Leaf litter hides post-dispersed seeds from invasive alien removers in a sclerophyll forest in central Chile

Aim of study: Seed removal may be a determining filter for regeneration. Factors such as leaf litter, tree cover and seed density affect seed removal. This aims to assess three aspects involving on post dispersal seed removal of four woody species ( Lithraea caustica, Maytenus boaria, Quillaja saponaria and Retanilla ephedra ) of a sclerophyll forest. Area of study: Andean Mediterranean sclerophyll forest of Quillaja saponaria and Lithraea caustica located in the Río de Los Cipreses National Reserve, Chile. Material and methods: Seven experimental plots were set up, in which seeds were offered in an experiment with variations in leaf litter, tree cover and seed density. Generalized Linear Mixed Models (GLMM) were used for the statistical analysis. Camera traps were installed to identify the seed removers. Main results: Q. saponaria seeds were more removed than the seeds of the other species (0.84, observed mean). Whereas that M. boaria and R. ephedra did not show significant differences on the removal of their seeds (0.77 and 0.67 respectively), both were more removed than L. caustica (0.59). The removal of M. boaria , Q. saponaria and R. ephedra seeds was lower in the presence of leaf litter. No factor influenced the removal of L. caustica seeds. Seed removers were identified as invasive alien species such as Rattus sp . and Oryctolagus cuniculus and native species such as Lycalopex culpaeus. Research highlights: Three woody species of this forest suffer severe seed removal by invasive alien fauna with major implications for the resilience of these forests. The coverage of leaf litter is key to hide away the seeds, increasing survival and could promote germination.


Introduction
The Central Chile forest is part of the five regions in the world with Mediterranean ecosystems. These regions hold almost 20% of the world's flora (Cowling et al., 1996), have a high level of endemism (Naveh & Wittaker, 1979;Arroyo & Uslar, 1993;Cowling et al., 1996) and cover less than 5% of the world's surface (Cowling et al., 1996). Also, these regions are recognized as biodiversity hotspots due those important biodiversity characteristic and the large number of threatened species by anthropogenic impacts (Myers et al., 2000).
The forest in Central Chile is considered the area that has received the most severe anthropogenic disturbances in South America (Salazar et al., 2015) and in the country (Rundel, 1998), causing deep transformations to the landscape (Elizalde, 1970;Schulz et al., 2011) and putting this region's remaining biodiversity at severe and increasing risk (Hobbs et al., 1995;Armesto et al., 1998). In the past, the forest was dense and biologically diverse (Aronson et al., 1998). Thus, today the forest is heavily fragmented and it has been suggested it has problems with its regeneration process, with little recuperative capacity (Jiménez & Armesto, 1992;Schulz et al., 2010).
Inside of plant communities, species have different requirements and responses at any stage in the regeneration process (Grubb, 1977). For the case of some species in a sclerophyll forest, one of the main forest types in Central Chile (Armesto et al., 2007;Salazar et al., 2015), it has been suggested that the transition from seed to seedling could be one of the key processes on the recruitment of tree species (Simonetti, 1983a;Bustamante, 1995). Specifically, seed removal may be a crucial filter for the recruitment of some species if these seeds are predated (Bustamante, 1995), dispersed to sites without the environmental conditions for their germination and establishment (Del Pozo et al., 1989;Bustamante et al., 1992;León-Lobos & Arroyo, 1994) or if these do not resist the mastication and passage through the animal's digestive tract (Castro et al., 1994;Castro et al., 2008).
However, it has been suggested that factors like leaf litter, seed density and tree cover may affect the removal of these dispersed seeds and therefore species recruitment (Forget, 1992;Forget, 1993;Myster & Pickett, 1993;Bustamante & Vásquez, 1995;Cintra, 1997;Dalling, 2002, Yu et al., 2015Dudenhöffer et al., 2016). For example, Janzen (1970) and Connell (1971) suggested that density dependent seed removers prefers areas with high seed densities becoming this areas in less safe to the species regeneration. Cintra (1997) suggested that the presence or not of leaf litter can influence the time and energy they invest in the search for seeds due to the difficulty to detect the seeds that are not visible (Cintra, 1997). Finally, Díaz et al. (1999) reported that areas with tree cover are safe areas for seed removers due the less visibility for to be predated. Nevertheless, how seed density, the presence of leaf litter and tree cover influences postdispersal seed removal in a sclerophyll forest is scarcely understood (Bustamante & Vásquez, 1995).
On the other hand, it is highly known the existence of invasive alien species inhabiting the sclerophyll forest and different effects we can found in the literature about seed removal by invasive alien species. For example, Castro et al. (2008) suggested that Oryctolagus cuniculus are dispersing seeds of Lithraea caustica. On the contrary, Barceló & Simonetti (2020) have suggested that Rattus rattus are predating seeds of Gomortega keule, endangered tree species of Central Chile. Thus, researches highlighting the potential negative effects of seed removal by invasive alien species are also scarce.
The objective of this study were to assess seed removal after dispersal from four woody species of a sclerophyll forest, to assess the influence of the factors leaf litter, seed density and tree cover and to identify seed removers with the motivation to know if these ecological interactions are contributing to the recruitment problems on the main species of this forest. To do so, the following questions were asked: How many seeds can be removed after dispersal? Could seed removal be an important filter for the recruitment of these species? How do the study factors affect seed removal? Which fauna are predator-dispersers of these seeds?

Study area
The experiment was conducted in the Río de Los Cipreses National Reserve ( Fig. 1; 34°15'56.55"S; 70°27'56.97"W). The reserve is located in the Andes mountains at a height that ranges from 900 to 3500 m.a.s.l. There are different plant associations at this altitudinal gradient, with the sclerophyll forest being predominant. Generally, the soils at the reserve are lithosol and to a lesser extent alluvial and colluvial soils. The climate is Mediterranean with an annual average precipitation of 1200 mm, which occurs particularly in winter. The area has a dry period of between 3 and 4 months. The annual average temperature is 5.9 ºC (Uribe et al., 2012). These trees and shrub (R. ephedra) are typical species of the Andean Mediterranean sclerophyll forest dominated by Q. saponaria and L. caustica (Luebert & Pliscoff, 2017). Approximately 6,500 seeds were collected, stored for their conservation in dry, cool containers until being offered to the fauna.

Seed species features
As we can see, in general the seeds are small and light (Table 1). Seed as L. caustica and R. ephedra are heavier than the others and its dispersal mechanism are barochory-endozoochory and barochory respectively. Whereas Lycalopex culpaeus and O. cuniculus have been described as dispersers for L. caustica (Medan & Aagesen 1995;Catro et al., 2008;Morales-Paredes et al., 2015), M. boaria seeds are dispersed by Colorhamphus parvirostris, by Xolmis pyrope and by Anairetes parulus (Reid & Armesto, 2011). The lightest seed of this study is a winged seed of Q. saponaria and its dispersal mechanism is meteoranemochory ( Table 1).
After exploring and checking the area of occurrence of the sclerophyll forest in the Río de Los Cipreses National Reserve, seven experimental plots were set up with similar slope (between 3 and 8%), altitude (1100 to 1200 m.a.s.l.) and contiguous areas with and without tree cover (Fig. 1).

Experimental desing
On each experimental plot 224 seed removal units (SRUs) were installed after dispersal with their respective treatments (Fig. 1). The SRUs followed recommendations described by Herrera et al. (1994), Alcántara et al. (2000), García et al. (2005) and Jaña (2012). The SRUs were transparent and odorless plastic containers, 13 cm in diameter and five cm deep (102.1 cm 3 ). These dimensions were due to the small size of the seeds of the species being studied (< 12 mm). The SRUs were fixed to the ground with 3 cm nails. Holes were made in the lower part of SRUs, which were filled with 3 cm of stones and earth from the sector, leaving the seeds 2 cm from the upper edge of the container. Holes facilitate water infiltration in the event of possible precipitation, and the edge above the seeds avoids their loss due to overspill or wind. Seeds of M. boaria, L. caustica and R. ephedra were marked with a gray permanent and odorless marker (ZIG marker brand; model PS-220/5V) for recognition in case other seeds were accidentally added to the SRU. The seeds of Q. saponaria (winged seeds) were stuck with odorless, colorless glue on a mesh placed over the filling to avoid loss by wind.
SRUs were arranged in a 2 x 2 x 2 completely randomized factorial experiment, with seven replicates (experimental plots) per treatment for each species (Fig. 1). Treatments consisted of combination of two levels of tree cover (SRUs without tree cover and with tree cover), two levels of leaf litter cover (SRUs without leaf litter and with seeds covered by leaf litter) and two levels of seed density (low density with five seeds in each SRUs  and high density with 50 seeds). Thus, each experimental plot consisted on two 4x4 grids (the first one without tree cover and the second one with tree cover) where we placed each SRU with leaf litter treatments and seed density treatments (Fig. 1). The monitoring took place from April to September 2015. The SRUs were revisited on five separate occasions (approximately once a month). The monitoring involved making a count of the seeds present in the SRUs, without replenishment, to determine the number of seeds removed. During each of the visits, the number of seeds removed was calculated as the difference between the initial quantity of seeds placed on each removal unit (50 seeds or 5 seeds, according to the seed density).

Identification of seed removers after dispersal
The seed removers after dispersal were identified by installing four Bushnell camera traps with motion sensors (Trophy Cam HD). The cameras were installed in conditions with and without tree cover in two sectors of the study area in order to visualize possible predator differences between the two covers. Four SRUs were placed in front of the camera traps, offering seeds from each species being studied. The four camera traps were programmed so that they remained active all day, taking photographs at three-second intervals whenever they perceived movement. The camera traps were installed in June and worked until September. Later, the photographs were reviewed and selected those that showed animals taking seeds from the removal units. Descriptive statistics were used to analyze both the predator sightings as well as the species whose seed were removed. The rodents in the photographs were recognized at genus level. Other mammals were recognized at species level.

Statistical analysis
We conducted two analyses to detect significant differences (p>0.05) among species and factors on seeds removed. We used the complete data set to fitted two Ge-neralized Linear Mixed Model (GLMM) with a binomial distribution given the nature of the response variable (seeds removed or not; "rem" and "1-rem", Formula 1 and 2). Due our nested experimental design, date and plot were set as random effects to include plot specific effects and to take into account for temporal pseudoreplication (Model 1, Eq. 1 and Model 2, Eq. 2). With the "lme4" package for GLMM in R Software (Bates et al., 2015;R Core Team, 2019), the first analysis was set with species ("spp") as a factor plus random effects factors (Model 1, Eq. 1).
( rem , 1 -rem ) ~ spp + ( 1 | date ) + ( 1 | plot ) Eq. 1 The second analysis was set with the seeds removed per each specie ("rem") resulted at the end as the response variable. We set as factors the tree cover (TC), leaf litter cover (LL) density (D) and its interactions plus random effects factors (Model 2, Eq. 2). Eq. 2 We applied on both models the analysis of deviance for GLMM to detect significant differences with "car" package in R Software (Fox & Weisberg, 2019; R Core Team, 2019), "lsmeans" package for post hoc analysis in R Software (Lenth, 2016; R Core Team, 2019) and "multcomp" packages in R Software for pairwise comparisons of least-squares means (Westfall et al., 1999;R Core Team, 2019).

Identification of seed removers
Thirty photographs were obtained from the camera traps, where the main seed removers were identified: individuals of O. cuniculus (Fig. 2a), rodents of the genus Rattus sp. (Fig. 2b) and individuals of L. culpaeus (Fig. 2c). On site, seeds removed by O. cuniculus and Seed removal in sclerophyll forests in central Chile individuals of Rattus sp. was confirmed, as remains of seeds nibbled by these individuals were found in the removal units.
73 % of the photographs that were taken at sites with tree cover showed rodents of the genus Rattus removing the seeds offered, 18 % showed individuals of O. cuniculus and 9 % L. culpaeus (Table 2). In sites without tree cover, 69 % of the photographs showed individuals of O. cuniculus removing the seeds offered and 31 % rodents of the genus Rattus (Table 2).
The SRUs that contained M. boaria seeds were preferred by individuals of O. cuniculus (Table 3) and those that contained R. ephedra seeds by rodents of the genus Rattus. L. culpaeus was detected removing only L. caustica seeds. The least visited SRUs by both O. cuniculus and rodents of the genus Rattus were those that contained Q. saponaria seeds (Table 3).

Post-dispersed seeds removed of each species
In general, seeds of the four species were located and removed from SRUs since the first monitoring (Fig. 3a). The greatest proportion of seeds removed in the first monitoring was, on average, from R. ephedra (0.52 of seeds removed). The seeds removed for each specie increased gradually towards the end of the monitoring (day 185, Fig.  3a). In general, our first analysis showed significant differences among species on post-dispersed seeds removed (p < 0.001; see Table S1 [suppl.] for the GLMM summary). After, the comparisons of means showed that the removal proportion of Q. saponaria seeds (0.84 ± 0.04, observed mean ± standard error respectively) was significantly greater than the seeds removed proportion of M. boaria (0.77 ± 0.04, p < 0.002), R. ephedra (0.67 ± 0.05, p < 0.030) and L. caustica (0.59 ± 0.04, p < 0.001, Fig. 3b). In contrast, the removal proportion of R. ephedra seeds did not show significant differences with M. boaria (p > 0.830) and it was significantly greater than the removal proportion of L. caustica seeds (p < 0.010). Also, the removal proportion of M. boaria was significantly higher than these for L. caustica (p < 0.006). Finally, the seeds removed proportion of L. caustica was statistically lower than the seeds removed proportion of the other three species (Fig. 3b).

Factors influencing seeds removed on each species
The analysis of deviance applied to the Model 2 showed that seeds removed was significantly higher without leaf litter than those seeds covered with leaf litter for M. boaria (p < 0.019), Q. saponaria (p < 0.005) and R. ephedra (p < 0.009) ( Table 4; See Table S2 [suppl.] for GLMM summary of each specie). There were no influences of tree cover, seed density and factors interactions on seeds removed (Table 4).
Leaf litter cover showed an influence on seeds removed since the first monitoring, except for L. caustica seeds (Fig. 4a). The proportion of seeds removed without leaf litter cover was significantly higher than seeds removed covered by leaf litter for M. boaria (Fig. 4b), Q. saponaria (Fig. 4c) and R. ephedra (Fig. 4d).

Discussion
Our study reports on the relevance of the removal of seeds dispersed of a sclerophyll forest in Central Chile. It was documented that the invasive alien fauna are removing the seeds and that the absence of leaf litter cover is determinant on the fate of seeds after dispersal (Cintra, 1997). At least two of the 10 most dangerous invasive alien mammals in the world (e. g. O. cuniculus and individuals of the genus Rattus; Lowe et al., 2000;European Environment Agency, 2012;Scanes, 2018) were detected removing a large part of the seeds offered. It is known that rodents of the genus Rattus are not hoarding rodents (Meyer & Butaud, 2009) and that they consume on site small seeds (Lobo et al., 2013). In fact, seed predation by rodents of the genus Rattus was demonstrated by the remains of seeds chewed and nibbled in the SRUs (Fig.  S1 [suppl.]) and also by photographs obtained with camera traps. By contrast, it has been suggested that O. cuniculus could act as a seed predator (Martin et al., 2007) and as a primary and secondary seed-dispersing agent (Pakeman et al., 2002;Dellafiore et al., 2006;Dellafiore et al., 2010). Thus, despite Castro In a) vertical bars on the proportion of seeds removed corresponds to their respective standard error at each date. In b) circle is an outlier and triangles are the mean of the proportion of seeds removed per specie. Different vertical letters (a, b and c) shows statistically significant differences of seeds removed among species (p < 0.05). Table 4. Influence of the factors on the mean proportion of seeds removed of the species studied. The factors shown vertically were tree cover (TC), presence of leaf litter (LL) and seed density (D). Different letters (a, b) show significant differences in the mean proportion of seeds removed between levels of a factor. * p < 0.05; ** p < 0.01; *** p < 0.001 Nevertheless, seed removal and predation (as in this case) by these agents has a negative effect on the regeneration of these tree species, the seedling recruitment, the forest structure and the forest dynamics (Simonetti, 1983a;Díaz et al., 1999;Jaksic, 2001;Baldini & Pancel, 2002;Fernandez & Sáiz, 2007;Castro et al., 2008;Barceló & Simonetti, 2020). It is possible to assume that the removal of tree seeds from species in sclerophyll forests (and others vulnerable ecosystems) is affected along the distribution of the invasive alien fauna in Chile (Fernandez & Sáiz, 2007;Barceló & Simonetti, 2020). Hence the importance of controlling invasive aliens fauna for forest conservation and resto-ration (Fernandez & Sáiz, 2007;Gutiérrez et al., 2007;Barceló & Simonetti, 2020), particularly when the forest already has severe problems with the regeneration process (Fuentes et al., 1983;Fuentes et al., 1986;Fuentes & Muñoz, 1995Jiménez & Armesto, 1992Armesto et al., 1995;Armesto et al., 2007;Schulz et al., 2010) and even more when these are inside protected areas (Barceló & Simonetti, 2020). Nevertheless, seed removal and seed predation has scarcely been evaluated since the introduction of invasive alien fauna in Chile, such as the case of Rattus rattus (introduced in 1540) or O. cuniculus (mid-18 th century; Simonetti, 1983b;Baldini & Pancel, 2002;Camus et al., 2008;Barceló & Simonetti, 2020).

Factor
L. culpaeus cannot handle the seed precisely with its paws (unlike many herbivorous mammals) (González-Varo et al., 2015) and, therefore, it is suggested that it acts as a seed predator-disperser of L. caustica seeds (Castro et al. 1994). However, the dispersal of L. caustica seeds by L. culpaeus is neither effective nor efficient for their subsequent germination (Bustamante et al., 1992;León-Lobos & Arroyo, 1994).

Forest Systems
December 2020 • Volume 29 • Issue 3 • e022 Contrasting the results obtained by Myster & Pickett (1993), our results suggest that most of the fauna that remove seeds have a density-independent behavior similar to that found by Hulme & Hunt (1999), Myster (2003, Marino et al. (2005) and Haught & Myster (2008). Thus, seed removers did not differentiate between SRUs with high and low seed density possibly due a variety of circumstances that could have acted on the influence of this factor in the removal of dispersed seeds. For example, the consumption of the different densities of seeds offered could also be due to the greater time invested by the fauna that looked for and removed it, possibly for not perceiving danger of being predated. In addition, there could have been a large number of fauna demanding the seeds being studied, removing both densities indiscriminately (Haught & Myster, 2008). The suggestion by Forget (1993) is also a possibility, in which there may be years of low fruition and thus little food available for granivores. This would lead granivores to consume the greatest amount of seeds offered regardless of the density.
Leaf litter cover affected the decrease in the removal of dispersed seeds in three of the four species studied. These results are consistent with those obtained by other authors (Shaw, 1968;Radvanyi, 1970;Harper, 1977;Schupp, 1988;Willson, 1988;Cintra, 1997;Reed et al., 2004), who suggest that the influence of this factor on seed removal is a pattern that occurs cross-sectionally in forests. Harper (1977) showed that the mere presence of one dead leaf that covers a seed is enough to reduce the possibility of detection by fauna. This means that the presence of leaf litter cover hides the seeds, even when the fauna removes them they make an effort to find these seeds. On the other hand, in addition to hiding the seeds from the fauna, leaf litter may contribute to creating suitable microsites (humidity and temperature) for subsequent seedling germination and recruitment (Cintra, 1997). While a few amount of seeds without leaf litter cover were not removed, these showed to be desiccated (even more the seeds located without tree cover). Instead, seeds with leaf litter seemed well conserved (seeds of all four species) and some of them had started to germinate (e.g. M. boaria and Q. saponaria).
Tree cover did not influence seeds removed suggesting the presence of predators in both habitats (Jaksic, 2001). This is supported by evidence (e.g., experiment with camera traps) that the rodents are restricted to dense spaces of cover as a response to the risk of predation in open areas between shrubs (Fuentes et al., 1983;Simonetti, 1983a;Díaz et al., 1999), whereas O. cuniculus lives in both places but they could remove in areas without tree cover (Gálvez-Bravo, 2017).
Under the conditions in which this study was conducted, it is an interesting finding that seed removal of R. ephedra, M. boaria and Q. saponaria was severe that suggest possible implications for the recruitment success. Never-theless, the presence of leaf litter prevents seed removal to the alien invasive seed removers and could also generate good microsite conditions for germination and the subsequent establishment of seedlings of the species studied. This suggests that this factor is an indispensable resource to be protected and important to future forest restoration and regeneration plans. Following our findings and others (e. g. Simonetti, 1983a), a restoration management that could be suitable is to exclude individuals trees (the future nurse tree; Root-Bernstein & Svenning, 2017) to the alien invasive seed removers and seedling predators, with leaf litter inside, allowing good and safe microsite conditions. Thus, there could be multiple clumps (i.e. tree exclusions) that are passively restored in a forest stand that can lead their formation through the process of succession (Newton & Cantarello, 2015). As regards to alien invasive fauna to alien invasive fauna, only control methods on O. cuniculus and rodents of the genus Rattus can be applied (instead eradication; but see Allendorf & Lundquist, 2003).
Although this study increases the scientific evidence of the removal of dispersed seeds of woody species in a sclerophyll forest in central Chile, this line of research must continue, as it remains an area with many assumptions awaiting evaluation. For example: How much are the seed predation and dispersal of the other tree species in a sclerophyll forest? What are the effects of climate change and site on the generation of flowers and subsequent seed production? What are the best tools for controlling and eradicating invasive alien fauna? Are invasive alien fauna the main problem for sclerophyll forest regeneration? An extensive body of evidence will provide better restoration and conservation tools for this important ecosystem both nationally and internationally.