Transfer Function Estimation for Natural Electromagnetic Fields

Four techniques are compared for estimating the transfer function between fields measured at one site with fields measured at another site, or with different fields measured at the same site. The techniques are:

1. Linear, point-to-point time domain estimation (LPP),
2. Conventional frequency domain transfer function estimation (FD),
3. Multichannel Wiener estimation in the time domain (WE), and
4. Adaptive filter processing in the time domain (AFP).

These four transfer function estimation methods have been applied to high precision SQUID magnetometer array data collected in Grass Valley, Nevada and Long Valley, California, using natural fields in the frequency band 10^-4 to 1.0 Hz. To monitor the ‘signal’ introduced by motion of the sensors in the Earth’s large static field, high sensitivity cryogenic tiltmeters were used in each magnetometer. The transfer function estimation methods above have accounted for more than 50 db of the signals received at one site using fields measured at another site. Removal of motion-induced magnetic signals provides as much as 10 db further reduction in the residual signal compared to using solely electromagnetic field signals.