Bias in vegetation databases? A comparison of stratified-random and preferential sampling.