A previous attempt to integrate geological, geodetic, and observed seismici
ty data into a probabilistic-hazard source model predicted a rate of magnit
ude 6 to 7 earthquakes significantly greater than that observed historicall
y. One explanation was that the discrepancy, or apparent earthquake deficit
, is an artifact of the upper magnitude limit built into the model. This wa
s controversial, however, because removing the discrepancy required earthqu
akes larger than are seen in the geological record and larger than implied
from empirical relationships between fault dimension and magnitude. Althoug
h several articles have addressed this issue, an alternative, integrated so
urce model without an apparent deficit has not yet appeared. We present a s
imple geologically based approach for constructing such a model that agrees
well with the historical record and does not invoke any unsubstantiated ph
enomena. The following factors are found to be influential: the b-value and
minimum magnitude applied to Gutenberg-Richter seismicity; the percentage
of moment released in characteristic earthquakes; a round-off error in the
moment-magnitude definition; bias due to historical catalog incompleteness;
careful adherence to the conservation of seismic moment rate; uncertainty
in magnitude estimates obtained from empirical regressions; allowing multi-
segment ruptures (cascades); and the time dependence of recurrence rates. T
he previous apparent deficit is shown to have resulted from a combination o
f these factors. None alone caused the problem nor solves it. The model pre
sented here is relatively robust with respect to these factors.