Thanks for the kind words Petra.
Seems to me that the kind of forecasting that Don is doing is on the right track, but
perhaps we are looking at this from the wrong direction. Most of us on this board
now know that Don has a good track record in being able to locate within a few
days the epicenters of many earthquakes that do then occur. He has, and has
admitted, that there is a problem with the magnitudes - that is, he can't seem to
be able to tell when an earthquake is going to be large and when it will be just a
"popcorn" quake.
If Don is right, and it appears that he is, forecasting may be possible right now.
But, instead of saying when an earthquake is likely to happen, we ought to be
saying when it is NOT likely to happen. With the technique Don has, he could
make a forecast with a reasonable sense of accuracy based on past track record,
that if his data does not show anything in San Francisco between day X and day Y
then it is very likely that nothing will happen. On the other hand, if something
does show, something like a weather forecast could be made - perhaps something
along the following lines:

Data: Don has data showing possible quakes in the region of Parkfield but not
in San Francisco or Los Angeles for the next 3 days. He could issue the following
earthquake advisory with relative confidence:

" There is a 50% chance of an earthquake in the Parkfield area in the next three
days of Ml>=2. San Francisco and Los Angeles each have less than 10% chance
of such an event. If you live in the Parkfield area and feel an earthquake beginning,
be prepared to take immediate precautions should the shaking last more than 5
seconds"

A daily probability map for the state could also be produced and added to the
daily weather report - perhaps first as something to be talked about, but later
as something to be considered as reliable, if the data supports this conclusion.