NTSB Aviation Accident Reports and Docket PDF Scraper

Scrape aviation accident and incident investigations from the NTSB CAROL database. Each investigation comes back as a structured record with the probable cause, findings, narratives, aircraft, injuries and the text of the investigation report PDF. Every document in the Public Docket (factual reports, witness statements, ATC records, maintenance records) comes back as its own record with the PDF text extracted. Coverage runs from 1962 to the latest events.

---

NTSB Aviation Accident Scraper Features

• Pull recent events (last 90 days), a whole year range, or specific NTSB numbers such as ENG22LA049
• Probable cause statement, finding statements, and the factual and analysis narratives as clean text
• Aircraft registration, make, model, serial number, engine type and damage level for every aircraft involved
• Operator, regulation the flight was flown under (Part 91, 121, 135 and others), flight phase and the defining event
• Fatal, serious, minor and uninjured counts
• Full text of the newest investigation report PDF, preliminary or final
• Every Public Docket attachment as its own row, with page count, title, type, download link and extracted text
• OCR fallback for scanned PDFs such as logbooks, ATC strips and handwritten statements
• Docket PDF links collected on each event record, so you can fetch the originals yourself

---

Who Uses NTSB Accident Data?

• Aviation insurance underwriters price operators and aircraft types from probable causes and findings across years of accidents
• Aviation litigators pull the factual reports, witness statements and maintenance records filed in a case docket
• Safety and risk teams benchmark their own incident rates against NTSB trends by flight phase, regulation part or aircraft model
• Fleet and maintenance analysts search findings for component failures on a specific make and model
• Aviation journalists get the text of a probable cause or docket document without paging through the site

---

How the NTSB Scraper Works

1. Choose an event source: the last 90 days, a year range, or a list of NTSB numbers.
2. The scraper lists every matching aviation event, newest first.
3. For each event it collects the case details and, when you ask for it, extracts the text of the report PDF.
4. When a Public Docket exists, it lists the attachments and extracts the text of each PDF.
5. You get one event row per investigation and one docket_document row per attachment.

Rows from events and from docket attachments both count toward maxItems. A docket with 40 attachments uses 41 rows.

---

Input

{
  "eventMode": "year_range",
  "yearFrom": 2022,
  "yearTo": 2023,
  "includeReportPdfText": true,
  "includeDocketPdfs": true,
  "ocrFallback": true,
  "maxItems": 100
}

Field	Type	Default	Description
eventMode	string	recent	recent (last 90 days), year_range, or event_ids.
yearFrom	integer	2024	First year when eventMode is year_range. Coverage starts in 1962.
yearTo	integer	current year	Last year when eventMode is year_range.
eventIds	array	[]	NTSB numbers such as ENG22LA049, used when eventMode is event_ids.
includeReportPdfText	boolean	true	Extract the text of the newest investigation report PDF.
includeDocketPdfs	boolean	true	Return every Public Docket attachment as its own row with extracted text.
ocrFallback	boolean	true	OCR image-only PDFs. Limited to the first 30 pages of each file.
maxItems	integer	20	Maximum rows returned. Event rows and docket attachment rows both count.

Look up one known investigation:

{
  "eventMode": "event_ids",
  "eventIds": ["ENG22LA049"],
  "maxItems": 10
}

Events only, no PDF text, for a fast trend dataset:

{
  "eventMode": "year_range",
  "yearFrom": 2015,
  "yearTo": 2025,
  "includeReportPdfText": false,
  "includeDocketPdfs": false,
  "maxItems": 5000
}

---

NTSB Aviation Accident Scraper Output Fields

An event row:

{
  "record_type": "event",
  "event_id": "ENG22LA049",
  "case_key": "106762",
  "event_type": "INC",
  "event_date": "2022-03-31",
  "event_datetime": "2022-03-31T19:05:00Z",
  "location": { "city": "Riyadh", "state": null, "country": "SA", "latitude": 24.715122, "longitude": 46.658257, "airport_id": null, "airport_name": null },
  "aircraft": [{ "registration": "9H-VIG", "make": "BOMBARDIER", "model": "Global 7500", "serial_number": "70086", "category": "AIR", "engine_count": 2, "engine_type": "TF", "damage": "Minor", "amateur_built": false }],
  "operator": "VistaJet Ltd",
  "far_part": "NUSC",
  "flight_phase": "Initial Climb",
  "defining_event": "Fire/smoke (non-impact)",
  "injury_summary": { "fatal": null, "serious": null, "minor": null, "none": null },
  "probable_cause_text": "An in-flight engine fire resulted from a fuel leak from fuel nozzle No. 18 pigtail-to-fuel manifold b-nut connection...",
  "findings": ["Aircraft Fuel distribution - Design", "Personnel issues Installation - Maintenance personnel"],
  "has_safety_recommendations": false,
  "investigation_status": "Completed",
  "latest_report_type": "Final",
  "report_pdf_url": "https://data.ntsb.gov/carol-repgen/api/Aviation/ReportMain/GenerateNewestReport/106762/pdf",
  "report_text": "Aviation Investigation Final Report ...",
  "report_pages": 12,
  "docket_url": "https://data.ntsb.gov/Docket?ProjectID=106762",
  "docket_document_count": 2,
  "public_docket_pdf_urls": ["https://data.ntsb.gov/Docket/Document/docBLOB?ID=15334415&FileExtension=pdf&FileName=GE%20Passport%20Fire%20-%20Saudi%20Arabia-Rel.pdf"]
}

A docket_document row:

{
  "record_type": "docket_document",
  "event_id": "ENG22LA049",
  "case_key": "106762",
  "event_date": "2022-03-31",
  "document_number": 1,
  "document_title": "POWERPLANTS GROUP CHAIRMAN'S FACTUAL REPORT",
  "document_type": "Text/Image",
  "document_pages": 43,
  "document_url": "https://data.ntsb.gov/Docket/Document/docBLOB?ID=15334415&FileExtension=pdf&FileName=GE%20Passport%20Fire%20-%20Saudi%20Arabia-Rel.pdf",
  "document_text": "NATIONAL TRANSPORTATION SAFETY BOARD Office of Aviation Safety ...",
  "text_source": "pdf_text",
  "text_truncated": false,
  "docket_url": "https://data.ntsb.gov/Docket?ProjectID=106762"
}

Field	Type	Description
record_type	string	event or docket_document
event_id	string	NTSB event number
case_key	string	NTSB internal case key used by CAROL and the Public Docket
event_type	string	ACC accident or INC incident
event_date	string	Event date, YYYY-MM-DD
event_datetime	string	Event date and time as published (UTC)
location	object	City, state, country, latitude, longitude, airport id and name
aircraft	array	One entry per aircraft: registration, make, model, serial number, category, engines, damage
operator	string	Operator name when published
far_part	string	Regulation the flight was conducted under, for example 091, 121, 135
flight_phase	string	Phase of flight of the defining event
defining_event	string	First occurrence in the event sequence
injury_summary	object	Fatal, serious, minor and uninjured counts
probable_cause_text	string	Probable cause statement, once the investigation reaches its final determination
findings	array	Finding statements
factual_narrative	string	Factual narrative from the investigation record
analysis_narrative	string	Analysis narrative from the investigation record
has_safety_recommendations	boolean	Whether the NTSB issued safety recommendations
investigation_status	string	In work or Completed
latest_report_type	string	Prelim or Final
report_pdf_url	string	Newest investigation report PDF
report_text	string	Extracted text of that PDF
report_pages	integer	Page count of that PDF
docket_url	string	Public Docket page for the event
docket_document_count	integer	Number of docket attachments
public_docket_pdf_urls	array	Links to every PDF attachment in the docket
document_number	integer	Position of the attachment in the docket (docket_document rows)
document_title	string	Attachment title
document_type	string	Attachment type as listed in the docket
document_pages	integer	Page count listed in the docket
document_url	string	Download link of the attachment
document_text	string	Extracted attachment text
text_source	string	pdf_text, ocr, too_large (file over 25 MB, not parsed), or none when nothing could be extracted
text_truncated	boolean	true when the text was cut at 900,000 characters
source_url	string	Page the record was taken from

Newly opened investigations have only the basics: aircraft, location, injuries and the opening report. The probable cause, findings and narratives appear once the NTSB publishes its final determination. Docket attachments that are not PDFs (photos, spreadsheets) are listed with their link and no text. Foreign-state investigations often have fewer fields than NTSB-led ones.

---

FAQ

How do I scrape NTSB aviation accident reports?

Choose an event source and run the actor. No NTSB account is needed. Use event_ids when you already know the event numbers, year_range for a historical pull, or recent for the latest 90 days.

What is in the NTSB Public Docket?

Factual reports from each investigative group, witness interview summaries, ATC records, maintenance and training records, photos and test reports. Larger accidents carry dozens of attachments. This actor returns each one as a row.

Does it read scanned PDFs?

Yes. Text PDFs are read directly. Image-only PDFs go through OCR when ocrFallback is on, up to 30 pages per file. A file that cannot be read, or is over 25 MB, comes back with text_source of none or too_large and its link, so you can still download the original.

How far back does the data go?

Coverage starts in 1962. Older records carry fewer fields than recent ones, and most have no docket.

Why did my run return fewer rows than maxItems?

maxItems is a ceiling. A run returns fewer rows when fewer events match your filters, or when docket attachments are switched off.

---

Need More Features?

Need custom fields, filters, or other NTSB transportation modes? File an issue or get in touch.

Why Use the NTSB Aviation Accident Scraper?

• Reports and docket in one run - the investigation record, the report text and every docket attachment, not just the summary page
• Clean, structured fields - probable cause, findings and aircraft come back as fields and arrays, ready for a database
• Full history - from the latest event back to 1962, by year range or by event number