Determine which points are farthest from each other set of coordinates in data.table R

Question

I have data like this:

 library(data.table)
 dtorig <- data.table(x=1:100,lat=(sample(c(800:900),100))/100, lon=(sample(c(3800:3900),100))/100)

I'd like to get the location of rows in dtorig that are farthest away from one another. My attempt (which doesn't work) is below:

 ###NOT WORK
  library(geosphere)
  extremep <- chull(dtorig[, c(3,2)])
  extremepoints <- dtorig[extremep,]
  wmax <- sapply(1:NROW(extremepoints), function(i) which.max(distm(extremepoints[i, c(3,2)],fun=distHaversine)))

Answer 1

Does this help?

library(data.table)
library(geosphere)
dtorig <- data.table(x=1:100,lat=(sample(c(800:900),100))/100, lon=(sample(c(3800:3900),100))/100)
x <- distm(x = as.matrix(dtorig[, .(lon, lat)]), fun=distHaversine)
which(x == max(x), arr.ind = TRUE)
#      row col
# [1,]  50  27
# [2,]  27  50

Row 27 and 50 are the furthest apart.

Answer 2

Not using data.table or geosphere , but here is an sf alternative that illustrates finding the right distances from the dense matrix given by st_distance . Here you can see:

1. Use of set.seed so you should be able to reproduce this
2. Converting coordinate columns to sf points with st_as_sf and an appropriate crs
3. Getting distance between all rows with by_element = FALSE in st_distance
4. Removing the duplicate lower triangle of the distance matrix with lower.tri
5. Adding a rowid column so we can gather up the rest of the matrix into a single distance column
6. Sorting by distance to see the most separated points with arrange(desc()) .
7. Highlighting the most distant pair of points on a plot .

library(tidyverse)
library(sf)
set.seed(12345)
dtorig <- tibble(
  x = 1:100,
  lat = (sample(c(800:900), 100)) / 100,
  lon = (sample(c(3800:3900), 100)) / 100
)

sforig <- st_as_sf(dtorig, coords = c("lon", "lat"), crs = 4326)

distances <- sforig %>%
  st_distance(by_element = FALSE) %>%
  unclass %>% 
  `[<-`(lower.tri(., diag = TRUE), NA) %>%
  as_tibble() %>%
  rowid_to_column %>%
  gather(colid, distance, starts_with("V"), na.rm = TRUE) %>%
  arrange(desc(distance))
distances
#> # A tibble: 4,950 x 3
#>    rowid colid distance
#>    <int> <chr>    <dbl>
#>  1    30 V68    138244.
#>  2    30 V75    137957.
#>  3    19 V30    135068.
#>  4    48 V78    131849.
#>  5    40 V90    131280.
#>  6    48 V90    130946.
#>  7    40 V78    130035.
#>  8    50 V68    128851.
#>  9    45 V48    128805.
#> 10    40 V45    128531.
#> # ... with 4,940 more rows

sforig %>%
  st_geometry %>% 
  plot(col = "red", pch = 19)

sforig %>% 
  filter(x %in% c(30, 68)) %>%
  st_geometry %>% 
  plot( add = TRUE, col = "blue", pch = 19)

Created on 2018-07-25 by the reprex package (v0.2.0).