The accuracy of machine learning models is affected by hyperparameters when classifying different types of urban green spaces. To investigate the impact of hyperparametric algorithms on model optimization, this study used the Marine Predators Algorithm (MPA) to optimize three models: K-Nearest Neighbor (KNN), Support Vector Machines (SVM), and Random Forest (RF). The feasibility of the algorithm was illustrated by extracting and analyzing park green space and attached green spaces within the fifth-ring road of Beijing. A dataset of urban green space type labels was constructed using SPOT6. Three optimized models, MPA-KNN, MPA-SVM and MPA-RF, were constructed. The optimum hyperparameter combination was chosen based on the accuracy of the validation set, and the three optimized models were compared in terms of the Area Under Curve (AUC) value, accuracy on the test set, and other indicators. The results showed that applying MPA improves the accuracy of the validation set of the KNN, SVM, and RF models by 4.2%, 2.2%, and 1.2%, respectively. The MPA-RF model had an AUC value of 0.983 and a test set accuracy of 89.93%, indicating that it was the most accurate of the three models.

Climate-induced lake expansion on the Tibetan Plateau (TP) has led to two distinct hazard types: outburst floods and passive inundation. However, the divergent driving mechanisms behind these hazards remain insufficiently understood. This study analyzes the spatiotemporal trends of 1352 non-glacial lakes (>1 km²) on the TP from 2014 to 2023 using high-resolution Gaofen-1 (GF-1) and Gaofen-2 (GF-2) imagery. By integrating geomorphic analysis with hazard mechanisms, we screened and categorized lakes prone to outburst floods and inundation using a classification and assessment framework proposed in this study. The results indicate that the net area of these lakes expanded by 2839.53 km² (6.07%), with the Inner TP Basin contributing the largest absolute area gain (1960.60 km²). We identified 21 potentially hazardous lakes (10 outburst-prone and 11 inundation-prone) and systematically categorized them by risk level. Field investigations of high-risk candidates, such as Rulei Co and Xiao Qaidam Lake, validated the accuracy of the hazard classification and risk assessment methodology. Preliminary attribution analysis further suggests that the two hazard types may be associated with distinct climatic factors. Overall, this study provides a scientific basis for disaster mitigation and lake management on the TP.

Geological hazard risk assessment provides essential scientific support for geological disaster prevention and governance. The selection of appropriate evaluation factors is crucial to the accuracy and practicality of the risk assessment results. The existing factors for geological hazard risk assessment often suffer from issues such as poor timeliness and insufficient completeness. Interferometric Synthetic Aperture Radar (InSAR) technology, which offers large-scale, high spatiotemporal resolution monitoring of surface deformation, can effectively compensate for the shortcomings of existing risk assessment factors. How to effectively integrate time-series InSAR deformation results into geological hazard risk assessment has become a focus of research. This study fully considers the time-series InSAR deformation information; both the ascending and descending orbit results of the time-series InSAR deformation are introduced as two categories of evaluation factors in the risk assessment model. Subsequently, 11 types of assessment factors are selected by the Pearson correlation coefficient method, while the Information Volume Model and Evidence Weight Model are applied in the partitioning and assessment of risks in Xiaojin County, China. Finally, ROC (Receiver Operating Characteristic Curve) analysis is utilized to compare the accuracy of model evaluations before and after incorporating time-series InSAR deformation results. The results indicate that: (1) after incorporating time-series InSAR deformation monitoring results as evaluation factors into the information volume model and evidence weight model, the evaluation accuracy of the two models improved by 9.69% and 11.26%, respectively; (2) there are differences in risk partitioning among different evaluation models. From the risk partitioning result of Xiaojin County in this study, the evaluation accuracy of the information volume model is higher than that of the evidence weight model, and the performance is more prominent after adding the time-series InSAR deformation results.

Accurate lake boundary extraction from optical remote sensing imagery remains challenging in high-altitude regions such as the Tibetan Plateau due to ice cover, snow, shadows, and spectrally similar backgrounds. Although recent deep learning models achieve strong region-overlap performance, they often fail to ensure stable shoreline localization. To address this issue, we propose a Global–Local Residual Refinement Network (GLR-Net) for boundary-aware lake extraction from remote sensing imagery. The proposed framework first captures large-scale semantic context through a global branch and subsequently performs patch-level residual refinement to improve local shoreline geometry. A global-to-local guidance mechanism is further introduced to incorporate structural priors into local refinement. Experiments on a manually annotated Tibetan Plateau lake dataset demonstrate that the proposed method achieves competitive region-level segmentation performance while improving geometric shoreline accuracy. Compared with representative semantic segmentation baselines, including U-Net, SegFormer-B0, SegFormer-B4, and OCRNet, the proposed method achieves the highest Boundary F1 score of 0.811 under a 3-pixel tolerance and the lowest mean BDE of 13.19 pixels. The results indicate that conventional overlap-based metrics alone are insufficient for evaluating shoreline delineation quality in complex alpine environments.