Using terra with geotargets

library(targets)
library(geotargets)

The geotargets package extends targets to work with geospatial data formats, such as rasters and vectors (e.g., shape files). In particular, geotargets aims to support use of the terra package, which tend to cause problems if used in targets created with tar_target() . If you are new to targets, you should start by looking at the targets manual to get a handle on the basics.

The design of geotargets is to specify target factories like so: tar_<pkg>_<type>.

In this vignette we will demonstrate the use of the terra R package, and we will demonstrate how to build raster (rast), vector (vect), raster collection (sprc), and raster dataset (sds) targets with:

How to run targets examples from vignettes

The example code in this vignette is designed for you to be able to just copy and paste into the R console. However, this is not a typical way to run a targets workflow.

The examples make use of targets::tar_script(), which creates or overwrites a _targets.R file (See for example, the _targets.R file in the demo-geotargets repo). In these examples, everything inside of tar_script({}) is what would go inside a _targets.R file defining a workflow. When running the tar_script code, it will ask you each time if you want to overwrite the _targets.R file. This means if you are exploring these examples and copying the entire examples, it is worthwhile doing this in a separate directory/project/repository to avoid overwriting your own _targets.R file.

So, when building your own targets workflow it is not recommended to use tar_script(), but instead to create _targets.R and edit _targets.R directly.

tar_terra_rast(): targets with terra rasters

targets::tar_script({
  library(targets)
  library(geotargets)
  tar_option_set(packages = "terra")
  geotargets_option_set(gdal_raster_driver = "COG")
  list(
    tar_target(
      tif_file,
      system.file("ex/elev.tif", package = "terra"),
      format = "file"
    ),
    tar_terra_rast(
      r,
      {
        rast <- rast(tif_file)
        units(rast) <- "m"
        rast
      }
    ),
    tar_terra_rast(
      r_agg,
      aggregate(r, 2)
    )
  )
})

Above is a basic example showing the use of tar_terra_rast() in a targets pipeline. The command for tar_terra_rast() can be any function that returns a SpatRaster object.

In this example, we’ve set the output to a cloud optimized geotiff (“COG”), but any GDAL driver that works with terra::writeRaster() should also work here. By default, we use “GTiff”. You can also set this option on a target-by-target basis with the filetype argument to tar_terra_rast().

Running the pipeline:

tar_make()
#> + tif_file dispatched
#> terra 1.8.42
#> ✔ tif_file completed [5ms, 7.99 kB]
#> + r dispatched
#> ✔ r completed [1ms, 10.47 kB]
#> + r_agg dispatched
#> ✔ r_agg completed [1ms, 6.30 kB]
#> ✔ ended pipeline [100ms, 3 completed, 0 skipped]
tar_read(r)
#> class       : SpatRaster 
#> dimensions  : 90, 95, 1  (nrow, ncol, nlyr)
#> resolution  : 0.008333333, 0.008333333  (x, y)
#> extent      : 5.741667, 6.533333, 49.44167, 50.19167  (xmin, xmax, ymin, ymax)
#> coord. ref. : lon/lat WGS 84 (EPSG:4326) 
#> source      : r 
#> name        : elevation 
#> min value   :       141 
#> max value   :       547 
#> unit        :         m
tar_read(r_agg)
#> class       : SpatRaster 
#> dimensions  : 45, 48, 1  (nrow, ncol, nlyr)
#> resolution  : 0.01666667, 0.01666667  (x, y)
#> extent      : 5.741667, 6.541667, 49.44167, 50.19167  (xmin, xmax, ymin, ymax)
#> coord. ref. : lon/lat WGS 84 (EPSG:4326) 
#> source      : r_agg 
#> name        : elevation 
#> min value   :    166.75 
#> max value   :    529.50

Raster metadata

You may have noticed the units for the r target above have gone missing. This is due to limitations of terra and targets—terra saves some metadata in “sidecar” aux.json files and targets enforces a strict one file per target rule.

You can get around this by setting preserve_metadata = "zip" in tar_terra_rast() to save the output files, including the metadata, as a minimally compressed zip archive.

You can also set this for all raster targets with geotargets_option_set(terra_preserve_metadata = "zip").

Note: there are likely performance costs associated with this option.
As an alternative, you can encode information in the layer names by setting names(r) <- which are retained even with the default preserve_metadata = "drop".

targets::tar_script({
  # contents of _targets.R:
  library(targets)
  library(geotargets)
  tar_option_set(packages = "terra")
  geotargets_option_set(gdal_raster_driver = "COG")
  list(
    tar_target(
      tif_file,
      system.file("ex/elev.tif", package = "terra"),
      format = "file"
    ),
    tar_terra_rast(
      r,
      {
        rast <- rast(tif_file)
        units(rast) <- "m"
        rast
      },
      preserve_metadata = "zip"
    )
  )
})
tar_make()
#> + r dispatched
#> terra 1.8.42
#> ✔ r completed [2ms, 10.26 kB]
#> ✔ ended pipeline [83ms, 1 completed, 1 skipped]
terra::units(tar_read(r))
#> [1] "m"

tar_terra_vect(): targets with terra vectors

For terra SpatVector objects, use tar_terra_vect() in the pipeline. You can set vector specific options with geotargets_option_set() or with the filetype and gdal arguments to individual tar_terra_vect() calls.

targets::tar_script({
  # contents of _targets.R:
  library(targets)
  library(geotargets)
  geotargets_option_set(gdal_vector_driver = "GeoJSON")
  list(
    tar_target(
      vect_file,
      system.file("ex", "lux.shp", package = "terra"),
      format = "file"
    ),
    tar_terra_vect(
      v,
      terra::vect(vect_file)
    ),
    tar_terra_vect(
      v_proj,
      terra::project(v, "EPSG:2196")
    )
  )
})
tar_make()
#> + vect_file dispatched
#> ✔ vect_file completed [0ms, 64.69 kB]
#> + v dispatched
#> ✔ v completed [8ms, 117.63 kB]
#> + v_proj dispatched
#> ✔ v_proj completed [20ms, 213.55 kB]
#> ✔ ended pipeline [128ms, 3 completed, 0 skipped]
tar_read(v)
#>  class       : SpatVector 
#>  geometry    : polygons 
#>  dimensions  : 12, 6  (geometries, attributes)
#>  extent      : 5.74414, 6.528252, 49.44781, 50.18162  (xmin, xmax, ymin, ymax)
#>  source      : v
#>  coord. ref. : lon/lat WGS 84 (EPSG:4326) 
#>  names       :  ID_1   NAME_1  ID_2   NAME_2  AREA       POP
#>  type        : <num>    <chr> <num>    <chr> <num>     <num>
#>  values      :     1 Diekirch     1 Clervaux   312 1.808e+04
#>                    1 Diekirch     2 Diekirch   218 3.254e+04
#>                    1 Diekirch     3  Redange   259 1.866e+04
tar_read(v_proj)
#>  class       : SpatVector 
#>  geometry    : polygons 
#>  dimensions  : 12, 6  (geometries, attributes)
#>  extent      : -69990.51, -13879.85, 5484907, 5566555  (xmin, xmax, ymin, ymax)
#>  source      : v_proj
#>  coord. ref. : ETRS89 / Kp2000 Jutland (EPSG:2196) 
#>  names       :  ID_1   NAME_1  ID_2   NAME_2  AREA       POP
#>  type        : <num>    <chr> <num>    <chr> <num>     <num>
#>  values      :     1 Diekirch     1 Clervaux   312 1.808e+04
#>                    1 Diekirch     2 Diekirch   218 3.254e+04
#>                    1 Diekirch     3  Redange   259 1.866e+04

tar_terra_sprc(): targets with terra raster collections

Targets that produce a SpatRasterCollection can be created with tar_terra_sprc(). The various rasters in the collection are saved as subdatasets to adhere to targets one file per target rule.

targets::tar_script({
  # contents of _targets.R:
  library(targets)
  library(geotargets)
  elev_scale <- function(raster, z = 1, projection = "EPSG:4326") {
    terra::project(
      raster * z,
      projection
    )
  }
  tar_option_set(packages = "terra")
  geotargets_option_set(gdal_raster_driver = "GTiff")
  list(
    tar_target(
      elev_file,
      system.file("ex", "elev.tif", package = "terra"),
      format = "file"
    ),
    tar_terra_rast(
      r,
      rast(elev_file)
    ),
    tar_terra_sprc(
      raster_elevs,
      # two rasters, one unaltered, one scaled by factor of 2 and
      # reprojected to interrupted good homolosine
      terra::sprc(list(
        elev_scale(r, 1),
        elev_scale(r, 2, "+proj=igh")
      ))
    )
  )
})
tar_make()
#> + elev_file dispatched
#> terra 1.8.42
#> ✔ elev_file completed [5ms, 7.99 kB]
#> + r dispatched
#> ✔ r completed [2ms, 8.52 kB]
#> + raster_elevs dispatched
#> ✔ raster_elevs completed [14ms, 37.90 kB]
#> ✔ ended pipeline [136ms, 3 completed, 0 skipped]
tar_read(raster_elevs)
#> class       : SpatRasterCollection 
#> length      : 2 
#> nrow        : 90, 115 
#> ncol        : 95, 114 
#> nlyr        :  1,   1 
#> extent      : 5.741667, 1558890, 49.44167, 5556741  (xmin, xmax, ymin, ymax)
#> crs (first) : lon/lat WGS 84 (EPSG:4326) 
#> names       : raster_elevs, raster_elevs

tar_terra_sds(): targets with terra raster datasets

A terra SpatRasterDataset is very similar to a SpatRasterCollection except that all sub-datasets must have the same projection and extent

targets::tar_script({
  # contents of _targets.R:
  library(targets)
  library(geotargets)
  tar_option_set(packages = "terra")
  list(
    tar_target(
      logo_file,
      system.file("ex/logo.tif", package = "terra"),
      format = "file"
    ),
    tar_terra_sds(
      raster_dataset,
      {
        x <- sds(rast(logo_file), rast(logo_file) / 2)
        names(x) <- c("first", "second")
        x
      }
    )
  )
})
tar_make()
#> + logo_file dispatched
#> terra 1.8.42
#> ✔ logo_file completed [5ms, 22.46 kB]
#> + raster_dataset dispatched
#> ✔ raster_dataset completed [18ms, 54.73 kB]
#> ✔ ended pipeline [89ms, 2 completed, 0 skipped]
tar_read(raster_dataset)
#> class       : SpatRasterDataset 
#> subdatasets : 2 
#> dimensions  : 77, 101 (nrow, ncol)
#> nlyr        : 3, 3 
#> resolution  : 1, 1  (x, y)
#> extent      : 0, 101, 0, 77  (xmin, xmax, ymin, ymax)
#> coord. ref. : Cartesian (Meter) 
#> source(s)   : raster_dataset 
#> names       : raster_dataset, raster_dataset