The present research was carried out to retrieve dissolved oxygen (DOX) using the Copernicus Marine Services products from the Irish transitional and coastal waters. To achieve the research goal, the study developed and validated 2101 machine learning (ML)/artificial intelligence (AI) (supervised learning, stacking-ensembles, equations, and voting-based ensembles) and statistical models using multi-level (Level-3 and Level-4) Sentinel-3 OLCI (S3-OLCI) and Multi-sensor (MS) remote sensing (RS) datasets in conjunction with in-situ and modelled DOX datasets. While supervised models (e.g., K-nearest neighbours, Gradient boosting, and Extra decision trees) excelled in the training phase (EPA: MSE ≤ 0.03 with CI ± 0.02; Modelled: MSE ≈ 0 with CI ± 0) but showed limited generalizability on independent validation datasets (2022-2023), indicating poor model accuracy and sensitivity (EPA-2022: R2 = -0.03 - 0.16; EPA-2023: R2 = -0.09 - 0.1; Modelled-2022: R2 = 0.37 - 0.53; Modelled-2023: R2 = -1.39 - -0.26). In terms of product, S3-OLCI outperformed MS data with low uncertainty, whereas spatio-temporal analysis showed the highest DOX in inshore/semi-enclosed bays and the lowest offshore. Overall, the results underscore that model performance is determined by methodological characteristics rather than model quantity. Despite the validation challenges, the results highlight key difficulties in retrieving optically inactive water quality (WQ) indicators like DOX using RS and ML/AI approaches. The findings of the research could be effective for supporting the mapping of baseline oxygen conditions, the application of ML/AI techniques to retrieve WQ indicators from RS products and their further technological advancement, such as managing the anthropogenic water cycle (i.e., human-altered hydrological and nutrient dynamics).

Land use and land cover (LULC) change is a major anthropogenic factor influencing flood behavior and hydrological processes. This systematic review synthesizes two decades (2005-2025) of research on hydrological modeling approaches used to assess flood responses under LULC transitions. A total of 114 publications were retrieved from the Scopus database, and after applying PRISMA-based screening, 78 peer-reviewed studies were analyzed using bibliometric and content mapping. The review categorizes hydrological models by spatial scale, process representation, and sensitivity to LULC dynamics. Findings consistently indicate that urban expansion, deforestation, and vegetation loss intensify surface runoff, peak flow, and flood frequency. Despite advancements, significant challenges remain particularly related to data scarcity, model calibration, and the limited integration of socio-economic variables. Emerging tools such as Remote Sensing (RS), Geographic Information Systems (GIS), and machine learning especially within platforms like Google Earth Engine (GEE) enhance LULC detection accuracy and flood prediction capability. The study proposes an integrated decision framework linking bibliometric trends with model selection strategies, enabling researchers to align model choice with data availability and landscape characteristics. Overall, this review emphasizes the importance of interdisciplinary, data-driven modeling to strengthen flood resilience in rapidly transforming land systems.

High-resolution daily air temperature estimation over China: An explainable stratified stacking ensemble approach

The safety and stability of dams depend strongly on their foundation conditions, particularly in seismically active regions where loose sandy and silty soils are prone to liquefaction. This study evaluates the liquefaction susceptibility of the Sultan Dam foundation in Wardak Province, Afghanistan, where a partial collapse occurred in 2005. Fourteen exploratory cores were investigated using in situ standard penetration tests, supported by satellite data analysis of active faults, geomorphology, seismicity, and geology. Subsurface characterization revealed bedrock composed of quartzite and granodiorite overlain by silty sand and clay with variable density. Liquefaction analysis identified three borehole locations and an estimated 3,000 m2 of foundation area at risk. These results provide essential input for seismic hazard assessment, rehabilitation of the Sultan Dam, and planning of new infrastructure in earthquake-prone regions.

Understanding smart thermostat adoption: Housing, HVAC, and socio-economic traits in the U.S.

Large-size very-high-resolution (VHR) remote sensing imagery has emerged as a critical data source for high-precision vector mapping of multi-scale geographical elements such as building, water, road and etc. When dealing with the large-size image, due to the limited memory of GPU, the deep learning-based vector mapping methods often employ the sliding block strategy. This inevitably leads to the degenerated performance because of the stitching difficulty of the sliding blocks' vector mapping results. Therefore, it is necessary to conduct full-scope vector mapping via mining the consistent cue in large-size remote sensing imagery. To this end, this paper presents a novel global context-aware local point optimization method. To leverage the global context, this paper proposes a novel pyramid fusion network (PFNet) to conduct semantic segmentation of the large-size image in an end-to-end manner. Under the constraint of the global semantic segmentation result, a new inflection-point perception network (IPNet) is proposed to generate a set of stable points to depict the boundary of each element. Extensive experiments on building, water and road datasets, where each image has over 100 million pixels, show that our method obviously outperforms the existing methods.

Large-scale monitoring networks employing remote sensors have harnessed big data to evaluate conservation efforts on a global scale. While dryland ecosystems are anticipated to expand, and management activities like rewilding and habitat restoration are increasing, the use of data streams and modern analytical methods to plan conservation interventions and quantify their effectiveness remains limited. We recommend establishing a global network for dryland practitioners to bridge critical data gaps, providing two real-world examples. A pilot study in the Kingdom of Saudi Arabia shows how coordinated use of multiple remote sensors and a standardised data pipeline could improve interoperability and facilitate the use of more accurate ecological models. Likewise, the Wildlife Observatory of Australia demonstrates that robust metadata and shared analytical frameworks enable the effective integration of diverse datasets using hierarchical occupancy models. Key steps to build this network include forming a steering committee, engaging stakeholders from various backgrounds, piloting projects in different regions, agreeing on protocols and exploring seed funding opportunities.

Precise extraction of buildings from high-resolution remote sensing images is essential for urban analysis and land management. However, accurately extracting buildings as a region of interest (ROI) from remote sensing (RS) images remains challenging. This difficulty arises from the spectral similarity of other objects, such as roads, cars, or trees, along with limited information on building boundaries and small buildings. Traditional image segmentation methods often rely on a fixed threshold value, making optimisation difficult in cases of over-segmented regions. As a result, region merging is subsequently performed on the region adjacency graph (RAG). Consequently, building segmentation in RS images becomes problematic and can lead to inaccurate boundary delineation or region classification. To overcome these limitations, we propose a novel segmentation approach that incorporates an adaptive thresholding optimisation technique and a merging criterion (MC) based on deep features extracted via a convolutional neural network (CNN)-based AttentionU-Net architecture. This ensures that merging decisions are guided by intrinsic region-level characteristics and refined through deep feature representations. Beginning with initial segmentation generated by the simple linear iterative clustering (SLIC) algorithm, the AttentionU-Net architecture is applied to high-resolution RS images to extract deep features, respectively. As a result, our approach combines both low and high-level feature information, reducing misalignment during merging and enhancing traditional region merging strategies. To validate this approach, the WHU buildings’ RS images dataset was utilised. Experimental results demonstrate that our approach achieves superior segmentation accuracy in building delineation while eliminating the need for rigid thresholds. Finally, the results were compared with those obtained using the multiresolution segmentation (MRS) algorithm implemented in eCognition software on the same WHU buildings RS images, where our approach performs better. Specifically, the proposed approach attained a higher segmentation accuracy, with an F-measure of 0. 91 and a goodness of segmentation score Gs of 0.92, compared to 0.52 and 0.83, respectively, achieved by the MRS algorithm.

Remotely sensed evapotranspiration-based ensemble streamflow modeling in an ungauged watershed under climate and land use/cover change, North Korea.

Multi-source remote sensing data classification refers to the process of categorizing ground objects by integrating complementary strengths of multiple remote sensing data, such as hyperspectral image (HSI), light detection and ranging (LiDAR) and synthetic aperture radar (SAR) data. However, current Mamba-based multisource remote sensing data classification approaches rely on fixed scanning patterns that are inadequate in characterizing spectral-spatial information. Additionally, current fusion techniques adopt concatenation or attention-based fusion rules without considering the complementary characteristics between different modalities. To address these limitations, we propose a spectral-spatial dynamic scan Mamba (SDSM) for multi-source remote sensing data classification. Specifically, a dynamic scan Mamba network is proposed to extract the spectral-spatial features of multi-source remote sensing data, in which a dynamic scan module is designed to adaptively capture the important spatial and spectral information. Furthermore, a bidirectional cross-modal fusion rule is proposed to merge the extracted features, in which a global-local frequency feature extraction module is designed to extract the salient structural features of multi-source remote sensing data as clues to guide heterogeneous feature fusion. Comprehensive experiments on four multi-source remote sensing datasets, i.e., MUUFL, Augsburg, Italy and Yellow River, demonstrate that the proposed method outperforms other state-of-the-art methods with respect to quantitative and qualitative results. The code of this article is available at https://github.com/PuhongDuan/SDSM.

Label-free optical absorption microscopy techniques continue to evolve as promising tools for label-free histopathological imaging of cells and tissues. However, critical challenges relating to specificity and contrast, as compared to current gold-standard methods continue to hamper adoption. This work introduces Photon Absorption Remote Sensing (PARS), a new absorption microscope modality, which simultaneously captures the dominant de-excitation processes following an absorption event. In PARS, radiative (auto-fluorescence) and non-radiative (photothermal and photoacoustic) relaxation processes are collected simultaneously, providing enhanced specificity to a range of biomolecules. As an example, a multiwavelength PARS system featuring UV (266nm) and visible (532nm) excitation is applied to imaging human skin, and murine brain tissue samples. It is shown that PARS can directly characterize, differentiate, and unmix, clinically relevant biomolecules inside complex tissues samples using established statistical processing methods. Gaussian mixture models (GMM) are used to characterize clinically relevant biomolecules (e.g., white, and gray matter) based on their PARS signals, while non-negative least squares (NNLS) is applied to map the biomolecule abundance in murine brain tissues, without stained ground truth images or deep-learning methods. PARS unmixing and abundance estimates are directly validated and compared against chemically stained ground truth images, and deep learning based-image transforms. Overall, it is found that the PARS unique and rich contrast may provide comprehensive, and otherwise inaccessible, label-free characterization of molecular pathology, representing a new source of data to develop AI and machine learning methods for diagnostics and visualization.

ABSTRACT Current recommendations for supervised machine learning classification in Remote Sensing advocate for using only high-quality reference data, which is often associated to pure pixels or endmembers. This focus, however, overlooks a fundamental challenge: real-world environmental images are rarely composed primarily of pure pixels. As a result, using only pure pixels as training data can leave a significant portion of the image unrepresented and hinder adequate classification. To demonstrate this effect, we introduce the Reference Sample Selection (RSS) approach. RSS systematically varies pixel purity in the training dataset to assess its impact on classification results. Pixel purity is determined based on a finer spatial resolution image. In this study, we present a case study within a select region of the Brazilian Amazon Rainforest. We applied RSS to the commonly used medium spatial resolution data from the Sentinel-2 MultiSpectral Imager (MSI) to target four land cover types: forest, water, grassland, and bare soil. The analysis used three common shallow classifiers: K-Nearest Neighbours (KNN), Support Vector Machines (SVM), and Random Forests (RFR). Our results demonstrate that including pixels with varied purity levels can significantly alter classification accuracy, depending on the land cover class. This finding challenges the conventional definition of reference data quality and highlights the need for using training samples that represent the entire image, not just its purest components. This method is readily applicable to a wide range of Remote Sensing studies. The source code and used data are available at https://github.com/paeslemesa/rss_approach/ and https://www.kaggle.com/datasets/sabrinacorra/uruara-s2msi-samples.

To the best of our knowledge, this paper is the first to integrate fractal signal processing with vision graph neural networks, establishing a new graph representation learning paradigm consistent with fractal dynamics. Building on this foundation, we propose a Fractal-domain Vision Graph Neural Network (FD-ViG). Specifically, FD-ViG includes: (i) a Fractal-Domain Learning Module that maps images into the fractal-domain using local Hölder exponents and the Singularity Power Spectrum (SPS), enabling fractal-spatial feature fusion; (ii) a Fractal Graph Construction Module that adaptively generates a topology by combining semantic attention with fractal similarity in the fractal feature space; and (iii) a Graph Propagation Module with power-law multi-scale propagation to realize cross-scale diffusion and aggregation, enabling coupled texture-structure learning. Experiments on UCMerced, RSSCN7, and SIRI-WHU achieve overall accuracies of 91.75%, 89.52%, and 92.78%, respectively. Compared with representative vision graph models such as ViG, WiGNet, and ViHGNN, our method achieves consistent improvements over prior methods across all three datasets, while remaining lightweight (2.6M parameters). Moreover, despite having far fewer parameters than ResNet-18, our model yields competitive or better performance on two datasets, and further demonstrates strong generalization ability in cross-dataset evaluation on SAR imagery. This work provides a principled and effective bridge between fractal theory and graph deep learning, benefiting interpretable remote sensing scene understanding under complex textures and structures.

Computer Vision–Based Crop Disease Detection with Potential Integration into Remote Sensing Systems

ABSTRACT Purpose This study assessed the accumulation of heavy metals (Pb, Cd, Cu, Zn, Cr, Fe, and Ni) in six halophytic species and evaluated their phytoremediation potential. Additionally, remote sensing indices, including NDVI (Normalized Difference Vegetation Index), SAVI (Soil Adjusted Vegetation Index), NDSI (Normalized Difference Salinity Index), and land surface temperature, were used to analyze soil salinity patterns and vegetation distribution. Materials and methods Polluted saline soils and plant samples were collected from ten sites along Egypt’s Mediterranean coast. Heavy metals were determined after acid digestion, and bioconcentration (BCF) and translocation (TF) factors were calculated. Remote sensing data were used to assess vegetation health and environmental stress. Results Showed significant spatial variability in salinity and metal concentrations, with the highest salinity recorded at Site 3. Mesembryanthemum cordifolium accumulated high levels of Cu, Fe, Ni, and Cr in shoots, while Suaeda vermiculata showed high Cd and Zn accumulation in roots. Bassia indica roots contained the highest Pb levels. Most BCF values were below 1, except for Cd and Zn in S. vermiculata, while TF values generally exceeded 1. Soil alkalinity likely reduced the mobility of some metals. Conclusion These findings highlight the role of halophytes as bioindicators and their potential in phytoremediation, supporting their use in sustainable restoration of saline, contaminated ecosystems.

This study evaluates flood vulnerability in the Chennai City Region, Tamil Nadu, using remote sensing and GIS techniques to guide urban development planning. With rapid urbanization and recurrent flooding, Chennai faces heightened risks from heavy monsoon rains, inadequate drainage, and encroachment on natural floodplains. Sentinel-2 and Landsat satellite imagery, combined with GIS data such as digital elevation models (DEM) and land-use maps, were used to classify land cover, map flood extents, and assess flood vulnerability. A multi-criteria evaluation using Analytical Hierarchy Process (AHP) identified key vulnerability factors, including population density, elevation, land use, and proximity to water bodies and drainage infrastructure. The study also conducted sensitivity analyses, including map-removal sensitivity analyses, to quantify the impact of individual parameters on flood vulnerability mapping. The findings reveal significant urban expansion (85% of the area) and widespread impermeable surfaces contributing to high surface runoff and limited infiltration. Topographic Wetness Index (TWI), drainage density, slope, and distance from streams were used to assess flood-prone zones further. The Normalized Difference Vegetation Index (NDVI) was calculated to evaluate the extent and health of vegetation affected by flooding. At the same time, DEMs and terrain analysis provided insights into low-lying areas with higher flood vulnerability. The research identified flood-prone zones classified into low, medium, and high-risk areas, covering 24.4%, 50.2%, and 25.4% of the study region, respectively. These results underscore the need for sustainable land-use management, improved drainage infrastructure, and climate-resilient urban development strategies to mitigate flood vulnerability in Chennai. The comprehensive assessment aims to support flood vulnerability management efforts and urban resilience planning in the region.

Urbanization and industrial development have intensified pressure on forest reserves in cities and settlement areas, increasing the vulnerability of urban forests to land-use change. This study examines the effects of land-use changes on forest encroachment in Akure City, Ondo State, Nigeria. A remote sensing and GIS-based approach was employed using satellite imagery obtained from the United States Geological Survey (USGS) for 2000, 2010, and 2020. The images were processed and analyzed using ArcMap 10.5 and Environment for Visualizing Images (ENVI) software to assess changes in land-use and land-cover patterns over the study period. The findings reveal notable changes in Akure City’s land-use structure, with less dense forest accounting for 41%, 66%, and 58% in 2000, 2010, and 2020, respectively, while built-up areas increased from 1.53% in 2000 to 3.28% in 2010 and 5.28% in 2020. The loss of natural vegetation also increased substantially from 4.4 km² in 2000 to 9.34 km² in 2010 and 12.01 km² in 2020. Surface water bodies were nearly absent in 2000 but accounted for 0.93 km² in 2010 and 0.66 km² in 2020. These findings indicate that continued urban expansion and associated land-use changes are likely to accelerate forest encroachment in Akure City. The study contributes to urban environmental management by highlighting the need for government authorities at all levels to implement sustainable forest conservation strategies and integrated city management policies.

This work reports the experimental demonstration of a dual-wavelength L-band fiber ring laser for remote sensing applications. The system incorporates a polarization-sensitive semiconductor optical amplifier as the gain medium and two fiber Bragg gratings placed 25 km away from the laser cavity using standard single-mode fiber that serve both as wavelength-selective elements and sensing heads. Wavelength switching between single- and dual-channel lasing configurations is enabled by a simplified two-paddle motorized polarization controller. The system achieves optical signal-to-noise ratios exceeding 55 dB and power differences between lasing lines as low as 0.01 dB. To ensure long-term stability, an automatic control algorithm dynamically adjusts the polarization state in real time, compensating for environmentally induced polarization drift. The proposed setup provides a compact and robust solution for polarization-based wavelength switching in fiber lasers, with applications in the field of remote optical sensing.

Geographic Information Systems (GIS) are adequate for analyzing complex scientific and spatial phenomena in geography education. Google Earth is a geographic information tool for GIS-based learning in phase E with the topic of Remote Sensing Mapping and GIS Basics. This tool allows students to engage in simpler map creation. This study aims to determine the use of Google Earth media based on the Project Based Learning model on student learning outcomes at SMAN 6 Solok-Selatan. The research method used is quantitative with a quasi-experimental approach. Sampling was carried out using a purposive sampling technique. The sample used consisted of a control class and an experimental class with a total of 72 students. Data collection used pretest and posttest instruments in the form of essays. Data analysis used an independent sample t-test to determine the difference in pretest-posttest results between the two classes. The results showed a difference in learning outcomes of 87.75 for the experimental class, and 87.27 for the control class. This is proven by the value of the difference test results (t), from the calculation above obtained (2-tailed) of 0.000 smaller than 0.05, so it can be concluded that H₀ is rejected and Hₐ is accepted. The researcher's conclusion is that Google Earth media is more effective in improving student learning outcomes compared to conventional presentation media in Geography learning.

Fine-grained object detection (FGOD) is crucial for identifying visually similar sub-categories in remote sensing imagery. However, existing detectors suffer from severe supervision imbalance because static label assignment strategies assign a fixed number of positive samples to all sub-categories and targets. To address this challenge, this paper presents Cumulative Quality-based Dynamic Assignment (CQDA), a fine-grained aware label assignment algorithm that dynamically calculates the optimal positive budget for each instance based on its cumulative alignment quality. Moreover, to further resolve feature-space confusion, this paper introduces two modules: a frequency-decoupled enhancement algorithm to sharpen discriminative features, and an orthogonal classification head to maximize inter-class separability. Integrated into the KFIoU framework, extensive experiments demonstrate that the proposed method consistently achieves performance improvements of 4.2, 15.8, and 35.3 in mAP@0.5 on the fine-grained oriented object detection datasets FAIR1M-v2, MAR20, and ShipRSImageNet, respectively.