The model of GPS coordinate time series consists of two contributions: One is a model of "steady-state" relative plate motion on the Guerrero plate interface, in which (1) fault slip must be between zero and the relative plate motion rate (a necessary condition of steady-state on interseismic timescales); (2) dip-slip coupling and strike-slip coupling are separate parameters, and strike-slip coupling must be greater than or equal to dip-slip coupling (suggested by observations that strain at oblique subduction zones partitions into dip slip on the plate interface and strike slip in vertical faults further from the trench); and (3) the variation of coupling is smooth (slip on neighboring discrete patches of the plate interface differ by less than half the relative plate motion rate). The plate interface was discretized with a fine mesh (20 km) within the main Guerrero network region and coarser mesh outside.
The second contribution to the model comes from the transients. Here, transient slip in 1996, 1998, 1999, 2000, 2001 2002 and 2003 is modeled as uniform slip in a rectangular region, with time dependence for the transient slip displacement parameterized as a hyperbolic tangent function. The models of the 1998 and later transient events have eight parameters: (1) the mid-time of transient slip; (2) the timescale parameter tau; (3) the width of the slip event; (4) the length; (5) and (6) the center location of the slip event, and (7) and (8) the total anomalous dip-slip and strike-slip during the event. The latter two can be estimated by linear least-squares inversion, but the other six are found via grid-search. The 1995 transient has only six parameters because there were no continuous sites operating at that time and thus the two time parameters are fixed a priori. The 2002 event was modeled by two dislocations with eight parameters each, because the large quantity of continuous data are much better fit by two "sub-events" than by a single dislocation. I also solve for two slip parameters of the 1995 Copala earthquake.
A more detailed description of modeling and inversion as it was applied to the 1995 and 1998 transients only is given in Larson et al. [2004]. A cursory description of the single-patch modeling of all transients presented here (and also of the discretized model type) is given in Lowry et al. [2006].