More information about ground motion can be extracted from seismoscope records than a single point on a response spectrum. To demonstrate this, the relation between seismoscope response and Wood-Anderson instrument output and peak horizontal ground velocity has been studied by simulating the various responses for a range of distances and magnitudes. The simulations show that the relation used by Jennings and Kanamori (1979) to convert from peak seismoscope readings to the peak response of a Wood-Anderson instrument has a distance- and magnitude-dependent systematic error. The error is negligible, however, for modern seismoscopes at distances of a few tens of kilometers. At several hundred kilometers, the relation underestimates the Wood-Anderson response by as much as a factor of two. The spread in Jennings and Kanamori's estimate of ML for the 1906 San Francisco earthquake, recorded on seismoscopes having relatively low natural frequencies (0.26 and 0.5 Hz), is reduced by the results in this paper—the upper value, from a seismoscope in Carson City, Nevada, at 290 km from the fault, going from ML = 7.2 to ML = 7.0 and the lower value, from Yountville, California (R ≈ 60 km), going from about 6.3 to 6.4. About 0.3 units of the remaining spread may be due to local geologic site conditions. If the 0.3 units is distributed equally between the Yountville and Carson City recordings, the estimates of ML for the San Francisco earthquake then become 6.5 and 6.8, somewhat lower than Jennings and Kanamori's final estimates of to 7. Although the error in using the relation of Jennings and Kanamori to estimate Wood-Anderson response was at most a factor of 1.6 for the 1906 earthquake, the error can be substantially larger for smaller earthquakes recorded on similar low frequency seismoscopes.
The relation between Wood-Anderson and seismoscope response used by Jennings and Kanamori can be combined with an empirical relation between peak horizontal velocity and Wood-Anderson response to predict peak velocity from seismocope recordings. The simulations show that this relation (vmax = 8.1Awa, where vmax is the peak horizontal velocity in centimeters/second and Awa is one-half the range of the Wood-Anderson motion in meters) forms a lower bound for estimates of peak velocity from seismoscope recordings. The relation is good for stations within about 100 km of earthquakes with moment magnitudes of about 4.5 to 6.5, and it underestimates peak velocity by factors up to 2 or 3 for larger earthquakes at distances within 100 km. An application of the simulation method to the 1976 Guatemala earthquake (moment magnitude = 7.6) results in 37 cm/sec as a lower bound to vmax, with 66 cm/sec as a more likely value, from the seismocope recording in Guatemala City (approximately 25 km from the Motagua fault).