Plant–soil feedbacks contribute to coexistence when considering multispecies assemblages over a soil depth gradient

Abstract

Plant–soil feedbacks (PSFs) may determine plant–species coexistence. They may stabilize coexistence, but frequently destabilize it by increasing fitness differences between species. Most studies focus on pairwise models in greenhouses. Thus, whether PSFs favor or deter coexistence is still unclear, especially in multispecies field contexts. We analyzed pairwise and multispecies coexistence over a hydric gradient in a semiarid grassland. Using PSF strength estimates between 17 species, we measured stability and fitness differences between all species pairs, and built all possible multispecies communities to test computationally whether they were stabilized by PSFs. We analyzed whether coexistence probability diminishes with species richness, as previously hypothesized. Because PSFs change with environmental conditions, we investigated their contribution to overall diversity maintenance over the hydric gradient. Strong PSF increased fitness differences, hindering pairwise coexistence. As expected, the probability that an assemblage was stable diminished with its richness, with the largest stable community containing 12 of the 17 species. However, all species coexisted with others in at least one assemblage, highlighting the importance of multispecies analyses. Positive PSFs promoted coexistence in pairwise analyses, but were associated with species‐poor communities. Contrastingly, negative PSFs predominated in species‐rich associations, perhaps due to indirect positive interactions (an “enemy of my enemy is my friend” scenario) that are known to maintain diversity in this grassland. Changes in the density of different species over the hydric gradient predicted from PSF‐stabilized communities matched observations in nature. This seems to promote species turnover and thus coexistence along the gradient. As such, the interplay between environmental conditions and PSFs may be an important driver of diversity. Our results emphasize the need to move beyond pairwise coexistence models. In multispecies systems, crucial indirect interactions may arise. The interplay between environment and PSF under field conditions may provide important insights into coexistence in nature.

INTRODUCTION

Diversity maintenance within guilds of species that compete for the same resources, such as plants, has long puzzled ecologists (Hutchinson, 1961). Long‐term coexistence is possible if the interspecific density‐dependent regulation outweighs interspecific effects. Under this scenario, coexistence is stabilized because a species whose numbers decrease is released from its major antagonist—itself—leading it to population growth and thus to avoid extinction (Chesson, 2000). Nevertheless, stable coexistence is only possible if the strength of the stabilization, that is, the difference between intra‐ and interspecific effects, is sufficiently strong to overcome the fitness advantage of the strongest competitor. The issue of coexistence in competitive guilds is particularly puzzling in species‐rich communities because, in contrast to empirical studies, models tend to indicate that coexistence in rich communities is unstable (Eppinga et al., 2018; Mack et al., 2019; May, 1973). If intraspecific facilitation is weak, as happens in plants (Adler et al., 2018; Bonanomi et al., 2005), positive plant–soil feedbacks (PSFs) may stabilize pairwise coexistence because they ameliorate negative interspecific interactions, changing the balance between intra‐ and interspecific negative density dependence (Bimler et al., 2018). However, stronger negative PSFs allow for greater species richness and longer coexistence periods (Dudenhöffer et al., 2022; Eppinga et al., 2018; Mack et al., 2019). PSFs may also allow for long‐term species persistence even if they destabilize coexistence: Intransitive networks, where the effects of plant species can be visualized as cycles of winners and losers as in the rock–paper–scissors game, may cause plant population densities to fluctuate indefinitely without any becoming extinct. Intransitive PSFs have been rarely analyzed but do not seem to be common (Eppinga et al., 2018; Pajares‐Murgó et al., 2024).

Coexistence may be affected by PSFs. PSFs can be negative (e.g., if plants deplete soil resources or they promote the development of pathogens) or positive (e.g., by cultivating mutualists in their rhizosphere; Bennett et al., 2017; Crawford et al., 2019; Domínguez‐Begines et al., 2021; Teste et al., 2017). PSFs are widespread and affect vegetation at all levels of ecological organization. They affect the performance of individual plants (van der Putten et al., 2013). Feedback sign and strength are positively correlated with population density (Reinhart et al., 2021, but the opposite has also been found, see Maron et al., 2016). PSFs may enhance community diversity by stabilizing coexistence, because conspecifics frequently have more negative effects on each other than on heterospecifics (Crawford et al., 2019). However, PSFs usually widen fitness differences, hindering coexistence despite their stabilizing effect (Yan et al., 2022).

Environmental gradients can aid coexistence because different sets of species may occur over different segments of the gradient (Martínez‐Blancas et al., 2022; Martínez‐Blancas & Martorell, 2020; Martorell et al., 2015). Because plant–soil interactions depend on the environment, gradients can also promote plant coexistence due to changes in PSFs. For instance, in productivity gradients, PSFs may shift from negative when resources are abundant to positive when resources (particularly water) are scarce (In't Zandt et al., 2019; Jiang et al., 2024; O'Brien et al., 2017; Porter et al., 2020; Revillini et al., 2016). Soil mutualists also confer tolerance to drought stress (Augé, 2001; Bennett & Klironomos, 2018; Fitzpatrick et al., 2018; Porter et al., 2020; Revillini et al., 2016). In contrast, when more water is available, there is a greater motility of soil pathogens and reduced protection from mutualists against them (Bennett & Klironomos, 2018). However, the consequences of such shifts in PSFs on coexistence are yet unclear.

Most studies on species coexistence focus on pairwise models and are performed in greenhouses. However, in natural communities multispecies interactions occur. Moreover, there are several reports that PSF estimates from greenhouse and field studies differ greatly, so results obtained from controlled conditions may not shed light on how diversity is maintained in nature (Forero et al., 2019; Kulmatiski, 2018).

In this contribution, we analyzed the effect of PSFs on coexistence, community composition, and diversity over a hydric‐stress gradient. To do so, we used previously published (Martorell et al., 2021) pairwise intra‐ and interspecific PSF strengths for a set of 17 grassland plant species quantified over a hydric gradient. With these data, we determined whether PSFs promoted coexistence by calculating their effects on pairwise stabilization and fitness differences over the hydric gradient. If, as previously observed, strong PSFs lead to large fitness differences (Kandlikar et al., 2021; Yan et al., 2022), we may expect that (1) the possibility of coexistence diminishes with PSF strength and thus (2) with stress, under which PSFs are stronger at the study site (Martorell et al., 2021). To consider more realistic multispecies scenarios, we analyzed computationally the PSF‐mediated stability of all possible communities that can be assembled from our 17 species. From this analysis, we (3) tested whether the probability of coexistence diminishes with species richness, as happens in most models, and whether different species tend to occur in species‐poor or species‐rich communities depending on PSF. If so, we may expect that (4) species that are involved in negative PSFs occur in more speciose communities (Eppinga et al., 2018), although the same has been argued for facilitative effects. Another mechanism underlying the effects of PSF on diversity may be their interaction with environmental heterogeneity. Because PSF varies with stress, different species may coexist stably with each other over the hydric gradient, allowing (5) for a greater diversity overall in the system even if just few species can coexist in a specific portion of the gradient. Finally, we tested whether some of these computational results have a correspondence in nature. We used independent field data to test whether changes in plant density can be explained by stabilizing effects and fitness differences derived from PSF, and whether the distribution of species over the hydric gradient (species turnover) predicted by PSF resembles that observed in nature.

METHODS

This study was conducted in a semiarid grassland in Concepción Buenavista, southern Mexico. The mean temperature is 16°C, and precipitation is 578 mm. Soil depth rarely exceeds 20 cm and reflects water availability: Shallow soils are drier than deep ones (Martorell et al., 2015; Martorell & Martínez‐López, 2014).

Occupancy–survival relationships

Our analyses are based on occupancy–survival relationships (OSRs). Positive OSRs indicate that the life expectancy of a seedling—the successor—increases as a result of previous occupancy by the same or a different species in a plot—the predecessors. They consider survival, not germination probability, and thus are a proxy for positive PSFs. Negative OSRs indicate negative PSF. Data for OSR calculation were obtained by (1) recording the abundance of naturally occurring plants over 6 years—the predecessors, which may affect soil properties or biota, and (2) evaluating on the seventh year whether previous occupancy affected the survival of the successors. To do so, we randomly placed 471 0.1 × 0.1 m quadrats in the field in 2012. We monitored the abundance of all plant species until 2017. In June 2018, vegetation was removed and transplanted seedlings of 17 species (Aristida adscensionis, Bouteloua chondrosioides, B. hirsuta, Crusea simplex, Digitaria ternata, Heterosperma pinnatum, Hilaria cenchroides, Microchloa kunthii, Muhlenbergia phalaroides, Plantago nivea, Sanvitalia procumbens, Stevia ephemera, Tagetes micrantha, Thymophylla aurantiaca, Tridax coronopifolia, Tripogandra purpurascens, and Tripogonella spicata) to the quadrats. We checked survival weekly for a month, and fortnightly until the growth season ended in October.

To obtain OSR, the longevity l _i,j of the seedlings of each successor species i was regressed on soil depth and previous occupancy by each predecessor species j. The model was $\ln l_{i,j}={\mathrm{\beta }}_{0,i}+{\mathrm{\beta }}_{1,i,j}{P}_{i,j}+{\mathrm{\beta }}_{2,i}d+{\mathrm{\beta }}_{3,i,j}{P}_{i,j}d$ where $P_{i,j}={\sum }_{t=0}^5{w}_{i,j}^t{n}_{j,t}$ is previous occupancy, d is soil depth, n _j,t is the number of individuals of species j at time t (years elapsed before 2017), and w determines the persistence of PSF. w is bounded between 0 if the effect vanishes after one year, and 1 if it endures indefinitely. The short‐term OSR for a given depth, $S_{\text{short}i,j}\left(d\right)={\mathrm{\beta }}_{1,i,j}+{\mathrm{\beta }}_{3,i,j}d$, is the effect of one individual of species j growing in the plot the year before the successor germinates. In cumulative OSR, $S_{\text{cumul}i,j}\left(d\right)=S_{\text{short}i,j}\left(d\right){\sum }_{t=0}^5{w}_{i,j}^t$, one predecessor grows in the quadrat for six years, not just one. See Martorell et al. (2021) for details.

OSR and community stability and diversity

Following Kandlikar et al. (2019), we calculated the short‐term pairwise stabilization (I _i,j ) of the coexistence of species i and j mediated by PSF as

$$I_{i,j}\left(d\right)=-\frac{1}{2}\left(S_{\text{short}i,i}\left(d\right)-S_{\text{short}i,j}\left(d\right)-S_{\text{short}j,i}\left(d\right)+S_{\text{short}j,i}\left(d\right)\right),$$ (1)

and fitness differences (F _i,j ) mediated by previous occupancy as

$$F_{i,j}\left(d\right)=\frac{1}{2}\left(S_{\text{short}i,i}\left(d\right)+S_{\text{short}i,j}\left(d\right)-S_{\text{short}j,i}\left(d\right)-S_{\text{short}j,j}\left(d\right)\right).$$ (2)

The same was done using S _cumul. PSF tends to stabilize coexistence if I _i,j is positive, but really does so only if stabilization is greater than fitness differences, that is, if their net effect

$$N_{\mathit{ij}}=I_{\mathit{ij}}-\left|F_{\mathit{ij}}\right|>0.$$ (3)

To test Prediction (1), that is, whether coexistence is hindered by strong PSF, we plotted I _i,j , F _i,j , and N _i,j against OSR for all species pairs. These same values were plotted over the depth gradient to assess Prediction (2), that coexistence is hampered in shallow soils. Because stability depends on the difference between intra‐ and interspecific PSFs (Equation 1), we compared intra‐ and interspecific OSRs at different depths as an aid to understand stability changes over the gradient.

Eppinga et al. (2018) generalized Equation (1) to multispecies situations. Their community‐wide measure of stabilization $I_c=-2I_{\mathit{ij}}$ in the two‐species case. Negative I _c values indicate that OSRs contribute to stabilize coexistence because they reflect the strength of the plant–plant feedbacks that preclude species in the community from becoming extinct even if their numbers oscillate over time (Eppinga et al., 2018; Pajares‐Murgó et al., 2024). There is no equivalent multispecies measure of fitness differences, but the relative frequencies of all species at the equilibrium can be calculated. A community is feasible if all species have positive frequencies at equilibrium. Feasibility implies that stabilization overcomes fitness differences (i.e., that N _i,j  > 0 in the two‐species case) because if a species has a negative number of individuals, it is not coexisting with the others (Kandlikar et al., 2021).

Thus, to assess the effect of PSFs on coexistence in multispecies scenarios, we determined whether they contribute to maintaining a stable species composition (i.e., I _c  < 0) with feasible (i.e., positive) densities at equilibrium. We will refer to assemblages that satisfy both conditions as stabilized and feasible communities (SFCs). Note negative I _c values contribute to stabilizing coexistence but do not grant it. In Appendix S1, we consider the more restrictive conditions required for species persistence, but the results are similar to those presented here.

We searched for SFCs out of all the possible communities having between 2 and 17 species. Many SFCs have a subset of the species found in richer SFCs, and thus are less stable in their composition. Consider two SFCs, one containing species A and B, and another with species A, B, and C. The first community is likely to be invaded by species C from nearby patches and would remain in such a state, making it unlikely to be found in the long run. Thus, we define nsSFC as SFCs whose species are not subsets of other SFCs. This was done for all soil depths between 4 and 28 cm every centimeter.

To assess whether the probability of coexistence diminishes with species richness (Prediction 3), we calculated the fraction of possible communities with different numbers of species that were SFCs. To determine whether species involved in positive PSFs tend to occur in richer communities (Prediction 4), we first determined the “preference” of each species for poor or rich communities. This was done by calculating the difference between the mean richness of the nsSFCs where each species occurs and the mean richness of all nsSFCs. Positive differences indicate that the species tends to occur in rich communities, whereas negative ones suggest a preference for species‐poor nsSFCs. Preference values for each species were then regressed on OSR calculated either as an average of the species' effects on others or as an average of the impact that other species have on it.

PSF‐mediated coexistence over the hydric gradient

To analyze whether environmental heterogeneity promotes the coexistence of species through changes in PSF, we ordered all nsSFC using a principal components analysis (PCA) based on the equilibrial relative abundances of the species they contain. We then plotted all the nsSFC that occur in each soil depth in a stacked area plot, in which similar communities (based on PCA scores) have similar colors and are on top of each other in the stack. If there is a replacement of communities over depth, different colors should predominate in different portions of the gradient.

OSR and population density

To assess the effects of OSR on population density, we used data from 1614 0.1 × 0.1 m quadrats (over 11 sites) in which we recorded the abundance of all species between 2012 and 2017. In 2018, instead of removing all plants and introducing transplants as in our experimental plots, we recorded the density of naturally occurring plants. We then calculated the summed OSR (S _OSR) from the observed density of the 17 study species in the six years before 2018 to obtain a single PSF estimate for every successor species i in each quadrat k. For short‐term OSR, this was calculated as follows:

$$S_{\mathrm{OSR}\text{short}i,k}=\sum _j^{17}\left({\mathrm{\beta }}_{1,i,j}+{\mathrm{\beta }}_{3,i,j}{d}_k\right)n_{j,k,t=0},$$ (4)

where d _k is the depth of quadrat k and $n_{j,k,t=0}$ is the abundance of the predecessor in that quadrat at elapsed time t = 0, that is, in 2017. For cumulative effects,

$$S_{\mathrm{OSR}\text{cumul}i,k}=\sum _j^{17}\sum _{t=0}^5\left({\mathrm{\beta }}_{1,i,j}+{\mathrm{\beta }}_{3,i,j}{d}_k\right)w_{i,j}^t{n}_{j,k,t}.$$ (5)

The clonal B. chondrosioides and H. cenchroides as successors were not analyzed because, unlike the rest of the species, they had been recorded as presence–absence and thus were not comparable. We regressed the densities observed in 2018 on their respective S _OSR using the package gamlss (Rigby & Stasinopoulos, 2005) for R (R Core Team, 2024). Because densities in most quadrats were zero, we used a zero‐altered gamma distribution where zeros occur with probability q. We fitted models in which we included S _OSR, depth, and their interaction, the null model with none of these, and every possible combination between these two extremes. This was repeated four times, allowing q to vary as a function of depth, PSF, both, or to remain constant. Models were compared using Akaike information criterion (AIC). See Appendix S2 for statistical details.

To determine whether the changes predicted from PSF in community composition over the depth gradient resembled those observed in nature, we used the equilibrial relative abundance of each species in all SFCs for soil depths between 4 and 28 cm to calculate the mean depth in which each species occurred over the gradient. These mean depths were compared with those in the field between 2001 and 2023 at our study site (see Martínez‐Blancas et al., 2022 for dataset description).

RESULTS

Effects of previous occupancy on stability, fitness differences, and species richness

Positive interspecific OSR tended to stabilize pairwise coexistence; the opposite was true for negative OSR (Figure 1A,D). Fitness differences were larger when OSRs were strong, regardless of their sign (Figure 1B,E). Thus, net effects indicate that negative OSR typically hindered coexistence. As OSR approached zero, the destabilizing N _ij weakened. However, stable coexistence remained uncommon even with very positive OSR, with average N _ij values remaining below 0 except in very deep soil (Figure 1C,F). The median stabilization I _i,j was near zero at all soil depths, indicating that OSRs tended to stabilize coexistence in as many species pairs as they were destabilizing (Figure 1G,J). In general, fitness differences caused by OSR were larger than stabilization (Figure 1H,K), so the net effect of OSR was to destabilize pairwise coexistence. As expected, this effect was more pronounced in shallow soils where OSR is stronger (Figure 1I,L). This may result from increasingly more negative intra‐ than interspecific OSR in terms of both frequency and magnitude as depth increases (Table 1).

FIGURE 1.

Pairwise coexistence statistics (stabilization [upper panels], fitness differences [middle panels], and net effect [lower panels]) related to plant–soil feedbacks. The left panels show the effect of short‐term (A–C) and cumulative (D–F) occupancy survival relationships on coexistence statistics at three soil depths. A thin‐plate spline was fitted to show trends. The right panels show the changes in coexistence statistics over soil depth based on short‐term (G–I) and cumulative (J–L) occupancy survival relationships. Thick line: median; thin lines: quartiles; dashed lines: 0.1 and 0.9 quantiles.

TABLE 1.

Changes in intra‐ and interspecific occupancy–survival relationships (OSRs) with soil depth.

Type of OSR	Depth (cm)	Fraction < 0		Mean
		Intra	Inter	Intra	Inter
Short term	4	0.353	0.474	−0.093	−0.092
	16	0.471	0.548	−0.085	−0.054
	28	0.647	0.581	−0.076	−0.017
Cumulative	4	0.353	0.474	−0.273	−0.164
	16	0.471	0.548	−0.238	−0.108
	28	0.647	0.581	−0.202	−0.053

Note: The fraction of OSRs that are negative and their mean is reported.

In line with the expectation that stability diminishes with diversity, the fraction of the possible communities that were SFCs diminished rapidly with species richness (Figure 2). No SFCs for 12 or more species were possible. The OSR–richness relationship was complex. It could be expected that predecessors that exert positive OSR increase species richness, occurring in speciose SFCs more frequently than expected by chance (Bimler et al., 2018). This was only observed in deep soils when short‐term effects were considered. When depth diminished, the trend reversed (Figure 3A). Successors suffering negative short‐term (Figure 3B) or cumulative (Figure 3D) OSRs also occurred preferentially in species‐rich SFCs, especially in deep soil. Patterns were unclear when considering the cumulative effects of plants on their successors (Figure 3C).

FIGURE 2.

Fraction of the possible communities with different species richness that would be stabilized and feasible in different soil depths. Purple bars indicate the fraction of communities considering short‐term effects; white bars indicate cumulative plant–soil feedbacks. The width of the bar in the top‐left of the figure indicates the scale of the horizontal bars.

FIGURE 3.

Species preference for stabilized and feasible communities depending on their richness as a function of plant–soil feedbacks (PSFs) measured as the average effects that species exert on others (upper panels), or the average effects that a species receives from others (lower panels). Positive slopes indicate that species involved in positive PSF tend to occur in rich communities, and the opposite occurs if slopes are negative. Results are shown for short‐term (left panels) and cumulative PSFs (right panels), and for different soil depths (colors).

At any point in the depth gradient, there were hundreds of different SFCs and several tens of nsSFCs, and all species were always present in at least one of them. Species composition changed throughout the gradient. A group of communities with similar composition (in blue in Figure 4) comprised a large fraction of the nsSFCs occurring in shallow soil, whereas in deep soil the predominant type of nsSFC (in red) was very different. Such changes are not due to differences in species richness but mainly due to composition (see Appendix S3). They are the result of some species predominating in shallow or deep soils as a result of changes in OSR. Importantly, there was a correspondence between such changes in abundance and those observed in nature. The regression of the mean depth where each species occurred in the field on the respective figure predicted by OSRs had much more support than the null model (which had a difference in AIC between models [ΔAIC] of 5.41 for short‐term OSR and 2.17 for cumulative ones).

FIGURE 4.

Species composition of all stabilized and feasible communities that occur at different soil depths for short (A) and cumulative (B) plant–soil feedbacks. Communities shown in similar colors have similar compositions based on a principal components analysis. Changes in the dominance of different colors over the depth gradient show a turnover in community composition.

Effects of previous occupancy on density

Plant density in nature depended on OSR and soil depth, and the probability q that a species was absent increased with soil depth. All the other models had little support in the data (ΔAIC > 2, see Appendix S2 for the full results). The abundance of the successors that experienced negative OSRs increased as OSRs became less negative, especially in shallow soil. Successors that experienced strong positive OSR had very low densities, regardless of soil depth (Figure 5). Patterns for cumulative OSR were similar, with the difference that the evidence for the interaction between OSR and soil depth was inconclusive (Appendix S2).

FIGURE 5.

Number of plants in 0.1 × 0.1 m quadrats depending on the sum of the occupancy–survival relationships (OSRs), an estimate of plants soil feedbacks, for all the species that previously occupied the quadrat. Different colors correspond to soil depth. Lines correspond to the fitted model estimated for shallow (8 cm) and deep (25 cm) soils. (A) Short‐term OSR and (B) cumulative feedbacks.

DISCUSSION

At first glance, PSFs do not seem to be an important driver of coexistence at our study site. For most species pairs, stabilization was not strong enough to overcome the fitness differences imposed by PSFs. However, multispecies coexistence analyses revealed several SFCs, where all species coexisted with others in at least one assemblage. Moreover, there was a turnover of assemblages over the gradient and particular groups of species were more likely to occur at certain soil depths, increasing the likelihood of coexistence due to the interaction of PSFs with environmental heterogeneity. Counterintuitively, negative interspecific PSFs were associated with greater richness, suggesting that indirect positive effects may promote coexistence through the “enemy of my enemy is my friend” scenario that has been reported at our study site (Martínez‐Blancas et al., 2022).

PSF and community stability

As observed in previous studies on PSF (Yan et al., 2022), pairwise OSRs hampered coexistence at our study site. Although we found stabilizing effects, particularly when interspecific feedbacks were positive, OSRs increased fitness differences for most species pairs, overcoming stabilization. Net stabilization, when present, was weaker for short‐term rather than for cumulative OSR, suggesting that short‐term studies underestimate the role of PSF in species coexistence as occurs in other grasslands (Chung et al., 2019; Dostálek et al., 2022). The effects of PSF on coexistence were more variable in shallow soils, reflecting stronger and more variable OSRs under such conditions (Martorell et al., 2021). As expected, coexistence was less stable there, reflecting stronger fitness differences despite more positive OSR. Such strong fitness differences may be driven by the greater performance of some species in shallow soil (Martínez‐Blancas & Martorell, 2020).

Our observations in the field were more or less consistent with the results of the pairwise analyses: Density was low where there were strong summed OSRs (S _OSR) and increased when they were weak (Figure 5). This reduction in density reflects greater fitness differences due to strong OSRs (Figure 1), as expected. In addition, S _OSR are calculated by adding the OSRs of all predecessor species, so very positive and very negative S _OSR occur likely in quadrats that were previously occupied by many species. This may be responsible for the observed pattern, because species‐rich communities were rarely stable (Figure 2). Such instability seems to be common in rich communities with positive feedbacks (Dudenhöffer et al., 2022; Eppinga et al., 2018; Mack et al., 2019), perhaps explaining the very large number of plots where density was zero when feedbacks were large and positive (Figure 5).

Despite that only a small fraction of species pairs would be able to coexist stably through PSFs, as many as 12 species, or 70% of our total, could coexist in a single SFC (although this figure diminished to 41% using the more restrictive conditions for persistence in Appendix S1). Thus, PSF may contribute to stabilizing coexistence in species‐rich communities, even if the opposite occurs when considering only pairwise effects, as has been previously suggested (Lankau et al., 2011). This emphasizes the need to move beyond pairwise analysis of PSFs and coexistence (Kandlikar, 2024; Mack et al., 2019; Miller et al., 2022; Ranjan et al., 2024; Senthilnathan & D'Andrea, 2024).

Stability–diversity relationship

Stable coexistence due to PSF has been found to become more unlikely, and the return to equilibrium becomes slower as species richness increases (Eppinga et al., 2018; Mack et al., 2019; May, 1973; Miller et al., 2022). This reflects a number of constraints to multispecies coexistence mediated by PSF: Negative feedbacks need to be stronger (Eppinga et al., 2018; Pajares‐Murgó et al., 2024), or the action of additional biological mechanisms or of intransitive cycles is required (Mack et al., 2019; Miller et al., 2022; Pajares‐Murgó et al., 2024). Our results are in line with these antecedents. The fraction of possible communities that were stabilized and feasible decreased with richness (Figure 2), and their return rates to the stable configuration after a disturbance became longer (see Appendix S1). Similarly to our pairwise analysis, where stable coexistence was less likely to occur in shallow soil due to greater fitness differences, a smaller proportion of SFCs occurred there (Figure 2). This is in line with a previous finding at the study site where lower species richness occurs in shallow soils (Martorell et al., 2015). Regarding intransitivity, it rarely resulted in persistent fluctuations of species densities in our system (it was observed in 4.5% of the persistent communities under short‐term OSR, and 2.1% for cumulative OSR; see Appendix S1). This fraction did not show any trend with species richness. Fluctuations in abundance always occur in fully intransitive PSF networks. However, in partially intransitive networks, strong negative pairwise feedback among some species (positive I _ij values in our framework) dampens such fluctuations (Mack et al., 2019). PSF intransitivity has been reported to be uncommon and weak elsewhere, but its relevance increases quickly with richness (Mack et al., 2019; Pajares‐Murgó et al., 2024). It is most likely that our SFCs contain at least some intransitive loops that are stabilized by pairwise feedbacks, contributing to coexistence.

The relationship between PSF sign, magnitude, and richness

There are two conflicting expectations on the relationship between PSF sign and diversity. First, positive interspecific PSFs should promote diversity because coexistence is favored when interspecific PSFs are less negative than intraspecific ones, as is likely to occur with interspecific facilitation (Bimler et al., 2018; Senthilnathan & D'Andrea, 2024). Accordingly, positive OSR increased stability in our pairwise analysis. Under stressful conditions, positive PSFs are more likely (Brooker et al., 2008; Martorell et al., 2021; Revillini et al., 2016). Additionally, positive PSFs may increase the range of conditions that species tolerate (Revillini et al., 2016). This may be the case with water stress if predecessors cultivate mutualist microorganisms that confer successors with drought resistance (Augé, 2001; Bennett & Klironomos, 2018; Fitzpatrick et al., 2018; Porter et al., 2020; Revillini et al., 2016). Thus, positive OSRs should be associated with higher diversity in drier, shallow soils. However, the opposite occurred: Predecessors that exerted positive OSRs increased diversity in the more benign deep soil but reduced it in shallow soil (Figure 3A). While this was observed in SFCs only when considering short‐term OSRs, the same pattern was observed for cumulative OSR under the stability criteria used in Appendix S1.

The second expected pattern is based on recent findings that negative interspecific feedbacks allow for more speciose communities (Eppinga et al., 2018, Mack et al., 2019, Dudenhöffer et al., 2022; but see Senthilnathan & D'Andrea, 2024). This was observed in our results, especially when considering the OSRs experienced by successor species (Figure 3B,D). Perhaps the large fitness differences observed for strong positive OSRs reduced diversity, although it is unclear why negative OSRs, which also increased fitness differences, behave differently. Alternatively, positive feedbacks amplify random perturbations, leading to the exclusion of some species (Dudenhöffer et al., 2022), and lower richness. As already mentioned, intransitivity may play a (seemingly secondary) role in maintaining diversity in our system (Eppinga et al., 2018; Mack et al., 2019), but it is unclear whether it should be preferentially associated with negative rather than positive interactions.

Coexistence can also occur through indirect effects that ameliorate direct negative pairwise PSFs (Kulmatiski et al., 2011). This scenario is possible when a species constrains another that would otherwise thwart a third one: the “enemy of my enemy is my friend” scenario. Intransitive networks are a special case of this effect (rock is friends with paper because it destroys scissors), but this is a more general phenomenon (see fig. 4 in Mack et al., 2019). Indirect positive interactions have been found to allow coexistence at the study site, seemingly without involving strong intransitivity (Martínez‐Blancas et al., 2022), as it happened with our PSFs. However, whether PSFs drive such indirect interactions is, to our knowledge, yet to be explored (but see Kulmatiski et al., 2011, Mack et al., 2019).

The role of space and PSF in maintaining diversity

Space may promote diversity through two mechanisms. First, each of our study species persisted in at least one SFC at every depth. Given the small scale of our quadrats, random local extinctions or colonizations from nearby plots are likely to cause the local community to shift between stable states, allowing for the simultaneous co‐occurrence of SFCs with different species in nearby plots. Fluctuations in species abundance due to intransitivity may also facilitate such shifts, as the relative abundances of species in an assemblage approach those of other SFCs over time. It is unclear how subcommunity stability affects the stability of the metacommunity, or whether it suffices to maintain species that only occur in rare SFCs. However, in another semiarid grassland, patches with more dynamic species composition were characterized by neutral to positive intraspecific PSFs with the potential to destabilize coexistence (Chung et al., 2019).

Second, the composition of SFCs varied with spatial heterogeneity. Groups of SFCs with similar compositions were more likely to occur at specific depth intervals. Thus, OSRs contributed to community replacement over the gradient, and probably to coexistence by allowing different species to predominate under different conditions. Species turnover has also previously been observed in this grassland (Martínez‐Blancas et al., 2022; Martínez‐Blancas & Martorell, 2020; Martorell et al., 2015). Moreover, PSFs seemed to generate sharp transitions between groups of similar communities at particular points in the depth gradient, especially when considering cumulative OSR. These cutoff points occurred at 12, 17, and 26 cm, matching the depths where transitions between clusters of species were reported by Martínez‐Blancas et al. (2022). These clusters seem to be driven by indirect facilitation that stabilizes coexistence (Martínez‐Blancas et al., 2022). It may be that such indirect interactions can be mediated by PSFs, allowing for species coexistence within soil depth intervals. This is in line with the prediction by Senthilnathan and D'Andrea (2024) where positive PSFs can promote coexistence by bringing species toward soil conditions that favored more species, leading to species clustering. The fact that the distribution of different species over the depth gradient predicted by OSRs resembles that observed in nature suggests that soil feedbacks play a role in species turnover and thus diversity maintenance.

Considerations in model implementation

Environmental heterogeneity may bias OSRs. Plants may live longer in plots where conspecifics or species with shared environmental preferences grew previously because local conditions benefit both predecessors and successors, resembling positive PSFs. This does not seem to drive interspecific OSRs, which are not correlated with spatial associations between species (Martorell et al., 2021). However, intraspecific positive OSRs, although more negative than interspecific ones, were still frequent. In deep soils, the frequency of intraspecific positive OSRs was small and much lower than that of interspecific OSR, but the opposite occurred in shallow soils. These could be spurious positive PSFs, but only if environmental requirements are more specific in shallow soil. This does not seem to be the case. Greater environmental specificity means less niche overlap and less competition (Keddy, 2012), but at the study site competition has repeatedly been found to intensify as soils become thinner (Martínez‐Blancas et al., 2022; Villarreal‐Barajas & Martorell, 2009). On the other hand, true positive OSRs should predominate in thin soil because plants invest heavily in cultivating mutualistic soil biota in harsh environments. Also, the same biota may change from pathogenic in benign sites to mutualistic in adverse ones (Johnson et al., 2008; Revillini et al., 2016; Sachs & Simms, 2006; Veresoglou & Rillig, 2012). In any case, a bias toward intraspecific positive values would lead to an underestimation of the contribution of PSFs to diversity maintenance because positive (or less negative) intraspecific OSR destabilizes coexistence (Equation 1; Bever et al., 1997, Kandlikar et al., 2019). Reducing such bias would not undermine the conclusion that PSFs contribute to coexistence in our study system; it would only increase their importance.

Another issue is that OSRs only consider survival, while coexistence analysis is based on population growth rates that also depend on plant size and fecundity. This adds noise to our results. However, the fact that species turnover with soil depth was predicted from OSRs suggests that they capture fundamental aspects of population dynamics.

Conclusions

Our results respond to the call of previous authors that we need to move beyond the study of species coexistence using pairwise models. Perhaps the most important contributions of PSFs to coexistence became apparent in multispecies analyses: Persistence of all species would be favored by PSFs in different assemblages and in different environmental conditions. In pairwise models, facilitation seemed to promote coexistence; however, it hindered species richness in the more realistic multispecies scenarios perhaps because negative PSFs are the ones that promote coexistence through intransitive loops or “enemy of my enemy is my friend” dynamics. Integration over different spatial scales is also important, as PSFs may enhance diversity when environmental heterogeneity and multiple stable states are considered.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Appendix S1.

Appendix S2.

Appendix S3.

Martorell, Carlos , and Martínez‐Blancas Alejandra. 2025. “Plant–Soil Feedbacks Contribute to Coexistence When Considering Multispecies Assemblages over a Soil Depth Gradient.” Ecology 106(3): e70052. 10.1002/ecy.70052

DATA AVAILABILITY STATEMENT

Historic data for each subquadrat and longevity data for all plants (Martínez Blancas et al., 2021) are available in Figshare at https://doi.org/10.6084/m9.figshare.14767494.v1. Abundance data in each subplot (Martorell, 2025) are available in Figshare at https://doi.org/10.6084/m9.figshare.28153094.v1. Code and data for calculating the effect of short‐term and cumulative plant–soil feedback (Martorell & Martínez‐Blancas, 2025) are available in Zenodo at https://doi.org/10.5281/zenodo.14563291.