Remote sensing imagery contains rich information about geographical targets, and performing super-resolution (SR) reconstruction on such images requires greater feature representation capabilities. Convolutional neural network (CNN)-based methods excel at extracting intricate local features but fall short in terms of capturing global representations. While transformer methods are capable of learning long-distance dependencies, they often overlook local feature details, which can diminish the discriminability between the background and the foreground. Moreover, the distinctive architectures of transformers, their extensive parameter counts, and their reliance on large-scale training datasets impose constraints on transformer applications in remote sensing image feature extraction tasks. To address these challenges, this study introduces a novel hybrid CNN-Transformer network model named RepCHAT for remote sensing single image reconstruction, which incorporates a structural re-parameterization technique and a hybrid attention mechanism. This method leverages the strengths of transformers in terms of learning long-distance dependencies (global features) and CNNs with respect to extracting local features. The proposed approach achieves SR reconstruction for remote sensing images with fewer parameters and less computational overhead than those of traditional transformers and high-performance CNN models. We develop a multiscale feature extraction module that integrates both spatial- and frequency-domain features and employs structural re-parameterization theory to increase the inference efficiency of the model. Furthermore, we incorporate depthwise-separable convolution into the transformer block to bolster the local feature learning capabilities of the transformer. The method we propose achieves the optimal performance for remote sensing single-image super-resolution reconstruction and outperforms the competing methods by 0.28–1.05 dB (×4 scale) in terms of signal-to-noise ratio (PSNR). Experimental results indicate that the RepCHAT model proposed in this study maintains a high performance with significantly reduced complexity, making it suitable for deployment on edge devices.

The grassland ecosystem in the Inner Mongolia Autonomous Region (IMAR) serves as a vital ecological barrier in northern China, and the vegetation productivity in the grasslands exhibits considerable temporal and spatial variations. However, few studies have examined the long-term variations in the NPP in the IMAR and quantified the effects of natural factors and human activities on the NPP. The study modeled the net primary productivity (NPP) of the IMAR’s grasslands using the Carnegie–Ames–Stanford approach (CASA) model and employed linear regression, trend analysis, and spatial statistics to analyze the spatio-temporal patterns in vegetation productivity and explore the impact on the NPP of natural and socio-economic factors over the past two decades. The results reveal that the average NPP value from 2001 to 2021 was 293.80 gC∙m⁻² a⁻¹, characterized by spatial clustering of a relatively high NPP in the east, a low NPP in the west, and an annual increase of 3.26 gC∙m⁻² over the years. The NPP values varied significantly across different vegetation cover types, with meadows having the highest NPP, followed by typical steppe and desert grasslands. The spatial distribution pattern and temporal changes in the grassland productivity are the result of both natural factors and human activities, including topographical properties and socio-economic indicators such as gross domestic product, night-time light, and population. The results for the NPP in the IMAR were based solely on the CASA model and, therefore, to achieve improved data reliability, exact measurements in real field conditions will be conducted in the future. The findings from the spatial clustering and temporal trajectories of the NPP and the impacts from the factors can provide useful guidance to planning grassland vegetation protection policies for the IMAR.

As the most basic physical geographic elements, basin terrain and river networks have high spatial complexity and are closely related. However, there is little research on the correlation between terrain and river networks. In this paper, the Yellow River Basin was selected as the study area. Topographic factors of multiple dimensions were calculated. The influence of different topographic factors on the river network structure at different scales and their correlation from a multifractal perspective based on geographical detectors and a geographically weighted regression model were determined. The explanatory power of topography on the river network structure at different scales was: multifractal spectrum width > multifractal spectrum difference > slope > average elevation > elevation maximum > elevation minimum, which generally indicated that the topographic factor that has the greatest influence on the river network structure is the complexity and singularity of the terrain. The second-order clustering of regression coefficients from the results of the geographically weighted regression model revealed that the Yellow River basin was divided into three types of high-aggregation areas, which are dominated by the Qinghai-Tibet Plateau, the Loess Plateau, and the Huang-Huaihai Plain, respectively. The clustering results also revealed that the river network structure was affected by different key topographic factors in the different types of areas. This research studies and quantifies the relationship between basin topography and river network structure from a new perspective and provides a theoretical basis for unraveling the development of topography and river networks.

A major source of carbon dioxide emissions (CO₂) arises from the household sector. Recent studies have reported increasing household CO₂ emissions (HCO₂) in many countries. Cities represent a key administrative level in China and can be managed to mitigate HCO₂ if spatial and temporal variations in HCO₂ are understood at fine scales. Here, we applied panel data analysis to map HCO₂ at a pixel scale of 1 km in China using remotely sensed time series nighttime light data, grid population density data, and provincial energy consumption statistics from 2000 to 2020. Spatial and temporal variations in HCO₂ were observed with four growth modes, including high growth (HG), low growth (LG), negative growth (NG), and high negative growth (HNG), for different periods, i.e., 2000–2010, 2010–2020, and 2000–2020. We proposed a local scale analysis of HCO₂ growth patterns within a global context to assess the contribution roles of 372 municipal cities to the changes in the national total HCO₂ (T-HCO₂). The results indicated that T-HCO₂ has tripled in the last two decades, but the roles of the contribution to the increase varied among cities. The local scale analysis revealed that more cities contributed to the rise in T-HCO₂ through HG and LG than those that suppressed it through NG and HNG. The majority of the cities displayed contributions to the rise in T-HCO₂ through two or more of the growth modes, confirming a significant variation in HCO₂ across locations, even within a city. This study provides a new approach to understanding the roles cities play in the long-term dynamics of T-HCO₂. We recommend increased efforts to encourage HCO₂ mitigation in cities that have contributed to the rise in T-HCO₂ to help neutralize carbon emissions at the national level.

Due to the various types of land cover and large spectral differences in remote sensing images, high-quality semantic segmentation of these images still faces challenges such as fuzzy object boundary extraction and difficulty in identifying small targets. To address these challenges, this study proposes a new improved model based on the TransDeepLab segmentation method. The model introduces a GAM attention mechanism in the coding stage, and incorporates a multi-level linear up-sampling strategy in the decoding stage. These enhancements allow the model to fully utilize multi-level semantic information and small target details in high-resolution remote sensing images, thereby effectively improving the segmentation accuracy of target objects. Using the open-source LoveDA large remote sensing image datasets for the validation experiment, the results show that compared to the original model, the improved model’s MIOU increased by 2.68%, aACC by 3.41%, and mACC by 4.65%. Compared to other mainstream models, the model also achieved superior segmentation performance.