Quantifying uncertainty due to sampling error in a social vulnerability index: A Monte Carlo approach