ABSTRACT Vegetation critically regulates permafrost stability and carbon-water cycles in the climate-sensitive Da Xing’anling Mountains (DXL), Eurasia’s southern permafrost margin. Warming threatens ecosystem integrity and ground ice here. Detailed distribution of vegetation type is essential for accurately identifying the extent of permafrost. Moreover, it is also necessary for modelling ecosystem dynamics and carbon cycles in this vulnerable region. This study generated a high-resolution (30 m) distribution of vegetation type for a representative cold region within the DXL using multi-source remote sensing data from 2018 to 2023, combined with field surveys conducted in July – August 2023 and July – August 2024. We focused on the Genhe River Basin (GRB) on the western slope of the DXL. This area exemplifies the region’s diverse vegetation types, including forests, wetlands, and grasslands, alongside extensive permafrost. Integrating extensive field vegetation surveys with multi-source remote sensing data, we employed a Random Forest (RF) approach to systematically incorporate and evaluate three feature selection methods for performance optimization. The optimal classification result reached an overall accuracy of 0.78 and a Kappa coefficient of 0.73. The result demonstrates high user accuracy for key types such as wetland (0.94), deciduous broad-leaved forest (0.82), mixed forest (0.65), and deciduous needle-leaved forest (0.74). This provides reliable foundational data for research requiring precise wetland delineation, such as permafrost distribution mapping and ecosystem studies. This study has some limitations. Specifically, sampling ambiguities arise from inconsistencies in multi-source data integration, particularly pronounced in dynamic vegetation areas. Furthermore, discrepancies in mixed forest classification outcomes stem from conflicting international and domestic standards, which respectively set the thresholds for dominant species at 60% and 75%. This product facilitates ecosystem restoration, carbon source-sink analysis, regional permafrost mapping, and water cycle management, while furnishing foundational data for classifying discontinuous permafrost vegetation in the DXL.

ABSTRACT Accurate 3D building reconstruction is crucial for advancing urban digital twinning, city planning, and sustainable development. As a key architectural component, rooftops facilitate urban energy management and inform urban morphological analysis. Consequently, achieving precise and scalable rooftop reconstruction has emerged as a key research focus in recent years. Point clouds, with their ability to preserve detailed geometric structures, are well suited for this task. However, existing methods predominantly target synthetic rooftop datasets, which lack architectural diversity and often require high-quality point clouds as input. These limitations hinder their applicability to large-scale, real-world urban environments characterized by varied rooftop designs and noisy or sparse data. To address these challenges, we propose a novel end-to-end framework for rooftop wireframe reconstruction from airborne laser scanning (ALS) point clouds. Our approach introduces a multi-scale local feature descriptor optimized for rooftops to enhance per-point geometric feature extraction. Then, a hypergraph-based attention fusion module integrates these features. After comprehensive feature learning by a robust backbone, initial corner detection is followed by a Transformer- and EdgeConv-enhanced edge classification mechanism that models topological relationships through long-range dependencies. Experiments on the large-scale real-world Building3D dataset demonstrate significant improvements over the baseline, with corner accuracy improved by 35% on the Entry-level subset and 41% on the Tallinn subset. Qualitative comparisons further reveal superior wireframe fidelity, underscoring the method’s potential to support digital twinning, urban management, and economic development in smart city initiatives.

ABSTRACT This study examines the efficiency of three segmentation algorithms – Felzenszwalb, Simple Non-Iterative Clustering (SNIC), and Multiresolution Segmentation (MRS) – in conjunction with four deep learning models (2D CNN, 3D CNN, Hybrid CNN, and Vision Transformer (ViT)) for evaluating building damage after the 2020 Beirut port explosion. To validate the generalizability of the models, an additional case study was conducted using the large-scale xBD dataset. Utilizing WorldView-2 imagery for the Beirut case, the segmentation performance and classification accuracy were assessed across four damage categories: destroyed, major damage, minor damage, and no damage. The methodology first divides the pre-disaster image into segments using these three algorithms. Subsequently, image classification was performed on the stacked pre- and post-disaster image patches. The final classification is performed using decision-level fusion, which is based on the majority vote of all pixels classified by the deep learning models within the image objects obtained from segmentation. The segmentation results showed that SNIC provided the most balanced performance, achieving the highest F1-score (0.5589). Regarding damage classification for the Beirut dataset, the Vision Transformer (ViT) significantly outperformed all CNN variants. The combination of ViT with SNIC achieved the highest overall accuracy of 95.1% and an overall F1-score of 0.950. This approach yielded superior class-wise F1-scores, reaching 0.978 (destroyed), 0.943 (major), 0.957 (minor), and 0.945 (No Damage). On the xBD dataset, the ViT model confirmed its robustness by achieving the highest accuracy (81.4%) and F1-score (0.789), outperforming the CNN models and other state-of-the-art benchmarks. These results emphasize the efficacy of the ViT combined with SNIC for quick, accurate, and generalizable damage assessments, providing essential insights for disaster management and recovery efforts.

ABSTRACT Sensible heat flux is a key component of the surface energy flux that directly influences the urban energy balance. Changes in Land Use and Land Cover (LULC) affect heat transfer between the surface and the atmosphere, thereby influencing the distribution of available energy into latent, sensible, and ground heat. This study aimed to analyse the impact of LULC changes on sensible heat flux in the Itupararanga Environmental Protection Area, a remnant of the Brazilian Atlantic Forest biome, from 1986 to 2021. The significant increase in LULC categories such as temporary crops and urbanized areas, along with the reduction in pasturelands, substantially impacted sensible heat in the study area. The expansion of temporary crop areas contributed to a decrease in sensible heat during the study period, as these areas tend to retain more moisture due to the irrigation demands of the crops. Conversely, the growth of urbanized areas exhibited an unusual pattern, with sensible heat showing a decline despite these areas having the highest sensible heat averages overall. Notably, the year 1994 recorded the lowest sensible heat values, attributed to atypical weather conditions.

ABSTRACT Accurate wall-to-wall mapping of tree species composition (TSC) is essential for effective forest management. However, distinguishing species-level information from satellite imagery remains a challenge due to the coarse spatial resolution of open-access satellite imagery. In this study, we present the first systematic evaluation of spatial resolution enhancement and multi-seasonal data fusion for deep learning (DL)-based TSC mapping using Sentinel-2 imagery. Specifically, we assessed: (1) the impact of different spatial resolutions and enhancement methods, comparing native 20 m Sentinel-2 imagery against bilinear resampled imagery at 10 m and 5 m, super-resolution (SR)-enhanced imagery at 10 m and their combined use; (2) the contributions of multi-seasonal imagery and auxiliary environmental data (climate, topography); and (3) the effectiveness of a novel multi-source multi-seasonal fusion (MSMSF) method for integrating seasonal and environmental datasets. Our results demonstrated substantial improvements (7% higher R a d j 2 ) when increasing spatial resolution from 20 m to 10 m and achieved the best result (RMSE = 0.120, R a d j 2 = 0.731) by combining bilinear resampled 5 m and SR-enhanced 10 m datasets. Additionally, our proposed MSMSF module and multi-seasonal data outperformed the best single-season model by >5% in terms of R a d j 2 . These findings establish a new benchmark for DL-based TSC mapping and highlight the novelty of combining resolution enhancement with a detail-preserving fusion strategy to enable scalable, high-precision forest inventories using freely available satellite data.

ABSTRACT In recent years, the use of Unmanned Aerial Vehicles (UAVs) for remote sensing and aerial photogrammetry has surged, owing to their affordability and ability to capture high-resolution data in hard-to-reach areas. However, the effectiveness of these platforms can be limited by environmental factors such as turbulence and wind gusts, which may destabilize the aircraft and compromise the proper exposure of images. To address this issue, this work presents a terrain survey planner that considers the environmental conditions derived from a meteorological forecast. The system employs Computational Fluid Dynamics (CFD) simulations to generate high-resolution wind predictions for a given study area. This data is integrated into a state-of-the-art UAV simulator to estimate aircraft behaviour at various locations. Additionally, it incorporates an empirically calibrated camera model to predict sensor performance based on solar radiation estimates. With this information, a multi-objective optimization is performed, computing the optimal path and camera settings to mitigate the impact of wind on photogrammetry. Results highlight the significant impact of wind and poor lighting on motion blur, emphasizing the need to carefully plan not only the inspection path but also the time and date for correct image exposure.

ABSTRACT In relatively dry areas, due to lower soil moisture and less vegetation coverage, land surface microwave radiation comes from a certain depth of soil, while infrared skin temperature is only sensitive to the thin layer of the land surface. The inconsistencies in the detection depth of the microwave and infrared can lead to significant differences in retrieved emissivity between day and night. To improve the instantaneous microwave emissivity retrieval over barren areas, the land surface effective temperature calculation method was proposed based on AMSR2, which constructed the relationship between effective temperature and microwave brightness temperatures (TBs) in hours, approximating that monthly mean skin temperature equals monthly mean effective temperature and ignoring the change in emissivity over a month. The results showed that the effective temperature from 10.65 to 89 GHz had significantly smaller diurnal amplitude than skin temperature, and the lower the frequency, the smaller the amplitude. The effective temperature was then applied to the instantaneous emissivity inversion, and it was shown that this method remarkably reduced the emissivity difference between day and night, with the mean difference at the magnitude of 10^-3.

ABSTRACT Phytoplankton underpin marine food webs and carbon cycling, converting dissolved carbon dioxide into organic matter and exporting it to deeper layers. However, these organisms are sensitive to environmental changes that affect their growth and community structure differently, which may be represented by their taxonomic structure or cell size categories. Consequently, there is increasing interest in developing and improving satellite-based models for estimating the abundance of phytoplankton size classes (PSCs) and different taxonomic groups. Satellites can reliably estimate two key properties related to phytoplankton biomass and ocean dynamics: chlorophyll-a concentration (Chla), the primary pigment of phytoplankton, and sea surface temperature (SST), which is associated with water masses and often related to nutrient availability. In this study, we tested different approaches and developed regional models to retrieve PSCs from satellite data. The regional models were fitted to the South Brazil Bight (SBB) in the Southwestern Atlantic Ocean. The in situ training and validation datasets were obtained from oceanographic cruises conducted in the SBB during 2019–2022. We applied different model parameterisation schemes to compare SST-independent and SST-dependent models with both global and regional fits. The models were applied to both in situ data and satellite observations from Ocean and Land Colour Instrument (OLCI) sensors on board Sentinel 3A and 3B satellites, alongside the Multi-scale Ultra-high Resolution (MUR) SST product. The regional SST-dependent approach consistently outperformed alternatives across all size classes, achieving correlation coefficients (ρ) greater than 0.7, bias less than 0.14, and mean absolute error (MAE) of less than 0.36. By comparison, the regional SST-independent approach (ρ > 0.54, bias < 0.17, MAE < 0.38) and the global SST-dependent approach (ρ > 0.59, bias < 0.11, and MAE < 0.40) showed weaker performance. These results highlight the importance of regional SST-dependent models for improving PSC estimation accuracy in the SBB and similar regions where SST variability affects nutrient availability, phytoplankton biomass, and community structure.

ABSTRACT Currently, hyperspectral image classification (HSIC) faces two major challenges: the attenuation of knowledge transfer efficiency caused by cross-domain distribution differences, and the insufficient generalization ability of representation learning under a single visual modality. Although scholars have attempted to address these challenges, there are still shortcomings in achieving deep cross-domain alignment and multimodal collaboration. Therefore, this paper proposes a dual stage adversarial domain adaptation (DSADA) HSIC framework which incorporates multi-modal learning. Specifically, this article proposes a dual stage adversarial learning framework that significantly alleviates the distribution shift between the source domain and the target domain and enhances the model’s cross-domain adaptability. In addition, the introduction of label text modality optimizes the similarity between image and text prototypes through cross-modal alignment mechanism to fully explore the complementary information between modalities and enhance feature discriminative power. This paper conducts systematic experimental verification on three hyperspectral image datasets, and the results show that DSADA can effectively alleviate cross-domain migration challenges in few shot scenarios and significantly improve the accuracy of HSIC.

ABSTRACT With the advancement of real-time object detection technology, maintaining high detection accuracy for small objects across multiple scales remains challenging. Conventional convolutional neural networks (CNNs) struggle to effectively capture multi-scale features, often failing to meet detection requirements. This study proposes RMRN-DETR, an optimized remote sensing image detection network based on multi-dimensional real-time detection and domain adaptation. First, we introduce a Multi-dimensional Real-time detection module (MR) to achieve efficient end-to-end accuracy improvement. Second, a Multi-dimensional Domain Adaptation module is proposed to address feature fusion across different scales, effectively capturing both low-level and high-level semantic information in a multi-scale hierarchy. Finally, a novel loss boundary regression module is introduced to enhance bounding box regression accuracy, precisely reflecting the discrepancy between predicted and ground-truth boxes. Experimental results demonstrate a 1.8% accuracy improvement over the baseline on the ROSD dataset and a 2.9% gain on the DIOR dataset. The proposed method significantly enhances the detection accuracy and efficiency of small objects in remote sensing images, demonstrating strong adaptability to complex multi-scale scenarios.