Skip to contents

Convert a netcdf array to long format as data.table

Source: R/nc_grid_to_dt.R
nc_grid_to_dt.Rd

Extracts a variable from netcdf, and returns a data.table with cell index, date, values, and optionally: coordinates.

Usage

nc_grid_to_dt(
  filename,
  variable,
  icell_raster_pkg = TRUE,
  add_xy = FALSE,
  interpolate_to_standard_calendar = FALSE,
  date_range,
  verbose = FALSE
)

Arguments

filename

Complete path to .nc file.

variable

Name of the variable to extract from filename (character).

icell_raster_pkg

Boolean, if TRUE, cell indices will be ordered as if you were extracting the data with the raster package.

add_xy

Boolean, if TRUE, adds columns with x and y coordinates.

interpolate_to_standard_calendar

Boolean, if TRUE will use map_non_standard_calendar to interpolate values to a standard calendar.

date_range

(optional) two-element vector of class Date (min, max), which will be used to extract only parts of the netcdf file

verbose

Boolean, if TRUE, prints more information.

Value

A data.table with columns:

  • date: Date of class Date, if file has a standard calendar. Date as character, if it has a non-standard calendar (360, noleap) and if interpolate_to_standard_calendar is set to FALSE. If interpolate_to_standard_calendar is TRUE, it's always of class Date.

  • variable: Values, column is renamed to input variable

  • (optional) x,y: Coordinates of netcdf dimensions, will be renamed to dimension names found in array named after input variable

Details

Coordinates are usually not put in the result, because it saves space. It is recommended to merge them after the final operations. The unique cell index is more efficient. However, if you plan to merge to data extracted with the raster package (assuming the same grid), then cell indices might differ. Set icell_raster_pkg to TRUE, to have the same cell indices. Note that raster and ncdf4 have different concepts of coordinates (cell corner vs. cell center), so merging based on coordinates can produce arbitrary results (besides rounding issues).

Warning

Netcdf files can be huge, so loading everything in memory can rapidly crash your R session. Think first about subsetting or aggregating (e.g. using CDO: https://code.mpimet.mpg.de/projects/cdo/).

See also

The raster and terra packages can also open netcdf files and create data.frames with raster::as.data.frame or terra::as.data.frame. But, it does not handle non-standard calendars, and returns a data.frame, which is slower than data.table.

Examples

# example data from EURO-CORDEX (cropped for size)
fn1 <- system.file("extdata", "test1.nc", package = "eurocordexr")
dat <- nc_grid_to_dt(fn1)
str(dat)
#> Classes ‘data.table’ and 'data.frame':	25550 obs. of  3 variables:
#>  $ icell : int  64 65 66 67 68 69 70 57 58 59 ...
#>  $ date  : Date, format: "1950-01-01" "1950-01-01" ...
#>  $ tasmin: num  269 269 271 271 272 ...
#>  - attr(*, ".internal.selfref")=<externalptr>