Corridor detection

mt_corridor() identifies segments of a track where the animal was moving quickly in a spatially consistent direction. It is a port of the legacy move::corridor() function (LaPoint et al. 2013) to the modern move2 / sf stack — the concept is unchanged.

When to use corridor detection

The typical context is a migratory or commuting animal that travels between two (or several) regions via recognisable routes. Within those routes, successive segments point the same way and are traversed at speed; elsewhere on the track the animal forages or rests at slow, directionally inconsistent speeds. Corridor detection asks: which segments of the track belong to the route, and which belong to the foraging / resting stretches at each end?

The algorithm in one paragraph

For each segment i of the track, compute its speed and azimuth. Build a search circle centred at the segment midpoint with radius equal to half the segment length. Collect all other segment midpoints inside that circle. Compute the circular variance of pseudo-azimuths ((180 + azimuth) * 2 mod 360) within the neighbourhood — the doubling is what lets two animals walking the same corridor in opposite directions count as consistent. A segment is classified as corridor if (a) its speed exceeds the global speed_quantile, (b) its neighbourhood’s circular variance is below the global circ_quantile, and (c) the neighbourhood contains at least min_segments qualifying neighbours, which must outnumber the non-qualifying ones.

Load data

The function requires geographic coordinates (EPSG:4326) so that the search radius can be computed in metres on the sphere with geosphere. Re-project if necessary.

library(move2)
library(sf)
library(move2utils)

## fisher data (already in WGS84)
fishers <- mt_read(mt_example())
fishers <- fishers[!st_is_empty(fishers), ]
leroy <- filter_track_data(fishers, .track_id = "M1")
st_crs(leroy)$input
#> [1] "EPSG:4326"

Run corridor detection on a fisher

out <- mt_corridor(leroy)
#> Found 22 corridor segments (2.4% of 918 segments).
table(out$corridor)
#> 
#>     corridor not corridor 
#>           22          897

Five columns are added to the input move2:

corridor — factor with levels "corridor" and "not corridor".
corridor_speed — per-segment speed (m/s), last row NA.
corridor_azimuth — per-segment azimuth (degrees), last row NA.
corridor_circvar — circular variance of pseudo-azimuths within the segment’s neighbourhood.
corridor_n_neighbours — number of segments found within the search radius.

A quick map

corr   <- as.character(out$corridor)

out$segments <- mt_segments(out)
plot(out$segments, col = c("firebrick","grey80")[as.factor(out$corridor)], asp = 1,
     lwd=c(2,0.7)[as.factor(out$corridor)],
    main = sprintf("Leroy (fisher): %d corridor segments of %d",
                    sum(corr == "corridor"), length(corr)))
points(out, col = c("firebrick","grey40")[as.factor(out$corridor)], pch = 16, 
       cex = c(0.6,0.3)[as.factor(out$corridor)])
legend("topright", pch = 16, col = c("grey40", "firebrick"),
       legend = c("not corridor", "corridor"), bty = "n")

Leroy’s track with the minority of segments that meet the default corridor criteria (fast, directionally consistent) highlighted in firebrick. Most of the track — slower, tightly looping foraging movement — stays grey. The corridor segments here are the sparse commuting segments between forage patches.

Tightening speed_quantile and circ_quantile makes the detector more conservative — only the fastest, most directionally consistent portions qualify.

Tuning

Argument	Default	What to tune it for
`speed_quantile`	0.75	Raise for stricter speed cutoff (more conservative). Lower to catch slower but directionally consistent travel (e.g. a plodding long-distance walker).
`circ_quantile`	0.25	Lower for tighter directional consistency. Raise if the corridor genuinely winds.
`min_segments`	2	Minimum neighbourhood support. Raise if you want only well-populated corridor stretches.