A practical example: walking from public transport in Singapore

Here we will calculate the walking travel time from the nearest mass transit station across the island nation of Singapore — specifically Mass Rapid Transit (MRT) and Light Rail Transit (LRT) stations — and create a map of this.

Prepare the data

For this exercise, we need two items of data:

our points to calculate travel time from — here the locations of Singapore’s MRT and LRT stations, and
our area of interest — in this case a map of Singapore.

Points

Our points of interest will be the stations data set included in traveltime; a 563 row, 2 column matrix containing the longitude (x) and latitude (y) of all LRT and MRT station exits in Singapore from ¹:

library(traveltime)
head(stations)
#>             x        y
#> [1,] 103.9091 1.334922
#> [2,] 103.9335 1.336555
#> [3,] 103.8493 1.297699
#> [4,] 103.8508 1.299195
#> [5,] 103.9094 1.335311
#> [6,] 103.9389 1.344999

Area of interest

To obtain our area of interest, we download a national-level polygon boundary of Singapore using the geodata package. Here we download only the national boundary (level = 0) at a low resolution (resolution = 2). Our boundary singapore_shapefile is a SpatVector class object from the package terra.

library(terra)
library(geodata)

singapore_shapefile <- gadm(
  country = "Singapore",
  level = 0,
  path = tempdir(),
  resolution = 2
)

singapore_shapefile
#> class       : SpatVector
#> geometry    : polygons
#> dimensions  : 1, 2  (geometries, attributes)
#> extent      : 103.6091, 104.0858, 1.1664, 1.4714  (xmin, xmax, ymin, ymax)
#> coord. ref. : lon/lat WGS 84 (EPSG:4326)
#> names       : GID_0   COUNTRY
#> type        : <chr>     <chr>
#> values      :   SGP Singapore

Friction surface

Now that we have the two items of data that we require initially, the next step is to prepare a friction surface for our area of interest.

We will use the friction surface from Weiss et al. (2020)² that can be downloaded by traveltime with the function get_friction_surface(). This function takes extents in a variety of formats and returns the surface for that extent only.

We can pass in our basemap singapore_shapefile, a SpatVector, directly as the extent. We’re interested in walking time from a station, so we’ll download the walking friction surface by specifying surface = "walk2020".

(Alternatively, we could use surface = "motor2020" for motorised travel).

We’re only interested in walking on land, so we then mask out areas outside of the land boundary of singapore_shapefile:

friction_singapore <- get_friction_surface(
    surface = "walk2020",
    extent = singapore_shapefile
  )|> 
  mask(singapore_shapefile)

Thus we have our friction surface as a SpatRaster:

friction_singapore
#> class       : SpatRaster
#> size        : 37, 57, 1  (nrow, ncol, nlyr)
#> resolution  : 0.008333333, 0.008333333  (x, y)
#> extent      : 103.6083, 104.0833, 1.166667, 1.475  (xmin, xmax, ymin, ymax)
#> coord. ref. : lon/lat WGS 84 (EPSG:4326)
#> source(s)   : memory
#> varname     : Accessibility__202001_Global_Walking_Only_Friction_Surface_1.1664,103.6091,1.4714,104.0858
#> name        : friction_surface
#> min value   :            0.012
#> max value   :         0.061927

Input data

Below we plot the friction surface raster friction_singapore, with the vector boundary singapore_shapefile as a grey line, and stations as grey points. traveltime takes resistance values of friction (see paper for more details), so higher values of friction indicate more time travelling across a given cell.

library(tidyterra)
library(ggplot2)

ggplot() +
  geom_spatraster(
    data = friction_singapore
  ) +
  geom_spatvector(
    data = singapore_shapefile,
    fill = "transparent",
    col = "grey50"
  ) +
  geom_point(
    data = stations,
    aes(
      x = x,
      y = y
    ),
    col = "grey60",
    size = 0.5
  ) +
  scale_fill_viridis_c(
    option = "A",
    na.value = "transparent",
    direction = -1
  ) +
  labs(
    fill = "Resistance",
    x = NULL,
    y = NULL
  ) +
  theme_minimal()

Friction surface raster of Singapore, showing Singapore boundary in grey, and station locations as grey points.

Calculate travel time

With all the data collected, the function calculate_travel_time() takes the friction surface friction_singapore and the points of interest in stations, and returns a SpatRaster of walking time in minutes to each cell from the nearest station:

travel_time_singapore <- calculate_travel_time(
  friction_surface = friction_singapore,
  points = stations
)

travel_time_singapore
#> class       : SpatRaster
#> size        : 37, 57, 1  (nrow, ncol, nlyr)
#> resolution  : 0.008333333, 0.008333333  (x, y)
#> extent      : 103.6083, 104.0833, 1.166667, 1.475  (xmin, xmax, ymin, ymax)
#> coord. ref. : lon/lat WGS 84 (EPSG:4326)
#> source(s)   : memory
#> varname     : Accessibility__202001_Global_Walking_Only_Friction_Surface_1.1664,103.6091,1.4714,104.0858
#> name        : travel_time
#> min value   :           0
#> max value   :   165.48359

Plot results

We present the resulting calculated travel times below where (as expected) the travel times are lowest near station exits and progressively higher further away. Note that some cells in travel_time_singapore are NA. The islands to the south and north-east are shown as filled cells in the figure above, i.e., they are not masked out by singapore_shapefile. But because those islands are not connected to any cells with a station, they cannot be reached, so their travel time is NA and these cells do not appear in the figure below.

ggplot() +
  geom_spatraster(
    data = travel_time_singapore
  ) +
  scale_fill_viridis_c(
    option = "A",
    direction = -1,
    na.value = "transparent"
  ) +
  theme_minimal() +
  labs(fill = "Minutes") +
  geom_spatvector(
    data = singapore_shapefile,
    fill = "transparent",
    col = "grey20"
  )

Map of walking travel time in Singapore, in minutes from nearest MRT or LRT station.