CODEWITHSUNDEEP
rterra
djas
djas

Reputation: 1013

How to aggregate categorical SpatRaster

I am new to the terra package. I am trying to aggregate a categorical raster (or a SpatRaster, to be more precise) that has only one layer. The result should be a raster with as many layers as categories in the original raster; the cells' values should have the number of original (smaller) cells in each category.

Here's an example showing what I trying to achieve:

    library(terra)

    # the SpatRaster with 3 categories: 1, 2, 3
    set.seed(0)
    r <- rast(nrows=4, ncols=4)
    values(r) <- sample(3, ncell(r), replace=TRUE)

    # create one layer per category with binary indicators
    r1 <- subst(r, from=c(2,3), 0)
    r2 <- subst(r, from=c(1,3), 0); r2 <- subst(r2, from=2, 1)
    r3 <- subst(r, from=c(1,2), 0); r3 <- subst(r3, from=3, 1)
    
    # stack
    s <- c(r1, r2, r3)
    names(s) <- c("cat1", "cat2", "cat3")
    
    # aggregate
    a <- aggregate(s, fact = 2, fun = "sum")

This works for this example. But it is not practical nor efficient. It is probably(?) not feasible with large raster datasets (orders of magnitude 1GB-10GB) and many categories.

So, how would a terra pro do this?

Upvotes: 3

Views: 991

Answers (1)

Robert Hijmans
Robert Hijmans

Reputation: 47026

Here is a more streamlined approach

Your example data:

library(terra)
set.seed(0)
r <- rast(nrows=4, ncols=4)
values(r) <- sample(3, ncell(r), replace=TRUE)

solution:

s <- segregate(r)
a <- aggregate(s, 2, sum)

It is also possible to do this:

b <- list()
for (i in 1:3) {
    b[[i]] <- aggregate(r, 2, function(v) sum(v==i,na.rm=TRUE))
}
b <- rast(b)

Which you can also write like this (I wouldn't recommend it)

bb <- lapply(1:3, \(i) aggregate(r, 2, \(v) sum(v==i,na.rm=TRUE))) |>
      rast()

Upvotes: 2

Related Questions