0

I am trying to do some calculation on raster using following code

library(terra)

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

r_n <- (r - global(r, fun = "min", na.rm=TRUE))/
  (global(r, fun = "max", na.rm=TRUE) - global(r, fun = "min", na.rm=TRUE))

It throws the following error

Error in data.frame(value, row.names = rn, check.names = FALSE) : duplicate row.names: 374, 356, 348, 378,...

But if I do it using raster R package, it runs fine like

library(raster)

r_raster <- stack(r)
r_n_raster <- (r_raster - minValue(r_raster))/
  (maxValue(r_raster) - minValue(r_raster))

What should I do to make the code run with terra package?

UseR10085
  • 7,120
  • 3
  • 24
  • 54
  • The first thing you should do when encountering an error is to do the computations line by line so that you can see where the error occurs. It then becomes much easier to understand where your code breaks, and why. If you ask a question you should include the output (error message) exactly as-is and where it occurs. – Robert Hijmans Jun 23 '22 at 15:59

1 Answers1

2

This happen because the ouput of global is a data.frame.You can simply convert the output of global to numeric with as.numeric or use the original summary function on the values of the raster:

rmin <- as.numeric(global(r, fun = "min", na.rm=TRUE))
rmax <- as.numeric(global(r, fun = "max", na.rm=TRUE))

r_n <- (r - rmin)/
  (rmax - rmin)

or

rmin <-  min(values(r), na.rm=T)
rmax <-  max(values(r), na.rm=T)

EDIT

A microbenchmark of the two proposed method with the example data:

library(microbenchmark)
microbenchmark(global=as.numeric(global(r, fun = "min", na.rm=TRUE)),
               summary=min(values(r),na.rm=T))

output

Unit: microseconds
    expr    min      lq     mean median      uq     max neval cld
  global 2071.2 2157.40 2437.687 2223.6 2309.85 21725.9   100   b
 summary  729.2  756.35  844.381  794.4  824.15  5579.1   100  a
Elia
  • 2,210
  • 1
  • 6
  • 18
  • It will be great if you can show which option is faster `global` or `values`? Then it will be easier to choose one option over other. – UseR10085 Jun 23 '22 at 04:39
  • 1
    The "values" method may be faster when it works, but it is not a generally safe method. It should not be recommended except for the rare situations where minimal performance improvements are essential. – Robert Hijmans Jun 23 '22 at 15:57