Adaptive separation of regular and irregular magnetic activity for K indices

Abstract A numerical data adaptive separation method is developed for the discrimination of quasi-regular magnetic variations from the irregular variations for the determination of the K indices. The method is based on a pattern comparison and discrimination technique using harmonic analyses and is more objective than the hand-scaling method. The wave characteristics of a quiet regular or reference day are compared with those of the disturbed day being examined. The harmonic components that are consistent with those of the reference day are determined by a cosine weighting function. The Fourier coefficients are calculated from these modified harmonics and are then transformed back to the time domain. This ‘regular variation’ is subtracted from the data before the ranges are determined for the semi-logarithmic K scale. The method is tested on digital data from the Canadian magnetic observatories Victoria, Meanook and Ottawa over a 4 month period. The data are at a 10 s sampling interval and encompass the full range of K values (0–9). These K values are compared and checked with hand-scaled values, which were carefully and independently determined from reconstructed magnetograms from the same data. The two methods were found to agree ∼ 97% of the time, with most errors occurring during very irregular periods. This is comparable with values carefully determined by highly skilled observers. It is proposed that the method is suitable for use on a regular basis.