I actually started this a long time ago, but got swamped with other work and buried by snow. Our seismon code does not seem to give an accurate prediction of ground velocities. Attached plot compares the timeseries .03-.1hz rms ground velocity measured by the ITMY STS in Z (solid blue line) and the seismon predicted ground velocities for 5000 minutes of data from this month. Each dot represents the predicted velocity and arrival time for each earthquake seismon found over this time period (seismon can give up to five active predictions, eq chan 1,2 & 3 are the first 3 predictions, typically the "live" ones if seismon has given us early warning). The numbers by each prediction point are the magnitude and distance in 1000km of the earthquake (so 5.8, 10 is a 5.8 magnitude earthquake 10,000 km away). The earthquake at 4000 minutes actually has a  arrival time prediction , but the predicted velocity is up around 60 micron/s, which made the rest of the plot hard to see, so I cut it off.

I would expect that if seismon was accurately predicting ground velocities, the dots would follow the blue line in some way. If there is a systematic relationship between the dots and the line, I don't see it.

The good new is that for other time periods when there were large earthquakes, seismon usually gave us early warning, which is the primary goal. But it would help deciding what the response to a notification should be if we could get a believable prediction of the velocities.