Spatial point-pattern methods can be used by ecologists to infer competitive or facilitative interactions among plants. Interactions between neighboring plants are inferred from point patterns of distances between plant centers. However, the spatial extent of effects of large but distant individuals, such as a parent’s canopy on clumps of seedlings beneath it, or belowground root interactions, may be difficult to detect with point-patterns. Methods that examine the distribution of vegetation cover can improve ecologists’ ability to make inferences about local interactions.
We have developed a new method, the neighbor-area function, that analyzes the continuous distribution of vegetation cover For each individual plant on a landscape, this function integrates the total area occupied by neighboring plant canopies as a function of radius from the canopy center of the focal plant. Summed over all individuals, this measures the mean neighboring canopy density across the landscape.
As a case study, we applied this method to mapped distributions of two shrub species with contrasting spatial distributions in Joshua Tree National Park. We compared results from the neighbor-area function with Ripley’s K function, which is a commonly used point-pattern method, and the mark-correlation function, which permits inclusion of a point attribute, such as size.
Results/Conclusions
Results from the point pattern methods were different from the neighbor-area method. Ripley’s K confirmed a locally aggregated point pattern of plant locations for Ambrosia dumosa (Asteraceae), and a random distribution of locations for Larrea tridentata (Zygophyllaceae). However, the neighbor-area method found that the distribution of Ambrosia canopy densities was less aggregated at all spatial scales than the point pattern of plant centers. For Larrea, the distribution of canopy area was overdispersed, not random. For both species, the neighbor-area function and the univariate mark-correlation function (where plant size is the mark) are identical at large distances, as expected, but differ at spatial scales less than the mean canopy diameter of each species, because the neighbor-area function can detect canopy overlap.
This new neighbor-area method, presented here as the univariate case of intraspecific interactions, can be extended to a multivariate analysis of interspecific interactions by integrating the neighboring canopies of every other species in the community separately. Furthermore, because this new method works with continuous distributions of the density of vegetation across a landscape, our implementation of the neighbor-area method can be directly used with image data, providing a potentially powerful tool to analyze vegetation distributions at the landscape scale.