Data

To help you solve this year’s Data challenge, we provide the following (logical) datasets:

  1. Movements (arrival and departure) information for 2025 at 11 major European airports including
    1. taxi-out time for each movement at the airport and
    2. scheduled and actual times
    This is related to departing (PHASE = DEP) or arriving (PHASE = ARR) airport.
  2. Flight information for each movement. This is what is matched in the Network Manager (NM) records.

The movements table rows have been (left) joined with the flight table ones.

There are a total of 4,167,797 movements at the 10 airports as shown in in Table 1.

Table 1: List of airports for which movements data is provided
ICAO IATA Name
EDDF FRA Frankfurt Main
EDDM MUC Munich
EGLL LHR London Heathrow
EHAM AMS Amsterdam Schiphol
LEBL BCN Josep Tarradellas Barcelona-El Prat
LEMD MAD Adolfo Suárez Madrid–Barajas
LFPG CDG Charles de Gaulle
LIRF FCO Rome–Fiumicino Leonardo da Vinci
LTFM IST İstanbul
LSZH ZRH Zürich
NoteNote 1

We have removed Military, Head of State, and Sensitive movements. These should not have a disruptive impact on the possibility to model and predict taxi-out time.

NoteNote 2

Real world data is messy. You will find some inconsistencies between movement and flight information. We haven’t tried to reconcile them because they are probably due to some limitations of the underlying processes trying to match different sources of data.
As per all our data challenges, we do not play with synthetic/artificial datasets.

Buckets

The 2024 PRC Data Challenge website has some instructions about how to use Minio to access OpenSky buckets at

https://ansperformance.eu/study/data-challenge/dc2024/data.html#using-minio-client

You will need to generate access keys for your OSN account as shown below (thanks to JavierIOM from Discord channel):

MinIO access key generation on OSN

The Training Dataset

The training dataset is split into monthly files. Each file contains the movements for the airports of the challenge:

  • training_2025-01-01_2025-02-01.parquet (21M)
  • training_2025-02-01_2025-03-01.parquet (19M)
  • training_2025-03-01_2025-04-01.parquet (22M)
  • training_2025-04-01_2025-05-01.parquet (23M)
  • training_2025-05-01_2025-06-01.parquet (25M)
  • training_2025-06-01_2025-07-01.parquet (24M)
  • training_2025-07-01_2025-08-01.parquet (25M)
  • training_2025-08-01_2025-09-01.parquet (25M)
  • training_2025-09-01_2025-10-01.parquet (24M)
  • training_2025-10-01_2025-11-01.parquet (25M)
  • training_2025-11-01_2025-12-01.parquet (22M)
  • training_2025-12-01_2026-01-01.parquet (22M)

The Ranking Dataset

The ranking dataset contains the same columns as the training data with:

  • Movements for Jan and Jul 2026.
  • BLOCK_TIME_UTC_mvt and TAXITIME_SEC_mvt for departures (movements with PHASE = "DEP") have been blanked out.

It consists of a single file:

  • ranking.parquet (27M)

The Submitting Dataset

The submitting dataset is the template to be filled with the predicted taxi-out times, i.e. TAXITIME_SEC_mvt = MVT_TIME_UTC_mvt - BLOCK_TIME_UTC_mvt (for the departure). It only contains the columns MVT_ID_mvt and TAXITIME_SEC_mvt where MVT_ID_mvt values match those of the departures in the ranking dataset. It consists of only one file:

  • submitting.parquet (1.1M)

See Submission Instructions.

Column Description

Columns postfixed with _mvt are associated with the reporting airport.
Columns postfixed with _flt are associated with information from NM flight list.

Movement

  • MVT_ID_mvt: unique identifier for the movement record
  • FLIGHT_ID_mvt: unique identifier for the NM flight (if matched)
  • FLIGHT_mvt: flight number, i.e. what you see on your boarding pass
  • FLIGHT_RULE_mvt: which sets of regulations the flight is operated under. Possible values are:
    • I for Instrument Flight Rules (IFR)
    • V for Visual Flight Rules (VFR)
    • NA if unknown
  • ADEP_mvt: the (ICAO code of the) Aerodrome of DEParture (ADEP).
  • ADES_mvt: the (ICAO code of the) Aerodrome of DEStination (ADES).
  • PHASE_mvt: movement phase. DEP=departure, ARR=arrival.
  • MVT_TIME_UTC_mvt: (best available) movement time (takeoff if PHASE = DEP, landing if PHASE = ARR.)
  • BLOCK_TIME_UTC_mvt: block time (off-block if PHASE = DEP, in-block if PHASE = ARR.)
  • SCHED_TIME_UTC_mvt: scheduled time (of departure if PHASE = DEP, of arrival if PHASE = ARR.)
  • AIRCRAFT_TYPE_mvt: the ICAO code for the aircraft type, for example A21N for Airbus A321neo.
  • RUNWAY_mvt: the Runway (RWY) ID (of departure if PHASE = DEP, of arrival if PHASE = ARR.)
  • STAND_mvt: the stand ID (of departure if PHASE = DEP, of arrival if PHASE = ARR.)
  • TAXITIME_SEC_mvt: amount of seconds for taxi time (taxi-out if PHASE = DEP, taxi-in if PHASE = ARR.)

Flight

The following columns come from data recorded by NM.

  • LOBT_flt: last known off-block time
  • CALLSIGN_flt: the callsign of the relevant flight, e.g. BAW6VB.
  • ADEP_flt: the ICAO code of the ADEP.
  • ADES_flt: the ICAO code of the ADES.
  • ADES_FILED_flt: the ADES initially filed in the Flight Plan (FPL). If different from ADES_flt, then the flight has been diverted.
  • MARKET_SEGMENT_flt: the market segment type as defined in the Market Segment Rules, it can be:
    • “Mainline”
    • “Regional”
    • “Low-Cost”
    • “Business Aviation”
    • “All-Cargo”
    • “Charter” (Non-Scheduled)
    • “Military”
    • “Other”
    • “Not classified”
  • IOBT_flt: Initial Off-Block Time (IOBT).
  • FLIGHT_RULE_flt: which sets of regulations the flight is operated under (see FPL Item 8). Possible values are:
    • I for IFR
    • V for VFR
    • Y first IFR thereafter VFR
    • Z first VFR thereafter IFR
  • FLIGHT_TYPE_flt: flight type (see FPL Item 8). Possible values are:
    • S for scheduled air service
    • N for non-scheduled air service
    • G for general aviation
    • M for military (note: filtered out)
    • X for other than the preceding categories
  • AIRCRAFT_TYPE_flt: the ICAO code for the aircraft type, for example A30B for an Airbus A-300B2-200.
  • WK_TBL_CAT_flt: wake turbulence category (see FPL Item 9), can be:
    • L LIGHT, i.e. maximum certificated takeoff mass of 7000 kg (15_500 lbs) or less.
    • M MEDIUM, i.e. maximum certificated takeoff mass less than 136_000 kg (300_000 lbs), but more than 7_000 kg (15_500 lbs).
    • H HEAVY, i.e. maximum certificated takeoff mass of 136_000 kg (300_000 lbs) or more (except those specified as J).
    • J SUPER, presently only the AIRBUS A-380-800.
  • AIRCRAFT_OPERATOR_flt: the (anonymized) ICAO Airline Designator.
  • EOBT_1_flt: the Estimated Off-Block Time (EOBT) for FPL-based (M1) trajectory.
  • ARVT_1_flt: the arrival time for FPL-based (M1) trajectory.
  • AOBT_3_flt: the Actual Off-Block Time (AOBT) for flown (M3) trajectory.
  • ARVT_3_flt: the arrival time for flown (M3) trajectory.

List of Acronyms

ADEP: Aerodrome of DEParture

ADES: Aerodrome of DEStination

AOBT: Actual Off-Block Time

EOBT: Estimated Off-Block Time

FPL: Flight Plan

IFR: Instrument Flight Rules

IOBT: Initial Off-Block Time

NM: Network Manager

RWY: Runway

VFR: Visual Flight Rules