extractLarge: Memory-wise extraction of data from a raster or spatial...

In adamlilith/enmSdm: Tools for Modeling Niches and Distributions of Species

Memory-wise extraction of data from a raster or spatial object

Description

This function is the same as the extract function in the terra package, except that it is built to better handle cases when the points/lines/polygons where the extractions are are very large in memory. Essentially, the function is a wrapper for extract with a for loop that allows you to discard unwanted columns. Note that this function is not intended to make things fast... just feasible for cases where memory may be a problem. A few tips for memory management:

• If you do not need need specific fields from an object you will be extracting from, remove them (e.g., x$doNotNeed <- NULL).

• Likewise, removing rasters in a raster stack that you don not need may help.

• On a Windows machine, you can give R more memory using memory.limit(memory.limit() * 2^x) where n is an integer. On my machine, which has 32 GB of RAM, I use n = 29.

• Run R through the terminal, not through RStudio or the R GUI. This can reduce overhead because these programs take more memory.

• Close other programs you may not be using.

• Start in a fresh R session and load only what you need to get the extraction done. Save the output before proceeding.

• Go get a coffee, because it may take a while.

Usage

extractLarge(x, y, keep = NULL, atATime = 10000, verbose = TRUE, ...)

Arguments

x	sf, Raster*, SpatRaster, Spatial*, or SpatVector object from which data will be extracted from.
y	sf, Spatial* or SpatVector with points/lines/polygons denoting locations where data will be extracted from.
keep	Names or indices of the fields or values in x to keep. If NULL (default), then all columns will be kept. If you are unsure of these, try just extracting a small number items from y and see what the names in the output are.
atATime	Number of records in y to extract at one time. If the object x has just a few things you are extracting (e.g., a few layers for a raster set or a few fields from a polygon), you can probably use a larger number; maybe, say, 100000. If your raster set or polygon has a lot of fields, then a smaller number may be required.
verbose	Report progress (TRUE) or not (FALSE).
...	Arguments to pass to extract.

Value

Data frame.

Examples

# This example does *not* require this function because the raster
# and points are so small in memory. But it illustrates the process.

data(mad0)
data(lemurs)

ll <- c('longitude', 'latitude')
lemurs <- lemurs[ , ll]

# using extractLarge()
ex1 <- extractLarge(mad0, lemurs, atATime=10)

# using just extract()
ex2 <- extract(mad0, lemurs)

adamlilith/enmSdm documentation built on Jan. 6, 2023, 11 a.m.