Flags locations that imply fast round trips
flag_roundtrips(x, delta, epsilon, time_unit = "secs", ...)# S3 method for track_xyt
flag_roundtrips(x, delta, epsilon, time_unit = "secs", ...)
Returns x (a track_xyt) with a flagging column added
(x$fast_roundtrip_).
[track_xyt] A track_xyt object.
[numeric] The threshold SDR for flagging. Locations that imply
both legs of a round trip with SDR > delta/epsilon are flagged. See details.
[numeric] A scaling factor to create the threshold for
flagging.
[character] Character string giving time unit. Should be
"secs", "mins", or "hours". See details.
Addtional arguments. None currently implemented.
Brian J. Smith, based on code by Johannes Signer and Tal Avgar
Locations implying a single fast step can be flagged using
flag_fast_steps(). However, it is more likely that a single
location is imprecise if it implies an unrealistically fast out-and-back round
trip. In that case, the user might be willing to scale the threshold SDR.
In this function, delta gives the base SDR and epsilon is the scaling
factor, such that locations are considered for flagging if the SDR from the
previous location (location i - 1) to the focal location (i) [call it sdr1]
and the focal location (i) to the next location (i + 1) [call it sdr2] both
have SDR > delta/epsilon.
In that case, the SDR from the previous location (i - 1) to the next location
(i + 1) is computed; i.e., the SDR assuming we omit the focal location (i)
[call it sdr3]. The remaining locations should be closer together than
they are to the omitted location. Thus the focal location is flagged if
(sdr1 > epsilon * sdr3) & (sdr2 > epsilon * sdr3).
Note that epsilon both decreases delta in the out-and-back case and
increases sdr3 (between the remaining neighbors).
Internally, flagged locations are dropped from future consideration.
The time_unit should be the same time unit with which the SDR threshold
was calculated. SDR is typically calculated in m^2/s, so time_unit defaults
to "secs". The spatial unit is determined by the CRS, which should typically
be in meters. The epsilon parameter is unitless.
flag_duplicates(),
flag_fast_steps(),
flag_defunct_clusters()