R: Extract or summarize values from rasters

exact_extract {exactextractr}

R Documentation

Extract or summarize values from rasters

Description

Extracts the values of cells in a raster (RasterLayer, RasterStack RasterBrick, or SpatRaster) that are covered by polygons in a simple feature collection (sf or sfc) or SpatialPolygonsDataFrame. Returns either a summary of the extracted values or the extracted values themselves.

Usage

## S4 method for signature 'Raster,sf'
exact_extract(
  x,
  y,
  fun = NULL,
  ...,
  weights = NULL,
  append_cols = NULL,
  coverage_area = FALSE,
  default_value = NA_real_,
  default_weight = NA_real_,
  include_area = FALSE,
  include_cell = FALSE,
  include_cols = NULL,
  include_xy = FALSE,
  force_df = FALSE,
  full_colnames = FALSE,
  stack_apply = FALSE,
  summarize_df = FALSE,
  quantiles = NULL,
  progress = TRUE,
  max_cells_in_memory = 3e+07,
  grid_compat_tol = 0.001,
  colname_fun = NULL
)

## S4 method for signature 'Raster,SpatialPolygonsDataFrame'
exact_extract(x, y, ...)

## S4 method for signature 'Raster,SpatialPolygons'
exact_extract(x, y, ...)

## S4 method for signature 'Raster,sfc_MULTIPOLYGON'
exact_extract(
  x,
  y,
  fun = NULL,
  ...,
  weights = NULL,
  append_cols = NULL,
  coverage_area = FALSE,
  default_value = NA_real_,
  default_weight = NA_real_,
  include_area = FALSE,
  include_cell = FALSE,
  include_cols = NULL,
  include_xy = FALSE,
  force_df = FALSE,
  full_colnames = FALSE,
  stack_apply = FALSE,
  summarize_df = FALSE,
  quantiles = NULL,
  progress = TRUE,
  max_cells_in_memory = 3e+07,
  grid_compat_tol = 0.001,
  colname_fun = NULL
)

## S4 method for signature 'Raster,sfc_POLYGON'
exact_extract(
  x,
  y,
  fun = NULL,
  ...,
  weights = NULL,
  append_cols = NULL,
  coverage_area = FALSE,
  default_value = NA_real_,
  default_weight = NA_real_,
  include_area = FALSE,
  include_cell = FALSE,
  include_cols = NULL,
  include_xy = FALSE,
  force_df = FALSE,
  full_colnames = FALSE,
  stack_apply = FALSE,
  summarize_df = FALSE,
  quantiles = NULL,
  progress = TRUE,
  max_cells_in_memory = 3e+07,
  grid_compat_tol = 0.001,
  colname_fun = NULL
)

## S4 method for signature 'Raster,sfc_GEOMETRY'
exact_extract(
  x,
  y,
  fun = NULL,
  ...,
  weights = NULL,
  append_cols = NULL,
  coverage_area = FALSE,
  default_value = NA_real_,
  default_weight = NA_real_,
  include_area = FALSE,
  include_cell = FALSE,
  include_cols = NULL,
  include_xy = FALSE,
  force_df = FALSE,
  full_colnames = FALSE,
  stack_apply = FALSE,
  summarize_df = FALSE,
  quantiles = NULL,
  progress = TRUE,
  max_cells_in_memory = 3e+07,
  grid_compat_tol = 0.001,
  colname_fun = NULL
)

## S4 method for signature 'Raster,sfc_GEOMETRYCOLLECTION'
exact_extract(
  x,
  y,
  fun = NULL,
  ...,
  weights = NULL,
  append_cols = NULL,
  coverage_area = FALSE,
  default_value = NA_real_,
  default_weight = NA_real_,
  include_area = FALSE,
  include_cell = FALSE,
  include_cols = NULL,
  include_xy = FALSE,
  force_df = FALSE,
  full_colnames = FALSE,
  stack_apply = FALSE,
  summarize_df = FALSE,
  quantiles = NULL,
  progress = TRUE,
  max_cells_in_memory = 3e+07,
  grid_compat_tol = 0.001,
  colname_fun = NULL
)

## S4 method for signature 'SpatRaster,sf'
exact_extract(
  x,
  y,
  fun = NULL,
  ...,
  weights = NULL,
  append_cols = NULL,
  coverage_area = FALSE,
  default_value = NA_real_,
  default_weight = NA_real_,
  include_area = FALSE,
  include_cell = FALSE,
  include_cols = NULL,
  include_xy = FALSE,
  force_df = FALSE,
  full_colnames = FALSE,
  stack_apply = FALSE,
  summarize_df = FALSE,
  quantiles = NULL,
  progress = TRUE,
  max_cells_in_memory = 3e+07,
  grid_compat_tol = 0.001,
  colname_fun = NULL
)

## S4 method for signature 'SpatRaster,SpatialPolygonsDataFrame'
exact_extract(x, y, ...)

## S4 method for signature 'SpatRaster,SpatialPolygons'
exact_extract(x, y, ...)

## S4 method for signature 'SpatRaster,sfc_MULTIPOLYGON'
exact_extract(
  x,
  y,
  fun = NULL,
  ...,
  weights = NULL,
  append_cols = NULL,
  coverage_area = FALSE,
  default_value = NA_real_,
  default_weight = NA_real_,
  include_area = FALSE,
  include_cell = FALSE,
  include_cols = NULL,
  include_xy = FALSE,
  force_df = FALSE,
  full_colnames = FALSE,
  stack_apply = FALSE,
  summarize_df = FALSE,
  quantiles = NULL,
  progress = TRUE,
  max_cells_in_memory = 3e+07,
  grid_compat_tol = 0.001,
  colname_fun = NULL
)

## S4 method for signature 'SpatRaster,sfc_POLYGON'
exact_extract(
  x,
  y,
  fun = NULL,
  ...,
  weights = NULL,
  append_cols = NULL,
  coverage_area = FALSE,
  default_value = NA_real_,
  default_weight = NA_real_,
  include_area = FALSE,
  include_cell = FALSE,
  include_cols = NULL,
  include_xy = FALSE,
  force_df = FALSE,
  full_colnames = FALSE,
  stack_apply = FALSE,
  summarize_df = FALSE,
  quantiles = NULL,
  progress = TRUE,
  max_cells_in_memory = 3e+07,
  grid_compat_tol = 0.001,
  colname_fun = NULL
)

## S4 method for signature 'SpatRaster,sfc_GEOMETRY'
exact_extract(
  x,
  y,
  fun = NULL,
  ...,
  weights = NULL,
  append_cols = NULL,
  coverage_area = FALSE,
  default_value = NA_real_,
  default_weight = NA_real_,
  include_area = FALSE,
  include_cell = FALSE,
  include_cols = NULL,
  include_xy = FALSE,
  force_df = FALSE,
  full_colnames = FALSE,
  stack_apply = FALSE,
  summarize_df = FALSE,
  quantiles = NULL,
  progress = TRUE,
  max_cells_in_memory = 3e+07,
  grid_compat_tol = 0.001,
  colname_fun = NULL
)

## S4 method for signature 'SpatRaster,sfc_GEOMETRYCOLLECTION'
exact_extract(
  x,
  y,
  fun = NULL,
  ...,
  weights = NULL,
  append_cols = NULL,
  coverage_area = FALSE,
  default_value = NA_real_,
  default_weight = NA_real_,
  include_area = FALSE,
  include_cell = FALSE,
  include_cols = NULL,
  include_xy = FALSE,
  force_df = FALSE,
  full_colnames = FALSE,
  stack_apply = FALSE,
  summarize_df = FALSE,
  quantiles = NULL,
  progress = TRUE,
  max_cells_in_memory = 3e+07,
  grid_compat_tol = 0.001,
  colname_fun = NULL
)

Arguments

`x`	a `RasterLayer`, `RasterStack`, `RasterBrick`, or `SpatRaster`
`y`	a `sf`, `sfc`, `SpatialPolygonsDataFrame`, or `SpatialPolygons` object with polygonal geometries
`fun`	an optional function or character vector, as described below
`...`	additional arguments to pass to `fun`
`weights`	a weighting raster to be used with the `weighted_mean` and `weighted_sum` summary operations or a user-defined summary function. When `weights` is set to `'area'`, the cell areas of `x` will be calculated and used as weights.
`append_cols`	when `fun` is not `NULL`, an optional character vector of columns from `y` to be included in returned data frame.
`coverage_area`	if `TRUE`, output pixel `coverage_area` instead of `coverage_fraction`
`default_value`	an optional value to use instead of `NA` in `x`
`default_weight`	an optional value to use instead of `NA` in `weights`
`include_area`	if `TRUE`, and `fun` is `NULL`, augment the data frame for each feature with a column for the cell area. If the units of the raster CRS are degrees, the area in square meters will be calculated based on a spherical approximation of Earth. Otherwise, a Cartesian area will be calculated (and will be the same for all pixels.) If `TRUE` and `fun` is not `NULL`, add `area` to the data frame passed to `fun` for each feature.
`include_cell`	if `TRUE`, and `fun` is `NULL`, augment the data frame for each feature with a column for the cell index (`cell`). If `TRUE` and `fun` is not `NULL`, add `cell` to the data frame passed to `fun` for each feature.
`include_cols`	an optional character vector of column names in `y` to be added to the data frame for each feature that is either returned (when `fun` is `NULL`) or passed to `fun`.
`include_xy`	if `TRUE`, and `fun` is `NULL`, augment the returned data frame for each feature with columns for cell center coordinates (`x` and `y`). If `TRUE` and `fun` is not `NULL`, add `x` and `y` to the data frame passed to `fun` for each feature.
`force_df`	always return a data frame instead of a vector, even if `x` has only one layer and `fun` has length 1
`full_colnames`	include the names of `x` and `weights` in the names of the data frame for each feature, even if `x` or `weights` has only one layer. This is useful when the results of multiple calls to `exact_extract` are combined with `cbind`.
`stack_apply`	if `TRUE`, apply `fun` independently to each layer or `x` (and its corresponding layer of `weights`, if provided.) The number of layers in `x` and `weights` must equal each other or `1`, in which case the single layer raster will be recycled. If `FALSE`, apply `fun` to all layers of `x` (and `weights`) simultaneously.
`summarize_df`	pass values, coverage fraction/area, and weights to `fun` as a single data frame instead of separate arguments.
`quantiles`	quantiles to be computed when `fun = 'quantile'`
`progress`	if `TRUE`, display a progress bar during processing
`max_cells_in_memory`	the maximum number of raster cells to load at a given time when using a named summary operation for `fun` (as opposed to a function defined using R code). If a polygon covers more than `max_cells_in_memory` raster cells, it will be processed in multiple chunks.
`grid_compat_tol`	require value and weight grids to align within `grid_compat_tol` times the smaller of the two grid resolutions.
`colname_fun`	an optional function used to construct column names. Should accept arguments `values` (name of value layer), `weights` (name of weight layer), `fun_name` (value of `fun`), `fun_value` (value associated with `fun`, for `⁠fun %in% c('quantile', 'frac', 'weighted_frac)⁠` `nvalues` (number of value layers), `weights` (number of weight layers)

Details

exact_extract extracts the values of cells in a raster that are covered by polygonal features in a simple feature collection (sf or sfc) or SpatialPolygonDataFrame, as well as the fraction or area of each cell that is covered by the feature. Pixels covered by all parts of the polygon are considered. If an (invalid) multipart polygon covers the same pixels more than once, the pixel may have a coverage fraction greater than one.

The function can either return pixel values directly to the caller, or can return the result of a predefined summary operation or user-defined R function applied to the values. These three approaches are described in the subsections below.

Returning extracted values directly

If fun is not specified, exact_extract will return a list with one data frame for each feature in the input feature collection. The data frame will contain a column with cell values from each layer in the input raster (and optional weighting raster) and a column indicating the fraction or area of the cell that is covered by the polygon.

If the input rasters have only one layer, the value and weight columns in the data frame will be named values or weights. When the input rasters have more than one layer, the columns will be named according to names(x) and names(weights). The column containing pixel coverage will be called coverage_fraction when coverage_area = FALSE, or coverage_area when coverage_area = TRUE. Additional columns can be added to the returned data frames with the include_area, include_cell, and include_xy arguments.

If the output data frames for multiple features are to be combined (e.g., with rbind), it may be useful to include identifying column(s) from the input features in the returned data frames using include_cols.

Predefined summary operations

Often the individual pixel values are not needed; only one or more summary statistics (e.g., mean, sum) is required for each feature. Common summary statistics can be calculated by exact_extract directly using a predefined summary operation. Where possible, this approach is advantageous because it allows the package to calculate the statistics incrementally, avoiding the need to store all pixel values in memory at the same time. This allows the package to process arbitrarily large data with a small amount of memory. (The max_pixels_in_memory argument can be used to fine-tune the amount of memory made available to exact_extract.)

To summarize pixel values using a predefined summary option, fun should be set to a character vector of one or more operation names. If the input raster has a single layer and a single summary operation is specified, exact_extract will return a vector with the result of the summary operation for each feature in the input. If the input raster has multiple layers, or if multiple summary operations are specified, exact_extract will return a data frame with a row for each feature and a column for each summary operation / layer combination. (The force_df option can be used to always return a data frame instead of a vector.)

The following summary operations are supported:

min - the minimum non-NA value in any raster cell wholly or partially covered by the polygon
max - the maximum non-NA value in any raster cell wholly or partially covered by the polygon
count - the sum of fractions of raster cells with non-NA values covered by the polygon
sum - the sum of non-NA raster cell values, multiplied by the fraction of the cell that is covered by the polygon
mean - the mean cell value, weighted by the fraction of each cell that is covered by the polygon
median - the median cell value, weighted by the fraction of each cell that is covered by the polygon
quantile - arbitrary quantile(s) of cell values, specified in quantiles, weighted by the fraction of each cell that is covered by the polygon
mode - the most common cell value, weighted by the fraction of each cell that is covered by the polygon. Where multiple values occupy the same maximum number of weighted cells, the largest value will be returned.
majority - synonym for mode
minority - the least common cell value, weighted by the fraction of each cell that is covered by the polygon. Where multiple values occupy the same minimum number of weighted cells, the smallest value will be returned.
variety - the number of distinct values in cells that are wholly or partially covered by the polygon.
variance - the population variance of cell values, weighted by the fraction of each cell that is covered by the polygon.
stdev - the population standard deviation of cell values, weighted by the fraction of each cell that is covered by the polygon.
coefficient_of_variation - the population coefficient of variation of cell values, weighted by the fraction of each cell that is covered by the polygon.
weighted_mean - the mean cell value, weighted by the product of the fraction of each cell covered by the polygon and the value of a second weighting raster provided as weights
weighted_sum - the sum of defined raster cell values, multiplied by the fraction of each cell that is covered by the polygon and the value of a second weighting raster provided as weights
weighted_stdev - the population standard deviation of cell values, weighted by the product of the fraction of each cell covered by the polygon and the value of a second weighting raster provided as weights
weighted_variance - the population variance of cell values, weighted by the product of the fraction of each cell covered by the polygon and the value of a second weighting raster provided as weights
frac - returns one column for each possible value of x, with the the fraction of defined raster cells that are equal to that value.
weighted_frac - returns one column for each possible value of x, with the fraction of defined cells that are equal to that value, weighted by weights.

In all of the summary operations, NA values in the the primary raster (x) raster are ignored (i.e., na.rm = TRUE.) If NA values occur in the weighting raster, the result of the weighted operation will be NA. NA values in both x and weights can be replaced on-the-fly using the default_value and default_weight arguments.

User-defined summary functions

If no predefined summary operation is suitable, a user-defined R function may be provided as fun. The function will be called once for each feature and must return either a single value or a data frame. The results of the function for each feature will be combined and returned by exact_extract.

The simplest way to write a summary function is to set argument summarize_df = TRUE. (For backwards compatibility, this is not the default.) In this mode, the summary function takes the signature ⁠function(df, ...)⁠ where df is the same data frame that would be returned by exact_extract with fun = NULL.

With summarize_df = FALSE, the function must have the signature ⁠function(values, coverage_fractions, ...)⁠ when weights are not used, and ⁠function(values, coverage_fractions, weights, ...)⁠ when weights are used. If the value and weight rasters each have a single layer, the function arguments will be vectors; if either has multiple layers, the function arguments will be data frames, with column names taken from the names of the value/weight rasters. Values brought in through the include_xy, include_area, include_cell, and include_cols arguments will be added to the values data frame. For most applications, it is simpler to set summarize_df = TRUE and work with all inputs in a single data frame.

Value

a vector, data frame, or list of data frames, depending on the type of x and the value of fun (see Details)

Examples

rast <- raster::raster(matrix(1:100, ncol=10), xmn=0, ymn=0, xmx=10, ymx=10)
poly <- sf::st_as_sfc('POLYGON ((2 2, 7 6, 4 9, 2 2))')

# named summary operation on RasterLayer, returns vector
exact_extract(rast, poly, 'mean')

# two named summary operations on RasterLayer, returns data frame
exact_extract(rast, poly, c('min', 'max'))

# named summary operation on RasterStack, returns data frame
stk <- raster::stack(list(a=rast, b=sqrt(rast)))
exact_extract(stk, poly, 'mean')

# named weighted summary operation, returns vector
weights <- raster::raster(matrix(runif(100), ncol=10), xmn=0, ymn=0, xmx=10, ymx=10)
exact_extract(rast, poly, 'weighted_mean', weights=weights)

# custom summary function, returns vector
exact_extract(rast, poly, function(value, cov_frac) length(value[cov_frac > 0.9]))

[Package exactextractr version 0.10.0 Index]