The Loess Plateau in northwest China features fragmented terrain and is prone to landslides. However, the complex environment of the Loess Plateau, combined with the inherent limitations of convolutional neural networks (CNNs), often results in false positives and missed detection for deep learning models based on CNNs when identifying landslides from high-resolution remote sensing images. To deal with this challenge, our research introduced a CNN–transformer hybrid network. Specifically, we first constructed a database consisting of 1500 loess landslides and non-landslide samples. Subsequently, we proposed a neural network architecture that employs a CNN–transformer hybrid as an encoder, with the ability to extract high-dimensional, local-scale features using CNNs and global-scale features using a multi-scale lightweight transformer module, thereby enabling the automatic identification of landslides. The results demonstrate that this model can effectively detect loess landslides in such complex environments. Compared to approaches based on CNNs or transformers, such as U-Net, HCNet and TransUNet, our proposed model achieved greater accuracy, with an improvement of at least 3.81% in the F1-score. This study contributes to the automatic and intelligent identification of landslide locations and ranges on the Loess Plateau, which has significant practicality in terms of landslide investigation, risk assessment, disaster management, and related fields.

Strong soil erosion and increasing human activities have made Loess Plateau areas ecologically fragile regions. Constructing the ecological security pattern (ESP) is imperative to maintain their ecosystem functions and sustainable development. However, it is still challenging to establish the ESP in such an unstable and scattered ecological environment. In this study, we take Shanxi Province, which suffers severe ecological problems in Loess Plateau areas, as an example to construct the ESP in a pattern of “source-resistance-corridor”. The proposed methods include the following steps: (1) potential ecological sources are selected with important ecosystem functions based on contributions of soil and water conservation, habitat quality, and carbon storage; (2) ecological sources are determined by considering core areas at the landscape scale based on morphological spatial pattern analysis (MSPA) along with stability based on dynamic assessment on previous sources; (3) the comprehensive resistance surface is constructed by multiple resistance factors and remotely sensed nighttime light data; (4) ecological corridors are simulated and extracted based on circuit theory. As a result, the proposed ESP in our study area mainly includes 13,592 km² of ecological sources, 8519.64 km of ecological corridors, and 277 ecological nodes. Meanwhile, an ecological framework of “two axes, three belts, and three zones” was proposed based on the optimization and reorganization of ecological components within the ESP. Our research lays a methodological and practical foundation for regional ESP construction and sustainable development in Loess Plateau areas.

Accurate assessment of carbon storage dynamics and their driving factors is important for ecological sustainability and land management on the Loess Plateau under China’s dual carbon goals. In this study, the InVEST and PLUS models were integrated to evaluate carbon storage changes from 2000 to 2020 and simulate future carbon storage patterns for 2030 under four development scenarios, including natural development (ND), rapid development (RD), cropland protection (CP), and ecological protection (EP). In addition, the XGBoost-SHAP framework was employed to identify the dominant drivers and nonlinear response relationships controlling spatial variation in carbon storage. During 2000–2020, ecosystem carbon storage across the Loess Plateau generally increased, rising from 5.780 Pg to 5.893 Pg. Spatially, carbon storage displayed a pronounced pattern characterized by higher levels in the southeast and lower levels in the northwest, aligning with forest–grassland restoration belts. Scenario simulations showed that EP produced the largest carbon storage gain, with total carbon storage projected to reach 5.962 Pg in 2030. In contrast, RD reduced carbon storage to 5.858 Pg because of intensive construction land expansion. XGBoost-SHAP results identified net primary productivity (NPP) as the most influential factor controlling spatial variation in carbon storage, accounting for 57.3% of the total explanatory importance, whereas soil erosion (SE) exhibited a strong negative effect on carbon storage. Population density (POPD) also exerted a negative effect, whereas gross domestic product (GDP) showed positive contributions in economically developed counties. These findings enhance understanding of the spatial response characteristics of carbon storage under environmental gradients and human disturbance across the Loess Plateau. They further provide scientific support for differentiated ecological management and regionally adapted carbon mitigation planning.