Seawater motion-induced electromagnetic noise reduction in marine magnetotelluric data using current meters
KeywordsMarine magnetotelluric Adaptive correlation noise-canceling filter Seawater motion-induced electromagnetic noise Current meter
controlled source electromagnetic
ocean bottom electromagnetic receiver
robust multivariate errors-in-variables
short-time Fourier transform
Since the 1970s, a wide range of electromagnetic (EM) methods, such as the magnetotelluric (MT) method, have been used for deep marine detection (Vozoff 1972) and have proven successful for studying volcanoes (Constable and Heinson 2004), mid-ocean ridges (Baba et al. 2006), subduction zones (Naif et al. 2015), and in petroleum exploration (Key et al. 2006). Marine MT surveys use ocean bottom electromagnetic (OBEM) receivers on the seafloor to measure natural MT source signals (Filloux 1980). Recently, due to advancements in semiconductor and information technology, these instruments now have broader bandwidths, lower power consumption, and a lower level of internal sensor noise, especially in the effective bandwidth that is extendable to 0.1 Hz in deep water (approximately 1000 m) (Constable et al. 1998).
While an OBEM is deployed on the seafloor, it records the MT signal, which is attenuated due to highly conductive seawater; internal instrument noise from the electrode, induction coil, and data logger; and EM noise from the environment. EM noise from the environment includes noise from nearby vessels, onshore cultural noise, biological noise, and seawater motion-induced noise. In deep water settings, both vessel and onshore cultural noise are negligible, and there is a low probability of biological noise. EM noise is caused by seawater current EM induction and by rocking of the instrument. It is well known that mechanical motion of an induction coil in the Earth’s magnetic field will generate an additional magnetic field, and hence correlated noise, but this should not affect the electrical field unless v × B is comparable to the ambient field. Lezaeta et al. (2005) previously showed that some sites display electric as well as magnetic field variations that correlated with tilt, suggesting EM induction due to water motion. Seawater motion-induced EM noise is therefore possibly a large source of noise in deep water. This type of noise not only rocks the instrument, but also generates EM noise that couples with the Earth’s magnetic field. The coupling of seawater motion-induced EM noise with the Earth’s magnetic field will produce an obvious magnetic field and, hence, correlated EM noise (Sanford, 1971; Chave and Luther, 1990) that affects the electric field (E = v × B). This is comparable to the locally induced ambient magnetic field in the surrounding medium, which occurs in the highly conductive ocean and is subject to various water disturbances.
Previous studies have unsuccessfully attempted to reduce seawater motion-induced EM noise. Key and Constable (2002) used a multi-station process code developed by Egbert (1997) to process marine MT data, which is a robust multivariate errors-in-variables (RMEV) approach. When EM noise is too strong, improvement is limited, and data contaminated by the induced EM noise must be removed. Lezaeta et al. (2005) performed noise corrections on shallow-water EM data affected by instrument motion. They measured tilt data to correct for instrument motion noise. Instrument motion noise contaminates the magnetic file channel and induces EM noise in all channels that contain an E- and B-field channel. Neska et al. (2013) reduced motion noise in marine magnetotelluric data by using tilt records, and applied a standard motion-removal technique as well as a newly developed method to correct for motion-induced noise. The tilt measurement method can be used for instrument motion and is less useful for seawater motion-induced EM noise. Modifications to buoyancy design has increased the mechanical stability of the instrument and decreased instrument motion noise, however induced EM noise is still an issue.
Water movement not only occurs at the sea surface, but also along the seafloor. Correspondingly, seafloor current-induced EM noise exists both in shallow and deep water. To analyze the seafloor currents, we developed an OBEM with a current meter that simultaneously measured current data along with the OBEM data logger. If we can obtain seafloor current speeds, we can reduce induced EM noise by using the current meter data to apply an adaptive correlation noise-canceling filter to the OBEM data.
This paper mainly focuses on (1) the discussion of seawater motion-induced EM noise which affects marine MT data and (2) the development of a time-domain separation technique for correlated noise, using an adaptive correlation noise-cancellation filter, before estimating the frequency-domain MT apparent resistivity and impedance phase.
Between August and September 2018, two gas hydrate exploration surveys were conducted in the Shenhu and Qiongdongnan areas of the northern South China Sea. The objective of these surveys was to characterize the background geology and electrical conductivity structure of the region prior to continued exploration in the area, which is a prospective gas-hydrate area (JianEn et al. 2018). Shenhu is located at the edge of the continental shelf, which produces strong currents, and water depths range between 900 and 1700 m. The Qiongdongnan area is a marine basin at the base of the continental shelf with a water depth of approximately 1750 m. Seawater motion in the basin area is quiet compared to the continental shelf/slope area.
The OBEM records variations in both the Earth’s natural-time MT field and the synthetic artificial source derived from the transmitter, as well as any noise generated by seawater motion. After instrument recovery, we downloaded the raw data from the OBEM data logger and current meter and calculated the five EM component power spectra using a short-time Fourier transformation (STFT) algorithm. We set the length of the STFT to 4096 and the effective frequency band from approximately 0.05 to 75 Hz. All of the raw time series were calculated using this method.
At SH-R11, current speed peaks over 20 cm/s are ten times what is expected in waters of comparable depth. We observed no 12- or 24-h period tidal modulations, but shorter period variations had a similar magnitude and there were long-term variations in speed that may have been modulated by tides. Noise existed throughout the entire deployment period, where approximately one-third of the data appear noisy. However, as observed from the E- and B-field power spectra, seafloor currents are characterized by a very rapid increase in magnitude on a very short timescale. The synchronous noise effect from the E- and B-fields can explain noise that is mainly due to seawater motion-induced magnetic fields. There is a strong correlation between EM noise and velocity: when velocity was faster than 2 cm/s, we observed clear EM noise signals in the data.
Figures 3 and 6 show that whenever current values are high, there are spectral maxima in at least two of the five measured EM components at the same frequency and at its harmonics. Figure 6 shows that such enhanced signals at this peak frequency can occur in the magnetic field, yet be absent from the electric components at the same time. The explanation for these effects is that the instruments are rocking, and it is their motion that matters for these characteristic frequencies. The EM noise source in Shenhu area was mainly caused by current-induced noise and instrument’s rocking, while the EM noise source in Qiongdongnan was mainly caused by the instrument rocking.
Adaptive correlation noise-canceling filter
Because of inevitable and uncontrollable seawater motion, and because the velocity of seawater motion is highly variable over time, the electromagnetic noise caused by seafloor currents is also highly variable over time. Therefore, it is almost useless to use the frequency domain transfer function method to implicitly rely on a stationary assumption to remove noise. An adaptive filtering noise removal method for non-stationary time-domain data signals is therefore proposed. The most commonly used method is adaptive correlation noise removal (Haykin 2002). The adaptive correlation canceller combines the time-varying filter coefficients, which change in the best sense, and eliminates the unknown interference contained in the main signal by using a noise reference (where the main signal is weak). In our research, the primary polluted signal is a component of the magnetic field and the electric field, and the reference (noise) signal is the two horizontal components of current velocity. Each adaptive filter is a linear transverse finite impulse response (FIR) filter whose coefficients are updated by the least mean squares (LMS) algorithm at each time step.
MT data-processing results
To improve data quality over the entire frequency range for other sites, nearby reference processing by setting up the current data as a nearby reference site for other sites in the Shenhu area was applied, up to approximately 1 km from the current meter measurement site (SH-R11). The filter successfully improved the MT impedance of both local sites and other nearby sites.
The data were processed using the previously mentioned adaptive correlation noise-canceling filter. Finally, the MT impedances were estimated using the SSMT2000 software package from Phoenix Geophysics. The MT impedances were computed for each frequency band, and the apparent resistivity and the phase curves for the entire frequency range were all obtained.
The data quality was obviously improved and is comparable to the results obtained without the above-mentioned filter. Even though further work is still required, these preliminary data-processing results demonstrate that utilizing adaptive correlation noise-cancellation filters with current data is an effective method for MT data processing.
Discussion and conclusions
Depending on offshore conditions, marine MT signals are “polluted” by seafloor current-induced EM noise at varying levels. The raw MT time series from the continental shelf in the Shenhu area shows such strongly correlated noise that we could not obtain useable impedance estimates using traditional data-processing techniques. The basin stations in the Qiongdongnan area show much less current-related EM noise and have longer quiet segments in the data than in the Shenhu area. The maximum current speeds in Shenhu and Qiongdongnan were approximately 20 cm/s and 10 cm/s, respectively, and EM noise was apparent whenever the current speed was faster than 2 cm/s.
In this paper, we show that an adaptive correlation noise-cancellation filter is an effective method for MT data processing and can be used to reduce induced EM noise by using local current data. Therefore, this method is applicable to datasets with co-located current data, and has the potential to be an effective filter for MT data processing. The adaptive correlation noise-cancellation filter improved data quality significantly when the data had been contaminated by seawater motion-induced noise. However, installing a current meter on each OBEM would increase the cost for marine MT data acquisition. One current meter data used for multiple nearby sites may be more feasible. The effective distance of local and distal reference sites may be decided by the complexity of currents in a particular study area.
In the future, we will combine the adaptive correlation noise-cancellation filter with robust processing procedures. In addition, adaptive correlation noise-cancellation filters may have the potential to separate noise from raw time series for marine MT data processing, where the MT signal is often contaminated by strong noise caused by induced EM noise due to seawater motions.
Thanks for the extensive support offered by the respective captains, the crew of the ship, and the marine technicians aboard the R/V “Haiyangsihao”. Professor Xia gave some suggestions on adaptive canceling filters. Professor Wu provided valuable comments regarding data processing. We particularly thank Drs. Zhang and Tu for their helpful suggestions and support on the current meter. The SSMT2000 processing software was provided by Phoenix Geophysics Ltd. Comments provided by the reviewers helped to improve the clarity of the manuscript.
KC developed the required hardware and software, led the overall design, and performed the offshore experiments. MD was the noise reduction technology consultant. QZ and XL assisted with the marine EM survey. JJ participated in the MT data processing. All authors read and approved the final manuscript.
The general study and experiment were funded by the National Key R&D Program of China (2016YFC0303100, 2017YFF0105700). The development of the OBEM was supported by the National Science Foundation of China (41804071, 61531001).
The authors declare that they have no competing interests.
- Egbert GD (1997) Robust multiple-station magnetotelluric data processing. Geophys J Int 130(2):475–496. https://doi.org/10.1111/j.1365-246X.1997.tb05663.x CrossRefGoogle Scholar
- Haykin SS (2002) Adaptive filter theory, 4th edn. Prentice Hall, New JerseyGoogle Scholar
- JianEn J, QingXian Z, Ming D, XianHu L, Kai C, Meng W, GuangHong T (2018) A study on natural gas hydrates and their forming model using marine controlled-source electromagnetic survey in the Qiongdongnan Basin. Chin J Geophys 61(11):4677–4689. https://doi.org/10.6038/cjg2018L0660 (in Chinese) CrossRefGoogle Scholar
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.