Data Selection For Imaging Near-surface Passive Surface Waves

With the rapid development of urbanization,urban underground space investigation plays an increasingly significant part in the utilization of underground space.Passive surface-wave methods,as a powerful tool of urban underground space investigation,have been gradually obtained attention by urban geophysicists in recent years.In near-surface applications,active surface-wave survey is able to solve many shallow geoengineering problems but cannot satisfy specific demands of deeper exploration depth.Passive surface-wave methods can extract surface waves from long noise recordings to obtain near-surface shear-wave velocity structure.Compared with active surface-wave survey,passive surface-wave methods do not require excitation sources and have ability to increase exploration depth.Passive surface-wave methods face many challenges in practical applications,however,such as the heterogeneous source-distribution of ambient noise and the complicated anthropogenic environment,which make it difficult to obtain high-quality dispersion images of passive surface waves.Therefore,this thesis aims to improve the accuracy of passive surface-wave imaging by data selection.Firstly,three passive surface-wave methods suitable for linear deployment are introduced in detail,namely refraction microtremor(ReMi),roadside passive multichannel analysis of surface waves,and multi-channel analysis of passive surface waves(MAPS).Numerical experiments indicate that surface-wave energy generated by the MAPS can correctly match the theoretical dispersion curve in full-frequency band.The effects on passive surface-wave imaging due to noise-source distribution using the MAPS are further discussed.Results show that to select the anti-causal or causal part of cross-correlation depends on noise-source distribution.As for one-side noise-source distribution,phase velocity of surface waves can be easily picked on the dispersion image without artifacts as long as the anti-causal or causal part of cross-correlation is reasonably selected.For bidirectional noise-source distribution,it becomes more complicated.Specifically,the average of two parts should be selected for noise-source distribution that left-side sources are equal to right-side sources;the anti-causal part should be selected for noise-source distribution that left-side sources is weaker than right-side sources;the causal part should be selected for noise-source distribution that left-side sources is stronger than right-side sources.Selecting the anti-causal or causal part of cross-correlation properly can not only attenuate artifacts but also enhance surface-wave energy on the dispersion image.Furthermore,traffic noise along a railway is used as a typical example in order to prove the reliability of the numerical results.It is difficult to obtain accurate and reliable surface-wave velocity information from the dispersion image for low-quality noise recordings by currently available methods.With data selection proposed by this research,according to the imaging feature of each segment,all noise segments are manually classified and selected.Surface-wave energy is clearly seen on the dispersion image after data selection and interest frequency band is effectively broadened.Results show that data selection can significantly improve the accuracy of surface-wave image generated by the previous mentioned three passive surface-wave methods.Finally,according to the previous numerical experiments and filed examples,this thesis presents an automatic data selection technique based on signal-to-noise ratio(SNR)in the time domain for selective stacking of cross-correlations.Three examples demonstrate that the proposed data selection technique can preserve the time segments with coherent signals,and reject the time segments polluted by non-stationary noise sources and spatial aliasing in order to improve the accuracy of passive surface-wave imaging and achieve the same imaging as manual selection.Besides,the influence of velocity window parameter for SNR calculation on the performance of the proposed technique is further discussed.Results suggest that the velocity window should be appropriately wide to obtain clear and stable surface-wave energy on the dispersion image after data selection.