|Paper title||Geomagnetic Virtual Observatories for investigating sub-annual core field variation|
|Form of presentation||Poster|
The Earth's magnetic field changes continuously both spatially and temporally. A measurement of the magnetic field at or above the Earth's surface is the summation of numerous different sources, each with a different spatial and temporal behaviour. On short time scales of seconds to months, the changes are driven primarily by the interaction of the ionosphere and magnetosphere with the solar wind. Seasonal changes are also influenced by the variation of the tilt of the magnetic field with the respect to the ecliptic plane. On longer timescales of years to centuries, changes of the core field (known as secular variation, SV) alter the morphology of the observed field at the surface.
With the plethora of Swarm satellite data, it is now possible to examine field sources in detail on a global basis. However, in contrast to a ground observatory where time series can be produced at a fixed location allowing the time change of the field to be deduced precisely, the orbital velocity of a satellite (at ~8km/s at 500 km altitude) makes source separation more difficult as measurements are a combination of both spatial and temporal variations of the field. The solution to this is often achieved by using a small subset of the data and modelling the expected geophysical extent of each source in space or time, or both. For example, main field models provide a large spatial scale representation with a smoothed time dependence typically fitted to six-monthly splines. However, such modelling approaches do not capture the more rapid variations of the core field, making it more difficult to robustly detect features such as geomagnetic jerks in satellite data compared to ground observatory data. Such rapid processes are believed to hold vital new information regarding the behaviour of the outer core.
Geomagnetic Virtual Observatories (GVOs) are a method for processing magnetic satellite data in order to simulate the observed behaviour of the geomagnetic field at a static location. As low-Earth orbit satellites move very quickly but have an infrequent re-visit time to the same location, a trade off must be made between spatial and temporal limits, typically between one month and four months with a radius of influence of 700 km chosen for the Swarm mission.
We build a global network of geomagnetic main field time series derived from magnetic field measurements collected by satellites, with GVOs placed at 300 approximately equally spaced locations, at the mean satellite altitude. GVO time series are derived by fitting local Cartesian potential field models to along-track and east-west sums and differences of data collected within a radius of 700 km of each grid point, over a given time period. For the Swarm mission, two Level 2 data products are now available: (a) time series of `Observed Field' GVOs, where all observed sources contribute to the estimated values, without any data selection or correction, and (b) time series of `Core Field' GVOs, where additional data selection and external field model corrections are applied.
These products are derived at one- and four-monthly sampling. We focus on the de-noising that is carried out on the one-monthly data set, the aim being to reduce the contamination due to magnetospheric and ionospheric signals, and local time (LT) sampling biases. It has been found that the secular variation of residuals of GVO time series data at a single location will be strongly correlated with its neighbours due to the influence of large-scale external sources and the effect of local time precession of the satellite. Using Principal Component Analysis (PCA) we can remove signals related to these noise sources to better resolve internal field variations on short timescales. This reduces the negative effects of using a time bin shorter than the local time precession rate of the orbit in terms of LT bias, improving the temporal and spatial resolution of more rapid SV. The PCA also allows the use of more data to build each GVO sample, accounting for external signals without the need for stringent data selection, a useful feature as there is a minimum number of data needed to stably resolve a local cubic potential in a given spatial and temporal GVO bin size. We describe the process developed as part of the ESA Swarm Level 2 GVO product, and also the application of this method to GVO series derived from observations of the Oersted, CHAMP and CryoSat-2 missions.
This method can be used on other magnetic missions or those with ESA platform magnetometers. Our denoised GVO data set covers November 2013 to 2021 for Swarm, and has been extended for Ørsted 1999 to 2005, for CHAMP 2000 to 2010, and for CryoSat-2 2010 to 2018.
In addition, the methodology can be used to model the improvements possible using additional satellite missions such as NanoMagSat. The availability of data from a wider range of local times along with more rapid repeat periods allows denser grids of GVO and higher cadences, for example, reducing from 4.2 months to three weeks. This would allow very rapid core signals to be identified in a more robust manner, broadening the extent to which we can probe the outer core while relying on Swarm as a backbone that ensures absolute accuracy over time.