In the spring of 1997 the Long Valley Caldera began a new phase of unrest
that
lasted through early 1998.
This unrest initiated in May with reinvigored inflation of the resurgent dome observed on
Long Valley geodetic networks (Langbein et al., 1998). In June the seismicity rate began to increase
markedly, the deformation rate continued to be high and the flux of *CO*_{2}
release increased between September to December.
The seismic unrest culminated on November 22 with a series of four
magnitude 4.6-4.9 earthquakes located in the south moat of the caldera.
Preliminary moment tensor analysis of these events revealed anomalous
seismic radiation. Deviatoric moment tensors have large non-double-couple(NDC) components, and full moment tensor inversions reveal a component of
coseismic volumetric expansion in all four events (Figure 13.1).
These are the first direct observations of coseismic volumetric
expansion in Long Valley seismicity.

The double-couple and full moment tensor inversions have 5 and 7 free
parameters respectively. We evaluated whether the full moment tensor
solutions provides a significant
improvement in fit considering the increase in model parameters using an
F-test (e.g. Menke, 1989). The F statistic is formed from the ratio of
the two data-model variances.
It is possible to
compute critical values, *F*_{crit}, for specific levels of confidence
and degrees of freedom in the inversion (e.g. Menke, 1989). If observed
F exceed *F*_{crit} then the improvement in fit is statistically
significant at the prescribed level.
In our inversions of complete, three-component,
long-period (0.02 to 0.05 Hz) displacement waveforms we have 3000 data
points for 5 stations (200 points per component with dt=1s). The F-test assumes that each time point is uncorrelated,
which is clearly not valid. For example a number of points
are required to define the pulse width of the P-wave. We use a 20s period
low pass butterworth filter in processing the data and therefore assume that the
data is correlated over a 20s time scale. This reduces the number of
uncorrelated data to 150 points. The
degrees of freedom are then 145 and 143 for the double-couple and full moment tensor inversions, respectively.
The *F*_{crit} for these degrees of freedom and 95% significance level
is 1.32. Observed F
values for events 1, 2, 3 and 4 are 1.96, 2.06, 3.59, and 2.33 respectively,
indicating that the improvement in fit afforded by the full moment tensor
parameterization is statistical significant above the 95% level. We note
further that even in the extreme case of two uncorrelated points per
trace the observed F are statistically significant at the 95% level.

A Jackknife test was performed to investigate the sensitivity of the isotropic components and general orientation of the principle axes of the seismic moment tensors to variations in the input data. This test evaluates whether lateral heterogeneity or whether a single station or groups of stations may be biasing the results. The Jackknife test revealed that the isotropic components for the anomalous events are significantly separated from double-couple reference earthquakes, and that the isotropic components that have been calculated are stable for all combinations of 1-, 2-, 3-, 4- and 5-stations indicating that there is no station specific or propagation path bias.

Figure 13.1 shows that the solutions for the anomalous events have significant isotropic components indicating volumetric expansion at the source. With the exception of EVT1 the deviatoric parts of the moment tensors are primarily CLVDs. The combined isotropic-CLVD solutions and anomalously long source durations apparent in the P-waveforms of these events together with the observation of a compressional strain transient recorded at the Devil's Post Pile borehole strainmeter suggest that fluids were involved in the source process (Dreger et al., 1999). Langbein et al (1998) report that the strain transient and other geodetic data may be modeled as either an inflation source beneath the south moat of the caldera or as right-lateral strike-slip on a west-northwest trending fault, which is consistent with background seismicity. The seismic moment tensor observations and the strain transient suggest that seismicity was triggered by the injection of high pressure fluid or possibly by pressurization of water saturated faults in the south moat by a pressure pulse generated by magmatic heating.

Langbein, J., S. Wilkinson, M. Johnston, J. Feinberg, R. Bilham, The 1997-98 Inflation of Long Valley Caldera and comparison with the 1989-95 episode, EOS, 79, 949, 1998.

Menke, W., Geophysical Data Analysis: Discrete Inverse Theory, Academic Press Inc., pp289, 1989.

The Berkeley Seismological Laboratory, 202 McCone Hall, UC Berkeley, Berkeley CA 94720

Questions and comments to www@seismo.berkeley.edu

Copyright 1999, The Regents of the University of California.