A number of global and regional catalogues of historical tsunamis have been compiled in recent years and are summarized in the following table. These are of varying reliability, particularly for the period before the beginning of the 20th Century and the advent of networks of tide gauges that could record tsunamis themselves and of seismometers that could locate the commonest type of tsunamigenic events, that is to say earthquakes. Prior to this time the catalogues rely on contemporary accounts that are of varying quality and completeness. In some regions the catalogues extend back many centuries, most notably in Japan, China and the Mediterranean. In others, most notably New Zealand, Hawaii, and the sector of the Pacific rim from the Aleutians through to California, the duration of the catalogues is only 150 - 200 years. Furthermore the older parts of the catalogues (especially those for regions in which tsunamis and tsunamigenic earthquakes are relatively rare events) are often unreliable, confusing such phenomena as river floods, exceptional tides, and meteorological waves and storm surges with tsunamis. Even where a tsunami origin for an event is likely, the available information often permits a large degree of uncertainty as to the magnitude of the event. These uncertainties are typically reflected in a numerical scale (0 to 4, representing "not-a-tsunami" through to "certain tsunami", in the NGDC catalogue, for example). The development of the instrumental networks noted above and also the appearance of photography and telegraph and newspaper networks which allowed the reporting of tsunami events in greater detail mean that these problems largely (but not completely) disappeared around the beginning of the 20th Century.

Certain of the catalogues also contain separate sections for tsunami event records and tsunami runup records (note that the latter are much more numerous than the former, because an individual tsunami is likely to be recorded in many places) and these can be used to generate statistics on the frequency of events on the one hand and the approximate scales and distribution of damage on the other.

The fundamental problem with all of these catalogues, however, is that their limited length means that they will rarely, if ever, contain the most extreme event that is likely to affect a region. Their use is therefore likely to be particularly misleading in the production of Probable Maximum Loss scenarios. In the cases of the regions with short (150 - 200 year) historical catalogues this will be the case even if the pragmatic view is taken that attention should be restricted to 250-year periods (for example, for the purposes of assessing insurance company solvency). Perhaps the most notable case that has come to light in recent years is that of the Pacific North West of the United States of America, where the historical record contains only minor tsunami of local origin but geological and archaeological evidence indicates the occurrence at 300 - 700 year intervals of major subduction zone earthquakes and associated large-magnitude tsunami. Since the last such earthquake and tsunami has been identified as occurring in January 1700 (see Cascadia Tsunami ) a possible future recurrence of a comparable earthquake and tsunami should definitely be taken into account in PML scenarios for the Pacific Northwest. More fundamentally, the non-seismic tsunami sources, particularly volcano collapses and impacts, are statistically likely to be underrepresented in the historical catalogues, whilst submarine landslides and volcano collapses pose further problems because their frequency of occurrence may vary through time (see Future Directions in Tsunmai Research ).