The hazard is expressed in terms
of the
probability of exceeding a certain level
of shaking in 50 years — not only because the hazard in places like California is not expected to change much
over that
time period, but also because 50 years — the typical life
span of a building — is a useful period
of time for engineers.
Were the hypothesis that warming will increase at least 1C / decade averaged
over a millennium at 95 % confidence, nineteen
times in twenty, given the noise in the signal, all other things being equal, we'd first need 17 years at least to get some kinda sketchy data, and then could begin calculating from the set
of subsequent running or independent 17 year
spans (a different calculation for each, depending on the PDF) the
probability that a -20 C decade would be consistent with a +1 C / decade hypothesis.