The recovery of large carnivores offers unique opportunities to study their cascading impacts on plant population dynamics. Medium-sized carnivores, both prey and seed dispersers, are suppressed by apex predators, indirectly increasing seed-eating rodent’s populations and potentially altering plant establishment. We investigated how natural variation in the presence of the Iberian lynx (Lynx pardinus), a top predator in southern Spain, triggered cascading effects on the recruitment of the Iberian pear (Pyrus bourgaeana) through altered seed dispersal patterns by mesopredators and post-dispersal seed predation by rodents. To assess whether and how the seed-dispersal effectiveness of the Iberian pear was influenced by lynx presence across different habitats (open, forest) and microsites (shrub, rock and open), we conducted field experiments and observations spanning multiple life-cycle stages of this fleshy-fruited tree mainly dispersed by carnivorous mammals. Path analysis revealed that lynx presence decreased seed dispersal by 80% and biased it toward forests, where seedling survival was extremely low (1%). Most of the seeds were delivered in open microsites (61%), particularly in lynx absence by the red fox. Although we detected no direct effect of lynx presence on post-dispersal seed predation, rodents removed 49% and 116% more seeds under shrubs than in rock and open interspaces, respectively, negatively affecting plant recruitment. Since shrubs provided the most favourable conditions for seedling survival, particularly in open habitats, these results expose a seed-seedling conflict, whereby microsites with the highest seed predation are also those that maximize seedling establishment. This may limit the expansion potential of the Iberian pear, and likely other fleshy-fruited species, under the current scenario of apex predators rewilding. Reintroduction programs of threatened carnivores should account for trophic cascades that may disrupt frugivory interactions and ultimately shape plant recruitment and establishment. This is especially relevant in defaunated ecosystems, where plant–animal mutualisms are often compromised.

Collection/generation of data:

This study was conducted in the Sierra de Andújar Natural Park (SANP), southern Spain (38°14’27.71” N, 4°4’45.03” W), home of one of the largest wild populations of the Iberian lynx that has recovered from near-extinction in the 1990s without the need for reintroductions. We collected data of four different life-cycle stages of *Pyrus bourgaeana, *a mammal-dispersed tree (Fedriani and Delibes 2009): 1) seed dispersal, 2) post-dispersal seed predation, 3) seedling emergence and 4) 1-year seedling survival. 

1) To assess the impact of lynx presence on Pyrus bourgaeana seed dispersal, a mammal-dispersed tree, we used the seed dispersal data from Burgos et al. (2024) (https://doi.org/10.5061/dryad.b2rbnzsph). This study assessed seed dispersal by frugivorous meso-carnivores in five localities with permanent territorial presence of reproductive Iberian lynxes and five control localities without lynxes in SANP. Burgos et al. (2024) measured seed dispersal in two distinct habitat types: a) mature forest with a dense shrub layer, and b) open landscapes with sparse shrub cover and isolated or small clumps of trees. Meso-carnivore scat sampling was carried out by two mammal scat experts along 1.5 km x 3 m transects in each habitat type during two consecutive fruiting seasons, from October to March of 2018-19 and 2019-20. Given the importance of microsite type for plant recruitment (Schupp 1993, Escribano-Ávila et al. 2014), we differentiated and recorded three microsites of scat deposition: a) beneath or on a small rock (rock), b) beneath or on a shrub (shrub), and c) on open ground (open). We recorded scats from red foxes (Vulpes vulpes) and stone martens (Martes foina). Pyrus bourgaeana seeds in faecal samples were visually identified based on morphology, then extracted and manually counted from the dried scats.

2) To examine whether lynx presence indirectly affects seed survival through changes in rodent-mediated seed predation, we conducted a field experiment simulating natural seed deposition by mesopredators. We estimated post-dispersal seed predation of P. bourgaeana seeds by rodents in November 2020, coinciding with the species' seed-dispersal peak (Fedriani and Delibes 2009). Seed removal rates were assessed through a field experiment in two localities: one with Iberian lynx presence and one control site without lynxes. We placed 1,152 P. bourgaeana seeds in 144 Petri dish depots (eight seeds per depot). As seed removal intensity varies across habitats and microsites, depots were distributed across six combinations of habitat type (forest and open) and microsite (rock, shrub, and open), simulating dispersed seeds by meso-carnivores (e.g. Garrote et al. 2019). In each habitat-microsite combination, we set up depots in six spatially independent plots (15 x 15 m) spaced 200 m apart—sufficient distance given rodents' limited home ranges (~1 ha; Rosalino et al. 2011). We placed eight seeds in each Petri dish at the start of the experiment, monitored them over five consecutive days, and recorded the total number of seeds removed. To account for potential seed removal by other consumers (e.g. ants, birds; Warzecha and Thomas Parker 2014, Suárez-Esteban et al. 2018), we installed 36 camera traps (Scoutguard SG562-C; white LED) at a subset of 25% of the seed depots. The camera traps confirmed that rodents were the sole seed predators in our experiment. Seed removal rates per depot were calculated as the total number of seeds removed after five days divided by the number of seeds initially placed.

3 and 4) To assess how environmental conditions influence Iberian pear recruitment probability, we carried out a sowing experiment tracking seedling performance over one year. Assuming that lynx presence does not directly affect P. bourgaeana seedling emergence and survival, we estimated these rates in a single locality within our study area. This experiment was replicated across ten spatially independent plots (200 m apart) of 15 x 15 m. The experiment began in November 2020 and was monitored until February 2022, with visits to record seedling emergence and survival approximately every two months. We sowed 360 P. bourgaeana seeds (12 per depot) across the same seed-dispersal habitats (forest and open) and microsites (rock, shrub, and open) as mentioned earlier. We calculated emergence rates as the number of seedlings emerged divided by the total number of seeds sown, and survival rates as the number of seedlings surviving at the end of the experiment divided by the total number of seedlings emerged.

Processing the data:

To generate the input data for the structural equation models (see below), we estimated stage-specific transition probabilities (TPs) under contrasting ecological scenarios based on our empirical data. TPs were calculated for seed survival, seedling emergence, and seedling survival as the ratio of individuals completing a stage to those entering the stage (e.g., Rey and Alcántara 2000, Balcomb and Chapman 2003, Garrote et al. 2022b). Given the limited sample sizes used to estimate some mean observed TPs, we fitted Bernoulli distributions (n = 100) to the observed TPs for each lynx scenario, habitat, and microsite. We then resampled randomly without replacement from each distribution, truncated by the standard deviation of the observed TP (Manly 2006), to introduce moderate stochasticity into the path analysis (see below). To estimate how many individuals were retained at each stage under each ecological scenario, we multiplied the stage-specific TPs by the corresponding number of individuals entering that stage. Specifically, post-dispersal seed survival rates were multiplied by the number of dispersed seeds to estimate the number of surviving seeds. Following the same approach, emergence rates were applied to the number of surviving seeds, and seedling survival rates to the number of emerged seedlings, thus estimating the number of recruits (i.e., surviving seedlings). Finally, we created a dataset with the number of dispersed seeds, seeds that survived post-dispersal predation, emerged seedlings, and seedlings that survived their first year of life, for each habitat type, microsite, and lynx scenario.

Statistical analyses:

To integrate all ecological filters and assess the direct and indirect effects of lynx presence, habitat, and microsite on the Iberian pear recruitment, we applied structural equation modelling (SEM). This analysis covered seed deposition by mesopredators through to seedling establishment, based on confirmatory path analysis (Shipley 2000). Prior to the SEM, we first fitted separate mixed models (GLMMs) with lme4 package (Bates et al. 2020) for each life-cycle stage to report the general patterns observed in each experiment. We then employed the piecewiseSEM R package (Lefcheck 2016) to fit structural equations with GLMMs, specifying lynx presence/absence, habitat type (forest and open), and microsite type (rock, shrub, and open) as exogenous variables. Endogenous variables included: number of i) mesopredator scats, ii) dispersed seeds, iii) post-dispersal surviving seeds, iv) emerged seedlings, and v) first-year surviving seedlings. Poisson errors and the random effect ‘Locality’ were specified in all GLMMs. Standardised path coefficients were used to compare direct effects across different scales. We calculated standardised path coefficients separately for lynx presence and absence scenario in order to explore interactions between habitat, microsite type, and lynx presence. Indirect and total effects were also computed to assess the relative importance of each exogenous variable on seedling survival (Shipley 2000; see details in Appendix S2). Marginal and conditional pseudo-R² were calculated to estimate the variance explained by each model. Residuals were inspected using the DHARMa package v. 0.4.6 (Hartig and Lohse 2022).

References:

• Balcomb, S. R. and Chapman, C. A. 2003. Bridging the gap: Influence of seed deposition on seedling recruitment in a primate-tree interaction. - Ecol. Monogr. 73: 625–642.
• Burgos, T. et al. 2024. Apex predators can structure ecosystems through trophic cascades: Linking the frugivorous behaviour and seed dispersal patterns of mesocarnivores. - Funct. Ecol. 00: 1–13.
• Fedriani, J. M. and Delibes, M. 2009b. Seed dispersal in the iberian pear, pyrus bourgaeana: A role for infrequent mutualists. - Écoscience 16: 311–321.
• Garrote, P. J. et al. 2019. Extrinsic factors rather than seed traits mediate strong spatial variation in seed predation. - Perspect. Plant Ecol. Evol. Syst. 38: 39–47.
• Garrote, P. J. et al. 2022. Coping with changing plant–plant interactions in restoration ecology: Effect of species, site, and individual variation. - Appl. Veg. Sci. 25: e12644.
• Hartig, F. and Lohse, L. 2022. Package “DHARMa” Residual Diagnostics for Hierarchical (Multi-Level / Mixed) Regression Models. - CRAN-R Packag.
• Lefcheck, J. S. 2016. piecewiseSEM: Piecewise structural equation modelling in r for ecology, evolution, and systematics. - Methods Ecol. Evol. 7: 573–579.
• Manly, B. F. 2006. Randomization, bootstrap and Monte Carlo methods in biology (Vol. 70). - CRC press.
• Rey, P. J. and Alcántara, J. M. 2000. Recruitment dynamics of a fleshy-fruited plant (Olea europaea): Connecting patterns of seed dispersal to seedling establishment. - J. Ecol. 88: 622–633.
• Rosalino, L. M. et al. 2011. Usage patterns of Mediterranean agro-forest habitat components by wood mice Apodemus sylvaticus. - Mamm. Biol. 76: 268–273.
• Shipley, B. 2000. Cause and Correlation in Biology. A User’s Guide to Path Analysis, Structural Equations and Causal Inference. - Cause Correl. Biol. in press.
• Suárez-Esteban, A. et al. 2013. Unpaved road verges as hotspots of fleshy-fruited shrub recruitment and establishment. - Biol. Conserv. 167: 50–56.
• Warzecha, B. and Thomas Parker, V. 2014. Differential post-dispersal seed predation drives chaparral seed bank dynamics. - Plant Ecol. 215: 1313–1322.