regridR: re-grid raster layers

Most global-coverage raster layers containing variables used in species distribution models (e.g. WorldClim, CHELSA) come in unprojected longitude-latitude coordinates, with a nominal spatial resolution like “approximately 1x1 km² at the equator”. One problem is that longitude-latitude pixels are not equal-area, and they’re not 1x1-km² either (or whatever their nominal resolution is): as the longitude meridians converge towards the North and the South Pole, longitude-latitude pixels cover progressively smaller areas as we move away from the Equator, and they are already considerably smaller than their nominal resolution across temperate regions like Europe. Their actual sizes can be checked e.g. with the terra::cellSize() function.

If we want our raster variables on a grid of pixels that matches an actually equal-area grid within a given region (such as the EEA reference grid in Europe), we can use the regrid() function of package regridR to convert raster layers into such equal-area grid. Below is a worked example. We start by downloading some climate layers from the CHELSA website, using the downloadif() function also included in the regridR package. This will download files if they haven’t already been (completely) downloaded and saved in the destination folder:

# LOAD REQUIRED PACKAGES ----

# install.packages("terra")
# remotes::install_github("AMBarbosa/regridR")

library(regridR)
library(terra)
#> terra 1.9.50


# DOWNLOAD SOME VARIABLES ----

# get CHELSA climate links for a couple variables:
links <- regridR::linkbuild(c("bio1", "bio12", "scd"))

# create a folder for receiving downloads:
dir.create("outputs/variables", recursive = TRUE)
#> Warning in dir.create("outputs/variables", recursive = TRUE):
#> 'outputs/variables' already exists

# allow longer download times:
options(timeout = 6000)

# download variables if not already there:
regridR::downloadif(links, destdir = "outputs/variables")
#> 1
#> CHELSA_bio1_1981-2010_V.2.1.tif
#> 2
#> CHELSA_bio12_1981-2010_V.2.1.tif
#> 3
#> CHELSA_scd_1981-2010_V.2.1.tif

# import variables from downloads folder:
layers <- terra::rast(list.files("outputs/variables", full.names = TRUE))

terra::plot(layers)

Next, we will import a vector polygon map of a 10x10-km² equal-area grid recommended by the European Environment Agency (EEA), using Belgium as an example:

# IMPORT EQUAL-AREA VECTOR GRID ----

EEAgrid <- terra::vect(system.file("extdata/eea10_belgium.gpkg", 
                                   package = "regridR"))

terra::plot(EEAgrid)

We can project the vector grid to overlay a part of the climate layers and confirm that they don’t align (and the pixels are not equal-area, or square, or 1-km²), so simply aggregating the raster into groups of 10x10 pixels wouldn’t be ideal:

terra::plot(layers[[1]], maxcell = ncell(layers), ext = c(5, 5.4, 49.5, 49.7))

terra::plot(terra::project(EEAgrid, layers), add = TRUE)

So, we’ll use the regrid() function of regridR to get the climate layers on a raster grid whose pixels match those polygon grid cells:

# RE-GRID LAYERS ----

layers_regrid <- regridR::regrid(layers = layers, grid = EEAgrid, 
                                 na.rm = TRUE, touches = TRUE)

terra::plot(layers_regrid)

By default, regrid() will use the mean() function to summarize the values of the pixels falling within each polygon grid cell, and the terra::zonal() function to do this summarizing. However, installing also the exactextractr package and running regrid() with the argument exactextract = TRUE can make the computation much faster for large grids, albeit with a slightly different algorithm:

# RE-GRID LAYERS FASTER ----

layers_regrid <- regridR::regrid(layers = layers, grid = EEAgrid, 
                                 exactextract = TRUE)
#> projecting 'grid' to overlay 'layers'
#> extracting 'layers' to projected 'grid' (can take a while for dense grids...)
#> rasterizing input 'grid' with extracted 'layers' values
#> finished!

terra::plot(layers_regrid)

We can visually check that the output (re-gridded) layers’ pixels align with the input EEA grid:

terra::plot(layers_regrid[[1]])

terra::plot(EEAgrid, lwd = 0.3, add = TRUE)

So, you can use the re-gridded layers e.g. for modelling species distribution data using environmental rasters whose pixels match the equal-area squares recommended for your study region, such as the EEA reference grid in Europe.