We demonstrate that the approximate source kinematics of the San Fernando,
1971 earthquake can be back-predicted by analysing its macroseismic intensi
ty data set (felt reports) objectively and quantitatively. This is done by
inverting either the data set of the intensity values observed in all sites
, or the intensities tessellated with the Voronoi polygons technique. It is
shown that the kinematic characteristics found following our method (epice
ntral coordinates, source depth, seismic moment, rupture length, Mach numbe
r, fault plane solution) match those determined by other authors, via instr
umental measurements, rather well. The prerequisite for obtaining these res
ults is that local amplification must not affect groups of neighboring site
s. It was possible to invert the U.S.G.S. "felt reports'' for the source be
cause this data set is sufficiently uncontaminated by local site responses,
and retains relevant regional traces of source effects. Isoseismal maps ca
nnot be safely used for this task, because qualitative drawing criteria giv
e subjective results. Isoseismals, based on incomplete space frequency samp
lings, give rise to spurious effects, whereas the Voronoi polygons produce
easy-to-grasp, quantitative and objective, representations of macroseismic
intensity data. The tests performed, up to now on a series of earthquakes,
suggest that the combined use of tessellation and of our KF model is promis
ing mostly for inverting intensities of preinstrumental earthquakes.