One of the peculiarities of statistics is that sampling error is not dependent on the size of the population from which the sample is drawn. Sampling error decreases relatively little after the 30th sample. However, if it is known beforehand that a population consists of more than one subpopulation, it may be necessary to "stratify" the sample to ensure that each of the subpopulations is sampled. For example, if you wish to sample a species that includes two subspecies, you may need to take a sample of 30 from each subspecies. Of course, if you wish to measure traits that are rare in a population, you may have to increase your sample size beyond 30 to take that rarity into account.