Assembly history alters alpha and beta diversity, exotic–native proportions and functioning of restored prairie plant communities


Correspondence. E-mail


  1. Restorations provide a test of community assembly history theory, and practitioners require information on how assembly history might help to restore diverse native species communities. Variation in community assembly history (historical order of species arrival) is hypothesized to generate beta diversity by producing alternate states, but restorations are hindered because there have been few tests using long-term field experiments.
  2. We experimentally altered assembly history of native species into formerly exotic-dominated grassland sites while simultaneously removing dispersal limitation to test whether alternate states or a single equilibrium would develop and whether alternate states would generate varying ecosystem-level effects. Assembly history was altered by varying the identity of early-emerging species, timing of seed additions after disturbance (early-emerging species added in spring or summer) and priority effects (common 30-species seed mixture added either at the same time or after an early-emerging species canopy developed). The experiment was conducted at two sites that differed in productivity.
  3. Altering timing and priority effects during assembly history had large effects on species composition and diversity. On average, diversity was highest, and the proportion of exotic species was lowest in plots seeded in spring and without priority effects. Identity of early-emerging species did not significantly affect community structure.
  4. Differences in species composition affected fuel mass, fire temperatures and peak above-ground primary productivity, key ecosystem processes in tallgrass prairie.
  5. Synthesis and applications. Our results indicate that grassland communities can reach alternate exotic- or native-dominated states in uniform environments when perennial exotic species are present. These states were strongly affected by timing of native species arrival and priority effects. Thus, assembly history is a key process that can give rise to beta diversity, and our results suggest that native species should be established early in the restoration process before exotics become fully established.


Beta diversity, or turnover in species composition across space, is an important component of diversity (Wilsey, Martin & Polley 2005), yet the relative importance of processes generating turnover is poorly understood. Without a better understanding of these processes, our ability to restore beta diversity is hindered. Variation in abiotic conditions, dispersal limitation and community assembly history are main hypotheses for how communities differentiate (Belyea & Lancaster 1999; Chase 2003). First, species may sort deterministically according to variation in abiotic conditions (e.g. similar soil types or climate), consistent with an abiotic niche model (Whittaker 1960). This hypothesis implies that species composition should reach a single stable equilibrium within a uniform environment but exhibit high beta diversity among environments (Chase 2003). Observational studies suggest, however, that much variation in species composition is left unexplained after relating it to abiotic conditions (e.g. Condit et al. 2002; Legendre et al. 2009). Second, differences in dispersal among species, where some species fail to reach all suitable locations, may generate beta diversity. With unlimited dispersal and similar abiotic conditions, all species would be found everywhere, and beta diversity would approach zero (Mouquet & Loreau 2003). Finally, assembly history theory predicts that if the historical order of species arrival differs, then multiple stable equilibria can arise within uniform environments, thus generating high beta diversity (Drake 1991; Chase 2003). Although abiotic filters, dispersal processes and species arrival order are often combined into the general term, ‘community assembly’, here, we use ‘community assembly history’ to refer solely to the historical order of species arrival. Seasonal differences in abiotic conditions during establishment, identity of earlier arriving species and variations in arrival order could lead to variation in species composition within an area. Chase (2003) hypothesized that a single stable equilibrium is more likely to arise when the regional species pool is small, dispersal and disturbance rates are high, and local productivity is low. Alternatively, multiple stable equilibria are hypothesized to develop within a uniform environment when the regional species pool is large, dispersal and disturbance rates are low, and local productivity is high (Chase 2003).

Community assembly history is poorly understood because long-term field experiments incorporating many species are few in number (Anderson et al. 2011), and this is particularly true for plant communities (but see e.g. Fargione, Brown & Tilman 2003; Fukami et al. 2005; Collinge & Ray 2009; Myers & Harms 2011). Here, we test whether the identity of early-emerging species, the timing of seed addition and priority effects alter the outcomes of tallgrass prairie restorations in uniform environments with high exotic species propagule pressure. Differences in the identity of early-emerging species could increase beta diversity if species permit establishment from other functional groups more readily than their own during the assembly history process (Diamond 1975; Fox 1987; Gotelli & McCabe 2002; Fargione, Brown & Tilman 2003). The seasonal timing of disturbances by ungulates or by digging fossorial mammals could enhance beta diversity if species exploit disturbances at different times of the year according to temporal niche requirements (Questad & Foster 2008). Bazzaz (1996) hypothesized that early species may alter local abiotic conditions and affect subsequent community composition. For example, spring disturbance could favour C4 plant species because soil temperatures are increasing, but summer disturbance could favour C3 species because temperatures are decreasing (Sage & Monson 1999). Priority effects occur when an early-establishing species is able to attain large size before its competitors arrive, which enables it to out-compete later-arriving species (Alford & Wilbur 1985). Priority effects could therefore alter eventual species composition and ecosystem processes. Alternatively, competition–colonization theory predicts that a species with the highest competitive ability will eventually out-compete those with a better colonizing ability, despite differences in priority effects or timing of disturbance (Tilman 1994). If competition–colonization processes dominate, then it would not matter when propagules are added to the system during a restoration. However, if the assembly history processes described above are more important, then the order of introduction is important.

Evidence is emerging that priority effects differ between native and exotic species, and this could be important in human-altered systems where novel, exotic-dominated systems can exist as alternate states to native systems (Christian & Wilson 1999; Hobbs et al. 2006; Wilsey, Daneshgar & Polley 2011). Exotic species have been shown to green up several weeks before comparable native species and to have stronger priority effects than natives (Marushia, Cadotte & Holt 2010; Wilsey, Daneshgar & Polley 2011; Wainwright, Wolkovich & Cleland 2012; Dickson, Hopwood & Wilsey 2012). Meta-analyses showed that above-ground growth rates are higher among exotic than native species (Leishman et al. 2007; van Kleunen, Weber & Fischer 2010). These results imply that fundamentally different communities, or alternate states, may develop if exotics establish before natives, and this is important in restorations if the goal is to maximize native species diversity (Martin, Moloney & Wilsey 2005).

We used a long-term field experiment to test the importance of assembly history in generating beta diversity. We varied assembly history of native prairie species into systems with high exotic propagule pressure by altering identity of early-emerging species, timing of seed additions after disturbance (spring or summer) and priority effects (diverse prairie mix added with early-emerging species or the following spring) in more than one environment. These factors have not been varied simultaneously before. Manipulations took place at an intermediate scale between small-scale disturbances (e.g. gopher mounds, ant hills) and field-size restoration in grasslands. We measured species composition and diversity to assess community responses to treatments. We also assessed fire temperature, fuel mass and above-ground net primary productivity to determine whether community changes had ecosystem-level ramifications that are important in grasslands. We predicted that if aspects of community assembly history were important, generation of multiple states would arise by altering (i) identity of early-emerging species, which would allow members of other functional groups into the plots more readily than their own; (ii) timing of disturbance and seed additions, where spring timing would favour C4 establishment and summer timing would favour C3 establishment; and (iii) priority effects, where adding a diverse native prairie mixture with either early-emerging species or after they established would alter species composition. Alternatively, if abiotic site conditions, dispersal limitation and competition–colonization mechanisms were most important, then the historical order of species arrival would not affect communities. We found that timing and priority effects can generate alternate states of native- and exotic-dominated communities and that these states have ecosystem-level consequences.

Materials and methods

Study Sites

The experiment was conducted on two separate Iowa State University-owned field sites that differed in abiotic conditions and above-ground net primary productivity in Iowa, USA. Conducting the experiment in two sites allowed us to test whether assembly history processes were stronger in the more productive site as predicted by Chase (2003), although this is an unreplicated test. Previous measurements verified that the central site (Horticulture Research Station, hereafter HRS) is approximately two times more productive than the western site (Western Research Farm, hereafter WRF) in a separate study at the same sites (Blong 2007). The WRF is located in the Iowa Loess Hills region (lat. 42°03′N, long. 95°49′W) and is more xeric relative to HRS, which is located in the Des Moines Lobe (lat. 42°6′N, long. 93°35′W). The 30-year annual precipitation average prior to the study was 777 ± 201 mm for WRF and 855 ± 208 mm for HRS. Both sites were within normal ranges during 2005 and 2006, the establishment years of the experiment. WRF contains Ida silt loam, well-drained calcareous loess, with 14–20% slopes and 2·5% organic matter, while HRS contains Storden loam soils with 9–14% slopes and 2·5–3·5% organic matter. Plots at both sites were located on abandoned pasture dominated by the perennial exotic species Bromus inermis Leysser (nomenclature follows Eilers & Roosa 1994) and were established by disking the area to bare ground twice in early spring 2005 prior to adding treatments. Plots had no B. inermis present at the beginning of the study, but it reestablished from seed in the second year. A previous study at the same two sites found that the seed bank was predominately exotic in origin (Blong 2007), thus allowing a test of hypotheses under high exotic propagule pressure.

Experimental Design

We used a split-plot experiment to test the influence of three community assembly history factors. Factors included variation in the identity of early-emerging species, timing of seed addition following disturbance and priority effects. Six treatments were randomly assigned to 5 × 5 m main plots with five replicates at each site to alter the identities of early-emerging tallgrass prairie species (Fig. 1). Species were selected because they establish early compared to other members of their functional groups. Main plot (‘identity’) treatments included the C3 annual Chamaecrista fasciculata (Michx.) Greene, C3 perennial Elymus canadensis L., C3 biennial Rudbeckia hirta L., C4 perennial Bouteloua curtipendula (Michx.) Torrey, a mixture of all four species, and a control (no early-emerging species). Main plot identity treatments were seeded at a rate of 11·5 kg ha−1. We randomly assigned subplot treatments of seed timing and priority effects in a 2 × 2 factorial design to four, 2 × 2 m subplots (with 1-m wide alleyways) within each main plot (Fig. 1). The timing treatment simulated differences in species propagule arrival to bare ground following disturbance. Timing was altered by adding seeds either in April 2005 (spring timing) after initial disking of the entire area or in August 2005 (summer timing) after removing negligible amounts of biomass that established to ground level using a gas-powered, hand-held string trimmer. Plants that were trimmed did not regrow, and thus, seeds were added to bare ground in both spring and summer. Priority effects were altered by adding a common 30-species prairie seed mixture (hereafter ‘prairie mix’) simultaneously with early-emerging species in either 2005 (no priority effect) or a year later in the following spring (April) 2006 (with priority effect). Thus, plots with priority effects allowed time for early-emerging species (or volunteers in control plots) to establish before the prairie mix was added. The 30-species prairie mix, which included 32 seeds per species of tallgrass prairie forbs and grasses totalling 960 seeds per subplot, was broadcast by hand to all subplots (see Table S1 in Supporting Information). There were six main plot treatments with five replicates across two sites giving 60 plots in total and four subplots in each of the 60 main plots for 240 subplots in total. However, one main plot at each site was removed from all analyses because of accidental mowing or incorrect seeding, which reduced main plots to = 58 and subplots to = 232.

Figure 1.

Schematic of assembly history treatments at a mesic site, HRS and a dry site, WRF. Main plots (5 × 5 m) with six early-emerging species identity treatments (labelled above main plots) and subplot treatments (2 × 2 m with 1-m alleys) with altered timing and priority effects: spring timing–no priority (Sp-NP), early-emerging species added in spring with the 30-species prairie mix; spring timing–with priority (Sp-P), early-emerging species added in spring with the 30-species prairie mix added the following year; summer timing–no priority (Su-NP), early-emerging species added in summer with the 30-species prairie mix; summer timing–with priority (Su-P), early-emerging species added in summer with the 30-species prairie mix added the following year.

Priority effect treatments are confounded with time to varying degrees in all priority effect studies. This is because when seeds are added before or after a canopy of other species develops, they are also added at two different time periods, which could vary in granivore activity, rainfall or temperature. However, we hypothesize that this led to minimal confounding in the current study for three reasons. First, there was very little recruitment from the prairie mix during the time period in question (i.e. recruitment occurred in 2006 and possibly afterwards). Second, there was little difference in temperature and precipitation between 2005 and 2006 in this study. Finally, similarity in results at both sites (as we found) would suggest that any confounding factor, such as granivore activity or rainfall, would also need to be similar at both sites. We suggest that this is unlikely because sites are located 185 km apart in different geologic regions. Thus, our treatments should be interpreted as primarily a priority effect, but other potentially important factors might have also differed between the time periods when seedlings established.

Plots were not weeded to allow communities to realistically assemble in a restoration context. Plots were immediately surrounded by fields dominated by exotic pasture grasses in a larger agricultural matrix (corn and soybean fields) and were separated from native, unploughed remnant grasslands by 5 and 13 km at HRS and WRF, respectively. Consequently, we assumed that prairie species establishment was solely from the seeded prairie mix.

Sampling Design

Species diversity, proportion of exotic species and establishment from the prairie mix were estimated in each subplot using abundance data from a non-destructive point-intercept sampling technique, which is highly correlated with biomass (Jonasson 1988; Wilsey, Daneshgar & Polley 2011). At the end of 2005, spring timing plots were dominated by annual species that were not present in the prairie mix (HRS 82% and WRF 89% annual species, based on cover estimates) and that did not persist in appreciable quantities. Establishment of species from the prairie mix did not begin until the 2006 growing season in all plots. Therefore, sampling was conducted in July 2006–2008 and 2010, the second to fourth and sixth years of the experiment. Species abundances were estimated by dropping 1-m-long point-intercept pins 24 times in a systematic fashion within a 50 × 100 cm point-intercept frame placed in the centre of each subplot and counting the number of times pins hit each species, allowing multiple hits per species. Points were raised to contact plants >1 m tall in a few cases. Plant species present in the frame that were not hit were given a value of 0·5 hits (Wilsey, Daneshgar & Polley 2011). Species diversity was calculated using Simpson's diversity (inline image). Species were designated as either native or exotic (Eilers & Roosa 1994), and the proportion of exotic species was estimated by dividing the number of exotic species hits by total number of hits. Establishment of species from the prairie mix was estimated by summing total number of hits from the mix. Early-emerging species were omitted from response variables because we were interested only in responses of developing communities.

Species establishing early may alter abiotic conditions for future species, providing a mechanism to explain long-term community structure (Bazzaz 1996). We measured litter mass and soil water at both sites and soil temperature at HRS in subplots in June 2006 to test this hypothesis when plant seedling establishment was highest. Litter was collected in a 20 × 50 cm quadrat in one random location in each subplot, dried and weighed. Soil water was estimated gravimetrically [(wet mass − dry mass)/dry mass] by taking one, 2·5-cm-diameter soil core to 10 cm depth from each subplot and drying cores at 105 oC. Soil temperature was estimated using a Hobo® data logger (Onset Computer Corporation, Bourne, MA, USA) and thermocouples.

Ecosystem-Level Effects

Fuel mass, fire temperatures and peak above-ground primary productivity were estimated in each subplot to determine whether assembly history altered ecosystem-level effects among communities. Fuel mass was estimated by clipping biomass in a 20 × 50 cm quadrat prior to burning all plots in spring, 2008. Fire temperatures were estimated by placing ceramic tiles painted with Omegalaq® Liquid Temperature Lacquers (Omega Engineering Inc., Stamford, CT, USA) on the ground in each subplot prior to burning and comparing developed colours to a standard. Fire temperature results were qualitatively similar to results from thermocouples (D. Schwilk, unpublished data). Peak above-ground biomass was estimated by clipping biomass in a 20 × 50 cm quadrat in late August or early September 2010, the sixth year of establishment. Green biomass was dried at 65 °C for 48 h to constant mass and weighed.

Statistical Analyses

Response variables measured during 2006–2008 and 2010 were analysed using split-plot repeated measures anova. All plots at HRS were burned, and the original seed mixture was added to half the main plots in spring, 2010 for an additional study not described here, but these treated plots did not differ from untreated plots in 2010 (seeded versus non-seeded contrast: diversity, abundance of prairie mix and proportion of exotics all F1,17 < 1·00, > 0·30), so seeded and unseeded plots were combined for analyses. Site and early-emerging species identity were tested with the main plot error term, and timing and priority effect treatments and interactions were tested using the subplot error term using type III sums of squares in Proc Mixed in SAS (Littell, Stroup & Freund 2002). A priori contrasts were conducted on timing (spring vs. summer) and priority effects (with vs. without) and timing x priority interactions. Split-plot anova was also used to analyse litter, soil water, soil temperature, fuel mass, fire temperature and ANPP. Proportion of exotics was logit-transformed (log(p+ɛ/1-p+ɛ), where p is proportion of exotics, and ɛ is the lowest nonzero proportion, to improve normality prior to analysis (Warton & Hui 2011). Abundance from the mix contained many zeros in 2006 because establishment was low in some treatments; therefore, it was analysed without the 2006 data, and remaining data were (ln + 1)-transformed prior to analyses to reduce heteroscedasticity and to improve normality. Data from 2006 are presented visually. Litter, fuel mass and ANPP were ln-transformed prior to analyses to reduce heteroscedasticity and to improve normality.

We used multivariate tests to assess overall variation in community composition (i.e. beta diversity) among treatments (Anderson et al. 2011). We omitted early-emerging species and used Bray–Curtis dissimilarity, which excludes joint absences, on species relative abundances for all multivariate analyses. We used nonparametric permutation-based manova (permanova) to test for overall community composition differences between main and subplot treatments using the adonis function in package vegan version 2.0-1 in R, which uses type I sums of squares (Anderson 2001), using data averaged across years to test for differences. Main plot treatment differences also used data averaged across subplots within each main plot. We included site and early-emerging species treatments with all interactions in the subplot model, and site and interactions with early-emerging species in the main plot model so that multivariate and univariate tests would use the same error d.f. Data were permuted 999 times within each main plot for subplot tests and within each site for main plot tests. Finally, if permanova results were significant, we conducted NMDS analysis using the metaMDS function in vegan version 2.0-1 in R to visualize variation in species composition among treatments at each site for each year of sampling (Anderson et al. 2011).


HRS had higher diversity and proportion of natives than WRF (diversity 3·2 at HRS, 2·5 at WRF, SE = 0·09; exotics 0·60 at HRS, 0·69 at WRF, SE = 0·01) (Table 1). Identity of early-emerging species did not significantly affect any of the community response variables (Table 1).

Table 1. anova results (F, (P)) for community assembly history experiment effects on diversity (Simpson's 1/D), proportion of exotic species abundance and abundance from the 30-species prairie mixture. Values significant at the α ≤ 0·05 level are in bold
Sourced.f.aDiversityProportion exoticPrairie mix
  1. a

    Different d.f. for prairie mix are reported in parentheses because of removing the first year (2006) of data because of the large number of zero values.

Site (S)1 30·66 (<0·01) 19·12 (<0·01) 2·41 (0·13)
History (Identity) (HI)51·64 (0·17)2·23 (0·07)1·63 (0·17)
S × HI52·35 (0·06)2·01 (0·10)1·21 (0·32)
Error (Plot × S × HI)46   
History (Timing and Priority) (HTP)3 14·33 (<0·01) 81·40 (<0·01) 108·69 (<0·01)
S × HTP3 3·30 (0·02) 1·14 (0·34)1·51 (0·21)
HI × HTP150·53 (0·92)1·02 (0·44)1·39 (0·16)
S × HI × HTP150·47 (0·95)1·09 (0·37)1·62 (0·08)
Error (Plot × S × HI × HTP)138   
Time3 (2) 44·56 (<0·01) 123·69 (<0·01) 40·64 (<0·01)
S × Time3 (2)1·02 (0·39) 17·30 (<0·01) 13·82 (<0·01)
HI × Time15 (10)1·63 (0·06) 2·54 (<0·01) 1·74 (0·07)
S × HI × Time15 (10)1·00 (0·46)1·09 (0·36)0·76 (0·67)
HTP × Time9 (6) 5·22 (<0·01) 17·11 (<0·01) 2·26 (0·04)
S × HTP × Time9 (6) 2·13 (0·03) 3·60 (<0·01) 6·70 (<0·01)
HI × HTP × Time45 (30)0·83 (0·77)0·86 (0·73)1·02 (0·43)
S × HI × HTP × Time45 (30)0·87 (0·72)0·67 (0·95)0·79 (0·78)
Error (Plot × S × HI × HTP × Time)552 (368)   
Priority123·00 (<0·01)82·57 (<0·01)137·07 (<0·01)
Timing10·11 (0·74)64·59 (<0·01)74·11 (<0·01)
Timing × Priority119·88 (<0·01)97·04 (<0·01)114·90 (<0·01)
Error (Plot × S × HI × HTP)138   

Diversity, relative abundance from the prairie mix and proportion of exotics were all significantly affected by timing and priority effects. Interactions between these two effects produced differences that were stronger than between sites or identity treatments. Removing dispersal limitation did not lead to convergent species compositions, and communities diverged similarly among assembly history treatments at both sites. Abundance from the prairie mix was four times higher and proportion of exotics was 20% lower in spring than summer timing treatments. All variables were strongly affected by priority effects, with diversity and abundance from the prairie mix being 16% and 68% lower, respectively, and proportion of exotics being 20% higher when the prairie mix was added later than early-emerging species (i.e. with priority effects) (Table 1, Fig. 2). Timing and priority effects interacted, with diversity being 37% higher and relative abundance of species from the prairie mix being six times higher in the spring timing–no priority effect treatment compared to all other treatments by the end of year six (Table 1, Fig. 2). The spring timing–no priority effect treatment was 47% native by year six, whereas other treatments were dominated by exotics (82–86% exotic) (Table 1, Fig. 2). All treatments were dominated by perennials (>94%), and the proportion of C3 biomass was 35% and 26% lower in spring timing–no priority effect treatments compared to other treatments at WRF and HRS, respectively, by the end of year six (see Table S2).

Figure 2.

Simpson's diversity at a mesic site, HRS (a), a dry site, WRF (b), abundance from the 30-species prairie mixture at HRS (c) and WRF (d) and proportion of exotic abundance at HRS (e) and WRF (f) in community assembly history treatments (LS means ± SE).

Species composition varied among timing and priority effect treatments (permanova F3,138 = 22·2, < 0·01), but early-emerging species did not result in changes to species composition (F5,46 = 0·88, = 0·58). Ordinations revealed that the spring timing–no priority effect treatment diverged sharply from other treatments (Fig. 3).

Figure 3.

Non-metric multidimensional scaling ordinations using Bray–Curtis dissimilarity on species relative abundances from the prairie mix and invaders, showing the four assembly history treatments at HRS and WRF in 2006 (a, e), 2007 (b, f), 2008 (c, g) and 2010 (d, h). Points represent spring timing–priority (black circles), spring timing–no priority (red circles), summer timing–priority (inverted green triangles) and summer timing–no priority (yellow triangles) plots.

Abiotic Conditions

HRS had higher litter and soil moisture than WRF at the beginning of 2006 (litter F1,46 = 76·96, < 0·01; soil water F1,46 = 84·86, < 0·01), but identity treatments had no significant effect on abiotic variables (litter F5,46 = 0·69, = 0·63; soil water F5,46 = 1·17, = 0·34; soil temperature F5,23 = 0·79, = 0·57). Timing treatments altered all abiotic variables at the beginning of 2006 (Fig. 4). Litter was much higher in spring than summer timing treatments at HRS and WRF (timing contrast F1,138 = 309·8, < 0·01, site x timing and priority interaction F1,138 = 3·42, = 0·02). Litter was not significantly affected by priority effects (F1,138 = 0·24, = 0·62) on average, but litter was increased slightly by priority effects in spring timing treatments (timing x priority interaction = 6·7, = 0·01). Soil water content was 29% higher in spring than summer timing treatments overall, (timing contrast F1,138 = 94·64, < 0·01), and this was largely driven by HRS, which had 49% higher soil moisture in spring than summer (site x timing and priority interaction F1,138 = 16·31, < 0·01). Soil moisture was not affected by priority effects or interactions (priority F1,138 = 1·49, = 0·22, timing x priority F1,138 = 0·05, = 0·83). Average soil temperature at HRS was 11% lower in spring than summer timing treatments (F1,69 = 112·91, < 0·01) and was not affected by other treatments (priority F1,69 = 0·40, = 0·53, timing x priority F1,69 = 1·40, = 0·24).

Figure 4.

Litter (a) and soil water content at HRS and WRF (b) and soil temperature at HRS in the second year (c) and ecosystem processes at both sites including fuel mass (d) and fire temperatures in the fourth year (e) and peak above–ground biomass in the sixth year (f) of the community assembly history experiment (LS Means ± SE). ‘Spring’ and ‘Summer’ refer to the timing treatments.

Ecosystem-Level Effects

Timing and priority effects significantly altered ecosystem variables consistently at both sites (all site × timing and priority interactions F < 1·9, all  0·15). None of these were affected by identity treatments (fuel mass F5,46 = 0·68, = 0·64, fire temperature F5,45 = 1·28, = 0·29, peak ANPP F5,46 = 0·59, = 0·71). Fuel mass and fire temperatures were 34% and 19% higher, respectively, in spring timing treatments (fuel mass F1,134 = 47·8, < 0·01; temperature F1,134 = 50·4, < 0·01, Fig. 4). The spring timing–no priority effect treatments had 51% higher peak ANPP than other treatments (timing × priority F1,138 = 12·35, < 0·01, Fig. 4). Peak ANPP was 47% lower at WRF than HRS (F1,46 = 18·65, < 0·01).


Community assembly history in our long-term field experiment generated alternate states characterized by considerable differences in diversity and native–exotic proportions, and consequently in ecosystem processes. Long-lived perennial species dominated all communities, suggesting that a future disturbance or death of long-lived individuals would be required to reinitiate the assembly history process. We found higher biodiversity overall at the more productive site, similar to Chase (2010). However, in contrast to predictions by Chase (2003), we found that alternate states were just as likely to develop in the site with low productivity. Our sites with different productivity levels were not replicated, so we were unable to determine whether the interesting trend we found with alternate states and productivity was general. Therefore, we suggest that further tests of this hypothesis should be conducted across replicated sites. Wilsey (2010) reported in a forum article on diversity partitioning that beta diversity was higher among assembly history treatments within sites than between the two sites in this experiment. We found here that the high beta diversity developed early, and it was caused by interacting timing and priority effects, with the spring timing–no priority effect treatment having vastly lower proportions of exotic species and greater species diversity on average. Although priority effects can be confounded with year in priority effect studies, these effects were minimal in the current study because virtually no recruitment from the seed mix was found until the second year, and there were similarities in precipitation and temperature between years. After communities diverged, they remained significantly different through five growing seasons, and these differences affected fuel mass, fire temperatures and ANPP. Altering identity of early-emerging species had no significant effect on any of the community or ecosystem variables. Our results suggest that identity of the early invader during community assembly history may not be as important as timing of species arrival and priority effects in altering species composition.

We predicted that timing of disturbance in subplots (creating bare ground in spring or summer) would differentially affect C3–C4 proportions, but our results suggest that exotic invasions complicated matters when exotic propagule pressure was high as it is in our system. Specifically, summer disturbance and concurrent seeding favoured C3 species as predicted. Disturbance and seeding in spring favoured C4 species, but only when they established before C3 species (which were mostly exotic). When C4 species did not establish first, C3 species eventually dominated communities. Overall, the percentage of C4 species was much higher in spring timing plots when the prairie mix was added simultaneously compared to all other treatments. Our results are consistent with Dickson et al. (2010), who found that an exotic legume was able to out-compete native perennial C4 prairie grasses because it established earlier. If exotic species green up earlier, have stronger priority effects and have higher above-ground growth rates than natives (Leishman et al. 2007; Wilsey, Daneshgar & Polley 2011; Dickson, Hopwood & Wilsey 2012), then exotics may have large and underappreciated consequences for predicting community assembly history outcomes (Wolkovich & Cleland 2011).

Variation in abiotic conditions between spring and summer timing treatments suggests that early invaders can modify their environments and contribute to community divergence. Localized species–environment feedbacks were more important than site conditions in generating community structure. Polley et al. (2006) found that early-season annuals regulated perennial plant composition and ecosystem functioning in grasslands. In our study, spring disturbance and seeding favoured early invasion of annual volunteer species. These annuals produced higher litter mass, which generated cool, moist soil conditions favourable for prairie species to grow early in the second year of the study, when most establishment from prairie species seeded without a priority effect occurred. Conversely, summer seeding led to fewer annuals, less litter mass and warmer, drier soil conditions. These conditions, along with priority effect treatments, favoured early establishment and resulting dominance of exotic perennial C3 species and the suppression of native prairie species.

Our results provide stronger support for assembly history than for competition–colonization mechanisms in structuring restored tallgrass prairie communities in areas of high exotic propagule pressure (Dickson et al. 2010). Competition–colonization theory predicts that a single equilibrium will develop as long as propagules of a competitive species are available. Native perennial C4 grass species have some of the highest competitive abilities (Wedin & Tilman 1993), yet these species established in only one assembly history treatment despite being seeded in all plots. Pacala & Rees (1998) suggest, as an alternative model to competition–colonization trade-offs, that niche differences permit early-colonizing species to establish and maintain dominance even when the better competitor is present. Miles & Knops (2009) observed that exotic C3 perennials persisted over very long time periods in the presence of competitive native species. In our study, perennial exotic species established early in all treatments, and they continued to dominate plots when they established fully before natives.

The community differences that we found resulted in important long-term effects on ecosystem variables. Greater fuel mass and higher fire temperatures were found in spring timing treatments, and these variables are important to functioning (e.g. Stronach & McNaughton 1989). Above-ground productivity was higher in the more diverse plots containing a greater proportion of native species that resulted from the assembly history treatments. Our results support the few existing studies (Foster & Dickson 2004; Körner et al. 2008; Fukami et al. 2010; Dickson et al. 2010) that found that community assembly history has ecosystem-level ramifications.

Community assembly history theory can link a range of questions in community, invasion and restoration ecology (Weiher & Keddy 1999; Suding, Gross & Houseman 2004; Temperton et al. 2004), and it is an underappreciated mechanism contributing to beta diversity. We found that alternate states of perennials can occur by altering assembly history. These states persisted into the sixth year, and only longer-term monitoring will determine whether they are transient or stable (Fukami & Nakajima 2011). For example, Collinge & Ray (2009) found that strong priority effects of vernal pool communities diminished after four to five years and suggested that assembly may be transient. Furthermore, our results imply that generation of beta diversity in restorations may be undesirable if it results in exotic dominance in some areas. We suggest that establishing native species before exotics is crucial for restoring diverse native prairie communities in situations where perennial exotics are present and that using cover crops (early-emerging species in our study) may not be as beneficial as expected (Padilla & Pugnaire 2006) in less stressful environments.


Our results indicate that variation in species composition among communities that appears random and idiosyncratic in field conditions may in fact be the predictable result of historical assembly processes (Drake 1991). Beta diversity arose when disturbance and priority effects were altered during assembly history in the field, and communities of perennial exotic versus native species established deterministically, not stochastically. Further work on how timing of disturbance and priority effects alter native–exotic proportions, especially in the face of global climate change and other human impacts, will help us develop an understanding of when communities will assemble into diverse native communities and when they will collapse into low-diversity communities of exotics. Our management recommendation is for restoration projects to establish native species as early as possible when in situations with high exotic species abundance.


We thank Adam Asche, Dylan Schwilk, Chris Steege, Chris Swenson and Jenny Richter for their assistance with data collection. Andrea Blong and Kim Wahl helped to establish and maintain the experiment. Tim Dickson, Wayne Polley, Lauren Sullivan and two anonymous reviewers provided helpful comments on an earlier version of this manuscript. The Iowa Department of Transportation LRTF provided funding.