How to replace NA's in a raster object

Question

I need to replace the NA's in the raster object (r) from the example below.

library(raster)
filename <- system.file("external/test.grd", package="raster")
r <- raster(filename)

I also tried to remove these these (and place the result in a data.frame), but to no avail.

dfr <- as.data.frame(r, na.rm=T)
summary(dfr)
# test       
# Min.   : 128.4  
# 1st Qu.: 293.2  
# Median : 371.4  
# Mean   : 423.2  
# 3rd Qu.: 499.8  
# Max.   :1805.8  
# NA's   :6097

Blimey

Yes, summary(as.data.frame(r, na.omit=T)) says that there are 6097 NA's. – Blimey

Welcome to StackOverflow. +1 for reproducible example. – Maddeu

As you can see in the answer, na.omit is a function not an argument of as.data.frame. – Ghana

I saw that, @seancarmody, thanks! – Blimey

Try SpatialPixelsDataFrame for sparse grids without the NAs, but this is pretty open ended, what do you really need? – Urania

Answer 1

I'm not sure it makes sense to remove NA values from a raster object, but you can easily replace it.

For example:

oldpar <- par(mfrow=c(1, 2))
plot(r)
r[is.na(r)] <- 250
plot(r)
par(oldpar)

---

If you really want to, you can extract the raster values into a vector and then remove the NA values. (Although, since you lose the spatial information, I can't see how this can be helpful.)

r <- raster(filename)

r <- values(r)
head(r)
[1] NA NA NA NA NA NA

head(na.omit(r))
[1] 633.686 712.545 654.162 604.442 857.256 755.506

Maddeu

Answer 2

Here are three ways in which you can do this with terra

library(terra)
f <- system.file("ex/elev.tif", package="terra")
r <- rast(f)
x <- classify(r, cbind(NA, -99))
y <- subst(r, NA, -99)
z <- ifel(is.na(r), -99, r)

With raster, the best (memory safe approach that also works for large files) is to use reclassify:

library(raster)
filename <- system.file("external/test.grd", package="raster")
r <- raster(filename)
rna <- reclassify(r, cbind(NA, 250))

Anaclitic

This doesn't seem to work for NaN (i.e. reclassify(r, cbind(NaN, NA))) – Gogh

Using the "in-place" function set.values, we can also write: set.values(r, which(is.na(values(r))), -99), which is perhaps the fastest way to substitute values in a terra raster. – Marable

That may be, but using values is not memory safe, so I won't recommend it. – Anaclitic

What exactly do you mean by "not memory-safe"? Are you talking about the fact that all the other rasters pointing to the modified raster are also modified? Or is there another danger? – Marable

values attempts to read all raster cell values into memory. That may not be possible. – Anaclitic