Can tectonic processes be recovered from new gravity satellite data?

The goal of this study is to investigate whether temporal variations of the gravity field caused by tectonic processes (hereafter geodynamic signals) can be recognized in satellite gravity data, including the currently operating GRACE satellites and future systems. We restricted our study to subduction zones, calculating possible gravity field variations associated with elastic stress accumulation in locked areas and with stress release by earthquakes. We used fault-plane solutions for the Alaska-1964, Chile-1960 and Hokkaido-2003 earthquakes, and GPS-based strain accumulation data in locked areas of the Alaska subduction zone. Vertical displacements of the Earth's surface were calculated using a model of a rectangular fault in an elastic half-space. We developed and applied a statistical signal-recognition technique to identify signals caused by displacements of unknown magnitude on fault planes of given position and dimension. Our goal is thus to detect and analyse in satellite gravity data a signal constrained by ground geophysical and geodetic data. We assumed different levels of data accuracy, ranging from the first GRACE model GGM-01S to two orders of magnitude lower, corresponding to the target accuracy for GRACE and GOCE data. We concluded that using the developed technique, gravity field variations similar to those caused by Alaska-1964 earthquake should be recognizable in GRACE data at the accuracy level of the model GGM-01S. If forthcoming satellite gravity models have an accuracy one order of magnitude better, then the signal recognition probability will be about 99% using our approach. The required accuracy is close to the errors due to imperfect corrections for atmospheric effects. For the Chile-1960 earthquake we considered different fault-plane models and found that one can distinguish between these models with a probability approaching 70% at present level of GRACE accuracy. Increasing the data accuracy by one order of magnitude makes this probability very high. Because the gravity signal from the Hokkaido-2003 earthquake was rather weak, it would only be recognized if the data accuracy increases by two orders of magnitude thus approaching the target GRACE accuracy. If forthcoming gravity models are one order of magnitude more accurate compared to the first GRACE model then 5 years of data will allow recognition of time varying gravity signal associated with locked areas of the Alaska subduction zone. Our method may be easily applied to ot