Phase Reading Flags
The MNF file format for earthquake arrival time data files includes the concept of “usage flags”, in column 3 of phase reading records. Usage flags are used by mloc (in combination with explicit commands and built-in logic) to determine which phase readings are available to be used for relocation. If there is a valid flag in column 3 of a phase reading, the reading will not be used; the availability of different flags is simply informational. Some flags (e.g., most flags for duplicate readings) are added by mloc during processing. Much of the work in a relocation with mloc, done by the user between runs, is identifying and flagging outlier readings (i.e., cleaning). This section describes the meanings of the available flags.
The absence of a flag (blank in column 3 of an MNF phase readings record) indicates that there are no known reasons to avoid using it, but this does not guarantee that a particular reading will be used in the relocation. For example, the hypocentroidal decomposition algorithm requires that a station-phase be observed more than once among the events in a cluster, so that a travel-time difference can be calculated, in order to use it to estimate the cluster vectors. If a reading is the only sample of a particular station-phase in the data set, it might still be used for estimating the hypocentroid, but if it does not meet the criteria (e.g., epicentral distance) for use in the hypocentroid, it will play no role in the relocation analysis. The usage (or not) of every reading can be found in the ~.phase_data output file.
Any phase reading with one of the flags defined below in column 3 of its phase record will not be used by mloc for either the hypocentroid or cluster vectors.
There is some rudimentary logic in mloc that attempts to identify duplicate readings automatically, but it is far from thorough. Duplicates can also be flagged manually. Multiple samples of the same station-phase from the same event are not duplicates if they are actually independent estimates, but it can be very difficult to determine from the data itself. This has become a serious problem at the ISC in recent years, as the multiplicity of communication channels for seismic data leads to them receiving reports of the same events from many sources, some of which may have made their own pick from the waveform.
If the user knows in advance that she does not have station coordinates for a station in the arrival time dataset, she can use this flag to explicitly state that fact. It is mainly a matter of good record-keeping because mloc will drop a reading anyway if it cannot find station coordinates.
This flag indicates that the phase name of the reading is problematic (indecipherable) or the seismic phase itself is problematic. It could be that the phase identification algorithm in mloc fails to associate the reading with a known seismic phase, or it may be that the phase, while known, is one for which it is not possible to calculate a theoretical travel time (e.g., PPP in ak135). This flag is applied automatically to depth phases reported from epicentral distances less than 26°, which are known to be subject to bias from upper mantle structures, and for depth phases reported by stations suspected of reporting bogus depth phase readings (command bdps).
Using the skip command, readings can be flagged on the basis of phase name, station, author or a combination of these parameters.
This flag is used if the station is known or suspected to have timing problems. There is documentation of timing problems for a few stations during known periods, and mloc has logic to check for corresponding readings and flag them automatically. The flag can also be applied manually.
One use of the timing problem flag has been to deal with readings from strong motion instruments that do not have calibrated timing systems. The P and S arrivals can be read from nearby earthquakes and a new phase record for S-P can be added to the MNF file while the original P and S phase records are given the t flag.
Another use has been in dealing with suspected one-minute errors (and on occasion, one-hour errors) in reported arrival times, which used to be quite common in the analog record era. To correct a suspicious reading, the original phase record is flagged and a copy of the record is added, in which the arrival time is (hopefully) corrected and the author of the reading is changed to be the user. This provides a clear record of what was done.
It is obviously a very dangerous procedure, but the timing of a specific station can be systematically altered with the terr command. Unless there is information to corroborate its use, terr should only be used in an experimental mode.
This is the most important and commonly-used flag. A reading may be determined as an outlier and flagged manually during analysis of empirical reading errors with the utility program rstat or through inspection of the ~.phase_data file, or it may be flagged semi-automatically as a result of running the lres or xdat utilities. The concept “outlier” can be either relative (to other samples of the same station-phase) or absolute (with reference to a theoretical travel time model).