Satellite gravity gradient grids for geophysics

The Gravity field and steady-state Ocean Circulation Explorer (GOCE) satellite aimed at determining the Earth’s mean gravity field. GOCE delivered gravity gradients containing directional information, which are complicated to use because of their error characteristics and because they are given in a rotating instrument frame indirectly related to the Earth. We compute gravity gradients in grids at 225 km and 255 km altitude above the reference ellipsoid corresponding to the GOCE nominal and lower orbit phases respectively, and find that the grids may contain additional high-frequency content compared with GOCE-based global models. We discuss the gradient sensitivity for crustal depth slices using a 3D lithospheric model of the North-East Atlantic region, which shows that the depth sensitivity differs from gradient to gradient. In addition, the relative signal power for the individual gradient component changes comparing the 225 km and 255 km grids, implying that using all components at different heights reduces parameter uncertainties in geophysical modelling. Furthermore, since gravity gradients contain complementary information to gravity, we foresee the use of the grids in a wide range of applications from lithospheric modelling to studies on dynamic topography, and glacial isostatic adjustment, to bedrock geometry determination under ice sheets.


Validation of regional gravity field recovery using spherical basis functions
The tesseroid grids were validated with an alternative method: We applied a regional modeling approach using spherical basis functions (SBF) [2] based on series expansions in terms of Legendre polynomials up to spherical harmonic degree L = 560. The related unknown scaling coefficients were estimated by relative weighting of the six GOCE/GRACE gravity gradients, composing each a separate observation group, using variance component estimation. The less accurate gradients V XY and V YZ are included here and get low relative weights compared with the accurate gradients. In advance, the GOCE/GRACE observations were reduced by the reference model GOCO03S up to degree L = 60. Rotating the observation equations for the analysis process into the GRF enables the use of the original GOCE/GRACE gravity gradients [1]. We estimated the coefficients up to degree L = 560, and multiplied them in the synthesis with Blackman SBFs, where the signal is represented up to degree L = 511, but smoothed between degree L = 256 and 511 to avoid erroneous edge effects. We located them finally in the LNOF on a regular grid at mean orbit height of 225 km.
Differences SBF (up to degree L = 511) -GOCO03s (L = 250) are shown in Figure S1 for V NN , V WW , and V UU for a region east of the Philippines. The differences are a few mE at 225 km height, comparable with the case for the tesseroid grids, and we see that the SBF grids add the same spatial detail to GOCO03s. The differences between tesseroids and SBFs are small and generally below 1 mE, see the right column in Figure S1. We therefore conclude that the SBF and tesseroid grids are in good agreement with each other.

Spherical and ellipsoidal LNOF
Grids of gravity gradients are provided in the Local North-Oriented Frame (LNOF) above the ellipsoid. The gradient grids are given on a homothetic ellipsoid that has the same eccentricity as the WGS84 ellipsoid and a semi-major axis aH = aWGS84 + H, where aWGS84 = 6378.137 km and H is 225 km or 255 km. It should be noted that different LNOF definitions are possible, which depend on whether the vertical axis is taken along the direction of the ellipsoidal normal or along the direction of the spherical normal. The GOCE gravity gradient grids are given in the spherical LNOF but can easily be transformed to the ellipsoidal LNOF. The exact definitions are given below as well as an assessment of the differences. How the transformation can be done is shown as well.

LNOF -Local North Oriented Frame (spherical)
The Local North Oriented Frame (LNOF) is a right-handed North-West-Up frame with the X-axis pointing North, the Y-axis pointing West and the Z-axis Up. which thus slightly differs from the spherical definition because geocentric and geographic latitude are slightly different.

Transformation from spherical to ellipsoidal LNOF
The relation between geocentric latitude and geographic latitude is with the inverse relation
In each grid point one can then rotate the gradient tensor from the spherical LNOF to the ellipsoidal LNOF using ).
Given the geographic latitude of the grid points, the geocentric latitude can be computed and the gradient tensor can be rotated.

Gradient differences ellipsoidal normal and radial direction
When computing gravity gradients in a regular ellipsoidal grid, the direction of the ellipsoidal normal and the direction of the spherical normal are not equal. To assess the these differences, we computed the spherical coordinates of a grid with 0.2° spacing at 225 km above the ellipsoid. The North-East Atlantic region was used for the test computations. The figure below shows the difference between the two latitudes.
Next we computed gravity gradients at 225 km above the ellipsoid for the NEA region using GOCO03s with respect to GRS80. The lower left panel in the figure below shows the vertical gravity gradient, which has maximum amplitude of about 0.6 E. The other panels show the difference between the gradients in the LNOF (ellipsoidal) and LNOF (spherical). Because the rotation is around the y-axis (east-west) the T YY differences are zero. Largest differences occur for T ZZ and these are 3 mE or less. Thus, although there may be correlation between these differences and the signal, the maximum amplitude is small, but may be above the accuracy of the gradient grids. 0

Zooming in on gravity gradient grids around the world @ 225 km altitude
The gravity gradient signal at 225 km altitude is shown for longitude-latitude blocks of 60° × 30° for longitudes -180° ≤ λ ≤ 180° and latitudes -75° ≤ φ ≤ 75°. The North and South Pole are shown separately. The gradient signal with respect to the WGS84 reference ellipsoid is shown in the order as shown in the table below, starting in the south-west. The colour scale in each region and for each gradient is adapted to the min/max values in that region. Topography and bathymetry contour lines are shown every 1000 m and were derived from ETOPO1 [47]. The Lambert projection is used for the patches centred at φ = ±60°, a stereographic projection is used for the North and South Pole, and all other patches use the Mercator projection.

Gravity gradient grids @ 225 km altitude with topographic reduction
The gravity gradient signal after topographic mass reduction at 225 km altitude is shown for longitude-latitude blocks of 60° × 30° for longitudes -180° ≤ λ ≤ 180° and latitudes -75° ≤ φ ≤ 75°. The North and South Pole are shown separately. The topographic mass reduction has been done using a spherical harmonic model for rock, water and ice density [44], where we used a maximum spherical harmonic degree of L = 360 to be consistent with the gravity gradient resolution. This global correction enhances the signal of the internal structure of the Earth and is the equivalent to a Bouguer gravity anomaly. The gradient signal with respect to the WGS84 reference ellipsoid is shown in the order as shown in the table below, starting in the south-west. The colour scale in each region and for each gradient is adapted to the min/max values in that region. Topography and bathymetry contour lines are shown every 1000 m and were derived from ETOPO1 [47]. The Lambert projection is used for the patches centred at φ = ±60°, a stereographic projection is used for the North and South Pole, and all other patches use the Mercator projection.