Interaction between Ailanthus altissima and Native Robinia pseudoacacia in Early Succession: Implications for Forest Management

The goal of this study was to discover the nature and intensity of the interaction between an exotic invader Ailanthus altissima (Mill.) Swingle and its coexisting native Robinia pseudoacacia L. and consider management implications. The study occurred in the Mid-Appalachian region of the eastern United States. Ailanthus altissima can have a strong negative influence on community diversity and succession due to its allelopathic nature while R. pseudoacacia can have a positive effect on community diversity and succession because of its ability to fix nitrogen. How these trees interact and the influence of the interaction on succession will have important implications for forests in many regions of the world. An additive-replacement series common garden experiment was established to identify the type and extent of interactions between these trees over a three-year period. Both A. altissima and R. pseudoacacia grown in monoculture were inhibited by intraspecific competition. In the first year, A. altissima grown with R. pseudoacacia tended to be larger than A. altissima in monoculture, suggesting that R. pseudoacacia may facilitate the growth of A. altissima at the seedling stage. After the second year, R. pseudoacacia growth decreased as the proportion of coexisting A. altissima increased, indicating inhibition of R. pseudoacacia by A. altissima even though the R. pseudoacacia plants were much larger aboveground than the A. altissima plants. In early successional sites A. altissima should be removed, particularly in the presence of R. pseudoacacia in order to promote long-term community succession.


Introduction
The negative impacts of invasive plant species on native plant populations, plant community structure, and ecosystem nutrient cycling are well documented. Invasive plants have displaced native plant species [1], deteriorated the quality of native forests [2,3], changed soil chemical properties [4] and altered soil microflora [5]. The success of invasive species depends on the nature of the habitat at the time of invasion, the functional traits of introduced plants, and the interactions with other species after the first establishment of the invasive species. Some invasive plants such as Microstegium vimineum (Trin.) A. Camus and Alliaria petiolate (M.Bieb.) Cavara & Grande become dominant in forest systems by outcompeting other native species [6,7]. However, other invasive plants are less-dominant (distributed sparsely in the forest), but can still have measurable effects on forest regeneration during succession and understory diversity following succession, such as Lythrum salicaria L. and Ailanthus increase as the plants aged from seedlings to saplings. Moreover, we hypothesized that the native species would facilitate the growth of the invasive species at the seedling stage, but compete with the invasive species at the sapling stage.
In this study, we evaluate the type and intensity of the interaction between the two selected tree species from seedling to sapling. Both species are known to be "ecosystem engineers" with potentially opposite effects on community development. As mentioned above, R. pseudoacacia has a documented positive effect on succession [25,26], while A. altissima has a documented negative effect on succession [8]. Our approach was to establish a common garden study, within the native range of R. pseudoacacia, in which seedlings of both species were planted in an additive-replacement series design and their growth patterns were followed for three years, during which time the two species became saplings. We discuss the ecological and forest management implications of the results. This field site was originally a post-agricultural field that had been fallow for at least 10 years. The site was sprayed with herbicide (Round-up, Monsanto, St. Louis, MO, USA; active ingredient = Glyphosate (N-(phosphonomethyl) glycine) twice over a three-week period to control the previous plants before the site was prepared for planting. The site was fenced during the three years of the experiment to reduce herbivory by deer and rabbits.

Study Species and Site Description
A weather station was installed in the experimental site for continuous monitoring of ground and leaf-height air temperature, and precipitation. The average growing season (May-October) monthly high/low temperature and total precipitation in the three years of the study were: 24.5 • C/12.9 • C, 65.6 cm in 2011; 25 • C/11.9 • C, 81.7 cm in 2012; 24.4 • C/11.4 • C, 73.4 cm in 2013.

Experimental Design
The experimental design was an additive-replacement series ( Figure 1A). This competition approach is also known as a response surface design, which enables us to evaluate the effects of different proportions of each species on each other (without the confounding effect of total density) across a set of different total densities [27]. Each experimental unit was a 2 m by 2 m square plot, separated by at least 1 m on all sides of each plot. The plots were laid out into six rows with each row separated by 2 m from the rows on either side. The experimental plots were randomly located within and across rows, resulting in 14 treatments that included four different tree densities (one to four plants plot), all possible species combinations under each density, and five replicates for each treatment. Within the plots that had more than one tree, a constant distance of 1 m was kept between neighboring trees in order to allow growth over three years and to reflect common distances between trees in local early successional stands. Although the distance between each plant was constant, the number of plants per 1 m 2 in the center of the plot reflects a change in plant density. Our data on mapping both A. altissima and R. pseudoacacia seedlings (currently unpublished) in a southern Appalachian forest indicate some clumping, but distances of 1 m between seedlings are common. The trees were arranged in the 1 m 2 center of the plot as far away from the plot border as possible ( Figure 1B). The spatial In addition to the experimental plots, 12 extra plots were established to grow trees that were destined to be harvested, to formulate a dimensional analysis after each growing season. We included these extra plots because the experimental trees could not be harvested for biomass estimates during the three-year study. The extra plots were monocultures of three or four plants per plot, resulting in 21 extra trees from each species. In the early fall of each year, seven plants of each species were randomly selected for harvest in order to formulate annual dimension analyses.
The experiment ran from June 2011 to November 2013. No additional watering or nutrients were applied to the plots except weekly watering during the first month after planting. Weeds were mowed around the experimental plots and removed from within the plots on a weekly basis during the growing season. After three seasons of growth the trees were large enough for the canopies of the adjacent plots to touch. Therefore, the experiment was terminated after three years because the plots would not be independent aboveground in future years.

Dimension Analysis, Competition Indices and Data Analysis
The annual, estimated, aboveground, biomass increment, defined as the end biomass minus the beginning biomass in each growing season, was chosen as the primary measurement of plant fitness for the competition analysis. At the beginning and end of each growing season, the height and basal diameter were measured for every tree in all plots. The height of each plant was determined by comparing the plant height to a pole graduated in cm or a meter stick the first year, and the basal diameter was determined with a caliper. Also, at the end of each growth season, before leaf fall, the seven randomly selected extra trees of each species were harvested at ground level, separated into leaf and stem, and put into a 60 °C oven until completely dry (typically four to seven days for leaves, 15-35 days for stems) to get dry biomass measurements. Subsamples of the leaves were taken from randomly selected plants of the seven extra trees to calculate the specific leaf area before the first year harvest. After obtaining the best regression equation for each species, the estimated, aboveground biomass (referred to as "biomass" from this point on) for each individual of A. altissima and R. pseudoacacia in the experimental plots was calculated based on its height and basal diameter In addition to the experimental plots, 12 extra plots were established to grow trees that were destined to be harvested, to formulate a dimensional analysis after each growing season. We included these extra plots because the experimental trees could not be harvested for biomass estimates during the three-year study. The extra plots were monocultures of three or four plants per plot, resulting in 21 extra trees from each species. In the early fall of each year, seven plants of each species were randomly selected for harvest in order to formulate annual dimension analyses.
The experiment ran from June 2011 to November 2013. No additional watering or nutrients were applied to the plots except weekly watering during the first month after planting. Weeds were mowed around the experimental plots and removed from within the plots on a weekly basis during the growing season. After three seasons of growth the trees were large enough for the canopies of the adjacent plots to touch. Therefore, the experiment was terminated after three years because the plots would not be independent aboveground in future years.

Dimension Analysis, Competition Indices and Data Analysis
The annual, estimated, aboveground, biomass increment, defined as the end biomass minus the beginning biomass in each growing season, was chosen as the primary measurement of plant fitness for the competition analysis. At the beginning and end of each growing season, the height and basal diameter were measured for every tree in all plots. The height of each plant was determined by comparing the plant height to a pole graduated in cm or a meter stick the first year, and the basal diameter was determined with a caliper. Also, at the end of each growth season, before leaf fall, the seven randomly selected extra trees of each species were harvested at ground level, separated into leaf and stem, and put into a 60 • C oven until completely dry (typically four to seven days for leaves, 15-35 days for stems) to get dry biomass measurements. Subsamples of the leaves were taken from randomly selected plants of the seven extra trees to calculate the specific leaf area before the first year harvest. After obtaining the best regression equation for each species, the estimated, aboveground biomass (referred to as "biomass" from this point on) for each individual of A. altissima and R. pseudoacacia in the experimental plots was calculated based on its height and basal diameter measurements. The annual biomass increment for each tree was determined by subtracting the spring biomass value from the fall biomass value.
A repeated measurement ANOVA was used to detect the impacts of: (1) tree density (total for both species); (2) proportion of the other species in the plot; and (3) year (one, two, three) on biomass increment. Students' t-tests were applied to pairwise comparisons of the different tree densities and proportions of other species for each year separately. We used the Students' t-test instead of the Tukey post-hoc test because the Tukey post-hoc test assumes independency of the data across the groups, which is not appropriate for repeated measure ANOVA.
The final biomass after the three growing seasons was analyzed by two-way ANOVA with the density and proportion of the other species in the plot as two factors. Students' t-tests were applied to pairwise comparisons of the final total biomass values.
The interpretation of the outcome of competition depends critically on how the competition was measured and which indices were used [28]. We evaluated two ways of quantifying the neighboring plant effect on target species biomass increment. The first and most common method used in the literature, was an evaluation of biomass increment in relation to the number of neighboring plants of the other species. The second way was to evaluate the target species biomass increment in relation to the total biomass of the neighboring species. The second method has been considered more appropriate when two species with a large size difference are compared [29,30].
To evaluate the competition effect and its intensity, we reviewed the existing indices in the literature and chose two competition indices that were appropriate for our experimental design-(1) substitution rate/competition coefficient (S), and (2) relative interaction intensity (RII) to evaluate the type and intensity of the interaction between the two species [31][32][33].
The substitution rate is computed based on multivariate regression coefficients to evaluate the relative intensity of intraspecific and interspecific competition, which are obtained by the following equations: where D aa : density of A. altissima in the plot D rp : density of R. pseudoacacia in the plot B o : estimated intercept in the model B aa : estimated coefficient for intraspecific competition of A. altissima B rp : estimated coefficient for intraspecific competition of R. pseudoacacia B aarp : estimated coefficient for interspecific competition of R. pseudoacacia on A. altissima B rpaa : estimated coefficient for interspecific competition of A. altissima on R. pseudoacacia The substitution rate (also called the competition coefficient) for A. altissima: S aa = |B aarp /B aa | and for R. pseudoacacia S rp = |B rpaa /B rp | was calculated [31,32]. To measure the interaction type and intensity for these two plant mixtures, the relative interaction intensity (RII) was calculated.
where B w : Biomass observed for target plant when growing with other plants B o : Biomass of target plant growing alone in the absence of inter/intraspecific interactions RII was chosen over other frequently used competition indices because it has robust mathematical and statistical properties, and allows for clear interpretation. RII has a continuous range from −1 (competitive exclusion) to 1 (symbiosis), and a positive number indicates facilitation while a negative number suggests competition. The more negative the RII, the greater the interspecific competition. RII offers the most consistent results among the interaction indices that are commonly used [33].
Since there were only five replicates for plots with only one plant per plot for each species in our experimental design, the usual calculation of a mean B o could result in a bias when calculating the RII for the other 45 plots. In order to reduce the possibility of this bias, we used a non-parametric bootstrap to make sure that the B o we used was randomly selected [34]. For each mixture plot, we first computed the average of calculated B w , then computed RII 2000 times by bootstrapping the B o value from the five different biomass increment values from the monoculture plots, and obtained the average RII value for this plot. After computing the RII for each mixed species plot, a repeated measurement two-way ANOVA was applied to analyze the competition indices. We also used a rank order technique to analyze the competitive effect of various treatments. For each year, we ranked the treatments by RII values from the lowest (more competition) to the highest (possibly facilitation) to perform a rank order analysis. The three ranked numbers (from 2011 to 2013) were added together for each treatment to obtain the overall rank, which was used to indicate competition intensity.

Growth Patterns and Dimension Equations
At the end of the first growing season the plants were relatively small, with an average (mean ± standard deviation) height of 1.24 ± 0.71 m and 0.52 ± 0.43 m for R. pseudoacacia and A. altissima, respectively. The plants had not formed a multi-branched architecture at this stage. After the second growing season the trees attained 3.85 ± 1.49 m and 1.48 ± 1.63 m height on average for R. pseudoacacia and A. altissima, respectively. The final average heights of the saplings after three years of growth were 4.58 ± 2.04 m for R. pseudoacacia and 2.03 ± 2.12 m for A. altissima. Following a late freeze in the spring of the second year, A. altissima became partially coppiced. Therefore, we measured the height and diameter for all the main stems of each plant (coppice) to increase the accuracy of the biomass estimates. Consequently, the height and diameter of A. altissima had increased variance in later years.
The dimensional data from 2011 to 2012 were combined to formulate our regressions between biomass, height and diameter for both species (Table 1). The fitting power of the combined regressions was improved compared with the regressions formed using the dimensional data from each year alone. The combined regression was used to predict the biomass for the first and second year. In order to fit the regression equations for biomass, a square root transformation was needed for A. altissima in all three years, while log transformation was needed to fit linear regressions for R. pseudoacacia in the first two years to meet the normality assumption ( Table 1). The adjusted R 2 for these four equations were all ≥0.95, which indicated that our biomass estimate was robust. The basal diameter was more important for predicting biomass than the height, which is reasonable because the canopies of both species spread widely. Table 1. Dimension analysis of total above-ground dry weight (biomass) for Ailanthus altissima and Robinia pseudoacacia based on the basal diameter and height for three growing seasons 2011-2013.  There was a significant difference in the biomass increment between the two species (p < 0.0001) and across the three years (p < 0.0001, Table 2). The biomass increment for R. pseudoacacia in the first growing season was approximately double that of A. altissima (Figure 2). In the second growing season, the biomass increment for both species increased significantly (p < 0.0001) by six to eight times that of the first growing season. However, the biomass increment during the third growing season was slightly smaller than that of the second growing season when data of both species were combined (p = 0.0017). The significant down-turn in annual biomass increment in the third year was primarily due to R. pseudoacacia. However, the effect of density, species proportion and their interaction were not significant (Table 2).   growing season was approximately double that of A. altissima (Figure 2). In the second growing season, the biomass increment for both species increased significantly (p < 0.0001) by six to eight times that of the first growing season. However, the biomass increment during the third growing season was slightly smaller than that of the second growing season when data of both species were combined (p = 0.0017). The significant down-turn in annual biomass increment in the third year was primarily due to R. pseudoacacia. However, the effect of density, species proportion and their interaction were not significant (Table 2).

Intraspecific Competition
In monocultures, the largest individual biomass increment (Figure 2A,B) was found in the lowest density plots (density = 1) in all three years (data for both species combined), but the effect of density was not significant at the α = 0.05 level. Pairwise comparison for A. altissima indicated that the biomass increment was significantly larger in the lowest density monoculture than that of all other monoculture densities (density = 2, 3, or 4 trees per plot) in year 1 (p = 0.032, 0.047, and 0.037 respectively) and year 2 (p = 0.030, 0.023, and 0.013 respectively). For year 3, the difference in biomass increment was not significant between any pair of monoculture density for A. altissima (Figure 2A).
Surprisingly, in monoculture, R. pseudoacacia showed a trend of increasing biomass increment as the density increased in year 1, but the trend was not significant. This trend was reversed in the next two growing seasons ( Figure 2B). The difference in individual biomass increment between the monoculture densities was only significant between the highest density and the lowest density group in year 2 (p = 0.032). There was a significant or barely significant difference in biomass increment in the pairwise comparison between the lowest density monoculture and other density monocultures for R. pseudoacacia in year 3 (p = 0.026, 0.053, 0.035, respectively) ( Figure 2B). In general, increasing density from two trees per plot to three and four trees per plot in monoculture did not affect the biomass increment significantly in all three years for either species (Figure 2).

Interspecific Facilitation or Competition
The individual biomass increment of A. altissima tended to be higher in the presence of R. pseudoacacia than in the presence of itself at density 2 in all three years, suggesting the facilitation of A. altissima by the presence of R. pseudoacacia ( Figure 3A-C). However, this trend was not statistically significant (p = 0.65, 0.42, 0.57 in year 1, 2 and 3). There was no significant difference in the biomass increment of A. altissima among the different proportions of R. pseudoacacia treatments at higher densities. The effect of the density and proportion of R. pseudoacacia on A. altissima were not significant at the α = 0.05 level when all treatments and years were combined ( Table 2).
The individual biomass increment of R. pseudoacacia tended to increase in the presence of A. altissima at density 2 in all three years ( Figure 3D-F), which suggests facilitation at this density, but this increase was not statistically significant (p = 0.27, 0.31, 0.42, in the respective three years). However, average biomass increment of R. pseudoacacia individuals decreased slightly as the proportion of A. altissima increased at density 3 ( Figure 3D-F). There was one tree with a negative biomass increment for both species in the density 4 monoculture group, which caused the monoculture group to have a lower observed average value than the actual value. The effect of the number of neighboring A. altissima on the biomass increment of R. pseudoacacia individuals was not significant at the α = 0.05 level when all treatments and years were combined (Table 2). It is important to note that the biomass increments of individuals in the mixture treatments were the result of the combined effects of both intra-and inter-specific interactions. A. altissima by the presence of R. pseudoacacia ( Figure 3A-C). However, this trend was not statistically significant (p = 0.65, 0.42, 0.57 in year 1, 2 and 3). There was no significant difference in the biomass increment of A. altissima among the different proportions of R. pseudoacacia treatments at higher densities. The effect of the density and proportion of R. pseudoacacia on A. altissima were not significant at the α = 0.05 level when all treatments and years were combined (  The final, total biomass per individual of both species after three years of the experiment (Figure 4) was analyzed to evaluate the cumulative effect of three years of plant interaction. Ailanthus altissima plants in density 1 had a significantly greater biomass per individual compared with plants in density 2, 3 and 4 (p = 0.043, 0.022 and 0.008 respectively). The final biomass of A. altissima individuals in monocultures of density 1 was significantly greater than monocultures with the other three densities (p < 0.0001), but there were no significant differences among the other three monocultures (Figure 4). For R. pseudoacacia, density was a significant factor. Plants in density 1 (Figure 4) had significantly greater biomass per individual when compared with plants in density 2, 3 and 4 (p = 0.003, 0.018 and 0.001). Robinia pseudoacacia grown in 50% A. altissima (combining densities 2 and 4) had significantly greater final biomass per individual (p = 0.024, 0.035) than those grown with 33% or 67% A. altissima (combining data from densities 3 and 4).

Competition Indices
In all three years, all the coefficients of regression had a negative sign ( Table 3), indicating that competitive interactions occurred either between or within both species. The intraspecific competition intensity for both A. altissima and R. pseudoacacia tended to increase over time ( Table 3). The intraspecific competition intensity for R. pseudoacacia, increased from years 1 to 2, but then decreased in year 3 ( Table 3). The interspecific competitive effect of R. pseudoacacia on A. altissima, indicated by Baarp, tended to increase over the three years ( Table 3). The interspecific competitive effect of A. altissima on R. pseudoacacia, indicated by Brpaa also increased over the three years, suggesting strong interspecific competition between A. altissima on R. pseudoacacia, and that competitive effect increased through time.

Competition Indices
In all three years, all the coefficients of regression had a negative sign (Table 3), indicating that competitive interactions occurred either between or within both species. The intraspecific competition intensity for both A. altissima and R. pseudoacacia tended to increase over time ( Table 3). The intraspecific competition intensity for R. pseudoacacia, increased from years 1 to 2, but then decreased in year 3 ( Table 3). The interspecific competitive effect of R. pseudoacacia on A. altissima, indicated by B aarp , tended to increase over the three years ( Table 3). The interspecific competitive effect of A. altissima on R. pseudoacacia, indicated by B rpaa also increased over the three years, suggesting strong interspecific competition between A. altissima on R. pseudoacacia, and that competitive effect increased through time.
The substitution rate (S) values suggested that intraspecific competition was dominant in A. altissima, but interspecific competition became more important for A. altissima as time passed (S aa was smaller than 1 and increased over the three years, Table 3). For R. pseudoacacia, interspecific competition was significant and the dominant factor over all three years, especially years 1 and 3 (S rp was bigger than 1, Table 3). Robinia pseudoacacia experienced an equally significant and stronger intraspecific and interspecific competitive effect in year 2, but the intraspecific competitive effect was slightly stronger (Table 3).
Despite the general high p-values (greater than 0.05), multiple comparisons of RII did reveal consistent patterns for both species across treatments (Tables 4 and 5). Most RII values were negative for A. altissima, suggesting that a competitive effect, either intraspecific or interspecific, had occurred. Two-way ANOVA results suggested that density had a significant effect on the RII of A. altissima in year 1. Specifically, RII at density 2 was significantly lower than at density 3 (p = 0.0376). On the other hand, the proportion of the neighboring R. pseudoacacia trees was not a significant factor in year 1 (p = 0.091). Neither the effect of density nor proportion of the R. pseudoacacia was significant in years 2 and 3. There was, however, a trend towards higher density groups having lower RIIs (Table 4), suggesting more intense competition as the tree density increased in both years. The ranking of the average RII value of A. altissima on a yearly basis across all treatments (Table 5) from highest to lowest was year 3 (RII = −0.061), year 1 (RII = −0.130) and year 2 (average RII = −0.175). Table 3. Coefficients of interspecific and intraspecific competition intensity between Ailanthus altissima and Robinia pseudoacacia growing in a common garden experiment in Boyce, VA, USA, over a three-year period (2011-2013). 2.33 a B aa = coefficient of intraspecific competition for A. altissima; B rp = coefficient of intraspecific competition for R. pseudoacacia; B aarp = interspecific competitive effect of R. pseudoacacia on A. altissima; B rpaa = the interspecific competitive effect of A. altissima on R. pseudoacacia; S aa = substitution index for A. altissima; S rp Substitution index for R. pseudoacacia; b Negative values for coefficient of competition indicate competition occurred. Substitution indices greater than one indicate that the interspecific competition effect is larger than the intraspecific competition effect. Table 4. Mean relative interaction intensity (RII) for three densities of Ailanthus altissima and Robinia pseudoacacia growing in a common garden additive-replacement series experiment in Boyce, VA, USA, over a three-year period (2011-2013).

Is the Interaction between Species Competition, Facilitation or Coexistence?
The type of interaction between these two species was mainly competition rather than facilitation. The negative effect of competition on plant growth affected R. pseudoacacia more than it did A. altissima. In this three-year period, the influence of R. pseudoacacia on the vegetative growth of A. altissima was either neutral or slightly facilitated, while, in most treatments, the growth of R. pseudoacacia trees was inhibited by neighboring A. altissima. For R. pseudoacacia, the competitive impact of A. altissima was stronger than competition with itself, but only when A. altissima made up greater than 50% of the plants in the plot. For A. altissima, there was less intraspecific competition than that for R. pseudoacacia, which would promote higher density natural stands of A. altissima relative to R. pseudoacacia.
The interspecific competition outcomes for the biomass increment from this experiment were not as clear as the intraspecific competition results because of greater variation within each treatment group, and because the outcomes reflected the combined effects of both intraspecific and interspecific competition. The competition coefficient result supports an interspecific competitive interaction between these two species. Overall, our results were consistent with previous theories of density-dependent regulation in plants. Based on our findings, we predict that R. pseudoacacia will be negatively impacted by neighboring A. altissima, wherever these two species co-occur. Our experimental design used a consistent 1 m spacing between trees. Although we changed the density on an area basis (1-4 m −2 ), the distance between the trees was the same in all density treatments. It is possible that the constant distance between trees caused our biomass increment per tree to be similar for the treatments in year 2 and 3.
The growth of R. pseudoacacia was most inhibited by A. altissima when the latter was dominant in the neighborhood. When R. pseudoacaia was dominant there was not a similar competitive effect on A. altissima. Otherwise, intraspecific competition was more important than interspecific competition for R. pseudoacacia. This exemplifies the importance of early-stage removal of the invasive species A. altissima in forests, (i.e., before A. altissima increases in abundance via sprouting or seeds). Such removal is especially important near R. pseudoacacia, in order to eliminate the negative impact of this invasive species on the native nitrogen-fixing symbioses associated with the native R. pseudoacacia, their influence on ecosystem nitrogen cycling and their effect on succession. This finding supports the early detection and rapid response approach for invasive species control [35]. Overall, competition was the main interaction between these species. Moreover, we found that the competitive effect stronger from A. altissima on R. pseudoacacia than in reverse. Although some suggestion of facilitation was observed at the seedling and sapling levels, there is not likely to be any facilitation of A. altissima by R. pseudoacacia in Appalachian forests.

Does the Interaction between Species Change from Seedling to Sapling?
Our results suggest that studying different developmental stages of the interactions between species is important. Most previous studies of the competition between native and invasive plants have focused on grasses and forbs, because of their short life cycle and because they are much easier to manipulate compared with trees [36]. However, interactions with invasive trees could be very different from the results of studies with invasive herbaceous species [37,38]. Tree interactions may also shape the community structure and function and impact other species more significantly than understory species [39][40][41]. The impact of invasive trees on native trees can occur for a long time period after a disturbance [42][43][44]. Studies like ours should occur over a longer time period to better understand the interactions between invasive and native trees. Although we only conducted this experiment for three years, we discovered a shift of interaction type and variation in competition intensity for trees from the seedling stage to the sapling stage. In relation to our second question, we found that the negative effect of the invasive tree A. altissima on native R. pseudoacacia became stronger as the plants transitioned from seedlings to saplings. One possible reason would be that A. altissima grew larger and as a result had a stronger interaction with neighbors in year 2 and year 3. Also, intraspecific competition became stronger for both species as time passed, especially during the second growing season when growth was most rapid. However, we recognize that our results may be because the spacing per plant size was large at the seedling stage (resulting in low competition), but the same distance per plant size was small at the sapling stage (resulting in higher competition).
Our findings indicate that any negative impact of A. altissima on R. pseudoacacia is dependent on the size of A. altissima. These results further support an early control strategy for the invasive A. altissima. Alternatively, these data suggest that some native species could be found that out-compete A. altissima in the early years of succession, but by the third year of succession it may be too late to use any native species to reduce the growth of A. altissima.
Previous studies have shown that the type of interaction between a pair of plant species can switch quickly in response to environmental conditions, even on a daily basis [45]. At any particular time, the overall result of competition or facilitation is a cumulative effect. Therefore, the results from this study complement previous research [46], which showed that seedlings of R. pseudoacacia have a facilitative effect (due to the effect of nitrogen fixation on soil nitrogen availability) on the growth of A. altissima seedlings when the soil nutrient level may be low. In locations with equal numbers of both species, R. pseudoacacia will also facilitate the growth of A. altissima. Moreover, at times or places where the population of locust leaf miners are low, the competitive effects of A. altissima may also be low. Although the intensity of competition or the possibility for facilitation may change within a season, our analyses were limited to the overall effect of a whole growing season each year. Thus, over the three-year period that the plants transition from seedling to sapling, competition is the dominant interaction between R. pseudoacacia and A. altissima and A. altissima is the better competitor.

What Are the Forest Management Implications of This Research?
Our results suggest that very early in succession, at the seedling stage, R. pseudoacacia may facilitate the growth of A. altissima, as it has been shown to do with other native and exotic species [26]. The facilitation may be enhanced at sites with low nitrogen availability [46], but not at high nutrient concentrations [47]. The intensity of the intraspecific competition for both species was highly correlated with plant size, which suggests that indirect competition for the same resources, such as light and nutrients, is the cause of intraspecific competition. When the two species co-exist, our findings suggest that the growth of invasive A. altissima would not be negatively affected by R. pseudoacacia; even when R. pseudoacacia's density is greater than that of A. altissima; conversely, the native R. pseudoacacia would be suppressed by A. altissima, particularly at high densities of A. altissima. Given A. altissima's prolific seed production and ability to sprout, we would expect this negative competitive effect to increase with time. However, R. pseudoacacia saplings accumulate much greater amounts of above-ground biomass and leaf area than A. altisisma saplings, suggesting that in early successional sites when trees attain sapling size R. pseudoacacia is a better competitor for above-ground resources than A. altissima. Consequently, A. altissima's ability to negatively affect R. pseudoacacia's growth rate in our study may be due to A. altissima's ability to better compete for below-ground resources, which was also found in greenhouse studies of seedlings [47]. Thus, the cause of inhibition by A. altissima on the much larger, above-ground R. pseudoacacia ought to be the result of either more effective below-ground competition or a negative effect of allelopathy on root growth and the nitrogen-fixing bacteria associated with R. pseudoacacia [46,48].
Thus, it is important to consider the management of both species simultaneously. These two species are likely to co-occur in post-harvest or early secondary succession forest sites world-wide. They have opposite effect on forest soil development (particularly nitrogen availability and litter depth) and opposing effects on forest development. Ailanthus altissima can reduce both canopy tree regeneration and understory diversity while R. pseudoacacia, in its native habitat, typically gives way to forest succession [24,49]. The effects of both species on the forest become accentuated with time. Therefore, early post-harvest management plans need to minimize the negative effects of A. altissima on R. pseudoacacia in order to optimize long-term forest regeneration. Indeed, despite a shared nonnative status and lack of enemies, A. altissima has spread faster than R. pseudoacacia in Northern Italy over the last thirty years [22], though planted R. pseudoacacia has increased the dominance of nitrophilous species in some European forests in part due to stalled succession caused by the short-rotation cycles typical of energy tree plantations [50].
Ailanthus altissima predominantly colonizes rapidly after forest harvests or at other sites with high light, such as road cuts [18,[51][52][53]. In fact, across South Africa, most A. altissima trees colonized within the first five years of a disturbance [53]. In R. pseudoacacia's native range, it is important to remove A. altissima early in forest development, particularly when the A. altissima trees are proximal to R. pseudoacacia trees, because the presence of A. altissima in early succession forests will inhibit the positive effect of R. pseudoacacia on forest regeneration. The removal of A. altissima must be carried out at an early developmental age because following root system establishment, cutting and injecting herbicide may not fully eradicate the plants due to extensive root growth and subsequent root suckering [54]. Once A. altissima plants attain reproductive age (post-sapling age), there will be little opportunity to reduce the negative effect of these trees on forest health due to their high seed production [55], strong legacy effects on soil development [23] and propensity to form monoculture stands.
The coexistence of R. pseudoacacia and A. altissima seedlings following harvest is likely to enhance the successful recruitment of A. altissima because the R. pseudoacacia may facilitate the growth of A. altissima at the seedling stage. Digging or carefully directed herbicide application of the A. altissima seedlings out of post-management sites, particularly around R. pseudoacacia trees, may be one of the most effective management plans for reducing the negative effect of A. altissima on forest regeneration in R. pseudoacacia's native range. Waiting until later to manage A. altissima will have a negative effect on R. pseudoacacia and will not be successful in managing A. altissima or promoting forest health. Moreover, the management of A. altissima becomes a higher priority if R. pseudoacacia is also in the early post-disturbance site.

Conclusions
The results of a replacement-series competition study clarified the interaction between two important invasive trees that have opposite effects on forest regeneration. Ailantus altissima and R. pseudoacacia are likely to coexist in high light and early disturbance sites world-wide. When coexisting, R. pseudoacacia may facilitate the growth of A. altissima at the seedling stage. However, at the sapling stage and thereafter, A. altissima will inhibit the growth of R. pseudoacacia, particularly where A. altissima has a higher density than R. pseudoacacia. Therefore, A. altissima will likely override the positive effects that R. pseudoacacia has on forest development and forest health. It is crucial to remove A. altissima at the seedling stage, by digging or carefully directed herbicide application, before it can have negative effects on forest health, particularly in the presence of R. pseudoacacia.