Southwest China has abundant grassland resources, but they are mainly scattered across fragmented mountainous terrain with frequently cloudy and rainy weather, making their accurate identification by remote sensing challenging. Therefore, the goal of this study was to generate prefecture-level city-scale mountainous grassland distribution data to support the development of sustainable grassland husbandry. Here, we proposed a sample selection method and comprehensively utilized multi-source data to obtain the quasi-10 m southwest grassland distribution data. The sample selection method was to first determine the sample selection range based on multi-source land use/cover database, and then to randomly select the samples under the constraint of secondary land use types, multiple factors of terrain and pure pixels. This method can deal with the difficulty in identifying the fragmented grassland distribution caused by steep mountains and hills. In addition, a multispectral time series dataset was constructed based on the fusion of Landsat 8 OLI and Sentinel-2A/B data due to cloudy and rainy weather and was used as one of the input features along with synthetic aperture radar Sentinel-1 time series data and the terrain multi-factor data. Finally, a remote sensing method to accurately identify grassland distribution in southwest China was constructed based on the Google Earth Engine (GEE) platform. Taking Zhaotong City, a prefecture-level city in Yunnan Province, as an example, a thematic map of grassland distribution with an overall accuracy of 88.21% was obtained using the above method. This map has been used by the local government of Zhaotong City in their planning of the development of sustainable grassland husbandry.

Street trees are of great importance to urban green spaces. Quick and accurate segmentation of street trees from high-resolution remote sensing images is of great significance in urban green space management. However, traditional segmentation methods can easily miss some targets because of the different sizes of street trees. To solve this problem, we propose the Double-Branch Multi-Scale Contextual Network (DB-MSC Net), which has two branches and a Multi-Scale Contextual (MSC) block in the encoder. The MSC block combines parallel dilated convolutional layers and transformer blocks to enhance the network’s multi-scale feature extraction ability. A channel attention mechanism (CAM) is added to the decoder to assign weights to features from RGB images and the normalized difference vegetation index (NDVI). We proposed a benchmark dataset to test the improvement of our network. Experimental research showed that the DB-MSC Net demonstrated good performance compared with typical methods like Unet, HRnet, SETR and recent methods. The overall accuracy (OA) was improved by at least 0.16% and the mean intersection over union was improved by at least 1.13%. The model’s segmentation accuracy meets the requirements of urban green space management.

The introduction of an attention mechanism in remote sensing image segmentation improves the accuracy of the segmentation. In this paper, a novel multi-scale KAN-based linear attention (MKLA) segmentation network of MKLANet is developed to promote a better segmentation result. A hybrid global–local attention mechanism in a feature decoder is designed to enhance the ability of aggregating the global–local context and avoiding potential blocking artifacts for feature extraction and segmentation. The local attention channel adopts MKLA block by bringing the merits of KAN convolution in Mamba like the linear attention block to improve the ability of handling linear and nonlinear feature and complex function approximation with a few extra computations. The global attention channel uses long-range cascade encoder–decoder block, where it mainly employs the 7 × 7 depth-wise convolution token mixer and lightweight 7 × 7 dilated deep convolution to capture the long-distance spatial features field and retain key spatial information. In addition, to enrich the input of the attention block, a deformable convolution module is developed between the encoder output and corresponding scale decoder, which can improve the expression ability of the segmentation model without increasing the depth of the network. The experimental results of the Vaihingen dataset (83.68% in mIoU, 92.98 in OA, and 91.08 in mF1), the UAVid dataset (69.78% in mIoU, 96.51 in OA), the LoveDA dataset (51.53% in mIoU, 86.42% in OA, and 67.19% in mF1), and the Potsdam dataset (97.14% in mIoU, 92.64% in OA, and 93.8% in mF1) outperform other advanced attention-based approaches in terms of small targets and edges’ segmentation.

Change detection (CD), as a special remote-sensing (RS) segmentation task, faces challenges, including alignment errors and illumination variation, dense small targets, and large background intraclass variance in very high-resolution (VHR) remote-sensing images. Recent methods have avoided the misjudgment caused by illumination variation and alignment errors by increasing the ability of global modeling, but the latter two problems have still not been fully addressed. In this paper, we propose a new CD model called SFCD, which increases the feature extraction capability for small targets by introducing a shifted-window (Swin) transformer. We designed a foreground-aware fusion module to use attention gates to trim low-level feature responses, enabling increased attention to the changed region compared to the background when recovering the changed region, thus reducing background interference. We evaluated our model on two CD datasets, LEVIR-CD and CDD, and obtained F1 scores of 91.78 and 97.87, respectively. The experimental results and visual interpretation show that our model outperforms several previous CD models. In addition, we adjusted the parameters and structure of the standard model to develop a lightweight version that achieves an accuracy beyond most models with only 1.55 M parameters, further validating the effectiveness of our design.

Chinese Named Entity Recognition (CNER) is a fundamental and crucial task in information extraction. In recent years, pre-trained language and lexicon-based models have proven more powerful than the previous character-based models in CNER tasks. However, existing lexicon-enhanced BERT models neither integrate lexical knowledge into the fundamental layers of the bidirectional transformer model nor explicitly align character features with lexicon features. In this paper, we propose a spatial-aware lexicon adapter (SALA), a neural adapter capable of dynamically integrating character and lexical representations through spatial-aware attention. SALA is incorporated between the layers of BERT to inject lexical information into the deep contextual representations of corresponding character sequences. The resulting fused vectors are further trained in SALA-BERT to enhance CNER. We evaluate SALA-BERT on various Chinese NER tasks. Compared to previous state-of-the-art models, it achieves comparable or better performance.

Outdoor aerosol processes are often associated with disasters and diseases, which threaten human life and health. Outdoor aerosols are a fluid system affected by meteorological conditions and three-dimensional complex terrain. Their variable wind speed and direction and complex terrain boundary conditions make simulating advection processes difficult. Based on incompressible flow conditions, we designed an adaptive time step algorithm for forward advection for the rapid simulation of aerosol processes. The method is based on the first-order forward semi-Lagrangian advection method with unconditional mass conservation. The first-order truncated error coefficient function theory generates an adaptive time step to control the accuracy of forward advection. Smoke aerosol simulation experiments in two small outdoor scenes were designed, and the effects of the traditional backward advection and forward fixed step methods were compared with the algorithm in this study. The proposed simulation method showed improved accuracy compared with the other two methods in experimental scenarios; moreover, compared with those of the traditional backward method, the computation time was significantly reduced and the conservation of mass was significantly improved. Thus, the proposed method is a fast simulation method for outdoor aerosol numerical prediction.