ENHANCEMENT OF SIGNAL‐TO‐NOISE RATIO IN MAGNETOTELLURIC DATA

Abstract

A new technique has been developed which can be employed for the enhancement of the signal‐to‐noise ratio of magnetotelluric data and/or for the establishment of a confidence limit as a criterion for final data selection. This technique functions in a consecutively cyclic process which can finally remove the noise from the autopower estimates of MT data if the crosspower estimates are noise free and, consequently, the predicted coherencies can be improved. When noise exists in both autopower and crosspower estimates, the cyclic operation can establish a confidence limit which reveals a direct measure of incoherent noise content and thus is a more reliable criterion than the predicted coherency itself for data selection. The technique is tested using actual MT data and the result shows the adequacy of the technique. It selects useful data from those originally having moderate or low predicted coherencies; thus it appears to be important when the predicted coherency of the majority of the collected data is not high.