Calculating pixelwise correlation between two multi-layer rasters in R with the terra package

Question

Is there a function in the R terra package that is equivalent to the raster package's corLocal() syntax below

corLocal(x, y, method=c("spearman"), test=FALSE)

I'd like to calculate the pixelwise correlation between two multi-layer SpatRaster objects. The SpatRasters have the same extent, resolution, and number of layers. I also need to find the p-values for each pixel. I've found layerCor(), but it only takes in one SpatRaster as input.

Thanks

Answer 1

With terra 1.7-69 you can use xapp

This currently is the development version of therra that you can install with:

install.packages('terra', repos='https://rspatial.r-universe.dev')

Example data

library(terra)
#terra 1.7.69
r <- rast(ncols=10, nrows=10, nlyr=5)
set.seed(1)
r <- init(r, runif)
s <- init(r, runif)

Solution

x <- xapp(r, s, fun=cor)
#class       : SpatRaster 
#dimensions  : 10, 10, 1  (nrow, ncol, nlyr)
#resolution  : 36, 18  (x, y)
#extent      : -180, 180, -90, 90  (xmin, xmax, ymin, ymax)
#coord. ref. : lon/lat WGS 84 (CRS84) (OGC:CRS84) 
#source(s)   : memory
#name        :      lyr.1 
#min value   : -0.9923990 
#max value   :  0.9808943 

Likewise, you can use cor.test, but you have to extract the value(s) you want from what that function returns. For example

xtest <- xapp(r, s, fun=\(x, y) cor.test(x, y)$p.value)

Unlike cor, cor.test fails when the input is NA. To work around this you can do

naproof.cor.test <- function(x, y) {
    if (any(is.na(x)) || any(is.na(y))) return(NA)
    cor.test(x, y)$p.value
}

xtest <- xapp(r, s, naproof.cor.test)

And you can get both the correlation coefficient and the p-value in one step like this

my.cor <- function(x, y) {
    if (any(is.na(x)) || any(is.na(y))) return(c(NA, NA))
    unlist(cor.test(x, y)[c("estimate", "p.value")])
}

z <- xapp(r, s, my.cor)

z
#class       : SpatRaster 
#dimensions  : 10, 10, 2  (nrow, ncol, nlyr)
#resolution  : 36, 18  (x, y)
#extent      : -180, 180, -90, 90  (xmin, xmax, ymin, ymax)
#coord. ref. : lon/lat WGS 84 (CRS84) (OGC:CRS84) 
#source(s)   : memory
#names       : estimate.cor,      p.value 
#min values  :   -0.9923990, 0.0007945862 
#max values  :    0.9808943, 0.9992525680