Hi Brian: unless you ask me not to, I'm going to just paste in AGU abstracts. These are public but I don't think you can link to them.

Just in case the link works, I'll test it here.

CONTROL ID: 1809708
TITLE: Failures and suggestions in Earthquake forecasting and prediction.

AUTHORS (FIRST NAME, LAST NAME): Selwyn I Sacks1
INSTITUTIONS (ALL): 1. Carnegie Inst Washington, Washington, DC, United States.

ABSTRACT BODY: Seismologists have had poor success in earthquake prediction. However, wide ranging observations from earlier great earthquakes show that precursory data can exist.
In particular, two aspects seem promising. In agreement with simple physical modeling, b-values decrease in highly loaded fault zones for years before failure. Potentially more usefully, in high stress regions, breakdown of dilatant patches leading to failure can yield expelled water-related observations. The volume increase (dilatancy) caused by high shear stresses decreases the pore pressure. Eventually, water flows back in restoring the pore pressure, promoting failure and expelling the extra water. Of course, in a generally stressed region there may be many small patches that fail, such as observed before the 1975 Haicheng earthquake. Only a few days before the major event will most of the dilatancy breakdown occur in the fault zone itself such as for the Tangshan, 1976 destructive event. Observations of “water release” effects have been observed before the 1923 great Kanto earthquake, the 1984 Yamasaki event, the 1975 Haicheng and the 1976 Tangshan earthquakes and also the 1995 Kobe earthquake. While there are obvious difficulties in water release observations, not least because there is currently no observational network anywhere, historical data does suggest some promise if we broaden our approach to this difficult subject.

• test link Sacks abstract