Road sinkhole detection with 2D ambient noise tomography

Seismic methods are often used for detection of pre-collapsed sinkholes (voids) under roadway for remediation to minimize the risk to the safety of the traveling public. While the active-source seismic methods can provide accurate subsurface profiles, they require closing the traffic flow for hours during testing and potentially cause sinkhole collapse due to ground perturbation by source excitation. To address these issues, we present a new 2D ambient noise tomography (2D ANT) method for imaging voids under roadway. Instead of using the approximated Green’s function, whose required assumption of energy balance at both sides of each receiver pair is rarely satisfied, the cross-correlation function of traffic noise recordings is inverted directly to obtain velocity structures. To adopt the concepts of seismic interferometry and derive the model structural kernel, passing-by vehicles are assumed as moving sources along the receiver array. The source power-spectrum density is determined via the reverse-time imaging approach to approximate the source distribution. The 2D ANT method is first demonstrated on a realistic synthetic model with the accurate recovery of the model variable layers and a buried void. To demonstrate its effectiveness to the real-world problems, we successfully applied it to field data for assessment of a repaired sinkhole under the US441 highway, Florida, USA. The field experimental result shows that the method is capable of resolving the subsurface S-wave velocity ( VS) structure and detecting a low-velocity anomaly. The inverted VS profile from the 2D ANT generally agrees with that of 2D active-source full-waveform inversion, including the VS value and depth of the anomaly. To our best knowledge, this is the first study to directly invert the waveform cross-correlation of traffic noise recordings to extract material property at the engineering meter scale (<30 m depth).