| The purpose of this brief document is to describe the quality control tests and
procedures which the Albuquerque Seismological Laboratory (ASL) Data Collection Center
(DCC) applies to tape based data from the IRIS/USGS Global Seismographic Network. The primary goal of the ASL DCC quality control (QC) effort is to ensure that ASL-operated stations produce the maximum possible amount of high-quality, problem-free, data. Thus, our highest priority is given to eliminating problems at the source: catching problems as quickly as possible, working with the field engineers to get these problems solved, then verifying the solutions are successful. The secondary goal of the QC effort is to repair data and/or document problems such that problem data becomes usable data. The third goal of our QC work is to document historical data problems. Such work typically results in flagging questionable or defective data. Our QC practices can be roughly categorized as examining the data for problems in the following areas:
In addition, we put extensive QC time and effort into:
This document is organized from the perspective of what QC procedures are performed, what accuracies are obtainable, and what percentage of the total data flow are subject to these QC reviews. Routine Data Quality Control Tests
interval: each station is tested once per 10 days of data, on average channels: BH (from broadband sensors) and SH (from short period sensors) description: If a single station has instruments with overlapping pass bands (e.g. Geotech KS54000 and Streckeisen STS-2), then comparisons are made between the appropriate channels of the different instruments. Sensitivity comparisons are done in the frequency domain (using power spectral density) with corrections for the different instrument responses (which has the added benefit of testing the specification of the instrument response in our database). Orientation comparisons are usually performed in the time domain, by bandpass filtering and overplotting the corresponding channels from different sensors.
interval: every station is reviewed after installation and after service visits channels: VH description: All of the very long period (0.1 Hz sampling) data are reviewed for consistency with the amplitude and polarity of synthetic earth tides. This can be a very accurate test of polarity. However, as the earth tide synthetics are not highly precise, this is just a rough check on sensitivity. In addition, at many stations the horizontal recordings at tidal frequencies are significantly contaminated by tilt and the tide recordings are nearly useless. At many stations it is necessary to low-pass filter the data (e.g. corner at 0.001 Hz) before making these time domain comparisons.
interval: each station is tested on one to three earthquakes per every ten days of data channels: BH and SH description: We attempt to extract three earthquake windows (using earthquakes appearing in the Harvard rapid CMT bulletin) out of every ten days of data. A simple three component particle motion analysis is performed to verify the earthquake location is indicated for the proper quadrant. In most cases, one clear earthquake with usable P waves yields a satisfactory result. More earthquakes are used if necessary or if problems are suspected (in some cases, the P wave is plotted on the focal sphere if a recording is suspected to be near-nodal). Mis-orientations as small as 5° to 10° have been detected through routine particle motion analysis. Smaller mis-orientations can be verified, but probably would not be caught by the routine checks.
interval: each station is tested on one to three earthquakes per every ten
days of data
interval: each station is tested on one to three earthquakes per every ten
days of data
interval: continuous
channels: VM or UM from Streckeisen STS-1 broadband sensors interval: every magnetic tape received from stations running Ultrashear software (stations fill tapes about every ten days, on average) description: The data acquisition computer writes a console log to the data
tape. This log is extracted and parsed for various classes of errors and/or warnings.
Plots are made of some variables. The description: This is a new series of tests, installed in August and September,
1997. These tests check all SEED records in our system. Minor problems in SEED data-record
headers are repaired and unexpected discrepancies are flagged. Unusual data record
duplication problems caused by the Quanterra's are repaired. Seismic data are useless unless one knows where the data were recorded and by what
instrument. Thus, we maintain a very large database of coordinates and complete instrument
response information for every channel of every sensor and data acquisition system ever
operated at any of our stations since their initial installation. The database information
is entered and updated by our database specialist. The specialist reviews all information
for a station after every modification to the database. All routine interactions with the
database are performed via a graphical user interface (GUI) to reduce the possibility of
data entry errors. A new dataless SEED volume is made whenever the database is modified.
These dataless SEED volumes are run through VERSEED to look for possible errors or
inconsistencies. In addition, the dataless SEED volumes are parsed into RESP files using
the RDSEED software. These RESP files are used for all QC analysis where an instrument
response is required. Thus, many of our other QC tests are utilizing database information
and provide a test of database consistency and accuracy. When we detect or suspect unusual problems at a station, we perform whatever tests
and/or checks we think may shed light on the problem. For example, we may select a number
of earthquakes where we can compare recordings made at a suspect station with recordings
from a nearby station. We may compare the seismic channels to temperature or wind
channels. We might compare noise spectra to long term (or published) values for a station.
We will perform a detailed examination of the data acquisition log. In some instances it
is necessary to examine data records in the order they were written to tape, one record at
a time, to verify data acquisition related problems Calibrations are regularly performed on IRIS/USGS stations and the resulting signals are analyzed. These calibrations do not assess the absolute calibration of the sensors, but are capable of detecting variations in instrument response as a function of frequency, and detecting long-term drift in the instrument response. Data Sources Quality control tests are performed on data which are arriving in a variety of ways.
Perhaps the most important QC tests are those performed on data retrieved via dialup or by LISS while
field engineers are on site at a station. Such data review has the greatest impact on preventing
problems. Problems not detected during installation onsite visits can result in months of
problematic data. |
sensitivity
and orientation test: comparison between co-located sensors