Skip to contents

Converting raster data

Source: vignettes/pg_from_raster.Rmd
pg_from_raster.Rmd

pg_from_raster() accepts a terra::SpatRaster or one or more raster-file paths. It aligns the source to a configured PRIO-GRID and can return a raster, a cell-level table, or both.

Raster transformation uses the working directory managed by priogrid. In a real project, configure a persistent location once with, for example, the following code before running the conversion examples:

priogrid::pg_set_rawfolder("data/priogrid")

The conversion chunks are not evaluated while the vignette is built because the working directory is a persistent, user-specific setting.

A small raster example 

config <- priogrid::pg_config(
  nrow = 18,
  ncol = 36,
  start_date = as.Date("2020-12-31"),
  end_date = as.Date("2021-12-31"),
  verbose = FALSE
)

temperature <- terra::rast(
  nrows = 18,
  ncols = 36,
  xmin = -180,
  xmax = 180,
  ymin = -90,
  ymax = 90,
  crs = "EPSG:4326"
)
terra::values(temperature) <- seq_len(terra::ncell(temperature)) / 10

Convert the raster and request tabular output:

temperature_pg <- pg_from_raster(
  x = temperature,
  variable = "temperature",
  aggregation = "mean",
  return = "data",
  config = config
)

head(temperature_pg)

Choosing an aggregation rule

Available rules are sum, mean, min, max, median, and modal. Choose the rule based on the variable’s meaning:

  • use mean for continuous surfaces such as temperature;
  • use sum for quantities that should accumulate across source cells;
  • use modal for categorical rasters; and
  • use minima, maxima, or medians only when those summaries answer the research question.

When the source raster is coarser than the target grid, disagg_method controls interpolation. Nearest-neighbor interpolation is appropriate for categories, while bilinear or cubic interpolation may suit continuous data.

Multiple layers and time

Multi-layer rasters require one date or year per layer unless dates are handled by pg_from_netcdf().

temperature_series <- c(temperature, temperature + 1)

temperature_time <- pg_from_raster(
  x = temperature_series,
  variable = "temperature",
  aggregation = "mean",
  time = c(2020, 2021),
  config = config
)

head(temperature_time)

Use layer to select a subset before conversion and na_values to translate source sentinel values, such as -9999, to NA.

Return and write results 

result <- pg_from_raster(
  x = temperature,
  variable = "temperature",
  aggregation = "mean",
  return = "result",
  config = config
)

result
pg_write_result(result, "temperature.csv")
pg_write_result(result, "temperature.tif")

For NetCDF variables and automatic layer-time extraction, see vignette("pg_from_netcdf", package = "priogridR").