Recruitments in the regeneration habitats depend on bio- bio-interactions between animal dispersal and spatiotemporal dynamics of

seed rain

Abstract

The seed-to-seedling transition is crucial during tree regeneration but highly variable, revealing how the reproductive success is affected by seed production and dispersal. The seed production and dispersal suffer pre-dispersal and post-dispersal predation, especially for animal-dispersed species. We thereby expect that the uncoupling in seed-to-seedling transition may depend on animal dispersal. To explore this transition, we conducted a 5-year seed rain survey (2008–2009, 2010–2012), and associated these data with the seedling recruitments in a montane cloud forest, Taiwan. Most of the tree species reproduced periodically, correlating to local climatic seasonality; however, their reproductions were uncoupled to seedling recruitments that were aperiodic. Only for the large-seeded species whose seeds were mostly predated, a concordant seedling recruitment responded to the amount of diaspore fragments; thus the uncoupling could be partly explained by pre-dispersal seed predation. In contrast, the large-seeded, mammal-dispersed species had a disproportionately high recruitments compared to their relatively few seeds in the seed rain. For small-seeded species, the seed availability was substantially abundant, and bird-dispersed species had much more seedling recruiments compared to wind-dispersed species. The differences in seed-to-seedling transition can be further explained by the the nearest possible dispersal distance, and the ability to germinate from soil seed bank, both of which facilitate bird-dispersed species colonizing regeneration habitats across years. For animal-dispersed species, both colonizing a new

habitat and dying in established habitats had a yearly increasing trend. The seed-to-seedling transition thus was variable depending on the bio-interactions between animal dispersal and spatiotemporal dynamics of seed rain.

Key words

Dispersal syndrome, pre-dispersal predation, recruitment limitation, regeneration habitat, reproductive phenology

Introduction

A central question of recruitment dynamics is whether the seed production can effectively contribute to seedling establishment in a long-term tree regeneration process (García et al., 2005; Rother et al., 2015; Wright et al. 2016). The seed-to-seedling transition represents the reproductive efforts of a tree species as its life-history strategies to adapt to the microenvironments and sustain its subpopulation in the forest (Wang and Smith, 2002; Swamy et al., 2011; Muscarella et al., 2013; Robledo-arnuncio et al., 2014).

However, the uncoupling from seed to seedling can occur during every intermediate stage, including pre-dispersal, dispersal, and post-dispersal stages, before a seedling finally recruits in the regeneration habitat (Nathan and Muller-Landau 2000).

It is difficult to determine which factor is more crucial in affecting the seed-to-seedling transition in pre-dispersal and post-dispersal loss of seeds (seed removal from trees, or from forest ground, respectively) (Hulme, 1998; Wright et al., 2005; Kolb et al., 2007).

Post-dispersal loss of seeds can erase the dispersion patterns of seed rain. However, post-dispersal loss of seeds is largely caused by generalist predators (Hulme 1998). If such post-dispersal predation is homogeneous among tree species and the heterogeneous seed dispersion patterns still can be kept even under post-dispersal predation, spatial concordance from the seed to seedling stage may occur (García et al. 2005). In contrast, pre-dispersal predation do not change the dispersion patterns of seed rain. But if most seeds were predated and not dispersed in the pre-dispersal stage, it could be a great impact on reproductive efforts invested by the mother trees and may reduce fitness (Bonal et al., 2007; Sun et al., 2007). The “economy of scale” hypothesis explains that the great reproductive efforts may lead to the satiation of specialist predators which favors the pre-dispersal stage (Hulme, 1998; Wright et al., 2005), and consequently results in more survivals of seeds, as well as more chances for the seed-to-seedling transition.

The concordance from seed to seedling thereby depends on plant-animal interactions in seed dispersal (so called “dispersal syndromes”) with both animal dispersers and seed predators. Whether one animal is a frugivore or a granivore can possibly be both the seed disperser and seed predator. Frugivores can transport seeds for longer distances, whereas some granivores can also be effective dispersers by moving seeds in microsites (scatter-hoarding) that are often better regeneration habitats (Levey et al., 2002; Vander Wall and Beck, 2012). These two animal dispersal syndromes (frugivory and scatter-hoarding) which involve vertebrates also cause seed loss by predation (Vander Wall and Beck, 2012). Animal dispersal syndromes affect not only primary seed dispersion (i.e., seed rain) (Helene et al. 2008), but also indirect or multiple dispersal through synzoochory (Wenny, 2000; Vander Wall et al., 2005). For examples, seeds can be cached several times by scatter-hoarding animals (e.g., rodents) (Jansen et al., 2012), or caught by mouthparts of animals (e.g., via monkeys’ cheek pouch, Yumoto et al., 1998). Primates are both important seed dispersers and predators. They constitute a large proportion of the frugivore biomass in a forest, and through their seed-handling behaviors, including seed splitting, dropping, or carrying that, facilitate both multiple dispersal and seed survivals (Chapman, 1995; Seufert et al., 2010). Such deposition processes (during post-dispersal) by primates increase spatiotemporal variations in recruitments of some large-seeded species, and cause uncoupling in the seed-to-seedling transition (Balcomb and Chapman, 2003; Jansen et al., 2004).

Interannual seed production may increase the variability of seed-to-seedling transition by affecting the satiation of predators (Wright et al., 1999; Bonal et al., 2007), negative density dependency (Wright et al. 2005), or interspecific competition (Usinowicz et al., 2012). Seed rain dynamics thus constitutes a community phenological pattern that has implications for reproductive efforts of trees, and plant adaptations to biotic (e.g.,

predation) or abiotic (e.g., local climates) pressure (Rathcke eta Lacey, 1985; van Schaik et al., 1993; Sakai, 2001), and may drive species spatial distribution (Chuine 2010).

In this study, we conducted a 5-year seed rain survey to explore the causes and consequences of seed-to-seedling transition in recruitment dynamics in a subtropical montane cloud forest, Taiwan. We anticipate that the phenological patterns of seed rain is greatly constrained by the climatic seasonality here at an around 2,000 m elevation. If so, any synchrony, such as clumping in fruiting time before the winter, is worth investigating its effects on the seed-to-seedling transition. Moreover, for there are so many mammalian frugivores and granivores here, we expect that the coupling (or uncoupling) of seed-to-seedling transition would be affected by animal dispersal. Here the seed dispersal modes include anemochory (wind dispersal), and two major types of animal dispersal syndromes; that is, frugivory and scatter-hoarding dispersal (Vander Wall and Beck, 2012). We then subdivided seed dispersal by frugivory into bird and mammal dispersal, thus including both endozoochory and synzoochory. Besides dispersal modes, still some factors or species traits may explain the uncoupling from seeds to seedlings, such as dispersal distance from mother trees (Zuidema et al. 2010), forest gaps (Puerta-Piñero et al. 2013), the ability to germinate from soil seed bank (seed-bank trait, hereafter) (Alvarez-Buylla eta Martínez-Ramos, 1990; Dalling eta Hubbell, 2002), etc.

To explore the seed-to-seedling transition, and its consequences on recruitments in the regeneration habitats, we (1) examine the seed availability in terms of periodicity in reproductive phenology of various plant species from seed rain data in relation to the local climatic factors; (2) evaluate the pre-dispersal seed predation from diaspore fragments in the seed rain, in relation to both seed and seedling abundance (density) and diversity of different dispersal modes; (3) assess seed-to-seedling transition in each sampling site for each species (species-specific regeneration habitats) in relation to various influential

factors, such as dispersal modes, dispersal distance, seed-bank trait, etc. We also examine the consistency of seed-to-seedling transition in species-specific regeneration habitats across years, so that we can explore the potential of seed rain that can contribute to forest community assembly.

Material and Methods

Study area

The study was conducted in the Nanhsi forest dynamics plot. The geographic location and local climate of this plot are stated in Chapter 1 Study area. Recall that Typhoon Morakot (August, 2009) caused landslide here, and created gaps and a decline of canopy cover. In addition, the monthly and yearly variations in rainfall were great. During our survey years, the mean annual rainfall was 5,554 mm (2008–2009), and 4,159 mm (2010–

2012) (Fig. S4.2c).

Vegetation

The forest vegetation and tree data according to Yang et al. (2008) are stated in Chapter 1 Vegetation. The tree seedling assemblages, censused in 10 parallel transects with a total of 994 2 m × 2 m quadrats in 2009 (before Typhoon Morakot, Fig. S4.1), consisted of 19 families, 29 genera and 36 tree species with 3,687 individuals (Weng et al. 2017b). The seedling species were mostly found in the families: Lauraceae, Pentaphylacaceae, Adoxaceae, and Fagaceae. A consecutive seedling survey for three years (2010–2012, final census in January, 2013), in 324 units of 1 m² quadrats (Fig. S4.1), contained 953 tree seedlings of 36 species that belong to 20 families and 33 genera (Weng et al. 2017a).

Most seedlings were found in the families: Lauraceae, Rutaceae, and Magnoliaceae.

Field surveys on seed rain dynamics

In order to explore the spatiotemporal dynamics of seed rain, we used seed traps that were constructed from 1 mm nylon-mesh net with 0.5 m² surface area, and suspended from a PVC frame at 0.75 m high above ground, to collect plant reproductive parts dispersed from forest canopy. In November 2007, 50 traps were set and separated by a distance of 20 m (Fig. S4.1). Most of the traps were destroyed by Typhoon Morakot in 2009. In 2010, we started our seedling dynamic census (Weng et al. 2017a), and meanwhile, reconstructed 50 traps. Each trap was separated by a distance of 20 m and could be associated with nearby seedling quadrats (Fig. S4.1). We use “year 2008–2009 cohort”

and “2010–2012 cohort" to refer to the two different sets of traps in the year groups.

The collections in the traps were censused bi-weekly from November 2007 to August 2009, and resumed from Feburary 2010 to January 2013. All plant reproductive parts in the collections were identified to the species or genus level, including both tree species and dependent plant species (mainly, climbing vines and parasitic plants). The reproductive parts were categorized as flowers, immature fruits, fruit appendages (e.g., cupules for nuts, receptacles for follicles, capsules, hypocarps or other tissues that species-specifically can be treated as fruit identity), diaspore fragments (majorly due to pre-dispersal predation), and seeds. The collections of these reproductive parts may exhibit some lags compared to the “real” phenological events (on trees) due to bi-weekly censuses and delayed release from trees. Note that the absence of seeds in traps may arise from (i) few or no mother trees fruited this year, (ii) the fruiting trees were too far, and (iii) seed removal majorly occurred on trees in pre-dispersal stage, especially due to some animal-dispersed species, and by some fruit-manipulating behaviors of frugivores.

Dispersal modes by seed dispersers

To explore how seed dispersal may affect the seed-to-seedling transition, we categorized seeds into four types (hereafter, dispersal modes) by their major dispersers or dispersal agents in our plot:

(i) Wind, which is the dispersal agent for those seeds with wings or pappus. Top abundant wind-dispersed seeds belong to Alnus formosana (Betulaceae), Schima superba (Theaceae), and Acer kawakamii (Sapindaceae).

(ii) Birds, which disperse seeds after fruit ingestion (endozoochory) for those small fleshy fruits and dehiscent fruits with small seeds. Top abundant bird-dispersed seeds belong to Tetradium glabrifolium (Rutaceae), Michelia formosana (Magnoliaceae), Callicarpa formosana (Lamiaceae), and Eurya spp. (Pentaphylacaceae).

(iii) Mammals, which disperse seeds for those large-seeded drupes or large fleshy fruits whose seeds can survived after ingestion (endozoochory), or can be carried by mouthparts (synzoochory). Specifically, from our field observations, a sizable proportion of seeds of Lauraceae species, such as Litsea, Machilus, and Neolitsea spp., can be dispersed by Formosan macaque. Previous studies revealed such seeds dispersed via monkeys’ cheek pouches, so called cheek-pouch dispersal (Yumoto et al., 1998).

(iv) Scatter-hoarding types, for those large seeds that are moved and are majorly predated by rodents or other animals, such as nuts of Fagaceae species. These seeds can survived if they were dropped or cached (but forgotten) by scatter-hoarding animals.

Climatic factors and other environmental factors

To examine the reproductive phenology in relation to local climatic factors, we adopt a long-term climatic records from the nearby A-li mountain weather station at similar elevation. The temperature and rainfall there are highly correlated to those in the Nanhsi plot (both R > 0.99, P < 0.01), but the meteorologic instruments in the Nanhsi plot stop

working during survey years. The climatic factors include monthly mean temperature, minimum temperature, daily temperature range, sunshine hours, rainfall, rainy days, and minimum relative humidity. Other biotic factors (e.g., canopy cover, tree density, overstory basal area, species diversity of adult trees) and abiotic factors (e.g., topographic factors) are derived from our previous studies (Weng et al. 2017a; Weng et al. 2017b).

Data analysis

Testing periodicity in reproductive phenology

To explore the phenological patterns of seed rain and test the significance of periodicity in reproductive phenology intra-specifically, we use circular vector algebra to represent yearly recurrence of phenological patterns (Wright and Calderón, 1995), including flowering, fruit setting (by fallen immature fruits), fruiting (by fallen mature fruits or seeds), and accompanied fruit appendages and diaspore fragments (each as a

“phenophase”). The census date t of each record for one phenophase i of one species j can be expressed as the angle ϕi, j, t of a vector on a 360° circle, with vector length ri, j, t

equals the count (in this record) in percentage of a yearly total. Here, the count of seeds includes those seeds within a mature fruit. The count of flowers is calculated by the sum of presence in every trap. The annual mean date of one phenophase can therefore be

with is the vector length, representing the concentration of the phenological pattern.

The significance of periodicity in phenological patterns is tested by 1,000 permutations, randomly shuffling yearly records of each phenophase, and re-calculating the mean dates.

The observed mean date of a phenophase in one year (yr) is periodic if the median of difference with mean dates in other years, multiplied by the concentration, i.e., cos ( _{, ,} − _{, ,} ) , is significantly greater than expected by chance (one-tailed test, with P < 0.05).

To explore reproductive phenology in relation to the climatic factors at the community level, a cross correlation is used to examine temporal relationships between community reproductive patterns and climatic factors (in 0–6 lags). The community reproductive patterns were evalued for the monthly abundance of flowering species, and that of fruiting species. The seasonal components of the community reproductive patterns, as well as that of climatic factors, are removed by using a seasonal-trend decomposition procedure based on Loess regression (STL, Cleveland et al., 1990) prior to applying cross correlations.

Examining the relationships between seeds and pre-dispersal seed loss.

We treat diaspore fragments as the pre-dispersal seed loss by predation, and examine the correlation between diaspore fragments and seeds. Seeds of animal-dispersed species are expected to come with more diasprore fragments. In addition, if the pre-dispersal predation was accompanied with more animal dispersal, it might affect the local recruitments of animal-dispersed species. We then examine the correlation between diaspore fragments (in traps) and seedling density and diversity of animal-dispersed species in nearby quadrats. Here, we manually associate each seed trap with nearby seedling quadrats (Fig. S4.1), so called a “site” (i.e., with both seeds dispersed (in trap) and seedling recruitments nearby).

The fruit appendages in a trap imply that the diaspores might have being dispersed to the local microhabitats near the trap. The temporal relationships between fruit appendages and seeds therefore could be as close as that between diaspore fragments and seeds. To examine this, we apply cross correlations between community fruiting pattern and the fruit appendages, as well as diaspore fragments, after removing seasonal components in them. We treat the presence of the seeds and fruit appendages in each trap as the potential seed arrival in each site in the following analyses.

Exploring the factors affecting the seed-to-seedling transition in regeneration habitats To tackle the seed-to-seedling transition in the regeneration habitats, we first define that a species-specific “regeneration habitat” is a site that had at least one seedling (of that specific species) recruited. Thus, if at least one potential seed (of that species) arrived at the associated trap, it is a species-specific regeneration habitat with seed subsidizing. We categorize sites for each species as three regeneration habitat types: (i) subsidized sites – the sites with seedlings established as well as seeds arriving (of a given species); (ii) established sites – the sites with seedlings established but no seeds subsidized; (iii) seed-arriving sites – the sites with only seeds seed-arriving but still no seedlings established, as the

“potential” regeneration habitats. For each species, each site can be recognized as subsidized, established, or seed-arriving site according to if any seed/seedling of that species arrived/recruited at that site. Thus, every site may be different regeneration habitat type for every species. We use the presence or absence of seeds/seedlings of one species in a site throughout each year cohort (for 2 or 3 years) to determine the regeneration habitat type, so that the tendency of one species had more of any of the three regeneration types can be hardly due to chance alone. For example, if one species had nine seed-arriving sites but only one subsidized site (9/1), its seed-to-seedling transition must be

more uncoupled than that of the species who had 1/9. Such distributions of regeneration habitat types among species are compared. In seed-arriving sites (without seedlings) and established sites (without seeds), the seed-to-seedling transition is uncoupled. However, note that the seedlings in established sites stem from the seeds that had been here by storing in the soil seed bank, or by some post-dispersal.

We used Generalized Linear Mixed Models (GLMM) to examine if each site for each species belongs to the subsidized site (1, otherwise 0), established site (1, otherwise 0), and seed-arriving site (1, otherwise 0) (i.e., three response variables) in each year cohort.

The sites are random effects in the three GLMM models. A negative binomial distribution with logit link function in each GLMM (performed by the R package “glmmADMB”) was used to link with various explanatory variables. We examine the following explanatory variables: (1) seed rain effects, including dispersal modes, pre-dispersal seed predation, nearest possible dispersal distance and nearest possible mother-tree basal area (distance from the trap to nearest possible species-specific adult tree, with its basal area) for each species; (2) seed-bank trait; (3) seed and seedling diversity of animal-dispersed species for each site; (4) adult-tree factors, including tree diversity, density and canopy cover for each site; (5) topographic factors, including altitude and slope for each site; (6) the mean climatic conditions (aforementioned climatic factors) of three consecutive months for species-specific mean fruiting date; (7) year cohort effect.

Assessing the consistency of a species-specific regeneration habitat across year cohorts By linking the regeneration habitat types in the overlapping sites across the two year cohorts (Fig. S4.1), we test the three models for the consistency of a species-specific regeneration habitat by GLMM (with aforementioned random effects, link function, and explanatory variables):

(i) Colonized model – the seed-arriving site (without seedlings) in 2008–2009 cohort, which had turned into a subsidized or established site (with seedlings) in 2010–2012 cohort for each species (1, otherwise 0). The colonized model suggests that the potential of a seed-arriving site is a new regeneration habitat (colonization) for one species.

(ii) Discontinuous model – the subsidized or established site in 2008–2009 cohort, which had no seedlings recruited in 2010–2012 cohort for each species (1, otherwise 0). The discontinuous model suggests the discontinuity of a regeneration habitat, i.e., a seedling species cannot sustain in this regeneration habitat even with subsidizing seeds.

(iii) Persistent model – the subsidized or established site across the two year cohorts for each species (1, otherwise 0). The persistent model suggests the persistence of a species-specific regeneration habitat.

Results

A total of 5,924 flowering records and 249,223 seeds for tree species were censused during 119 bi-weekly surveys. The flowering records contained 23 species, 22 genera, and 12 families. While there were more tree species found in the seeds (33 species, 29 genera, and 17 families), two small-seeded, wind-dispersed species in Betulaceae (86.4%) and Theaceae (5.4%) accounted for most of the seed abundance (Table S4.1). The seeds had a long-tailed distribution among other species, especially for animal-dispersed species (mostly less than 1%, Table S4.1).

Phenology: Most species reproduced annually and its reproduction was related to climatic seasonality

We find aperiodic seedling recruitments in our plots (Fig. 4.1) which had local peaks in recruit rate (in July 2010, November to December 2010, and October 2011). In contrast,

the reproductive phenology of tree species was more periodic during survey years. At the

the reproductive phenology of tree species was more periodic during survey years. At the