Under the joint influence of climate change and human production and living activities on land and sea, the morphology and use function of the coastline have undergone notable changes, triggering more significant environmental and ecological effects. In this paper, we quantified the evolution characteristics of the coastline over the past 30 years and measured the possible environmental pollution and ecological degradation by means of spatiotemporal coupling analysis in terms of spatial form and land-use function. The Chinese Bohai Rim region was taken as an example, and the Google Earth Engine was applied to achieve a long time series of interannually continuous land use function classification of the coastal zone. The study shows that: (1) from 1987–2020, the coastline of Bohai Rim showed an overall trend of seaward expansion, with the length of the coastline increasing by 15.6%, most significantly from 2003 to 2011; (2) the proportion of construction function coastline increased from 14.66% to 42.8%, while the proportion of cropland coastline decreased from 52.01% to 18.16% during 2000–2020, with the natural ecological coastline decreasing in the early stage and recovering in the later stage, thus becoming more stable overall; (3) 82.73% of coastal districts and counties showed a linear correlation between changes in coastline function and water quality level, and 77.8% showed a consistent degree of change of coastline function and loss of ecological land, indicating that changes in the coastline have triggered significant problems in terms of environmental pollution and ecological degradation. With the exception of localized areas such as the Yellow River Delta, where coastline change is somewhat more significantly influenced by environmental climate change, most other coastline changes are primarily due to human land-based development; therefore, coastal policy constraints bring obvious mitigation.

Cropland non-agriculturalization (CNA) threatens food security, ecosystem services, and sustainable development amid accelerating global urbanization. However, existing monitoring methods are often retrospective and lack adequate spatial and temporal resolution for proactive management. This study proposes GS-GAT, a graph-based deep learning framework for predicting CNA susceptibility at the meso-spatial scale. A spatial graph was constructed for the non-central districts of Wuhan, China, and multisource features were extracted across four dimensions: imagery, land cover, topography, and socioeconomics. A comprehensive intensity index is developed to compute susceptibility levels at the street-block level based on multi-year land use data from 2018 to 2022. To address class imbalance, GraphSMOTE is employed to enhance minority node representation. The key model of GS-GAT is trained across four temporal snapshots using attention-based feature aggregation and joint optimization of classification and structural reconstruction losses. Experimental results show that GS-GAT demonstrated an average AUC of 85.6% and an F1 score of 82.6%, which increased to 93% and 91%, respectively, under relaxed evaluation criteria, whereby baseline models such as SVM and XGBoost were outperformed. Ablation studies confirm the contributions of feature fusion and GraphSMOTE to model robustness and minority class detection. The proposed framework offers a scalable and interpretable approach for early identification of cropland conversion risks, supporting more targeted land-use management and cropland protection strategies.

As an important national resource and the foundation of food production, cropland is crucial for ensuring farmers' income and food security. However, amid rapid urbanization and industrialization, developing nations like China confront the significant challenge of cropland non-agriculturalization. Driven by multiple factors, this process involves converting agricultural land to non-agricultural uses and has multiple negative social, economic, and environmental impacts. To effectively manage and reduce the scale of it, widely adopted methods currently include remote sensing, machine learning and deep learning, spatiotemporal analysis, as well as combinations of these methods (Liu M et al., 2022; Lu D et al., 2024). Among them, remote sensing change detection technology can accurately identify the cropland non-agriculturalization patches from dual-phase high-resolution satellite imagery (Zhang H et al., 2024). Interpretable machine learning methods, such as XGBoost and SHAP (Zhang G et al., 2024), are used to identify and analyze the significance of driving factors. Change detection methods based on deep learning are able to be applied to the cropland non-agriculturalization monitoring (Wu Q et al., 2024), but suffer from the reflectance variability and the lack of dataset (Sun Z et al., 2024). Spatiotemporal analysis methods can reveal long-term dynamic changes and factors of the cropland non-agriculturalization (Chen Y et al., 2022), providing more precise and reliable data to support policy formulation for it (Wu X et al., 2024). Although these methods reach a high accuracy in respective fields, they can only detect areas where the cropland non-agriculturalization has already occurred. Correcting this issue is labor-intensive and requires significant material resources for rehabilitation. In addition, current research typically operates within time scales of 5-10 years and lacks a meso-scale approach to bridge the gap between patch-level and county-level scales, which does not meet monitoring and regulation needs. Thus, predicting and preventing high-risk areas is vital. Short-term, meso-scale predictions can offer a better basis for management and policy-making, improving efficiency and reducing costs.