基于相干目标的干涉图叠加方法监测天津地区地面沉降

Monitoring Subsidence in Tianjin Area Using Interferogram Stacking Based on Coherent Targets

利用ENVISATASAR数据,采用基于相干目标的干涉图叠加方法,对天津地区的地面沉降现象进行了DInSAR监测试验。差分干涉处理采用"两轨法",使用校正了高程异常的SRTM DEM数据消除高程相位。以相干系数为标准选取了相干目标,解缠过程中运用了Delaunay三角剖分和权重最小费用流算法。本文获得的季度平均沉降速率图有效揭示了试验区地面沉降的空间展布及相对形变量,但其获得的绝对形变量尚需地面实测数据校验。

The land subsidence of Tianjin area,mainly induced by withdrawal of ground water,has been measured by leveling measurements for many years.Although the monitoring results of leveling technique are reliable,the sparse leveling data grid in the area and the long time span on a single bench mark prevents us from understanding the process of deformation.Spaceborne differential radar interferometry(DInSAR) has been proven a remarkable potential for mapping ground deformation phenomena over tens-of-kilometers-wide areas with centimeter-scale accuracy on a more dense space grid and time series than leveling and GPS technique.As well known,geometrical and temporal decorrelation is an important factor that prevent DInSAR from being an operational tool for displacement monitoring.Moreover,atmospheric inhomogeneities produce an atmospheric phase screen(APS) on every SAR image,which can contaminate the results of the deformation monitoring.Interferogram stacking is a technique to improve the relative accuracy of SAR interferometric surface displacement mapping based on a combination of multiple interferograms.Under the assumption of statistical independence of the atmospheric distortions,the displacement terms add up linearly whereas the error term increases only with the square root of the number of pairs considered.Using ASAR images and the approach of interferogram stacking,the subsidence phenomena of Tianjin area has been mapped.9 ENVISAT ASAR images covering the period from October 2003 to August 2004 have been selected to retrieve the process of subsidence in Tianjin area.To reduce the geometric decorrelation and topographic errors,13 pairs with perpendicular baselines minor than 300 meters are chosen from the possible combinations.In the process of removing topographic component from the inerferometric images,SRTM DEM data are used.Since the orbit data of ENVISAT have been referenced to the WGS84 ellipsoid,the SRTM DEM height values the EGM96 geoid as the reference choose.The geoid height in the area has been compensated.Among the 13 differential interferograms,only one of them is believed to be severely affected by the atmospheric artifacts,the others show almost the same deformation phase model over the work area.Since it is difficult to discriminate displacement phase contributions from the atmospheric signature only by using individual interferogram,the approach of interferogram stacking is used.5 unwrapped phase images are summed and the time span is 350 days.Before phase unwrapped,the coherence values of pixels in all the interferograms are taken into consideration while coherent targets are selected.For the problem at hand,where in a large area high coherence urban patches are surrounded by vegetation-covered field with low coherence,the method of phase unwrapping firstly triangulates the unmasked pixels and then unwraps the phases based on the minimum cost flow algorithm with the averaged coherence map as a weight file.The geocoded subsidence map of test area during 90 days shows the distribution and the relative deformation value of the displacement field.Compared with previous measurements from DInSAR and in-situ GPS,the subsidence cones characterized by the method are reliable.But the temporal decorrelation and the atmospheric distortion are not completely overcome,and the deformation estimations derived from this method still need validation and correction by in-situ measurements mainly observed in Tianjin urban area.In the future work,with more ENVISAT ASAR data acquired,we plan to apply the PS InSAR and derived methods to exploit the data with long spatial and temporal baseline as much as possible.