Conspecific Negative Density Dependence and Forest Diversity

See allHide authors and affiliations

Science  18 May 2012:
Vol. 336, Issue 6083, pp. 904-907
DOI: 10.1126/science.1220269

Keep Your Distance

Conspecific negative density dependence (CNDD), whereby the abundance of a species is limited by negative interactions between individuals of the same species, is thought to have an important influence on the composition and dynamics of forest communities, but studies have generally been limited to few species and small areas. Johnson et al. (p. 904) analyzed CNDD in over 200,000 plots from a database of more than 3 million individuals of 151 species spanning 4 million square kilometers across forests in the eastern United States and found that the strength of CNDD strongly predicted the relative abundance of tree species. Because tree seedlings are unlikely to become established where conspecific adults are common, CNDD may provide a general mechanism maintaining diversity in forests.


Conspecific negative density-dependent establishment, in which local abundance negatively affects establishment of conspecific seedlings through host-specific enemies, can influence species diversity of plant communities, but the generality of this process is not well understood. We tested the strength of density dependence using the United States Forest Service’s Forest Inventory and Analysis database containing 151 species from more than 200,000 forest plots spanning 4,000,000 square kilometers. We found that most species experienced conspecific negative density dependence (CNDD), but there was little effect of heterospecific density. Additionally, abundant species exhibited weaker CNDD than rarer species, and species-rich regions exhibited stronger CNDD than species-poor regions. Collectively, our results provide evidence that CNDD is a pervasive mechanism driving diversity across a gradient from boreal to subtropical forests.

The factors that affect the distribution and abundance of species include abiotic influences, predators, parasites, competition, and mutualisms. A major mechanism proposed for the maintenance of diversity is conspecific negative density-dependent mortality, whereby proximity to adults of the same species reduces seedling survival through attack by host-specific enemies [also known as the Janzen-Connell hypothesis (JCH) (1, 2)]. The JCH was proposed with particular reference to species-rich tropical forests. The relative importance of the JCH in influencing diversity in plant communities more widely has ramifications for accurate modeling and appropriate conservation and management.

Most studies of conspecific negative density dependence (CNDD) in plant communities have focused on a single site or a single species, [however, see (35) for comparisons of CNDD from multiple tropical forest sites]. Recent studies of individual tropical and subtropical forest sites suggest that forest composition is influenced by CNDD at the individual tree scale, which in turn influences the relative abundance of species at the community scale [(68), but see (9)]. In comparison, less is known about the efficacy of this mechanism in forests outside the tropical sites that were examined. Some temperate forest species display clear CNDD (1013), and a comparative analysis of 13 temperate and tropical forest sites suggests that the proportion of species experiencing CNDD is not related to latitude (10). Despite the potential importance of CNDD in forest dynamics, we have a limited understanding of its role across forest communities.

Assessing CNDD at broader spatial scales is necessary for a more comprehensive understanding of the importance of CNDD to structuring forest communities. Here, we analyzed the U.S. Forest Service Forest Inventory and Analysis (FIA) database ( (14). The volume and scope of the FIA data provided an opportunity to explore ecological patterns at wider scales than before. Plots were located from the Canadian border south to Florida and from the Atlantic coast west to the 100th meridian, approximately the center of the continental United States. We analyzed data on tree composition from fully forested plots (circular area of 168.33 m2) and seedling establishment from nested plots (13.50 m2) (fig. S1). FIA defines stems larger than 12.7 cm in diameter at breast height (dbh) as trees, whereas individuals smaller than 2.54-cm dbh and taller than 30.5 cm (12.24 cm for conifer species) are defined as seedlings. In total, we analyzed data from 207,444 paired tree and seedling plots, which contained 1,334,347 trees and 1,709,314 seedlings representing 151 species. Data were aggregated into 2° latitude-by-longitude regional cells to examine regional patterns of CNDD (fig. S2). Species richness per cell ranged from 18 to 119 (mean ± SE: 72.97 ± 2.26), and species richness per plot ranged from 1 to 21 (4.2 ± 0.005).

As CNDD is manifested by a negative relation between the abundance of conspecific trees versus seedlings, we estimated the strength of CNDD by regressing seedling abundance (S) on tree abundance (T) using generalized linear regression. The specific equation used to estimate the strength of CNDD by maximum likelihood was the negative exponential function S = aebT, where a is the y intercept and b controls the inflection of the curve (fig. S3). This provides a single parameter, b, to represent the strength of CNDD, with larger, more negative values of b indicating stronger CNDD. We repeated this analysis using basal area, a proxy for biomass, as an alternative measure of tree abundance.

In separate analyses, we fit the model for both conspecific and heterospecific trees for each species in each regional cell. We restricted our analyses to species that occurred in a minimum of 30 plots per cell where at least one plot had co-occurrence of trees with conspecific seedlings. Regional cells lacking co-occurrence of trees and seedlings represent the strongest cases of CNDD, making our analyses conservative. Moreover, the FIA definition of a seedling excludes most species’ first-year seedlings and, therefore, does not capture early mortality, also making our analyses conservative. We examined the effect of heterospecific density to determine the role of density-dependent effects that were independent of conspecific abundance (for instance, shading by canopy trees or competition for soil nutrients).

Species by regional cell estimates of conspecific density dependence were significantly more negative and variable than heterospecific density dependence (Fig. 1, A and B), indicating much stronger and more varied effects of conspecific tree density on seedling establishment than heterospecific tree density. Analyses of basal area produced qualitatively similar results (Fig. 1, C and D). The stronger negative effect of conspecifics on seedling density occurred, despite the fact that seed density is expected to be higher near conspecific trees (15). Moreover, the reduction in conspecific seedling density cannot be totally explained by shading or soil-nutrient depletion because we did not find equivalent responses to heterospecific tree density or basal area (Fig. 1).

Fig. 1

Seedlings are significantly more likely to experience negative effects on establishment from conspecific neighbors than from heterospecific neighbors. The strength of density-dependent establishment was estimated for each species by regional cell combination on seedlings by (A) conspecific tree count (2848 combinations) and (B) heterospecific tree count (3229 combinations); these distributions were significantly different (Wilcoxon rank sum test, W = 790,869, P < 2.2 × 10−16). We performed the same analysis with (C) conspecific basal area (2768 combinations) and (D) heterospecific basal area (3234 combinations) and found that these distributions followed the same patterns and are significantly different (Wilcoxon rank sum test, W = 1,362,121, P < 2.2 × 10−16).

Conspecific negative density dependence of seedling establishment could provide a strong local filter favoring species diversity. For example, species experiencing strong CNDD will have fewer opportunities to establish, potentially reducing their relative abundance. Conversely, species that exhibit weak CNDD could establish near conspecifics, thus allowing them to increase in relative abundance. We tested the prediction that the strength of density dependence influences relative tree abundance and found that species undergoing strong CNDD had reduced relative abundance in that regional cell relative to species experiencing weak CNDD (Fig. 2). We examined the intraspecific pattern of CNDD across the range of all species and found that the majority of species behave similarly, where individual species experiencing stronger CNDD are less common regionally (fig. S4). In contrast, heterospecific density dependence had a far weaker relationship with regional relative abundance, as expected with little variation in the strength of heterospecific density dependence. A gradient in the strength of CNDD and mainly weak heterospecific density dependence is consistent with observations at the Forest Dynamics Plot on Barro Colorado Island, Panama (7, 8, 16), and in greenhouse trials of abundant temperate species (11). Whereas our data represent one slice in time of longer-term community dynamics (17), our results conform to a recent theory (8, 18, 19) that predicts that the equilibrium relative abundance of a species will covary with the strength of conspecific density dependence.

Fig. 2

Species experiencing strong conspecific density dependence have reduced relative abundance. The strength of conspecific density dependence had a significant positive correlation with species relative abundance (Spearman’s rank correlation, ρ = 0.3978, P < 2.2 × 10−16; N = 2848 species by regional cell combinations). The dashed line represents the median of the strength of conspecific density dependence. Heterospecific density dependence (fig. S12) had a weaker trend in the opposite direction than conspecific density dependence (Spearman’s rank correlation, ρ = –0.1278, P = 3.121 × 10−13; N = 3229 species by regional cell combinations). See fig. S5 for the corresponding conspecific basal area relation.

The average strength of conspecific density dependence was strongly correlated with tree species richness at the regional cell level (Fig. 3). This finding suggests that local biotic interactions are underlying regional species richness, in contrast to the prevailing explanations of the gradient in tree-species richness in North America, which focus primarily on physical aspects of the environment [such as temperature and precipitation (20)]. Species-level differences are discounted by neutral theory as unnecessary to describe the patterns of species abundances (21). Additionally, a recent statistical analysis has called into question the necessity of local interactions to describe patterns of diversity (22). Our results run counter to these arguments, as we found support for regional species richness patterns being driven by local species-specific ecological interactions and a local mechanism to explain variation in regional species richness.

Fig. 3

Regional species richness is predicted by the strength of negative density dependence. The regional strength of conspecific density dependence, the arithmetic mean of all density dependence estimates in each regional cell, had a significant negative linear relation with species richness (correlation coefficient r2 = 0.42, F1,106 = 75.4, P = 5 × 10−14; N = 108 regional cells) such that species richness tended to increase when negative density dependence was strongest. There was no significant relation between regional strength of heterospecific density dependence and species richness (r2 = 0.004, F1,106 = 0.5, P = 0.48, N = 108 regional cells). See fig. S6 for the corresponding conspecific basal area relation.

It is possible that the patterns found here were generated by mechanisms unrelated to conspecific density dependence that could create spatial separation of adults and conspecific seedlings [e.g., timber harvesting, succession, the mass effect (23)]. For example, recruitment differences between early successional and late successional species could imitate patterns of CNDD in forests. To test whether CNDD varies with forest age, we reanalyzed the data set by stratifying the data into early (0 to 39 years), middle (40 to 79 years), and later (80+ years) successional forests. The patterns of CNDD were robust and consistent between age classes, indicating that our results are not contingent on successional dynamics or indirectly on timber harvesting, which has the effect of setting back forest age (figs. S8 to S11).

Janzen (1) and Connell (2) originally hypothesized that CNDD generated by host-specific seed predators could help maintain the high species richness in tropical forests. We found that CNDD is a strong mechanism maintaining species richness in eastern U.S. forests, but CNDD may also explain the latitudinal gradient in species richness if CNDD becomes stronger with decreasing latitude. We tested this hypothesis in eastern North America, where there is a latitudinal gradient of tree species richness that peaks in the southern Appalachian region (20). We found evidence that CNDD could maintain this gradient in tree species richness, as the average regional strength of CNDD was significantly negatively correlated with latitude, ranging from boreal to subtropical forests (Fig. 4). Our results suggest that the strength of CNDD would increase with decreasing latitude into species-rich tropical forests.

Fig. 4

The strength of conspecific density dependence becomes stronger with decreasing latitude. (A) The regional strength of conspecific density dependence, the arithmetic mean of all conspecific density dependence estimates in a 2° latitude-by-longitude regional cell, had a significant relation with increasing latitude (Spearman’s rank correlation, ρ = 0.3, P = 0.0003; N = 108) (B). The pattern of tree species richness in the eastern United States does not follow a linear latitudinal gradient south of ~30° latitude because of reduced species richness resulting from the peninsular effect (Florida), reduced topographic heterogeneity, and arid southern Texas. See fig. S7 for the corresponding basal area relationship.

Our analyses of the FIA database provide robust evidence that CNDD is pervasive in forest communities and can significantly affect species relative abundance and species richness within and between forests. Further, our results show that species-specific processes acting on seedlings translate into patterns in the abundance and diversity of trees. Several potential interactions could generate CNDD, including intraspecific competition, autotoxicity, seed predators, and soil pathogens. Much research has demonstrated that the soil microbial community can drive CNDD in multiple plant communities, including tropical forests, temperate forests, grasslands, and sand dunes (18, 24). In particular, two studies measuring soil community feedbacks, presumably driven by soil-borne pathogens, have identified a positive relation between strength of CNDD in the greenhouse and relative abundance in the field (8, 25).

Local interactions have previously been considered a local filter on species diversity, but our findings indicate that local interactions feed back to regional species richness and abundance. Further, the prevalence of CNDD across many forest types and diverse species indicates the pervasive importance of these interactions. Our results show that CNDD is a general mechanism structuring forest communities across a wide gradient of forest types and can maintain the latitudinal gradient of tree species richness.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S12

Table S1

References (2628)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: We acknowledge contributions with database consulting from N. Long from the Indiana Univ. Research Database Complex, as well as statistical consulting from S. Dickinson and E. Hernandez at Indiana Statistical Consulting Center and C. Huang at the Indiana Univ. Department of Statistics. Helpful comments were provided by two anonymous reviewers, D. Civitello, C. Her, C. Lively, S. Mangan, S. McMahon, and S. Neumann. We also thank all of the FIA employees that collected and compiled the data that made these analyses possible. The data used for these analyses are publicly available from under FIA Data Mart. See the supplementary materials for specific dates and conditions applied to the data set.
View Abstract

Navigate This Article