Acquiring road information is important for smart cities and sustainable urban development. In recent years, significant progress has been made in the extraction of urban road information from remote sensing images using deep learning (DL) algorithms. However, due to the complex shape, narrowness, and high span of roads in the images, the results are often unsatisfactory. This article proposes a Seg-Road model to improve road connectivity. The Seg-Road uses a transformer structure to extract the long-range dependency and global contextual information to improve the fragmentation of road segmentation and uses a convolutional neural network (CNN) structure to extract local contextual information to improve the segmentation of road details. Furthermore, a novel pixel connectivity structure (PCS) is proposed to improve the connectivity of road segmentation and the robustness of prediction results. To verify the effectiveness of Seg-Road for road segmentation, the DeepGlobe and Massachusetts datasets were used for training and testing. The experimental results show that Seg-Road achieves state-of-the-art (SOTA) performance, with an intersection over union (IoU) of 67.20%, mean intersection over union (MIoU) of 82.06%, F1 of 91.43%, precision of 90.05%, and recall of 92.85% in the DeepGlobe dataset, and achieves an IoU of 68.38%, MIoU of 83.89%, F1 of 90.01%, precision of 87.34%, and recall of 92.86% in the Massachusetts dataset, which is better than the values for CoANet. Further, it has higher application value for achieving sustainable urban development.

The Loess Plateau is an ecologically fragile area in China; furthermore, loess landslides are typical forms of geological disasters, which severely limit the sustainable development of the local societies and the economy. Studying the automatic detection of landslides can facilitate disaster prevention and mitigation in the Loess Plateau, and help realize the climate action goal (SDG 13) of the United Nations Sustainable Development Goals (SDGs). This paper takes typical loess areas in China as the research object, and establishes a historical loess landslide sample database based on Google Earth (GEE) image data, with a total of 1451 loess landslides. The automatic detection of loess landslides is implemented by improving the You Only Look Once X (YOLOX) algorithm. The results show that the average accuracy of landslide detection in this method is 95.43%, and the accuracy rate is 96.32%, which effectively combines the earth’s big data to realize the automatic detection of loess landslides. The research results provide technical support for the promotion of disaster prevention and mitigation in China’s loess regions, the realization of sustainable development goals, and the improvement of natural disaster prevention–resistance–reduction systems.

Long-term degradation of black soil has led to reductions in soil fertility and ecological service functions, which have seriously threatened national food security and regional ecological security. This study is motivated by the UN’s Sustainable Development Goal (SDG) 2—Zero Hunger, specifically, SDG 2.4 Sustainable Food Production Systems. The aim was to monitor the soil organic matter (SOM) content of black soil and its dynamics via hyperspectral remote sensing inversion. This is of great significance to the effective utilization and sustainable development of black soil resources. Taking the typical black soil area of Northeast China as an example, the hyperspectral data of ground features were compared with SOM contents measured in soil samples to correlate SOM with spectral features. Based on their quantitative relationship, a dynamic fitness inertia weighted particle swarm optimization (DPSO) algorithm is proposed, which balances the global and local search abilities of a particle swarm optimization algorithm. The DPSO algorithm is applied to the parameter adjustment of an artificial neural network (BPNN), which is used instead of a traditional error back propagation algorithm, to build a DPSO-BPNN model. Then a global optimal analytical expression of hyperspectral inversion is obtained to improve the generalization ability and stability of the remote sensing quantitative inversion model. The results show that DPSO-BPNN model is more stable and accurate than existing models, such as multiple stepwise regression, partial least squares, and BP neural network models (adjust complex coefficient of determination = 0.89, root mean square error = 1.58, relative recent deviation = 2.93). The results of DPSO-BPNN inversion are basically consistent with the trend in SOM contents measured during surface geochemical exploration. As such, this study provides a basis for hyperspectral remote sensing inversion and monitoring of the SOM contents in black soil.

As China’s railways continue to expand into the Qinghai–Tibet Plateau, the number of deep-buried long tunnels is increasing. Tunnel-damaging geothermal disasters have become a common problem in underground engineering. Predicting the potential geothermal disaster areas along the Yunnan–Tibet railway project is conducive to its planning and construction and the realization of the United Nations Sustainable Development Goals (SDGs)—specifically, the industry, innovation and infrastructure goal (SDG 9). In this paper, the Yunnan–Tibet railway project was the study area. Landsat-8 images and other spatial data were used to investigate causes and distributions of geothermal disasters. A collinearity diagnosis of environmental variables was carried out. Twelve environmental variables, such as land surface temperature, were selected to predict potential geothermal disaster areas using four niche models (MaxEnt, Bioclim, Domain and GARP). The prediction results were divided into four levels and had different characteristics. Among them, the area under receiver operating characteristic curve (AUC) and kappa values of the MaxEnt model were the highest, at 0.84 and 0.63, respectively. Its prediction accuracy was the highest and the algorithm results are more suitable for the prediction of geothermal disasters. The prediction results show that the geothermal disaster potential is greatest in the Markam-Deqen, Zuogong-Zayu and Baxoi-Zayu regions. Through jack-knife analysis, it was found that the land surface temperature, active faults, water system distribution and Moho depth are the key environmental predictors of potential geothermal disaster areas. The research results provide a reference for the design and construction of the Yunnan–Tibet railway project and associated sustainable development.

The prediction of geothermal high-temperature anomalies along the plateau railway will be helpful in the construction of the project and its later management. Taking the Sichuan–Tibet railway as the study area and based on Landsat8 thermal infrared images, map data, and measured data regarding the cause and distribution of geothermal high-temperature anomalies, through correlation analysis, we selected six impact factors including the LST, combined entropy of geological formation, fault density, buffer distance to rivers, magnetic anomaly, and earthquake peak acceleration as the input maps of the model. The index-overlay information model, the weights of the entropy information model, and the weights of the evidence information model were established to quantitatively predict the geothermal anomaly in the study area, and the prediction maps were divided into four classes. The results show that the weights of the evidence information model achieved a high prediction accuracy; the success index and the ratio of the high anomaly area reached 0.0053% and 0.872, respectively, and the spatial distribution of the geothermal points is basically consistent with the prediction results. This research can act as a reference for the design and construction of the Sichuan–Tibet railway.