Technical Comments

Response to Comment on “Conspecific Negative Density Dependence and Forest Diversity”

Science  26 Oct 2012:
Vol. 338, Issue 6106, pp. 469
DOI: 10.1126/science.1225996


Dickie, Hurst, and Bellingham question some of the methods of our recent study on conspecific density dependence in forests. Here, we reanalyze our data set with the inclusion of joint absence plots of each species. We find that our results are robust to further analyses and that patterns of abundance and richness correlate with our measure of density dependence, supporting our original conclusions.

Forest systems provide key tests for exploring hypotheses for the maintenance of diversity in ecological communities. One major hypothesis is conspecific negative density-dependent mortality, where seedling establishment is reduced with increasing density of conspecific neighbors. The goal of our study was to analyze patterns of seedling establishment in relation to tree density across eastern U.S. forests to determine the extent and influence of density dependence on forest communities (1).

Dickie, Hurst, and Bellingham (2) argue that our analyses of seedling establishment patterns are biased and spuriously correlate with species abundance because sample plots lacking both trees and seedlings of the focal species were excluded (1). In simulations using randomly generated independent and identically distributed tree and seedling abundances, they found that excluding joint absence plots resulted in b values (the strength of density dependence estimated from the negative exponential function, S = aebT, where a is the y intercept, S is the number of seedlings, and T is the number of trees) that were significantly positively related to species abundance but not when joint absence plots were excluded (2).

Wholesale inclusion of joint absence data at the regional-cell level (2° latitude by longitude) numerically overwhelms presence data because most species do not occur everywhere (94.6% of species by cell samples are joint absences). Absence data often represent locations outside of species’ physiological range and therefore are not biologically relevant to patterns of regeneration. Additionally, sample locations in which a species is not present provide no information about the interaction between conspecific individuals (1).

However, to address Dickie et al.’s Comment and explore whether excluding joint absences biased our results, we analyzed the data with a subset of joint absence plots included when the species were known to occur locally. Specifically, we included joint absences on plots where at least one of the four nearest neighbor plots (1) (fig. S1) contained the species of interest. This provides biologically meaningful absences by including joint absences when the species is known to occur in the vicinity and is therefore within geographical and physiological range of the species. Moreover, there is a high probability that the plot is within the dispersal distance of that species. The relation of the strength of conspecific density dependence to species abundance and richness remained significant and in the same respective directions as in our original study (Fig. 1).

Fig. 1

Strength of density dependence when analyzed with joint absence U.S. Forest Service Forest Inventory and Analysis data shows the same pattern as originally reported. (A) Histogram of strength of density dependence by species-cell combination. (B) Relation between strength of density dependence and the relative abundance of the species in the regional cell (Spearman’s rank correlation, ρ = 0.138, P = 3.69 × 10−14, N = 2988). (C) Latitude relation to average strength of density dependence per regional cell (Spearman’s rank correlation, ρ = 0.313, P = 0.00093, N = 110). (D) Regression of the average strength of density dependence on species richness (correlation coefficient r2 = 0.1714, F1,108 = 22.35, P = 6.91 × 10−6).

Further, using a model-free approach, we found that the vast majority (96.7%) of seedling density is nonindependently associated with conspecific tree density with two-dimensional Kolmogorov Smirnov (2DKS) tests (3). These tests indicate that there are significant nonrandom negative relations in these bivariate data. Reanalyzing the strength of density dependence versus relative abundance with nonindependent species-cell combinations, determined by 2DKS tests, resulted in qualitatively identical results to those originally reported. Our results indicate that conspecific negative density dependence (CNDD) is widespread in forests, in support of the hypothesis that CNDD can maintain forest diversity.

Dickie et al. report an overall slight positive correlation between conspecific seedlings and trees in New Zealand (2). Local dispersal of seeds in the absence of ecological interactions will generate positive correlations. Tests of whether the correlations observed in New Zealand are more positive than the null expectation of local dispersal would require detailed information on the dispersal kernels of the individual tree species in New Zealand forests. Therefore, although negative correlations, as we observed in North America, can be confidently interpreted as resulting from negative conspecific density dependence, positive correlations could be stronger or weaker than null expectations and are therefore ambiguous as tests of conspecific density dependence. In North America, we observed weaker CNDD in higher latitudes. The generality of this pattern across the globe in general and specifically whether the slight positive correlation of seedlings with conspecific adults observed in the climates of New Zealand is consistent with this pattern is an interesting possibility that will require further research. Our results indicate that positive conspecific density dependence is much less common than negative density dependence in forest systems and agree with empirical patterns reported from plant communities (4).

References and Notes

  1. Acknowledgments: The data used for these analyses are publicly available from
View Abstract


Navigate This Article