With the launch of the Gravity Recovery and Climate Experiment (GRACE) mission in 2002 and continued with GRACE Follow-on (GRACE-FO) since 2018, it is nowadays possible to monitor important mass variations in the Earth system. Nevertheless, validating these observations is a challenging task due to the lack of alternative methods to obtain directly comparable in-situ measurements. The most appropriate approach for this endeavor consists of comparing the GRACE derived Total Water Storage (TWS) residuals against Superconducting Gravimeter (SG) residuals, which provide long term stability.
The in-situ data used for this project are the gravity residuals obtained after removing the effects of solid Earth tides and ocean tidal loading, atmospheric loading, instrumental drift, polar motion and length‐of‐day induced gravity changes, from nine SG stations between January 2010 and March 2017. Such residuals were then compared with GRACE retrieved TWS residuals obtained from the Gravity Information System (GravIS) portal (gravis.gfz-potsdam.de).
In this project, three decomposition methods were used for the comparisons: Principal Component Analysis (PCA), Spatiotemporal Independent Component Analysis (stICA) and Multivariate Singular Spectral Analysis (MSSA). The main aim was to assess the impact of the GRACE data corrections applied by GravIS to the coefficient C20, the coefficients of degree/order one, and the Glacial Isostatic Adjustment (GIA) effect. Moreover, the Gaussian, DDK and VDK filtering techniques were evaluated as well.
The tested methods proved to cope with the residual hydrological effects on SG measurements up to an extend that allows an objective evaluation of the data. The results obtained from this analysis indicate that the most optimal solution is achieved by correcting the C20 and degree/order 1 coefficients. The most effective filters are DDK1, VDK2 and Gaussian with a 500 km bandwidth, in that order. Furthermore, the GIA correction demonstrates to be relevant for northern locations like Onsala.
Concerning the decomposition methods, MSSA demonstrates to be a powerful tool, synthesizing the most important common trends among the in-situ measurements of different stations, and displaying the local differences of the signals. The common signals extracted from PCA represent a good overview of the trends from the data but is not detailed at the individual locations. Finally, the stICA decomposition is not able to extract these common signals when the input data is significantly different across the individual variables for SG data. This is explained by the Blind Source Separation (BSS) nature of the methodology, which intends to identify differences among the signals, and is not useful in this case where the signals are affected by the local hydrology.
The importance of this study lies in the versatility that the successfully tested methods show for the purpose of GRACE data comparison. Furthermore, the methodology applied in this project can be extended to analyze the current GRACE-FO mission as well other gravimetric satellite missions in the future.
