ISSN 1000-3665 CN 11-2202/P
    徐文正,卢书强,林振,等. 联合InSAR与神经网络的范家坪滑坡形变监测及预测研究[J]. 水文地质工程地质,2024,51(0): 1-13. DOI: 10.16030/j.cnki.issn.1000-3665.202308028
    引用本文: 徐文正,卢书强,林振,等. 联合InSAR与神经网络的范家坪滑坡形变监测及预测研究[J]. 水文地质工程地质,2024,51(0): 1-13. DOI: 10.16030/j.cnki.issn.1000-3665.202308028
    XU Wenzheng, LU Shuqiang, LIN Zhen, et al. Combination of InSAR and neural networks for the deformation monitoring and prediction of Fanjiaping landslide[J]. Hydrogeology & Engineering Geology, 2024, 51(0): 1-13. DOI: 10.16030/j.cnki.issn.1000-3665.202308028
    Citation: XU Wenzheng, LU Shuqiang, LIN Zhen, et al. Combination of InSAR and neural networks for the deformation monitoring and prediction of Fanjiaping landslide[J]. Hydrogeology & Engineering Geology, 2024, 51(0): 1-13. DOI: 10.16030/j.cnki.issn.1000-3665.202308028

    联合InSAR与神经网络的范家坪滑坡形变监测及预测研究

    Combination of InSAR and neural networks for the deformation monitoring and prediction of Fanjiaping landslide

    • 摘要: 目前,传统滑坡地表形变监测手段仍存在监测范围小、复杂地形信息获取难度高、经济成本投入量大等缺点,且大型复杂滑坡变形时间序列的非线性、不确定性变化特征也一直是滑坡形变监测及预测研究中亟待解决的难题。将三峡库区范家坪滑坡作为研究对象,利用短基线集(The Small BAseline Subset InSAR,SBAS-InSAR)技术结合地表GPS监测数据进行滑坡形变监测,基于SBAS-InSAR时间序列数据及长短时记忆网络(Long Short Term Memory,LSTM)开展滑坡形变预测研究。结果表明:研究时段内,范家坪滑坡SBAS-InSAR形变监测结果与地表GPS监测数据所反映出的形变区域及形变量级基本保持一致,与现场调查情况相吻合;范家坪滑坡在不同高程段,其位移变形并非总是表现出受浮托减重效应影响,当库水位低于160m左右时,渗流压力占主导作用,水位下降阶段的位移变形总体明显大于水位上升阶段,库水位下降速率对范家坪滑坡的位移变形产生重要影响,且木鱼包滑坡区相较于谭家河滑坡区对库水位下降速率的变形响应更为强烈(注:根据滑坡蠕滑方向及变形响应规律将范家坪滑坡进一步划分为木鱼包滑坡区及谭家河滑坡区);将LSTM神经网络模型与传统神经网络模型的预测结果进行效果对比、置信区间估计及相关性检验,结果显示,LSTM神经网络模型的预测结果始终保持较高的预测精度,预测效果最佳,验证了该方法能够为三峡库区地质灾害防治提供重要的数据参考和信息支撑。

       

      Abstract: At present, the traditional means of monitoring the surface deformation of landslides are still limited by the small monitoring area, the difficulty of obtaining information in complex terrain, and the high economic cost. The non-linearity and uncertainty of the deformation time series of large and complex landslides are difficult to solve in landslide deformation monitoring and forecasting. In this study, For the Fanjiaping landslide in the Three Gorges Reservoir area, The Small BAseline Subset InSAR (SBAS-InSAR) is used to monitor the landslide deformation in combination with the surface GPS monitoring data, and the landslide deformation prediction is carried out based on the SBAS-InSAR time-series data and Long Short-Term Memory (LSTM) network. The results indicate that during the study period, the deformation area and deformation magnitude monitored by SBAS-InSAR are basically consistent with that from surface GPS monitoring and field investigation. The displacement and deformation of the Fanjiaping landslide at different elevations are not always affected by the buoyancy reduction. When the reservoir water level is below approximately 160 m, seepage pressure dominates the deformation. The overall displacement and deformation during the water level decrease period are significantly greater than during the water level rise period. The decrease rate in reservoir water level has a significant impact on the displacement and deformation of the Fanjiaping landslide. Additionally, the Muyubao landslide area shows a stronger deformation response to the decrease rate of reservoir water level compared to the Tanjiahe landslide area. (Fanjiaping landslide is further divided into Muyu Bao landslide area and Tanjiahe landslide area based on the direction of landslide creep and deformation response). Multiple methods, including a comparison with traditional methods, confidence interval estimation, and correlation tests, were conducted to evaluate the prediction results of the LSTM model. It demonstrates that the LSTM model always has high precision and reliability. This study can provide important data and information support for the prevention and control of geological disasters in the Three Gorges reservoir area.

       

    /

    返回文章
    返回