Clarify output-stream support terminology
Browse files
data/label/README_all_model_phase_picks.md
CHANGED
|
@@ -30,7 +30,7 @@ by observed `k/m`, is released as
|
|
| 30 |
|
| 31 |
The current build contains 49,689,373 candidates from 124,403,181 source picks
|
| 32 |
produced by nine non-standardized picker runs. Of these candidates, 35,638,688
|
| 33 |
-
have one contributing
|
| 34 |
counts describe agreement among the released outputs; they do not establish
|
| 35 |
that a candidate is a true seismic arrival.
|
| 36 |
|
|
@@ -39,10 +39,10 @@ that a candidate is a true seismic arrival.
|
|
| 39 |
P, Pg, and Pn are mapped to canonical P; S, Sg, and Sn are mapped to canonical
|
| 40 |
S. Picks are grouped only when they have the same exact
|
| 41 |
`network.station.location`, the same canonical phase, and a strict maximum time
|
| 42 |
-
span smaller than 1.5 s. The candidate time is the mean of the per-
|
| 43 |
-
pick times, so overlapping-window duplicates from one
|
| 44 |
-
|
| 45 |
-
count is the number of unique contributing
|
| 46 |
are retained explicitly.
|
| 47 |
|
| 48 |
Here a "model" field identifies a source output stream. It is not necessarily a
|
|
@@ -100,7 +100,7 @@ Each decompressed line has the following keys:
|
|
| 100 |
| `p` | Canonical `P` or `S` phase. |
|
| 101 |
| `t` | Candidate UTC epoch time in integer microseconds. |
|
| 102 |
| `sd` | Sample standard deviation of per-model mean times, in microseconds. |
|
| 103 |
-
| `n` | Number of unique contributing
|
| 104 |
| `x` | Member arrays `[model_id, original_phase, dt_us, probability_q]`. |
|
| 105 |
|
| 106 |
For each member, `dt_us` is the source pick time minus `t`.
|
|
|
|
| 30 |
|
| 31 |
The current build contains 49,689,373 candidates from 124,403,181 source picks
|
| 32 |
produced by nine non-standardized picker runs. Of these candidates, 35,638,688
|
| 33 |
+
have one contributing output stream and 14,050,685 have at least two. These support
|
| 34 |
counts describe agreement among the released outputs; they do not establish
|
| 35 |
that a candidate is a true seismic arrival.
|
| 36 |
|
|
|
|
| 39 |
P, Pg, and Pn are mapped to canonical P; S, Sg, and Sn are mapped to canonical
|
| 40 |
S. Picks are grouped only when they have the same exact
|
| 41 |
`network.station.location`, the same canonical phase, and a strict maximum time
|
| 42 |
+
span smaller than 1.5 s. The candidate time is the mean of the per-output mean
|
| 43 |
+
pick times, so overlapping-window duplicates from one output cannot give that
|
| 44 |
+
output additional weight. All source records are retained. The stored support
|
| 45 |
+
count is the number of unique contributing outputs, and single-output candidates
|
| 46 |
are retained explicitly.
|
| 47 |
|
| 48 |
Here a "model" field identifies a source output stream. It is not necessarily a
|
|
|
|
| 100 |
| `p` | Canonical `P` or `S` phase. |
|
| 101 |
| `t` | Candidate UTC epoch time in integer microseconds. |
|
| 102 |
| `sd` | Sample standard deviation of per-model mean times, in microseconds. |
|
| 103 |
+
| `n` | Number of unique contributing source outputs. |
|
| 104 |
| `x` | Member arrays `[model_id, original_phase, dt_us, probability_q]`. |
|
| 105 |
|
| 106 |
For each member, `dt_us` is the source pick time minus `t`.
|