Tree - Open Grassland Mosaics Drive the Herbaceous Structure and Diversity in Mediterranean Holm Oak Meadows

Background: Mediterranean holm oak meadows are semi ‐ natural savannah ‐ like agroecosystems that result from traditional silvo ‐ pastoral practices, which shaped these systems into a mosaic of trees and open grassland. However, traditional silvo-pastoral uses are declining with the implications that this may have on the herbaceous layer, a very biodiverse and valuable resource of these systems. Here, we aim at assessing the inuence of the tree – open grassland mosaic on the structure, diversity, and composition of the herbaceous layer. Specically, assessing the canopy effect (a) under representative Iberian canopy types, considering traditional Quercus species stands and Pinus pinea plantations at different locations; and (b) along seasonality. Results: The different components of the herbaceous layer performed differential responses to the presence/absence of tree canopies, as for instance shows the dominance of grasses under the canopy, while legumes and forbs were favoured in the open grassland. Also, there was a certain a reduction in the species richness in P. pinea dominated plots compared to plots dominated by Quercus species. There was a reduction of the aboveground biomass under the canopy at the more environmentally constrained location. Such canopy effects were generally more pronounced in spring that in autumn. Conclusion: It is highly advisable preserve the tree – open grassland mosaic and traditional Quercus species stands to maximize and preserve plant specic and functional diversity. The the optimum tree coverage might be dependent, not only on the primary ecosystem service (i. e. forage provision), but also on local conditions.


Background
Mediterranean holm oak meadows, also called dehesas in Spain and montados in Portugal, are seminatural savannahlike agroecosystems that result from silvopastoral practices, in which an herbaceous and an arboreal layer -mostly Quercus species -coexist (Ibañez et al. 2021). They are one of the largest agroforestry systems in Europe (Eichhorn et al. 2006), and are particularly abundant in the SouthWest of the Iberian Peninsula (Olea et al. 2005).
Mediterranean holm oak meadows have traditionally provided pasture and acorns for livestock, timber, and cork, and those traditional silvo-pastoral uses shaped holm oak meadows into a mosaic of trees and open grassland creating diverse ecological niches (Sankaran et al. 2004). In particular, holm oak meadows are extraordinarily rich in plant diversity (Marañon 1985; Moreno et al. 2016) and have been typi ed as habitat of community interest by the EU Habitats directive (6310 Dehesas with evergreen Quercus spp) for their singularity and potential to preserve biodiversity. However, those traditional silvo-pastoral uses are nowadays changing. Mediterranean holm oak meadows are suffering simultaneously processes of intensi cation and abandonment (Peco et al. 2000). Traditional grazing practices are replaced by intensive farming and plantations of fastgrowing trees, mostly Eucalyptus and Pinus species (Costa Pérez et al. 2006; Costa et al. 2011); while less productive meadows are abandoned, which results in shrub encroachment and loss of diversity (Peco et al. 2000). However, although some research has been done on the in uence of tree canopies on the herbaceous layer at the ecosystem scale, in this study our objective is to deepen such canopy effect under representative Iberian canopy types, comparing traditional Quercus species stands and Pinus pinea plantations; at different locations; and along seasonality. For that purpose, we considered insightfully all the herbaceous layer compartments (above and belowground biomass, and litter); speci c and functional diversity and composition; and we asked the following questions: (i) are the different components of the herbaceous layer structure, composition and  (Peel et al. 2007) with warm, dry summers, and mild winters. However, SM is slightly cooler and wetter than DN, with mean annual temperature in SM of 16.8 ºC and in DN of 18.1 ºC, and mean annual precipitation in SM of 648 mm and in DN of 543 mm.
SM soils have a texture between sandy clay loam and clay. DN soils are sandier than SM, with a sandy loam texture. Total soil nitrogen (N, 0-30 cm depth) is quite low in both locations, but lower in DN (0.06-0.20% N) than in SM (0.85% N Ibañez et al., 2021). Grasslands in both locations are dominated by herbaceous annual species, including grasses (i. e. Bromus hordeaceus), non-legume forbs (i. e. Erodium moschatum), and legume forbs (i. e. Trifolium subterraneum). Both locations are extensively grazed at similar stocking rates: SM grazed by cattle and Iberian pigs (0.36 LSU ha − 1 ), and DN grazed by cattle and goat (0.40 livestock units (LSU) ha − 1 ), both typical stocking rates in Mediterranean grazing systems.
Study plots were selected according to their tree composition, representing typical canopy types of Iberian holm oak meadows. One pure Quercus ilex stand, in the SM location (SM-ilex), and one pure Quercus suber stand in the DN location (DNsuber), both the most abundant stands in the Iberian context (Costa Pérez et al. 2006); one Q. ilex and Q. suber mixed stand (DNmixed), the next most abundant; and a pure P. pinea stand (DN-pinea. Figure 1.c), a common tree plantation replacing traditional Quercus canopies (Costa Pérez et al. 2006). Plot tree densities (trees ha − 1 ) are on the average for the region (Costa Pérez et al. 2006): 34 ± 1 in SM-ilex, 26 ± 1 in DNmixed, 26 ± 4 in DN-suber, and 48 ± 6 in DN-pinea. In the DNmixed plot, we discriminated between both Quercus species (Q. suber and Q. ilex) to establish sampling points. However, preliminary comparative analysis in the DNmixed plot on microenvironmental conditions and vegetation characteristics under the canopy of both Quercus species indicated no relevant differences. DNmixed plot results are then always presented, combining both tree species.
Field work was carried out in spring (05/04/2016 − 10/04/2016) and autumn (13/12/201617/12/2016), coinciding with the most productive moments of the system, to capture vegetation seasonal variability and canopy effects that may be season dependent. Study treatments were, therefore, established according to plot (SMilex, DN-mixed, DN-suber, and DNpinea), season (spring and autumn), and canopy (open grassland, OG, and under the canopy, UC). Sampling points of the UC treatment were always placed at 1 m distance from the selected tree trunk, and sampling points of the OG treatment were placed at a minimal distance of 3 m from the selected tree, clearly outside the canopy. Sampling points were systematically placed following the north orientation with respect to the tree trunks (Ibañez et al. 2021). For each treatment level, we selected 3-4 samples, totalling 73 sampling points.

Vegetation sampling
At each sampling point, we hammered a metal collar (diameter = 25 cm) that de ned the sampling area, and we harvested the herbaceous vegetation at ground level rooted within each metal collar. Thereafter in the laboratory, we separated aboveground biomass (AGB) from litter (dead plant material detached from the herbaceous vegetation and tree leaves on soil surface). Also, we separated the AGB into plant species. Species were then attributed to the given plant functional type (PFT): grasses, non-legume forbs (hereafter "forbs"), and legume forbs (hereafter "legumes"). Where S is the number of species in the community matrix, and P i P j the pairwise interactions between species, meaning the product of the relative abundance of the ith and jth species (Eq. 1). SR and evenness were also modelled as function of plot, season, and canopy, and the corresponding interactions using linear models. Final models were selected by a stepwise procedure based on the Akaike information criterion (AIC) using the stepAIC function, MASS package (Venables and Ripley 2002).
Also, we described the in uence of plot, season, and canopy on species composition by canonical correspondence analysis (CCA) to represent the community on a given number of dimensions (Sandau et al. 2014). We performed the CCA on species absolute abundance using the cca function of the vegan package (Oksanen et al. 2018). Signi cance of the CCA terms (plot, season, and canopy) and signi cance of the CCA axes were both tested using the anova.cca function, also of the vegan package. Afterwards, means and standard errors of the rst three signi cant (p < 0.001) axes were calculated and plotted for each level of the terms included as predictors (plot, season, and canopy). Also, we selected the species with the highest CCA scores within each axis (CCA1CCA3, top ve negative and top ve positive), and the four species closest to the axes (0, 0), for representation purposes, and to unravel speci c similarities and dissimilarities in the species composition between treatments.

General structure of the herbaceous layer
The structure of the herbaceous layer was dependent on the particular plot, season, and the presence/absence of the tree canopy (Fig. 2). The AGB was lower in all DN plots compared to the SM-ilex plot (plot effects, Table 1); AGB decreased in autumn compared to spring (season effect, Table 1); and AGB was lower under the canopy than in the open grassland in all DN plots (canopy effect, Table 1), but not in the SM-ilex plot (Fig. 2).
Litter was markedly higher under the canopy than in the open grassland in spring, but this difference between the canopy and the open grassland almost disappeared in autumn (season x canopy effect, Table 1). BGB was quite variable among treatments and neither season, nor plot or canopy explained its variability. Linear modelling on BGB is not shown. (season x canopy effect, Table 2), when the biomass of grasses equalized between both microenvironments (Fig. 3). Finally, legumes appeared mostly in spring and in the open grassland (season x canopy effect, Table 2).

Herbaceous species diversity and composition
Species evenness (Fig. 4.a) was only dependent on season, being lower in autumn than in spring (season effect, Table 3). Species richness (SR) decreased in the DN-pinea plot compared to the other plots (plot DNpinea effect), specially in autumn ( Fig. 4.b). SR was also lower in autumn than in spring (season effect,  (Fig. 4.b). showed that the rst four axes out of six were also signi cant (CCA1 -CCA3, p < 0.001, and CCA4, p = 0.04).
The second axis (CCA2, 24% of total explained variance) explained mainly differences in species composition between plots: on one side SM-ilex and DNsuber, and on the other side DN-mixed and DNpinea plots (Fig. 5.a).
SM-ilex and DN-suber being especially abundant in Lolium perenne (negative side of the CCA3), and DN-pinea being especially abundant in Stachys arvensis (positive side of the CCA2. Figure 6.b).
Finally, the third axis (CCA3, 15% of total explained variance), captured differences in species composition between under the canopy and the open grassland dependent on season (CCA3, Fig. 5.b). In the open grassland legumes as M. doliata and T. glomeratum (Fig. 6.c) were highly abundant in spring (negative side of the CCA3 and positive side of the CCA1, Fig. 6.c); while in autumn there was presence of a group of forbs, including Taraxacum o cinale, Leontodon longirrostris, and Crepis capillaris (negative side of the CCA3 and negative side of the CCA1, Fig. 6.c). Conversely, under the canopy and in spring Carex divulsa was specially abundant, and some other species as Urtica urens, L. cicera, Brassica barrelieri, G. aparine, were also present, but in low proportion (positive side of the CCA3, Fig. 6.c). Under the canopy and in autumn, Geranium molle was specially abundant, together with many other species in lower proportion.

The tree -open grassland mosaic as driver of the herbaceous layer composition
Tree canopies were important spatial drivers of the herbaceous layer, both in terms of structure (specially increasing the litter under the canopy, Fig. 2) and composition (Figs. 3-5). However, in line with our rst question, the different herbaceous components presented some differences in their response to tree canopies, which was especially notable on the PFT (Fig. 3) and species distribution (Figs. 5-6). Thus, the microenvironment created under the canopy favoured the dominance of some species, mainly grasses, while the open grassland favoured the presence of legume and nonlegume forbs (Fig. 3). Previous studies have also

The canopy effect under representative canopy types of Iberian holm oak meadows
In line with our second question, the canopy effect differed between Quercus species and P. pinea plantations. This was shown by the SR decrease in the plot dominated by P. pinea (DNpinea) in comparison to plots dominated by Quercus species (specilally in autumn, Table 3 and Fig. 4.b). Fact that may be related with the litter characteristics of P. pinea, which may be driving soil properties, and this in turn SR. The litter of P. pinea is known for its mulching capacity, and allelopathic properties, both factors lowering the understory growth (Valera-Burgos et al. 2012). Also, the litter of P. pinea has been reported to be poorer in N content than litter of Quercus species (Fioretto et al. 2008;Sheffer et al. 2015), which could be lowering soil N content and N availability. Indeed, N availability for plants in the DNpinea plot was reported the lowest among the study plots (Ibañez 2019). Factors that combined could be lowering SR in the DN-pinea, driving this difference in the SR between P. pinea and Quercus species dominated plots.
The different tree canopies also drove species composition, in addition to the variability associated to seasonality ( Fig. 5 and Fig. 6). The mosaic of trees drove a heterogeneous distribution of the herbaceous species, with some species being dependent on the speci c tree species and microclimatic conditions ( Fig. 5 and Fig. 6). This combined with the PFT distribution mediated by the presence/absence of tree canopies (discussed in Sect. 4.1), interestingly, indicates that although SR decreased under the canopy (Fig. 4.b), On the other hand, the canopy effect differed between the two locations (DN vs. SM), as showed the neutral canopy effect on the AGB in the SM-ilex plot in contrast with the AGB decrease observed under the canopy in all DN plots (Fig. 2). This suggests that the mechanisms unravelling the canopy effect on the AGB production might differ depending on local conditions, including environmental conditions and competition/facilitation professes.
A main driver for the AGB reduction observed in all DN plots under the canopy (especially in spring, Fig. 2 The DN location is drier than SM, with higher temperatures, lower precipitation, and sandier soils (Sect. 2.1); and competition for water resources could be here a limiting factor (combined with light availability and litter mulching the soil). Thus, water is especially scarce in DN during dry periods within the growing season, which are frequent under the irregular Mediterranean climate. Conversely, in the SM location, the reduced environmental constraints (slightly cooler and wetter than DN) may allow a certain compensation between the drivers that might be reducing the biomass production under the canopy (i. e. reduced light availability), and the drivers favouring productivity (i. e. increased soil fertility), which results on a neutral canopy effect on the AGB (Fig. 2).
Also, the presence of grazer animals may be driving the general structure of the herbaceous layer, and some of those differences in the canopy effect between locations (DN vs. SM). The stocking rate was very similar in both locations (Sect. 2.1), but the productivity in the DN location was lower than in SM (Fig. 2), and the livestock impact on the AGB might be higher in DN, wherein livestock visiting the under the canopy microenvironment, looking for shadow, acorns, and fresh herb, could be more frequent.
These differential canopy effect on the AGB production between locations (DN vs. SM), interestingly links with the results reported on greenhouse gas exchange (Ibañez et  the more environmentally constrained location of DN; while such canopy effect was neural at the cooler and fresher location of SM. Generally, such canopy effects were more pronounced in spring that in autumn. Overall, our results suggest that it is highly advisable preserve the tree -open grassland mosaic and traditional Quercus species stands to maximize and preserve plant speci c and functional diversity; and that the optimum tree coverage might be dependent, not only on the primary ecosystem service (i. e. forage provision), but also on local conditions. Facts that may be considered to manage and guarantee ecosystem services provision and conservation in these systems.

Declarations
Ethics approval Not applicable.

Consent for publication
Not applicable.

Availability of data and material
The datasets generated during the current study are available from the corresponding author on reasonable request.