How to analyze Pearson's correlation and VIF between raster layers using R?

Question

I have 30 raster layers (.tiff) on oceanic and climatic environmental variables from all over the world downloaded from global databases.

I need to analyze the relationship that may exist between these layers of information, prior to carrying out other analyzes (species distribution models). I have read that I could perform two tests: Pearson Correlation and Variance Inflaction Factor (VIF). The idea is to stay only with the variables (layers) that have a Pearson correlation < 0.7 and a VIF < 10.

The aim is to output a correlation matrix that shows the coefficient for each of the combinations.

I have tried this code:

# Load the physical variables and saved as raster brick
list <- list.files(path="D:/layers", pattern='.tif$', full.names=TRUE)
predsGlobal.raw <- brick(stack(list))
save(predsGlobal.raw,file="D:/Correlation/predsGlobal.raw.Rda")
layersGlob <- predsGlobal.raw@data@names

# Applied a threshold of 0.7 for Pearson correlation
vc <- vifcor(predsGlobal.raw,th=.7)
correl.groups <- cor(predsGlobal.raw) # matrix
vc@results
ex <- exclude(predsGlobal.raw,vc)
save(ex,file="D:/Correlacion/ex.Rda")

# Save in an excel
wb <- createWorkbook()
addWorksheet(wb, "predsGlobal")
writeDataTable(wb, 'predsGlobal', ex, startCol = 1, startRow = 1, colNames = TRUE, rowNames = F)
saveWorkbook(wb, file = 'D:/Correlation/predsGlobal.xlsx')

But it has not worked for me, I get the following errors:

Error in cor (predsGlobal.raw):
supply both 'x' and 'y' or a matrix-like 'x'

Error in writeDataTable(wb, "predsGlobal", ex, startCol = 1, startRow = 1,  : 
x must be a data.frame.

Answer 1

Perhaps you should try taking the values from the rasterbrick using:

data<-values(predsGlobal.raw)

Then removing any NAs and turning it into a dataframe as: data<-as.data.frame(data)

Then take the correlation: cor(data) given that it seems to only work with matrix-like data, and you can select the x and y variables from the dataframe to test for colinearity, as it seems to not work effectively with rasterbricks.

Then you can use writeDataTable.