Historical data concerning more than 130 LPG rail and road transport accide
nts were critically examined, identifying the scenarios, following their ev
olution into the final accidental events, and determining their theoretical
probabilities of occurrence. In principle, rail accidents are rather hazar
dous, the most probable scenario being a major release followed by an UVCE.
However, in order to discriminate among road and rail transport, the relev
ant number of trips and accident rates should be taken into account. In fac
t, the application, concerning the transport of 5700 ton/year of LPG in Ita
ly along one rail and two different road itineraries, showed that the risk
for rail transport was more than one order of magnitude lower than that fbr
those on the roads. The population density along the route and the acciden
t rate, rather than the length of the route, appear the most important fact
ors for discriminating between different itineraries. (C) 1999 Elsevier Sci
ence Ltd. All rights reserved.