Spatial-temporal heterogeneity of land subsidence evolution in Beijing based on InSAR and cluster analysis

Thursday, 18 December 2014
Yingchen Li, Yinghai Ke, Huili Gong, Xiaojuan Li, Lin Zhu and Beibei Chen, CNU Capital Nornal University, Beijing, China
Land subsidence is a common natural hazard occurring in extensive areas in the world. In Beijing, the capital city of China, there has been serious land subsidence due to overexploitation of ground water during the recent decades. Five major subsidence tunnels have formed. Across the Beijing plain area, the ground is sinking at the rate of 30-100mm/year. Uneven subsidence leads to ground fissure and building destruction, and has caused great economical and property loss. To better characterize and understand regional land subsidence evolution, it is critical to monitor the time-series dynamics of subsidence, and capture the spatial-temporal heterogeneity of the subsidence evolution. Interferometric SAR technique, as it provides high spatial resolution and wide range of observation, have been successfully used to monitor regional ground deformation.

The objective of this study is to derive time-series regional land subsidence dynamics in Beijing, and based on which, analyze and assess the spatial-temporal heterogeneity of the evolution using cluster analysis. First, ENVISAT ASAR (2003-2009 years, 28 scenes, track number: 218) datasets during 2003-2010 covering Beijing plain area were utilized to obtain time-series subsidence rate using Persistent Scatter InSAR (PS-InSAR) technique provided in SARProz software. Second, time-series subsidence characteristics of the PS points were analyzed and the PS points were clustered based on Self-Organization feature Maps (SOM) algorithm considering environmental factors such as groundwater level and lithologic characters.

This study demonstrates that based on InSAR measurements and SOMs algorithm, the spatial-temporal heterogeneity of land subsidence evolution can be captured. Each cluster shows unique spatial-temporal evolution pattern. The results of this study will facilitate further land subsidence modeling and prediction at regional spatial scale.