PCA using raster datasets in R

Question

I have several large rasters that I want to process in a PCA (to produce summary rasters). I have seen several examples whereby people seem to be simply calling prcomp or pr

Answer 1

Answer to my own question: I ended up doing something slightly different: rather than using every raster cell as input (very large dataset), I took a sample of points, ran the PCA and then saved the output model so that I could make predictions for each grid cell…maybe not the best solution but it works:

rasters <- stack(myRasters)

sr <- sampleRandom(rasters, 5000) # sample 5000 random grid cells

# run PCA on random sample with correlation matrix
# retx=FALSE means don't save PCA scores 
pca <- prcomp(sr, scale=TRUE, retx=FALSE) 

# write PCA model to file 
dput(pca, file=paste("./climate/", name, "/", name, "_pca.csv", sep=""))

x <- predict(rasters, pca, index=1:6) # create new rasters based on PCA predictions