Guanren Wang
Until recently, magnetic field data measurements at geomagnetic observatories have been made at one-minute mean cadence. With the demands of space weather research and forecasting, one-second data values are now desired. Fluxgate magnetometers are widely used in geophysical observatories because of their ability to measure the Earth’s main field and their long-term stability. However, fluxgates have an inherent instrumental noise floor, which is large enough to “mask” ambient natural signals. Consequently, in the older generation of fluxgates, data in the high-frequency band between 1 and 10 seconds are noisy and do not satisfy the international standard for one-second data. induction coil magnetometers are optimised to capture field signals at the high-frequency spectrum (e.g. 0.1-100 Hz) with minimal instrument noise, but lack any long-term stability that characterises fluxgate magnetometers. The obvious solution is to combine the data from both types of instrument.
We merge time-series data from both an induction coil and fluxgate magnetometer, sited at Eskdalemuir Geophysical Observatory (UK), to produce an "improved" one-second time series for frequencies higher than the fluxgate noise level, in the 0.2 - 0.5 Hz band (periods of 2-10 seconds). Our goal is to combine the long-term stability of the fluxgate magnetometer with the low-noise of the induction magnetometer to capture the natural magnetic field for frequencies down to 1 second in the two horizontal components of the magnetic field.
We investigate the improvement by examining the merged time-series, computing the coherence and phase of the one-second merged data with that of the input induction coil and fluxgate magnetometer data. Our poster describes the data, methodology and results of our efforts to merge the data from these two complementary instruments. Only the north-south (X) component results are included in this poster.
The Induction Coil (IC) magnetometer has excellent high-frequency sensitivity, but long-term drifts. The fluxgate (FGE) magnetometer has excellent long-term stability, but high-frequency noise issues. At present, one-minute mean values are processed from 1Hz FGE data. The FGE data are not good enough for one-second data on their own. We want to make one-second data by combining data collected from these two different systems with complementary properties. Eskdalemuir (ESK) geophysical observatory is equipped with IC and DMI FGE magnetometers.
Slide 2. Top two figures on this slide shows example of minute mean data plotted on magnetograms, displayed as traces of changing magnetic values over a full-day. The top left figure shows periods of sharp magnetic field disturbances on a noisy day caused by a very large solar flare event. One can observe sine-wave (diurnal) variation on the magnetogram traces during a geomagnetically quiet day on the top right.
To combine FGE and IC data to produce better one-second data at ESK, we use an approach developed by Brunke, et al (2017) based on the concept that the signal of an IC is proportional to the changing magnetic field component along the axis of the coil. We can use this mathematical relationship to derive the merged one-second data for a given day. This process can be visualised by fitting the integrated IC curve to the measured (‘raw’) FGE curve shown in the lower right plot. At time t=0s of the integration process represents the noise reduced value Bx at t=0s. The shape of the fitted green curve results from the IC measurements.
Slide 3. The top right figure shows that the coherence between the resulting merged data and FGE data are perfect from 100 to 33s period with a minimal phase shift. The merged and IC data show a mostly good coherence between 30 to 10s period, but the inexact coherence is due to the low energy content of the signal at frequencies between 30mHz and 0.5Hz. Below 33s period, the IC and merged coherence shows a peak coherence value at 10s before dropping at periods below 10s.
The lower plot shows the power spectra of the “raw” DMI FGE, IC and their merged data along the X-component. From this, we can observe a natural decay of the FGE power spectra trace from lower frequencies up to a frequency around 30mHz, where it then flattens at a constant noise level of 0.2nT^2 Hz^-1. The IC trace extends to a lower constant noise level, but at a significantly higher frequency than the FGE trace. The merged trace follows the continuing decay of the natural field for 05-Jun-2019 above 30mHz. Towards the left-hand side of this plot, the merged and FGE trace follows an identical path below 30mHz, but the IC trace drifts at lower frequencies as expected.
Slide 4. Power spectra density spectrograms are overlain with time-series of the natural field in black. Left and right panels allow for direct comparison between the spectrogram computed from FGE only and from the merged one-second data in the X-component, for the same day, between 0.02 Hz to 0.5 Hz. At frequencies above 30mHz, the merged data spectrograms suggest FGE instrument noise are cleansed, from which we can observe that the merged data can better resolve changes up to the Nyquist frequency (0.5 Hz) for noisy and quiet days.