Epicentral parameter estimation from intensity data of uncertain accuracy

In parts of the world where seismic productivity is sparse we must rely on the historical record to provide estimates of future seismic hazard. Paleoseismic data can constrain location and magnitude for those earthquakes whose ruptures breach the surface, but for most earthquakes prior to 1900 almost all that is known of them comes from felt intensity data. Empirical approaches have hitherto been used to estimate both location and magnitude from these data, but isoseismal contouring methods, with or without computer aided interpolation are often biased by the user. Numerical methods to avoid isoseismal estimation appear to circumvent this bias but where we have been able to test such methods with sparse data we have found the location and magnitude can be critically dependent on a few key observations. The inclusion or rejection of these critical data again introduces a user bias. The most acceptable approach is to provide an uncertainty to each intensity observation, and to include this uncertainty in the calculations of probable location and magnitude. Examples, of this approach are provided for the Allah Bund 1819 and Kashmir 1555, 1885, and 2005 earthquakes in India.