Calibration and filtering strategies for frequency domain electromagnetic data

Repeat frequency-domain electromagnetic (FDEM) surveys have been acquired over agricultural fields in the Powder River Basin (PRB), Wyoming, where subsurface drip irrigation is being utilized for the beneficial dispersal of coalbed methane produced water. The purpose of the FDEM surveys is to monitor changes in subsurface electrical properties due to the injection of the produced water. In order to quantitatively interpret the data, however, both systematic and random errors must be accounted for. A calibration procedure, adapted from airborne geophysical data processing, corrects for systematic errors by making the FDEM data consistent with the results of a direct current resistivity survey that is coincident with a portion of the FDEM data. Calibration is shown to improve the inter-frequency relationships within the data, resulting in reduced misfit when the data are inverted and therefore added confidence in the inversion results. A filtering approach that is based on principal component analysis is used to attenuate random errors in the data. This type of filter is advantageous because it has a physical-basis in the fact that FDEM data are highly correlated across frequencies, and does not require an arbitrarily-defined spatial filter window length. The calibration and filtering methods are successfully applied to approximately 15 line-km of data in the PRB. It is apparent, however, that calibration parameters may drift in time and should be re-assessed at regular intervals throughout a survey.

► We introduce data processing strategies for frequency-domain electromagnetic data.
► Systematic errors are corrected by calibrating to a dc resistivity survey.
► Random errors are attenuated using a principal component analysis filtering approach.
► Application to a field survey in Wyoming results in improved inversion data misfit.