International Symposium on Applied Geoinformatics (ISAG2019), İstanbul, Türkiye, 7 - 09 Kasım 2019, ss.109
GRACE (Gravity Recovery and Climate Experiment) satellite system, operated by NASA and DLR, has
provided high-resolution information (up to 150 km) about the static and dynamic structure of the Earth's gravity field
between 2002 and 2016. The GRACE mission, especially known for its sensitivity to the hydrological events and their
temporal changes, has been used in a variety of geodetic and geodynamic studies, such as monitoring variations in the total
water storage on the land, the ocean bottom pressure etc. One of the most important products of the GRACE is the socalled Equivalent Water Thickness (EWT). This data is used to determine the total mass storage. The EWT data can be
derived from monthly spherical harmonic coefficients (L2 data) or as the final solutions (L3 data: filtered solutions) on the
1ox1o or 3ox3o sized physical surface parts provided by the process centres such as CSR (Centre for Space Research), JPL
(Jet Propulsion Laboratory), NASA-GSFC (Goddard Space Flight Centre). The most recent L3 data is the Mascon (Mass
concentration) solutions which are provided region by region, basin by basin or Mascon by Mascon. For analysing the
EWT time-series, commonly, a harmonic regression model including overall trend, annual and semi-annual periods is used.
Generally, it is assumed that only white noise exists in the data. However, the EWT time-series may contain different
coloured noises, such as flicker noise, random walk noise, etc., as in every kind of geodetic time-series. Considering these
noises in the harmonic regression model is essential because neglecting them in the analysis misleads the statistical
decisions about the deduced parameters. In this study, therefore, we investigate the noise characteristic of the basin-bybasin GSFC-Mascon EWT data between 2003 and 2016. For this aim, the log-log Power Spectral Density (PSD) plot of
the EWT time-series associated with each basin of the continents except for the Polar Regions is examined. It is known
that the slope of the log-log PSD plot gives information about the spectral index of the noise in the time-series: the spectral
index is 0, -1 and -2 for white, flicker and random walk noise, respectively. The estimated slope of each log-log PSD plot
of the EWT time-series used in this study is significantly different from zero. The average value of the spectral indexes for
the basins are about -1.3±0.3. In other words, the EWT data contains coloured noises (dominantly flicker noise) rather than
only-white noise. These estimated spectral indexes and their confidence regions are given in the spatial domain to discuss
the noise characteristics of the GRACE-Mascon EWT solutions.