Temporal and spatial patterns of seed dispersal of four shrubs in a Cistus-Erica shrubland from central Spain

Mediterranean shrublands are composed of species that have different regeneration strategies after fire and soil seed bank types. However, differences over the years in seed dispersal temporal and spatial patterns of the various plants composing a community have been little investigated. Here, we studied the temporal and spatial patterns of seed dispersal in four shrubs of an old (> 40 years) shrubland in central Spain. Three of them are seeders (Cistus ladanifer, Erica umbellata, and Salvia rosmarinus), and one is a resprouter (Erica arborea); the first two have persistent soil seed banks, and the latter two, transient. A 15 × 10 m plot was chosen and divided into a 0.5 × 0.5 m grid, where plant cover and density were measured. At 106 quadrats, seed traps were set and periodically (1–2 monthly) monitored for 3 years. S. rosmarinus dispersed in late spring-early summer, E. arborea dispersed during the summer, and C. ladanifer and E. umbellata dispersed from early summer to nearly late spring of the next year. Globally, seeds were being dispersed all year round. The seed crop size of a given species varied between years, although species differed in the year their seed crop was largest, despite large differences in climate. Seed rain and plant cover of each species were poorly related in terms of the variance explained by the models. Semivariogram analysis showed that seed dispersal expanded beyond that of the plant cover of each species by a few meters. No association between seed crop size and spatial dependence was ascertained. While species dispersal in space tended to be negatively related to one another, E. arborea seeds tended to dominate underneath the majority of the other species. S. rosmarinus dispersed before the fire season, which is consistent with seeds avoiding fire while on the plant; C. ladanifer and E. umbellata dispersed mostly after the fire season, which is coherent with a bet-hedging strategy against seed predators; E. arborea dispersed before the rainy season, which is expected for a plant that germinates readily after imbibition. Seed dispersal in time was compatible with the type of soil seed bank and post-fire regeneration of the species. The evidence of such a relationship with spatial patterns was weak. The dominance of E. arborea seeds underneath most of the other species suggests that this mid-successional species might dominate when openings form due to the deaths of standing plants of the seeders between two fires, given their lower longevity.

Conclusiones S. rosmarinus dispersó antes de la temporada de incendios, lo que es consistente con que las semillas eviten el fuego mientras estén en la planta; C. ladanifer y E. umbellata dispersaron principalmente después de la temporada de incendios, lo que es coherente con una estrategia de apuesta contra los depredadores de semillas; E. arborea dispersó antes de la temporada de lluvias, lo que es consistente con una especie que germina fácilmente tras la imbibición.La dispersión de semillas fue compatible con el tipo de banco de semillas del suelo y la estrategia de regeneración tras el fuego de la especie.La evidencia de esta relación con el patrón espacial fue débil.El predominio de las semillas de E. arborea debajo de las otras especies sugiere que esta especie mediosucesional dominará cuando se formen claros debido a la muerte de las plantas en pie de las semilladoras, que son menos longevas.

Background
Fire evolutionary pressures on seed dispersal temporal and spatial patterns in fire-prone ecosystems must have varied depending on the post-fire regeneration strategy of the plant.Pressures by fire to select particular temporal or spatial patterns must have been stronger on species whose persistence depends solely on a seed bank rather than on resprouting (e.g., seeders, S + , as opposed to resprouters, R + (Pausas et al. 2004)).Seed dispersal in fire-prone ecosystems, like those of Mediterranean-type areas, has been studied mostly in serotinous species, with a focus on how fire interacts with other biotic or abiotic factors to alter dispersal after fire (Lamont et al. 1991;Saracino et al. 1997;Parr et al. 2007;Broncano et al. 2008;Groom 2010;Dunker et al. 2019).Few studies, however, have addressed dispersal temporal and spatial patterns between fires in non-serotinous species whose persistence rests on soil seed banks.
Mediterranean seeder shrubs are dominated by Cistaceae, Lamiaceae, and Ericaceae, among others, and have small seeds that are dispersed alone or with their fruits or other floral structures by barochory (Paula et al. 2009).
Here, we will focus on four species that are dominant in the shrublands of central and western Spain.Three of them are seeders (R-S +) (C.ladanifer, S. rosmarinus, and E. umbellata), and the fourth is an obligate resprouter (R + S-) (E.arborea).Dispersal in C. ladanifer occurs from early summer (June) over the next 10 months or longer (Bastida and Talavera 2002).Similar patterns were found for C. monspeliensis, C. salviifolius, and C. creticus (Skourou and Arianoutsou 2004).These studies were carried out during one single year, so no information is available regarding the temporal variability of these patterns.Furthermore, seeds are predated by granivorous ants rather efficiently, decreasing seed density in the upper soil (Bastida and Talavera 2002;Bastida et al. 2009).S. rosmarinus has a short period of dispersal around late spring (Herrera 1986).Its seeds are prolifically predated by ants unless they are moistened and glued to the soil by a mucilage secreted by the seed (Engelbrecht and García-Fayos 2012).No study is available regarding temporal patterns of dispersal of both Erica species.
These four species differ also in their seed type, response to fire cues, and soil seed bank type: C. ladanifer has hard seeds, and its germination is promoted by a heat shock (Chamorro and Moreno 2019) and forms persistent soil seed banks that prolifically germinate after a fire (Quintana et al. 2004;Moreno et al. 2011).S. rosmarinus has soft seeds that imbibe readily when moistened, its germination can be enhanced by cold temperatures (Madeiras, Boyle et al. 2009), its viability in the soil is very low and forms transient seeds banks (Clemente et al. 2007).It germinates after a fire even if not in prolific numbers (Quintana et al. 2004;Moreno et al. 2011).E. umbellata seeds are soft but rather recalcitrant to germinate, its germination is either negatively affected by heat (Chamorro et al. 2018) or not (Vasques et al. 2012), but is positively affected by smoke (Moreira et al. 2010).Fire negatively affects seed density in the soil seed bank, particularly in the litter layer, but not so much in the deeper soil layers (Maia et al. 2016).Having seeds that remain in the deeper soil and are able to germinate is a sign of the capacity to form at least short-term persistent seed banks (Teketay and Granström 1995;Holmes and Newton 2004).The species germinates profusely after fires, even 2 or 3 years past the blaze (Quintana et al. 2004;Moreno et al. 2011;Céspedes et al. 2014).E. arborea germinates readily when moistened (Chamorro et al. 2017), heat either does not affect germination or reduces it (Mesléard and Lepart 1991;Valbuena and Vera 2002), and it does not germinate after a fire under Mediterranean conditions (Mesléard and Lepart 1991;Quintana et al. 2004;Moreno et al. 2011).We consider the soil seed bank of this species to be transient.

Framing dispersal in a context of fire and soil seed bank type
For species regenerating from a soil seed bank, the type of seed is important since it may condition the type of soil seed bank it forms.Some seeds germinate readily once imbibed, others have physiological dormancy and require chilling (Baskin and Baskin 1998), and others need to be exposed to fire cues (heat or smoke) (Staden et al. 2000;Moreira et al. 2010;Chamorro and Moreno 2019;Pausas and Lamont 2022).Recalcitrant seeds can persist for many years until germination is promoted by the appropriate cues, particularly those related to fire, and can form long-term persistent soil seed banks.In contrast, seeds that germinate readily or need just to be exposed to chilling may stay only for shorter periods in the soil and form transient soil seed banks (Pierce and Cowling 1991a, b;Holmes and Newton 2004;Clemente et al. 2007).
In species that form persistent soil seed banks, the density and spatial patterns of seedling emergence after a fire may reflect the long-term accumulation of seeds plus the input from the current year's seed crop (Keeley 1977;Kilian and Cowling 1992;Zammit and Zedler 1993).In contrast, species that form transient soil seed banks, emergence after a fire may mostly reflect inputs from the prefire crop (Musil 1991;Pierce and Cowling 1991a, b).It has been shown that fires before seed ripening and dispersal due to fires early in the season can affect population persistence and alter the plant community (Miller et al. 2019).Thus, for the current year's seed crop to be effective after a fire, dispersal must occur before the fire season, and this will be of utmost importance for species forming transient soil seed banks, not so much for those forming persistent soil seed banks (Le Maitre 1988;Harrington and Driver 1995;Wright and Clarke 2018).
In addition to fire, other stressors may drive seed dispersion patterns.Granivorous ants are the main seed predators in this region (Bastida and Talavera 2002;Bastida et al. 2009;Engelbrecht and García-Fayos 2012).Ants are active throughout the year until temperatures fall below 10 • C in autumn and winter (Retana and Cerdá 2000).Thus, seeder species with a persistent soil seed bank would be expected to delay dispersal well into autumn and winter, since there would be no benefit in dispersing when predators are most active, while their permanence after a fire is insured by a persistent seed bank.On the other hand, seeder species with transient soil seed banks would be expected to disperse before the fire season to maximally benefit from the current seed crop in case of fire.If chilling would be required, this would just delay the germination in time until late winter or early spring, which might compromise establishment (Luis et al. 2008;Moreno et al. 2011), but would not affect dispersal temporal patterns.Obligate resprouter species (R + S −) with transient seed soil seed banks would be expected to disperse before the wet season would start in autumn.Investing in dispersal before the fire season would be not selected since regeneration after a fire will rest only on resprouting.
Most fires are now lit by people and occur during the summer; natural, lightning fires occur during the summer, but with less temporal spread, with the largest area burned in recent decades centered in the month of July (Vázquez and Moreno 1998;Rodrigues, Jiménez-Ruano et al. 2023).Fires in these shrublands burn as highintensity crown fires that fully scorch the crowns.This type of fire has been shown to kill non-dispersed seeds (Harrington and Driver 1995;Schmidt et al. 2005).Thus, if fire had selected the timing of dispersal, this should occur before the onset of the lightning season, assuming that current lightning patterns have been prevalent under Mediterranean climate conditions.
Yearly seed crops fluctuate largely within and between species over the years (Herrera et al 1998) (Nathan and Muller-Landau 2000).This is partly determined by weather patterns during the year, with precipitation being an important driver in semiarid ecosystems like Mediterranean-type ones (Keeley 1977(Keeley , 1987)).Seed crops may differ among species owing to their sensitivity to the current weather and the timing of flowering and seed formation, among others.Therefore, seed crop size in a given year may vary among species of the same community.Furthermore, the size of the seed crop can affect seed dispersal spatial patterns (Gratzer, Pesendorfer et al. 2022).Therefore, total seed input and where seeds land may be subject to the vagaries of the weather during a given year.
Dispersal spatial patterns are important in view of a future fire and subsequent competition among species (Yeaton and Bond 1991).For instance, dispersing underneath the mother plant ensures spatial persistence, since these plants will be killed by fire, thus favoring the recolonization of space occupied prior to the fire since these sites were probably the most suitable ones.Yet, this would be more important for plants with persistent seed banks than for those with transient ones.In the latter case, greater exploration of the space at any year might be important to reduce competition with the many more seedlings that might emerge in areas occupied by species with persistent seed banks that accumulate seeds over the years.Additionally, seeders, or species regenerating by seeding in general, should avoid dispersing underneath resprouters, since resprouters will likely survive the fire (Parra and Moreno 2018) and might outcompete seedlings.
While seed dispersion spatial patterns can vary with the size of the seed crop (Gratzer, Pesendorfer et al. 2022), interannaul variability may depend on the soil seed bank of a species.This variability may be greater in species forming persistent seed banks than in those forming transient ones.The reason is that persistent soil seed banks would leverage in part the effects of a year with a low crop.In contrast, species with transient soil seed banks would be expected to vary less in their spatial relationships as a function of the yearly seed crop in order to maximize occupancy of open spaces after fire that will mostly rely on the last crop.Consequently, greater stability would be expected in the latter species than in the first ones.The pressure at any year over resprouting species to ensure its colonizing capacity would be minimal, since recruitment will occur over long periods of time, between fires.For these species, what is most important is to expand the colonizable space to occupy the empty spaces that will emerge due to the death of the seeders' standing plants owing to their lower longevity.Thus, pressures to maintain certain dispersion spatial patterns with seed crop size would be minimal.All of the above rest on the assumption that primary dispersal is critical for determining emergence after a fire and that no other post-dispersal mechanism is as important, although this may not apply to all species (Broncano et al. 2008;Warzecha and Thomas Parker 2014).

Objectives
Our goal was to study the seed temporal and spatial dispersal patterns over the course of 3 years in four Mediterranean species that dominate Mediterranean shrublands over acidic substrates.Until now, few studies have addressed dispersal at the level of the whole community in these shrublands.Additionally, we wanted to assess whether the patterns found would be compatible with a dispersal framework in which the type of soil seed bank could be important.Since the number of species in this community is limited, rigorous hypothesis testing is not possible given the variability in soil seed bank types and regeneration strategies.Nevertheless, addressing dispersal under such framework could provide insights as to what factors have selected dispersal patterns in these shrublands.

Methods
The study site was located at the Quintos de Mora Range Station, Montes de Toledo, Central Spain (39° 23′ N, 4° 00′ W).An old (> 40 years).A shrubland was chosen in a south-facing slope, at 700 m elevation.It was dominated by the shrubs Cistus ladanifer L. (Cistaceae), Salvia rosmarinus Spenn.(Labiatae) (former Rosmarinus officinalis L.), Erica umbellata Loefl.ex L., Erica arborea L. (Ericaceae), and Phillyrea angustifolia L. (Oleaceae).The first three species are seeders (S +), that is, they regenerate after a fire from seeds, while the last two are resprouters, that is, they regenerate after a fire by resprouting (R +).The soils are acidic, derived from quartzite; the mean annual temperature is 14.7 °C, and the mean annual precipitation is 689 mm (data from Los Cortijos meteorological observatory, period of observation 1971-2000, 10 km south from the area).We used data from this observatory plus the 12-month standardized precipitation evapotranspiration index (SPEI) (Vicente-Serrano et al. 2010) to characterize the climate and drought conditions, respectively, of the years of study.
In March 1997, a 15 × 10 m plot was chosen adjacent to an area used for experimental burning, and a 0.5 × 0.5 m grid established.Shrub density (individuals, m −2 ) was counted at each square of the grid and cover was measured for each species by the line-intercept method along the middle of each 0.5 × 0.5 m grid, with a total of 600 quadrats sampled.Plant cover was expressed as a percentage of the intersected species within each quadrat.In P. angustifolia, we considered as an individual all sprouts that had no clear and visible connection to any other.
Seed traps (106) were set in the middle of the 0.5 × 0.5 m quadrat following a systematic nested design.First, 40 plots were set at the upper left-hand-side corner of each 2 × 2 m.Then, 8 of these 2 × 2 m were randomly chosen, and 7 locations within the original 0.5 × 0.5 m were randomly selected for a total of 8 × 7 = 56 additional locations.This design was chosen to maximize sampling of the plot while having a range of distances between traps that would allow to rigorously apply geostatistical analysis (see later).Finally, 10 additional locations were randomly selected within the original grid.The seed traps consisted of a funnel (12 cm diameter) with a filter paper set onto a pot, with a hole to allow evacuating water, and fixed to the ground with a peg.Filters were sprayed with ant repellent, to avoid seed predation.Seeds were collected periodically, every 1-2 months, depending on the season, from April 1997 to April 2000 (3 years of observations).Bags were brought to the laboratory, cleaned, and seeds identified under a stereoscopic microscope.Seed counts in each trap were converted to seeds per square meter.In many instances, whole fruits fell into the traps; these were dislodged and seeds counted.In the case of S. rosmarinus, the fruit is dispersed as a unit, composed of four nuculae.We counted the nuculae as seeds.The fruits of P. angustifolia are drupes that are dispersed by birds, and few of them fell into our traps; hence, its seed production will not be treated further.

Temporal patterns
Differences in the heterogeneity of the seed crop within a species across years were tested by the coefficient of variation (Gratzer, Pesendorfer et al. 2022).Additionally, differences in the seed crop between years for each species were tested by GLM repeated measures ANOVA.When sphericity was not met (Mauchly's test significant at P ≥ 0.05), the Huynh-Feldt correction was used.Post hoc comparisons between years were corrected by Bonferroni for multiple comparisons.
The analysis of the temporal pattern of cumulative dispersal of each species was done considering the seed rain produced by the species during a 12-month period.The beginning and end of this period varied among species.The first month of the period was considered the one in which seed rain was null or minimum, since in some cases there were a few seeds being dispersed over time.The periods used were January to December (S.rosmarinus), June to May of the following year (C.ladanifer and E. umbellata), and April to March of the following year (E.arborea).We summed all the seeds dispersed during the corresponding period and calculated the percentage that was accumulated over time.We then fitted non-linear functions (power, exponential, and logistic, with 3, 4, and 5 parameters) to the cumulative seed rain over the selected period.The model selected based on Akaikes's Information Criterion (AIC) was: A1, A2, x o and P are parameters of the model.The relationships between seed rain among pair of species over the 3 years were studied by the non-parametric test of Spearman rank correlation coefficient (ρ), due to all four species being not normally distributed (Shapiro-Wilk test, P < 0.001).

Spatial patterns
The spatial relationship between standing species was investigated by calculating the Pearson correlation coefficient among pairs of species using cover (%) values from all 0.5 × 0.5 quadrats (n = 600) in the plots.Plant density in each quadrat was transformed to the presence/absence data, and then we assessed the association between pair of species using the Pearson χ 2 test.Statistical significance was set in both analyses at P ≤ 0.05.
For each species, we calculated omnidirectional, experimental semivariograms of annual and 3-year accumulated seed rain up to a distance of 10 m, roughly half the maximum possible distance in the study plot (Rossi et al. 1992).Empirical semivariograms were modeled with spherical models, unless no clear autocorrelation was evident, when a pure nugget model was used or when the spherical model fitted poorly.The maximum distance of autocorrelation (range, A 0 ) was extracted from the models and spatial dependence (SD, the variance explained by distance among samples) was calculated (Robertson and Gross 1994).To better ascertain where the seeds were landing over the years, 3-year accumulated seed rain maps were generated by ordinary kriging.
The relationships between the cover and total seed crop (n = 106 quadrats) of the respective species were studied by linear regression after Log 10 + 1 transformation of the seed crop variable in all species except S. rosmarinus, which could not be made normal.In this case, the equivalent relationship was explored by the non-parametric Spearman rank correlation test.
The relationship between total seed crop (n = 106 quadrats) among pairs of species was also studied by Spearman rank correlation.P-values of the correlation tests were corrected for multiple comparisons using the Benjamini-Hochberg test.
To determine the relationship between the size of the yearly seed crop and the range (A 0 , m) and spatial dependence (SD [%]) of the semivariograms obtained for each species and year, absolute values within each species were ranked from 1 to 3 from the largest to the smallest values.Lack of spatial dependence was ranked 3, since it was assumed that the range would be larger than what it was empirically obtained.Then, a contingency test was run for all four species together to assess whether seed crop was significantly associated to either range (A 0 ) or spatial dependence (SD %).Statistical analyses were carried out with SPSS vs. 28 (IBM Corp., Armonk, NY, USA).Modeling and graphing of cumulative rain over time were done with Origin-Pro 2019 (OriginLab Corp., Northampton, MA, USA).Geostatistical analyses were done with the gstat package (Pebesma 2004;Gräler et al. 2016) in R version 4.1.3(R Core Team 2022).

Results
The plot was dominated by C. ladanifer and P. angustifolia, followed by E. arborea and E. umbellata; the least abundant species was S. rosmarinus (Table 1).
The climate of the years of study was very different; the first two hydrological years (1996)(1997)(1998) were rather wet, with both temperatures and, particularly, precipitations well above average, while the third one (1998-1999) was rather dry, particularly in the vegetatively active months, in which precipitation was nearly half of the long-term average, with temperatures being also higher than usual.This is well reflected in the SPEI that showed the first two hydrological years well above the average drought conditions, the reverse being for the third (Fig. 1).
Seed crop was very variable over the 3 years, with E. arborea being the species that fluctuated the most (CV = 93%), followed by C. ladanifer (CV = 55.2%), while E. umbellata (CV = 44.0%)and, particularly, S. rosmarinus fluctuated the least (CV = 22.4%) (Fig. 2).In absolute terms, the seed crop was highest in both Erica, which, in some years, produced tens of thousands of seeds per square meter.C. ladanifer was also a prolific seed producer, with up to nearly thirty thousand seeds per square meter in some years.The species producing the least seeds was S. rosmarinus, not quite reaching one thousand seeds per square meter (Fig. 3).Averaged over the 3 years, E. arborea, despite the variability between years, was the largest seed producer, with C. ladanifer and E. umbellata producing about a third of such high number of seeds; S. rosmarinus produced the least of all, with two orders of magnitude less than E. arborea.These figures varied when correcting for the differences in cover and density.The relative rank persisted, but the differences were attenuated, particularly due to the larger cover of C. ladanifer and the smaller one of S. rosmarinus.In terms of seeds produced per individual, the values ranged from about 1000 in S. rosmarinus to 50 times higher in E. arborea, with the other species in between.When converting the figures to biomass, it is worth noting that the largest effort was done by C. ladanifer, followed by E. arborea, the two other species being at a distance, with S. rosmarinus being always the species that invested the least in seeds even after correcting for cover or density (Table 2).
Seed rain temporal dynamics were very different among species (Figs. 2 and 4).Seed dispersal in C. ladanifer started in early summer (June) and lasted nearly a whole year.During this time, accumulated seed rain was nearly constant over time, with a pattern that was almost linear during most of the months, there being some variability among years (Fig. 4A).The cumulative seed rain of S. rosmarinus had a clear S-pattern and was concentrated in spring and early summer (from April to June).There was very little variability in the cumulative temporal pattern of seed rain of this species (Fig. 4B).The cumulative seed rain temporal pattern of E. umbellata was similar to C. ladanifer, being nearly constant over most of the dispersing period, which started in early summer (late June/early July) and lasted nearly 10 months (Fig. 4C).This was the species with highest variability from year to year.The cumulative seed rain temporal pattern of E. arborea was also of an S-type, with little variability between years, and very much centered in the summer (Fig. 4D).The temporal patterns of seed rain of C. ladanifer and E. umbellata were positively correlated among each other, while both were negatively correlated with S. rosmarinus.The temporal pattern of seed rain of E. arborea was uncorrelated to any of the other three species (Table 3).Taken all four species together, seed rain was continuous over the years, there being no period without seeds falling onto the ground (Fig. 5).
The relationship among plant cover (%) of the five species indicated that C. ladanifer was positively correlated with S. rosmarinus and E. umbellata and negatively with E. arborea and P. angustifolia.S. rosmarinus was not correlated with any other species except as already indicated.E. umbellata was negatively correlated with both resprouters, while both were positively correlated to one another.The absolute values of the correlation were never very high (Table 4).The association between species rooted within each quadrat showed that there were fewer significant relationships except for the positive one between both Erica species and both resprouters (Table 3).
The geostatistical analysis of the seed rain that fell over the 3 years of study showed that the size of the patches of the seeds was larger than those of the plants, with ranges of 3.5 m in C. ladanifer and a strong spatial dependence, 2.9 m in E. umbellata, and 4.4 m in E. arborea, although with reduced spatial dependencies (55% and 36%, respectively).The largest patches of seeds were found for S. rosmarinus that showed a range (A 0 ) of 13.1 m and a moderate spatial dependence (66%) (Fig. 6, Additional file 1: Table S3).Seed rain maps obtained by ordinary kriging showed that seed rain over the plot of C. ladanifer was widespread, there being some clusters of seed accumulation and few areas devoid of seed rain.In contrast, S. rosmarinus had its seed rain concentrated in one large area, much of the rest of the plot having little seed input.E. umbellata had a pattern of dispersal much concentrated on one half of the plot, there being distinct patches of concentration of seeds, while few seeds were dispersed over much of the other half of the plot.  1) Fig. 3 Annual seed crop (mean ± SEM) in A Cistus ladanifer, B Salvia rosmarinus, C Erica umbellata, and D Erica arborea over the course of the 3 years studied in an old shrubland at Quintos de Mora Range Station (central Spain) (n = 106 seed traps).Note the different scales for each species as well as the different abundances (see Table 1).A year was considered from April (traps were set 1 month earlier) until March.Within a graph, significant differences among years are indicated with different letters (see text) E. arborea had seed input over nearly the whole plot, the only areas being devoid of them being the larger patches that had not plants on them (Fig. 7).
The variability in spatial patterning over the years changed with the species and size of the seed crop (Additional file 1: Fig. S2, Table S4): C. ladanifer was the    species that presented more consistent spatial patterns over the years, with ranges of 3.4 to 4.0 m and strong spatial dependence (> 79%).On the other side was S. rosmarinus, whose ranges were always much larger (9.7 to 11.7 m) and low spatial dependence (34%) at some years.E. umbellata had some years with ranges between 2.7 and 4.0 m, and variable spatial dependence, including a very strong one, to others in which there was not spatial dependence (second year investigated) despite being the year with the largest seed crop.Patterning in space of E. arborea also fluctuated quite a bit, with 1 year having a range similar to 3-year accumulated seed rain (4.0 m), and moderate spatial dependence (65%), and another one in which it produced a minimal seed crop without spatial structure (second year).
There was not a significant association between the yearly rank (1 to 3) of the year seed crop size and the ranks of either range (A 0 , m) or spatial dependence (SP [%]) as obtained in the relevant semivariograms for all four species analyzed together (Table 5).
Dispersal in space of C. ladanifer tended to be negatively correlated with both Erica species, particularly with E. arborea.Dispersal in space of S. rosmarinus was uncorrelated with the other three species.Dispersal in space of both Erica species was negatively correlated among each other (Table 6).
In general, for the three seeder species, there was a significant relationship between cover of the species and its total seed crop over the 3 years studied at the scale of the 0.5 × 0.5 quadrats (n = 106) (Table 7).Nonetheless, when significant relationships were detected, the correlation coefficients were not very large.No significant correlation was detected for E. arborea cover and its seed crop (Table 7).
Dispersal in either open spaces (no rooted plant within the 0.5 × 0.5 quadrat) or underneath particular species or combination of two of them was dominated by E. arborea, which produced that largest seed inputs in all areas but underneath E. umbellata, which was  7).If we consider only the seeders species, E. umbellata dominated in open spaces and underneath its own space and on spaces shared with C. ladanifer.In contrast, C. ladanifer dominated seed input in quadrats where it dominated and where S. rosmarinus or P. angustifolia did so (Table 8).

Discussion
Our results document that seed dispersion temporal and spatial patterns were very different among the species dominant in this shrubland.We found that C. ladanifer has a staggered dispersal, starting in summer and lasting for nearly a whole year.These results confirm earlier findings either with this species (Bastida and Talavera 2002) or with other species of the same genus (Skourou and Arianoutsou 2004).Such a temporal dispersion pattern suggests that it is driven by factors not related to fire.Indeed, dispersing during nearly a whole year, much of which took place in the cool months, suggests that evading predator pressure might have been behind the selection of this pattern, as suggested by (Bastida and Talavera 2002;Bastida et al. 2009).The fact that its seeds are hard and persist in the soil for years insures regeneration after fire.Under such a setup, the largest pressure on seed dispersal would be in ensuring the largest soil seed bank Table 5 Contingency test of the ranks (1 to 3, from more to less) of the yearly seed crop and those of the range (A 0 ) and spatial dependence (SD [%]) of the semivariograms obtained for each of the four species and the three years studied   Table 8 Mean seed crop (seeds m −2 ) per year that felt onto the traps located in 0.5 × 0.5 quadrats that either had no rooted individual (open space) or had individuals of one or two species.Note that we did not differentiate between one and more individuals per quadrat.No data was shown for the rest of combinations (13.4%) due to low numbers for any of them.Note that Phillyrea angustifolia did not produce seeds during the whole study period but was very abundant in the plot.The species that dominated the most underneath any species or where no plant was rooted is shown in bold possible, not on dispersing at a particular time of the year in anticipation of fire, since fires may not occur for years to decades.Thus, escaping predator pressure seems to be the most pressing factor for ensuring regeneration after fire.This response is compatible with that expected for a species with a persistent soil seed bank.Albeit more variable over the years, E. umbellata dispersed its seeds in a very similar way to C. ladanifer, the dispersal of both being highly correlated in time.Unlike C. ladanifer, its seeds are much smaller and soft, and the need for fire cues to germinate is less clear.While it is negatively affected by heat, one study reported a significant effect of smoke in promoting germination (Moreira et al. 2010).Fires of low severity tend to produce low emergence and cover over the years (Reyes and Casal 2008;Fernández and Vega 2014), in contrast to others of greater fire severity (Fernández and Vega 2016;Moreno et al. 2011).This species has generally low initial germinability, a trait that is often related to persistent soil seed banks (Holmes and Newton 2004).Their persistence in the soil seed bank has not been investigated, but it is probably long enough to be able to germinate even 3 years after a fire (Moreno et al. 2011;Céspedes et al. 2014).All of this is consistent with being at least shortterm persistent in the soil seed bank.Additionally, studies in vitro simulating various seasons have shown that south populations tend to germinate more readily than northern ones, which postpone their germination until past winter-type conditions, suggestive of physiological controls.Indeed, in post-fire environments, the species tends to germinate in spring (Quintana et al. 2004) but, depending on precipitation patterns, can even miss germinating during the first year after fire (Céspedes et al. 2014).Therefore, this species would gain little by dispersing prior to the fire season when it would be more exposed to its predators.Delaying dispersal until fall and winter would ensure that physiological dormancy might be broken, thus leaving the seeds ready to germinate when a next fire occurs due to the stimulatory effects of smoke.Thus, the strategy of escaping predators fits with this pattern, and is compatible with having a persistent soil seed bank.
S. rosmarinus showed a very different temporal pattern of seed release, since it dispersed mostly in late spring and early summer, a time that would be compatible with anticipating the fire season.This is consistent with being a seeder that has soft seeds that do not persist for long in the soil, and thus forms transient soil seed banks.Anticipating fires every year would allow this species to have their seeds in the soil and be more prone to escaping the blaze than if they had been in the canopies.Yet, survival in the canopy and soil during the passage of fire has not been investigated.Such early dispersal before the autumn rainy season would likely entail a high predation risk by ants, which are very efficient at doing so unless the seeds are moistened, in which case a mucilage is secreted by the seed that allows it to attach to the soil particles and largely reduces the risk of predation by these animals (Engelbrecht and García-Fayos 2012).Seed predation is also high in the plant, since whole calyx sections are cut and transported by the ants, which further adds to the pressure to disperse them (Engelbrecht and García-Fayos 2012).Late spring or early summer rains might reduce predator pressure on this species while providing the opportunity to remain in the soil until the rainy season.
The resprouter E. arborea had a pattern of dispersal similar to S. rosmarinus, although more delayed towards late in the summer and early fall.This would allow the seeds to be dispersed before the autumn rainy season when they could germinate readily once moistened.By shifting the dispersal timing more towards the middle of the summer, it reduces predation risk.Still, the period of risk of seed predation by ants was high, but no study has investigated this yet.
The fact that dispersal of C. Ladanifer and E. umbellata occurred much along the year and, to some extent, the separation in time between S. rosmarinus and E. arborea led to a system in which there was a continuous seed input into the soil, a finding that contrasts with the marked seasonality of this habitat.What role does such continuous input of seeds plays on their predators is unknown but deserves further investigation.
Our results also document that the factors behind producing smaller or larger seed crops were idiosyncratic for each species.Of the four species investigated, the largest seed crop was obtained in a different year for three of them (C.ladanifer, E. umbellata, and E. arborea), with one (S.rosmarinus) virtually not varying over the years.This is surprising given the large variability in the climate of each of the years investigated.
Provided that the last seed crop may contribute to the regeneration of the population after fire, more so for the species whose seeds do not persist for long in the soil seed bank, this can have variable implications for each of the species.Since S. rosmarinus disperses mostly before the fire season, the current crop could affect the postfire population emerging during a fire that occurred in that season.However, since C. ladanifer and E. umbellata disperse most of their seeds after the fire season, the effect of the current year's seed crop after a fire will be smaller, assuming that seeds stored in the canopy will be killed.Instead, it will be the seed crops of previous years the ones that would play a role in the subsequent recruitment after a fire.This can have important implications because a fire in a dry year, in which the seed crop might have been compromised (but see the high resilience of C. ladanifer to drought) (Chamorro et al. 2016), might benefit from the seed crop of the previous year that may or may not have been subject to negative climate pressures.This uncoupling between the current year's seed production and the population emerging after a fire can be advantageous to overcome the limitations imposed by the vagaries of climate, an issue that had not been noted until now and should be considered in how changes in fire seasonality can affect the plant community (Miller et al. 2019).
We documented that seed input into the soil varied very greatly among species and reached thousands of seeds per square meter.These figures for one single year are much larger than the number of seedlings that emerge after fire in the case of the three seeders.The maximum/minimum values of emergences after fire observed at adjacent stands or in similar old stands in this area are 274/61 (No. m −2 ) for C. ladanifer, 20/6 for R. officinalis, and 156/13 for E. umbellata (Quintana et al. 2004;Moreno et al. 2011;Parra and Moreno 2018) (Maia et al. 2016).These results confirm similar findings from other Mediterranean-type shrublands that suggest that large losses of seeds occur after dispersal and that emergence after fire may in part reflect seed inputs from the current year (Keeley 1977;Zammit and Zedler 1988;Pierce and Cowling 1991a, b;Meney et al. 1994).In contrast, annual seed input in S. rosmarinus was half of what has been reported at an old stand with a large seed bank [3061 No. m −2 ] ( Clemente et al. 2007).It must be noted, however, the large rate (55%) of non-viable seeds in this species, and the reduced viability (< 3%) when buried for 1 year, in contrast to species of Cistus most of whose seeds remained viable (> 70%) after 3 years burial (Clemente et al. 2007).
We found that dispersion in space was dominated by E. arborea, which ruled in squares that had on them any of the other species or combinations of them except in the case of E. umbellata.In this case, E. umbellata dominated underneath itself.Being E. arborea of much larger size than E. umbellata and given the positive relationship between the density of standing individuals of both species hint at a displacement of the second by the first.Moreover, E. arborea does not germinate after a fire, but between fires.The life cycle of the seeder species is in the range of a maximum of 15-30 years, which means that, even if few germination events occur, the likelihood that one E. arborea will occupy the open spaces is high, given its overall seed dominance in space.This suggests that this shrubland will not be stable over the years and that, little by little, E. arborea will be taking over the stand (Santana et al. 2011).While we did not collect sufficient seeds of P. angustifolia, the fact that this species was abundant in terms of spatial occupancy suggests that birds (Herrera 1995) are dispersing them from adjacent areas.The positive association with standing plants of E. arborea and not with the rest of the species suggests that birds are perching on this species.P. angustifolia is also a mid-successional species and suggests that the old stand is well on its way to becoming dominated by these two species.In case of fire, both species resprout rather well, with few losses even under severe drought conditions (Parra and Moreno 2018).Thus, even though abundant germination might occur from the other three species, a fire will deter for some time the successional vector of change that tends to make these two resprouters dominant.
Our results document that although standing plants were well-structured in space (i.e., space explained large portions of the spatial variability of seed dispersion), with very small patches, the patches formed by the total seed input during the 3 years were much larger, indicating for each species the rain shadow that dispersal can produce, which amounted to a very few meters.S. rosmarinus was different from the other three species: the patches produced by the standing plants were the largest, and those produced by the fallen seeds were even larger, which is surprising given that it was the species that produced the least seeds.Inspection of the seed dispersion map suggests that these larger patches might correspond to conglomerates of single patches of plants, rather than specific patches of the rain shadow produced by the individual plants.This patterning was consistent over the years which lends support to the contention that S. rosmarinus would be the one that would change the least over the years to ensure the greatest space in case a fire would occur.Notwithstanding, C. ladanifer presented also patches that changed little over the years, unlike what we initially expected.In contrast, the two Erica species, which were also large seed producers, had spatial patterns that changed very much and presented similar patterns among the two over the years despite the large differences in seed crop.Overall, we did not find support for a relationship between seed crop size and spatial patterning at the level of the whole community.The fact that we studied only 3 years necessarily limits the generalizations that can be made from this study, and more years (Gratzer, Pesendorfer et al. 2022) are needed to establish the relationships here depicted in Mediterranean shrublands for the first time.

Conclusions
Seed dispersal temporal and spatial patterns differed among the three species of seeders investigated.C. ladanifer and E. umbellata dispersed during long periods since early summer into the winter and next spring, indicative of a strategy that might have been selected to reduce the pressure of seed predators.In contrast, S. rosmarinus dispersed during a short period before the summer, consistent with the idea of completing dispersal before the summer season.The resprouter E. arborea dispersed during the summer and fall before the rainy season.These temporal patterns were compatible with the type of soil seed bank of these species.Nevertheless, the validity of these findings must be taken with care, given the limited sample of species that were investigated.Dispersion patterns in space produced seed rain shadows that exceeded a few meters of the plant cover.Spatial patterns of seed dispersion varied with seed crop size.We found support for a conserved spatial pattern over the years for S. rosmarinus, independent of seed crop size, but we did not find support for an overall relationship between spatial patterning and seed crop size for this community.Thus, the evidence for the role of soil seed bank type on spatial patterns was weak.The dominance of E. arborea in the seeds falling where the other species were rooted suggests that, as seeder plants die in the period between two consecutive fires, this species might take over the stand, together with P. angustifolia, both of them being mid-successional resprouters.

Fig. 1 Fig. 2
Fig. 1 Maximum and minimum temperature (T max , T min ) and precipitation (PRE) from Los Cortijos meteorological observatory (top) and 12 months Standardized Precipitation-Evapotranspiration Index (SPEI)(bottom) for the 1° × 1° cell of the study area (bottom) during the years of study of seed dispersal of four shrubs of a Cistus-Erica shrubland at Quintos de Mora Range Station (central Spain)

Fig. 4
Fig. 4 Cumulative seed rain of A Cistus ladanifer, B Salvia rosmarinus, C Erica umbellata, and D Erica arborea during the 3 years studied.Seed dispersal starts in January for S. rosmarinus, in April for E. arborea, and in June for C. ladanifer and E. umbellata.Modeling was done by fitting a logistic function (y = A2 + (A1 − A2)/(1 + (x/x 0 ) P ), where A1, A2, x 0 , and P are parameters of the model selected based on the minimum AIC

Fig. 6
Fig. 6 Semivariograms (semivariance as a function of distance) fitted to the total number of seeds fallen at 0.5 × 0.5 m quadrats (n = 106) during the 3 years of study in a Cistus-Erica shrubland at Quintos de Mora Range Station (central Spain).Top left: Cistus ladanifer; top right: Salvia rosmarinus; bottom left: Erica umbellata; bottom right: Erica arborea

Fig. 7
Fig. 7 Plant cover (left) and 3-year accumulated seed rain (right) maps at each 0.5 × 0.5 m over the whole plot (15 × 10 m) obtained by ordinary kriging in a Cistus-Erica shrubland at Quintos de Mora Range Station (central Spain).Top to bottom: Cistus ladanifer, Salvia rosmarinus, Erica umbellata, and Erica arborea

Table 1
Cover, density (mean, SE), and frequency (%) of the main species in the study Cistus-Erica shrubland at Quintos de

Table 2
Average seed crop per year (mean ± SE) or after correcting it for the actual cover (assuming 100% cover for each species [100% Co]) or density (Per Indiv.) of each species based on data from Table1.Similar results but expressed as biomass of the seed crop, after multiplying seed crop by seed biomass.Seed biomass data were provided by Dr. D. Chamorro

Table 3
Spearman rank correlation coefficient (ρ) among species for the temporal seed rain (No. m −2 ) at each sampling over the 3 years (n = 29).Significant correlations (α ≤ 0.05) after correcting P-values for the Benjamini-Hochberg test for multiple comparisons are marked in bold Fig. 5 Total seed rain (C.ladanifer, S. rosmarinus, E. umbellata, and E. arborea) (log seeds m. −2 ) during the 3 years studied at an old shrubland in Quintos de Mora Range Station (Toledo, Spain)

Table 4
Pearson correlation coefficient (r) and its statistical significance (P) for plant cover (upper-right panel) and Pearson χ 2 test (d.f. 1) for the presence/absence data of rooted plants (lower-left panel) in 0.5 × 0.5 quadrats (n = 600) between all five dominant species in a Cistus-Erica shrubland at Quintos de Mora (central Spain).Statistical significance (P ≤ 0.05) is marked in bold

Table 6
Spearman rank correlation coefficient (ρ) among species for accumulated seed rain (No. m −2 ) during the 3 years of study at each sampling position (n = 106).Significant correlations (α ≤ 0.05) after correcting P-values for the Benjamini-Hochberg test for multiple comparisons are marked in bold