Research And Application Of Full Resolution Fast InSAR Time-series Analysis Method

Sentinel-1 data are widely used in in earthquake and geological disaster monitoring and rescue,such as seismic deformation field and fracture sliding inversion,landslide normal motion monitoring and accelerated motion warning,which are important for natural disaster prevention and mitigation.However,due to the high resolution and short revisit period of Sentinel-1 data,the execution of large-scale InSAR processing can lead to high computational costs,and traditional processing methods cannot meet the demand of providing deformation results in a short time for realistic emergency rescue scenarios.In addition,in non-urban areas with high vegetation coverage,interference decoherence is likely to occur,and the number of high-quality coherence points obtained by existing time-series analysis methods is limited,resulting in the loss of phase unwrapping accuracy.To solve these problems,this paper presents a new InSAR time-series analysis workflow and applies it to the study of ground subsidence and tailings pond failure,and achieves the following results:(1)High-performance computing framework construction for massive InSAR data.The rise and development of high performance computing(HPC)methods have provided new ways of solving large-scale InSAR processing problems.When it is applied to the large-scale SAR data processing,it can significantly improve the data processing speed.Graphics processing units(GPUs),as a representative technology have received increased attention because of their robust parallelism,large memory bandwidth,and low power consumption,in addition to their excellent performance in solving parallel processing and calculation-intensive problems.Based on the SBASInSAR data processing flow,this paper proposes a fast time-series analysis method for full-resolution InSAR based on the GPU high-performance computing framework.For Sentinel-1 full-resolution single-look SAR data,GPU parallel algorithms are used to accelerate the key time-consuming steps in InSAR processing,including image coregistration,resampling,and enhanced spectral diversity(ESD).Compared with the traditional InSAR processing flow,the efficiency of Sentinel-1 TOPS data processing is significantly improved,which is important for rapid post-disaster emergency response and massive data processing.(2)Extraction of homogeneous image elements based on optimal parameter statistics.The homogeneous samples selection algorithm selects spatial samples with similar spatial backward scattering properties by comparing the statistical properties of the intensity of the central pixel of the SAR sample with the pixels within the window,and then pools similar pixels for parameter estimation.It is shown that statistical homogeneous pixels(SHPs),identified by statistical tests and stochastic models,can improve the estimation accuracy of distributed targets.The statistical homogeneous pixel selection(SHPS)method uses the flush samples set selected by the LRT as the initial value,and then refines the reference pixels to control Type Ⅱ errors using narrower confidence intervals(CIs)and high quality observation iterations,which can minimise the heterogeneity of the samples.The results of the SHPS algorithm are used for interferogram masks to remove low coherence points in wetlands,vegetation cover areas and low reflectance areas to achieve high accuracy extraction of homogeneous coherence points.(3)Large-scale surface subsidence monitoring in the Yellow River Delta.The LOS deformation rate in the Yellow River Delta was obtained using SAR datasets acquired between January 2021 and February 2022.The InSAR results revealed a general stability in our study area during the monitoring period of this paper.However,there remained several severe areas with a certain degree of subsidence in the study area.Underground brine and hydrocarbon extraction,as well as sediment consolidation and compaction,all contribute to land subsidence in the Yellow River Delta when combined with the geological environment.(4)Time series analysis of tailings dam deformation.Using 91-period uplift Sentinel-1 satellite data from 8 January 2019 to 17 March 2022,surface deformation rates and time-series cumulative deformation at a spatial resolution of 20 m were obtained for three years before the dam failure occurred in a tailings pond in Luliang,Shanxi Province.The InSAR results show that the tailings dam exhibited significant deformation characteristics prior to the collapse,with the average cumulative line-ofsight(LOS)directional deformation approaching 80 mm for the secondary stockpile dam and over 140 mm LOS directional deformation for the primary stockpile dam.Analysis of the local rainfall and cumulative deformation variables shows that rainfall has a significant effect on the deformation of the tailings ponds,but the maximum deformation rate lags behind the peak rainfall by approximately one month.