Analysis of GRACE time series noise characteristics

Abstract

Total water storage variations have been monitored using the GRACE satellite system since its launch in 2002. In the analysis of total water storage time series, which are defined in terms of equivalent water thickness, the harmonic regression method, a functional model incorporating trend and seasonal signals, is used. In this model, the observations are presumed to be equally weighted and uncorrelated. In other words, only white noise is included in the stochastic model. However, a spectrum analysis of the time series reveals temporal correlations, which result in colored noise. In this study, the power spectral of these colored noise characteristics is examined using the Goddard Space Flight Center mascon dataset with monthly sampling from 2002 to 2022. The total water storage change time series shows a colored noise pattern with a negative spectral index. In global analyses, it has been shown that the mascon (mass concentration) time series representing terrestrial regions has a characteristic of colored noise defined in the fractional Brownian power spectrum, which is similar to flicker noise. Furthermore, the DDK (Decorrelation Filter) dataset, which is derived from monthly spherical harmonic coefficients, is also used, and similar results are obtained.