitations of individual sensors. Active SAR systems provide weather-independent measurements with high spatial resolution, whereas optical and thermal sensors offer valuable spectral indices but are limited by cloud cover and shallow penetration depth. Learning-based methods are the most frequent approach (54% of studies), using machine and deep learning to model complex relationships between soil moisture and remote-sensing variables. Principal challenges include vegetation interference, surface roughness, and limited in-situ calibration data. Mitigation strategies involve longer-wavelength SAR (L- and P-bands), multi-sensor fusion, downscaling, and integration of auxiliary datasets (soil texture, elevation, meteorology). By synthesizing recent advances and emerging trends, this review provides practical guidance for accurate, scalable, and operational soil-moisture monitoring in precision agriculture and environmental management.</div></div>","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"12 ","pages":"Article 100328"},"PeriodicalIF":5.2,"publicationDate":"2025-11-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145465974","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

Monitoring cropland cultivation, abandonment, fallowing and recultivation dynamics with regard to conflict intensity in war-affected Ukraine 监测受战争影响的乌克兰境内与冲突强度相关的耕地种植、撂荒、休耕地和复垦动态

IF 5.2 Q1 ENVIRONMENTAL SCIENCES

Science of Remote Sensing

Pub Date : 2025-11-04 DOI: 10.1016/j.srs.2025.100326

Josef Wagner , Shabarinath S. Nair , Sergii Skakun , Erik C. Duncan , Fangjie Li , Oleksandra Oliinyk , Françoise Nerry , Jean Rehbinder , Inbal Becker-Reshef

Three years of sustained shelling, mining, and active combat have caused major cropland abandonment in Ukraine, particularly along frontlines. Existing estimates of abandoned areas vary up to fourfold, due to inconsistent definitions, baselines and biased estimators of area. Many studies classify fallow land – temporarily unused but managed – as abandoned. In contrast, abandoned lands (neither cultivated nor managed) are often contaminated by unexploded ordnance, mines, or chemicals, requiring clearance before recultivation. Disentangling fallow from abandoned cropland is therefore crucial for post-war recovery planning and for determining tax relief for farmers unable to access their fields. We applied a two-level stratified random sampling design and unbiased estimators of area to quantify the extent of cultivated, fallow, and abandoned cropland. Four regions with distinct conflict dynamics were delineated using the Armed Conflict Location and Event Dataset, and within each region stratified random samples were drawn from a Planet-based cultivation status map. Between 2021 and 2024, full Ukraine's cultivated area declined by 2.5 Mha (−8.5 %). By 2024, 7 % of cropland (2.213 ± 0.256 Mha) was abandoned, of which 1.121 ± 0.148 Mha to 1.671 ± 0.169 Mha may be permanently lost to cultivation, primarily along frontlines. Fallow areas increased nationally, especially in territories reclaimed from occupation. In 2024 alone, up to 0.472 ± 0.143 Mha were recultivated, by far exceeding official land clearance figures and suggesting widespread reliance on informal or self-organized demining. These results establish a replicable framework for monitoring land-use dynamics in conflict zones, supporting evidence-based recovery, landmine clearance prioritization, and agricultural policy planning in post-war Ukraine.

三年的持续炮击、地雷和积极的战斗导致乌克兰大片农田被遗弃，特别是在前线。由于不一致的定义、基线和有偏差的面积估计，对废弃地区的现有估计相差高达四倍。许多研究将休耕土地——暂时未使用但得到管理的土地——归类为废弃土地。相比之下，被遗弃的土地（既没有耕种也没有管理）往往受到未爆弹药、地雷或化学品的污染，需要在重新耕种之前进行清理。因此，将休耕区从废弃农田中分离出来，对于战后恢复计划和确定无法进入农田的农民的税收减免至关重要。我们采用两水平分层随机抽样设计和无偏面积估计来量化耕地、休耕耕地和废弃耕地的范围。利用“武装冲突地点和事件数据集”（Armed conflict Location and Event Dataset）圈定了4个具有不同冲突动态的区域，并在每个区域内从基于行星的种植状态图中抽取分层随机样本。在2021年至2024年期间，乌克兰的耕地面积减少了2.5亿公顷（- 8.5%）。到2024年，将有7%的耕地（2.213±0.256 Mha）被撂荒，其中1.121±0.148 ~ 1.671±0.169 Mha可能永久退耕，主要集中在前线。休耕面积在全国范围内增加，特别是在收复的领土上。仅在2024年，就重新开垦了多达0.472±0.143公顷的土地，远远超过了官方的土地清理数据，表明广泛依赖非正式或自组织排雷。这些结果建立了一个可复制的框架，用于监测冲突地区的土地使用动态，支持战后乌克兰的循证恢复、排雷优先次序和农业政策规划。

{"title":"Monitoring cropland cultivation, abandonment, fallowing and recultivation dynamics with regard to conflict intensity in war-affected Ukraine","authors":"Josef Wagner , Shabarinath S. Nair , Sergii Skakun , Erik C. Duncan , Fangjie Li , Oleksandra Oliinyk , Françoise Nerry , Jean Rehbinder , Inbal Becker-Reshef","doi":"10.1016/j.srs.2025.100326","DOIUrl":"10.1016/j.srs.2025.100326","url":null,"abstract":"<div><div>Three years of sustained shelling, mining, and active combat have caused major cropland abandonment in Ukraine, particularly along frontlines. Existing estimates of abandoned areas vary up to fourfold, due to inconsistent definitions, baselines and biased estimators of area. Many studies classify fallow land – temporarily unused but managed – as abandoned. In contrast, abandoned lands (neither cultivated nor managed) are often contaminated by unexploded ordnance, mines, or chemicals, requiring clearance before recultivation. Disentangling fallow from abandoned cropland is therefore crucial for post-war recovery planning and for determining tax relief for farmers unable to access their fields. We applied a two-level stratified random sampling design and unbiased estimators of area to quantify the extent of cultivated, fallow, and abandoned cropland. Four regions with distinct conflict dynamics were delineated using the Armed Conflict Location and Event Dataset, and within each region stratified random samples were drawn from a Planet-based cultivation status map. Between 2021 and 2024, full Ukraine's cultivated area declined by 2.5 Mha (−8.5 %). By 2024, 7 % of cropland (2.213 ± 0.256 Mha) was abandoned, of which 1.121 ± 0.148 Mha to 1.671 ± 0.169 Mha may be permanently lost to cultivation, primarily along frontlines. Fallow areas increased nationally, especially in territories reclaimed from occupation. In 2024 alone, up to 0.472 ± 0.143 Mha were recultivated, by far exceeding official land clearance figures and suggesting widespread reliance on informal or self-organized demining. These results establish a replicable framework for monitoring land-use dynamics in conflict zones, supporting evidence-based recovery, landmine clearance prioritization, and agricultural policy planning in post-war Ukraine.</div></div>","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"12 ","pages":"Article 100326"},"PeriodicalIF":5.2,"publicationDate":"2025-11-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145528678","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

Diurnal dynamics and coastal challenges in chlorophyll-a monitoring using Himawari-8 satellite data 利用Himawari-8卫星数据监测叶绿素-a的日动态和沿海挑战

IF 5.2 Q1 ENVIRONMENTAL SCIENCES

Science of Remote Sensing

Pub Date : 2025-11-04 DOI: 10.1016/j.srs.2025.100320

Quang-Tu Bui, Stéphane Saux Picart, Jerome Vidot

Chlorophyll-a concentration estimated from satellite measurements is useful to monitor phytoplankton biomass. However, traditional methods of using polar orbit satellites only provide low frequency observation, inadequate to detect hourly changes that often trigger ecological events. With high temporal frequency data, geostationary sensors are expected to solve this problem. This research investigates whether high frequency data from Himawari-8, when applied rigorous quality control, can be reliable for oceanographic applications. A large dataset of Himawari-8 pixels was collocated with Sentinel-3 during four different three-day periods, corresponding to solstices and equinox in 2019. After applying a homogeneous filter (CV <0.2), full-disk analysis has exhibited high correlations between two datasets with

r \approx 0.82

in June, September, and December. However, the correlation decreases in March with r=0.685 due to misestimation of chlorophyll-a concentration at high viewing zenith angle. In low viewing zenith angle areas, Himawari-8,overestimates chlorophyll-a concentration compared to Sentinel-3, while in areas with higher viewing zenith angle of above 35°, chlorophyll-a concentration is underestimated by up to 12%–19% compared to Sentinel-3. These misestimations are mostly located in the viewing zenith angle range of 45°to 55°. For temporal analysis, one-hour timeseries analysis shows a very low day-to-day variation in offshore oligotrophic gyres (<0.05 mg/

m^{3}

). In the coastal area, the noon variation can peak up to 0.2 mg/

m^{3}

. Although these patterns are similar to the known diurnal variation, when phytoplankton absorbs available sunlight and nutrients to grow, there is no direct evidence to prove that observed chlorophyll-a variations are caused by the growth of phytoplankton.

通过卫星测量估算出的叶绿素-a浓度对监测浮游植物生物量是有用的。然而，使用极轨卫星的传统方法只能提供低频率观测，不足以探测经常引发生态事件的每小时变化。利用高时间频率数据，地球静止传感器有望解决这一问题。本研究探讨了在严格的质量控制下，来自Himawari-8的高频数据是否可以可靠地用于海洋学应用。将Himawari-8像素的大型数据集与Sentinel-3在四个不同的三天期间进行了配对，对应于2019年的至日和春分。在应用均匀过滤器（CV <0.2）后，全盘分析显示，6月、9月和12月两个数据集之间的相关性很高，r≈0.82。但在3月份，由于对高观测天顶角叶绿素a浓度估计错误，相关系数降低，r=0.685。在观测天顶角较低的地区，Himawari-8比Sentinel-3高估了叶绿素-a浓度，而在观测天顶角大于35°的地区，叶绿素-a浓度比Sentinel-3低估了12% ~ 19%。这些错误估计大多位于观测天顶角45°至55°范围内。对于时间分析，一小时时间序列分析显示，近海少营养环流的日变化非常小（0.05 mg/m3）。沿海地区中午变化最高可达0.2 mg/m3。虽然这些模式与已知的浮游植物吸收可利用的阳光和养分生长的日变化相似，但没有直接证据证明观测到的叶绿素-a变化是由浮游植物的生长引起的。

{"title":"Diurnal dynamics and coastal challenges in chlorophyll-a monitoring using Himawari-8 satellite data","authors":"Quang-Tu Bui, Stéphane Saux Picart, Jerome Vidot","doi":"10.1016/j.srs.2025.100320","DOIUrl":"10.1016/j.srs.2025.100320","url":null,"abstract":"<div><div>Chlorophyll-a concentration estimated from satellite measurements is useful to monitor phytoplankton biomass. However, traditional methods of using polar orbit satellites only provide low frequency observation, inadequate to detect hourly changes that often trigger ecological events. With high temporal frequency data, geostationary sensors are expected to solve this problem. This research investigates whether high frequency data from Himawari-8, when applied rigorous quality control, can be reliable for oceanographic applications. A large dataset of Himawari-8 pixels was collocated with Sentinel-3 during four different three-day periods, corresponding to solstices and equinox in 2019. After applying a homogeneous filter (CV <0.2), full-disk analysis has exhibited high correlations between two datasets with <span><math><mrow><mi>r</mi><mo>≈</mo><mn>0</mn><mo>.</mo><mn>82</mn></mrow></math></span> in June, September, and December. However, the correlation decreases in March with r=0.685 due to misestimation of chlorophyll-a concentration at high viewing zenith angle. In low viewing zenith angle areas, Himawari-8,overestimates chlorophyll-a concentration compared to Sentinel-3, while in areas with higher viewing zenith angle of above 35°, chlorophyll-a concentration is underestimated by up to 12%–19% compared to Sentinel-3. These misestimations are mostly located in the viewing zenith angle range of 45°to 55°. For temporal analysis, one-hour timeseries analysis shows a very low day-to-day variation in offshore oligotrophic gyres (<0.05 mg/<span><math><msup><mrow><mi>m</mi></mrow><mrow><mn>3</mn></mrow></msup></math></span>). In the coastal area, the noon variation can peak up to 0.2 mg/<span><math><msup><mrow><mi>m</mi></mrow><mrow><mn>3</mn></mrow></msup></math></span>. Although these patterns are similar to the known diurnal variation, when phytoplankton absorbs available sunlight and nutrients to grow, there is no direct evidence to prove that observed chlorophyll-a variations are caused by the growth of phytoplankton.</div></div>","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"12 ","pages":"Article 100320"},"PeriodicalIF":5.2,"publicationDate":"2025-11-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145528067","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

Combining Landsat optical/thermal and LiDAR High Definition data to estimate turbulent fluxes over Mediterranean forests 结合陆地卫星光学/热和激光雷达高清数据估算地中海森林上空的湍流通量

IF 5.2 Q1 ENVIRONMENTAL SCIENCES

Science of Remote Sensing

Pub Date : 2025-11-01 DOI: 10.1016/j.srs.2025.100323

Victor Penot, Olivier Merlin

<div><div>The performance of thermal-based models for estimating sensible (<span><math><mi>H</mi></math></span>) and latent (<span><math><mrow><mi>L</mi><mi>E</mi></mrow></math></span>) heat fluxes over semi-arid forests is still not well documented, largely due to the difficulty of accounting for canopy height (<span><math><mrow><mi>h</mi><mi>c</mi></mrow></math></span>) effects on satellite land surface temperature (<span><math><mrow><mi>L</mi><mi>S</mi><mi>T</mi></mrow></math></span>). This study addresses this limitation by integrating LiDAR-derived <span><math><mrow><mi>h</mi><mi>c</mi></mrow></math></span> into the classical contextual method, which combines Landsat-derived <span><math><mrow><mi>L</mi><mi>S</mi><mi>T</mi></mrow></math></span> and green vegetation fraction. Landsat <span><math><mrow><mi>L</mi><mi>S</mi><mi>T</mi></mrow></math></span> is first normalized to remove the influence of <span><math><mrow><mi>h</mi><mi>c</mi></mrow></math></span>, and the resulting normalized <span><math><mrow><mi>L</mi><mi>S</mi><mi>T</mi></mrow></math></span> is then used to estimate <span><math><mi>H</mi></math></span> and <span><math><mrow><mi>L</mi><mi>E</mi></mrow></math></span> using the classical contextual approach. The method is applied over a nine-year period in two Mediterranean forest sites with eddy covariance stations – Puechabon and Fontblanche. The threshold canopy height (<span><math><mrow><mi>h</mi><msub><mrow><mi>c</mi></mrow><mrow><mi>m</mi><mi>a</mi><mi>x</mi></mrow></msub></mrow></math></span>), above which <span><math><mrow><mi>L</mi><mi>S</mi><mi>T</mi></mrow></math></span> becomes insensitive to turbulent fluxes, is estimated as <span><math><mrow><mi>h</mi><msub><mrow><mi>c</mi></mrow><mrow><mi>m</mi><mi>a</mi><mi>x</mi></mrow></msub></mrow></math></span> = 42 ＋/− 4 m for Puechabon and <span><math><mrow><mi>h</mi><msub><mrow><mi>c</mi></mrow><mrow><mi>m</mi><mi>a</mi><mi>x</mi></mrow></msub></mrow></math></span> = 35 ＋/− 3 m for Fontblanche. For both sites, the normalization of <span><math><mrow><mi>L</mi><mi>S</mi><mi>T</mi></mrow></math></span> for <span><math><mrow><mi>h</mi><mi>c</mi></mrow></math></span> effect significantly improves the correlation between remotely sensed and in situ <span><math><mi>H</mi></math></span> (<span><math><mrow><mi>L</mi><mi>E</mi></mrow></math></span>) measurements from 0.40 (0.06) to 0.72 (0.43), respectively. Moreover, by setting the dry edge by a simple soil energy balance model, the bias between remotely sensed and in situ <span><math><mi>H</mi></math></span> (<span><math><mrow><mi>L</mi><mi>E</mi></mrow></math></span>) measurements is much reduced from −163 (＋132) W.m<sup>−2</sup> to −56 (＋25) W.m<sup>−2</sup>, and the slope of the linear regression much closer to 1 from 0.28 (0.07) to 0.84 (0.60), respectively. This is the first study to incorporate LiDAR-derived <span><math><mrow><mi>h</mi><mi>c</mi></mrow></math></span> into contextual methods, significantly improving thermal-based

基于热的模式用于估算半干旱森林的感热通量和潜热通量的性能尚未得到很好的记录，这主要是由于难以考虑冠层高度对卫星地表温度（LST）的影响。本研究通过将lidar导出的hc与经典上下文方法（结合landsat导出的LST和绿色植被分数）相结合，解决了这一限制。首先对Landsat的LST进行归一化以消除hc的影响，然后使用经典的上下文方法将得到的归一化LST用于估计H和LE。该方法在两个地中海森林站点——普查邦和丰特布兰奇——应用了9年的时间。阈值冠层高度（hcmax）估计为Puechabon的hcmax = 42 +/ - 4 m， Fontblanche的hcmax = 35 +/ - 3 m，超过该值LST对湍流通量不敏感。对于两个站点，地表温度的归一化显著提高了遥感和原位H （LE）测量值的相关性，相关性分别从0.40（0.06）提高到0.72（0.43）。此外，通过简单的土壤能量平衡模型设置干边缘，遥感和原位H （LE）测量之间的偏差从- 163 (+132)W.m−2大大减小到- 56 (+25)W.m−2，线性回归斜率分别从0.28（0.07）到0.84（0.60）更接近1。这是第一个将激光雷达衍生的hc纳入上下文方法的研究，显着改善了基于热的森林半干旱环境湍流通量估计。

引用次数: 0

Refining point-cloud neighborhood construction for improved classification 改进点云邻域构建，改进分类

IF 5.2 Q1 ENVIRONMENTAL SCIENCES

Science of Remote Sensing

Pub Date : 2025-11-01 DOI: 10.1016/j.srs.2025.100325

Max Hess, Aljoscha Rheinwalt, Bodo Bookhagen

3D point clouds enable us to capture the structure of our environment in high detail. In the context of urban areas, point clouds allow us to segment single objects, detect changes, and generate detailed maps of highly complex scenes. All of these tasks share a common challenge: they rely on classifying each point into a meaningful class. The first step of this process involves constructing the neighborhood of each point, which is relevant to the subsequent steps of feature calculation and class prediction. Constructing robust neighborhoods that capture meaningful point characteristics for diverse datasets remains a challenging task.

We propose a structure-aware neighborhood for calculating geometric features using a kd-tree-based region-growing approach. We construct neighborhoods by selectively adding points guided by local point connectivity, normal orientation, and distance from the seed point. In particular, the connectivity is determined by a nearest-neighbor graph, parameterized to connect only points belonging to the same object. Following this graph, points are iteratively added to the neighborhood if the angular difference between their normal orientations lies below a locally derived tolerance threshold.

We conducted experiments on three lidar-based datasets from Vaihingen, Berlin, and Paris, and a Structure-from-Motion dataset from Potsdam, which vary in size, point density, and class imbalance. The results show that the constraint neighborhood improves urban point cloud classification for the classes ground, buildings, vegetation, and cars across the different datasets. Our tests include parameter stability, generalization capability, and analysis of the derived feature spaces.

3D点云使我们能够以高细节捕捉环境的结构。在城市地区的背景下，点云允许我们分割单个物体，检测变化，并生成高度复杂场景的详细地图。所有这些任务都有一个共同的挑战：它们依赖于将每个点分类到一个有意义的类中。该过程的第一步是构建每个点的邻域，这与后续的特征计算和类别预测有关。构建鲁棒邻域以捕获不同数据集的有意义的点特征仍然是一项具有挑战性的任务。我们提出了一个结构感知邻域，用于使用基于kd树的区域生长方法计算几何特征。我们通过局部点连通性、法线方向和与种子点的距离来选择性地添加点来构建邻域。特别是，连通性由最近邻图决定，参数化后只连接属于同一对象的点。在此图中，如果法向之间的角差低于局部导出的公差阈值，则迭代地将点添加到邻域。我们在来自Vaihingen、Berlin和Paris的三个基于激光雷达的数据集以及来自Potsdam的Structure-from-Motion数据集上进行了实验，这些数据集在大小、点密度和类别不平衡方面有所不同。结果表明，约束邻域改进了城市点云在不同数据集上对地面、建筑物、植被和汽车等类别的分类。我们的测试包括参数稳定性、泛化能力和对衍生特征空间的分析。

{"title":"Refining point-cloud neighborhood construction for improved classification","authors":"Max Hess, Aljoscha Rheinwalt, Bodo Bookhagen","doi":"10.1016/j.srs.2025.100325","DOIUrl":"10.1016/j.srs.2025.100325","url":null,"abstract":"<div><div>3D point clouds enable us to capture the structure of our environment in high detail. In the context of urban areas, point clouds allow us to segment single objects, detect changes, and generate detailed maps of highly complex scenes. All of these tasks share a common challenge: they rely on classifying each point into a meaningful class. The first step of this process involves constructing the neighborhood of each point, which is relevant to the subsequent steps of feature calculation and class prediction. Constructing robust neighborhoods that capture meaningful point characteristics for diverse datasets remains a challenging task.</div><div>We propose a structure-aware neighborhood for calculating geometric features using a kd-tree-based region-growing approach. We construct neighborhoods by selectively adding points guided by local point connectivity, normal orientation, and distance from the seed point. In particular, the connectivity is determined by a nearest-neighbor graph, parameterized to connect only points belonging to the same object. Following this graph, points are iteratively added to the neighborhood if the angular difference between their normal orientations lies below a locally derived tolerance threshold.</div><div>We conducted experiments on three lidar-based datasets from Vaihingen, Berlin, and Paris, and a Structure-from-Motion dataset from Potsdam, which vary in size, point density, and class imbalance. The results show that the constraint neighborhood improves urban point cloud classification for the classes ground, buildings, vegetation, and cars across the different datasets. Our tests include parameter stability, generalization capability, and analysis of the derived feature spaces.</div></div>","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"12 ","pages":"Article 100325"},"PeriodicalIF":5.2,"publicationDate":"2025-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145465973","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

Spatiotemporal averaging resolution of high importance within Earth-observation-based light use efficiency models of gross primary production 在基于地球观测的初级生产总量光能利用效率模型中具有高度重要性的时空平均分辨率

IF 5.2 Q1 ENVIRONMENTAL SCIENCES

Science of Remote Sensing

Pub Date : 2025-11-01 DOI: 10.1016/j.srs.2025.100324

Torbern Tagesson , Paul Senty , Ousmane Diatta , Zhanzhang Cai , Aleksander Wieckowski , Ousmane Ndiaye , Jonas Ardö

Gross primary production (GPP) of the vegetation is the largest carbon exchange process of the global carbon cycle. Currently, within satellite-based remote sensing, GPP is generally modelled using linear light use efficiency models where GPP is related to photosynthetically active radiation absorbed by the green vegetation (APAR). These models work well on moderate to low spatiotemporal averaging resolutions. However, the relationship has been shown to rather follow an asymptotic curve at high spatiotemporal resolutions. The main aim of this study was to investigate at which spatial and temporal scale the GPP-APAR relationship converts from being asymptotic to linear. We used field data and satellite observations from the Dahra field site, a semi-arid savanna grassland in West Africa. At half-hourly to daily temporal resolution an asymptotic relationship gives the better fit, whereas for monthly and weekly data a linear relationship is preferred. A linear relationship was best when working with low spatial resolutions (>two and four Ha for daily and sub-daily GPP estimates, respectively), whereas if working with smaller pixel sizes, the asymptotic relationship was preferred. Hence, if studying GPP variability with satellite sensors such as AVHRR, MODIS, and Sentinel-3, a linear light use efficiency approach works well, whereas if using sensors such as Landsat and Sentinel-2, an asymptotic relationship is recommended. If we aim to improve our understanding of the GPP variability and its role within the carbon cycle, increasing the spatial and temporal resolution of Earth observation-based products is vital. This study provides an initial step toward the impact this may have, and future research across diverse ecosystems and over longer timescales is essential to expand upon these findings.

植被的总初级生产（GPP）是全球碳循环中最大的碳交换过程。目前，在卫星遥感中，一般采用线性光利用效率模型来模拟GPP，其中GPP与绿色植被吸收的光合有效辐射（APAR）有关。这些模型在中低时空平均分辨率下工作良好。然而，在高时空分辨率下，这种关系已被证明是一条渐近曲线。本研究的主要目的是探讨GPP-APAR关系在空间和时间尺度上由渐近向线性转换的过程。我们使用了西非半干旱稀树草原Dahra现场的野外数据和卫星观测。在半小时至每日的时间分辨率下，渐近关系提供了更好的拟合，而对于月和周数据，线性关系是首选。线性关系在低空间分辨率下是最好的（分别为每日和次每日GPP估计的>； 2和4 Ha），而如果使用较小的像素尺寸，则首选渐近关系。因此，如果使用AVHRR、MODIS和Sentinel-3等卫星传感器研究GPP变异性，线性光利用效率方法效果很好，而如果使用Landsat和Sentinel-2等传感器，则建议采用渐近关系。如果我们的目标是提高我们对GPP变率及其在碳循环中的作用的理解，提高基于地球观测的产品的时空分辨率至关重要。这项研究为这可能产生的影响提供了第一步，未来在不同生态系统和更长的时间尺度上的研究对于扩展这些发现至关重要。

{"title":"Spatiotemporal averaging resolution of high importance within Earth-observation-based light use efficiency models of gross primary production","authors":"Torbern Tagesson , Paul Senty , Ousmane Diatta , Zhanzhang Cai , Aleksander Wieckowski , Ousmane Ndiaye , Jonas Ardö","doi":"10.1016/j.srs.2025.100324","DOIUrl":"10.1016/j.srs.2025.100324","url":null,"abstract":"<div><div>Gross primary production (GPP) of the vegetation is the largest carbon exchange process of the global carbon cycle. Currently, within satellite-based remote sensing, GPP is generally modelled using linear light use efficiency models where GPP is related to photosynthetically active radiation absorbed by the green vegetation (APAR). These models work well on moderate to low spatiotemporal averaging resolutions. However, the relationship has been shown to rather follow an asymptotic curve at high spatiotemporal resolutions. The main aim of this study was to investigate at which spatial and temporal scale the GPP-APAR relationship converts from being asymptotic to linear. We used field data and satellite observations from the Dahra field site, a semi-arid savanna grassland in West Africa. At half-hourly to daily temporal resolution an asymptotic relationship gives the better fit, whereas for monthly and weekly data a linear relationship is preferred. A linear relationship was best when working with low spatial resolutions (>two and four Ha for daily and sub-daily GPP estimates, respectively), whereas if working with smaller pixel sizes, the asymptotic relationship was preferred. Hence, if studying GPP variability with satellite sensors such as AVHRR, MODIS, and Sentinel-3, a linear light use efficiency approach works well, whereas if using sensors such as Landsat and Sentinel-2, an asymptotic relationship is recommended. If we aim to improve our understanding of the GPP variability and its role within the carbon cycle, increasing the spatial and temporal resolution of Earth observation-based products is vital. This study provides an initial step toward the impact this may have, and future research across diverse ecosystems and over longer timescales is essential to expand upon these findings.</div></div>","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"12 ","pages":"Article 100324"},"PeriodicalIF":5.2,"publicationDate":"2025-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145465895","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

Debris covered glacier mapping using newly annotated multisource remote sensing data and geo-foundational model 基于新标注多源遥感数据和地理基础模型的碎屑覆盖冰川制图

IF 5.2 Q1 ENVIRONMENTAL SCIENCES

Science of Remote Sensing

Pub Date : 2025-10-30 DOI: 10.1016/j.srs.2025.100319

Saurabh Kaushik , Lalit Maurya , Elizabeth Tellman , Guoqing Zhang , Jaydeo K. Dharpure

The automated mapping of debris covered glaciers remains challenging due to spectral similarity between supraglacial debris (on-glaciers) and periglacial debris (off-glaciers). Deep learning offers promising capabilities, yet the lack of high-quality publicly available datasets and limited exploration of optimal model architecture constrain progress in this domain. To address this, we introduce the Global Supraglacial Debris Cover Dataset (GSDD), consisting of 1876 images (∼49,000.00 km²) collected globally from diverse glacierized regions, including High Mountain Asia, Andes, Western Canada, Alaska, and Swiss Alps, to incorporate the heterogeneity of glacial features and environments. This multisource remote sensing dataset includes 10 spectral bands—Blue, Green, Red, Near-Infrared, Shortwave Infrared (SWIR1 & SWIR2), Normalized Difference Rock Index (NDRI), Slope, Elevation, and Velocity—providing critical information to distinguish glacier debris. To evaluate the efficacy of deep learning models for mapping glacier debris, we compare Prithvi Geo-Foundational Model (GFM) combined with multiple decoders, CNN-based models (UNet, Attention U-Net, and DeepLabv3+), a Vision Transformer-based model (TransNorm), and variant of the Prithvi GFM (i.e., UViT). Our results show Prithvi GFM with UperNet decoder outperformed all, achieving mIoU = 0.80 and F1-score = 0.91, surpassing DeepLabv3+ (0.71 mIoU), Attention U-Net (0.73), U-Net (0.72), TransNorm (0.71), and UViT (0.70). Our results demonstrate significant methodological advances in accurately mapping glacier termini, along with the identification of glacier snouts. Feature analysis identified the optimal band combination (B-G-NIR-SWIR-Slope-Elevation) for debris mapping. The GSDD dataset enables direct comparison, development, and evaluation of deep learning models, supporting advancement in fast and reliable automated glacier mapping.

由于冰川上碎屑（冰川上）和冰周碎屑（冰川外）的光谱相似性，冰川覆盖碎屑的自动测绘仍然具有挑战性。深度学习提供了有前途的能力，但缺乏高质量的公开可用数据集和对最佳模型架构的有限探索限制了该领域的进展。为了解决这个问题，我们引入了全球冰川上碎屑覆盖数据集（GSDD），该数据集包括从全球不同的冰川化地区收集的1876张图像（~ 49,000.00 km2），包括亚洲高山、安第斯山脉、加拿大西部、阿拉斯加和瑞士阿尔卑斯山，以纳入冰川特征和环境的异质性。这个多源遥感数据集包括10个光谱波段——蓝色、绿色、红色、近红外、短波红外（SWIR1 & SWIR2）、归一化岩石指数（NDRI）、坡度、高程和速度——为区分冰川碎片提供了关键信息。为了评估深度学习模型在冰川碎片映射中的有效性，我们比较了Prithvi地理基础模型（GFM）与多个解码器、基于cnn的模型（UNet、Attention U-Net和DeepLabv3+）、基于视觉转换器的模型（TransNorm）和Prithvi地理基础模型的变型（即UViT）。我们的研究结果表明，Prithvi GFM与UperNet解码器的性能优于所有解码器，mIoU = 0.80, F1-score = 0.91，超过DeepLabv3+ (0.71 mIoU), Attention U-Net (0.73), U-Net (0.72), TransNorm（0.71）和UViT（0.70）。我们的研究结果表明，在准确绘制冰川末端以及识别冰川口部方面，方法上取得了重大进展。特征分析确定了最佳波段组合（B-G-NIR-SWIR-Slope-Elevation）。GSDD数据集可以直接比较、开发和评估深度学习模型，支持快速、可靠的自动冰川制图。

{"title":"Debris covered glacier mapping using newly annotated multisource remote sensing data and geo-foundational model","authors":"Saurabh Kaushik , Lalit Maurya , Elizabeth Tellman , Guoqing Zhang , Jaydeo K. Dharpure","doi":"10.1016/j.srs.2025.100319","DOIUrl":"10.1016/j.srs.2025.100319","url":null,"abstract":"<div><div>The automated mapping of debris covered glaciers remains challenging due to spectral similarity between supraglacial debris (on-glaciers) and periglacial debris (off-glaciers). Deep learning offers promising capabilities, yet the lack of high-quality publicly available datasets and limited exploration of optimal model architecture constrain progress in this domain. To address this, we introduce the Global Supraglacial Debris Cover Dataset (GSDD), consisting of 1876 images (∼49,000.00 km<sup>2</sup>) collected globally from diverse glacierized regions, including High Mountain Asia, Andes, Western Canada, Alaska, and Swiss Alps, to incorporate the heterogeneity of glacial features and environments. This multisource remote sensing dataset includes 10 spectral bands—Blue, Green, Red, Near-Infrared, Shortwave Infrared (SWIR1 & SWIR2), Normalized Difference Rock Index (NDRI), Slope, Elevation, and Velocity—providing critical information to distinguish glacier debris. To evaluate the efficacy of deep learning models for mapping glacier debris, we compare Prithvi Geo-Foundational Model (GFM) combined with multiple decoders, CNN-based models (UNet, Attention U-Net, and DeepLabv3+), a Vision Transformer-based model (TransNorm), and variant of the Prithvi GFM (i.e., UViT). Our results show Prithvi GFM with UperNet decoder outperformed all, achieving mIoU = 0.80 and F1-score = 0.91, surpassing DeepLabv3+ (0.71 mIoU), Attention U-Net (0.73), U-Net (0.72), TransNorm (0.71), and UViT (0.70). Our results demonstrate significant methodological advances in accurately mapping glacier termini, along with the identification of glacier snouts. Feature analysis identified the optimal band combination (B-G-NIR-SWIR-Slope-Elevation) for debris mapping. The GSDD dataset enables direct comparison, development, and evaluation of deep learning models, supporting advancement in fast and reliable automated glacier mapping.</div></div>","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"12 ","pages":"Article 100319"},"PeriodicalIF":5.2,"publicationDate":"2025-10-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145415891","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

Estimation of grassland canopy cover at quadrat and plot scales using multi-scale UAV imagery 基于多尺度无人机影像的样地和样地草地冠层覆盖度估算

IF 5.2 Q1 ENVIRONMENTAL SCIENCES

Science of Remote Sensing

Pub Date : 2025-10-30 DOI: 10.1016/j.srs.2025.100312

Yongcai Wang , Huawei Wan , Dongpo Wang , Jixi Gao , Zhuowei Hu , Junjie Wang , Fengming Wan , Bin Yang , Zhiru Zhang , Ma Bian , Jiqian Zhou

Grassland canopy cover acts as an essential metric for gauging the vitality and ecological functions of grassland. Unmanned aerial vehicles (UAVs) provide stable and reliable data for estimating grassland canopy cover. However, conventional approaches primarily rely on samples from ground surveys and visual assessments, where data consistency is often affected by variations in survey techniques and personnel expertise. By contrast, UAVs provide consistent multi-scale grassland canopy data. Thus, effectively harnessing the strengths of multi-scale UAV imagery can markedly improve the efficiency and precision of canopy cover estimation. This study uses high-resolution UAV imagery for semantic segmentation to derive precise quadrat-scale canopy cover as ground truth. Subsequently, a deep regression network is developed using UAV orthophotos to estimate canopy cover at the plot scale. The findings indicate that semantic segmentation models leveraging deep learning techniques provide accurate vegetation segmentation and canopy cover estimation at the quadrat level, with UNet++ delivering the highest performance, marked by a mean intersection over union (MIoU) of 0.81 and an F1-score of 0.88. The canopy cover results derived from UNet++ segmentation exhibit a coefficient of determination (R²) of 0.98 and a root mean square error (RMSE) under 3.6 %, surpassing conventional methods like Canopeo and Random Forest (RF). At plot scale, models based on convolutional neural networks (CNNs) and vision transformer (ViT) architecture show enhanced capabilities in predicting canopy cover, with the Swin transformer-based model achieving the greatest accuracy (R² = 0.90, RMSE = 5.48 %). In meadow, typical, and desert steppe, the Swin tansformer-based model consistently delivers high-precision canopy cover estimates. This study highlights the potential of integrating multi-scale UAV imagery with advanced deep learning techniques for efficient and accurate grassland vegetation monitoring. Future research should focus on optimizing model performance, extending applications to diverse ecosystems, and incorporating additional data sources to enhance robustness and precision.

草地冠层盖度是衡量草地活力和生态功能的重要指标。无人机为估算草原冠层覆盖度提供了稳定可靠的数据。然而，传统方法主要依赖于地面调查和目视评估的样本，其中数据的一致性经常受到调查技术和人员专业知识差异的影响。相比之下，无人机提供了一致的多尺度草地冠层数据。因此，有效利用多尺度无人机图像的优势，可以显著提高冠层覆盖度估算的效率和精度。本研究使用高分辨率无人机图像进行语义分割，以获得精确的方形尺度冠层覆盖作为地面真值。随后，利用无人机正射影像建立了一个深度回归网络，在样地尺度上估计冠层覆盖度。研究结果表明，利用深度学习技术的语义分割模型在样方水平上提供了准确的植被分割和冠层覆盖估计，其中UNet++提供了最高的性能，其平均交联（MIoU）为0.81，f1得分为0.88。UNet++分割得到的冠层盖度结果的决定系数（R2）为0.98，均方根误差（RMSE）小于3.6%，优于Canopeo和Random Forest （RF）等传统方法。在地块尺度上，基于卷积神经网络（cnn）和视觉变压器（ViT）架构的模型在预测冠层覆盖度方面表现出更强的能力，其中基于Swin变压器的模型的预测精度最高（R2 = 0.90, RMSE = 5.48%）。在草甸、典型草原和沙漠草原中，基于Swin变压器的模型始终提供高精度的冠层覆盖估算。该研究强调了将多尺度无人机图像与先进的深度学习技术相结合的潜力，以实现高效、准确的草地植被监测。未来的研究应侧重于优化模型性能，将应用扩展到不同的生态系统，并纳入额外的数据源以提高鲁棒性和精度。

{"title":"Estimation of grassland canopy cover at quadrat and plot scales using multi-scale UAV imagery","authors":"Yongcai Wang , Huawei Wan , Dongpo Wang , Jixi Gao , Zhuowei Hu , Junjie Wang , Fengming Wan , Bin Yang , Zhiru Zhang , Ma Bian , Jiqian Zhou","doi":"10.1016/j.srs.2025.100312","DOIUrl":"10.1016/j.srs.2025.100312","url":null,"abstract":"<div><div>Grassland canopy cover acts as an essential metric for gauging the vitality and ecological functions of grassland. Unmanned aerial vehicles (UAVs) provide stable and reliable data for estimating grassland canopy cover. However, conventional approaches primarily rely on samples from ground surveys and visual assessments, where data consistency is often affected by variations in survey techniques and personnel expertise. By contrast, UAVs provide consistent multi-scale grassland canopy data. Thus, effectively harnessing the strengths of multi-scale UAV imagery can markedly improve the efficiency and precision of canopy cover estimation. This study uses high-resolution UAV imagery for semantic segmentation to derive precise quadrat-scale canopy cover as ground truth. Subsequently, a deep regression network is developed using UAV orthophotos to estimate canopy cover at the plot scale. The findings indicate that semantic segmentation models leveraging deep learning techniques provide accurate vegetation segmentation and canopy cover estimation at the quadrat level, with UNet++ delivering the highest performance, marked by a mean intersection over union (MIoU) of 0.81 and an F1-score of 0.88. The canopy cover results derived from UNet++ segmentation exhibit a coefficient of determination (<em>R</em><sup>2</sup>) of 0.98 and a root mean square error (RMSE) under 3.6 %, surpassing conventional methods like Canopeo and Random Forest (RF). At plot scale, models based on convolutional neural networks (CNNs) and vision transformer (ViT) architecture show enhanced capabilities in predicting canopy cover, with the Swin transformer-based model achieving the greatest accuracy (<em>R</em><sup>2</sup> = 0.90, RMSE = 5.48 %). In meadow, typical, and desert steppe, the Swin tansformer-based model consistently delivers high-precision canopy cover estimates. This study highlights the potential of integrating multi-scale UAV imagery with advanced deep learning techniques for efficient and accurate grassland vegetation monitoring. Future research should focus on optimizing model performance, extending applications to diverse ecosystems, and incorporating additional data sources to enhance robustness and precision.</div></div>","PeriodicalId":101147,"journal":{"name":"Science of Remote Sensing","volume":"12 ","pages":"Article 100312"},"PeriodicalIF":5.2,"publicationDate":"2025-10-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145528076","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