Yifan Wang

Deep learning-based parameterizations of subgrid-scale processes have become a major research focus in recent years, offering the potential to remedy inaccuracies inherent in traditional physics-based schemes. However, their integral stability within numerical models remains insufficiently explored. In this study, we develop deep learning parameterizations for the tropical cyclone boundary layer and implement them in the WRF model. We find that one-dimensional convolutional neural network fails to integrate stably, whereas a fully connected network succeeds. Further analysis shows that the limited receptive field of the convolutional network makes its outputs overly sensitive to certain input perturbations, ultimately causing integral instability. We examine three stabilization strategies—training data augmentation with Gaussian noise, spectral norm regularization, and L2 regularization—and find that all three methods effectively mitigate the network’s output sensitivity to input perturbations, enabling stable integration in WRF and yielding physically reasonable tropical cyclone simulations.

Extracting powerline point clouds from airborne LiDAR data and conducting 3D reconstruction has become a critical technical support for automatic transmission corridor inspection. To enhance data processing efficiency, this paper proposes an automatic method for span segmentation of powerline point clouds that accounts for adjacent powerline interference, aiming to provide “clean” data for the automatic reconstruction of powerline catenary curve models of each span. This method tackles a key challenge in airborne LiDAR data: interference from adjacent or cross-over powerlines when automatically extracting main-line pylon positions and powerline points. Leveraging the spatial relationship between pylons and powerlines in LiDAR point clouds, we developed a fast density clustering algorithm based on a novel point-counting grid (PCGrid), which greatly accelerates DBSCAN clustering while adaptively extracting main-line pylons and powerline point clouds. The method proceeds in three steps: first, using 2D density clustering to extract reliable pylon positions and 3D density clustering to filter out non-main-line point clouds; second, verifying pylon connection combinations via main-line point clouds and identifying the longest line in the connection matrix as the pylons of the main powerline; and third, assigning powerline points to their corresponding spans for segmented reconstruction. Experimental results demonstrate that the proposed PCGrid structure not only significantly improves clustering efficiency, but also enables a fully automated span segmentation process that effectively suppresses adjacent powerline interference, highlighting the novelty of integrating efficient PCGrid-based clustering with spatial-relationship-driven pylon verification into a unified framework for reliable 3D powerline reconstruction.

The Global Forecast System (GFS) of the National Center for Environmental Prediction (NCEP) can produce an hourly forecast output for estimating zenith tropospheric delay (ZTD) at any altitude on grid points. This study proposes a new ZTD estimation model for areas with large elevation differences by integrating GFS forecasts with global navigation satellite system (GNSS) observations. First, the vertical variation characteristic of ZTD is modeled using an exponential function based on GFS forecasts. Subsequently, the extreme gradient boosting (XGBoost) method is applied to calibrate the remaining GFS ZTD residuals using GNSS observations, establishing a mapping relationship between GFS ZTD residuals and GNSS ZTD residuals. By combining the exponential function model with the pre-trained XGBoost model, high-accuracy ZTD forecasts can be achieved. The Yunnan Province in China was selected as a typical area with large elevation differences for experiments. GFS forecasts and GNSS ZTD products at 26 stations of the Crustal Movement Observation Network of China (CMONOC) were used as training data, while 130 GNSS stations from the continuously operating reference stations (CORS) network were used for verification. Results showed that separating the vertical variation characteristic of ZTD can improve the accuracy of the established XGBoost model by approximately 54.96%. Compared with GFS ZTD forecasts and the generalized regression neural network (GRNN) model, the accuracy of ZTD derived from the proposed model was improved by approximately 39.42% and 8.71%, respectively. Meanwhile, the accuracy of the proposed model remains stable across different times of the day and altitude ranges. For areas with few CMONOC stations participating in training, our proposed model also demonstrates good performance. These results verify that GFS forecasts and GNSS observations can complement each other to achieve high-accuracy ZTD forecasts in areas with large elevation differences, which is also important for real-time GNSS applications.

Semantic segmentation of LiDAR point clouds in powerline corridor environments is crucial for the intelligent inspection and maintenance of power infrastructure. However, existing deep learning methods often underperform in such scenarios due to severe class imbalance, sparse and long-range structures, and complex elevation variations. We propose EMPower-Net, an Elevation-Aware Multi-Resolution Network, which integrates an Elevation Distribution (ED) module to enhance vertical geometric awareness and a Multi-Resolution (MR) module to enhance segmentation accuracy for corridor structures with varying object scales. Experiments on real-world datasets from Yunnan and Guangdong show that EMPower-Net outperforms state-of-the-art baselines, especially in recognizing power lines and towers with high structural fidelity under occlusion and dense vegetation. Ablation studies confirm the complementary effects of the MR and ED modules, while transfer learning results reveal strong generalization with minimal performance degradation across different powerline regions. Additional tests on urban datasets indicate that the proposed elevation features are also effective for vertical structure recognition beyond powerline scenarios.

Wildfires pose a regularly increasing threat to ecosystems and critical infrastructure. The severity of this threat is steadily increasing. The growing threat necessitates the development of technologies for rapid and accurate early detection. However, the prevailing fire point detection algorithms, including several deep learning models, are generally constrained by the inherent hard threshold limitations in their decision-making logic. As a result, these methods lack adaptability and robustness in complex and dynamic real-world scenarios. To address this challenge, the present paper proposes an innovative two-stage, semi-supervised anomaly detection framework. The framework initially employs a Transformer-based autoencoder, which serves to transform raw fire-free time-series data derived from satellite imagery into a multidimensional deep anomaly feature vector. Self-supervised learning achieves this transformation by incorporating both reconstruction error and latent space distance. In the subsequent stage, a semi-supervised XGBoost classifier, trained using an iterative pseudo-labeling strategy, learns and constructs an adaptive nonlinear decision boundary in this high-dimensional anomaly feature space to achieve the final fire point judgment. In a thorough validation process involving multiple real-world fire cases in Yunnan Province, China, the framework attained an F1 score of 0.88, signifying a performance enhancement exceeding 30% in comparison to conventional deep learning baseline models that employ fixed thresholds. The experimental results demonstrate that by decoupling feature learning from classification decision-making and introducing an adaptive decision mechanism, this framework provides a more robust and scalable new paradigm for constructing next-generation high-precision, high-efficiency wildfire monitoring and early warning systems.

Hydrological runoff prediction serves as the core technological foundation for water resource management and flood/drought mitigation. However, the nonlinear, non-stationary, and multi-temporal scale characteristics of runoff data result in insufficient accuracy of traditional prediction methods. To address the challenges of single decomposition methods’ inability to effectively separate multi-scale components and single deep learning models’ limitations in capturing long-range dependencies or extracting local features, this study proposes an Informer-GRU runoff prediction model based on STL-CEEMDAN secondary decomposition and Gorilla Troops Optimizer (GTO). The model extracts trend, seasonal, and residual components through STL decomposition, then performs fine decomposition of the residual components using CEEMDAN to achieve effective separation of multi-scale features. By combining Informer’s ProbSparse attention mechanism with GRU’s temporal memory capability, the model captures both global dependencies and local features. GTO is introduced to optimize model architecture and training hyperparameters, while a multi-objective loss function is designed to ensure the physical reasonableness of predictions. Using daily runoff data from the Liyuan Basin in Yunnan Province (2015–2023) as a case study, the results show that the model achieves a coefficient of determination (R²) and Nash-Sutcliffe efficiency coefficient (NSE) of 0.9469 on the test set, with a Kling-Gupta efficiency coefficient (KGE) of 0.9582, significantly outperforming comparison models such as LSTM, GRU, and Transformer. Ablation experiments demonstrate that components such as STL-CEEMDAN secondary decomposition and GTO optimization enhance model performance by 31.72% compared to the baseline. SHAP analysis reveals that seasonal components and local precipitation station data are the core driving factors for prediction. This study demonstrates exceptional performance in practical applications within the Liyuan Basin, providing valuable insights for water resource management and prediction research in this region.

Small- and medium-sized reservoirs significantly alter natural flood processes, making it essential to understand their impact on runoff for effective water resource management. However, the lack of measured data for most small reservoirs poses challenges for accurately simulating their behavior. This study proposes a novel method that utilizes readily available satellite observation data, integrating hydraulic, hydrological, and mathematical formulas to derive outflow coefficients. Based on the Grid-XinAnJiang (GXAJ) model, the enhanced GXAJ-R model accounts for the storage and release effects of ungauged reservoirs and is applied to the Tunxi watershed. Results show that the original GXAJ model achieved a stable performance with an average NSE of 0.88 during calibration, while the NSE values of the GXAJ and GXAJ-R models during validation ranged from 0.78 to 0.97 and 0.85 to 0.99, respectively, with an average improvement of 0.03 in the GXAJ-R model. This enhanced model significantly improves peak flow simulation accuracy, reduces relative flood peak error by approximately 10%, and replicates the flood flow process with higher fidelity. Additionally, the area–volume model derived from classified small-scale data demonstrates high accuracy and reliability, with correlation coefficients above 0.8, making it applicable to other ungauged reservoirs. The OTSU-NDWI method, which improves the NDWI, effectively enhances the accuracy of water body extraction from remote sensing, achieving overall accuracy and kappa coefficient values exceeding 0.8 and 0.6, respectively. This study highlights the potential of integrating satellite data with hydrological models to enhance the understanding of reservoir behavior in data-scarce regions. It also suggests the possibility of broader applications in similarly ungauged basins, providing valuable tools for flood management and risk assessment.

Accurate semantic segmentation in transmission corridor scenes is crucial for the maintenance and inspection of power infrastructure, facilitating the timely detection of potential hazards. In this study, we propose SA-KPConv, an advanced segmentation model specifically designed for transmission corridor scenarios. Traditional approaches, including Random Forest and point-based deep learning models such as PointNet++, demonstrate limitations in segmenting critical infrastructure components, particularly power lines and towers, primarily due to their inadequate capacity to capture complex spatial relationships and local geometric details. Our model effectively addresses these challenges by integrating a spatial attention module with kernel point convolution, enhancing both global context and local feature extraction. Experiments demonstrate that SA-KPConv outperforms state-of-the-art methods, achieving a mean Intersection over Union (mIoU) of 89.62%, particularly excelling in challenging terrains such as mountainous areas. Ablation studies further validate the significance of our model’s components in enhancing overall performance and effectively addressing class imbalance. This study presents a robust solution for semantic segmentation, with considerable potential for monitoring and maintaining power infrastructure.

The ionosphere is one of the important error sources that affect the communication of radio signals. The international reference ionosphere (IRI) model is a commonly used model to describe ionospheric parameters. The driving parameter IG12 of the IRI-2016 model was optimally updated based on GNSS-TEC data from 2015 and 2019. The electron density profiles and NmF2 calculated by the IRI-2016 model (upda-IRI-2016) driven by the updated IG12 value (IG-up) were evaluated for their accuracy using ionosonde observations and COSMIC inversion data. The experiments show that both the electron density profiles and NmF2 calculated by upda-IRI-2016 driven by IG-up show significant optimization effects, compared to the IRI-2016 model driven by IG12. For electron density, the precision improvement (PI) for both MAE and RMSE at the Beijing station exceed 31.2% in January 2015 and 16.0% in January 2019. While the PI of MAE and RMSE at the Wuhan station, which is located at a lower latitude, both exceed 32.5% in January 2015, both exceed 42.1% in January 2019, which is significantly higher than that of the Beijing station. In 2015, the PI of MAE and RMSE compared with COSMIC are both higher than 20%. For NmF2, the PI is greater for low solar activity years and low latitude stations, with the Wuhan station showing a PI of more than 11.7% in January 2019 compared to January 2015. The PI compared to COSMIC was higher than 17.2% in 2015.