Signal analysis of mass variations from temporal gravity satellite solutions

The monitoring of the temporal variation of the Earth's gravity field with low earth orbit satellites has a history of 20 years. The GRACE satellite system, which is one of these satellites, has provided important information especially in the studies which are sensitive to hydrological changes such as monitoring the mass change in ground water resources, determining of trends in glacial mass, thanks to the data set which is published in monthly periods between 2002-2016. In this study, GSFC Mascon (2003-2016) and DDK (2002-2016) filtered monthly equivalent water thickness solutions were used. GSFC Mascon solutions were obtained from the CCAR data center on a global basin basis and as individual Mascons in regional terms. DDK filtered solutions are provided from ICGEM data center. The trends of the hydrological changes in global and regional scale were studied under the harmonic regression model. This model, which has trend, annual and semi-annual periodic signal effects, was considered as the standard functional model in this study. In addition, in the polar regions (Greenland and Antarctica basins) the linear trend model expanded by the acceleration parameter and examined how different functional approaches affect the trend values, standart deviations and the significance of other parameters in EWT analysis.Furthermore, different models of the stochastic part of the time series were also used. Because, there are temporal correlations in the EWT time series, that is, there are also colored noise types besides white noise in the data set. In this context, the data set was also analyzed in spectral manner with PSD-frequency graphics. In the time series, it was decided that there was flicker noise and then the amplitudes of the noise combination (white + flicker noise) were obtained by the estimation of variance components. Thus, the data set was weighted properly. As a result, the EWT values in a time-series were studied under two different stochastic approaches: the EWT values are 1) equally weighted and non-correlated, and 2) different weighted and correlated. The effects of color noise on global and regional EWT analysis were examined. According to the results, it was seen that the temporal correlations in the EWT time series significantly affected the annual trend estimations.