Edge influence and population aggregation: On point and interval statistical performances of Morisita patchiness index estimators in different sampling schemes

Usually, individuals in biological populations are spatially clumped and the measures used to quantify this pattern has caught attention from ecologists due to its many implications, from species diversity to epidemiology, landscape use and ecosystem services. The Morisita patchiness index (I-delta) is one measure of population aggregation and has proven to be useful in quadrat sampling, although the effects of edge influence and the study site shape over the index estimation are a pending issue. Thus, in this paper we aim at evaluating the statistical performances of the moment estimator as well as the bias-corrected, non-parametric bootstrap of I-delta for completely random and aggregated spatial patterns, considering the presence or absence of an edge effect in convex study sites using completely random selection of sampling units in a lattice. We used computer simulations to both creating the fragment and sampling it, whilst exerting increasing sampling efforts in lattices of different shapes (quadrats, circlets, rectangles and hexagonal) and sizes. We considered increasing levels of aggregation using both, Poisson and modified Thomas point processes of homogeneous and inhomogeneous intensities to respectively modeling the absence or presence of the edge influence. We found no substantial difference in point performance when comparing study site shapes in scenarios of no edge influence, although synergetic effects between aggregation and edge avoidance further increased bias in I-delta estimation. Despite useful for obtaining confidence intervals, resampling methods yielded the same results observed for the classical estimator. In addition, we observed that sampling effort was not robust against lack of precision and accuracy and, despite prior beliefs, quadrats did not outperform circlets or rectangles; indeed, they all had essentially the same point and interval performances. Finally, the best estimation performances occurred for big sampling units of any shape when compared to small ones, thus, given an appropriate scale, one should prefer bigger quadrats whenever conducting a field study. Our results ultimately pointed out that edge avoidance by a species drives its patchiness estimation even when they are randomly distributed into the core area, which might be the reason behind for so many claims of population aggregation. (AU)