Remote Sensing Applications-Society and Environment最新文献_第7页

Automated detection and classification of bike lanes using multimodal imagery 基于多模态图像的自行车道自动检测与分类

IF 4.5 Q2 ENVIRONMENTAL SCIENCES

Remote Sensing Applications-Society and Environment

Pub Date : 2026-01-01 Epub Date: 2025-12-02 DOI: 10.1016/j.rsase.2025.101817

Seung Jae Lieu , Bon Woo Koo , Uijeong Hwang , Subhrajit Guhathakurta

Bike lanes are a critical element of urban infrastructure that promote cycling and support sustainable transportation goals. Effective planning and evaluation require comprehensive inventory datasets that both identify the locations of bike lanes and classify their types. However, existing data collection is limited by inconsistent municipal documentation practices and resource constraints. This paper introduces a computer vision–based approach for the automated detection and classification of bike lanes using publicly available multimodal imagery. Each data sample integrates two street view images, captured from opposite directions, with a corresponding satellite image, enabling complementary perspectives. This approach allows the model to reliably detect bike lane presence and distinguish between designated (marked lanes without physical barriers) and protected (lanes separated from traffic by physical barriers) types. To optimize performance, we conduct ablation experiments across three architectural dimensions: stage of modality concatenation, fusion strategy, and label structure. We also construct a training dataset using Google Street View and satellite imagery from 28 major U.S. cities to ensure broad applicability. Applying the model to over 1000 road segments in Atlanta, Georgia, we demonstrate its scalability and accuracy in a real-world urban setting. By providing an automated, transferable method for developing bike lane inventories, this research addresses a critical gap in infrastructure documentation and supports more effective planning of bicycle networks.

自行车道是城市基础设施的重要组成部分，可以促进骑行和支持可持续交通目标。有效的规划和评估需要全面的库存数据集，既可以确定自行车道的位置，又可以对其类型进行分类。然而，现有的数据收集受到不一致的市政文件实践和资源限制的限制。本文介绍了一种基于计算机视觉的方法，利用公开的多模态图像自动检测和分类自行车道。每个数据样本将两张从相反方向拍摄的街景图像与相应的卫星图像相结合，从而实现互补视角。这种方法允许模型可靠地检测自行车道的存在，并区分指定（没有物理屏障的标记车道）和保护（物理屏障与交通隔开的车道）类型。为了优化性能，我们在三个架构维度上进行了消融实验：模态连接阶段、融合策略和标签结构。我们还使用谷歌街景和来自美国28个主要城市的卫星图像构建了一个训练数据集，以确保广泛的适用性。将该模型应用于佐治亚州亚特兰大市的1000多个路段，证明了其在现实城市环境中的可扩展性和准确性。通过提供一种自动化的、可转移的方法来开发自行车道清单，本研究解决了基础设施文档中的一个关键空白，并支持更有效的自行车网络规划。

{"title":"Automated detection and classification of bike lanes using multimodal imagery","authors":"Seung Jae Lieu , Bon Woo Koo , Uijeong Hwang , Subhrajit Guhathakurta","doi":"10.1016/j.rsase.2025.101817","DOIUrl":"10.1016/j.rsase.2025.101817","url":null,"abstract":"<div><div>Bike lanes are a critical element of urban infrastructure that promote cycling and support sustainable transportation goals. Effective planning and evaluation require comprehensive inventory datasets that both identify the locations of bike lanes and classify their types. However, existing data collection is limited by inconsistent municipal documentation practices and resource constraints. This paper introduces a computer vision–based approach for the automated detection and classification of bike lanes using publicly available multimodal imagery. Each data sample integrates two street view images, captured from opposite directions, with a corresponding satellite image, enabling complementary perspectives. This approach allows the model to reliably detect bike lane presence and distinguish between designated (marked lanes without physical barriers) and protected (lanes separated from traffic by physical barriers) types. To optimize performance, we conduct ablation experiments across three architectural dimensions: stage of modality concatenation, fusion strategy, and label structure. We also construct a training dataset using Google Street View and satellite imagery from 28 major U.S. cities to ensure broad applicability. Applying the model to over 1000 road segments in Atlanta, Georgia, we demonstrate its scalability and accuracy in a real-world urban setting. By providing an automated, transferable method for developing bike lane inventories, this research addresses a critical gap in infrastructure documentation and supports more effective planning of bicycle networks.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101817"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145665528","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

Where rivers sleep: mapping ephemeral sand rivers in the West African Sahel 河流沉睡的地方：绘制西非萨赫勒地区短暂的沙河

IF 4.5 Q2 ENVIRONMENTAL SCIENCES

Remote Sensing Applications-Society and Environment

Pub Date : 2026-01-01 Epub Date: 2025-12-12 DOI: 10.1016/j.rsase.2025.101838

Axel Belemtougri , Roland Yonaba , Claire I. Michailovsky , Tibor Stigter , Lawani Adjadi Mounirou , Pieter van der Zaag

This study presents a new geospatial framework for detecting and mapping ephemeral sand rivers (ESRs) across the West African Sahel, focusing on Burkina Faso, Mali, and Niger, where food security challenges are acute. ESRs, which remain dry most of the year, act as vital subsurface water reservoirs in arid environments. During the wet season, infiltrated streamflow is stored within sandy beds, minimizing evaporative losses and providing shallow groundwater with potential to support domestic, livestock, and agricultural needs during dry periods. The methodology integrates hydrological analyses, remote sensing, and machine learning. A high-resolution drainage network was derived from the 90 m MERIT DEM, based on national reference river networks and satellite-derived information. A Random Forest model predicted river flow intermittency and identified ephemeral rivers (flowing 1–4 months annually, catchment area ≥1000 km²), around which 500 m buffer zones were delineated for analysis. Two composite thresholds (CTs) combining NDESI–NDVI spectral indices achieved moderate accuracy: CT1 (42 %) and CT2 (72 %), with CT2 serving as a first-order tool for sandy riverbed detection. A multi-temporal supervised land use/land cover classification achieved high accuracy (92 %) and F1 scores >0.86, outperforming the spectral thresholds. Using vegetation presence as a proxy for shallow groundwater, 19 % of ESRs (402 km) were identified as areas of potentially accessible water storage near settlements representing about 3 million people (4.8 % of the population) across the three countries. These findings highlight the importance of ESRs for sustainable water management and climate-resilient livelihoods in the Sahel.

本研究提出了一个新的地理空间框架，用于探测和绘制整个西非萨赫勒地区的短暂沙河（ESRs），重点关注粮食安全挑战严峻的布基纳法索、马里和尼日尔。esr在一年中大部分时间保持干燥，在干旱环境中充当重要的地下水库。在雨季，渗透的水流被储存在沙质河床中，最大限度地减少蒸发损失，并提供浅层地下水，有可能在干旱时期支持家庭、牲畜和农业需求。该方法集成了水文分析、遥感和机器学习。基于国家参考河网和卫星衍生信息，从90 m MERIT DEM中获得了一个高分辨率的排水网络。随机森林模型预测了河流的间歇性，并确定了短暂河流（每年流动1-4个月，集水区面积≥1000 km2），并在其周围划定了500 m缓冲区进行分析。结合NDESI-NDVI光谱指数的两个复合阈值（ct）达到了中等精度：CT1（42%）和CT2(72%)，其中CT2作为砂质河床检测的一级工具。一个多时间监督的土地利用/土地覆盖分类获得了很高的准确性（92%），F1得分>；0.86，优于光谱阈值。使用植被存在作为浅层地下水的代表，19%的esr（402公里）被确定为潜在的可达储水区域，靠近三个国家约300万人口（占人口的4.8%）的定居点。这些发现突出了可持续水资源管理和气候适应型生计在萨赫勒地区的重要性。

{"title":"Where rivers sleep: mapping ephemeral sand rivers in the West African Sahel","authors":"Axel Belemtougri , Roland Yonaba , Claire I. Michailovsky , Tibor Stigter , Lawani Adjadi Mounirou , Pieter van der Zaag","doi":"10.1016/j.rsase.2025.101838","DOIUrl":"10.1016/j.rsase.2025.101838","url":null,"abstract":"<div><div>This study presents a new geospatial framework for detecting and mapping ephemeral sand rivers (ESRs) across the West African Sahel, focusing on Burkina Faso, Mali, and Niger, where food security challenges are acute. ESRs, which remain dry most of the year, act as vital subsurface water reservoirs in arid environments. During the wet season, infiltrated streamflow is stored within sandy beds, minimizing evaporative losses and providing shallow groundwater with potential to support domestic, livestock, and agricultural needs during dry periods. The methodology integrates hydrological analyses, remote sensing, and machine learning. A high-resolution drainage network was derived from the 90 m MERIT DEM, based on national reference river networks and satellite-derived information. A Random Forest model predicted river flow intermittency and identified ephemeral rivers (flowing 1–4 months annually, catchment area ≥1000 km<sup>2</sup>), around which 500 m buffer zones were delineated for analysis. Two composite thresholds (CTs) combining NDESI–NDVI spectral indices achieved moderate accuracy: CT1 (42 %) and CT2 (72 %), with CT2 serving as a first-order tool for sandy riverbed detection. A multi-temporal supervised land use/land cover classification achieved high accuracy (92 %) and F1 scores >0.86, outperforming the spectral thresholds. Using vegetation presence as a proxy for shallow groundwater, 19 % of ESRs (402 km) were identified as areas of potentially accessible water storage near settlements representing about 3 million people (4.8 % of the population) across the three countries. These findings highlight the importance of ESRs for sustainable water management and climate-resilient livelihoods in the Sahel.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101838"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145790938","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

MixerCA: An efficient and accurate model for high-performance hyperspectral image classification MixerCA：一种高效准确的高性能高光谱图像分类模型

IF 4.5 Q2 ENVIRONMENTAL SCIENCES

Remote Sensing Applications-Society and Environment

Pub Date : 2026-01-01 Epub Date: 2025-12-12 DOI: 10.1016/j.rsase.2025.101823

Mohammed Q. Alkhatib , Ali Jamali

Over the past decade, hyperspectral image (HSI) classification has drawn considerable interest due to HSIs’ ability to effectively distinguish terrestrial objects by capturing detailed, continuous spectral information. The strong performance of recent deep learning techniques in tasks like image classification and semantic segmentation has led to their growing use in HSI classification, due to their ability to capture complex spatial and spectral features more effectively than traditional methods. This paper presents MixerCA, a novel lightweight model for HSI classification that leverages depthwise convolution and a self-attention mechanism. MixerCA integrates depth-wise convolutions, token and channel mixing, and coordinate attention into a unified structure to decouple spatial and channel interactions, maintain consistent resolution throughout the network, and directly process HSI patches. Extensive experiments on four hyperspectral benchmark datasets reveal MixerCA’s clear advantages over several competing algorithms, including 2D-CNN, 3D-CNN, Tri-CNN, HybridSN, ViT, and Swin Transformer. The source code is publicly available at https://github.com/mqalkhatib/MixerCA.

在过去的十年中，由于高光谱图像（HSI）能够通过捕获详细的、连续的光谱信息来有效地区分地面物体，因此高光谱图像分类引起了相当大的兴趣。最近深度学习技术在图像分类和语义分割等任务中的强大表现导致它们在HSI分类中的应用越来越多，因为它们能够比传统方法更有效地捕获复杂的空间和光谱特征。本文提出了MixerCA，一种新的轻量级HSI分类模型，它利用深度卷积和自关注机制。MixerCA将深度卷积、令牌和通道混合以及协调注意集成到一个统一的结构中，以解耦空间和通道交互，在整个网络中保持一致的分辨率，并直接处理HSI补丁。在四种高光谱基准数据集上进行的大量实验表明，MixerCA比几种竞争算法（包括2D-CNN、3D-CNN、Tri-CNN、HybridSN、ViT和Swin Transformer）具有明显的优势。源代码可在https://github.com/mqalkhatib/MixerCA上公开获得。

{"title":"MixerCA: An efficient and accurate model for high-performance hyperspectral image classification","authors":"Mohammed Q. Alkhatib , Ali Jamali","doi":"10.1016/j.rsase.2025.101823","DOIUrl":"10.1016/j.rsase.2025.101823","url":null,"abstract":"<div><div>Over the past decade, hyperspectral image (HSI) classification has drawn considerable interest due to HSIs’ ability to effectively distinguish terrestrial objects by capturing detailed, continuous spectral information. The strong performance of recent deep learning techniques in tasks like image classification and semantic segmentation has led to their growing use in HSI classification, due to their ability to capture complex spatial and spectral features more effectively than traditional methods. This paper presents MixerCA, a novel lightweight model for HSI classification that leverages depthwise convolution and a self-attention mechanism. MixerCA integrates depth-wise convolutions, token and channel mixing, and coordinate attention into a unified structure to decouple spatial and channel interactions, maintain consistent resolution throughout the network, and directly process HSI patches. Extensive experiments on four hyperspectral benchmark datasets reveal MixerCA’s clear advantages over several competing algorithms, including 2D-CNN, 3D-CNN, Tri-CNN, HybridSN, ViT, and Swin Transformer. The source code is publicly available at <span><span>https://github.com/mqalkhatib/MixerCA</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101823"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145790942","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 of the smouldering coal-waste dump in Chorzów (Poland) using spectral indices: A UAV- and satellite-based approach 使用光谱指数监测Chorzów（波兰）的闷烧煤矸石堆：基于无人机和卫星的方法

IF 4.5 Q2 ENVIRONMENTAL SCIENCES

Remote Sensing Applications-Society and Environment

Pub Date : 2026-01-01 Epub Date: 2026-01-03 DOI: 10.1016/j.rsase.2025.101865

Anna Abramowicz, Michał Laska, Ádám Nádudvari, Oimahmad Rahmonov

The study aimed to evaluate the applicability of environmental indices in the monitoring of smouldering coal-waste dumps. A dump located in the Upper Silesian Coal Basin served as the research site for a multi-method analysis combining remote sensing and field-based data. Two UAV survey campaigns were conducted, capturing RGB, infrared, and multispectral imagery. These were supplemented with direct ground measurements of subsurface temperature and detailed vegetation mapping. Additionally, publicly available satellite data from the Landsat and Sentinel missions were analysed. A range of vegetation and fire-related indices (NDVI, SAVI, EVI, BAI, among others) were calculated to identify thermally active zones and assess vegetation conditions within these degraded areas. The results revealed strong seasonal variability in vegetation indices on thermally active sites, with evidence of disrupted vegetation cycles, including winter greening in moderately heated root zones – a pattern indicative of stress and degradation processes. While open-access satellite data, such as Landsat and Sentinel-2, proved useful in reconstructing the fire history of the dump, their spatial resolution was insufficient for detailed monitoring of small-scale thermal anomalies. The study highlights the diagnostic potential of UAV-based remote sensing in post-industrial environments undergoing land degradation but emphasises the importance of field validation for accurate environmental assessment.

本研究旨在评价环境指标在阴燃煤矸石堆场监测中的适用性。位于上西里西亚煤盆地的一个垃圾场作为研究地点，进行了结合遥感和实地数据的多方法分析。进行了两次无人机调查运动，捕获RGB、红外和多光谱图像。这些数据还补充了地下温度的直接地面测量和详细的植被测绘。此外，还分析了陆地卫星和哨兵任务的公开卫星数据。计算了一系列植被和火灾相关指数（NDVI、SAVI、EVI、BAI等），以确定热活跃区并评估这些退化地区的植被状况。结果显示，在热活跃区，植被指数具有强烈的季节性变化，植被循环被破坏，包括在中等热的根区冬季绿化，这是一种指示应力和退化过程的模式。虽然开放获取的卫星数据，如Landsat和Sentinel-2，在重建垃圾场的火灾历史方面被证明是有用的，但它们的空间分辨率不足以详细监测小规模的热异常。该研究强调了基于无人机的遥感在经历土地退化的后工业环境中的诊断潜力，但强调了实地验证对准确环境评估的重要性。

{"title":"Monitoring of the smouldering coal-waste dump in Chorzów (Poland) using spectral indices: A UAV- and satellite-based approach","authors":"Anna Abramowicz, Michał Laska, Ádám Nádudvari, Oimahmad Rahmonov","doi":"10.1016/j.rsase.2025.101865","DOIUrl":"10.1016/j.rsase.2025.101865","url":null,"abstract":"<div><div>The study aimed to evaluate the applicability of environmental indices in the monitoring of smouldering coal-waste dumps. A dump located in the Upper Silesian Coal Basin served as the research site for a multi-method analysis combining remote sensing and field-based data. Two UAV survey campaigns were conducted, capturing RGB, infrared, and multispectral imagery. These were supplemented with direct ground measurements of subsurface temperature and detailed vegetation mapping. Additionally, publicly available satellite data from the Landsat and Sentinel missions were analysed. A range of vegetation and fire-related indices (NDVI, SAVI, EVI, BAI, among others) were calculated to identify thermally active zones and assess vegetation conditions within these degraded areas. The results revealed strong seasonal variability in vegetation indices on thermally active sites, with evidence of disrupted vegetation cycles, including winter greening in moderately heated root zones – a pattern indicative of stress and degradation processes. While open-access satellite data, such as Landsat and Sentinel-2, proved useful in reconstructing the fire history of the dump, their spatial resolution was insufficient for detailed monitoring of small-scale thermal anomalies. The study highlights the diagnostic potential of UAV-based remote sensing in post-industrial environments undergoing land degradation but emphasises the importance of field validation for accurate environmental assessment.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101865"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145926023","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 dynamics of carbon, water, and energy balance in Bangladesh using multi-source remote sensing and climate data 基于多源遥感和气候数据的孟加拉国碳、水和能量平衡时空动态

IF 4.5 Q2 ENVIRONMENTAL SCIENCES

Remote Sensing Applications-Society and Environment

Pub Date : 2026-01-01 Epub Date: 2025-12-27 DOI: 10.1016/j.rsase.2025.101847

Nur Hussain , Md Saifuzzaman , Didar Islam , S.M. Shahriar Ahmed , Md Shamim Ahamed , Dara Shamsuddin

Exploring the complex interactions between climate variables and ecosystem processes is crucial for understanding long-term environmental changes. This study examines the spatiotemporal dynamics of carbon, water and energy fluxes and their impacts on ecosystem processes in Bangladesh from 2005 to 2022 utilizing multi-source remote sensing and ground-based meteorological data. Carbon dynamics are estimated through gross primary productivity (GPP), net primary production (NPP), and ecosystem respiration (RE). Water and energy balances are derived from evapotranspiration (ET), water use efficiency (WUE), net radiation (Rn), and latent heat (LE). Our estimates indicate that GPP varied from 2351.29 g C m⁻² y⁻¹ in 2009–2178.45 g C m⁻² y⁻¹ in 2020, while NPP ranged from 1248.13 g C m⁻² y⁻¹ in 2012 to 929.46 g C m⁻² y⁻¹ in 2020, reflecting temporal variations in photosynthetic efficiency and carbon storage. The ratio of LE/Rn was found to vary from 0.72 to 1.01, with an average of 83 %, indicating that a significant portion of the radiative energy was transferred to the atmosphere as turbulent flux. Validation of LUE-based GPP compared to FLUXCOM-GPP showed a moderate correlation (R² = 0.61, p < 0.005), supporting the reliability of the estimates. We also conducted multivariate regression analysis to assess the relationships between climate variables and carbon, water, and energy balance. The results indicate that photosynthetically active radiation (PAR) is the primary and dominant driver of GPP (R² = 0.97), while temperature and precipitation are key factors significantly influencing carbon uptake. This study presents a comprehensive, integrated assessment of carbon, water, and energy fluxes at the national scale across Bangladesh, emphasizing the crucial role of climate variables in shaping these fluxes and offering valuable insights for climate-resilient land management and sustainable carbon strategies in monsoon-dominated regions.

探索气候变量和生态系统过程之间复杂的相互作用对于理解长期环境变化至关重要。本研究利用多源遥感和地面气象数据，研究了2005 - 2022年孟加拉国碳、水和能量通量的时空动态及其对生态系统过程的影响。碳动态通过总初级生产力（GPP）、净初级生产力（NPP）和生态系统呼吸（RE）来估算。水分和能量平衡来源于蒸散发（ET）、水分利用效率（WUE）、净辐射（Rn）和潜热（LE）。我们的估计表明，GPP在2009年至2020年的2351.29 g C m−2 y−1之间变化，而NPP在2012年的1248.13 g C m−2 y−1到2020年的929.46 g C m−2 y−1之间变化，反映了光合效率和碳储量的时间变化。LE/Rn的比值在0.72 ~ 1.01之间变化，平均为83%，表明有很大一部分辐射能量以湍流通量的形式转移到大气中。与FLUXCOM-GPP相比，基于lue的GPP验证显示有中等相关性（R2 = 0.61, p < 0.005），支持估计的可靠性。我们还进行了多变量回归分析，以评估气候变量与碳、水和能量平衡之间的关系。结果表明，光合有效辐射（PAR）是GPP的主要和主导驱动因子（R2 = 0.97），而温度和降水是影响碳吸收的关键因素。本研究对孟加拉国全国范围内的碳、水和能源通量进行了全面、综合的评估，强调了气候变量在形成这些通量方面的关键作用，并为季风主导地区的气候适应型土地管理和可持续碳战略提供了有价值的见解。

{"title":"Spatiotemporal dynamics of carbon, water, and energy balance in Bangladesh using multi-source remote sensing and climate data","authors":"Nur Hussain , Md Saifuzzaman , Didar Islam , S.M. Shahriar Ahmed , Md Shamim Ahamed , Dara Shamsuddin","doi":"10.1016/j.rsase.2025.101847","DOIUrl":"10.1016/j.rsase.2025.101847","url":null,"abstract":"<div><div>Exploring the complex interactions between climate variables and ecosystem processes is crucial for understanding long-term environmental changes. This study examines the spatiotemporal dynamics of carbon, water and energy fluxes and their impacts on ecosystem processes in Bangladesh from 2005 to 2022 utilizing multi-source remote sensing and ground-based meteorological data. Carbon dynamics are estimated through gross primary productivity (GPP), net primary production (NPP), and ecosystem respiration (RE). Water and energy balances are derived from evapotranspiration (ET), water use efficiency (WUE), net radiation (Rn), and latent heat (LE). Our estimates indicate that GPP varied from 2351.29 g C m<sup>−2</sup> y<sup>−1</sup> in 2009–2178.45 g C m<sup>−2</sup> y<sup>−1</sup> in 2020, while NPP ranged from 1248.13 g C m<sup>−2</sup> y<sup>−1</sup> in 2012 to 929.46 g C m<sup>−2</sup> y<sup>−1</sup> in 2020, reflecting temporal variations in photosynthetic efficiency and carbon storage. The ratio of LE/Rn was found to vary from 0.72 to 1.01, with an average of 83 %, indicating that a significant portion of the radiative energy was transferred to the atmosphere as turbulent flux. Validation of LUE-based GPP compared to FLUXCOM-GPP showed a moderate correlation (R<sup>2</sup> = 0.61, p < 0.005), supporting the reliability of the estimates. We also conducted multivariate regression analysis to assess the relationships between climate variables and carbon, water, and energy balance. The results indicate that photosynthetically active radiation (PAR) is the primary and dominant driver of GPP (R<sup>2</sup> = 0.97), while temperature and precipitation are key factors significantly influencing carbon uptake. This study presents a comprehensive, integrated assessment of carbon, water, and energy fluxes at the national scale across Bangladesh, emphasizing the crucial role of climate variables in shaping these fluxes and offering valuable insights for climate-resilient land management and sustainable carbon strategies in monsoon-dominated regions.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101847"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145884536","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

Precise 3D crustal displacement retrieval in GCP-free environments: A geodetic and deep learning–assisted integration of InSAR and optical stereo data near the Denali Fault 无gcp环境下精确三维地壳位移反演：迪纳里断裂带附近InSAR和光学立体数据的大地测量和深度学习辅助整合

IF 4.5 Q2 ENVIRONMENTAL SCIENCES

Remote Sensing Applications-Society and Environment

Pub Date : 2026-01-01 Epub Date: 2026-01-28 DOI: 10.1016/j.rsase.2026.101894

Zahra Alizadeh Zakaria , Farshid Farnood Ahmadi , Hamid Ebadi

Retrieving accurate 3D deformation fields from Interferometric Synthetic Aperture Radar)InSAR(line-of-sight (LOS) measurements is challenging because LOS data provide only one-dimensional motion, and the absence of ground control points (GCPs) in remote regions makes reliable 3D reconstruction even more difficult. This study introduces a deep learning-assisted approach to map 3D deformation fields without GCPs, integrating Pleiades stereo imagery with InSAR techniques over a 20 km × 20 km segment of the Denali Fault, Alaska. Initial 3D displacements from Pleiades, derived through DEM differencing and COSI-Corr, eliminate GCPs and reveal vertical displacements of ±10 mm (uplift/subsidence), with horizontal displacements of ±6.4 mm (east-west) and ±5.9 mm (north-south), consistent with the fault's right-lateral strike-slip kinematics and local GNSS IGS14/NNR velocities (2019–2024). These Pleiades displacements, combined with PS-InSAR rates and geological features like slope and Terrain Ruggedness Index (TRI), serve as inputs to a U-Net model that transforms LOS data into 3D fields, expanding displacement ranges to ±20 mm across all components. We enhance U-Net estimates with geodetic optimization and use Monte Carlo Dropout (10 samples) to quantify uncertainties of 0.1–0.3 mm for east-west and 0.5–0.8 mm for north-south and vertical displacements. Validation of the model against Pleiades test data yields RMSEs of 1.17 mm (east-west), 1.46 mm (north-south), and 1.52 mm (vertical), with an RMSE of 1.98 mm against the vertical component of three local GNSS stations (6–50 km distance, IGS14/NNR frame, 2019–2024). This InSAR-Pleiades-deep learning method offers a scalable solution for 3D deformation monitoring, advancing seismic hazard assessment in GCP-free environments.

从干涉合成孔径雷达（InSAR）视距（LOS）测量中获取准确的3D变形场是具有挑战性的，因为LOS数据仅提供一维运动，并且在偏远地区缺乏地面控制点（gcp），使得可靠的3D重建变得更加困难。该研究引入了一种深度学习辅助方法来绘制没有gcp的3D变形场，将Pleiades立体图像与InSAR技术集成在阿拉斯加Denali断层的20公里× 20公里段上。Pleiades的初始三维位移通过DEM差分和cos - corr得到，消除了gcp，显示垂直位移为±10 mm（隆起/沉降），水平位移为±6.4 mm（东西方向）和±5.9 mm（南北方向），与断层的右侧走滑运动学和当地GNSS IGS14/NNR速度（2019-2024）一致。这些Pleiades位移，结合PS-InSAR速率和地质特征（如坡度和地形崎岖指数（TRI）），作为U-Net模型的输入，将LOS数据转换为3D场，将所有组件的位移范围扩展到±20毫米。我们通过大地测量优化来增强U-Net估计，并使用蒙特卡罗Dropout（10个样本）来量化东西位移0.1-0.3 mm，南北和垂直位移0.5-0.8 mm的不确定性。根据Pleiades测试数据验证模型的RMSE为1.17 mm（东西方向），1.46 mm（南北方向）和1.52 mm（垂直方向），其中三个本地GNSS站（6-50 km距离，IGS14/NNR框架，2019-2024）的垂直分量RMSE为1.98 mm。这种insar - pleades -深度学习方法为3D变形监测提供了可扩展的解决方案，推进了无gcp环境下的地震危害评估。

{"title":"Precise 3D crustal displacement retrieval in GCP-free environments: A geodetic and deep learning–assisted integration of InSAR and optical stereo data near the Denali Fault","authors":"Zahra Alizadeh Zakaria , Farshid Farnood Ahmadi , Hamid Ebadi","doi":"10.1016/j.rsase.2026.101894","DOIUrl":"10.1016/j.rsase.2026.101894","url":null,"abstract":"<div><div>Retrieving accurate 3D deformation fields from Interferometric Synthetic Aperture Radar)InSAR(line-of-sight (LOS) measurements is challenging because LOS data provide only one-dimensional motion, and the absence of ground control points (GCPs) in remote regions makes reliable 3D reconstruction even more difficult. This study introduces a deep learning-assisted approach to map 3D deformation fields without GCPs, integrating Pleiades stereo imagery with InSAR techniques over a 20 km × 20 km segment of the Denali Fault, Alaska. Initial 3D displacements from Pleiades, derived through DEM differencing and COSI-Corr, eliminate GCPs and reveal vertical displacements of ±10 mm (uplift/subsidence), with horizontal displacements of ±6.4 mm (east-west) and ±5.9 mm (north-south), consistent with the fault's right-lateral strike-slip kinematics and local GNSS IGS14/NNR velocities (2019–2024). These Pleiades displacements, combined with PS-InSAR rates and geological features like slope and Terrain Ruggedness Index (TRI), serve as inputs to a U-Net model that transforms LOS data into 3D fields, expanding displacement ranges to ±20 mm across all components. We enhance U-Net estimates with geodetic optimization and use Monte Carlo Dropout (10 samples) to quantify uncertainties of 0.1–0.3 mm for east-west and 0.5–0.8 mm for north-south and vertical displacements. Validation of the model against Pleiades test data yields RMSEs of 1.17 mm (east-west), 1.46 mm (north-south), and 1.52 mm (vertical), with an RMSE of 1.98 mm against the vertical component of three local GNSS stations (6–50 km distance, IGS14/NNR frame, 2019–2024). This InSAR-Pleiades-deep learning method offers a scalable solution for 3D deformation monitoring, advancing seismic hazard assessment in GCP-free environments.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101894"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146077729","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

High-resolution ground NO2 estimation at hyperlocal level using deep learning with Sentinel-2 and Sentinel-5P data 利用Sentinel-2和Sentinel-5P数据的深度学习在超局部水平上进行高分辨率地面二氧化氮估计

IF 4.5 Q2 ENVIRONMENTAL SCIENCES

Remote Sensing Applications-Society and Environment

Pub Date : 2026-01-01 Epub Date: 2025-12-05 DOI: 10.1016/j.rsase.2025.101819

Solaiman Khan , Anes Ouadou , Xing Song , Grant J. Scott

Nitrogen dioxide (NO₂) is a harmful air pollutant that can cause various health issues, including respiratory disease and lung infection. Monitoring of NO₂ is primarily dependent on expensive ground-based sensor systems. This research explores the potential of integrating imagery from Sentinel-2 and Sentinel-5P to estimate high-resolution ground NO₂ concentration at city and neighborhood levels. This study presents a two-stream deep learning model for NO₂ estimation. The model is flexible regarding data input, allowing the use of Sentinel-2 and Sentinel-5P in combination or as single inputs from either satellite. The model performance is assessed over Chicago using Microsoft Eclipse ground sensor data aggregated in three temporal frequencies: daily, monthly, and quarterly. The experimental results demonstrate that fusing both satellite sources outperforms single-source models, achieving R² = 0.66, MSE = 5.92, and MAE = 1.75 at the quarterly scale, compared to R² = 0.59 for Sentinel-2 only and R² = 0.31 for Sentinel-5P only models. The estimated NO₂ is found to be most reliable at the quarterly level, followed by the monthly. Performance decreases at finer temporal scales (R² = 0.61 daily), likely due to the short-term fluctuation of NO₂ concentration. This study reinforces the application of deep learning and remote sensing for air quality monitoring, especially in the absence of expensive ground sensor-based monitoring systems.

二氧化氮（NO2）是一种有害的空气污染物，可导致各种健康问题，包括呼吸系统疾病和肺部感染。二氧化氮的监测主要依赖于昂贵的地面传感器系统。本研究探索了整合Sentinel-2和Sentinel-5P图像的潜力，以估计城市和社区水平的高分辨率地面二氧化氮浓度。本文提出了一种用于NO2估计的双流深度学习模型。该模型在数据输入方面非常灵活，可以同时使用Sentinel-2和Sentinel-5P，也可以单独使用任一卫星的数据输入。模型性能在芝加哥使用微软Eclipse地面传感器数据进行评估，这些数据以三个时间频率聚合：每日、每月和每季度。实验结果表明，两卫星源融合优于单源模型，在季度尺度上实现R2 = 0.66, MSE = 5.92, MAE = 1.75，而仅Sentinel-2模型的R2 = 0.59，仅Sentinel-5P模型的R2 = 0.31。结果表明，季度NO2估算值最可靠，月度NO2估算值次之。在较细的时间尺度上（R2 = 0.61日），性能下降，可能是由于NO2浓度的短期波动。这项研究加强了深度学习和遥感在空气质量监测中的应用，特别是在缺乏昂贵的地面传感器监测系统的情况下。

{"title":"High-resolution ground NO2 estimation at hyperlocal level using deep learning with Sentinel-2 and Sentinel-5P data","authors":"Solaiman Khan , Anes Ouadou , Xing Song , Grant J. Scott","doi":"10.1016/j.rsase.2025.101819","DOIUrl":"10.1016/j.rsase.2025.101819","url":null,"abstract":"<div><div>Nitrogen dioxide (NO<sub>2</sub>) is a harmful air pollutant that can cause various health issues, including respiratory disease and lung infection. Monitoring of NO<sub>2</sub> is primarily dependent on expensive ground-based sensor systems. This research explores the potential of integrating imagery from Sentinel-2 and Sentinel-5P to estimate high-resolution ground NO<sub>2</sub> concentration at city and neighborhood levels. This study presents a two-stream deep learning model for NO<sub>2</sub> estimation. The model is flexible regarding data input, allowing the use of Sentinel-2 and Sentinel-5P in combination or as single inputs from either satellite. The model performance is assessed over Chicago using Microsoft Eclipse ground sensor data aggregated in three temporal frequencies: daily, monthly, and quarterly. The experimental results demonstrate that fusing both satellite sources outperforms single-source models, achieving R<sup>2</sup> = 0.66, MSE = 5.92, and MAE = 1.75 at the quarterly scale, compared to R<sup>2</sup> = 0.59 for Sentinel-2 only and R<sup>2</sup> = 0.31 for Sentinel-5P only models. The estimated NO<sub>2</sub> is found to be most reliable at the quarterly level, followed by the monthly. Performance decreases at finer temporal scales (R<sup>2</sup> = 0.61 daily), likely due to the short-term fluctuation of NO<sub>2</sub> concentration. This study reinforces the application of deep learning and remote sensing for air quality monitoring, especially in the absence of expensive ground sensor-based monitoring systems.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101819"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145738455","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

CLIP the landscape: Automated tagging of crowdsourced landscape images 剪辑景观：自动标记众包景观图像

IF 4.5 Q2 ENVIRONMENTAL SCIENCES

Remote Sensing Applications-Society and Environment

Pub Date : 2026-01-01 Epub Date: 2025-12-24 DOI: 10.1016/j.rsase.2025.101824

Ilya Ilyankou , Natchapon Jongwiriyanurak , Tao Cheng, James Haworth

We present a CLIP-based, multi-modal, multi-label classifier for predicting geographical context tags from landscape photos in the Geograph dataset—a crowdsourced image archive spanning the British Isles, including remote regions lacking POIs and street-level imagery. Our approach addresses a Kaggle competition task based on a subset of Geograph’s 8M images, with strict evaluation: exact match accuracy is required across 49 possible tags. We show that combining location and title embeddings with image features improves accuracy over using image embeddings alone. We release an efficient pipeline² that trains on a modest laptop, using pre-trained CLIP image and text embeddings and a simple classification head. Predicted tags can support downstream tasks such as building location embedders for GeoAI applications, enriching spatial understanding in data-sparse regions.

我们提出了一个基于clip的、多模态、多标签的分类器，用于从地理数据集中的景观照片中预测地理背景标签。地理数据集是一个跨越不列颠群岛的众包图像档案，包括缺乏poi和街道级图像的偏远地区。我们的方法解决了一个Kaggle竞赛任务，该任务基于geography的8M图像子集，并进行了严格的评估：需要在49个可能的标签中精确匹配精度。我们表明，与单独使用图像嵌入相比，将位置和标题嵌入与图像特征相结合可以提高准确性。我们发布了一个高效的pipeline2，它使用预训练的CLIP图像和文本嵌入以及一个简单的分类头，在一台普通的笔记本电脑上进行训练。预测标签可以支持下游任务，例如为GeoAI应用程序构建位置嵌入器，丰富数据稀疏区域的空间理解。

引用次数: 0

Quantifying high spatiotemporal impervious surface change with time series PlanetScope imagery 利用时间序列PlanetScope影像量化高时空不透水面变化

IF 4.5 Q2 ENVIRONMENTAL SCIENCES

Remote Sensing Applications-Society and Environment

Pub Date : 2026-01-01 Epub Date: 2026-01-31 DOI: 10.1016/j.rsase.2026.101890

Suiyuan Wang, Le Wang

Monitoring impervious surface change (ISC) at high spatiotemporal resolution is essential for understanding urban expansion and renewal, yet such studies are often limited by the high cost of obtaining high-resolution imagery. Additionally, there is a research gap between remote sensing-based ISC detection and urban topic-driven research, resulting in a lack of systematic methods that integrate spatial and temporal dynamics with quantitative metrics applicable to urban analysis. To address these challenges, we developed a two-component framework. We first detected ISC using low-cost CubeSat, PlanetScope (3 m, near daily) imagery, then aggregated ISC spatially and temporally to recognize urban change patterns. This involved developing a hybrid time series method where change locations and change types were identified with a U-Net model, followed by dating these changes using Continuous Change Detection. Next, we quantified ISC characteristics through four facets: distribution, speed, intensity, and structure within census block groups and across monthly and seasonal scales. We evaluated the framework using two study areas: Buffalo, New York, representing urban renewal, and Buckeye, Arizona, exemplifying urban expansion. Our first component demonstrated high performance in the ISC location (Overall Accuracy (OA) = 0.94) and good performance in change timing (OA = 0.77). Results from the second component highlighted contrasts between the two study areas, with Buffalo showing fragmented, self-motivated, and seasonally sensitive renewal, while Buckeye displayed spatially coherent, large-scale, and consistently expanding suburban growth. Quantitatively, Buckeye exhibited a ISC rate approximately ten times higher than that of Buffalo, whereas Buffalo showed greater seasonal variability in ISC activity, with peaks occurring in spring. In both cities, the majority of construction-related changes were concentrated in residential areas. We developed a two-component framework using PlanetScope imagery that offers a scalable, practical method for supporting high-resolution, time-sensitive urban studies.

高时空分辨率监测不透水面变化（ISC）对于理解城市扩张和更新至关重要，但此类研究往往受到高分辨率图像获取成本的限制。此外，基于遥感的ISC检测与城市主题驱动研究之间存在研究差距，导致缺乏将时空动态与适用于城市分析的定量指标相结合的系统方法。为了应对这些挑战，我们开发了一个由两部分组成的框架。我们首先使用低成本的CubeSat， PlanetScope（3米，近每日）图像检测ISC，然后在空间和时间上汇总ISC以识别城市变化模式。这涉及开发一种混合时间序列方法，其中使用U-Net模型识别变化位置和变化类型，然后使用连续变化检测确定这些变化的日期。接下来，我们通过四个方面量化了ISC的特征：分布、速度、强度和结构在人口普查块组和月度和季节性尺度。我们使用两个研究区域来评估该框架：代表城市更新的纽约州布法罗和代表城市扩张的亚利桑那州七叶树。我们的第一个组件在ISC位置上表现出高性能（总体精度(OA) = 0.94），在变化时间上表现良好（OA = 0.77）。第二部分的结果突出了两个研究区域之间的对比，布法罗显示出碎片化、自我激励和季节性敏感的更新，而七叶树则显示出空间连贯、大规模和持续扩展的郊区增长。从数量上看，七叶树的ISC率大约是水牛的10倍，而水牛的ISC活动表现出更大的季节性变化，春季出现峰值。在这两个城市，大部分与建筑相关的变化都集中在居民区。我们利用PlanetScope图像开发了一个双组件框架，为支持高分辨率、时间敏感的城市研究提供了一种可扩展、实用的方法。

{"title":"Quantifying high spatiotemporal impervious surface change with time series PlanetScope imagery","authors":"Suiyuan Wang, Le Wang","doi":"10.1016/j.rsase.2026.101890","DOIUrl":"10.1016/j.rsase.2026.101890","url":null,"abstract":"<div><div>Monitoring impervious surface change (ISC) at high spatiotemporal resolution is essential for understanding urban expansion and renewal, yet such studies are often limited by the high cost of obtaining high-resolution imagery. Additionally, there is a research gap between remote sensing-based ISC detection and urban topic-driven research, resulting in a lack of systematic methods that integrate spatial and temporal dynamics with quantitative metrics applicable to urban analysis. To address these challenges, we developed a two-component framework. We first detected ISC using low-cost CubeSat, PlanetScope (3 m, near daily) imagery, then aggregated ISC spatially and temporally to recognize urban change patterns. This involved developing a hybrid time series method where change locations and change types were identified with a U-Net model, followed by dating these changes using Continuous Change Detection. Next, we quantified ISC characteristics through four facets: distribution, speed, intensity, and structure within census block groups and across monthly and seasonal scales. We evaluated the framework using two study areas: Buffalo, New York, representing urban renewal, and Buckeye, Arizona, exemplifying urban expansion. Our first component demonstrated high performance in the ISC location (Overall Accuracy (OA) = 0.94) and good performance in change timing (OA = 0.77). Results from the second component highlighted contrasts between the two study areas, with Buffalo showing fragmented, self-motivated, and seasonally sensitive renewal, while Buckeye displayed spatially coherent, large-scale, and consistently expanding suburban growth. Quantitatively, Buckeye exhibited a ISC rate approximately ten times higher than that of Buffalo, whereas Buffalo showed greater seasonal variability in ISC activity, with peaks occurring in spring. In both cities, the majority of construction-related changes were concentrated in residential areas. We developed a two-component framework using PlanetScope imagery that offers a scalable, practical method for supporting high-resolution, time-sensitive urban studies.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101890"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147395829","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

High-resolution UAV-based multispectral imagery and machine learning modeling for water quality monitoring in eutrophic and oligotrophic reservoirs in Brazil 基于无人机的高分辨率多光谱图像和机器学习建模用于巴西富营养化和贫营养化水库水质监测

IF 4.5 Q2 ENVIRONMENTAL SCIENCES

Remote Sensing Applications-Society and Environment

Pub Date : 2026-01-01 Epub Date: 2026-01-30 DOI: 10.1016/j.rsase.2026.101907

Caio C.S. Mello , Daniel H.C. Salim , Gabriela R. Andrade , Alexandre F. Assunção , Gabriel Pereira , Camila C. Amorim

This study investigates the potential of Unmanned Aerial Vehicle (UAV)-based multispectral imagery to estimate water quality parameters in two Brazilian reservoirs with contrasting trophic states: the eutrophic Pampulha Reservoir and the oligotrophic Três Marias Reservoir. Random Forest regression models delivered strong predictive performance for chlorophyll-a (Chl-a) in both systems (R² > 0.86), while phycocyanin (PC) showed exceptional accuracy in Três Marias (R² = 0.98). Secchi depth also correlated well with spectral data (R² > 0.89), whereas turbidity was more difficult to model in the optically complex Pampulha Reservoir (R² = 0.76). Feature importance analysis provided critical insights into model behavior and the most influential spectral inputs. In Pampulha, Chl-a and Secchi depth relied on red-edge bands and derivative indices (e.g., 3BDA_1, 3BDA_MOD), while PC predictions were dominated by the red-edge band (64.4%). In Três Marias, PC estimation was overwhelmingly influenced by the KIVU index (89.2%), with green, red, and near-infrared bands contributing to other parameters. Blue bands and NDCI index supported FDOM and turbidity predictions in both reservoirs, although Pampulha required broader combinations due to higher optical complexity. The comparative analysis revealed that oligotrophic conditions in Três Marias favored more stable and interpretable spectral relationships, while Pampulha's elevated biological activity and variable constituents produced greater dispersion in feature relevance. These findings highlight the value of UAV-based multispectral sensing for generating high-resolution datasets and improving machine learning interpretability. The approach demonstrates strong potential for scalable water quality monitoring across diverse environmental conditions.

本研究探讨了基于无人机（UAV）的多光谱图像在两个营养状态不同的巴西水库（富营养化的Pampulha水库和贫营养化的Três Marias水库）中估计水质参数的潜力。随机森林回归模型在两个系统中对叶绿素-a （Chl-a）均具有较强的预测性能（R2 > 0.86），而藻蓝蛋白（PC）在Três Marias中表现出优异的准确性（R2 = 0.98）。Secchi深度与光谱数据的相关性也很好（R2 > 0.89），而浊度在光学复杂的Pampulha水库中更难建模（R2 = 0.76）。特征重要性分析提供了对模型行为和最具影响力的光谱输入的关键见解。在Pampulha， Chl-a和Secchi深度依赖于红边带和衍生指数（如3BDA_1、3BDA_MOD），而PC预测以红边带为主（64.4%）。在Três Marias中，PC估计绝大多数受到KIVU指数的影响（89.2%），绿色，红色和近红外波段对其他参数有贡献。蓝带和NDCI指数支持两个储层的FDOM和浊度预测，但由于光学复杂性较高，Pampulha需要更广泛的组合。对比分析显示，Três Marias的贫营养条件有利于更稳定和可解释的光谱关系，而Pampulha的生物活性升高和可变成分在特征相关性上产生了更大的分散。这些发现突出了基于无人机的多光谱传感在生成高分辨率数据集和提高机器学习可解释性方面的价值。该方法显示了在不同环境条件下可扩展的水质监测的巨大潜力。

{"title":"High-resolution UAV-based multispectral imagery and machine learning modeling for water quality monitoring in eutrophic and oligotrophic reservoirs in Brazil","authors":"Caio C.S. Mello , Daniel H.C. Salim , Gabriela R. Andrade , Alexandre F. Assunção , Gabriel Pereira , Camila C. Amorim","doi":"10.1016/j.rsase.2026.101907","DOIUrl":"10.1016/j.rsase.2026.101907","url":null,"abstract":"<div><div>This study investigates the potential of Unmanned Aerial Vehicle (UAV)-based multispectral imagery to estimate water quality parameters in two Brazilian reservoirs with contrasting trophic states: the eutrophic Pampulha Reservoir and the oligotrophic Três Marias Reservoir. Random Forest regression models delivered strong predictive performance for chlorophyll-a (Chl-a) in both systems (R<sup>2</sup> > 0.86), while phycocyanin (PC) showed exceptional accuracy in Três Marias (R<sup>2</sup> = 0.98). Secchi depth also correlated well with spectral data (R<sup>2</sup> > 0.89), whereas turbidity was more difficult to model in the optically complex Pampulha Reservoir (R<sup>2</sup> = 0.76). Feature importance analysis provided critical insights into model behavior and the most influential spectral inputs. In Pampulha, Chl-a and Secchi depth relied on red-edge bands and derivative indices (e.g., 3BDA_1, 3BDA_MOD), while PC predictions were dominated by the red-edge band (64.4%). In Três Marias, PC estimation was overwhelmingly influenced by the KIVU index (89.2%), with green, red, and near-infrared bands contributing to other parameters. Blue bands and NDCI index supported FDOM and turbidity predictions in both reservoirs, although Pampulha required broader combinations due to higher optical complexity. The comparative analysis revealed that oligotrophic conditions in Três Marias favored more stable and interpretable spectral relationships, while Pampulha's elevated biological activity and variable constituents produced greater dispersion in feature relevance. These findings highlight the value of UAV-based multispectral sensing for generating high-resolution datasets and improving machine learning interpretability. The approach demonstrates strong potential for scalable water quality monitoring across diverse environmental conditions.</div></div>","PeriodicalId":53227,"journal":{"name":"Remote Sensing Applications-Society and Environment","volume":"41 ","pages":"Article 101907"},"PeriodicalIF":4.5,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147395832","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