滑坡灾害监测技术、预测方法与风险评价研究综述

Review on Landslide Disaster Monitoring Technologies, Prediction Methods and Risk Assessment

摘要: 滑坡作为突发性强、破坏性大的地质灾害，在全球气候变化与人类工程活动加剧下，频发态势严重威胁社会经济与生命财产安全，成为多学科研究热点。本文系统梳理国内相关研究进展，聚焦监测技术、空间预测、触发机制及风险评价四大核心。监测技术方面，传统GPS、测斜仪精度高但成本高、覆盖有限；InSAR遥感技术结合多源数据融合，实现大范围、高精度监测，却受植被密集区与复杂大气条件制约。滑坡空间预测以GIS为核心，从定性专家打分法演进至定量统计、机器学习模型(如逻辑回归、随机森林)，多模型集成可降低不确定性，因子选取与ROC验证对精度至关重要。触发机制中，降雨研究聚焦强度–持续时间阈值模型，需结合区域条件校准；地震触发以汶川地震为案例，明确其与地震动参数、地形坡度的关联。风险评价基于易发性，综合承灾体脆弱性与暴露度，国内以静态区划为主，脆弱性量化待完善。研究表明，监测向空天地一体化发展，机器学习在预测中应用深化，触发机制需加强多因素耦合，风险评价需动态化。未来应攻关多源数据融合与深度学习技术，结合气候变化实现风险动态评估，推动 “监测–预警–防控”闭环转化，为防灾减灾提供科学支撑。

Abstract: As a sudden and destructive geological disaster, landslides have become a multidisciplinary research hotspot under the intensification of global climate change and human engineering activities. This paper systematically sorts out the relevant research progress in China, focusing on the four cores of monitoring technology, spatial prediction, trigger mechanism and risk assessment. In terms of monitoring technology, traditional GPS and inclinometers have high accuracy but high cost and limited coverage. InSAR remote sensing technology combines multi-source data fusion to achieve large-scale and high-precision monitoring, but it is limited by dense vegetation areas and complex atmospheric conditions. Landslide spatial prediction is based on GIS, evolving from qualitative expert scoring to quantitative statistics and machine learning models (such as logistic regression and random forests). In the trigger mechanism, the rainfall study focuses on the intensity-duration threshold model, which needs to be calibrated in combination with regional conditions. The Wenchuan earthquake is taken as an example to clarify its relationship with ground motion parameters and terrain slope. Risk assessment is based on susceptibility, comprehensively vulnerability and exposure of disaster-bearing bodies, but domestic static zoning is the mainstay, and the quantification of vulnerability needs to be improved. The results show that monitoring is developing towards the integration of space, space, and ground, and the application of machine learning in prediction is deepening, the trigger mechanism needs to strengthen multi-factor coupling, and the risk assessment needs to be dynamic. In the future, multi-source data fusion and deep learning technology should be tackled to achieve dynamic risk assessment in combination with climate change, promote the closed-loop transformation of “monitoring, early warning, prevention and control”, and provide scientific support for disaster prevention and mitigation.