The main result of this study is a series of tomographic models
obtained for data in a moving time window. We found that the optimal
length of the window is one year, which utilises a sufficient amount of
data for good coverage, but is short enough to resolve temporal variations.
To explore the effect of temporal variations, we performed 40
inversions in overlapping one-year windows with three-month steps.
In this paper, we focus our results and discussion on the calendar
years 2001 through 2008. Before 2001, several stations did not yet
exist resulting in poor ray coverage