PRELIMINARY FORECAST TEST FOR 2001:


Three Postscript diagrams show prospective test results for NW and SW Pacific
forecasts made for calendar year 2001. Presently we have the Harvard catalog in
final form for 2001/01/01-2001/08/31, but we obtained earthquake positions for
2001/09/01-2001/12/04 from Harvard e-mail messages. We will update the results
when the catalog for all of 2001 is final; that will probably occur around
March 2002. Our previous experience suggests that results for the final catalog
may differ perceptibly from those based on the preliminary one. 

The maps are in a format like that of Figure 8 of Kagan and Jackson (2000; see
RECENT PUBLICATIONS below in this WEB page). Parameters of the 2001 forecasts
are the same as used in 2000 for both regions. The likelihood test diagram is
similar to Figure 9 in the paper. The earthquakes in the SW Pacific are
consistent with the forecast results (that is, the catalog falls within the 95%
confidence intervals based on simulations using the model probabilities).
However, earthquakes in the NW Pacific  are concentrated closer to past
earthquakes than the model predicted. 

REFERENCES:

Jackson, D. D., and Y. Y. Kagan, 1999. Testable earthquake forecasts for 1999,
Seism. Res. Lett., 70, 393-403.

Kagan, Yan Y., and David D. Jackson, 2000. Probabilistic forecasting of
earthquakes, (Leon Knopoff's Festschrift), Geophys. J. Int., 143, 438-453.

FIGURE CAPTIONS:

Figure 1a,b.

Color scale tones show the probability of earthquake occurrence calculated
using the Harvard 1977-2000 catalogue; preliminary positions of earthquakes in
2001/01/01-2001/12/04 are shown in white.

(a) Northwest Pacific long-term seismicity forecast: latitude limits from 0.25S
to 60.25N, longitude limits from 109.75E to 170.25E; Events from 2001/1/1 to
2001/12/04.

(b) Southwest Pacific long-term seismicity forecast: latitude limits from 0.25N
to 60.25S, longitude limits from 109.75E to 169.75W; Events from 2001/1/1 to
2001/12/04.

Figure 2.

Likelihood test of the forecast in previous Figure. We use the 1977-2000
earthquakes as a control set. The solid line is the best Gaussian curve, having
the same mean and standard deviation as the simulations. The dashed curve
corresponds to simulation distributions for the NW-Pacific, while the
dash-dotted curve is for the SW-Pacific. Curves to the left of the Gaussian
curve indicate that the actual earthquakes were more concentrated in the red
zones (high probability) than predicted by the model, so that the model
forecasts over too broad an area. Curves to the right of the Gaussian curve
indicate that the earthquakes were not as concentrated in the red zones as
expected from the probability model.