3. stars tidyverse methods
In stars: Spatiotemporal Arrays, Raster and Vector Data Cubes

For a better version of the stars vignettes see https://r-spatial.github.io/stars/articles/

knitr::opts_chunk$set(echo = TRUE, collapse = TRUE, dev = "png")
ev = TRUE
knitr::opts_chunk$set(fig.height = 4.5)
knitr::opts_chunk$set(fig.width = 6)

This vignette shows how some of the tidyverse verbs can be used on stars objects.

The stars and tidyverse packages are loaded by

library(stars)
library(dplyr)

Methods now available for class stars are

methods(class = "stars")

We will work with a three-band section of a landsat image:

system.file("tif/L7_ETMs.tif", package = "stars") %>%
    read_stars -> x
x

`slice`

slice slices a sub-array out of the cube; this is done by specifying the dimension on which to act, and the slice number.

x %>% slice(band, 6) -> x6
x6

It returns a lower-dimensional array if a single element is selected along the slice dimension.

`filter`

Similar to slice, filter selects on dimensions but evaluates their values rather than their index: in

x %>% filter(x > 289000, x < 291000, band > 3) -> x7
x7

the subarray is created based on the x coordinate values.

Note that filter converts the object to a tbl_cube, and uses the dplyr filter method for tbl_cube objects. This has the limitation that stars objects with rectilinear, curvilinear or simple feature geometries cannot be handled. For such objects, using regular [ selection or using st_crop may be an alternative.

`pull`

pull pulls out an array from a stars object:

x %>% pull(1) -> x8
class(x8)
dim(x8)

`mutate`

x %>% mutate(band2 = 2 * L7_ETMs.tif) -> x2 
x2

`select`

select selects an attribute, or a set of attributes:

x2 %>% select(band2) -> x9
x9

`geom_stars`

geom_raster is a ggplot2 geom function that accepts stars objects as its data argument and

sets up the raster or vector spatial coordinates as plot dimensions, and the first attribute as the fill variable
allows for downsampling (without choosing a suitable downsampling level)
chooses between using geom_raster, geom_rect and geom_sf depending on whether the geometry is regular, rectilinear or has vector geometries

An example use is

library(ggplot2)
library(viridis)
ggplot() + 
  geom_stars(data = x) +
  coord_equal() +
  facet_wrap(~band) +
  theme_void() +
  scale_fill_viridis() +
  scale_x_discrete(expand = c(0, 0)) +
  scale_y_discrete(expand = c(0, 0))