Banghui Yang

Emergency forecasts benefit from scene detail when it constrains the relevant physical process and reaches decision-makers in time. This critical review examines flood inundation, wildfire spread and marine drift through 148 sources and 16 purpose-selected study families. Within this selected corpus, four families develop dedicated 3D conversion, chiefly for hydraulic-domain or fuel-material preparation; the coding rule excludes routine terrain input and existing ocean analyses. Forecast cases mainly update states, controls or initial trajectories. Hydraulic connectivity can alter local inundation. Corrected fire fronts remain exposed to propagation error, while wider marine search regions improve coverage at increased search cost. These mechanisms explain why geometric accuracy and analysis fit alone cannot establish predictive value. We propose a framework that selects scene attributes, update targets and observation timing jointly. Matched alternatives share a future response quantity, information cutoff and resource budget; deployment-specific tolerances connect their comparison to evacuation requirements and search capacity. Physically coupled neural representations require calibrated material properties and an online delivery-time account. The proposed tests assess the forecast effect of additional scene detail, the persistence of a correction and the time left for response.

3D Gaussian Splatting (3D GS) has revolutionized scene representation, efficient photorealistic rendering, and novel view synthesis. However, when applied to complex large-scale real-world scenes, 3D GS faces two major challenges which are geometric errors and rendering artifacts. To address these challenges, we propose FuseGS, a novel cross-modal Gaussian reconstruction framework that seamlessly fuses multi-view images and dense LiDAR point clouds, along with open-vocabulary semantics from large Vision-Language Models (VLMs) to achieve more accurate global geometry and realistic renderings than the existing methods. FuseGS introduces three key innovations. First, to overcome low geometric accuracy in image-only methods, we propose a novel LiDAR-guided Gaussian initialization method to ensure strong geometric supervision. Second, to deal with the difficulty of integrating 2D VLMs semantics into 3D space, we propose a novel category-specific dual-cue refinement method to improve semantics and geometry alignment. Third, to remove rendering artifacts caused by dynamic objects, we introduce a novel semantic-based adaptive density control method to effectively maintain static background reconstruction. To demonstrate the robustness of our framework, we evaluate FuseGS using both self-collected data from handheld and Unmanned Aerial Vehicle (UAV) platforms, as well as public datasets. Compared to the baseline 3D GS method, FuseGS unifies joint optimization of photometric, geometric and static-aware semantics, resulting in 50% improvement in geometric accuracy and 4.8 dB improvement in rendering quality. The method achieves 67.43% mean Intersection over Union (mIoU) without manual annotation. Ultimately, FuseGS has demonstrated excellent potential for Earth observation applications such as smart cities and autonomous driving.

With the rapid evolution of Digital Twins and Embodied AI, achieving fast, dense, and high-precision 3D perception in unknown environments has become paramount. However, existing Visual SLAM paradigms face a critical dilemma: geometry-based methods often fail in texture-less areas due to feature scarcity, while learning-based approaches frequently suffer from scale drift and unphysical deformations. To bridge this gap, we propose VGGT-Geo, a novel SLAM system that synergizes generative priors from Large Foundation Models with multi-modal geometric optimization. Distinguishing itself from simple cascaded architectures, we construct a Probabilistic Geometric Fusion framework, consisting of (1) Generative Warm-start, leveraging the holistic scene understanding capabilities of the VGGT, (2) Confidence-Aware Optimization to extract dense features via DINOv3 and predict their confidence map, and (3) a Multi-Modal Constraint Closure that fuses point-line features and metric depth priors to constrain rotational Degrees of Freedom in Manhattan Worlds. We conducted systematic evaluations on TUM, Replica, Tanks and Temples, and a challenging self-collected dataset featuring extreme lighting and texture-less walls. Experimental results demonstrate that VGGT-Geo exhibits superior robustness and accuracy in unseen environments. On our most challenging dataset, it achieves an Absolute Trajectory Error of 4–5 cm and a Relative Rotation Error of 0.79°, outperforming current state-of-the-art methods by approximately 50% in trajectory accuracy. This study validates that synergizing the intuition of Large Foundation Models with geometric rigor is a viable path toward next-generation robust SLAM.

Traditional remote sensing techniques face notable limitations in accurately estimating forest canopy height. Optical data often suffer from vegetation occlusion, while radar systems, though capable of penetrating foliage, show reduced accuracy in complex terrains. The Global Ecosystem Dynamics Investigation (GEDI), a spaceborne LiDAR mission, offers high-resolution measurements that address these challenges. However, the complexity of waveform processing and the influence of geolocation uncertainty demand rigorous assessment. This study employs GEDI Version 2.0 data, which demonstrates substantial improvement in geolocation accuracy compared to Version 1.0, and integrates airborne laser scanning (ALS) data from the Changbai Mountain forest region to simulate GEDI waveforms. A Monte Carlo-based approach was used to quantify and correct geolocation offsets, resulting in a reduction in the average relative error (defined as the mean of the absolute differences between estimated and reference canopy heights divided by the reference values) in canopy height estimates from 11.92% to 8.55%. Compared to traditional correction strategies, this method demonstrates stronger robustness in heterogeneous forest conditions. The findings emphasize the effectiveness of simulation-based optimization in enhancing the geolocation accuracy and canopy height retrieval reliability of GEDI data, especially in complex terrain environments. This contributes to more precise global forest structure assessments and provides a methodological foundation for future improvements in spaceborne LiDAR applications.

Single tree detection is essential in forest resource surveys. Efficient, accurate, and rapid extraction of individual trees facilitates the timely acquisition of forest resource information. Traditional tree surveys rely on manual field measurements, which are limited in coverage, costly, and time-consuming. To address these issues, this study utilizes multispectral imagery captured by unmanned aerial vehicles (UAVs) as the data foundation and applies deep learning methods for tree species segmentation in the study area. An enhanced ECA-Unet model, derived from the U-net architecture, integrates a channel attention mechanism in the Decoder to augment the model’s accuracy and precision in small-scale UAV-based tree segmentation, attaining an overall accuracy of 85.87%. The performance of the suggested model was compared with that of conventional semantic segmentation models within the same study region. Compared with the original U-net model, the model in this study improved the mAP and Accuracy by 2.1% and 1.23%, respectively, in terms of tree species segmentation accuracy. Compared with the traditional semi-automatic forest crown segmentation method, the model in this paper can output forest single tree segmentation results more quickly and effectively. The findings indicate that the proposed technique yields optimal segmentation accuracy and performance for individual trees within the dataset, facilitating the rapid and exact differentiation of trees in wooded settings utilizing limited sample data.

With the launch of the ICESat-2 satellite, global-scale forest parameter monitoring has entered a new phase. However, the background noise in ICESat-2 lidar data significantly impairs the accuracy of signal photon extraction. This study introduces a direction-adaptive DBSCAN method for denoising ICESat-2 photon point clouds, integrating elevation histogram-based coarse denoising with adaptive clustering for fine denoising. The method is applied to data from the Gongbella River Nature Reserve. An innovative aspect of this approach is the introduction of elliptical tilt angle adaptation, which dynamically adjusts the elliptical orientation of the photon point cloud to determine the optimal tilt angle, thus optimizing the denoising effect and reducing computational and memory demands. The direction-adaptive DBSCAN algorithm improves denoising accuracy by dynamically adjusting the neighborhood radius based on the elliptic tilt angle and the distribution of the point cloud. Additionally, the density threshold selection is optimized using the Otsu method, enhancing the accuracy of distinguishing noise photons from signal photons. The method was validated using data from the Gongbella River Nature Reserve, showing significant improvements in denoising accuracy. Compared to existing methods, recall (R) increased by 6.91%, precision (P) improved by 8.82%, and both the F1-score and accuracy rose by 9.52%. The photon point cloud denoising algorithm demonstrated substantial accuracy improvements across multiple data strips, making it particularly effective for processing complex data from ICESat-2, with broad application potential. Both quantitative and qualitative analyses confirm that the algorithm outperforms traditional methods in signal-to-noise ratio and denoising performance, providing reliable technical support for extracting photon point cloud elevation data from forest surfaces and canopies.

Drought is a critical hydrological challenge with ecological and socio-economic impacts, but its long-term variability and drivers remain insufficiently understood. This study proposes a deep learning-based framework to explore drought dynamics and their underlying drivers across China’s major basins over the past four decades. The Long Short-Term Memory network was employed to reconstruct gaps in satellite-derived soil moisture (SM) datasets, achieving high accuracy (R² = 0.928 and RMSE = 0.020 m³m⁻³). An advanced explainable artificial intelligence (XAI) approach was applied to unravel the mechanistic relationships between SM and critical hydrometeorological variables. Our results revealed a slight increasing trend in SM value across China’s major basins over the past four decades, with a more pronounced downward trend in cropland that was more sensitive to water resource management. XAI results demonstrated distinct regional disparities: the northern arid regions displayed pronounced seasonality in drought dynamics, whereas the southern humid regions were less influenced by seasonal fluctuations. Surface solar radiation and air temperature were identified as the primary drivers of droughts in the Haihe, Yellow, Southwest, and Pearl River Basins, whereas precipitation is the dominant factor in the Middle and Lower Yangtze River Basins. Collectively, our study offers valuable insights for sustainable water resource management and land-use planning.

Ozone is a crucial atmospheric trace gas that protects life on Earth from harmful ultraviolet radiation. However, at low altitudes in the troposphere, ozone adversely impacts climate, human health, and ecosystems. Therefore, monitoring atmospheric ozone concentration and spatial distribution is essential. Ground-based monitoring sites are sparse and do not provide comprehensive coverage. Satellites have been used for global ozone monitoring for decades. Polar-orbiting satellites have the disadvantage of low temporal resolution, providing data from one overpass per day under cloud-free conditions. Better coverage is obtained by using geostationary data offering multiple observations daily. This study proposes a Neural Network for Ozone retrieval using data from the Himawari-8 (H8) Geostationary satellite (NNO3-G), which has a strong ozone absorption spectral band at 9.6 μm. ERA5 Total Column Ozone (TCO) data is used to train a Fully Connected Neural Network (FCNN) model to retrieve TCO in cloud-free areas of H8 images. FCNN shows best accuracy compared with other machine learning models. The retrieved TCO product has a spatial resolution of 2 km × 2 km and a temporal resolution of 10 minutes. The NNO3-G-retrieved TCO data are well-correlated with Pandora ground-based measurements, with Pearson’s correlation coefficient R of 0.95, R² of 0.89, and mean absolute error of 8.88 DU. The retrieval accuracy is better in low-latitude regions than in high-latitude regions, with the best performance in summer. The major outcome of this study is the use of one geostationary satellite for retrieving total column ozone, offering both high time resolution and high precision.