Seed type, habitat and time of day influence post-dispersal seed removal in temperate ecosystems

Study area

Our study was conducted in the South of Hesse, Germany, in June and July 2018. Since we were interested in comparing the two common habitats forest and grasslands, we selected forest (coniferous and deciduous) and grassland sites in each of three regions which are 20 km apart on average: Darmstadt, Oberzent/Airlenbach and Bad König/Zell (see Supplemental Information 1 for site coordinates). In Darmstadt, all grassland sites are meadows that are mown once or twice per year. Forest sites are recreational areas comprising only a moderate amount of forestry; they are mixed forests with native deciduous trees. In Airlenbach, grasslands are intensively used: they are mown three times a year and fertilized with cattle manure. Forest sites are recreational areas comprising mixed and coniferous stands. In Zell, grassland use is not agricultural but sites are mown once a year to avoid bush encroachment. Sites are located close to the river Mümling and regularly flooded. Forest sites are recreational deciduous mixed forests.

Plant species and seed parameters

Four plant species with different dispersal modes and seed traits were used:

a) Chelidonium majus; small seeds with elaiosomes, dispersed by arthropods

b) Lotus corniculatus; small seeds lacking elaiosomes, dispersed by longitudinally dehiscent fruits

c) Tragopogon pratensis; medium sized seeds with a pappus, wind dispersed

d) Helianthus annuus; large seeds dispersed by animals (birds, arthropods) and wind, intact (with seed coat) and dehulled seeds were compared.

Accordingly, five seed types were tested in total, represented by a single seed species or in case of H. annuus two variants of seed species. All seeds were purchased at Rieger-Hofmann GmbH, Blaufelden-Raboldshausen, Germany. For each seed type, the dry mass was quantified by measuring the mass of ten randomly selected seeds and the mean and standard deviation for each seed type was calculated: C. majus = 0.7 ± 0.0 mg, L. corniculatus = 1.1 ± 0.1 mg, T. pratensis = 7.9 ± 2.5 mg, H. annuus intact = 83.1 ± 18.9 mg, H. annuus dehulled = 50.4 ± 7.0 mg.

Statistical analysis

Study 1: Seed removal, habitat and seed type

All collected data were analyzed with R 3.5.2 (R Core Team, 2018). Data were checked for normal distribution and log(x+1) or square root transformed to increase homogeneity of variances if necessary.

To test the effect of seed type on seed removal over time in different habitats (forest vs. grassland), we first compared the proportional removal of seeds after 48 h (Meyer et al., 2017). Removal rate was defined as R = log((N_removed/25)+1), with N_removed representing the number of seeds removed after 48 h. Results were analyzed using a linear mixed-effect model (lmer) with R as dependent variable, and region, habitat and seed type as independent variables, including the interaction between habitat and seed type. Plots were included as random factor, individual effects were analyzed by one-way ANOVA included in the car package (Fox & Weisberg, 2019).

We additionally investigated seed removal over time using linear regressions; first, we compared the fit of linear versus non-linear models using Integrated Nested Laplace Approximation (Inla) and results showed linear models to fit best. For each seed type, seed removal (dependent variable) was plotted versus time (predictor). Effects were analyzed using a linear mixed-effect model (lmer) with square root transformed number of seeds removed as dependent variable and region, habitat, time and seed-type as independent variables, including interaction terms between variables. Plots were included as random factor (see Supplemental Information 2 for raw data).

Study 1: Seed removal, habitat and seed type

Seed removal after 48 h differed significantly across seed types but not among regions and only slightly among habitats; however, there was no interaction between habitat and seed type (Fig. 1, Table 1). Both in forests and in grasslands, C. majus and dehulled H. annuus seeds were often removed entirely, and their removal was significantly higher than for L. corniculatus, intact H. annuus and T. pratensis seeds; the latter three species did not differ significantly (Fig. 1). Removal of L. corniculatus, C. majus, dehulled H. annuus and T. pratensis seeds did not differ significantly between habitats, whereas seeds of intact H. annuus were removed faster in grasslands than in forests.

For each seed type in each region, seed removal increased continuously over time (Fig. 2, Table 2). In both regions and both habitats, C. majus and dehulled H. annuus seed removal was fastest, corresponding to the results after 48 h. With one exception (forest, Darmstadt), C. majus seed removal had the highest increase over time, while the removal of dehulled H. annuus seeds was fastest in forests in Darmstadt. However, removal rates changed in time with habitat and seed type (Table 2); e.g., T. pratensis was removed at higher rates in grasslands in Darmstadt but in forests in Airlenbach.

Study 2: Seed removal and time of day

Post-dispersal seed removal at different times of day significantly differed with habitat and region (Table 3). The removal of C. majus was higher at night in grasslands in Darmstadt and forests in Zell, but more pronounced during the day in grasslands in Airlenbach and Zell (Fig. 3B). Night removal of dehulled H. annuus was higher in grasslands in Darmstadt (Fig. 3C) and night removal of intact H. annuus was more pronounced in forests in Zell (Fig. 3D).

In general, seed removal of L. corniculatus was generally low (Fig. 3A). C. majus was removed at higher rate at night in grasslands in Darmstadt and in forests in Zell, whereas day-removal seemed to be more pronounced in forests in Airlenbach and grasslands in Zell (Fig. 3B). The removal of dehulled H. annuus was generally low in Darmstadt, but higher in grasslands in Airlenbach and forests in Zell (Fig. 3C). Intact H. annuus was even less removed than dehulled seeds, but tended to be removed at higher rate at night (Fig. 3D).

Study 1: Seed removal, habitat and seed type

Post-dispersal seed removal after 48 h differed across plant species. In both habitats, forest and grassland, seed removal of C. majus and dehulled H. annuus seeds was significantly higher than the removal of L. corniculatus, intact H. annuus and T. pratensis seeds.

In contrast to L. corniculatus, C. majus seeds are myrmecochorous because of the presence of elaiosomes (Pemberton & Irving, 1990; Peters, Oberrath & Böhning-Gaese, 2003). These lipid-rich appendages are attractive to many ant species and promote burial and dispersal (Hughes, Westoby & Jurado, 1994; Fischer et al., 2008). Since the abandoned seeds maintain their ability to germinate, seeds are dispersed effectively (Kjellsson, 1985; Gorb & Gorb, 2003). We observed that ants removed seeds of both C. majus and L. corniculatus, but preferred seeds of C. majus.

Removal of dehulled H. annuus was significantly higher than of natural, intact H. annuus seeds. This result mirrors the physical protection of the nutritional seed components by the seed coat (Souza & Marcos-Filho, 2001) that are apparent in dehulled seeds, and may additionally be driven by increased energy requirements for handling such intact seeds (Pyke, Pulliam & Charnov, 1977). In our study, we observed that ants and slugs consumed dehulled but ignored intact seeds, which was most probably directly related to the accessibility of the seed material in dehulled seeds. In addition to physical protection, dehulled seeds may produce more attractive odors, resulting in higher seed predation by granivores guided by olfactory cues (Jaganathan, 2018; Vander Wall, 1998).

Tragopogon pratensis seeds are attached to a pappus and are primarily wind-dispersed (Casseau et al., 2015). Thus, we expected the removal by animals to be of minor importance. As predicted, removal of T. pratensis seeds was low and no T. pratensis seed removal by animals was observed (but predators/dispersers were not monitored over time). However, secondary seed dispersal by animals may be highly relevant also for primarily wind-dispersed plant species (Der Weduwen & Ruxton, 2019), e.g., when wind-dispersed pines (Pinus spp.) are dispersed by scatterhoarding rodents (Vander Wall, 2003).

In summary, the results of our study show that post-dispersal seed removal depends on seed specific traits that differ among the seed species we used. Smaller seeds (C. majus, L. corniculatus), seeds that are more easily accessible (dehulled H. annus) and those dispersed by animals (C. majus) are removed faster than large (intact H. annuus), protected (intact H. annuus) or those seeds that depend on abiotic conditions for dispersal (T. pratensis). Since C. majus were removed at the highest rate, the presence of elaiosomes may have strengthened the advantage of being small. Thus, C. majus may have achieved an increased germination probability in new habitats away from the mother plant. However, the variation of seed traits used in this study may not be exclusively causal for the observed results owing to the potential collinearity with other unmeasured traits.

Typically, habitat (forest vs. grassland) had no effect on seed removal, although removal of intact H. annuus was higher in grasslands. In general, as seed predators may differ in spatial foraging patterns (Brown et al., 1975), many studies indicate a relevant effect of habitat type on seed removal concerning both microhabitats (Notman, Gorchov & Cornejo, 1996; Ji-Qi & Zhi-Bin, 2004) and forest and grassland sites (Holl & Lulow, 1997). Previous studies comparing levels of seed removal in early- and late-successional habitats have shown variable results: a few studies indicated that seed predation is higher in early than in mature successional habitats (Uhl, 1987; Hammond, 1995) whereas other studies reported opposite results (Aide & Cavelier, 1994; Osunkoya, 1994). Other studies showed that the habitat with the highest level of seed predation varies with seed species (Willson & Whelan, 1990; Whelan et al., 1991; Holl & Lulow, 1997). However, the effect of habitat on post-dispersal seed removal seems inconsistent and may depend on seed type and predator. Therefore, a comprehensive seed study of the plant community in different habitats and seed predators therein may be necessary to solve that problem.

Study 2: Seed removal and time of day

The effects of time of day on post dispersal seed removal of seed types differed with habitat and region. Contrary to this finding, several former studies indicated that seed predators differ not only in their preferences for seed types but also in their temporal patterns of foraging activities (Brown et al., 1975). Granivorous rodents are mainly nocturnal (Abramsky, 1983; Miller, 1994; Xiao, Zhang & Wang, 2005), whereas harvester ants are mostly diurnal (Abramsky, 1983; Díaz, 1992), as ectothermy of ants renders foraging activities temperature-dependent (Whitford et al., 1981; MacKay & MacKay, 1989). Thus, we expected that time of day influences seed removal, and that removal of myrmecochorous C. majus seeds was higher at day which was true for grassland sites in Airlenbach and Zell. Contrary to this expectation, removal of C. majus seeds was even higher at night in grasslands in Darmstadt and forests in Zell. Possibly, unusually very high temperatures in the summer 2018, may have shifted foraging activities of diurnal granivores to the early morning and late evening hours, which were covered by the nighttime assessments.