Description
Time discretization
-
MTUQ uses a centroid origin time convention, in which t=0 corresponds to the mean source excitation (usually equivalent to the peak source excitation)
-
CAPUAF uses a rupture origin time convention, in which t=0 corresponds to the initial source excitation
-
MTUQ still works even if the time discretization varies from station to station, or if data and Green's functions have different time discretization. Certain code optimizations may be lost, however, making run times much longer. Also such cases are not yet well tested.
Time shifts
-
Systematic difference between CAP and MTUQ time shifts arise from the way CAP implements source-time function convolution. CAP's "conv" function results in systematic magnitude-dependent shifts between centroid origin times and arrival times.
-
The amount of the time shift is half the sum of the earthquake's rupture time and rise time, as given by relations in the CAP Perl wrapper, t_offset = (cap_rupture_time(Mw) + cap_rise_time(Mw))/2
-
Shifting a processed MTUQ trace relative to a processed CAP trace does not completely remove the effect of the idiosyncratic time shift; the two traces will still be tapered differently at the end points.
Distances
-
MTUQ uses obspy.geodetics to compute station-event distances, with results that differ slightly from CAPUAF.
-
Distances shown in CAPUAF figure labels are sometimes off by an integer compared with the actual values CAPUAF uses in computing synthetics.
-
CAPUAF rounds down when working with FK Green's functions, while MTUQ rounds to the nearest integer.
-
MTUQ also supports AxiSEM and SPECFEM3D Green's functions, in which case GLL interpolation is used rather than rounding.
Cross correlations
- MTUQ provides the option to pad synthetics prior to cross correlating them with data. CAPUAF does not, so cross correlation values become less accurate at large lag times.