Study On The Monitoring Of Mining-induced Subsidence In Loess Hilly Gully Region Based On Time Series DInSAR Technique

Interferometric Synthetic Aperture Radar(InSAR) and differential InSAR (DInSAR)techniques, as one of the key spatial techniques of the earth observing system, are rapidlydeveloped in recent decades, which have been widely applied in earth science especially insurface deformation monitoring, such as the monitoring of earthquake, volcano, minecollapse land subsidence, landslide and glacier etc., for their all-weather conditions, largecoverage, fast reaction and high precision characteristics. The land subsidence is closelyrelated with human activities, which concerns the security of human lives and property. So itis essential to study the theories and methods of the land subsidence monitoring, which hasgreat scientific and practical values for the fine explanation，mitigation and forecasting ofgeo-hazards．Aiming to overcome the temporal and spatial decorrelation problems of the conventionalDInSAR technique, time series SAR interferometry techniques such as permanent scatterinterferometry, small baseline subset analysis, interferogram stacking, have been focused onand confirmed success to detect the ground subsidence on the base of coherent pointtarget(CPT)during the long time interval. However, these techniques, which cannot begenerally used, are still to be optimized. The environmental differences of the study areamight have great impact on data preprocessing. In addition，it is necessary to adjust theworkflow for different data sources.Ningtiaota mining region, located in Shenmu county in the province of Shaanxi, isselected as the test region for its loess hilly gully characteristics. Based on9ENVISAT/ASAR images and the time serial DInSAR，several critical parts, such as timeseries differential interferogram formation, identification and selection of the coherent pointtarget, unwrapping of the three-dimensional phases, and separation of the noise phases causedby different factors were focused on and deeply analyzed, and the land subsidencedevelopment of Ningtiaota mining region during2009.04to2010.09is acquired. The resultswere then validated with the ground-based leveling data and the mining work.In this dissertation, several aspects were discussed and some satisfying results have beenachieved as the following:1) Based on the analyses of the basic principle, technical characteristics, error sources and data processing flow of DInSAR, the phase sensitivity of reference DEM error, thesatellite orbit error and ground deformation was analyzed, and their relevant formulas werededuced as well.2）Selection of the coherent point target．From the viewpoint of amplitude, phase andtheir conjunction, several methodologies for the identification of the coherent point targetswere discussed and validated in this dissertation, based on amplitude steadiness, phasesteadiness, temporal correlation and amplitude dispersion threshold. According to thecharacteristics of SAR datasets in this paper, the union set of temporal correlation coefficientthreshold and amplitude dispersion threshold were put forward, as well as the relevant model,to improve the accuracy evaluation for all the point target.3) The inversion model of the land subsidence rate. The model included two aspects.One is the phase unwrapping, the other is the separation of the noise phase parts from theland subsidence information. Since the phase information of the time series coherent pointscould be regarded as a three-dimensional phase sets，a step-wise minimum cost network flowmodel was used in this dissertation for the phase unwrapping based on the sparse points. Thecumulative coherence coefficient and temporal coherence were respectively used as costfunction for phase unwrapping model in time dimension and in the space dimension. Thespatial autocorrelation of the adjacent PS points was used for modeling. The ground lineardeformation velocity was estimated with the model correlation coefficients maximizing andlinear regression analysis. Meanwhile, based on the results of phase unwrapping and thecharacteristics of the remaining residual phase, the temporal high pass filtering and spatiallow pass filtering methods were adopted to separate the non-linear deformation componentand the atmospheric noise component.4) Based on9ENVISAT/ASAR images, the land subsidence development of Ningtiaotamining region during2009.04to2010.09is acquired and analyzed with the ground datameasured. As well, the DEM residual errors and the atmosphere noise are spiritedindividually.