0

I need to create some new rasters by assigning values using z <- rast(x, vals = y). I would like the new raster (z) to have NA values at the same location as the old raster (x), but with the newly assigned value(s) (y) in the non-NA cells.

What is the most efficient way to create a new raster with a matching NA pixels but with new values assigned to the non-NA pixels?

I don't see any argument to handle NA values listed in the documentation for rast(), so do I need to do a two step process where the value is first assigned to all cells and then NAs are transfered from the old raster to new?

In the following example I would like s to have the same outline as r but have a uniform value of 10 in the non-NA cells.

f <- system.file("ex/elev.tif", package="terra")
r <- rast(f)
s <- rast(r, vals=10) # This fill all rasters cells, including NAs
ia200
  • 255
  • 1
  • 9

1 Answers1

2

Here are four ways to do that:

library(terra)
f <- system.file("ex/elev.tif", package="terra")
r <- rast(f)

s1 <- classify(r, cbind(-Inf, Inf, 10))
s2 <- init(r, 10) |> mask(r)
s3 <- ifel(is.na(r), NA, 10)
s4 <- r * 0 + 10

I would expect approach 1 and 4 to be the most efficient.

Robert Hijmans
  • 40,301
  • 4
  • 55
  • 63
  • For those interested in efficiency, using both `Sys.time()` and the microbenchmark package, showed `classify()` was the fastest, followed by the multiplication method, then `init()` and `ifel()`. – ia200 Apr 15 '22 at 17:19