A comparison of methods for estimating the geoelectric field

The geoelectric field is the primary input used for estimation of geomagnetically induced currents (GICs) in conducting systems. We compare three methods for estimating the geoelectric field given the measured geomagnetic field at four locations in the U.S. during time intervals with average Kp in the range of 2-3 and when the measurements had few data spikes and no baseline jumps. The methods include using (1) a preexisting 1-D conductivity model, (2) a conventional 3-D frequency domain method, and (3) a robust and remote reference 3-D frequency domain method. The quality of the estimates is determined using the power spectrum (in the period range 9.1 to 18,725 s) of estimation errors along with the prediction efficiency summary statistic. It is shown that with respect to these quality metrics, Method 1 produces average out-of-sample electric field estimation errors with a variance that can be equal to or larger than the average measured variance (due to underestimation or overestimation, respectively), and Method 3 produces reliable but slightly lower quality estimates than Method 2 for the time intervals and locations considered.