A new assessment of temporal correlations in GRACE TWS time series

Güneş Ö., Klos A., Lenczuk A., Aydın C., Bogusz J.

International Union of Geodesy and Geophysics (IUGG) 2023 General Assembly, Berlin, Germany, 11 - 20 July 2023, pp.1

  • Publication Type: Conference Paper / Summary Text
  • City: Berlin
  • Country: Germany
  • Page Numbers: pp.1
  • Yıldız Technical University Affiliated: Yes


We focus on evaluating the time correlations present in the Total Water Storage (TWS) time series observed by the GRACE missions over the period 2002-2022. We subject the original TWS time series provided in the form of a mascon solution by the NASA, GSFC to analysis of temporal correlations using the maximum likelihood estimation. We assume a conventional deterministic model of trend plus annual and semi-annual periodicities. We find that in many areas around the world colored noise similar to random-walk noise dominates, the spatial pattern of which is inconsistent. In the next step, we change the conventional model to a new redefined deterministic model that, in addition to seasonal components, includes third-, fourth- or fifth-degree polynomials, depending on the location of the mascon to parameterize the nonlinearity of long-term changes. Once the definition of the deterministic model has been changed, we re-run the analysis. We show that the spatial pattern of temporal dependencies is much more consistent than with the conventional model. Many regions are characterized by a change in stochastic character from random-walk noise to one closer to white noise. The TWS time series for these regions is characterized by white noise, meaning that there are no short-period signals associated with the hydrosphere, and the previously observed random-walk noise was only associated with long-term changes that carried over into the stochastic part. The random-walk noise observed for TWS time series after redefining the deterministic model has a high correlation with the occurrence of short-term events such as droughts and floods.