A new approach for time-lapse data weighting in ERT

Applications of timelapse inversion of electrical resistivity tomography (ERT) allows
monitoring variations in the subsurface that play a key role in a variety of contexts. The
inversion of timelapse data provides successive images of the subsurface properties
showing the medium evolution. Images quality is highly dependent on the data weighting determined from the data error estimates. However, the quantification of
errors in the inversion of timelapse data has not yet been addressed. We propose a
methodology for the quantification of timelapse data error based on the analysis of the
discrepancy between normal and reciprocal readings acquired at different times. We
apply the method to field monitoring data sets collected during the injection of heated
water in a shallow aquifer. We tested different error models to show that the use of an
appropriate time-lapse data error estimate yields significant improvements in terms of
imaging. An adapted inversion weighting for time-lapse data implies that the procedure
does not allow an over-fitting of the data, so the presence of artifacts in the resulting
images is greatly reduced. Our results demonstrate that a proper estimate of time-lapse data error is mandatory for weighting optimally the inversion in order to obtain images that best reflect the medium properties evolution through time.