3.1. Pre-2005 uplifts
The vertical component of the 3D coordinate time series of CHON
and BANH are shown in Fig. 4, along with the estimated linear height
trend before and after the earthquake. Before the earthquake we do
not have continuous GPS near both stations. For this we are forced to
use the available GPS campaign data. As the measurements do not
exactly span the same period of the year, they could be biased from a
seasonal effect (likely of atmospheric origin). The seasonal cycles
were estimated from post-earthquake data of nearby CGPSs assuming
that the behaviour remained the same. A robust regression fit with
IRLS, which was modified to model the vertical land motion as a
combination of a linear trend, an annual and a semi-annual cycle, was
applied to simultaneously solve these parameters for the time series
(early 2005–end of 2009) of the weekly GPS solutions of RYNG
(established in early 2005, 39.4 km from CHON). Plots of the weekly
solutions and annual cycles can be found in the electronic supplement.
The same procedure was applied to the time series of CPNT
CGPS (beginning of 2005–end of 2009, 33.1 km from BANH), The
amplitudes of the annual cycle are 12.0 mm and 11.0 mm for RYNG
and CPNT respectively, with their positive peaks at 140 and 147 days
after January 1st. The height coordinates of CHON and BANH were
then compensated for these annual cycles