Balancing feasibility and statistical rigor while avoiding bias in estimating plant population sizes

Background/Question/Methods

Plant ecologists commonly measure density (e.g., plants/m2) rather than estimating the actual size of a population. However, for some applications, particularly in conservation, we need an estimate of actual population size. To obtain such an estimate, an unbiased estimate of density across the entire estimated area of suitable habitat is needed. It is rarely feasible to sample the entire area of suitable habitat with randomly placed plots. At the other extreme, haphazardly looking for plants is biased, and the bias worsens the more the searcher knows about the species. We developed a method that is both logistically feasible and unbiased for estimating plant population size, and successfully used it for a rare cactus species, Coryphantha ramillosa. We defined suitable habitat as the entire area of Big Bend National Park that had elevation, topography, geology, soil, and vegetation similar to that of the known occurrences of the species. Within this pre-selected sampling universe, we randomly located transects in a N-S direction. All plants of C. ramillosa were counted within a pre-selected distance of each transect.

Results/Conclusions

Simulation modeling confirmed that if (and only if) transects are located truly randomly, as ours were, and the maximum search distance from each transect strictly maintained, the estimate of density is unbiased. Using transects is much more efficient in time and effort than using randomly-located plots, but does sacrifice strict randomness, as plants along a given transect are not strictly independent. Another limitation is that the results apply only to the pre-selected sampling universe, which was all apparently suitable habitat for this species in Big Bend National Park. We surveyed a total of 5.94 ha out of an estimated 618.08 ha of suitable habitat and found 40 plants, which yielded an estimate of 4164 plants of this species in Big Bend National Park. This method can be readily adapted for many different species. Since transects are commonly used by plant ecologists, this method should also be widely acceptable.