~.hdf Files
Each run of mloc creates one or more output files that summarize the relocated hypocenters of all events in the cluster in a “one line per event” format. The reason for multiple instances of the file is explained below. These files are known generically as HDF files and they all contain hdf as part of the file name suffix. The three possible filename extensions are:
- ~.hdf
- ~.hdf_dcal
- ~.hdf_cal
Different Flavors of HDF
The need for several variations of the HDF file in mloc arises from the different scenarios that are possible:
- No calibration. Only a ~.hdf is created. Uncertainties are for relative location only (cluster vectors).
- Direct calibration. Only a ~.hdf_dcal is created. Uncertainties are for absolute location (cluster vector plus hypocentroid).
- Indirect calibration. Two HDF files are created, a ~.hdf file (with uncertainties for relative location) and a ~.hdf_cal file (with uncertainties for absolute location).
- Direct calibration, followed by indirect calibration in the same run. Two files are created, a ~.hdf_dcal file and a ~.hdf_cal file. Indirect calibration takes precedence.
All HDF files, regardless of flavor, may be read with the same code. The interpretation of fields dealing with uncertainty of a parameter vary, depending on the flavor.
Format Description
Columns | Description |
---|---|
1:4 | Origin year (i4) |
6:7 | Origin month (i2) |
9:10 | Origin day (i2) |
12:13 | Origin hour (i2) |
15:16 | Origin minute (i2) |
18:22 | Origin seconds (f5.2) |
24:32 | Geographic latitude (f9.5) |
34:43 | Geographic longitude (f10.5) |
45:50 | Focal depth, km (f6.2) |
52:52 | How starting depth was set (a1). See depth codes. |
53:53 | Free depth flag (a1) (Note) |
54:59 | Depth from input file (f6.2) (Note) |
61:63 | Magnitude (f3.1) (Note) |
64:65 | Magnitude scale (a2) (Note) |
67:76 | Event ID (a10) (Note) |
78:81 | Number of observations contributed to hypocentroid estimation (i4) (Note) |
83:86 | Number of observations used for cluster vector (i4) (Note) |
88:91 | Number of observations flagged as outliers, fcode = ‘x’ (i4) (Note) |
93:98 | Normalized sample variance for the cluster vector (f6.2) (Note) |
100:104 | Uncertainty in origin time, sec (f5.2) (Note) |
106:109 | + uncertainty in depth (deeper), in km (f4.1) (Note) |
111:114 | – uncertainty in depth (shallower), in km (f4.1) (Note) |
116:120 | Epicentral distance of nearest station for cluster vector (f5.1) (Note) |
122:126 | Epicentral distance of farthest station for cluster vector (f5.1) (Note) |
128:132 | Largest open azimuth for cluster vector (f5.1) (Note) |
134:136 | Semi-axis azimuth (i3) (Note) |
138:142 | Semi-axis length, km (f5.2) (Note) |
144:146 | Semi-axis azimuth (i3) (Note) |
148:152 | Semi-axis length, km (f5.2) (Note) |
154:159 | Area of confidence ellipse, km2 (f6.1) (Note) |
161:164 | Calibration code (a4) |
166:185 | Annotation (a20) (Note) |
Free Depth Flag
The column immediately after the focal depth constraint flag is used for a special flag “f” which indicates that the focal depth has been a free parameter in the relocation.
Depth from Input File
This variable holds whatever focal depth was specified for an event in the input data file. Depending on what other commands are issued in the command file or interactively this value may or may not be used as the starting depth for relocation. It is sometimes useful to have this for comparison with the depth actually set for (or determined in) relocation. Large discrepancies, more than, say, 10 or 15 km, may warrant investigation.
Magnitude and Magnitude Scale
A single representative (or preferred) magnitude is carried through the mloc analysis to assist in interpretation. It has no bearing on the relocation. MNF input files can carry multiple estimates of magnitude, but one will be specified (or selected by default) as the preferred magnitude. If no magnitude estimate is available, these two fields will be blank. In some cases a magnitude may be available but not a scale, in which case the magnitude scale field will be blank.
Event ID
This field is provided to carry an EVID that was assigned by some relational database, but as far as mloc is concerned it is simply a character variable of 10 characters. If the field has an entry it will have been read from an MNF input file. If an integer EVID is stored in the field it should be right-justified.
Number of Observations (phase readings)
The HDF format carries three different counts of number of readings:
- Number of observations contributed to hypocentroid estimation
- Number of observations used for the cluster vector
- Number of observations flagged as outliers
It is wise to review carefully the reliability of the relative location (cluster vector) for an event that contributes many readings to the hypocentroid in a direct calibration study. Conversely, events with relatively small number of readings contributing to the cluster vector should be reviewed carefully, because the relative location of such events is likely to be poorly determined.
The third variable in this set carries the number of flagged readings that are due to being judged to have a large residual (so-called “cluster residuals” in which the residuals are compared on the basis of mutual consistency rather than absolute value). These are the readings for which fcode = ‘x’. It is not uncommon for the number of outliers to be a significant fraction of the number of readings used for the cluster vector, but when the fraction becomes notably large (say, more than 25%) investigation is warranted.
There are other reasons why a reading may be flagged to prevent its being used in the relocation but these are not tabulated in the hdf file. Examples include duplicate readings, unknown phase types or phase types that are not used for relocation, and stations for which no coordinates are available.
Normalized Sample Variance
This field carries a measure of the statistical self-consistency of the error budget related to the cluster vector, the normalized sample variance. If the data followed our statistical model perfectly (i.e., data drawn randomly from a normal distribution with spread equal to our estimated empirical reading error) the expected value would be 1.0. Values both larger and smaller are actually observed, naturally, and the range of variation declines as the analysis proceeds and outlier readings are identified and removed by flagging.
A Bayesian term in the calculation of the normalized cluster sample variance represents our a priori state of knowledge about its expected variability and prevents it from going unrealistically small for events with few data points. This expectation on the spread of values for the normalized sample variance (about 0.35) provides a check on the internal consistency of the statistical model for each event. Values larger than about 2.0 (~3σ) may reveal the presence of readings that violate the statistical model (e.g., outliers). By the end of a relocation analysis there should be few such cases.
Uncertainty of Origin Time
In seconds. For uncalibrated clusters and direct calibration, it is taken from the covariance matrix of the relocation, including uncertainty of hypocentroid and cluster vector. For indirect calibration it is based on the uncertainty of the calibration shift plus the uncertainty of relative origin time from the cluster vector.
Uncertainty in Focal Depth
Uncertainty in focal depth can be read from the input file, from the command file, or from the relocation. Uncertainty in focal depth is carried in two fields because it is not uncommon for the uncertainty to be asymmetric. This can arise when inferring depths from teleseismic depth phases and there is uncertainty about the correct identification of the phase (pP, sP, or pwP). It can also arise in waveform analyses where the error vs depth curve is not symmetric. If focal depth has been a free parameter in the relocation it will be symmetric, and it will over-ride any specification in the input file and command file. For uncalibrated clusters and direct calibration with a free depth solution, it is taken from the covariance matrix of the relocation, including uncertainty of hypocentroid and cluster vector. For indirect calibration it is based on the uncertainty of the calibration shift plus the uncertainty of relative depth from the cluster vector. If no estimate of uncertainty in focal depth is available, the fields are blank.
Epicentral Distance Range
The epicentral distance, in degrees, of the nearest and farthest reading used for the cluster vector of an event is carried in these fields. This is useful for judging how an event is connected to the cluster, i.e., through local readings or teleseismic readings, or both.
Open Azimuth
The largest open azimuth for the readings used to estimate the cluster vector is carried in this field. The reliability of the relative location of an event should be questioned if this value is much greater than 180°.
Confidence Ellipse
The confidence ellipse (90% confidence level) for the epicenter is defined in four columns which give the azimuth and length of each semi-axis (half-length). The shorter semi-axis is given first. Azimuth is in integer degrees, clockwise from North. Semi-axis lengths are given in decimal km.
A fifth column carries the area of the 90% confidence ellipse, in km2. This is a convenient metric to monitor when searching for events with problems. A circle of 5 km radius (the canonical GT5 location) has an area slightly greater than 75 km2.
As discussed above, the interpretation of the confidence ellipse (relative vs absolute location) depends on whether the cluster has been calibrated or not, which is indicated by the file name suffix.
Annotation
It is possible to declare a comment or annotation for an event using the anno command. An annotation can also be read from an MNF input file which will take precedence over an annotation given in the command file. Any annotation will be appended to the end of the line in all HDF files. The comment is limited to 20 characters.
Example
The Ahar, Iran cluster was calibrated using direct calibration followed by indirect calibration so the run produced both a ~.hdf_dcal and a ~.hdf_cal file. The shift from “direct” to “indirect” results was small (0.9 km at 20° for the epicenters) and uncertainties are quite similar in both results, providing additional confidence in the results. Here are the first dozen lines from both HDF files:
ahar12.17.hdf_dcal
2012 8 11 12 23 14.07 38.39981 46.83729 12.60 m 12.60 6.2mb 8 1988 401 0.71 0.12 1.2 1.2 0.2 164.7 12.2 275 0.70 5 1.75 3.8 CH02 2012 8 11 12 30 11.39 38.44147 46.75668 17.30 m 17.30 4.5 8 75 28 1.35 0.16 1.3 1.3 0.1 78.3 57.3 275 0.99 5 2.12 6.6 CH02 2012 8 11 12 34 32.89 38.44719 46.76897 19.00 m 19.00 6.1mb 2 1884 192 0.56 0.11 1.2 1.2 0.1 133.5 12.8 274 0.69 4 1.74 3.8 CH02 Iran-Armenia-Azerbai 2012 8 11 12 49 14.32 38.41455 46.68159 15.90 m 15.90 4.8mb 9 197 76 1.21 0.14 1.2 1.2 0.1 92.5 29.2 275 0.85 5 1.89 5.1 CH02 Iran-Armenia-Azerbai 2012 8 11 13 5 53.12 38.43239 46.70070 16.00 m 16.00 4.3mb 5 100 64 1.00 0.14 1.3 1.3 0.3 76.6 47.9 276 0.89 6 1.88 5.3 CH02 2012 8 11 13 14 4.36 38.43790 46.68469 17.00 n 15.10 4.7mb 5 305 75 0.74 0.12 3.0 3.0 0.1 94.8 22.4 277 0.80 7 1.83 4.6 CH02 2012 8 11 13 42 10.33 38.42986 46.68102 16.90 m 16.90 4.0mb 5 107 65 1.04 0.14 1.3 1.3 0.3 76.6 22.5 274 0.88 4 1.88 5.2 CH02 2012 8 11 13 54 20.69 38.43172 46.80892 17.00 n 17.80 3.4mb 8 36 17 0.84 0.14 3.0 3.0 0.2 27.2 74.2 273 0.88 3 1.68 4.7 CH02 2012 8 11 14 10 2.64 38.45308 46.66595 19.00 n 18.80 3.6mb 8 29 29 0.86 0.14 3.0 3.0 0.3 26.3 77.3 278 0.88 8 2.04 5.6 CH02 2012 8 11 14 16 47.52 38.44926 46.70688 20.00 n 17.90 3.6 8 43 21 1.04 0.15 3.0 3.0 0.3 27.3 74.9 275 0.94 5 2.23 6.6 CH02 2012 8 11 14 25 14.24 38.43574 46.68371 15.00 n 15.40 4.5mb 4 202 67 0.80 0.12 3.0 3.0 0.3 85.7 22.5 277 0.82 7 1.87 4.8 CH02 2012 8 11 14 33 52.34 38.45821 46.80797 17.20 m 17.20 3.8 7 42 13 0.46 0.12 1.2 1.2 0.2 14.4 71.6 276 0.76 6 1.91 4.6 CH02
ahar12.17.hdf_cal
2012 8 11 12 23 14.05 38.40744 46.84089 11.60 m 12.60 6.2mb 8 1988 401 0.71 0.20 1.2 1.2 0.2 164.7 12.2 271 0.94 1 1.98 5.9 CH02 2012 8 11 12 30 11.37 38.44911 46.76028 16.30 m 17.30 4.5 8 75 28 1.35 0.23 1.3 1.3 0.1 78.3 57.3 272 1.18 2 2.31 8.6 CH02 2012 8 11 12 34 32.88 38.45483 46.77257 18.00 m 19.00 6.1mb 2 1884 192 0.56 0.20 1.2 1.2 0.1 133.5 12.8 271 0.94 1 1.97 5.8 CH02 Iran-Armenia-Azerbai 2012 8 11 12 49 14.31 38.42219 46.68519 14.90 m 15.90 4.8mb 9 197 76 1.21 0.22 1.2 1.2 0.1 92.5 29.2 272 1.06 2 2.11 7.0 CH02 Iran-Armenia-Azerbai 2012 8 11 13 5 53.10 38.44003 46.70430 15.00 m 16.00 4.3mb 5 100 64 1.00 0.21 1.3 1.3 0.3 76.6 47.9 272 1.10 2 2.09 7.2 CH02 2012 8 11 13 14 4.35 38.44554 46.68829 16.00 n 15.10 4.7mb 5 305 75 0.74 0.20 3.0 3.0 0.1 94.8 22.4 273 1.03 3 2.04 6.6 CH02 2012 8 11 13 42 10.31 38.43749 46.68462 15.90 m 16.90 4.0mb 5 107 65 1.04 0.22 1.3 1.3 0.3 76.6 22.5 271 1.08 1 2.09 7.1 CH02 2012 8 11 13 54 20.67 38.43935 46.81252 16.00 n 17.80 3.4mb 8 36 17 0.84 0.22 3.0 3.0 0.2 27.2 74.2 89 1.08 179 1.92 6.5 CH02 2012 8 11 14 10 2.63 38.46071 46.66955 18.00 n 18.80 3.6mb 8 29 29 0.86 0.22 3.0 3.0 0.3 26.3 77.3 274 1.09 4 2.24 7.7 CH02 2012 8 11 14 16 47.50 38.45690 46.71048 19.00 n 17.90 3.6 8 43 21 1.04 0.23 3.0 3.0 0.3 27.3 74.9 272 1.13 2 2.41 8.6 CH02 2012 8 11 14 25 14.22 38.44337 46.68731 14.00 n 15.40 4.5mb 4 202 67 0.80 0.21 3.0 3.0 0.3 85.7 22.5 273 1.04 3 2.09 6.8 CH02 2012 8 11 14 33 52.33 38.46585 46.81157 16.20 m 17.20 3.8 7 42 13 0.46 0.21 1.2 1.2 0.2 14.4 71.6 272 0.99 2 2.12 6.6 CH02
The input_depths shown here are uncharacteristically close to the final depths, because the event files for this cluster used an earlier calibrated relocation run of mloc for the preferred hypocenters.
No event IDs were used in this cluster. The annotations shown were taken from the geographic description provided with the ISC data, but they were truncated by the 20-character limit in the HDF format.