Quality parameters for passive image interferometry tested at the Groningen network

Passive image interferometry is widely used to retrieve velocity variations as function of time. The cross-coherence (spectrally normalized cross-correlation) of ambient noise recorded at two receivers is an estimate of the Green’s function that due to velocity changes will be stretched or compressed in time. The relative velocity change (v/v) is determined by the time stretching (−t/t) that yields highest correlation between the reference cross-coherence and stretched lapse cross-coherence. The estimation of v/v could be used, for example, to warn for hazardous situations developing in a volcano setting or due to degradation of a civil engineering structure. Before a warning would be issued, however, one would like to have a handle on the quality of the medium-change estimate.
In the Groningen area in the Netherlands a well-sampled network of seismometers exists.
This allows direct assessment of the quality of the v/v estimation by many different receiver paths sampling the same medium. We use this quality assessment of v/v to test other possible quality parameters that could also be used in settings with only a sparsely sampled seismic network. The quality of the v/v determination appears to be well described by the correlation coefficient which is used to determine the velocity change. Consequently, in order to measure medium changes with a high certainty, it is important to have a high consistency between the lapse and the reference cross-coherence. A quality estimate based on this correlation coefficient can also be applied when there is only one receiver pair available. If the quality is insufficient, it can be improved at the cost of temporal resolution.
This study also investigates possible causes of the measured medium changes: variations in temperature, soil moisture, air pressure, water table, gas production rate and subsidence. We find a weak anticorrelation with temperature, and a weak correlation with the gas production rate and subsidence. The observed medium change is likely a complicated combination of different processes taking place.