Seed harvesting is influenced by associational effects in mixed seed neighbourhoods, not just by seed density

Authors


Correspondence author. E-mail: sostoja@usgs.gov

Summary

  1. Rodents frequently forage in a density-dependent manner, increasing harvesting in patches with greater seed densities. Although seldom considered, seed harvesting may also depend on the species identities of other individuals in the seed neighbourhood. When the seed harvest of a focal species increases in association with another seed species, the focal species suffers from Associational Susceptibility. In contrast, if seeds of the focal species are harvested less when in association with a second species, the focal species benefits from Associational Resistance.
  2. To evaluate density dependence and associational effects among seeds in mixtures, we conducted seed removal experiments using a completely additive design patterned after a two-species competition experiment using seeds of either Achnatherum hymenoides (Indian ricegrass), Leymus cinereus (basin wildrye) or Pseudoroegneria spicata (bluebunch wheatgrass), all native perennial grasses, combined with seeds of Bromus tectorum (cheatgrass), a non-native annual grass. The experiment involved placing five fixed quantities of the native seeds mixed with five fixed quantities of B. tectorum seeds in a factorial design, resulting in 35 seed mixture combinations. The seed-eating rodent community at our study sites, in order of abundance, is composed of Peromyscus maniculatus (North American deer mouse), Dipodomys ordii (Ord's kangaroo rat) and Perognathus parvus (Great Basin pocket mouse).
  3. Native seed harvesting was density dependent, with a greater proportion of seeds being harvested as density increased. In the mixed density model, the presence of B. tectorum did not affect harvest of any of the native species' seeds when analysed individually. However, when all three native species were analysed together, increasing quantities of B. tectorum resulted in reduced harvest of native seeds, demonstrating weak but significant Associational Resistance. In contrast, harvest of B. tectorum seeds increased when in combination with any of the native seed species individually, indicating relatively strong Associational Susceptibility.
  4. These results demonstrate that seed harvest is determined not just by seed density, but also by the local seed neighbourhood and suggest that associational effects between native seeds and B. tectorum can occur in field conditions. The ecological implications of seed selection and associational effects on plant populations in natural and managed systems are also discussed.

Introduction

Foraging decisions likely involve balancing energy gain with the perceived risk of being eaten (Longland & Price 1991). The effect of consumers on plant populations depends on, among other factors, the density of propagules (Veech & Jenkins 2005), where increased seed quantity can promote increased rates of seed harvest (Price & Heinz 1984). Density-dependent foraging is predicted by optimal foraging theory (Charnov 1976), which suggests that animals should spend more time foraging at more dense resource patches, maximizing gain while reducing associated costs of foraging (e.g. cost of travel, predation risk, etc.). Veech & Jenkins (2005) define density-dependent foraging as the harvest of a greater proportion of seeds (of a given seed species) from high-density patches than from low-density patches and ‘overall’ seed density as the combined density of seeds of all species in a patch. Density-dependent foraging has been well demonstrated for granivorous rodents in North American deserts (Nelson & Chew 1977; Price & Heinz 1984; Brown & Mitchell 1989; McMurray, Jenkins & Longland 1997; Veech & Jenkins 2005), where greater local seed density can alter the foraging behaviour of consumers by attracting mobile consumers (Holt & Kotler 1987).

At the same time, resource-wide qualities such as seed neighbourhood characteristics can influence patterns of seed harvest (Ostoja, Schupp & Klinger 2013). Mixed-species plant (e.g. Callaway et al. 2005) and more recently seed (e.g. Veech 2001; Emerson et al. 2012; Ostoja, Schupp & Klinger 2013) neighbourhoods have been shown to result in associational effects among co-occurring resources that are mediated by generalist consumers. As seed predators generally encounter seeds in mixed-species rather than monospecific patches, such interactions between seed species in mixture can influence the harvest of individual seed species (Brown & Mitchell 1989; García, Martínez & Obeso 2007). After all, a resource's susceptibility to attack is not only influenced by its own palatability but also by the palatability of its neighbours (Emerson et al. 2012). Considering the broader range of herbivory studies, in some studies, palatable plants growing among unpalatable plants were attacked more than when growing with conspecifics (Bergvall et al. 2006), while in other studies, plants were consumed less when surrounded by lower-quality plants (Holt & Kotler 1987; Hjältén, Danell & Lundberg 1993; Callaway et al. 2005). Such differential patterns of consumer selectivity demonstrate that consumption of a resource is not only dependent on the ‘quality’ of the resource itself, but also on the overall range of ‘qualities’ of other resources in the neighbourhood.

In seed neighbourhoods with mixtures of species, some species may confer Associational Resistance (Tahvanainen & Root 1972) to seed predators on other seed species. This concept, initially founded on studies of herbivory, also has been referred to as apparent mutualism (Holt 1977), associational plant defence (Hay 1986) and associational plant refuge (Pfister and Hay 1988). In the context of seed consumption, Associational Resistance occurs when harvesting of one seed species is diminished by the presence of other seed species in the local neighbourhood. This might lead to linked plant population dynamics where an increase in one species leads to an increase in another species (Holt 1977; Veech 2000). Such Associational Resistance driven by herbivory has been shown to alter plant community structure. With increasing abundance of an undesirable grass species, herbivore-induced mortality of the highly palatable grass Themeda triandra decreased (McNaughton 1978). Similarly, two unpalatable native invaders in heavily grazed subalpine meadows of the Eurasian Caucasus Mountains provided strong Associational Resistance for more palatable plant species, leading to increased species richness (Callaway et al. 2005). However, scaling up from associational effects among seed mixtures to differential patterns of seed survival and plant establishment has not been demonstrated. While plants growing in dense or diverse vegetation are frequently less susceptible to herbivory than are isolated plants or plants growing in monospecific stands (Tahvanainen & Root 1972; Bach 1980; Rausher 1981; Callaway et al. 2005), we are aware of only two reports of Associational Resistance with seed removal experiments, one with squirrels (Emerson et al. 2012) and one with ants (Ostoja, Schupp & Klinger 2013).

Alternatively, being in a mixed seed neighbourhood might increase the susceptibility of a seed species to predation. Theoretical treatments suggest that a forager may generate a negative interaction among two resource types (e.g. seeds) when present together (Heller 1980; Holt & Kotler 1987). This is known as Associational Susceptibility (Hjältén, Danell & Lundberg 1993; also as short-term apparent competition, Holt & Kotler 1987; or shared doom, Emerson et al. 2012). Associational Susceptibility occurs when an increase in the quantity of seeds of one species leads to an increase in the harvesting of the seeds of another species. For example, Astragalus cicer suffered Associational Susceptibility in the presence of Achnatherum hymenoides due to reduced foraging in patches with only A. cicer compared to patches with both seed species (Veech 2001).

Patterns of seed predation in a temperate secondary forest in Spain were influenced by seed species identity, quantity and the background seed neighbourhood (García, Martínez & Obeso 2007). Brown & Mitchell (1989) showed that increasing the density of a resource type resulted in higher mortality on a second resource type, suggesting interplay between resource density and associational effects. Similarly, Emerson et al. (2012) showed that the harvest of less palatable sunflower seeds increased in the presence of more palatable sunflower seeds, but the actual rates of harvest depended on the relative densities of the two resource types. These results suggest that the effects of neighbours on associational outcomes likely depend not just on their identities, but also on the relative abundances of the different species in the neighbourhood.

Granivory is a critical process contributing to the composition and structure of arid and semi-arid plant communities (Morton 1985; Heske, Brown & Guo 1993; Marone, De Casenave & Cueto 2000). Seed dispersal networks in deserts sometimes involve highly evolved mutualisms with granivores (Longland et al. 2001), a resource–consumer interaction with keystone effects ecosystem wide (Brown & Heske 1990; Longland & Ostoja 2013). Many desert ecosystems, including those in Western North America, have been heavily invaded by and negatively impacted by non-native plant species, primarily annual grasses (Knapp 1996), which disrupt key natural feedbacks (D'Antonio & Vitousek 1992; Brooks et al. 2004). Among the consequences of invasive species, dominance is the disruption of granivore communities (Ostoja & Schupp 2009; Ostoja, Schupp & Sivy 2009; Litt & Steidl 2011). However, the extent to which invasive species affect seed dispersal/predation networks in these systems is not understood. Invasive species may affect native networks directly via the establishment of new seed–seed remover interactions with native biota and indirectly by altering the abundance, composition or behaviour of the native species (McConkey et al. 2012). In a temperate Mediterranean grassland, exotic Brassica nigra limited the establishment of native Nassella pulchra in their neighbourhood by increasing consumption by small mammal consumers (Orrock, Witter & Reichman 2008). Although apparently driven by adult B. nigra altering mammal foraging patterns by altering assessment of risk rather than a seed–seed interaction, this highlights that exotics can have strong indirect effects on native plants by altering consumption rates. Given the importance of granivory in these systems, considering basic dynamics associated with seed–seed consumer interactions in the face of invasion is critically important.

While both overall seed densities and relative proportions of different seed species in mixed-species neighbourhoods can affect seed harvesting, their effects on granivore foraging behaviour and seed removal have been poorly explored. Moreover, to our knowledge, no study has investigated seed–seed interactions between native and non-native species, yet in the face of widespread and abundant invasive species (e.g. cheatgrass; Bromus tectorum), these dynamics may have important restoration implications (McConkey et al. 2012). In this study, we examined the importance of seed density and, more importantly, of relative proportions of seed species in mixtures on seed removal of three common native perennial grasses (A. hymenoides, Leymus cinereus and Pseudoroegneria spicata) and one exotic annual grass (B. tectorum) by rodent granivores in a Great Basin big sagebrush community. This granivorous rodent community was dominated by Peromyscus maniculatus, Dipodomys ordii and Perognathus parvus.

We had two objectives. First, we tested whether seed harvest of the three native grasses was density dependent, using seed weight as a proxy for density. In particular, we assessed whether proportional harvest of the native species depended on their own density and on the density of B. tectorum. Second, we tested whether associational effects (Associational Resistance or Associational Susceptibility) occur between seeds of B. tectorum and the three native grasses when in mixture. For our first objective, based on extensive demonstration of density-dependent seed harvest (Price & Heinz 1984; Veech & Jenkins 2005), we predicted that with increased native seed availability, a greater proportion of seeds would be removed by rodents. For our second objective, we predicted that the harvest of native seeds would be less when in mixture with B. tectorum than when in monospecific patches (Associational Resistance) due to increased handling time to sort natives from B. tectorum. Conversely, we predicted that more B. tectorum seed would be harvested when present with native seeds than when alone (Associational Susceptibility) because the greater preference of native seeds would result in greater overall harvest in these patches. We further predicted that Ahymenoides would provide the greatest degree of Associational Susceptibility to B. tectorum because it (A. hymenoides) is highly preferred by rodent granivores (Kelrick et al. 1986; Veech 2001; S. M. Ostoja, unpublished data) and that B. tectorum would provide the weakest degree of Associational Resistance to A. hymenoides based on the same rationale.

Materials and methods

Study Sites and Species

The study was conducted in west-central Utah, USA, at Vernon Hills (12 384335E 4438482N) and Simpson Springs (12 350537E 4437129N), which are, respectively, about 155 and 172 km south-west of Salt Lake City, Utah, in Tooele County. Vegetation is typical of Great Basin Wyoming big sagebrush communities. Wyoming big sagebrush (Artemisia tridentata ssp. wyomingensis) is dominant, although other shrubs such as fourwing saltbush (Atriplex canescens), broom snakeweed (Gutierrezia sarothrae), yellow rabbitbrush (Chrysothamnus viscidiflorus) and Mormon tea (Ephedra viridis) are present. The understory is dominated by Indian ricegrass (A. hymenoides), bottlebrush squirreltail (Elymus elymoides), Sandberg bluegrass (Poa secunda), needle and thread grass (Hesperostipa comata), basin wildrye (L. cinereus), bluebunch wheatgrass (P. spicata) and some cheatgrass (Btectorum).

Native seeds were bought from Granite Seed Company, Lehi, UT, USA. Seeds of B. tectorum were collected near the research sites in 2005. The rodent community is numerically dominated by P. maniculatus, which accounted for >50% of all individuals captured (Ostoja & Schupp 2009). Other nocturnal granivorous rodents trapped were D. ordii and P. parvus (Ostoja & Schupp 2009). Camera trap observations of visitation to seed trays indicated that P. maniculatus and D. ordii were the primary foragers.

Seed Trays

To assess intraspecific and interspecific effects on removal rates in two-species seed mixtures, we used a cafeteria-style seed removal experiment with a completely additive design modelled after a two-species plant competition experiment using seeds of one of the native species in combination with B. tectorum. We placed 0, 1, 2, 3, 4 or 5 g of a native species' seeds with 0, 1, 2, 3, 4 or 5 g of B. tectorum seeds, resulting in 35 weight (density) combinations, omitting 0, 0 treatments. Preliminary trials were conducted to determine the range of seed weights needed to ensure that all seeds presented were not always completely removed across treatment/mixture and species combinations. Each of seven 1·2-ha plots (three at Simpson Spring, and four at Vernon Hills) had six parallel 120-m transects (20-m spacing between transects) with six permanent points on each transect. Six of the 35 seed combinations were randomly placed at permanent points along each of five transects and the remaining five were randomly placed along the last transect, and specific seed tray placement was re-randomized over successive trials for the duration of the experiment. Each native species-B. tectorum combination was replicated 15 times in each plot during the period 4 May to 9 September 2005.

Because we were interested in how variation in resource availability affects seed harvest in mixed-species neighbourhoods, we used seed weights rather than seed densities. Seed weight is suitable for assessing changes among seed mixtures in the relative quantities of resources harvested, and it was relative changes rather than absolute numbers that were the focus of this study. This approach has been successfully used in similar studies (Hay & Fuller 1981; Schmidt, Brown & Morgan 1998; Ostoja, Schupp & Klinger 2013).

Species differed in the number of seeds available for a given weight. Mean seed mass of each species was estimated by weighing thirty 25-seed portions of each species, determining the mean and then dividing by 25. Approximate numbers of seeds/gram were as follows: A. hymenoides, 256; B. tectorum, 182; L. cinereus, 169; and P. spicata, 227. Note that B. tectorum is intermediate in seed weight, but nearer the high end of the range. Plastic Petri dishes (14 cm diameter × 1·5 cm deep) were used to offer seeds to rodents. Seed tray experiments may overestimate absolute rates of seed harvest but are unlikely to affect relative intensity of harvest. Trays were set out at sunset and collected at or before sunrise the following morning, denying access to granivorous ants. Remaining seeds were separated by species and re-weighed to determine the amount removed. Twelve trays destroyed or damaged by pronghorn antelope (Antilocapra americana) or wild horse (Equus ferus) trampling were omitted from the analyses.

Statistical Analyses

Density dependence

We used Generalized Linear Mixed Model (GLMM; Zuur et al. 2009) to analyse the relationship between the mean proportion of native seeds removed per night as a function of their weight (‘density’) and of the weight (‘density’) of B. tectorum. We developed an initial set of nine models with different combinations of individual, additive and interactive effects of native seed density, B. tectorum seed density and species identity (A. hymenoides, L. cinereus and P. spicata; Table 1). We then tested random intercept and random intercept + random slope variations of each model, with site as the random factor, producing 17 total models. Akaike Information Criterion (AIC) was derived for each model, from which we calculated ΔAIC and AIC weights to compare the level of support for each model. Analyses were conducted with the NLME package in r (R Development Core Team; http://www.R-project.org).

Table 1. Level of support for 17 models analysing the influence of initial seed density on the proportion of seed removed for three native grass species (Achnatherum hymenoides, Leymus cinereus and Pseudoroegneria spicata) in the Great Basin, Tooele County, Utah from May through September 2005. BRTEden is the density of Bromus tectorum, and Specden is the density of the three native grasses. Intercept indicates a random intercept model and intercept + slope a random intercept and slope model. ΔAIC and wAIC are the difference and the relative weight of Akaike Information Criterion (AIC)
ModelAICΔAICwAIC
BRTEden (intercept)188·60·00·517
Specden (intercept)190·82·20·172
Specden + BRTEden (intercept)191·22·60·141
BRTEden (intercept & slope)192·64·00·070
Specden (intercept & slope)194·86·20·023
Species (intercept)194·86·20·023
Specden + BRTEden (intercept & slope)195·26·60·019
BRTEden + Species (intercept)195·26·60·019
Specden + Species (intercept)197·58·90·006
Specden + BRTEden + Species (intercept)197·89·20·005
BRTEden + Species (intercept & slope)199·210·60·003
Specden + Species (intercept & slope)201·512·90·001
Specden + BRTEden + Species (intercept & slope)201·813·20·001
Specden + BRTEden + Species + Specden*BRTEden + Specden*Species + BRTEden*Species (intercept)240·852·20·000
Specden + BRTEden + Species + Specden*BRTEden + Specden*Species + BRTEden*Species (intercept & slope)244·856·20·000
Specden + BRTEden + Species + Specden*BRTEden + Specden*Species + BRTEden*Species + Specden*BRTEden*Species (intercept)255·266·60·000
Specden + BRTEden + Species + Specden*BRTEden + Specden*Species + BRTEden*Species + Specden*BRTEden*Species (intercept & slope)259·270·60·000

Associational effects

We used multiple linear regression with one categorical predictor (focal species), two continuous predictor variables (native seed weight and B. tectorum seed weight) and their interactions. Two sets of analyses were conducted. First, we analysed the entire data set with native species combined to assess interactions between ‘native seeds’ and B. tectorum. Second, we analysed each native species separately to assess species-specific interactions. In all cases, we assessed the effect of both native seeds and B. tectorum seeds on (i) native seed harvest and (ii) on B. tectorum harvest. Due to blocking by plot and survey replicate and the nature of the study design, this multiple regression approach is embedded in a mixed design (because individual data points are not independent). Graphical analysis of residuals was used to assess assumptions of linearity, normality and homogeneity of variance. MIXED procedure in sas/stat for Windows Release 9.1.2 was used for model fitting (SAS Institute Inc. 2008).

We projected predicted 3-D response in seed harvest surfaces into 2-D (i.e. native and B. tectorum) seed weight planes using topographical isocline plots, where isoclines show all combinations of initial native seed weight and B. tectorum weight that resulted in the same quantity of harvest of the focal species. Isoclines with positive slopes show Associational Resistance; with increasing amounts of a second, associated seed species (x-axis), more of the focal seed species (y-axis) must be initially present for the same amount of that species to be harvested, indicating reduced harvesting of the focal species. Conversely, negative isoclines indicate Associational Resistance; an increase in focal seed harvest occurs with increasing quantities of the second, associated seed species, so that less of the focal species must be initially present for the same amount to be harvested.

Results

Density Dependence of Seed Removal

The model with the individual effect of B. tectorum on proportional native seed removal had the greatest support among the 17 candidate models (Table 1), but models for the individual effects of native species quantity and the additive effects of B. tectorum quantity and native species quantity also had appreciable support. Collectively, 83% of the potential relative support among the candidate models was comprised by these three models (Table 1). We interpret this to indicate that both B. tectorum and native species quantity affected native seed removal, but that B. tectorum quantity had the stronger effect.

Effects of B. tectorum and native species were consistent across species and ‘density’ levels, but the direction of their effects differed (Table 2). Increasing quantities of B. tectorum negatively influenced proportional native seed harvest, while increasing quantities of native seeds positively influenced proportional native seed harvest (Fig. 1). Mean nightly species-specific seed harvest rates ranged from proportions of 0·62 to 0·81.

Table 2. Parameter estimates from a Generalized Linear Mixed Model of the relationship between the mean proportion of seeds removed per night for three native grass species (Achnatherum hymenoides, Leymus cinereus and Pseudoroegneria spicata) and their own density (Specden) and the density of Bromus tectorum (BRTEden) in the Great Basin of Tooele County, Utah, from May through September, 2005. Site (the locations where the trials were conducted) was a random factor in the analysis
VariableParameterSEd.f. t P
  1. Variance from random slope model: Site 0·029 (38·6%).

Intercept0·7390·06857810·850·0000
Specden0·0220·0085782·700·0071
BRTEden−0·0210·0065783·140·0018
Figure 1.

The relationship between the mean proportion of seeds removed for the native species Achnatherum hymenoides (o symbol), Leymus cinereus (Δ symbol) and Pseudoroegneria spicata (+ symbol) and the quantity (gm) of their seeds (a) and the quantity (gm) of Bromus tectorum seeds (b). The experiment was conducted from May through September 2005 in the Great Basin, Tooele County, Utah.

Associational Effects – Native Seed Harvest

Considering intraspecific effects first, in the overall analysis combining all three native species, the weight of native seeds removed increased as the initial amount of native seeds increased (Table 3A); with more seeds available, more were harvested. This is shown by the ‘equal harvest’ isoclines increasing in value up the y-axis for all native species (Fig. 2). These patterns did not differ significantly among native species (Table 3A). In the individual species analyses, native seed quantity also positively affected the harvest of all three native seed species (Table 4, Fig. 2). In general, rodents tended to remove most of the available native seeds, but particularly harvested Pspicata over Lcinereus, and Lcinereus over A. hymenoides, as shown by the isoclines values.

Table 3. MIXED procedure regression results for weight of native seeds harvested (A) and the weight of Bromus tectorum seed harvested (B) as a function of varying B. tectorum and native seed quantities combining the data from all three species of natives. Significant P-values are shown in bold
Effectd.f. F P
(A) Native seed harvested
Species2,430·190·829
Initial native seed weight1,43626·14 <0·001
Initial native seed weight × species2,430·020·983
Initial B. tectorum seed weight1,436·29 0·016
Initial B. tectorum seed weight × species2,430·040·965
Initial native seed weight × initial B. tectorum seed weight1,12090·100·754
Initial native seed weight × initial B. tectorum seed weight × species2,12091·900·149
(B) Bromus tectorum seed harvested
Species2,431·990·149
Initial native seed weight1,4463·86 <0·001
Initial native seed weight × species2,444·85 0·013
Initial B. tectorum seed weight1,42107·62 <0·001
Initial B. tectorum seed weight × species2,420·560·577
Initial native seed weight × initial B. tectorum seed weight1,11959·75 0·002
Initial native seed weight × initial B. tectorum seed weight × species2,11953·17 0·043
Table 4. Results from the MIXED procedure for the weight of native seed harvested by species as a function of the quantity of Bromus tectorum seed and of native seed in the mixture. Significant P-values are shown in bold
Effectd.f. F P
Achnatherum hymenoides seed harvest
 A. hymenoides seed weight1,15273·48 <0·001
 B. tectorum seed weight1,152·560·129
 A. hymenoides seed weight × B. tectorum seed weight1,4150·100·755
Leymus cinereus seed harvest
 L. cinereus seed weight1,14164·67 <0·001
 B. tectorum seed weight1,142·160·164
 L. cinereus seed weight × B. tectorum seed weight1,3990·670·413
Pseudoroegneria spicata seed harvest
 P. spicata seed weight1,13215·66 <0·001
 B. tectorum seed weight1,131·790·203
 P. spicata seed weight × B. tectorum seed weight1,3963·610·058
Figure 2.

Native seed removal for each of the three native seed species as a function of varying native/Bromus tectorum seed densities. The figure does not depict regression lines, but rather topographical isoclines applied to fit a 2-D surface of predicted harvest of species 1 (y-axis) or 2 (x-axis) as a function of changing seed densities of either species 2 or 1, respectively. Each individual isocline indicates all density combinations of the two species that yield a constant level of seed harvest for a given native seed species (y-axis). The positive slopes show that with increasing B. tectorum seed, more native seeds must be initially present to maintain the same level of native harvest, suggesting associational resistance, although the effect was not significant for any species individually.

In contrast to the positive intraspecific effect, when all native species were analysed together, the amount of native seed harvested decreased slightly but significantly as the initial quantity of Btectorum increased (Table 3A), as shown by the shallow positive isocline slopes (Fig. 2). These positive slopes indicate that with increasing quantities of B. tectorum, more native seeds needed to be initially present to have the same quantity harvested. Therefore, B. tectorum provided relatively weak but significant Associational Resistance to the native seeds as a group (Fig. 2). However, in individual species analyses, native seed harvest was not significantly influenced by the amount of B. tectorum in the mixture for any of the three native species (Table 4), further testament to the weakness of the effect.

The near significance of the P. spicata weight × B. tectorum weight interaction (Table 4) suggests that with increasing amounts of P. spicata, B. tectorum provides increasing Associational Resistance to P. spicata harvest, as shown by the increasing steepness of the isoclines as one moves up the y-axis (Fig. 2).

Associational Effects – Bromus tectorum Seed Harvest

Harvest of B. tectorum seeds increased with intraspecific ‘density’ in both the overall analysis (Table 3B) and in each of the individual native species analyses (Table 5, Fig. 3). Although highly significant, the intraspecific effects on B. tectorum harvest did not appear to be as strong as those acting on the native seeds, as indicated by the more gradual increase in isocline values (Fig. 3).

Table 5. Results from the MIXED procedure of Bromus tectorum seed harvested by species as a function of the quantity of Btectorum and of the native seed in the mixture, by native seed species. Significant P-values are shown in bold
Effectd.f. F P
Bromus tectorum seed harvest when present with Achnatherum hymenoides
 A. hymenoides seed weight1,1515·87 <0·001
 B. tectorum seed weight1,1536·48 <0·001
 B. tectorum seed weight × A. hymenoides seed weight1,4090·120·734
Bromus tectorum seed harvest when present with Leymus cinereus
 L. cinereus seed weight1,1441·50 <0·001
 B. tectorum seed weight1,1334·26 <0·001
 B. tectorum seed weight × L. cinereus seed weight1,39312·61 0·004
Bromus tectorum seed harvest when present with Pseudoroegneria spicata
 P. spicata seed weight1,149·72 0·007
 B. tectorum seed weight1,1337·87 <0·001
 B. tectorum seed weight × P. spicata seed weight1,3922·250·134
Figure 3.

Bromus tectorum seed removal when present with each of the three native seed species as a function of varying B. tectorum/native seed densities. The figure does not depict regression lines, but rather topographical isoclines and fit a 2-D surface of predicted harvest of species 1 (y-axis) or 2 (x-axis) as a function of changing seed densities of either species 2 or 1, respectively. Each individual isocline indicates all density combinations of that two-species mixture that yields a constant level of seed harvest for B. tectorum (y-axis). The negative isoclines show that with increasing native seed, less B. tectorum seeds must be initially present to maintain the same level of B. tectorum harvest, indicating Associational Susceptibility, which was significant for all species individually.

The amount of native seed available also had a significant positive effect on B. tectorum seed removal by rodents in both the overall (Table 3B) and the individual species (Table 5) analyses, as shown by the negative slopes of the isoclines (Fig. 3). With increasing native species seed weight, less B. tectorum seeds needed to be initially present to have the same quantity harvested, indicating Associational Susceptibility. Further, in the overall analysis, the initial native seed weight × species interaction was significant (Table 3B), suggesting that although all native species promoted increased B. tectorum harvest, they differed in the strength of the effect, as seen in the different slopes of the isoclines, with L. cinereus having the stronger effect (Fig. 3). The significant initial native seed weight × initial B. tectorum seed weight interaction (Table 3B) reveals that the effects of both native seeds and of B. tectorum seeds on the amount of B. tectorum harvested vary with the amount of the other seed type present. The significant three-way interaction (Table 3B) shows that this relationship varied among the native species. This is supported by the individual analyses where only L. cinereus had a significant interaction between initial native seed weight and initial B. tectorum weight (Table 5), as seen in the more curvilinear isoclines for this species (Fig. 3). The isocline curvature suggests that the effect of increasing native density is very strong at low native seed quantities (steep isoclines slopes), but that with increasingly greater native seed quantities, an equivalent increase in natives leads to less and less effect on B. tectorum harvest (shallow slopes).

In general, more B. tectorum seed was removed when in combination with L. cinereus and Pspicata, than when with A. hymenoides (Fig. 3). More importantly, the consistent pattern of increased B. tectorum harvest when in mixture with native seeds is evidence that native species impose strong Associational Susceptibility on B. tectorum seeds.

Discussion

The effect of granivory on seed survival is an interaction among resource availability and quality (Pearson, Callaway & Maron 2011) and risk of foraging (Randall & King 2001), which may differ across spatial scales (Ostoja, Schupp & Klinger 2013). In this study, we demonstrated both intraspecific and interspecific effects in seed mixtures. The strongest effect detected was an Associational Susceptibility demonstrated with increased harvest of B. tectorum when present with native seeds. We also report a less pronounced result of Associational Resistance with a reduction of native seed harvest when present with B. tectorum seed. Further, our study as well as others (see Orrock, Witter & Reichman 2008; Pearson, Callaway & Maron 2011) demonstrates that non-native plant species can alter natural seed–seed consumer relations, which has direct relevance to conservation and restoration in invaded and disturbed landscapes.

Intraspecific Effects and Density Dependence

While the mixed density models demonstrated more native seeds were harvested when more were initially present, these models cannot assess density dependence as defined, which is why GLMMs were also constructed. With this approach, we demonstrated that when more native seed was available, proportionally more was taken by rodents, as predicted. Although we evaluated seed weight rather than numbers, this reflects intraspecific density dependence, which has been shown in other North American desert systems and in laboratory studies with related rodent species (Price & Heinz 1984; Bowers 1990; Veech 2001; Veech & Jenkins 2005). Foraging behaviours of individual species were not evaluated, but the seed-eating rodent community as a whole demonstrated density-dependent foraging. Density dependence is consistent with optimal foraging theory, where the forager should maximize energy intake per unit time spent foraging (Charnov 1976). Granivorous animals of the eastern Great Basin live in a seed-limited environment. These seeds come in annual pulses that can further intensify competition, which likely influences foraging decisions in the face of potential risk of attack. Thus, increased harvest is expected when increasingly greater quantities of seeds are available, especially for more preferred seeds like the native species used in this study. Interestingly, we did not detect a lower threshold at which the rodents ceased foraging for any of the three native seed species.

We did not assess density dependence of B. tectorum harvest as defined; that is a greater proportion removed as the number initially available increased. However, as with the natives, more B. tectorum seeds were removed when more were initially available, although rodents rarely completely removed the B. tectorum seeds as they more often did with the native seeds. This is compatible with other results indicating lower preference of rodents for Btectorum seed compared to native seeds (e.g. Kelrick et al. 1986).

The low removal of B. tectorum could simply be a function of the very high amounts of B. tectorum seeds relative to native seeds in the seed bank at our sites (S. M. Ostoja, unpublished data). Alternatively, B. tectorum seeds may be less preferred due to their morphology and/or nutritional quality compared to the native seeds. Large and nutritious seeds are attractive to granivores, stimulating harvest, whereas the cost associated with difficult to handle seeds may cause animals to ignore such seeds (Vander Wall 2010). Unlike the native species used, B. tectorum has long awns that would make collection and placement within cheek pouches more difficult and slow, and potentially increase the predation risk associated with B. tectorum seed processing. Considering nutrition, B. tectorum seed harvest by nocturnal rodents ranked fourth of six available seed species even though it had the second highest percentage of soluble carbohydrates (Kelrick et al. 1986). Soluble carbohydrates are thought to predict the relative preference of seeds in sagebrush communities (see Kelrick & MacMahon 1985; Kelrick et al. 1986) because they are a water-efficient energy source and their relative percentage is a good measure of available digestible energy. While the level of soluble carbohydrates may contribute to the preference ranking of B. tectorum, it cannot be the sole explanation; potentially, the positive role of soluble carbohydrates is offset by the increased handling time associated with the long awns.

Associational Effects

This study is one of the few to date clearly demonstrating associational effects on seed harvesting. In the present study, we have clearly shown that the seed neighbourhood influences seed harvest and that in mixtures there are both intraspecific and interspecific effects on harvesting. Associational effects can vary in the strength of the interaction, with results to date ranging from strong (Veech 2000) to nonexistent effects (Hulme & Borelli 1999). In the present study, we document moderate Associational Susceptibility of B. tectorum harvest exerted by native seeds and weak Associational Resistance of native harvest exerted by B. tectorum. The magnitude, consistency and overall impact of Associational effects vary temporally and spatially and may be most significant when direct effects are strong (see White, Wilson & Clarke 2006; Barbosa et al. 2009). Additionally, it seems likely that patterns will vary depending on resource type and consumer species (see Hjältén, Danell & Lundberg 1993), and may be influenced by the patch or spatial scale considered (Hjältén, Danell & Lundberg 1993; Emerson et al. 2012; Ostoja, Schupp & Klinger 2013).

Although B. tectorum is only moderately desirable, it was removed in greater amounts when in patches with native species, even in low quantities, demonstrating Associational Susceptibility. This finding is consistent with our prediction from the second objective. However, the greatest associational effect was found with L. cinereus and not A. hymenoides as we predicted. This is likely due to rodents harvesting more L. cinereus than A. hymenoides, contrary to our expectations based on previous studies.

Associational Susceptibility in the form of shared doom or associational damage has been documented in consumer–resource interaction studies (Wahl & Hay 1995; White & Whitham 2000; Barbosa et al. 2009), although not to the degree of Associational Resistance (see below). However, it may be more common than appreciated in the ecological literature (see White & Whitham 2000). In the forestry literature, during pest outbreaks, nonfavoured host species may be attacked more when in the vicinity of the pests' preferred host plant (Fedde 1964). For example, Populus angustifolia × P. fremontii individuals growing under Acer negundo suffered greater rates of defoliation by cankerworms (Alsophila pometaria) than did individuals growing under conspecifics or in the open (White & Whitham 2000).

As noted, examples of Associational Susceptibility in the literature are scant (see Barbosa et al. 2009) but broadly distributed, being found in marine (Wahl & Hay 1995), terrestrial herbivory (White & Whitham 2000) and granivory studies (Emerson et al. 2012). Interestingly, although infrequently documented overall, it is the associational effect most frequently demonstrated in the few experimental studies addressing associational effects on seed removal. Removal of less-preferred Nothofagus dombeyi seeds increased when present with Austrocedrus chilensis seeds (Caccia, Chaneton & Kitzberger 2006), while less palatable sunflower seeds were harvested more by squirrels (Sciurus spp.) when in mixture with more palatable sunflower seeds (Emerson et al. 2012). More Oryzopsis hymenoides (A. hymenoides) seed was harvested when mixed with any of six different species than when available alone, but the proportion harvested differed depending on the identity of the second seed species (Veech & Jenkins 2005). For example, Panicum seeds promoted greater rates of O. hymenoides harvest than did Stanleya seeds. However, it is difficult to separate density-dependent effects from interspecific effects as an explanation in that study because total seed density doubled in mixture. In the present study controlling for overall density, we clearly demonstrated that Btectorum seed suffered Associational Susceptibility when present with any of the three native seeds used, but the strength and shape of the response differed among the native species. Thus, both the identity of the neighbour and the quantity of the neighbour present are important in terms of the fate of associated seeds.

In contrast to the effects of natives on B. tectorum, the effects of B. tectorum on native seed harvest were weaker, less consistent and in the direction of Associational Resistance, again as predicted. However, as no individual species analysis was significant, our prediction that B. tectorum would have the weakest effect on A. hymenoides was not supported; all species had at best weak effects. In contrast to the weak results in these mixed-species models, the interspecific density-dependence model showed a moderate decrease in the proportion of native seeds harvested with an increase in B. tectorum. Despite a difference in strength, both approaches show that native seeds suffer less harvest in the presence of B. tectorum.

Associational Resistance in the form of plant defence guilds has been well demonstrated in herbivory studies (Atsatt & O'Dowd 1976). Palatable species are frequently subject to less herbivory when in mixture with others than when alone (e.g. Hay 1986; Callaway et al. 2005; Wang et al. 2010). However, documented patterns of Associational Resistance among co-occurring seeds and generalist consumers have rarely been shown. In a study of seed mixture effects on harvester ant foraging on six native species, proportional harvest was generally less when in combination with a second seed species (Ostoja, Schupp & Klinger 2013). Interestingly, the second seed species was B. tectorum, generally less desirable to ants as well as rodents, which exhibited only weak Associational Resistance with rodents in the present study. Similarly, Emerson et al. (2012) found that more palatable sunflower seeds were consumed less when at stations mixed with less palatable sunflower seeds.

Potential Mechanisms of Associational Effects

Beyond documenting the occurrence and conditions under which Associational Resistance or Associational Susceptibility occurs, potential driving mechanisms such as resource palatability, morphology, chemistry and behavioural plasticity should be investigated as well. Here, we briefly discuss potential mechanisms driving our documented associational effects (see Barbosa et al. 2009 for a more thorough review of associational effects in general).

The Associational Resistance B. tectorum exerted on native seeds could be due to native seeds being more difficult to locate within patches containing increasingly more B. tectorum seeds. Protection also may result from reduced foraging efficiency on desirable seeds in diverse seed mixtures, including less desirable seeds or seeds requiring increased handling time before the animal can perceive its relative preference. Our results showing Associational Susceptibility of B. tectorum when present with native seeds are possibly due to increased incidental seed harvest of the less desirable B. tectorum when present with native seeds. Rather than incur the risks associated with highly effective sorting of preferred from less-preferred seeds, consumers may sort seeds with less rigour, collecting some proportion of less desirable seeds in the process.

Implications for Plant Recruitment

Our results demonstrating Associational Resistance and Associational Susceptibility of seed harvest suggest that independent of the overall densities of seeds, how those seeds are distributed across the landscape likely influences the recruitment and dynamics of plant populations and the structure of plant communities (Schupp & Fuentes 1995; Schupp, Jordano & Gómez 2010). At a larger scale, in B. tectorum-invaded sagebrush grasslands, associational effects are expected to lead to greater native grass recruitment and less invasive grass recruitment than expected based simply on the number of seeds available and their germination ecologies. However, the actual impacts on plant recruitment will depend on how the seeds of the different species are distributed across the community as the associational effects occur at the scale of local resource patches.

Non-Native Species Effects

Non-native invasive plant species can promote significant shifts in seed consumer communities (Ostoja & Schupp 2009; Ostoja, Schupp & Sivy 2009; Litt & Steidl 2011), likely profoundly affecting seed survival and plant recruitment. Further, as demonstrated here, invasive plants and the seeds they produce can alter foraging behaviours and patterns of native seed harvest. Although native seeds are preferred in our study system, the presence of the non-native B. tectorum decreased the harvest of native seeds, a positive but subtle effect on natives. The ultimate implications of this probably depend strongly on the ecological context. In invaded sagebrush grasslands where B. tectorum is a minor component, the presence of the non-native could ultimately increase the recruitment of the native grasses. However, in B. tectorum-dominated communities, any such positive effect of increased native species recruitment is likely nullified by competition with B. tectorum, which has vastly greater seed numbers in the seed bank (Humphrey & Schupp 2001).

In contrast, it has been suggested in other studies that invasive plants can displace native plants at least partly by increasing the pressure of native consumers on native plants (Orrock, Witter & Reichman 2008; Pearson, Callaway & Maron 2011). In grasslands of western Montana, USA, seeds of eight native species and four weakly invasive species were all preferred by P. maniculatus over the strongly invasive Centaurea stoebe (Pearson, Callaway & Maron 2011); greater rates of native seed predation may reduce competition on establishing C. stoebe. Consumption of seeds of the California native bunchgrass N. pulchra increased in the presence of the invasive B. nigra, leading to a lack of reestablishment of the native (Orrock, Witter & Reichman 2008). The authors contend this may be due in part to the presumed relatively low preference of B. nigra seeds, although it appears to be primarily driven by established B. nigra altering foraging behaviour. Thus, while limited data suggest seeds of natives may be frequently preferred over those of invasive species, associational effects that may accompany those preferences and alter the outcomes are unknown.

Implications for Restoration

It is common to restore disturbed landscapes by applying large quantities of seed, with the intent to revegetate a site with desirable perennial plants. Associational effects have clear ecological applications for reseeding where granivores are important by lessening the impact of seed consumption (Longland & Ostoja 2013). Although native species were preferred over B. tectorum, B. tectorum weakly reduced native seed harvest. The stronger result of increased B. tectorum harvest with native seeds also has potential applications, but comes with practical limitations. However, our results suggest that increasing native seed survival by developing appropriate seed mixtures is a possibility. While B. tectorum would not be used in a restoration context, other seed species (‘diversionary seed’ see Longland & Ostoja 2013) that might also promote greater desirable seed survival should be explored experimentally. For example, Longland & Ostoja (2013) found lower A. hymenoides seed cache recovery at sites where a diversionary seed, millet (Panicum miliaceum), was applied. With increased understanding of the importance of seed mixture contexts for seed harvesting, the consideration of which species to put in seed mixtures and the relative amounts of each could become part of the decision when developing seed mixes for restoration. If seed mixes can be developed that fulfil traditional goals (e.g. rapid establishment, competitive ability) while reducing rodent harvesting of more desirable species, restoration could become easier.

Acknowledgements

Thanks to R. Barker, J. Burnham, D. Christensen and K. Sivy for field and/or laboratory assistance. K. Beard, C. Call, T. Esque, W. Longland, J. MacMahon, T. Monaco and an anonymous reviewer provided helpful comments. This research was supported by the Cooperative State Research, Education, and Extension Service, U.S. Department of Agriculture, under Agreement No. 2001-52103-11322; the Utah Agricultural Experiment Station (UAES); Utah State University (USU), Logan, UT, USA; a USU Ecology Center research fellowship (to SMO); and a USU School of Graduate Studies dissertation fellowship (to SMO). Any use of trade, product or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. government. Approved as UAES journal paper no. 8496.

Ancillary