Oops -- regarding my last reply, I see now that the approach with 'extract' using 'weights' was actually your original method at the start of this thread -- so, apparently it is slow. I don't know if the spsample+extract approach would be faster or not.