Premium
Passive‐ocean radial basis function approach to improve temporal gravity recovery from GRACE observations
Author(s) -
Yang Fan,
Kusche Jürgen,
Forootan Ehsan,
Rietbroek Roelof
Publication year - 2017
Publication title -
journal of geophysical research: solid earth
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.983
H-Index - 232
eISSN - 2169-9356
pISSN - 2169-9313
DOI - 10.1002/2016jb013633
Subject(s) - radial basis function , spherical harmonics , regularization (linguistics) , gravitational field , basis (linear algebra) , basis function , geodesy , geology , computer science , mathematics , physics , mathematical analysis , classical mechanics , geometry , machine learning , artificial intelligence , artificial neural network
We present a state‐of‐the‐art approach of passive‐ocean modified radial basis functions (MRBFs) that improves the recovery of time‐variable gravity fields from Gravity Recovery and Climate Experiment (GRACE). As is well known, spherical harmonics (SHs), which are commonly used to recover gravity fields, are orthogonal basis functions with global coverage. However, the chosen SH truncation involves a global compromise between data coverage and obtainable resolution, and strong localized signals may not be fully captured. Radial basis functions (RBFs) provide another representation, which has been proposed in earlier works to be better suited to retrieve regional gravity signals. In this paper, we propose a MRBF approach by embedding the known coastal geometries in the RBF parameterization and imposing global mass conservation and equilibrium behavior of the oceans. Our hypothesis is that with this physically justified constraint, the GRACE‐derived gravity signals can be more realistically partitioned into the land and ocean contributions along the coastlines. We test this new technique to invert monthly gravity fields from GRACE level‐1b observations covering 2005–2010, for which the numerical results indicate that (1) MRBF‐based solutions reduce the number of parameters by approximately 10% and allow for more flexible regularization when compared to ordinary RBF solutions and (2) the MRBF‐derived mass flux is better confined along coastal areas. The latter is particularly tested in the southern Greenland, and our results indicate that the trend of mass loss from the MRBF solutions is approximately 11% larger than that from the SH solutions and approximately 4%–6% larger than that of RBF solutions.