Ocean internal waves not only enhance ocean mixing and impact sediment resuspension, but also threaten the safety of marine engineering facilities and underwater navigation bodies. Therefore, the accurate detection and identification of ocean internal waves are crucial to ensure the safety of marine activities and optimize the development of marine resources. To improve the accuracy and efficiency of ocean internal wave identification, this paper proposes an automatic identification technique based on YOLOv7, which can quickly and accurately extract the feature information of ocean internal waves from a large amount of SAR data, and realize the efficient identification of ocean internal waves. First, in this paper, dynamic snake convolution (DSConv) is introduced into the efficient layer aggregation network (ELAN) module of the backbone network, so that the network can adaptively focus on the irregular strip-like morphology of the ocean internal waves. In addition, large separable kernel attention (LSKA) is introduced in Conv_BN_SiLU (CBS) of the two downsampling modules in the neck network to capture a wider range of contextual information and enhance the feature fusion process of ocean internal waves. The experimental results show that the F1, (mean average precision) mAP50, and mAP50:95 of the improved YOLOv7 network model are 91.3%, 94.3%, and 59.1%, respectively, which are 5.3%, 2.7%, and 2.8% higher compared to the baseline model.

High-precision semantic segmentation of remote sensing imagery is crucial in geospatial analysis. It plays an immeasurable role in fields such as urban governance, environmental monitoring, and natural resource management. However, when confronted with complex objects (such as winding roads and dispersed buildings), existing semantic segmentation methods still suffer from inadequate target recognition capabilities and multi-scale representation issues. This paper proposes a neural network model, LMVMamba (LoRA Multi-scale Vision Mamba), for semantic segmentation of remote sensing images. This model integrates the advantages of convolutional neural networks (CNNs), Transformers, and state-space models (Mamba) with a multi-scale feature fusion strategy. It simultaneously captures global contextual information and fine-grained local features. Specifically, in the encoder stage, the ResT Transformer serves as the backbone network, employing a LoRA fine-tuning strategy to effectively enhance model accuracy by training only the introduced low-rank matrix pairs. The extracted features are then passed to the decoder, where a U-shaped Mamba decoder is designed. In this stage, a Multi-Scale Post-processing Block (MPB) is introduced, consisting of depthwise separable convolutions and residual concatenation. This block effectively extracts multi-scale features and enhances local detail extraction after the VSS block. Additionally, a Local Enhancement and Fusion Attention Module (LAS) is added at the end of each decoder block. LAS integrates the SimAM attention mechanism, further enhancing the model’s multi-scale feature fusion capability and local detail segmentation capability. Through extensive comparative experiments, it was found that LMVMamba achieves superior performance on the OpenEarthMap dataset (mIoU 52.3%, OA 69.8%, mF1: 68.0%) and LoveDA (mIoU 67.9%, OA 80.3%, mF1: 80.5%) datasets. Ablation experiments validated the effectiveness of each module. The final results indicate that this model is highly suitable for high-precision land-cover classification tasks in remote sensing imagery. LMVMamba provides an effective solution for precise semantic segmentation of high-resolution remote sensing imagery.

The rapid industrialization and urbanization in China have exacerbated air pollution, particularly PM_2.5, posing significant threats to public health. This study focused on Lianyungang, an industrial city, to analyze the spatiotemporal variations in PM_2.5 concentrations from 2000 to 2023 and identify the influencing factors. Utilizing high-resolution PM_2.5 data from the ChinaHighPM_2.5 dataset and ERA5 meteorological data, the study employed Empirical Orthogonal Function (EOF) analysis to capture spatial variability and the Bayesian Estimator of Abrupt Change Seasonal and Trend (BEAST) to assess long-term trends and abrupt changes. The key findings include a marked seasonal pattern, with higher PM_2.5 levels during the winter months and lower concentrations in the summer, primarily driven by temperature, humidity, and precipitation. A significant decline in PM_2.5 levels was observed after 2014, following the implementation of pollution control measures. The study underscores the importance of continued environmental regulation and green technology adoption in mitigating air pollution in rapidly industrializing cities. This research provides a comprehensive analysis of PM_2.5 trends and highlights the critical role of natural and human factors, contributing valuable insights for policymakers and researchers aiming to improve air quality.

The arid region of Northwest China (ARNWC) faces severe desertification, posing a major threat to ecological sustainability and socio-economic development. However, systematic evaluation of desertification across the entire northwestern arid zone remains limited. To address the uncertainty caused by mixed pixels in sparsely vegetated drylands, this study innovatively integrates vegetation and soil indices to develop a robust machine learning-based system for classifying desertification levels in the ARNWC over three decades. In addition, the geographical detector method is employed to quantify the driving factors influencing desertification. The key findings are as follows: (1) Desertification expansion predominantly occurred between 1990 and 1995, followed by a gradual improvement from 1995 to 2020. Transitions between severe and moderate desertification were the most frequent, with approximately 15 × 10⁴ km² shifting from severe to moderate desertification. (2) Physiographic factors were the primary drivers of changes in desertification level, followed by climatic factors. Fractional Vegetation Cover (FVC) had the strongest influence, with an average q-value of 0.72. (3) The explanatory power of the drivers increased significantly through interactions, with the combination of FVC and evaporation (EVA) showing the most pronounced effect. Overall, the methods and findings of this study provide critical insights for targeted desertification control and ecological restoration strategies in arid regions. Although this approach primarily captures desertification symptoms related to surface cover, it offers a valuable long-term perspective on surface cover dynamics.

Coral reefs around Hainan Island in the northern South China Sea represent a marginal reef system exposed to interacting climatic and anthropogenic stresses. This study used an optimized MaxEnt model, remote-sensing-derived coral reef occurrence data, key environmental variables, and CMIP6 climate projections to assess habitat suitability, identify key environmental thresholds associated with suitability change, and examine areas with potential refugial significance. The optimized model showed high predictive performance (mean AUC = 0.947). Bathymetry was the dominant predictor of habitat suitability, while sea surface temperature (SST) and dissolved oxygen (DO) concentration were also important predictors. Predicted suitability declined markedly when water depth exceeded 8.9 m or when multiannual mean SST exceeded 26.8 °C. Under current climate conditions, suitable habitat was limited in extent and showed strong spatial heterogeneity. Future projections indicated severe habitat contraction under SSP2-4.5 and SSP5-8.5, whereas under SSP1-1.9 suitable habitat contracted sharply by the 2050s but partially re-emerged by the 2090s. Under SSP1-1.9, parts of eastern Hainan, especially the coastal waters of southern Wenchang, Qionghai, and Wanning, may retain refugial potential. These results help clarify future spatial patterns of habitat persistence and decline, providing a scientific reference for regional conservation prioritization and adaptive management.