The interaction of management practices with ecological processes re‑ stricts seedling establishment in different ways in each patch type. As Chapter 3 shows, the constraints to seedling establishment (namely, lim‑ ited seed sources, dispersal, germination, and establishment) were sig‑ nificantly more severe in shrublands than in oak forests and savannas, hence oak seedling recruitment in shrubland was impeded in multiple ways during the early phases of the tree recruitment cycle.
In Chapter 3 it was shown that forest patches had more cork oak natural regeneration and fewer limitations for oak recruitment. It was argued that the higher tree density in forests enables a more benign mi‑ croclimate with higher soil moisture and lower thermal stress, since oak requires shade for seed germination and seedling establishment (Calla‑ way, 1992; Pulido et al., 2001). The findings (Chapter 3) that dispersed seeds cached in forest patches germinated more easily, and that of all the sown acorns that germinated and grew into seedlings, the ones shaded by trees had the lowest mortality rates, imply that seedlings established in forest patches are more likely to survive severe summer droughts in future years. The lack of correlation between exposure and the transitions or persistence of forest patches also suggests that thermal stress is less im‑ portant for seedlings growing in such patch types (Chapter 2). The results presented in Chapter 5 indicate that seedling survival will be greater in forest patches on steep slopes rarely or never subjected to intensive and prolonged understorey use or to planting operations, and where wildfires have not occurred. But, as argued in Chapter 4, in the event of wildfires, the shrubby understorey (fuel load) of forest patches will result in intense wildfires, more fire damage, and consequently lower post ‑fire cork oak survival, especially on southern slopes. It was indeed found that the most important ecological variable triggering transitions from forest to shrub‑ lands was wildfire.
Patches with lower cork oak density result from former agro ‑silvopastoral systems whose understorey use has been abandoned, or whose under‑ storey is managed extensively. It was argued that because of the lower tree density the seedlings and saplings are exposed to higher irradiance and water stress than in forests. Indeed, established cork oak natural regeneration was found to be lower in savannas than in forest patches (Chapter 3). Savanna patch types were more persistent on aspects that did not face north, which, as argued in Chapter 2, implies that in drier sites the increased thermal stress on seedlings halts tree recruitment and forest recovery. The fuel load in savanna patches is usually much lower than in forests, particularly if the understorey is still managed. The lower fuel and larger space between the trees contributes to faster fires with lower intensity and a lower risk of crown fires, thus decreasing wildfire damage to seedlings and trees. Indeed, as reported in Chapter 2, wild‑ fire did not appear to be primarily responsible for transitions or persist‑ ence of savannas. Even though observed seedling recruitment was not strongly limited in the savanna patches (Chapter 3), from the data col‑ lected on survival over 2 years it seems likely that the mortality of seed‑ lings growing in full sun in savannas will increase in subsequent years due to summer drought.
Cistus shrubs usually invade cleared understorey patches, particu‑
larly on south ‑facing slopes exposed to long ‑term intensive land use and with low soil moisture and nutrient status (Gonçalves, 1991; Diniz, 1994; Latorre, 1996; Blanco et al., 1997; Seng and Deil, 1999). The results re‑ ported in Chapter 3 show that several mechanisms limit cork oak recruit‑ ment in shrubland patches. Low cork oak acorn production and high rates of weevil infestation limit the first phase of the tree recruitment cy‑ cle, followed by lower rates of buried acorns (scatter ‑hoarded by mice), thus decreasing the likelihood of acorn germination. The germination of acorns that are effectively dispersed in shrublands is limited by high rates of post ‑dispersal predation. The seedlings that do successfully over‑
come the ecological constraints are less likely to survive. The findings that seedling mortality was significantly higher in shrublands than in forests and savannas, and that seedlings shaded by shrubs had lower survival rates than seedlings in full sun indicate that seedling growth is more con‑ strained in Cistus shrublands than in open oak patches, and therefore competition between seedlings and Cistus is much stronger than facili‑ tation. Other studies have shown that competition between Cistus and seedlings prevails over facilitation, in contrast to the facilitative effect of other shrubs (particularly late ‑successional shrubs) in Mediterranean en‑ vironments (Gómez ‑Aparicio et al., 2004; Pulido and Díaz, 2005; Pérez‑ ‑Devesa et al., 2008). Competition is likely to favour the shrubs, since
Cistus spp. is physiologically well adapted to drought. It has also been
reported that Cistus shrubs have allelopathic effects, aggravated by high temperatures and long photoperiods, which inhibit seedling germination and survival (Chaves et al., 2001; Lobón et al., 2002), but no information is available on the quantification of allelopathic effects between oak seed‑ lings and Cistus shrubs.