guided_repick¶

Science¶

Run a deep-learning picker in a short window around a predicted arrival at one station, and return referenceable picks. The window is centred on a given time or on the predicted P or S of a referenced origin, and is capped at 30 s: a wide window defeats the point of guiding and reintroduces the false picks that blind low-threshold picking produces.

The menu is phasenet or eqtransformer, with a discrete picker_threshold and filter_band. The model sees the full three-component stream; channel only chooses which component's picks are reported. Each pick has a pick_index so a later relocate can add it by reference.

noise_context is returned unasked: the 1–20 Hz acceleration PSD of the 10 s before the searched window, the weekly baseline p50_db, and the percentile of that baseline. A low-threshold pick on a noisy station is the usual failure mode — the picker will mark noise as well as phases.

Finding nothing is informative. Finding something is not proof the phase is real.

References¶

Zhu, W. and Beroza, G. C. (2019). PhaseNet: a deep-neural-network-based seismic arrival-time picking method. Geophysical Journal International. doi:10.1093/gji/ggy423
Mousavi, S. M., Ellsworth, W. L., Zhu, W., Chuang, L. Y. and Beroza, G. C. (2020). Earthquake transformer—an attentive deep-learning model for simultaneous earthquake detection and phase picking. Nature Communications. doi:10.1038/s41467-020-17591-w
Woollam, J. et al. (2022). SeisBench—A toolbox for machine learning in seismology. Seismological Research Letters. doi:10.1785/0220210278

Contract¶

Generated from the live registry (ToolSpec). Field names, types and defaults come from the input and payload models; they are not hand-typed.


Group	`repick`
Class	4
Tool version	`1.0.1`
Highest declared tier	E
Timeout	20 s
Budget key	`guided_repick`
`readOnlyHint`	true
`idempotentHint`	false

Input¶

Field	Type	Required	Notes
`station_id`	string	yes
`target`	object (kind `origin` \| `time`)	yes	discriminated on `kind`
`picker`	string ∈ {`phasenet`, `eqtransformer`}	yes
`picker_threshold`	number ∈ {`0.1`, `0.2`, `0.3`, `0.5`}	yes
`channel`	string ∈ {`Z`, `horizontals`, `all`} \| null	no	default `null`
`window_s`	number	no	default `10.0`; ≥ `2.0`; ≤ `30.0`
`filter_band`	string ∈ {`none`, `0.5-5`, `1-20`, `2-8`, `2-15`}	no	default `"none"`
`hypothesis`	string \| null	no	default `null`

Payload¶

Field	Type	Required	Notes
`station_id`	string	yes
`channel_ids_used`	array[string]	yes
`target_time_utc`	string	yes
`searched_window`	object	yes
`searched_window.from_utc`	string	yes
`searched_window.to_utc`	string	yes
`picks`	array[object]	yes
`picks[].pick_index`	integer	yes
`picks[].phase`	string	yes
`picks[].time_utc`	string	yes
`picks[].probability`	number	yes
`picks[].offset_from_target_s`	number	yes
`picks[].channel_id`	string	yes
`noise_context`	object	yes
`noise_context.measured_noise_db`	number \| null	no	default `null`
`noise_context.baseline_p50_db`	number \| null	no	default `null`
`noise_context.percentile_of_baseline`	number \| null	no	default `null`
`noise_context.method`	string	yes
`noise_context.definitions`	object	yes
`model`	object	yes
`model.picker`	string	yes
`model.weights`	string	yes
`model.seisbench_version`	string \| null	no	default `null`

Example request¶

{
  "station_id": "YY.STA3",
  "channel": "Z",
  "target": {
    "kind": "origin",
    "origin_ref": {
      "kind": "catalog",
      "event_id": 1001
    },
    "phase": "P"
  },
  "window_s": 10,
  "picker": "phasenet",
  "picker_threshold": 0.2,
  "filter_band": "none"
}

Example payload¶

{
  "station_id": "YY.STA3",
  "channel_ids_used": [
    "YY.STA3..HHZ",
    "YY.STA3..HHN",
    "YY.STA3..HHE"
  ],
  "target_time_utc": "2024-06-15T08:12:12.100Z",
  "searched_window": {
    "from_utc": "2024-06-15T08:12:07.100Z",
    "to_utc": "2024-06-15T08:12:17.100Z"
  },
  "picks": [
    {
      "pick_index": 0,
      "phase": "P",
      "time_utc": "2024-06-15T08:12:12.900Z",
      "probability": 0.34,
      "offset_from_target_s": 0.8,
      "channel_id": "YY.STA3..HHZ"
    }
  ],
  "noise_context": {
    "measured_noise_db": -141.0,
    "baseline_p50_db": -146.3,
    "percentile_of_baseline": 88
  },
  "model": {
    "picker": "phasenet",
    "weights": "instance",
    "seisbench_version": "0.5.0"
  }
}

What this tool does not tell you¶

What this tool does not tell you: that a pick is real. If it finds nothing, that is informative. If it finds something at a low threshold, compare with the noise context before believing it — the picker produces picks on noise too.

Warnings named for this tool¶

Code	When
`station_silent`	the station is silent in the searched window
`data_gap`	the window is only partially covered
`clipped_trace`	amplitude reaches digitizer full scale
`low_confidence_fit`	a fitted quantity has poor statistics
`budget_low`	at most one call remains in the tool budget

Errors named for this tool¶

Code	When
`E_DATA_UNAVAILABLE`	`reason` is `no_waveform` or `station_absent`