stoscan: Spatio-temporal clustering from tdbscan outpout manipulated...

In mpio-be/tdbscan: tdbscan, a dbscan implementation applied to animal tracks.

Description

Spatio-temporal clustering from tdbscan outpout manipulated by dt2Convexhull

Usage

stoscan(
  d,
  pid = "pid",
  arrival = "arrival",
  departure = "departure",
  geometry = "geometry"
)

Arguments

d	a data.frame including points which belong to clusters (tdbscan output), latitude, longitude and datetime
pid	ID unique for points that belong to one cluster
arrival	arrival datetime in polygon
departure	departure time from polygon
geometry	convex hull of points of a unique cluster

Value

data.table including pid, s_clustID = ID unique for spatial overlapping clusters, st_clustID = ID unique for spatial temporal overlapping clusters

Examples

require(data.table)
require(ggplot2)

data(zbird)
z = tdbscan(zbird, eps = 12, minPts   = 5, maxLag = 5, borderPoints = TRUE )
z = z[, clustID := factor(clustID)]

d = data.frame(zbird) %>% data.table
d = merge(z, d, by.x = 'id', by.y = 'sp.ID')

d = rbindlist(list(copy(d[, tagID := 'bird1']), copy(d[, tagID := 'bird2'])), use.names = TRUE)
d[tagID == 'bird2', x := x + 5]
d[tagID == 'bird2', x := x + 5]
d[!is.na(clustID), ID := paste0(tagID, '_', clustID)]

dp = dt2Convexhull(d, pid = 'ID', projection = '+proj=utm +zone=4 +datum=WGS84')

s = stoscan(dp)
d = merge(d, s, by.x = 'ID', by.y = 'pid', all.x = TRUE)

# plot of spatial overlapping clusters
ggplot(d, aes(x, y, color = as.character(s_clustID) ) ) +
  geom_path(aes(color = NULL), col = 'grey', size = .5) +
  geom_point( alpha = .5, size = 2)

# plot of spatial and temporal overlapping clusters
ggplot(d, aes(x, y, color = as.character(st_clustID) ) ) +
  geom_path(aes(color = NULL), col = 'grey', size = .5) +
  geom_point( alpha = .5, size = 2)

mpio-be/tdbscan documentation built on March 10, 2021, 1:20 p.m.