The two primary goals of this project are to provide quality control
for waveform
data at the IRIS DMC and to routinely pick times for stations that may
not otherwise
be picked. Our focus is to pick P-wave travel times and amplitudes from
BHZ and SHZ channels of all stations from all Networks that are were in the
IRIS FARM/POND for events prior to 2000 and for all unrestricted BHZ channels after that.
Click on the following link to download the
phase picks. These picks are stored by event with each file
containing all the events for a given month. The IMS1.0 format is
described in two documents
that can be downloaded from the ISC . The two documents
describe the IMS1.0 format and the extensions known as the IASPEI
Seismic Format (ISF).
There are similar picking projects at the IRIS/USGS and IRIS/IDA Data Collection Centers. These groups pick times of phases as data comes in from the field as part of the IRIS quality control efforts. Our efforts are complementary to theirs in that we plan to pick PASSCAL data, and to pick times from older earthquakes. On the other hand, there is some duplication of effort in that we also pick recent events that have already been picked by these groups. A histogram comparing our picks with the IRIS/USGS picks made at the start of this project when there was a small number of picks shows that our picks generally agree well. The outliers on the histogram are picks that we assigned a lower quality.
The data we pick are being sent to the ISC for inclusion in their
bulletin as well as to R. Engdahl for reprocessing in his
bulletin. They are also sent to the IRIS DMC for inclusion in
their data base.
Absolute time picks and a description of file format are available for:
P-waves from the 1997 IRIS FARM and from the 1998 IRIS FARM.
P-waves in the epicentral distance range 0-100o from the xe94 PASSCAL experiment.
PKP-waves in the epicentral distance range 120-180o from the xe94 PASSCAL experiment.
Plots of travel-time residuals versus event date summarize the dates that each station is are operating, the number of useful P picks that are obtained for each station, and the variability of the residuals.
For all 219 earthquakes in the 1997 IRIS FARM ( MW > 5.6 and depth > 100 km or events with MW > 5.8) and all seismograms in the epicentral distance range 20 to 100o we attempted to pick P-wave arrival times. We made one pass through the waveforms picking times from the broadband vertical-component channels and a second pass picking times from the same waveforms after filtering to a narrow-band short-period response. The short-period seismograms were obtained by first deconvolving the instrument response (poles and zeros only) and then convolving the standard Digital World-Wide Standardized Seismograph Network (DWWSSN) short-period response. This report analyzes the differences, advantages and disadvantages among teleseismic P-wave picks from these two waveform filters. The picks themselves and a description of file format are available for the broadband waveforms and from waveforms filtered to short-period.
Each phase pick is assigned a quality number from 1 to 4. Quality 1 implies that the arrival is impulsive and the analyst is 99% sure of the correct cycle, quality 2 is picked to within 1 s with 95% confidence. Quality 3 is for emergent arrivals in which the analyst is 67% confident of the correct cycle, while quality 4 means energy is present on this seismogram, but a start time is difficult to identify. Only qualities 1 and 2 are included in the data files above.
For the highest quality data, picks from the broadband and short-period responses give nearly identical results. For this comparison, the quality of a pair of picks made from the two filters is the maximum of the two qualities. The standard deviations for the difference in pick times between broadband and short-period, quality 1, 2 and 3 picks respectively are 0.14, 0.38 and 1.14 s. Histograms of these time differences show that the picking errors are consistent with our picking quality criteria. In particular, about 99%, 95% and 67% of the pick differences are less than +/- 0.8 s for the quality 1, 2 and 3 picks respectively.
However, there appears to be a phase delay associated with the DWWSSN short period instrument response such that these arrivals are systematically picked 0.06 s later than the picks from the raw data. To avoid bias from outliers the mean is determined by the following procedure. First we calculate the mean and standard deviation. Then we recalculate the mean of the data that lie within 2 standard deviations of the first estimate of the mean. Three independent estimates of this phase delay are determined from the quality 1, 2, and 3 data to be 0.06+/-.01, 0.05+/-.01, and 0.06+/-.02 where the error estimates are standard error in the mean. For most applications 0.06 s is not an important difference, but it is statistically significant. The short-period picks are late by 0.06 s (about one sample) relative to the broadband picks, as would be expected because of the causal narrow band filter.
Our experience is that that for most of the teleseismic events we analyzed, it is easier to pick the waveforms from the broadband seismograms because the extra bandwidth at high frequencies allows sharper onsets and the bandwidth at the long periods can make for more distinctive arrivals (e.g. Seismograms on the right ). On the other hand, it is not uncommon, especially for the smaller events, for the microseismic noise (dominated by 6 s periods) to mask the arrival of a phase on the broadband records. In these cases, the short-period response filters out the microseismic noise allowing an easier identification of a seismic phase (e.g. Seismograms on the left ). In this figure the event origin date/time, latitude, longitude, depth and moment magnitude are given in the title along with the epicentral distance and station name. The filtered and unfiltered seismograms are also shown.
To our surprise, there is no clear advantage to picking teleseismic P-wave travel times from the broadband or the short-period response. We obtained about the same number of quality 1, 2, and 3 picks from each of the filters, and the scatter from each was about the same. Because the amplitudes used to make estimates of body wave magnitude should be estimated from short-period waveforms, and there is no clear advantage to the broadband seismograms, we plan to make our P-wave picks on short-period seismograms.