The weakest link hypothesis for adaptive capacity: An empirical test