Sapling growth rates reveal conspecific negative density dependence in a temperate forest

Abstract Local tree species diversity is maintained in part by conspecific negative density dependence (CNDD). This pervasive mechanism occurs in a variety of forms and ecosystems, but research to date has been heavily skewed toward tree seedling survival in tropical forests. To evaluate CNDD more broadly, we investigated how sapling growth rates were affected by conspecific adult neighbors in a fully mapped 25.6 ha temperate deciduous forest. We examined growth rates as a function of the local adult tree neighborhood (via spatial autoregressive modeling) and compared the spatial positioning of faster‐growing and slower‐growing saplings with respect to adult conspecific and heterospecific trees (via bivariate point pattern analysis). In addition, to determine whether CNDD‐driven variation in growth rates leaves a corresponding spatial signal, we extended our point pattern analysis to a static, growth‐independent comparison of saplings and the next larger size class. We found that negative conspecific effects on sapling growth were most prevalent. Five of the nine species that were sufficiently abundant for analysis exhibited CNDD, while only one species showed evidence of a positive conspecific effect, and one or two species, depending on the analysis, displayed heterospecific effects. There was general agreement between the autoregressive models and the point pattern analyses based on sapling growth rates, but point pattern analyses based on single‐point‐in‐time size classes yielded results that differed markedly from the other two approaches. Our work adds to the growing body of evidence that CNDD is an important force in temperate forests, and demonstrates that this process extends to sapling growth rates. Further, our findings indicate that point pattern analyses based solely on size classes may fail to detect the process of interest (e.g., neighborhood‐driven variation in growth rates), in part due to the confounding of tree size and age.


REGRESSIONS
. Sapling growth rate as a function of conspecific and heterospecific inverse distance-weighted basal area, with the deer exclosure omitted. A corresponding table with the deer exclosure included is provided in the main text (Table 2; see caption for additional details). Cells with brackets identify qualitative differences from Table 2 (in the main text). L. tulipifera could not be analyzed with the exclosure omitted because there was a very low number of saplings (5) outside of the exclosure.

Species
Conspecific effect Heterospecific effect n P-value Estimate Partial r2 Interpretation of differences between Table S1 (above) and Table 2 (main text): The qualitative p-value differences for C. tomentosa (conspecific and heterospecific) do not appear to be due to ecological differences between the areas inside and outside of the exclosure. Instead, we believe the discrepancy arises from the fairly subtle effects in the full analysis (low standardized slope estimates and partial r 2 in Table 2), combined with the substantial reduction in sample size when the exclosure is omitted (368 vs. 262). In support of this assertion, both slope estimates are similar (but not as pronounced) in the reduced data set, as compared to the full analysis. In addition, the mean conspecific and heterospecific IDW BA values are nearly identical across both analyses, indicating that the relative abundance of conspecific and heterospecific canopy trees is not notably different within the exclosure. Finally, an analysis of the exclosure alone (results not shown) also yields an absence of significant effects. Since both components of the full set (inside and outside of the exclosure) return non-significant results, it is highly likely that the significant effects in the full analysis are primarily due to a larger sample size.
For N. sylvatica, an analysis of the exclosure alone (results not shown) yields a highly non-significant p-value for the heterospecific effect (p=0.991), despite similar mean conspecific and heterospecific IDW BA values, suggesting that the shift from a non-significant p-value (in the full analysis) to a barely significant p-value (in the analysis omitting the exclosure) may reflect a real ecological distinction.
Note that direct comparisons between the areas inside and outside of the exclosure are not universally reported because sample sizes within the exclosure are very small for most species.

Conspecific adults
Heterospecific adults Fig. S1. Spatial patterns of two sapling categories (slow-and fast-growing) with respect to conspecific and heterospecific adult trees, with deer exclosure omitted. A corresponding figure with the deer exclosure included is provided in the main text (Fig. 1). Higher values represent increased clustering. The panels with confidence envelopes display slow values minus fast values; extensions above the envelope indicate that slow-growing saplings are more clustered than fast-growing saplings around adults (conspecific or heterospecific), and vice versa.
 The only qualitative difference between Fig. S1 (above) and Fig. 1 (main text) is that Fig. S1 indicates a tendency for slower growing Nyssa sylvatica saplings to be repelled from heterospecific adults (at intermediate and long distances), but Fig. 1 reveals no significant heterospecific effects for this species.
Supporting Information for: Sapling growth rates reveal conspecific negative density dependence in a temperate forest 3

POINT PATTERN ANALYSES (SIZE-BASED)
Conspecific adults Heterospecific adults Fig. S2. Spatial patterns of two small stem categories (1-5 cm DBH ["saplings"] and 5-10 cm DBH) with respect to conspecific and heterospecific adult trees, with deer exclosure omitted. A corresponding figure with the deer exclosure included is provided in the main text (Fig. 2). Higher values represent increased clustering. The panels with confidence envelopes display "1to5" values minus "5to10" values; extensions above the envelope indicate that 1-5cm DBH stems are more clustered than 5-10cm DBH stems around adults (conspecific or heterospecific), and vice versa.
 The only qualitative difference between Fig. S2 (above) and Fig. 2 (main text) is that Fig. 2 indicates a tendency for smaller Nyssa sylvatica juveniles (1to5) to cluster more around conspecific adults, as compared to larger N. sylvatica juveniles (5to10), from distances of about 5 to 15 meters, but Fig. S2 reveals no significant effects for this species.

Congeneric effects among Carya species
We found mixed but limited support for congeneric NDD among hickories (Table S2). Most pairwise comparisons were non-significant, but C. glabra had a negative effect on C. tomentosa and C. ovalis (which is sometimes considered a subspecies of C. glabra), and C. tomentosa negatively affected C. glabra and C. cordiformis (but the latter effect was borderline significant). When all Carya were pooled, only C. ovalis and C. tomentosa were negatively affected, matching the independent effects of C. glabra, and thus suggesting that C. glabra may be driving the overall pattern. No positive congeneric effects were detected. Overall, conspecific NDD seems to be stronger than congeneric NDD among hickories, suggesting that the underlying mechanisms (e.g. resource competition, shared natural enemies) are mostly species-specific. However, Zhu et al. (2015a) found -in a tropical forest -that negative effects beyond the species level (i.e. phylogenetic NDD) increased with life stage (and were only significant in size classes beyond the sapling stage), while conspecific NDD decreased with life stage. As such, it is possible that the relative importance of congeneric and conspecific NDD in hickories is dependent on the size class investigated. Table S2. Congeneric effects among Carya species (sapling growth rate as a function of IDW BA). In the four leftmost results columns, each cell is from a separate model with one Carya species as the focal species and another singled out as a heterospecific species of interest (so that each can be evaluated independently, one by one). These models also included conspecific IDW BA and IDW BA for all other heterospecific species (pooled together), but results are only shown for the Carya species of interest. The three rightmost columns are all from the same model, but each row represents a separate model. All other details match those described in Table 2 in the main text (e.g. spatial structure was included when appropriate). "0" = non-significant effect; "(-)" = borderline negative effect [0.05<p<0.10]; "-" = negative effect [0.01<p<0.05]; "--" = strong negative effect [p<0.01]; "X" = conspecific effect not shown here (instead see rightmost columns and Table 2).