Quantitative Analysis of Early Seismograph Recordings - Historical Seismology: Interdisciplinary Studies of Past and Recent Earthquakes - page 383

distortion in the flat part of the instrument response. Teleseismic body wave arrivals
for large events can be picked rather accurately, overcoming the notorious timing
inaccuracies of early seismographs, and permitting a consistency check between the
assumed component polarities and the direction of the incident ray. Consequently,
many studies focused on modelling or inverting teleseismic body waves to constrain
the orientation of faulting, scalar seismic moment, and source time histories (e.g.
Singh et al. 1984, Stein et al. 1988, Doser 1992, Doser et al. 1999, Alvarado and
Beck 2006). A combination of teleseismic data and either geodetic leveling data or
observed surface faulting was used to obtain finite slip distributions for several large
earthquakes (e.g. the 1906 San Francisco earthquake by Wald et al. 1993; the 1923
Kanto earthquake by Wald and Somerville 1995; the 1905 Mongolian earthquakes
at Tsetserleg and Bolnay by Schlupp and Cisternas 2007, and the 1944 Tonankai
earthquake by Ichinose et al. 2003). Many of those results are highly relevant for
science and society, such as the fault dimensions of early XX century subduction
earthquakes in Japan or Mexico.
When historical teleseismic waveforms are on scale and well resolved, the long
period component may be input into routine schemes for global source parameter
retrieval, as shown by Okal and Reymond (2003), who use long period (100-200 s)
mantle Love and Rayleigh waves to invert for the seismic moment tensor of the
1938, Mw 8.5 Banda Sea earthquake from the azimuthal pattern of spectral am-
plitudes, or Huang et al. (1998), who systematically applied the Harvard centroid
moment tensor technique (Dziewonski et al. 1981) to a global set of 35 pre-WWSSN
deep earthquakes (depth 330-670 km, Mw 6.3-7.9), benefiting from the comparably
even resolution of moment tensor elements and simple excitation kernels for deep
focus events.
For local and regional distance recordings, the small range, bandwidth, and sta-
tion sparseness may introduce more severe complications in the analysis of histor-
ical waveforms. Strong ground motion is off scale, except for few purpose-built
low gain instruments that may have recorded near-regional P waves of large earth-
quakes (Kikuchi et al. 2003, Ichinose et al. 2003). The limited frequency bandwidth
affects the reliability of restituted long period ground motion, and source analysis
must be often based on shorter period components that do not account for the entire
source process and, furthermore, are severely influenced by small-scale heterogene-
ity affecting regional wave propagation. Two strategies to stabilize source retrieval
suggest the use of as unaltered historical recordings as possible, that is substitut-
ing the deconvolution of the instrument response from the target waveforms by the
corresponding convolutions (see previous section and Rivera et al. 2002, Kikuchi
et al. 2003, Ichinose et al. 2003), or the direct processing of individual un-rotated
horizontal component seismograms instead of the usual radial and transverse wave-
forms (Stich et al. 2005), to avoid distortions introduced by rotation of pairs of
horizontal historical seismograms with incorrect alignment, uneven drum speed, and
imprecise instrumental correction.
Comparably stable approaches for analysing regional historical data include
the retrieval of scalar seismic moment from displacement amplitude spectra (e.g.
Teves-Costa et al. 1999, Pino et al. 2000), or seismic moment rate from empirical