Drought is an abnormal hydrological phenomenon characterized by high frequency, long duration, and extensive effect, exerting significant effects on both the ecological environment and the socioeconomic system. This study uses the Standardized Precipitation Index (SPI) to identify drought events in the Yellow River Basin from 1961 to 2021. The identified drought events were analyzed from a three-dimensional perspective to examine their spatial–temporal continuity. Furthermore, this study reveals the spatial and temporal evolution of key drought characteristics, including drought duration, drought area, intensity, density, and centroid position (latitude and longitude). The results indicate the following: (1) A total of 77 drought events have occurred in the Yellow River Basin since 1961, with an average duration of 8.7 months. Of these events, 55.8% lasted longer than the average duration. The average drought area was 50 855 km², and the average drought density was 0.95, indicating a severe overall drought situation. (2) All the drought characteristics showed increasing trends over time (except for drought density). Drought area, drought duration and drought intensity show a strong correlation. Drought events are becoming more severe in the basin. (3) The directions and trajectories of 55 typical drought events in the Yellow River Basin were identified. Overall, they show a reversed migration pattern between the northeastern and southwestern regions of the basin, reflecting a dynamic transition between the Loess Plateau region and Gansu–Ningxia region. This phenomenon is closely associated with variations in El Niño–Southern Oscillation (ENSO) events.

Satellite-derived bathymetry is an important method for obtaining shallow sea water depth. To address the problem that the traditional band log-ratio model exhibits significant discrepancies in bathymetric inversion accuracy in both shallow sea (<2 m) and deep sea (>12 m), where water depth is overestimated in shallow areas and underestimated in deep areas, this paper proposes a bathymetric inversion method from active–passive satellite remote sensing data based on a residual correction model. First, the proposed method uses the denoised Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2) ATL03 data set as precise depth control points to construct and implement the traditional band log-ratio model, generating initial bathymetric inversion results. Second, the initial inversion result is compared with the ICESat-2 depth control points, and the residuals between the initial water depth estimates and the true water depths in the local region are calculated. Then, based on the relationship between the spectral characteristics of remote sensing images and water depth residuals, a random forest model is constructed to fit the residual distribution across the entire study area. Finally, using the residual distribution derived from this fitting process, the initial bathymetric inversion results for the entire region are corrected, thereby improving the accuracy and precision of the water depth data. Experimental results demonstrate that the proposed method achieves a bathymetric inversion accuracy with the root mean square error better than 1.57 m and the mean absolute error better than 1.15 m for islands with varying seabed topographies, providing high-precision shallow sea bathymetric inversion results.

The detailed and precise reconstruction of building facades with semantic information conforming to the level-of-detail protocol has drawn increasing attention in recent years. However, despite windows being a major component of building facades, current research often oversimplifies them in the modeling process, primarily focusing on their location and size. This study proposes a joint approach to simultaneously address window detection and detailed parametric modeling. All required window information, including type, location, and interior structure parameters, is obtained via an end-to-end network. This information is then consolidated into predefined window syntax accessible via the SketchUp Ruby application programming interface (API), resulting in the construction of detailed three-dimensional window models. To resolve the issue of training the network with limited labeled data, an adaptive threshold method based on Soft Teacher is proposed, leveraging the pseudolabel technique and consistency regularization to enhance detection precision. Experiments performed on multiple datasets demonstrate that the proposed methods achieve improved window-detection precision with reduced cell estimation errors compared with existing Faster R‐CNN‐based methods. The mean average precision and average precision at the intersection of union = 0.5 (AP50) metrics increased from 47.9% and 61.9% to 69.0% and 83.1%, respectively. The mean absolute error metric for cell estimation improved from 0.103 to 0.076.

Applications of remote sensing images in both defense and civilian sectors have spurred substantial research interest. In the field of remote sensing, object detection confronts challenges such as complex backgrounds, scale diversity, and the presence of dense small objects. To address these issues, we propose an improved deep learning-based model, the Global Multi-scale Fusion Self-calibration Network, which is expected to contribute to alleviating the challenges. It consists of three main components: the hierarchical feature aggregation backbone, which uses improved modules such as the receptive field context-aware feature extraction module, the global information acquisition module, and the simple parameter-free attention module to extract key features and minimize the background interference. To couple multi-scale features, we enhanced the fusing component and designed the multi-scale enhanced pyramid structure integrating the proposed new modules. During the detection phase, especially when focusing on small object detection, we designed a novel convolutional attention feature fusion head. This head is constructed to integrate local and global branches for feature extraction by leveraging channel shuffling and multi-head attention mechanisms for efficient and accurate detection. Experiments on the Detection in Optical Remote Sensing Images (DIOR), Northwestern Polytechnical University Very High-Resolution‐10 (NWPU VHR‐10), remote sensing object detection (RSOD), and DOTAv1.0 data sets show that our method achieves mAP50(mean average precision at 50% intersection over union) of 69.7%, 91.3%, 94.2%, and 70.0%, respectively, outperforming existing comparative methods. The proposed network is expected to provide new perspectives for remote sensing tasks and possible solutions for relevant applications in the image domain.

Rapid urbanization and population growth have intensified the need for accurate and efficient identification of residential buildings in urban planning and disaster management. This paper presents a novel approach for identifying residential buildings by leveraging multiple source data and a transfer learning model, with a case study conducted in Galveston Island, Texas. We propose an integrated framework that combines very-high-resolution (VHR) remote sensing imagery, lidar data, points of interest (POI) data, and GIS-based land use information to extract features of residential buildings. We fine-tune a pretrained deep learning model with our data sets to enhance the model’s adaptability and efficiency in detecting residential buildings in the study area. The results demonstrate that the proposed approach achieves high accuracy in identifying residential buildings, with overall accuracy of 85.6% in the case study. This study also offers valuable insights into the potential of combining multiple data sources and transfer learning techniques for improving residential building identification and other related tasks in urban remote sensing applications.

In this study, we explore using satellite observations to assess community recovery from natural disasters such as fires and hurricanes, supplementing the Resilience Inference Model (RIM). The RIM model has been successfully used to quantify recoveries from hurricanes along the Gulf Coast, but it relies on long-term population changes over years or decades. Our approach integrates satellite observations to enhance recovery assessment with a shorter latency of weeks or months. Using fire, vegetation, and night light data from the Visible Infrared Imaging Radiometer Suite (VIIRS) with daily global observations, Sentinel-2, Landsat-8, and Geostationary Operational Environmental Satellite/Advanced Baseline Imager, we quantitatively evaluate fire intensity, light outage, and urban greenness changes, along with subsequent recovery, focusing on the 2023 Maui fire and selected hurricane cases along the Gulf Coast. This approach complements the RIM model by introducing quantifiable physical parameters with shorter latency, particularly beneficial in areas where census data are either unavailable or unreliable.

The reconstruction of buildings in level of detail–2, according to the CityGML standard, is an essential feature in applications such as urban planning, environmental simulations, and virtual reality. Existing methods work primarily only on aerial data, depend on an external digital terrain model, or do not accurately separate individual buildings. In this work, we present SAT2BUILDING, a method that predicts roof planes, building sections, and building heights in a single, fully convolutional neural network. The network relies on only orthorectified panchromatic imagery and a photogrammetric digital surface model. The three outputs are jointly processed in a level of detail–2 reconstruction pipeline that generates building models that are seamlessly connected, geometrically accurate and complete, and topologically correct. We use spatial embeddings that enable accurate segmentation of building sections and roof planes from satellite imagery. The model generalizes to data from Bonn, Germany, and Lyon, France, after being trained on data from Berlin, Germany. The training and test data differ in lighting conditions, architectural styles, and ground sampling distances. Thorough comparative evaluation shows the superiority of SAT2BUILDING over three baseline methods.

Object detection is crucial to extracting and analyzing information autonomously from high-resolution remote sensing images (HRSIs). To address ideal blocking for large-scale HRSI object detection, this study uses a novel adaptive orientation object-detection method for largescale HRSIs based on multi-scale block fusion. An adaptive orientation object-detection framework based on a convolutional neural network is applied to detect diverse objects of large-scale HRSIs through different block scales; average precision (AP) values of diverse object detection results are calculated at different block scales. Then, block scales matching the largest AP values of diverse objects are determined based on statistical results of the AP values of the diverse object at different block scales. Finally, object-detection results at block scales matching the largest AP values of diverse objects are fused by the non-maximum suppression algorithm to achieve large-scale HRSI object-detection results. Experimental findings reveal that the proposed method is better than any single block-scale object-detection method, resulting in satisfactory large-scale HRSI object-detection results.

This study proposes an automated tailings pond-detection method based on the YOLOv8-RVSW model to address the limitations of traditional surveys. A tailings pond dataset was created using high-resolution satellite images, and data quality was improved through data augmentation techniques. In the model, YOLOv8's backbone feature network was replaced with RepViT to effectively capture global and local information. Additionally, the C2f module was enhanced to QuadraSE_C2f to focus on essential feature channels, and WIoUv3 was used as the loss function to improve object localization and detection accuracy. Experimental results indicate that compared with the original model, accuracy increased by 3.9% to 97.4%, recall improved by 4.9% to 96.3%, and mean average precision rose by 2.4% to 98.5%. This method significantly enhances the automation and intelligence of tailings pond monitoring, providing an effective tool for emergency monitoring.

In recent years, synthetic aperture radar (SAR) vehicle detection has become a research hotspot. However, algorithms using horizontal bounding boxes can lead to redundant detection areas due to the varying aspect ratio and arbitrary orientation of vehicle targets. This paper proposes a morphology-based feature extraction network (MFE-Net), which fully uses the prior shape knowledge of the vehicle targets. Specifically, we adopt rotatable bounding boxes to predict the targets, and a novel rectangular rotation-invariant coordinate convolution (RRICC) is proposed to extract the feature, which can determine more accurately the convolutional sampling location of the vehicles. The adaptive thresholding denoising module (ATDM) is designed to suppress background clutter. Furthermore, inspired by the convolutional neural networks (CNNs) and self-attention, we propose the hybrid representation enhancement module (HREM) to highlight the vehicle target features. The experiment results show that the proposed model obtains an average precision (AP) of 93.1% on the SAR vehicle detection data set (SVDD).