Welcome to {tidyterra}
tidyterra is a package that add common methods from
the tidyverse for SpatRaster
and SpatVectors objects created with the {terra} package. It
also adds specific geom_spat*() functions for plotting
these kind of objects with {ggplot2}.
Why {tidyterra}?
Spat* objects are not like regular data frames. They are a different
type of objects, implemented via the S4 object system, and have
their own syntax and computation methods, implemented on the {terra}
package.
By implementing tidyverse methods for these objects, and more
specifically {dplyr} and {tidyr} methods, a useR can
now work more easily with Spat*, just like (s)he would do with tabular
data.
Note that in terms of performance, {terra} is much
more optimized for working for this kind of objects, so it is
recommended also to learn a bit of {terra} syntax. Each
function of {tidyterra} refers (when possible) to the corresponding
equivalent on {terra}.
A note for advanced {terra} users
As previously mentioned, {tidyterra} is not optimized in terms of
performance. Specially when working with filter() and
mutate() methods, it can be slow.
As a rule of thumb, {tidyterra} can handle objects with less than
10.000.000 slots of information(i.e.,
terra::ncell(a_rast) * terra::nlyr(a_rast) < 10e6).
Get started with {tidyterra}
Load {tidyterra} with additional libraries of the {tidyverse}:
library(tidyterra)
library(dplyr)
#>
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#>
#> filter, lag
#> The following objects are masked from 'package:base':
#>
#> intersect, setdiff, setequal, union
library(tidyr)
Currently, the following methods are available:
tibble::as_tibble() |
✔️ |
✔️ |
dplyr::filter() |
✔️ |
✔️ |
dplyr::mutate() |
✔️ |
✔️ |
dplyr::pull() |
✔️ |
✔️ |
dplyr::relocate() |
✔️ |
✔️ |
dplyr::rename() |
✔️ |
✔️ |
dplyr::select() |
✔️ |
✔️ |
dplyr::slice() |
✔️ |
✔️ |
dplyr::transmute() |
✔️ |
✔️ |
tidyr::drop_na() |
✔️ |
✔️ (questioned) |
tidyr::replace_na() |
✔️ |
✔️ |
ggplot2::autoplot() |
✔️ |
✔️ |
ggplot2::fortify() |
✔️ to sf with sf::st_as_sf() |
✔️ To tibble with coordinates |
Let’s see some of them in action:
library(terra)
f <- system.file("extdata/cyl_temp.tif", package = "tidyterra")
temp <- rast(f)
temp
#> class : SpatRaster
#> dimensions : 89, 116, 3 (nrow, ncol, nlyr)
#> resolution : 3856.617, 3856.617 (x, y)
#> extent : 2893583, 3340950, 2019451, 2362690 (xmin, xmax, ymin, ymax)
#> coord. ref. : ETRS89-extended / LAEA Europe (EPSG:3035)
#> source : cyl_temp.tif
#> names : tavg_04, tavg_05, tavg_06
#> min values : 0.565614, 4.294102, 8.817221
#> max values : 13.283829, 16.740898, 21.113781
mod <- temp %>%
select(-1) %>%
mutate(newcol = tavg_06 - tavg_05) %>%
relocate(newcol, .before = 1) %>%
replace_na(list(newcol = 3)) %>%
rename(difference = newcol)
mod
#> class : SpatRaster
#> dimensions : 89, 116, 3 (nrow, ncol, nlyr)
#> resolution : 3856.617, 3856.617 (x, y)
#> extent : 2893583, 3340950, 2019451, 2362690 (xmin, xmax, ymin, ymax)
#> coord. ref. : ETRS89-extended / LAEA Europe (EPSG:3035)
#> sources : memory
#> memory
#> memory
#> names : difference, tavg_05, tavg_06
#> min values : 2.786910, 4.294102, 8.817221
#> max values : 5.408157, 16.740898, 21.113781
On the previous example, we had:
Eliminated the first layer of the raster
tavg_04.
Created a new layer newcol as the difference of the
layers tavg_05 and tavg_06.
Relocated newcolas the first layer of the
SpatRaster
Replaced the NA cells on newcol with
3.
Renamed newcol to difference.
In all the process, the essential properties of the SpatRaster
(number of cells, columns and rows, extent, resolution and coordinate
reference system) have not been modified. Other methods as
filter(), slice() or drop_na()
can modify these properties, as they would do when applied to a data
frame (number of rows would be modified on that case).
Plotting with {ggplot2}
SpatRasters
{tidyterra} provides several geom_* for SpatRasters.
When the SpatRaster has the CRS informed
(i.e. terra::crs(a_rast) != ""), the geom uses
ggplot2::coord_sf(), and may be also reprojected for
adjusting the coordinates to other spatial layers:
library(ggplot2)
# A faceted SpatRaster
ggplot() +
geom_spatraster(data = temp) +
facet_wrap(~lyr) +
scale_fill_whitebox_c(
palette = "muted",
na.value = "white"
)
# Contour lines for a specific layer
f_volcano <- system.file("extdata/volcano2.tif", package = "tidyterra")
volcano2 <- rast(f_volcano)
ggplot() +
geom_spatraster(data = volcano2) +
geom_spatraster_contour(data = volcano2, breaks = seq(80, 200, 5)) +
scale_fill_whitebox_c() +
coord_sf(expand = FALSE) +
labs(fill = "elevation")
# Contour filled
ggplot() +
geom_spatraster_contour_filled(data = volcano2) +
scale_fill_whitebox_d(palette = "atlas") +
labs(fill = "elevation")
With {tidyterra} you can also plot RGB SpatRasters to add imagery to
your plots:
# Read a vector
f_v <- system.file("extdata/cyl.gpkg", package = "tidyterra")
v <- vect(f_v)
# Read a tile
f_rgb <- system.file("extdata/cyl_tile.tif", package = "tidyterra")
r_rgb <- rast(f_rgb)
rgb_plot <- ggplot(v) +
geom_spatraster_rgb(data = r_rgb) +
geom_spatvector(fill = NA, size = 1)
rgb_plot
# Change CRS automatically
rgb_plot +
coord_sf(crs = 3035)
{tidyterra} provides selected scales that are suitable for creating
hypsometric and bathymetric maps:
asia <- rast(system.file("extdata/asia.tif", package = "tidyterra"))
asia
#> class : SpatRaster
#> dimensions : 232, 432, 1 (nrow, ncol, nlyr)
#> resolution : 22550.66, 22512.94 (x, y)
#> extent : 7619120, 17361007, -1304745, 3918256 (xmin, xmax, ymin, ymax)
#> coord. ref. : WGS 84 / Pseudo-Mercator (EPSG:3857)
#> source : asia.tif
#> name : file44bc291153f2
#> min value : -10071.50
#> max value : 6064.73
ggplot() +
geom_spatraster(data = asia) +
scale_fill_hypso_tint_c(
palette = "gmt_globe",
labels = scales::label_number(),
breaks = c(-10000, -5000, 0, 2500, 5000, 8000),
guide = guide_colorbar(
direction = "horizontal",
title.position = "top",
barwidth = 20
)
) +
labs(
fill = "elevation (m)",
title = "Hypsometric map of Asia"
) +
theme_minimal() +
theme(legend.position = "bottom")