Both the deterministic and probabilistic methods of seismic hazard analysis
serve necessary purposes. Probability is needed to obtain operating basis
earthquakes, to perform risk analyses, to prioritize projects, and for assi
gning recurrence estimates to deterministic earthquakes. The probability fo
r these purposes is used as a relativistic measure.
The problem with seismic probability is that it relies on the Gutenberg-Ric
hter b-line, which has: severe shortcomings. There are corrections that can
be applied, which attempt to remedy the problems. Data are introduced for
paleoseismic events, characteristic earthquakes, and slip-rate, or judgment
s are introduced from logic trees, multiple expert opinions, and de-aggrega
tion of E-values. Unfortunately, none are equal to the task. The probabilis
tic seismic hazard analyses remain fundamentally limited in their dependabi
lity. However, the deterministic method can provide evaluations that are at
a practical level for engineering. Engineering design must be done determi
nistically if one is to have seismic safety coupled with good engineering j
udgement.
The design for critical structures, those for which failure is intolerable,
such as dams, nuclear power plants, hazardous waste repositories, etc., mu
st be based on maximum credible earthquakes, obtained by deterministic proc
edures, in order to assure their seismic safety.