Temporal gravity variations in GOCE release 6 gravitational gradients

As opposed to the level 1B release 5 GOCE gravitational gradient data, the newly reprocessed release 6 gradients provide reduced noise amplitudes in the low frequency-range, leading to reduced noise amplitudes of the derived gravity field models at large spatial scales, where temporal variations of the Earth’s gravity field have their highest amplitudes. This is the motivation to test the release 6 gradients for their ability to resolve temporal gravity variations.

For the gravity field processing, we apply a conventional spherical harmonics approach using the time-wise (TIM) processing method as well as a mass concentration (mascon) approach using point masses as base elements, which are grouped to land or ocean mascons by taking into account the coastlines.

By means of a closed-loop simulation study, we find that the colored instrument noise of the GOCE gravitational gradiometer introduces noise amplitudes into the derived gravity field models that lie above the amplitude of the gravity trend signal accumulated over 5 years. This indicates that detecting gravity variations taking place during the four-year GOCE data period from GOCE gradients only is challenging.

Using real GOCE data, we test bimonthly gradiometry-only gravity field models computed by both the spherical harmonic and the mascon approach for gravity signals that are resolved by GRACE data, being the temporal signals due to the ice mass trends in Greenland and Antarctica and the 2011 earthquake in Japan. Besides, corresponding GRACE/GOCE combination models are used to test whether the incorporation of GOCE data increases the resolution of temporal gravity signals.

We found that high-amplitude long-wavelength noise prevented the detection of temporal gravity variations among the bimonthly GOCE-only models. Using the SH approach, it was possible to detect the mean trend signal contained in the data by averaging multiple bimonthly models and considering their difference to a reference model. Using the mascon approach, trend signals contained in GOCE data could be recovered by including a GRACE model truncated to d/o 45 in a GRACE/GOCE combination model and thus let the GOCE data determine the short-scale signal structures instead of GRACE.

Finally, compared to the temporal gravity signal as resolved by GRACE data, no significant benefit of using or incorporating GOCE gravitational gradient data was found. The reason are the still rather high noise amplitudes in the derived models at large spatial scales, where the considered signal is strongest.

In order to detect temporal gravity variations in satellite gravitational gradiometry data, the measurement noise amplitudes in the low-frequency range would need to be reduced.