Analysis of Noise in Ranging Data From the GRACE Follow-On Satellite Mission

Just like its predecessor GRACE, GRACE Follow-On (GFO) satellite mission was mostly designed to monitor mass transport in the Earth's system. The mission consists of two co-orbiting satellites, which continuously collect inter-satellite ranging data. To that end, the satellites are equipped with two ranging systems: (i) a K-band Ranging (KBR) system, which is similar to that onboard GRACE and ensures a mkm-level precision; and (ii) an experimental Laser Ranging Interferometry (LRI) system, which allows the ranging precision to be improved by about two orders of magnitude, as compared to KBR. The presence of two alternative ranging systems facilitates a more in-depth analysis of noise in ranging data, as compared to GRACE.

The analysis we conducted consists of two steps: (i) identification and elimination of outliers in the data; (ii) spectral analysis of the KBR-LRI differences, which are interpreted as noise in KBR data due to a much lower precision of this observation technique, as compared to LRI. The following findings have been made.

LRI data suffer from thousands of outliers, which are visible as peaks in terms of range-rates. Most probably, they can be explained by cycle slips. Remarkably, 98% of the outliers occurred in a single day of March 19, 2019 (i.e., directly after re-starting the LRI system that followed the shut-down period from Feb. 7 to Mar. 17). Outliers have been revealed in the KBR data as well: KBR-based range-rates show 362 peaks, all of which occurred within a single two-day interval (Oct. 16-17, 2019). It is also found that 2.6% of the differences between KBR and LRI data stay at an abnormally high level. Most of the abnormal differences occur within only a few one-day time intervals. In these intervals, the observed differences show a pattern that resembles the total signal acquired by KBR and LRI systems. Presumably, these differences may be partly explained by timing errors.

Having eliminated all the abnormal data, we compute the Power Spectral Density (PSD) of the differences between the KBR-based and LRI-based range-rates (which is likely close to noise PSD of KBR-based range-rates). Among others, we reveal an increased noise level at the frequencies of 1, 2, and 3 cpr. This feature may partly explain an increased noise level in spherical harmonic coefficients C20 and C30, which are estimated from GFO data.