Evidence for antagonistic effects of climate change and exotic pathogens on regeneration of Mediterranean forests

174 | Journal of Ecology. 2024;112:174–188.wileyonlinelibrary.com/journal/jec Received: 23 February 2023 | Accepted: 3 August 2023 DOI: 10.1111/1365-2745.14227 RESEARCH ARTICLE Evidence for antagonistic effects of climate change and exotic pathogens on regeneration of Mediterranean forests Pablo Homet1,2 | Luis Matías3 | Oscar Godoy4 | Lorena Gómez-Aparicio1 This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. © 2023 The Authors. Journal of Ecology published by John Wiley & Sons Ltd on behalf of British Ecological Society. 1Instituto de Recursos Naturales y Agrobiología de Sevilla, Consejo Superior de Investigaciones Científicas (IRNASCSIC), Sevilla, Spain 2Estación Biológica de Doñana, Consejo Superior de Investigaciones Científicas (EBD-CSIC), Calle Américo Vespucio, Sevilla, Spain 3Departamento de Biología Vegetal y Ecología, Universidad de Sevilla, Sevilla, Spain 4Departamento de Biología, Instituto Universitario de Investigación Marina (INMAR), Universidad de Cádiz, Puerto Real, Spain Correspondence Lorena Gómez-Aparicio Email: [email protected] Funding information European Regional Development Fund; European Social Fund; Ministerio de Ciencia e Innovación, Grant/Award Number: CGL2014-56739-R, PID2019- 108288RA-I00, RTI2018-094394-B-I00 and RYC-2017-23666; Universidad de Sevilla Handling Editor: Catherine Preece Abstract 1. Understanding the interactive effects of global change drivers on tree demography is fundamental for realistic predictions of future forest dynamics. Multiple studies have shown increasing drought and exotic pathogens to severely threaten forest persistence by increasing mortality and decreasing growth of adult trees. However, much less is known about their effects on regeneration, and how they might affect seedling performance in additive and non-additive (synergistic or antagonistic) ways. 2. Here we aimed to fill this gap by experimentally exploring the effects of increasing drought and soil-borne pathogens on tree regeneration in two types of mixed oak forests (Quercus suber-Q. canariensis and Q. suber-Olea europaea) invaded by the exotic soil-borne oomycete Phytophthora cinnamomi, one of the most aggressive plant pathogens on earth. We conducted a seed sowing experiment with oomycete-specific fungicide taking advantage of rainfall exclusion infrastructures that excluded 30% of the annual rainfall, simulating predictions of climate change models for Mediterranean systems. Seedling emergence, survival and growth of the three tree species were followed during 3 years. 3. We found that neutral or positive drought effects on regeneration dominated over negative effects in the tree community. Moreover, most positive drought effects on the dominant species (Q. suber) were not direct, but indirectly mediated by soil-borne pathogens. This was shown by the fact that positive drought effects disappeared with fungicide application. 4. Synthesis: Overall, our results suggest that rainfall reductions predicted by climate change models for the Mediterranean region might have minor direct negative effects on early regeneration of tree species, but could play a major indirect role by limiting the negative effects of exotic pathogens on highly susceptible tree species. These findings highlight that antagonisms among global change drivers should be recognized as important forces that might slow down the current loss of tree health. | 175 HOMET et al. 1 | INTRODUCTION Global change is severely modifying tree demographic rates in forests world-wide (Allen et al., 2015; McDowell et al., 2020; Millar & Stephenson, 2015). In the last decades, multiple studies have aimed to disentangle how growth and survival of adult trees are being changed by higher CO2 levels, a warmer climate, increasing drought or the attack by exotic pests and pathogens (Carnicer et al., 2011; Seidl et al., 2017; Sperry et al., 2019). Most of these studies have found that global change drivers are increasing mortality rates and decreasing growth of adult trees, and that this reduction in tree health might threaten the role of forest ecosystems as biodiversity reservoirs and carbon sinks (Anderegg et al., 2020; Brienen et al., 2015; Trumbore et al., 2015). However, understanding forest dynamics under global change requires not only evaluating the capacity of adult trees to grow and survive, but also to regenerate. Establishment of new individuals is a fundamental process for the maintenance of tree populations and their post-disturbance recovery (Batllori et al., 2020; Davis et al., 2019). Despite its relevance, the number of studies that have explored the impacts of global change drivers on tree regeneration is still relatively low compared to the many studies focused on adult trees (García de Jalón et al., 2020; Lloret et al., 2004; Matías et al., 2012). Among forest ecosystems, Mediterranean forests are considered particularly threatened by global change (Peñuelas et al., 2017). On the one hand, the Mediterranean basin is a hot-spot of climate change, with a clear trend towards drier conditions due to the combination of higher temperatures and rainfall reduction (Giorgi & Lionello, 2008; IPCC, 2021). A drier climate is imposing a high stress level on a type of system where water is already a main limiting resource for plant species performance (Acácio et al., 2017; Peñuelas & Sardans, 2021). On the other hand, the Mediterranean basin is also a hot-spot of exotic species, due to its long history of human occupancy and high diversity of environments (Cherif et al., 2020). In Mediterranean countries, the number of invasive forest pathogens has increased exponentially in the last four decades influenced by an unprecedented growth of international travel and trade (Garbelotto & Pautasso, 2012; Santini et al., 2013). Among these new pathogens, oomycetes have shown the largest increase, and particularly those belonging to the genus Phytophthora, which represents one of the most important genera of plant pathogens on earth (Burgess et al., 2021; Jung et al., 2018; Scott et al., 2019). Exotic oomycetes have decimated populations of several Mediterranean woody species (Jung et al., 2018), and have been recently shown to be an important factor limiting forest regeneration (Domínguez-Begines et al., 2020, 2021). However, it is uncertain how regeneration dynamics of Mediterranean forests might be affected by the simultaneous, interactive effects of exotic pathogens and a drier climate. Conducting studies that simultaneously explore the effect of several global change drivers on tree demography is a complex and urgent challenge, since combined effects of global change drivers might not be always predicted from studies that explore their effects in isolation (Côté et al., 2016; Sage, 2020; Simler-Williamson et al., 2019; Tylianakis et al., 2008). From a theoretical perspective, three different scenarios of interaction among global change drivers might be expected: no interaction, synergistic interaction and antagonistic interaction (Figure 1). The lack of interaction among drivers would represent the simplest scenario, where the total effect of both drivers would be equal to the sum of their individual effects (Figure 1a). A synergistic interaction would happen when the combined effect of the two drivers is of larger magnitude KEYWORDS experimental drought, global change ecology, interactive effects, oomycetes, Phytophthora cinnamomi, rainout shelters, seedling performance, soil-borne pathogens FIGURE 1 Theoretical scenarios of interaction among two global change drivers (increasing drought and exotic pathogens): (a) no interaction, where the negative effect of increasing drought on seedling performance is the same with or without pathogens; (b) synergistic interaction, where the negative effect of increasing drought is larger in the presence of pathogens (as shown by steeper slope of the red line); (c) antagonistic (qualitative) interaction, where the negative effect of increasing drought is smaller in the presence of pathogens (as shown by slower slope of the red line); (d) antagonistic (quantitative) interaction, where the effect of increasing drought on seedling performance changes of sign (from negative to positive) depending on the presence of pathogens (as shown by the crossing of the red and blue lines). 13652745, 2024, 1, Downloaded from https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2745.14227 by Universidad De Sevilla, Wiley Online Library on [16/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 176 | HOMET et al. than the sum of their individual effects (Figure 1b). For example, in a forest affected by increasing drought and exotic pathogens, drought might amplify the negative effects of pathogens due to a weakening effect on trees that make them more vulnerable to biotic attack (Corcobado et al., 2014; Desprez-Loustau et al., 2006). Finally, an antagonistic interaction would happen when the combined effects of both drivers is of smaller magnitude than the sum of their individual effects (Figure 1c). In this case, negative effects of pathogens on tree performance could be smaller in a drier climate due to negative effects of drought on the population growth rate of pathogens, since most tree pathogens are fungi or oomycete that need water for the germination and dispersal of their spores (Erwin & Ribeiro, 1996; Homet et al., 2019). Antagonistic interactions could be not only quantitative, but also qualitative, if they involved a change of sign (from negative to positive, or vice versa) in the effect of one of the drivers (Figure 1d). This represents an extreme situation of antagonistic effects very rarely reported in the literature. In this study, we aimed to experimentally explore the individual and interactive effects of increasing drought and exotic pathogens on tree regeneration dynamics in mixed Mediterranean oak forests invaded by the exotic soil-borne oomycete Phytophthora cinnamomi, one of the most aggressive plant pathogens on earth (Lowe et al., 2000). To this end, we conducted a seed sowing experiment with oomycete-specific fungicide taking advantage of rainfall exclusion infrastructures installed in the southernmost European oak forests, located in the mountain range north of the Strait of Gibraltar. This area contains the largest and best conserved European forests of Quercus suber, but currently suffers important problems of oak decline due to P. cinnamomi invasion (Ávila et al., 2017; Gómez-Aparicio et al., 2012). The experiment was replicated in two different types of mixed oak forests: a closed forest of Q. suber and Q. canariensis, and an open woodland of Q. suber and Olea europaea. The three species differ in their susceptibility to drought (Q. canariensis > Q. suber > O. europaea; Gómez- Aparicio, Pérez-Ramos, et al., 2008; Ibáñez et al., 2017) and P. cinnamomi attack (Q. suber > O. europaea > Q. canariensis; Gómez et al., 2020). The rainfall exclusion infrastructures excluded 30% of the annual rainfall, simulating predictions of climate change models for southern Europe (IPCC, 2021). Seedling emergence, survival and growth of the three tree species were followed during 3 years to analyse their response to rainfall exclusion and the application of oomycete-specific fungicide. As in previous studies (Domínguez-Begines et al., 2020, 2021), we assumed that P. cinnamomi would be the main oomycete species driving fungicide effects due to its fundamental role as a driver of oak decline in southern Europe (Brasier, 1992; Camilo- Alves et al., 2013; Moricca et al., 2016) and its high abundance in the study area (Gómez-Aparicio et al., 2012; Serrano et al., 2022). This assumption is also supported by the fact that, although other Phytophthora species have been recently described in Spanish forests (e.g. P. quercina, P. psychrophila, P. gonapodyides, P. cactorum, Catalá et al., 2017; Mora-Sala et al., 2018; Ruiz-Gómez et al., 2019), pathogenicity tests consistently find that they are much less aggressive than P. cinnamomi to oak species (Corcobado et al., 2017; Mora-Sala et al., 2019; Pérez-Sierra et al., 2013; Seddaiu et al., 2020). We hypothesized that the two experimental factors (drought and pathogens) would have negative effects on regeneration, with the magnitude of such effects varying among tree species according to their differential susceptibility to either abiotic or biotic stress. Moreover, based on previous studies conducted under controlled conditions (Homet et al., 2019), we expected drought and pathogens to have interactive (antagonistic) effects, particularly on Q. suber, the most susceptible species to P. cinnamomi. Finally, because in mixed forests tree species have species-specific footprints on the above- and below-ground environment that determine spatial patterns of regeneration (Gómez-Aparicio, Canham, et al., 2008; Ibáñez et al., 2017), we hypothesized that the impact of a given global change stressor on the success of seedling establishment might differ under different canopy species. Overall, the results of this study will contribute to improve our ecological understanding of the interactive effects of two main global change drivers—increasing drought and exotic pathogens—that, to our knowledge, have never been experimentally explored before under natural conditions. 2 | MATERIALS AND METHODS 2.1 | Study site and species The study was conducted from 2017 to 2019 at Los Alcornocales Natural Park (Cádiz), in southern Spain (Figure S1). The climate is subhumid Mediterranean; rainfall concentrates (95%) from October to May, ranging from 620 mm to 2100 mm (Junta de Andalucía, 2017). This variability in annual precipitation was well represented during the study: 2017 was an average year (980 mm), 2018 was a rainy year (1384 mm) and 2019 was a dry year (735 mm). Soils are sandy, acidic and nutrient poor, derived from bedrock dominated by Oligo– Miocene sandstones. We selected two study sites located 4 km apart and representative of the main mixed Q. suber forests of the park: a closed forest where Q. suber coexists with the deciduous, shade-tolerant Q. canariensis (closed forest, hereafter) and an open woodland where Q. suber coexists with the evergreen and drought-tolerant Olea europaea var. sylvestris (open woodland, hereafter; Figure S1). Closed forests appear in wet areas on sandy soils, and usually have a sparse understorey dominated by tall shrubs, particularly Erica arborea. Open woodlands generally appear in the drier lowlands of the area on clayish soils prone to temporal waterlogging during the wet season, and have a dense and rich understorey composed by shrub species such as Pistacia lentiscus, Genista spp., Ulex spp., Erica spp. and Cistus spp. Soils at the two study sites were invaded by the oomycete soil-borne pathogen P. cinnamomi (Homet et al., 2021; Serrano et al., 2022), but only a few Q. suber trees showed above-ground symptoms of decline at the time of the experiment. Although P. 13652745, 2024, 1, Downloaded from https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2745.14227 by Universidad De Sevilla, Wiley Online Library on [16/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 177 HOMET et al. cinnamomi has the ability to infect the roots of the three tree species (at least during the seedling stage), Q. suber is considered more susceptible to the pathogen than O. europaea and Q. canariensis (Gómez et al., 2020; Moralejo et al., 2009). 2.2 | Rainfall exclusion general design This study was conducted taking advantage of a rainfall exclusion infrastructure built to evaluate the combined effect of climate change and the exotic pathogen P. cinnamomi on the dynamics and functioning of Mediterranean mixed oak forests. Permission to build the infrastructure and conduct the fieldwork in private property was provided by the Jaime González-Gordon Foundation. At each of the two study sites (closed forest and open woodland), we established six plots of 20 × 15 m during early spring 2016. Plots were distributed in three pairs (hereafter, blocks), with a mean distance of 30 m among blocks. From each pair, one plot was assigned to a rainfall exclusion treatment, whereas the other plot served as a control (Figure S2). All plots were fenced to avoid the interference of large wild animals on the experiment. The rainfall exclusion treatment was designed to reduce in 30% the amount of rainfall, reproducing future climate change scenarios for southern Spain (model CMIP5 for the scenario RCP 8.5 for the period 2081–2100; Stocker, 2013). Rainfall was excluded using PVC gutters that covered 30% of the plot area in order to exclude a similar amount of natural precipitation. Gutters were placed at least 1 m height over the forest floor. In control plots, the same gutters were placed upside down to simulate potential secondary effects of the infrastructure on the forest microenvironment while maintaining the natural precipitation regime (Figure S2). Tree density, basal area and light levels were similar for rainfall exclusion and control plots (Table S1). A buffer area of 1 m was considered within each plot in order to avoid any possible edge effects. All trees within the plots and outside the buffer area were individually marked (three to four trees per plot and species), making a total of 21 Q. suber and 22 Q. canariensis in the closed forest, and 21 Q. suber and 22 O. europaea in the open woodland (n = 86 trees in total). P. cinnamomi was isolated in the soil of all trees before the start of the experiment (Serrano et al., 2022, unpublished data). The effect of the rainfall exclusion treatment on soil moisture was quantified at four depths (10, 20, 30 and 40 cm) with a PR2 Profile Probe, Delta-T Devices (UK). Measurements were conducted every 3–4 weeks at six points per plot where access tubes were permanently installed for easy probe insertion and removal. Access tubes were installed in November 2015, to have a record of soil moisture in the different plots before the start of the rainfall treatments in March 2016. Soil texture (0–20 cm) was measured next to each access tube because previous studies in the area have shown a large variability in the sand content at small spatial scales (Gómez- Aparicio et al., 2012), which might affect soil moisture measurements and therefore the quantification of the effects of the rainfall exclusion treatment. 2.3 | Seed sowing experiment We conducted a sowing experiment using seeds of the two dominant species at each forest type: Q. suber and Q. canariensis in the closed forest, and Q. suber and O. europaea in the open woodland. In fall 2016, acorns of Quercus species were collected from at least 10 adult individuals per species in the study area. Due to the difficulties to collect and clean large quantities of O. europaea seeds, we bought them from a specialized supplier close to the study area (Semillas Silvestres S.L.). Seeds were stored in humid substrate at 4°C until sowing. Before sowing, healthy Q. suber acorns were selected by flotation (Gribko & Jones, 1995). To reduce differences in seedling performance due to acorn size, we selected only healthy acorns with a weight within the second and third quartile (25%–75%) of the whole acorn population (i.e. 5.77 ± 1.57 g for Q. suber and 5.16 ± 1.46 for Q. canariensis). Selected acorns were surface sterilized with a 10% bleach solution previous to sowing. O. europaea seeds were subjected to mechanical scarification and soaked in water for 5 h previous to sowing in order to break dormancy and improve germination (Baskin & Baskin, 1998; Costa & Sánchez, 2001). In December 2016, two 25 × 25 cm sowing quadrats (‘Control’ and ‘Fungicide’) were established under each of the 86 study trees separated at least 30 cm. At each quadrat we established 16 sowing points separated 5 cm from each other. In the closed forest, one acorn of Q. canariensis or Q. suber was sown per point at 2 cm depth, making a total of 688 seeds per species (8 seeds × 2 treatments × 43 trees). In the open woodland, one acorn of Q. suber was also sown per point (n = 688 acorns). However, we sowed a higher number of O. europaea seeds (17 seeds per point, 11,696 seeds in total) to compensate for the low germination rate of the species (see Domínguez-Begines et al., 2020 for a similar approach; Figure 2). Seeds were covered with a metal mesh to avoid seed predation by rodents. Seeds in the fungicide quadrats were treated monthly with an oomycete-specific fungicide (Armetil 25 WP, 25% metalaxyl; Industrias Químicas del Valles SA, Barcelona, Spain) applied as a spray (1.6 gr/L water solution) on seedlings and soil. Control seedlings were treated with an identical volume of water. Armetil is a metalaxyl-based, systemic fungicide that interferes with oomycete development through the inhibition of RNA (Schwinn & Staub, 1987; Sukul & Spiteller, 2000). Metalaxyl does not affect true fungi (Cohen & Coffey, 1986), so its use minimizes the likelihood of side effects on microbial biomass and nutrient dynamics associated with the application of broad-spectrum fungicides (Chen et al., 2001). Armetil has been previously used to successfully disentangle the role of oomycetes pathogens for tree seedling survival and growth under both natural (Domínguez-Begines et al., 2020) and controlled conditions (Domínguez-Begines et al., 2021). Seedling emergence, survival and growth were followed monthly during 34 months. At each census, the state of each seed (emerged/ not emerged) or seedling (alive/dead) was registered. In October 2019, all the living seedlings were harvested, including the root system to a standardized depth of 25 cm. Once in the laboratory, seedlings were divided into stem, leaves and roots. All plant material was dried at 70°C for a minimum of 48 h to estimate shoot and root 13652745, 2024, 1, Downloaded from https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2745.14227 by Universidad De Sevilla, Wiley Online Library on [16/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 178 | HOMET et al. biomass. Because Quercus species develop a deep tap root that was not possible to recover in its totality in the field, only fine roots (i.e. roots <2 mm in diameter) were considered for further analyses. 2.4 | Data analysis All analyses were conducted separately for each combination of forest type and seedling species. We used mixed effects Cox's proportional hazards models (Cox, 1972) to explore the effect of the experimental treatments on time to emergence along the first 5 months of the experiment (emergence rate), as well as on time to death along the whole duration of the experiment (survival rate). Rainfall exclusion, fungicide, canopy species and all double and triple interactions were included as fixed factors. Acorn weight was included as a covariate in all models of Quercus species to control for its large effect on seedling performance (Pérez-Ramos et al., 2013; Quero et al., 2007). To control for the nested structure of the experimental design, block was as included as a fixed factor and the sowing quadrat as a random factor nested in tree. Survival analyses were performed using packages ‘survival’ (Therneau, 2020a) and ‘coxme’ (Therneau, 2020b). Model fit was measured using root mean squared error (RMSE) and mean absolute error (MSA), with lower values being indicative of better fits. Because of the long duration of the experiment, we also used GLMMs to evaluate treatment effects on survival at different time points: after the first summer (the period of usually largest seedling mortality in Mediterranean systems; Gómez-Aparicio, Canham, et al., 2008, Ibáñez et al., 2017), after the second summer (i.e. after 22 months) and at the end of the study (i.e. after three summers, 34 months). The structure of the full model was the same as for the Cox's proportional hazards models. The same full GLMM was applied for the analysis of the shoot and fine root biomass of the seedlings harvested at the end of the study. Emergence and survival were analysed using a binomial distribution, whereas shoot and fine root biomass were analysed using a normal distribution. In all cases, we followed a process of model comparison and simplification using AICc to find the best model (i.e. lowest AICc). GLMMs were performed using package ‘lme4’ (Bates et al., 2015) and compared with the MUMIN package (Barton, 2022). The MUMIN package was FIGURE 2 Design of the seed sowing experiment. At each of the two forest sites (closed forest and open woodland), seeds were sown in three control and three rainfall exclusion plots. Within each plot, sowing squares were installed under the canopy of each canopy tree (21 Quercus suber and 22 Quercus canariensis in the closed forest, 21 Q. suber and 22 O. europaea in the open woodland, 43 trees per site). Two 25 × 25 cm sowing quadrats (‘Control’ in brown, ‘Fungicide’ in red) were established under each tree, with 16 sowing points. In the closed forest, one acorn of Q. canariensis or Q. suber was sown per point (i.e. eight seeds per species within each quadrat). In the open woodland, one acorn of Q. suber was also sown per point, whereas a higher number of O. europaea seeds was sown per point (17 seeds) to compensate for the low germination rate of the species (i.e. 136 Olea seeds within each quadrat). 13652745, 2024, 1, Downloaded from https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2745.14227 by Universidad De Sevilla, Wiley Online Library on [16/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 179 HOMET et al. also used to calculate the variance explained by the GLMM models (marginal and conditional R2), according to the method proposed by Nakagawa and Schielzeth (2013). All statistical analyses were conducted with R software version 4.1.1 (R Core Team, 2021). 3 | RESULTS 3.1 | Effect of rainfall exclusion on soil moisture Soil moisture levels showed some differences among drought and control plots before the installation of the rainfall exclusion infrastructure (Table S2). However, the rainfall exclusion treatment clearly magnified these differences across the different soil depths (Table S3). The largest differences occurred after large rainfall events, as those occurred during the spring of 2018, where soil moisture reached values 33% higher in control than drought plots (Figures S3 and S4). 3.2 | Seedling emergence Emergence of Quercus species occurred progressively during the first 5 months of the experiment, reaching values up to 85% for Q. suber (Figure 3a) and 75% for Q. canariensis (Figure 3b) in the closed forest, and 92% for Q. suber in the open woodland (Figure 3c). Emergence of O. europaea occurred during the first and second springs of the experiment, and was much lower than for Quercus species (10%) (Figure 3d). Best models of emergence included the effect of the rainfall exclusion treatment for the two Quercus species, but not for O. europaea (Tables 1 and 2). Effects of rainfall exclusion on oak emergence were species and site specific. In the closed forest, the rainfall exclusion effect on Q. suber emergence was strongly modulated by the fungicide (Rainfall exclusion × Fungicide interaction, Table 1), changing from slightly negative with fungicide to very positive without fungicide (Figure 4a). At this same forest, rainfall exclusion negatively affected emergence of Q. canariensis, which emerged slower and had lower final emergence in rainfall exclusion plots (0.73) than under control conditions (0.80, Figure 3b). In the open woodland, rainfall exclusion had a positive effect on the emergence rate of Q. suber (Figures 2c and 3e). Fungicide application increased the emergence rate in all species-site combinations, except for Q. suber in the open woodland (Tables 1 and 2, Figure 3). Among the three species, O. europaea showed the largest positive fungicide effects, particularly under conspecific Olea trees (Fungicide × Canopy species interaction; Table 2), where emergence increased by 39% with the application of fungicide (Figure S5d). FIGURE 3 Emergence rate of Quercus suber (a) and Quercus canariensis (b) in the closed forest, and of Quercus suber (c) and Olea europaea (d) in the open woodland. Emergence is given separately for each of the four experimental treatment combinations (RE, rainfall exclusion; C, control; F, with fungicide; NF, no fungicide). Germination of the two Quercus species took place during the first spring of the experiment, whereas germination of Olea europaea took place during the first and second spring. 13652745, 2024, 1, Downloaded from https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2745.14227 by Universidad De Sevilla, Wiley Online Library on [16/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 180 | HOMET et al. TABLE 1 Best models analysing the effect of rainfall exclusion (RE), fungicide (F) and canopy species (SP) on emergence and survival rates along the study period, survival at different time points (first, second and third year), and final shoot and fine-root biomass of experimental seedlings in the closed forest. Full models also included seed weight (SW) as a covariate and block (B) to control for the spatial structure of the experimental design. ∆AICc represents the differences between the AICc (Akaike information criterion corrected for small sample sizes) of the best and null models. The total number of parameters in the best model (‘Par’) is also given. Conditional R2 (Rc 2) reflects the variance explained by all factors (fixed and random) in the GLMMs, whereas marginal R2 (Rm 2) reflects the variances explained only by fixed effects. Seedling species Variable Best model NPar ∆AICc Rc 2Rm 2 Quercus suber Emergence rate y ~ SW + RE + F + RE × F 688 518.4 Survival rate y ~ SW + RE 581 311.9 Survival—first year y ~ SW + RE + F + SP + RE × F + RE × SP + F × SP + RE × F × SP 606 96.1 0.23 0.20 Survival—second year y ~ SW + RE + F + RE × F 581 511.8 0.29 0.17 Survival—third year y ~ SW + RE 581 39.9 0.25 0.13 Shoot biomass y ~ SW + RE + SP + RE × SP 273 510.1 0.25 0.17 Fine root biomass y ~ SW + RE + F + RE × F 169 59.7 0.33 0.16 Quercus canariensis Emergence rate y ~ SW + RE + F 688 419.1 Survival rate y ~ RE 504 210.5 Survival—first year y ~ F + SP + F × SP 529 31.1 0.24 0.10 Survival—second year y ~ RE + F 504 38.2 0.19 0.11 Survival—third year y ~ RE 504 26.1 0.21 0.14 Shoot biomass y ~ SW + SP 287 310.3 0.22 0.15 Fine root biomass y ~ RE + SP 256 33.1 0.17 0.12 Note: Models are mixed-effects Cox’ proportional hazards models for emergence and survival rate, and GLMMs for the other variables. Model fit of Cox proportional hazards models was measured using root mean squared error (RMSE) and mean absolute error (MSA): Q. suber emergence: 0.49 and 0.49; Q. suber survival: 0.59 and 0.52; Q. canariensis emergence: 0.48 and 0.47; Q. canariensis survival: 0.58 and 0.55. TABLE 2 Best models analysing the effect of rainfall exclusion (RE), fungicide (F) and canopy species (SP) on emergence and survival rates along the study period, survival at different time points (first, second and third years), and final shoot and fine root biomass of experimental seedlings in the open woodland. Full models also included seed weight (SW) as a covariate and block (B) to control for the spatial structure of the experimental design. ∆AICc represents the differences between the AICc (Akaike information criterion corrected for small sample sizes) of the best and null models. The total number of parameters in the best model (‘Par’) is also given. Conditional R2 (Rc 2) reflects the variance explained by all factors (fixed and random) in the GLMMs, whereas marginal R2 (Rm 2) reflects the variances explained only by fixed effects. Seedling species Variable Best model NPar ∆AICc Rc 2Rm 2 Quercus suber Emergence rate y ~ SW + RE 688 333.6 Survival rate y ~ B + SW + SP 636 44.9 Survival—first year y ~ RE + SP + RE × SP 636 43.4 0.51 0.12 Survival—second year y ~ B + SW + RE + SP + RE × SP 636 612.2 0.44 0.24 Survival—third year y ~ B + SW + RE + F + SP 636 612.7 0.42 0.20 Shoot biomass y ~ SW + RE + F + SP + RE × F + RE × SP 239 77.1 0.31 0.22 Fine root biomass y ~ SW + RE + F + SP + RE × F + F × SP 146 711.8 0.17 0.16 Olea europaea Emergence rate y ~ F + SP + F × SP 11,668 512.5 Survival rate y ~ F 1015 25.1 Survival—first year y ~ SP + F 808 33.1 0.33 0.13 Survival—second year y ~ SP + F 636 312.3 0.44 0.23 Survival—third year y ~ F 1015 23.2 0.31 0.20 Shoot biomass y ~ B + SP 146 317.2 0.34 0.29 Fine root biomass y ~ B + SP 142 313.9 0.35 0.26 Note: Models are mixed-effects Cox proportional hazards models for emergence and survival rate, and GLMMs for the other variables. Model fit of Cox proportional hazards models was measured using root mean squared error (RMSE) and mean absolute error (MSA): Q. suber emergence: 0.50 and 0.50; Q. suber survival: 0.56 and 0.53; O. europaea emergence: 0.55 and 0.54; O. europaea survival: 0.46 and 0.43. 13652745, 2024, 1, Downloaded from https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2745.14227 by Universidad De Sevilla, Wiley Online Library on [16/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 181 HOMET et al. 3.3 | Seedling survival Survival at the end of the experiment varied from 56% in Q. canariensis to 20% in O. europaea, with Q. suber showing intermediate values (47% in the closed forest, 37% in the open woodland). Most mortality occurred after the first summer, particularly at the open woodland (Figure 5). Best models of survival included the effect of rainfall exclusion for the two Quercus species, but not for O. europaea (Tables 1 and 2). Effects of rainfall exclusion on survival of Q. canariensis were consistently positive along the study period, final survival being 20% higher under rainfall exclusion than in control plots (0.64 vs. 0.51, Figure 5b). Effects on Q. suber were, however, more complex and modulated by both the fungicide and the tree canopy treatment at the two sites (i.e. best models included Rainfall exclusion × Fungicide and Rainfall exclusion × Canopy species interactions; Tables 1 and 2). At the closed forest, rainfall exclusion effects on first- and secondyear survival of Q. suber varied from slightly negative with fungicide to largely positive without the fungicide (Figure 4b), particularly under conspecific canopies (Figure 6a). Larger benefits of drought without fungicide (i.e. with natural pathogen populations) than with fungicide remained along most of the study (Figure 5a). In the open woodland, rainfall exclusion effects on Q. suber survival were also modulated by the identity of the canopy tree (Table 2), being positive only under heterospecific O. europaea canopies (Figure 6c). The application of oomycete-specific fungicide had positive effects on survival in the four species-site combinations (Tables 1 and 2, Figure 5). Fungicide was included in 50% of all best survival models and increased final survival of the four species, particularly O. europaea (32% increase, Figure 5d). Positive fungicide effects were modulated by the identity of the canopy tree in the closed forest (Fungicide × Canopy species interactions, Table 1), being particularly large under Q. suber trees (Figure S5a,b). 3.4 | Shoot and root biomass Best biomass models of the two oak species included the effect of rainfall exclusion, whereas only best models of Q. suber biomass included the effect of fungicide (Tables 1 and 2). Biomass of O. europaea seedlings was not affected but any of the two experimental treatments (Table 2). As for survival, effects of rainfall exclusion on biomass were more complex for Q. suber than for Q. canariensis. Rainfall exclusion increased the fine root biomass of Q. canariensis seedlings by 10% (0.076 ± 0.001 g vs. 0.068 ± 0.001 g), but it had no effect on shoot biomass (Table 1). Rainfall exclusion effects on Q. suber biomass were strongly modulated by both the fungicide treatment and the canopy species (Tables 1 and 2). Rainfall exclusion effects on shoot and root biomass of Q. suber seedlings varied from mostly negative with fungicide to positive without fungicide at the two sites (Rainfall exclusion × Fungicide interaction; Figure 4d,g,h). Rainfall exclusion effects on Q. suber shoot biomass also varied from negative to positive among coexisting canopy species (Figure 6b,d). FIGURE 4 Effects of rainfall exclusion and fungicide application on emergence rate (a, e), first-year survival (b, f), shoot biomass (c, g) and fine root biomass (d, h) of Quercus suber seedlings at the two study sites. Values represent mean ± SE. Lines linking fungicide levels are shown only for variables for which best models included the interaction between rainfall exclusion and fungicide application (see Tables 1 and 2). (a) 0.70 0.75 0.80 0.85 0.90 0.95 1.00 FungicideNo fungicide Emergencerate (c) 0.10 0.20 0.30 0.40 0.50 0.60 0.70 Fungicide Shootbiomass (g) (d) 0.00 0.02 0.04 0.06 0.08 0.10 Fungicide Fine root biomass(g) Closed forest (e) 0.70 0.75 0.80 0.85 0.90 0.95 1.00 Fungicide Emergencerate (f) 0.5 0.6 0.7 0.8 0.9 1.0 Fungicide First-year survival (g) 0.10 0.20 0.30 0.40 0.50 0.60 0.70 Fungicide Shootbiomass (g) (h) 0.00 0.02 0.04 0.06 0.08 0.10 Fungicide Fine root biomass(g) Open woodland Control Rainfall exclusion No fungicide No fungicide No fungicide No fungicide No fungicide No fungicide (b) 0.5 0.6 0.7 0.8 0.9 1.0 FungicideNo fungicide First-year survival Fungicidetreatment 13652745, 2024, 1, Downloaded from https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2745.14227 by Universidad De Sevilla, Wiley Online Library on [16/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 182 | HOMET et al. 4 | DISCUSSION In the last decade, rainfall exclusion experiments have become a very popular method to experimentally address climate change impacts on plant communities under natural conditions (Hoover et al., 2018). Despite the many rainout shelters established throughout the world, the number of studies that have analysed their impact on regeneration dynamics is still very limited, and only a handful have explored the interacting effects of increasing drought with other global change drivers (Classen et al., 2010; Lloret et al., 2009). FIGURE 5 Survival curves of Quercus suber (a) and Quercus canariensis (b) in the closed forest, and of Quercus suber (c) and Olea europaea (d) in the open woodland. Survival is given separately for each of the four experimental treatment combinations (RE, rainfall exclusion; C, control; F, with fungicide; NF, without fungicide). Dashed lines separate the 3 years of study. FIGURE 6 Interactive effects of rainfall exclusion and canopy species on first-year survival and shoot biomass of Q. suber. suber seedlings in the closed forest (a, b) and the open woodland (c, d). Emergence and fine-root biomass are not shown because no interactions were found (see Tables 1 and 2). Values represent mean ± SE. (a) 0.7 0.8 0.9 1.0 Q. canariensis Q. suber Firstyearsurvival (b) 0.0 0.2 0.4 0.6 0.8 Q. canariensisQ. suber Shootbiomass (g) (c) 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Q. suber Firstyearsurvival (d) 0.0 0.2 0.4 0.6 0.8 O. europaea Q. suber Shootbiomass (g) O. europaea Closed forest Open woodland Canopy species Control Rainfall exclusion 13652745, 2024, 1, Downloaded from https://besjournals.onlinelibrary.wiley.com/doi/10.1111/1365-2745.14227 by Universidad De Sevilla, Wiley Online Library on [16/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License