Understanding the relationship between ecosystem service value and ecological risk evolutions holds great theoretical and practical significance, as it helps to ensure the quality management of ecosystems and the sustainable development of human–land system interactions. We analyzed this relationship in the Dongting Lake area in China from 1995 to 2020 using data from remote sensing-interpreted land use with ArcGIS and Geoda. We used the equivalent factor method to estimate the ecosystem service value, constructed a landscape ecological risk index to quantitatively describe the ecological risk of Dongting Lake, and analyzed their correlation. The results show that: (1) over the last 25 years, the ecosystem service value decreased by 31.588 billion yuan, with higher values in the middle of the area and lower values in the surroundings—the highest value was found in forested land and the lowest was for unutilized land; (2) the ecological risk index also decreased slowly over time, from the perspective of single land use type, the ecological risk value of construction land was the lowest, followed by woodland, grassland, and cultivated land, with water area being the highest—the ecological risk level presents the distribution state of whole piece and local aggregation; and (3) the ecological risk index in Dongting Lake area demonstrated positive spatial correlation, and the spatial agglomeration of land with similar risk levels showed a decreasing trend. Areas with strong partial spatial correlations between ecosystem service value and ecological risk index are mainly distributed in the central water areas and their surrounding areas. This study investigates the rational utilization of land resources, and the sustainable development of regional ecological security in Dongting Lake area.

In China, the rural human settlement system (RHSS) reflects the relationship between rural people and land but is affected by land use and land cover change (LUCC). Maintaining a harmonious development between RHSS resilience (RHSSR) and LUCC is an important rural development issue. However, the spatial association between LUCC and RHSSR remains unclear, constraining effective land use and rural policymaking. The association between RHSSR and LUCC was assessed from a spatial perspective. Using county-level spatial panel data for Hunan Province, China (2000–2020), an evaluation indicator system was established to measure RHSS and analyze the spatiotemporal evolution of RHSSR using a geospatial analysis and geodetector model. The average RHSSR level increased from 0.158 to 1.406. The RHSSR generally presented a three-level stepped spatial distribution feature of high in the east and low in the west, with a belt-like distribution. Additionally, the RHSSR was consistent with land use intensity (LUI), but there was significant spatial heterogeneity in the spatial relationship between LUI and RHSSR. The evolution of RHSSR occurred through industrial development, medical service improvement, increased income, and environmental protection. This has important implications for future rural development strategies, the sustainable development of rural land, and the integration of RHSS into regional planning.

With rapid socio-economic development, excessive anthropogenic consumption and the exploitation of natural resources have impaired the self-healing, supply, and carrying capacities of ecosystems. The assessment and prediction of ecosystem service values (ESVs) are crucial for the coordinated development of ecology and economy. This research examines the Xiangjiang River Basin and combines land use data from 1995 to 2020, Landsat images, meteorological data, and socio-economic data. These data are incorporated into the PLUS model to simulate land use patterns in 2035 under the following five scenarios: natural development, economic development, farmland protection, ecological protection, and coordinated development. Additionally, this research analyzes the dynamics of land use and changes in ESVs in the Xiangjiang River Basin. The results show that between 1995 and 2020 in the Xiangjiang River Basin, urbanization accelerated, human activities intensified, and the construction land area expanded significantly, while the areas of forest, farmland, and grassland decreased continuously. Based on multi-scenario simulations, the ESV showed the largest and smallest declines under economic development and ecological protection scenarios, respectively. This results from the economic development scenario inducing a rapid expansion in construction land. In contrast, construction land expansion was restricted under the ecological protection scenario, because the ecological functions of forests and water bodies were prioritized. This research proposes land use strategies to coordinate ecological protection and economic development to provide a basis for sustainable development in the Xiangjiang River Basin and constructing a national ecological security barrier, as well as offer Chinese experience and local cases for global ecological environment governance.

Vehicle re-identification (Re-ID) aims to retrieve images with the same vehicle ID across different cameras. Current part-level feature learning methods typically detect vehicle parts via uniform division, outside tools, or attention modeling. However, such part features often require expensive additional annotations and cause sub-optimal performance in case of unreliable part mask predictions. In this paper, we propose a weakly-supervised Part-Attention Network (PANet) and Part-Mentored Network (PMNet) for Vehicle Re-ID. Firstly, PANet localizes vehicle parts via part-relevant channel recalibration and cluster-based mask generation without vehicle part supervisory information. Secondly, PMNet leverages teacher-student guided learning to distill vehicle part-specific features from PANet and performs multi-scale global-part feature extraction. During inference, PMNet can adaptively extract discriminative part features without part localization by PANet, preventing unstable part mask predictions. We address this Re-ID issue as a multi-task problem and adopt Homoscedastic Uncertainty to learn the optimal weighing of ID losses. Experiments are conducted on two public benchmarks, showing that our approach outperforms recent methods, which require no extra annotations by an average increase of 3.0% in CMC@5 on VehicleID and over 1.4% in mAP on VeRi776. Moreover, our method can extend to the occluded vehicle Re-ID task and exhibits good generalization ability.