postprocessing_seismo_lib

postprocessing_seismo_lib is a lightweight Python library for building and parsing structured API messages, especially for use with nested JSON structures used in event-based data systems. It also converts seismic pick and detection data between the SOA/SCSN JSON format and the ANSS format used by anss-formats.

This library is vetted and works against python 3.10.5. This library does not work with any python libraries below 3.10 (it has been specifically vetted against python 3.6.5 and python 3.8.10 and found not to work).

Table of Contents

• Features
• Use cases of this library
• Installation
• Core Utilities
  • Extraction of body
  • Creation of the request for a body object
  • Creation of full response format
• Pick / Detection Format Conversions
  • SOA vs ANSS field mapping (anss-formats==0.1.3)
  • Core Pick conversion functions
  • How to Run the Conversion Examples
  • SOA-envelope → ANSS Pick / Detection conversions (new)
  • AQMS / pick-filter-spec Detection JSON writer (new)
• Notes

Features

• Extract the body section from a structured JSON file using extract_body_from_file
• Create request for a body object using wrap_data, with provided associator or pickfilter files.
• Validates if input and output formats are to specification using wrap_data
• Builds a full message with status, headers and body using convert_file_to_json, with provided csv, arcout or quakeml files
• Converts seismic Pick and Detection data between the SOA/SCSN JSON format and the ANSS format (anss-formats==0.1.3), from several different upstream sources (a raw SOA pick list, the Phasenet lambda's response, pick-filter-spec's PickProcess/v2 response, or an AQMS-db / pandas-DataFrame pipeline)

This library also provides utilities for converting between different seismic pick and detection data formats, including:

• SOA → SOA (ANSS-informed)
• SOA → ANSS (Pick)
• ANSS → SOA (Pick)
• ANSS → SCSN (Detection)
• PhaseNet CSV → SOA
• Phasenet-lambda SOA → ANSS (Pick)
• pick-filter-spec SOA → ANSS (Pick)
• pick-filter-spec SOA → ANSS (Detection)
• AQMS db rows / pandas DataFrame → ANSS (Detection)

These functions are useful for normalizing pick and detection data across different pipelines and tools. See Pick / Detection Format Conversions for details.

Use cases of this library

1. Individual users
2. Pipeline scripts

Installation

pip install postprocessing-seismo-lib
OR
pip install --upgrade postprocessing-seismo-lib

This installs anss-formats==0.1.3 as a dependency (declared in setup.py). If you have code written against the older anss-formats==0.1.0-shaped Pick objects (pickerType, qualityInfo, machineLearningInfo, anssformats.association.Association), see SOA vs ANSS field mapping below -- those classes/fields were removed in 0.1.3.

After installation, we have provided sample files that can be vetted against the library's core functions. Run the below script to analyze the contents of each file, and see if the outputs are generated locally:

import json, importlib.resources
from postprocessing_seismo_lib import wrap_data, extract_body_from_file, convert_file_to_json

pick_file = json.load(importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('xxxx_file_containing_picks.json').open('r'))
print("Picks")
print(pick_file)

filtered_pick_file = json.load(importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('xxxx_file_containing_filtered_picks.json').open('r'))
print("Filtered picks")
print(pick_file)


json_path = importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('40584759_csv.json')
print("JSON file with body")
print(json_path)


#FIRST USE CASE: Extract body from a JSON file
body_data = extract_body_from_file(str(json_path))
print("Body extracted:")
print(body_data)


#SECOND USE CASE: Create RetrieveParameter wrapping around input data for various modules

## FOR THE ASSOCIATOR MODULE:

input_path = importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('xxxx_file_containing_filtered_picks.json')

wrap_data(
    input_file_path=str(input_path),
    output_file_path='output_associator.json',
    evid='evid_filtered_picks',
    module='associator'
)

## FOR THE PICK FILTER MODULE:

### THE BELOW SCENARIO filters picks under the default conditions:
[1] mode='hypoPN'
[2] testType='local'
[3] logging='False'

input_path = importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('xxxx_file_containing_picks.json')

wrap_data(
    input_file_path=str(input_path),
    output_file_path='output_pickfilter.json',
    evid='evid_picks',
    module='pickfilter'
)

### THE BELOW SCENARIO shows that we can adjust those conditions within the pickfilter:
[1] mode='st-proc'
[2] testType='local'
[3] logging='True'

wrap_data(
    input_file_path = [YOUR INPUT FILE PATH],
    output_file_path = [YOUR OUTPUT FILE PATH],
    module = 'pickfilter',
    evid = '[NAME OF EVID USED]',
    mode = 'st-proc',
    testType = 'local',
    logging = 'True'
)

#THIRD USE CASE: Create Response wrapping around known data
gamma_events = importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('40584759_gamma_events.csv')
gamma_picks = importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('40584759_gamma_picks.csv')

xml_file_nosignifier = importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('40584759_events_testGOUA')
xml_file_signifier = importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('40584759_events_test.xml')
arcout_file = importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('z78966423_api_stproc_9999.arcout')

print("GaMMa events")
print(gamma_events)
print(type(gamma_events))

print("GaMMa picks")
print(gamma_picks)
print(type(gamma_picks))

# For CSV
convert_file_to_json(
    input_file="",  # not used for CSV
    output_file="response_csv.json",
    id="id_testing",
    event_file=str(gamma_events),
    pick_file=str(gamma_picks),
    error_log_file="csv_error_log.txt"
)

# For QuakeML XML (this input file has no XML signifiers but was parsed successfully as XML here)
convert_file_to_json(
    input_file=str(xml_file_nosignifier),
    output_file="response_quakeml_nosignifiers.json",
    id="id testing",
    error_log_file="quakeml_error_log_one.txt"
)

#Conventional QuakeML XML here
convert_file_to_json(
    input_file=str(xml_file_signifier),
    output_file="response_quakeml_signifiers.json",
    id="id testing",
    error_log_file="quakeml_error_log_two.txt"
)


# For ArcOut
convert_file_to_json(
    input_file=str(arcout_file),
    output_file="response_arcout.json",
    id="id testing",
    error_log_file="arcout_error_log.txt"
)

Core Utilities

Extraction of body

The below function allows for extracting out the body from an output response file:

from postprocessing_seismo_lib import extract_body_from_file

body_data = extract_body_from_file("output_response_association.json")
body_data = extract_body_from_file("output_response_pickfilter.json")

where as an example, output_response_association.json is:

{
  "status": 404,
  "headers": {
    "Content-Type": "application/json"
  },
  "body": {
    "id": "78604159",
    "format": "none.noeventsfound",
    "data": []
  }
}

Creation of the request for a body object

The below function creates the request from the body object, which can be extracted from the above function. All four variables listed below need to be specified:

from postprocessing_seismo_lib import wrap_data

#creating the request for the associator input
wrap_data(
    input_file_path='[xxxx_file_containing_filtered_picks].json',
    output_file_path='output_associator.json',
    evid='[Name of choice]',
    module='associator'
)


#creating the request for the pickfilter input

## Pickfilter default settings:
[1] mode='hypoPN'
[2] testType='local'
[3] logging='False'

wrap_data(
    input_file_path='[xxxx_file_containing_picks].json',
    output_file_path='output_pickfilter.json',
    evid='[Name of choice]',
    module='pickfilter'
)

### Pickfilter, adjusting various settings:
[1] mode='st-proc'
[2] testType='local'
[3] logging='True'

wrap_data(
    input_file_path = [YOUR INPUT FILE PATH],
    output_file_path = [YOUR OUTPUT FILE PATH],
    module = 'pickfilter',
    evid = '[NAME OF EVID USED]',
    mode = 'st-proc',
    testType = 'local',
    logging = 'True'
)

The request format will be different across each module. Currently, the module takes in 'associator' and 'pickfilter' but this will be expanded in future updates.

Specifically, this function reads a list of pick dictionaries from a JSON file specified by input_file_path, validates them against a schema, wraps the data into a module-specific JSON structure, validates the output, and writes it to a new file specified by output_file_path. Any errors are logged to a file named wrap_data_errors.log.

wrap_data's input schema does not require every SOA pick key to be present -- for example, Onset (present on pick-filter output, generally absent on raw Phasenet output) is accepted either way, and undeclared extra keys are never rejected. See How to Run the Conversion Examples below for a worked example using a real pick-filter response (which does carry Onset).

As an example, our input_file_path='[xxxx_file_containing_picks].json' might look like this (as a list of dictionaries):

[
    {
        "Amplitude": {
            "Amplitude": 1039.6302490234,
            "SNR": 11.074
        },
        "Filter": [
            {
                "HighPass": 1.0,
                "Type": "HighPass"
            }
        ],
        "Onset": "emergent",
        "Phase": "S",
        "Picker": "deep-learning",
        "Polarity": "no-result",
        "Quality": [
            {
                "Standard": "PhaseNet",
                "Value": 0.851
            },
            {
                "Standard": "hypoinverse",
                "Value": 2
            }
        ],
        "Site": {
            "Channel": "HHE",
            "Location": "",
            "Network": "CI",
            "Station": "WOR"
        },
        "Source": {
            "AgencyID": "CI",
            "Author": "hypoPN"
        },
        "Time": "2025-04-22T21:51:15.148Z",
        "Type": "Pick"
    },
    {
        ...
    }
]

and its output would be the necessary format to POST into the associator API endpoint:

{
  "RetrieveParameters": {
    "pickFile": "Ryan_testingAgainPicks_picks.json",
    "pickDataStr": [
      {
        "Amplitude": {
          "Amplitude": 1039.6302490234,
          "SNR": 11.074
        },
        "Filter": [
          {
            "HighPass": 1.0,
            "Type": "HighPass"
          }
        ],
        "Onset": "emergent",
        "Phase": "S",
        "Picker": "deep-learning",
        "Polarity": "no-result",
        "Quality": [
          {
            "Standard": "PhaseNet",
            "Value": 0.851
          },
          {
            "Standard": "hypoinverse",
            "Value": 2
          }
        ],
        "Site": {
          "Channel": "HHE",
          "Location": "",
          "Network": "CI",
          "Station": "WOR"
        },
        "Source": {
          "AgencyID": "CI",
          "Author": "hypoPN"
        },
        "Time": "2025-04-22T21:51:15.148Z",
        "Type": "Pick"
      },
      ...
    ]
  }
}

Creation of full response format

Below shows how to build out the Response format for provided files. In all cases below, you provide an ID and an output file name (of type json). Also, provide the error log file, in case any errors occur. If any errors exist, a file of the name you specified will be generated. If no errors exist, the output JSON file will be generated at the path where you run the python script.

If you are converting from csv to json, you provide the _events.csv and _picks.csv that are generated from pinging the associator API, and set them to event_file and pick_file. Leave the input_file blank. For quakeML or arcout conversion to json, specify the input_file.

from postprocessing_seismo_lib import convert_file_to_json

# For CSV
convert_file_to_json(
    input_file="",  # not used for CSV
    output_file="[Output file name].json",
    id="[Name of choice]",
    event_file="[xxxx]_gamma_events.csv",
    pick_file="[xxxx]_gamma_picks.csv",
    error_log_file="csv_error_log.txt"
)

# For QuakeML XML (this input file has no XML signifiers but was parsed successfully as XML here)
convert_file_to_json(
    input_file="[xxxx]_events_test",
    output_file="[xxxx]_quakeml.json",
    id="[Name of choice]",
    error_log_file="quakeml_error_log.txt"
)

#Conventional QuakeML XML here
convert_file_to_json(
    input_file="[xxxx]_events_test.xml",
    output_file="[xxxx]_quakeml.json",
    id="[Name of choice]",
    error_log_file="quakeml_error_log.txt"
)


# For ArcOut
convert_file_to_json(
    input_file="[xxxx]_api_stproc_9999.arcout",
    output_file="[Output file name].json",
    id="[Name of choice]",
    error_log_file="arcout_error_log.txt"
)

Pick / Detection Format Conversions

This library also allows for conversion between SOA and ANSS Pick and Detection formats, using anss-formats==0.1.3.

Note that whether you are converting to the ANSS or SOA format, the following keys: Polarity, Onset, Picker, Filter, Amplitude and Quality are all optional, with Type, Site, Source, Time and Phase mandatory.

SOA vs ANSS field mapping (anss-formats==0.1.3)

anss-formats==0.1.3 changed the Pick schema significantly from the 0.1.0 version this library previously targeted. If you have existing code or data built against 0.1.0, note the following:

Because Pick.method is now a free string, soa_to_anss_pick_format no longer needs to coerce unrecognized Picker values (like deep-learning or machine-learning) down to "other" -- they pass straight through.

The SOA Pick format itself (Site/Source/Time/Phase/Polarity/Picker/Onset/Filter/Amplitude/Quality, capitalized keys) has not changed.

Core Pick conversion functions

from postprocessing_seismo_lib import (
    anss_to_soa_pick_format,
    soa_to_anss_pick_format,
    phasenet_csv_to_soa_pick_format,
    soa_to_soa_pick_format_using_anss_libraries
)

soa_to_anss_pick_format(pick_file_path, station_csv_path=None) looks up each pick's station coordinates by (Network, Station, Channel). station_csv_path is optional: if omitted, the archive_stations_loc_dates.csv bundled with this library under postprocessing_seismo_lib/example_data/ is used as the default station list; pass your own CSV (same NET,STA,CHA,LOC,LAT,LON,ELEV,ONDATE,OFFDATE columns) to override it.

How to Run the Conversion Examples

The library includes example datasets under postprocessing_seismo_lib.example_data. You can use these to test each conversion workflow.

Start by importing dependencies:

import json
import importlib.resources
import pandas as pd

Example 1: SOA → SOA (ANSS-Informed)

Enhances SOA picks using ANSS-informed logic.

pick_file_one_contents = json.load(
    importlib.resources.files('postprocessing_seismo_lib.example_data')
    .joinpath('79765767_picks.json')
    .open('r')
)

print("Inspecting the pick contents, example one")
print(pick_file_one_contents)

pick_file_two_contents = json.load(
    importlib.resources.files('postprocessing_seismo_lib.example_data')
    .joinpath('60209491_picks.json')
    .open('r')
)

print("Inspecting the pick contents, example two")
print(pick_file_two_contents)

pick_file_one = importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('79765767_picks.json')

print("Path for the pick file, example one")
print(pick_file_one)

pick_file_two = importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('60209491_picks.json')

print("Path for the pick file, example one")
print(pick_file_two)

soa_picks_informed_by_anss = soa_to_soa_pick_format_using_anss_libraries(pick_file_one)

print("SOA picks (ANSS-informed):")
print(soa_picks_informed_by_anss)

Example 2: SOA → ANSS

Converts SOA-formatted picks into ANSS format using station metadata.

archive_stations = importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('archive_stations_loc_dates.csv')

pick_file_two = importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('60209491_picks.json')

print("Path for the archive station file")
print(archive_stations)

df = pd.read_csv(archive_stations)
print("Archive station contents:")
print(df)

formatted_anss_picks = soa_to_anss_pick_format(pick_file_two, archive_stations)

print("ANSS picks, from SOA")
print(formatted_anss_picks)

# List of picks in ANSS format
picks_anss = formatted_anss_picks["picks"]

string_dict_picks = json.dumps(picks_anss, indent=4, default=str)

# saved file
with open("picks_ANSS_from_library.json", "w") as f:
    f.write(string_dict_picks)

Example 3: PhaseNet CSV → SOA

Converts PhaseNet TensorFlow CSV output into SOA pick format.

picks_phasenet_tensorflow = importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('picks_phasenet_tensorflow.csv')

print("Path for the picks_phasenet_tensorflow file")
print(picks_phasenet_tensorflow)

df = pd.read_csv(picks_phasenet_tensorflow)
print("Phasenet picks, tensorflow, contents:")
print(df)

highpass_filt = 1.0

pick_list = phasenet_csv_to_soa_pick_format(picks_phasenet_tensorflow, highpass_filt)

print("SOA picks, from Phasenet CSV")
print(pick_list)

Example 4: ANSS → SOA

Converts ANSS-formatted picks (anss-formats==0.1.3 shape) into SOA format.

anss_file_contents = json.load(importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('picks_ANSS.json').open('r'))
print("Inspecting the pick contents, ANSS format")
print(anss_file_contents)

anss_file_path = importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('picks_ANSS.json')
print("Path for the pick file, ANSS format")
print(anss_file_path)

soa_picks=anss_to_soa_pick_format(anss_file_path)

# Access list of picks
print("SOA picks, from ANSS")
picks_soa=soa_picks["picks"]
print(picks_soa)

SOA-envelope → ANSS Pick / Detection conversions (new)

These three functions consume the SOA-format JSON response envelopes produced further upstream in the pipeline (the Phasenet lambda, or pick-filter-spec's PickProcess/v2 endpoint) and convert them directly into anss-formats==0.1.3 Pick/Detection objects. All accept either a file path to the full {"status", "headers", "body"} envelope, a path to just the bare body, or an already-loaded dict/list -- and all accept an optional station_csv_path, defaulting to the bundled archive_stations_loc_dates.csv if not given.

from postprocessing_seismo_lib import (
    phasenet_soa_to_anss_pick_format,
    pickfilter_soa_to_anss_pick_format,
    pickfilter_soa_to_anss_detection_format,
)

Phasenet lambda SOA pick list → ANSS Pick

phasenet_soa_to_anss_pick_format(soa_source, station_csv_path=None) converts the SOA picks returned by the Phasenet lambda into a list of anss-formats==0.1.3 Pick dicts.

phasenet_soa_response = importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('picks_SOA_fromPhasenetLambda.json')

formatted = phasenet_soa_to_anss_pick_format(phasenet_soa_response)

print("ANSS picks, from Phasenet-lambda SOA response")
print(formatted["picks"])

Example input (picks_SOA_fromPhasenetLambda.json, one entry from the "body" list):

{
    "Type": "Pick",
    "Site": {"Station": "TKX", "Channel": "HNZ", "Network": "BC", "Location": ""},
    "Time": "2026-02-24T17:47:47.410Z",
    "Source": {"AgencyID": "BC", "Author": "hypoPN"},
    "Phase": "P",
    "Polarity": "up",
    "Picker": "deep-learning",
    "Filter": [{"Type": "HighPass", "HighPass": 1.0}],
    "Amplitude": {"Amplitude": 3.380613327026367, "SNR": 10.485},
    "Quality": [{"Standard": "PhaseNet", "Value": 0.906}]
}

pick-filter-spec SOA pick list → ANSS Pick

pickfilter_soa_to_anss_pick_format(pickfilter_source, station_csv_path=None) converts the SOA picks returned by pick-filter-spec's PickProcess/v2 endpoint (a "localpick"-style request) into a list of anss-formats==0.1.3 Pick dicts. Unlike a raw Phasenet pick, a pick-filter pick typically carries Onset and a two-entry Quality list (a PhaseNet probability and a hypoinverse pick-weight code 0-4) -- both are preserved (Onset -> Pick.onset, Quality -> Pick.analyticsInfo.predictions, one Prediction per Quality entry).

pickfilter_picks_response = importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('61146748_filteredpicks_picksToCONVSPECFORMAT_PICKK.json')

formatted = pickfilter_soa_to_anss_pick_format(pickfilter_picks_response)

print("ANSS picks, from pick-filter-spec SOA response")
print(formatted["picks"])

postprocessing_seismo_lib/example_data/61146748_picksSPECFORMAT_ANSS_from_library_0.1.3.json is the expected output for this fixture (bundled alongside the input so you can diff your own run against it -- every field matches except each pick's randomly-generated id).

pick-filter-spec SOA detection → ANSS Detection

pickfilter_soa_to_anss_detection_format(detection_source, station_csv_path=None) converts the SOA detection returned by pick-filter-spec's PickProcess/v2 endpoint (a "localdetection"-style request) into a single anss-formats==0.1.3 Detection dict.

pickfilter_detection_response = importlib.resources.files('postprocessing_seismo_lib.example_data').joinpath('61146748_filteredpicks_detectionSPECFORMAT_TEST_2026CHECK_DETECT.json')

formatted = pickfilter_soa_to_anss_detection_format(pickfilter_detection_response)

print("ANSS detection, from pick-filter-spec SOA response")
print(formatted["detection"])

postprocessing_seismo_lib/example_data/61146748_detectionSPECFORMAT_ANSS_from_library_0.1.3.json is the expected output for this fixture (again, matches except for pick ids).

AQMS / pick-filter-spec Detection JSON writer (new)

A separate module, postprocessing_seismo_lib.json_writers, builds anss-formats==0.1.3 Detection JSON directly from two other, distinct data sources -- not the SOA JSON envelopes used by the functions above:

• AQMS database rows, passed as (event, origin, netmag, arrivals) dicts.
• a pandas DataFrame of already-processed picks plus an evt dict, as produced by pick-filter-spec's pick_processing.py.

from postprocessing_seismo_lib import dumps_detection_anss, dump_detection_anss, map_arrival_quality_to_score, detection_anss_to_scsn

dumps_detection_anss(*data, agencyID=None, author=None) -> dict builds and returns the ANSS detection JSON. dump_detection_anss(*data, fp, agencyID=None, author=None, **json_kwargs) does the same and writes it to a file-like object fp (mirroring json.dump). map_arrival_quality_to_score(quality, channel) maps an ML confidence score to a 0-4 Hypoinverse pick-weight code, for callers building AQMS-style arrivals.

import pandas as pd
from datetime import datetime
from types import SimpleNamespace

picks_df = pd.DataFrame([
    {
        "id": "p1",
        "Phase": "P",
        "Time": datetime(2026, 7, 19, 12, 0, 0),
        "Site": {"Station": "ABC", "Channel": "HHZ", "Network": "CI", "Location": ""},
        "Source": {"AgencyID": "CI", "Author": "hypoPN"},
        "Picker": "deep-learning",
        "Onset": "impulsive",
        "Polarity": "up",
        "AssociationInfo": {"Distance": 0.5, "Residual": -0.02},
        "Quality": [{"Standard": "hypoinverse", "Value": 1}],
    },
])

evt = {
    "id": 99999,
    "loc": SimpleNamespace(longitude=-118.0, latitude=34.0, z=8.0),
    "time": datetime(2026, 7, 19, 11, 59, 55),
    "mag": {"Type": "l", "Value": 2.5},
}

anss_detection = dumps_detection_anss(picks_df, evt, agencyID="CI", author="hypoPN")
print(anss_detection)

The reverse direction, detection_anss_to_scsn(detection) -> dict (importable from postprocessing_seismo_lib, defined in utils.py), converts an ANSS Detection dict back into the legacy SCSN detection JSON format (Type/ID/Source/Hypocenter/Data/Magnitude, capitalized keys):

scsn_detection = detection_anss_to_scsn(anss_detection)
print(scsn_detection)

These two functions were ported from a separate contribution (pick-filter-spec / postprocessing-library commit 8cdb808, "writer for anss detection json initial") and are kept in their own json_writers module specifically because their input contract (AQMS rows / DataFrame) differs from the SOA-JSON-envelope converters documented above.

Notes

• All Pick conversion functions return a dictionary containing a "picks" key (a list); the Detection conversion function (pickfilter_soa_to_anss_detection_format) returns a dictionary containing a "detection" key (a single dict).
• You can directly manipulate or save this list/dict depending on your workflow.
• Ensure input files match expected formats (JSON or CSV as required).
• These utilities are especially useful for integrating:
  • PhaseNet outputs
  • ANSS datasets
  • SOA-based processing pipelines
  • pick-filter-spec / AQMS detection pipelines