Index Trips Whose Track Duplicates Another Trip

Flags trips whose GPS track is identical to another trip's, so the same fishing event is not counted once per trip identifier.

Usage

index_trip_duplicates(df, prefer = character(0), seen = NULL)

Arguments

df

Prepared fishing points with one row per ping and columns trip, timestamp, latitude, longitude.

prefer

Character vector of trip identifiers to keep in preference to others – normally the trips PDS still lists, which are the ones carrying device metadata. Defaults to none.

seen

Optional data frame of trips aggregated in earlier runs, with columns trip and fingerprint.

Value

A tibble with one row per trip in df: trip, fingerprint, n_points, duplicate_of (the surviving trip, NA when the trip is itself the survivor) and keep.

Details

Pelagic Data Systems re-segments trips retroactively. A trip identifier can be retired (its points reappear under a new identifier) or reassigned (its points move to a new trip while the old identifier comes to describe a different window). Since predict_pds_tracks() writes one file per trip, the predicted-track store ends up holding several identifiers that describe the very same track, and aggregate_pds_effort() counts each of them as a separate vessel: fishing hours, unique trips and pings are multiplied by the number of copies, while n_active_days (a set union of dates) is not, so the per-day rates in add_cell_effort_metrics() explode.

Each trip is reduced to a fingerprint of its (timestamp, latitude, longitude) triples, order-independent so that two files listing the same points always agree. Trips sharing a fingerprint – with each other, or with a trip aggregated in an earlier run (seen) – collapse to one survivor: a trip listed in prefer wins, then the trip with the most points, then the lowest identifier, so the choice is stable across runs.

A trip already in the effort store always outranks a new duplicate of it, so that a run cannot move effort between trips that are known to be the same. When a retired identifier is aggregated before the trip that supersedes it, the retired one therefore survives and its cells keep gear/country = "unknown"; the deletion pass in predict_pds_tracks() is what stops that happening.

See also

drop_overlapping_pings(), aggregate_pds_effort()