Abstract:

Typically, GRACE error reduction has been achieved via spatial filtering to average out errors. These filters are based on the assumption of randomly distributed errors. Errors in GRACE maps are often described by "striping". The implication is that the errors are not randomly distributed in space for a given month, but are correlated. We show that there exists a component of the GRACE errors that is correlated, that these correlated errors are responsible for the "striping" in the GRACE gravity fields, and that these errors can be, to some extent, isolated and removed from the data. As with other types of data filtering, there is a tradeoff between reduction of the inherent measurement error and potential signal corruption, and we assess the negative impact of this type of filter also. This analysis provides a clearer picture of the resolution of the GRACE gravity fields.