摘要
云南永胜、宾川一带新构造运动活跃,巨型、大型滑坡现象显著。为了分析区内滑坡的发育分布规律、动力成因及其与地质地貌的关系,在以目视解译为主获得615个滑坡点的基础上,进一步选取8个滑坡影响因子,包括断层、水系、岩性、NDVI、坡度、坡向、地形起伏度和高程,利用级联BP神经网络模型进行了计算分析,发现该区影响因子指数高的区域具有明显的聚集性,并据此划分出4个高指数区。经过进一步讨论分析认为,研究区的滑坡主要由地震滑坡、冲蚀型滑坡和崩积加载型滑坡3类构成,尤以地震滑坡为主,即活动断裂是研究区滑坡最主要的影响因素。结合对该区典型古地震滑坡的地表调查结果,进一步证明了该区的程海-宾川断裂带第四纪期间的长期强烈活动对永胜、宾川地区的滑坡发育起着重要控制作用,同时也表明模型结果具有良好的可信度。
The neotectonic movement is active in Yongsheng, Binchuan and their neighboring area in Yunnan Province, and the giant landslides in the area are significant. The objective of this paper is to analyze the distribution law and the dynamic genesis of the landslide and the relationship between the distribution of landslide sites and geological features. For this purpose, based on the visual interpretation of 615 landslides, eight influential factors of landslides (faults, drainage, lithology, NDVI, slope, aspect, Topographic relief, elevation) were selected to import to the cascade BP neural network model to perform analysis. Results show that, high influential factor regions had collective characteristics, and based on this, 4 high index regions were drawn. Further study suggested that, the gene- sis types of landslides in this area consisted of landslide triggered by earthquake, erosion, and colluvial load, respectively, and especially, the active fracture played an important role in the landslide. Combined with field inves- tigation on the typical ancient seismic landslides, it is proved that, the long- term intense activity of the Chenghai -Binchuan fracture zone in quaternary played an important control role in the landslide growth in Yongsheng and Binchuan area, and at the same time, it is proved that the model results were reliable.
出处
《自然灾害学报》
CSCD
北大核心
2016年第1期176-190,共15页
Journal of Natural Disasters
基金
中国地质调查局项目(1212011120163
12120114002101)
中国地质科学院地质力学研究所基本科研业务专项(DZLXJK201410)~~
关键词
青藏高原东南缘
滑坡影响因子
BP神经网络
程海—宾川断裂带
地震滑坡
southeastern margin of the Qinghai - Tibetan Plateau
influence factor of the landslide
BP neural net-work
Chenghai-Binchuan fault zone
seismic landslide