Journal list menu

Volume 8, Issue 10 e01987
Open Access

Between invaders and a risky place: Exotic grasses alter demographic tradeoffs of native forb germination timing

Diane M. Thomson

Corresponding Author

Diane M. Thomson

W. M. Keck Science Department, The Claremont Colleges, 925 N. Mills Avenue, Claremont, California, 91711 USA

E-mail: [email protected]Search for more papers by this author
Rachel A. King

Rachel A. King

W. M. Keck Science Department, The Claremont Colleges, 925 N. Mills Avenue, Claremont, California, 91711 USA

Search for more papers by this author
Emily L. Schultz

Emily L. Schultz

W. M. Keck Science Department, The Claremont Colleges, 925 N. Mills Avenue, Claremont, California, 91711 USA

Search for more papers by this author
First published: 30 October 2017
Citations: 12
Corresponding Editor: Michael P. Perring.


Priority effects are hypothesized to play an important role in community assembly and may promote suppression of native by exotic species. Work in a range of grassland systems has proved valuable for testing these effects, demonstrating that earlier germination by some exotic annual grasses contributes to their competitive dominance over natives. Yet few studies have measured native forb germination phenology under field conditions, and the demographic consequences of emergence timing for competitive interactions and native fitness are not well understood. We focused on three native annual species in a southern California grassland dominated by exotic Bromus spp. over three years, measuring (1) seedling emergence rates, for both early (October) and later (November and December) germinators; (2) effects of exotic grasses on native survival and reproduction, through a grass removal experiment; and (3) interactions between emergence timing and grass competitive effects on native mortality, survival, and flowering. We quantified tradeoffs of emergence timing, by estimating mortality experienced by early germinants until the late cohort emerged (early survival), and then for both cohorts from the time of late emergence to flowering (spring survival). The two most common focal natives, Amsinckia intermedia and Phacelia distans, varied substantially in germination phenology but primarily emerged early. The less abundant Clarkia purpurea germinated late. Late emergence reduced spring survival in control plots but not those where exotic grasses were reduced experimentally, supporting the importance of priority effects and benefits of early germination in competition with grasses. However, early emergence entailed a high cost of initial mortality risk in some years. We found no effect of emergence timing on size at flowering. Estimates of net survivorship to flowering suggest that late emergence consistently was associated with the highest survival when exotic grasses were reduced experimentally. Early emergence was more favored in control than in exotic grass reduction plots, but the survival tradeoffs differed substantially between years. These results suggest that priority effects contribute to suppression of native forbs, but may not consistently promote higher fitness for earlier germinators. Instead, exotic invasion may increase yearly variation in the fitness consequences of native germination phenology, with potential implications for bet hedging strategies.


Exotic plant invasions can lead to large reductions in native abundance, shifts in community structure, and alterations in ecosystem functions such as fire frequency and soil nutrient cycling (Mack et al. 2000, Vila et al. 2011, Simberloff et al. 2013). Interactions between exotic and native plants also provide important insights into community assembly processes; in turn, developing a clearer picture of the mechanisms underlying invasion is key to effective management and restoration (Funk et al. 2008). Intensive research over the last few decades has contributed to an improved understanding of invasion ecology in some important respects (Sax et al. 2007); nevertheless, key gaps remain (Moles et al. 2012). Experimental studies of invasions are still far less common than observational ones (Lowry et al. 2013), and work that tests mechanisms by quantifying demographic effects of exotics on native species is surprisingly scarce (Jauni and Ramula 2015, Mordecai et al. 2015, Vaughn and Young 2015).

Priority effects are one mechanism often cited as an important contributor to plant invasions (Gioria et al. 2016). Many exotic annuals are thought to colonize open space or germinate faster than native competitors, and so gain advantages through resource preemption and legacy effects (Wolkovich and Cleland 2011). Existing work demonstrates that earlier arrival in the community benefits many species in a range of systems, although effects can vary widely depending on the interaction partners (Ulrich and Perkins 2014) and abiotic conditions (Young et al. 2015). Some studies suggest that priority may be more important than species traits and niche mechanisms in determining competitive outcomes, supporting the idea that historical contingency changes community assembly (Szymura and Szymura 2016). This work raises the question of whether invasive exotic species necessarily are inherently stronger competitors than natives (Seabloom et al. 2003). In other cases, priority effects may act in tandem with niche differentiation and species competitive ability in determining community composition (Perkins and Hatfield 2014, Werner et al. 2016).

California grasslands are heavily invaded and now dominated by exotic annual grasses and forbs, primarily introduced from the Mediterranean. Large areas of southern California coastal sage scrub (CSS) also have undergone type conversion into exotic annual grasslands, parallel to similar changes in other shrub habitats (Cox and Allen 2008, Talluto and Suding 2008). Previous work suggests priority effects are important in helping drive these shifts, particularly the replacement of native perennial grasses by exotic annuals. Exotic annual grasses have been shown to germinate earlier than native perennial ones, and these differences affect competitive outcomes (Abraham et al. 2009, Grman and Suding 2010, Young et al. 2015). For example, giving native perennial grasses just a 2- to 4-week lead in planting time increased native cover and reduced exotic annuals, effects persisting up to 8 yr after community establishment (Vaughn and Young 2015).

Priority effects can have important consequences not just upon initial arrival of species in the community but within individual growing seasons; such effects are likely to be particularly important for native forbs (Cleland et al. 2015, Gioria et al. 2016, Wolf and Young 2016). Growth chamber trials for a range of California CSS species found that exotics germinated faster and at higher rates than most native forbs, particularly under warm conditions (Wainwright and Cleland 2013). Other studies likewise support that southern California native forbs cue germination to cooler temperatures or larger rain events more likely to occur late in fall (Levine et al. 2008, 2011, Wainwright et al. 2012). This difference in germination phenology may be costly for native species if exotics gain a priority advantage; such effects also could select for earlier native germination (Cleland et al. 2015). At the same time, early germination potentially increases mortality risk (Gioria et al. 2016), for example, due to herbivory (Waterton and Cleland 2016) or hot and dry conditions that can occur between the first few rainstorms of the growing season (Levine et al. 2011). These tradeoffs are likely to shift in important ways as both temperature and rainfall patterns respond to climate change (Wolkovich and Cleland 2011).

While previous studies document the potential for within-season priority effects, we know relatively little about the demographic consequences of germination phenology for interactions between exotic grasses and native forbs. Most research on priority effects has been carried out in mesocosms or greenhouses; further steps are needed to connect this work with patterns of abundance in the field (Helsen et al. 2016). Growth chamber trials can fail to fully capture important germination dynamics (Levine et al. 2011). Moreover, the lag times between plantings imposed experimentally may not reflect the range of variation in emergence under ambient environmental conditions. Finally, many priority effects experiments are watered; this likely obscures the survival tradeoffs of different germination phenologies and therefore their implications for native plant fitness and population growth. More field studies are particularly critical for assessing how germination timing influences both priority effects and risk of mortality due to abiotic stress.

In this study, we focused on three native annual species that span a range of relative abundance, in a southern California grassland dominated by exotic annuals. For each, we asked the following questions: (1) When do seedlings emerge? (2) How does competition with exotic grasses affect survival and reproduction? and (3) How do competitive effects and survivorship patterns vary with germination timing? Emergence rates and timing for focal natives were measured in the field during three different years, both under ambient conditions and with exotic grass abundance reduced experimentally. In addition, germination rates of planted seeds were measured under both ambient and reduced exotic grass abundance.

We hypothesized that early native emergence results in a tradeoff, between the benefits of greater survival and growth in competition with exotic grasses and the costs of increased early mortality risk that could be avoided by delaying germination. To quantify this tradeoff, we measured survivorship for two time periods. We determined first the mortality risk experienced only by the earliest cohorts of seedlings to emerge during fall, before later bouts of germination occurred (early survival). Then, we measured the subsequent survivorship of both early and late germinating cohorts, starting after late germinants emerged and continuing until flowering (spring survival). Finally, estimates for these two different periods were integrated to determine net effects of emergence timing on survival from germination to flowering, both with and without exotic grass reduction.


Study system

The study area consists of former shrub habitat now supporting exotic grassland in southern California, at the 34.8-ha Bernard Field Station (34°6′ N, 117°42′ W; 348 m elevation; Claremont, California, USA). This grassland is bordered by intact CSS remnants and road. Soils are made up of well-drained alluvial silt, sands, and gravel. The climate is Mediterranean, with first fall rains typically occurring in early October and most precipitation between December and March. Historical total growing season rainfall (October–April) averages 44.04 cm per year (1971–2000, Western Regional Climate Center data for Claremont, California). During the three study years, total precipitation varied from slightly below (2011, 34.2 cm) to well above average (2009, 58.2 cm; 2010, 74.9 cm).

Exotic annual grasses dominate the study plots, with Bromus diandrus the most abundant (proportion of cover 0.43 ± 0.03 [mean ± standard error, SE]). Bromus madritensis (0.04 ± 0.02), Bromus hordeaceous (0.10 ± 0.010), and Vulpia myuros (0.32 ± 0.04) are also present, along with the exotic annual forb Erodium cicutarium (0.07 ± 0.03; Thomson et al. 2016). Previous work documents that most annual exotic grass germination in California takes place after the first fall rainstorm (Bartolome 1979). We quantified grass emergence rates in each year to characterize variability in the competitive environment (Appendix S1).

The three focal native species are herbaceous annuals that vary in relative abundance within the study grassland. Amsinckia intermedia is the most abundant native forb (proportion of cover 2006–2010: 0.14 ± 0.03 [mean ± SE]). This species is widespread throughout western North America, tolerant of disturbance and common in southern California exotic annual grasslands (Pantone et al. 1995). Phacelia distans occurs at lower densities (0.06 ± 0.01), but has been consistently present in the community over ten years of monitoring (2006–2016). Clarkia purpurea ssp. quadrivulnera is rare in this habitat. A fourth native, Camissoniopsis bistorta, was part of the original experiment but almost never observed in plots, so is not included in this analysis (details in Appendix S1). All focal species are native to a diversity of habitats in California and found in adjacent intact CSS. Previous research at this site showed that the focal natives responded positively to reductions in exotic grass cover in one of two growing seasons (Thomson et al. 2016). All three also increased in abundance with seed sowing, although effects were lagged for A. intermedia and P. distans and did not persist for any of the species.

Experimental design

In each of three growing seasons (fall to spring: 2009–2010, 2010–2011, and 2011–2012; hereafter referred to as 2009, 2010, and 2011), we carried out a randomized block experiment crossing reduction of grasses (control or exotic grass removal) with native seed planting (control or seed planted). We used the planting treatment to estimate focal native germination rates; individuals from planted seed were not part of seedling emergence analyses and represented only a small fraction of those tagged to estimate survival (see Data analysis). Each block was 2.5 × 2.5 m, containing four 0.75 × 0.75 m plots separated by 1-m buffer strips; the four treatment combinations were randomized and applied at the plot scale. Eight new blocks (×4 = 32 plots) were established in each year, except in 2010 when we included two additional blocks (10 in total).

We reduced exotic grass abundance through regular clipping. Past experiments show that this approach significantly reduces exotic grass cover without soil disturbance and damage to native seedlings (Levine et al. 2010, Thomson et al. 2016). We clipped and removed dead grass biomass and thatch from the previous growing season at the first rainstorm in October, while leaving any dead stems from native annuals. Thatch removal exposed newly germinating exotic grasses, which likewise were clipped back to the soil surface. We repeated clipping two to three more times in each growing season, at intervals of 4–6 weeks. Our study did not aim to test for soil legacy effects. Clipping might have indirectly influenced but not removed below-ground effects of exotic grasses, and soil conditions likely reflected persistent legacy consequences of long-term grass dominance in the community (Corbin and D'Antonio 2012).

All seeds were collected at the study area in May of the year prior to planting and stored at room temperature; we did not use commercial seed, to ensure germination behavior reflected that of unplanted seeds at the site. Seeds were added to treatment plots once per growing year, in November (dates in Appendix S2: Table S1). We seeded prior to December rainstorms, which previous research suggested cue the most native germination. Scarcity of local seed prevented planting multiple times per growing season. We divided seeded plots into quadrants (0.375 × 0.375 m) and assigned one for each focal native (Appendix S2: Fig. S1). A plexiglass sheet with nine evenly spaced holes on a 3 × 3 grid was laid over a quadrant, and seeds for the assigned species planted just under the soil surface through each hole. Sowing densities varied slightly among years, as we aimed to maximize the numbers planted given seed availability (seeds per planting point in 2009, 2010, 2011: A. intermedia = 1, 2, 1; P. distans = 1, 2, 1; C. purpurea = 2, 2, 1; plot densities = 9 × number seeds/point).

Seedling emergence counts

We counted native annual forbs emerging at different times during the fall rains in three years (2009, 2010, and 2011). Beginning in late September, the study area was checked regularly for new germination, 3–7 d after rainfall events with over 5 mm of precipitation. We initiated a germination count after each substantial new bout of forb seedling emergence. Lags between new storms and emergence varied from about three to ten days. We began counts once most germinants were large enough for identification to species, based on cotyledons or development of true leaves. Precipitation and corresponding bouts of germination did not divide exactly into distinct cohorts (Fig. 1). Nevertheless, in all three years, the census dates captured contrasts between early germination (following the first major rainstorms in October) and late germination (after colder storms in late November and early December; Appendix S2: Table S2). In 2009, we carried out two germination counts, one in mid-October and the second in early December (Fig. 1). The second study year (2010) was the wettest, and we conducted four counts: two early and two late. Three counts were done in 2011: one early and two late (Fig. 1). The average times between censuses classified as early and late were 48, 56, and 61.5 d in 2009, 2010, and 2011, respectively.

Details are in the caption following the image
Patterns of fall precipitation and the timing of emergence counts during the three study years (2009, 2010, and 2011). Weather data are from the Western Regional Climate Center records for the Claremont, California station. Day 0 of the growing season was defined as 1 September. Solid orange bars represent the dates of emergence censuses where newly emerged individuals were classified as “early” germinators, and blue dashed bars the dates of censuses where newly emerged individuals were considered “late” germinators. The total fall precipitation in mm is given on the top left of each panel.

At every germination census, we estimated seedling densities in nine 0.1 × 0.1 m subplots, evenly spaced across the plot (Appendix S1: Fig. S1). We recorded abundance for every forb species present in a subplot. At the first census in each year, all seedlings were newly emerged. In subsequent counts, we separately estimated abundance of both new seedlings with cotyledons and established seedlings without cotyledons. The exact abundance was recorded if fewer than six individuals for a given species and seedling type (newly emerged or established) were present; otherwise, we assigned an abundance category (6–10, 11–25, 26–50, or >50 seedlings). For any species and seedling type found in fewer than three subplots, we estimated abundance for the entire plot using the same classifications. Seedlings not large enough to identify were recorded as unknown, and representative individuals were marked for subsequent identification. In 2010 and 2011, we also carried out counts in all blocks from the preceding year, to increase sample sizes and better characterize variation between years (e.g., in 2010, the 2009 blocks were also sampled again). Treatments were not reapplied to blocks in the second year of sampling.

We checked for emergence from planted seeds during all germination counts that took place after planting and in one additional survey during late December of fall 2009. The plexiglass planting grid was held over each quadrant in seeded plots, and numbers of seedlings for any of the four focal natives found at each of the nine planting points were recorded. We used data for species not planted in a given quadrant to estimate and control for background emergence rates near planting points.

Native forb survival

In total, we recorded the fates of 1377 A. intermedia and 1021 P. distans marked seedlings over the three years of the study. First, early-germinating A. intermedia and P. distans seedlings were marked at the initial census in each year. To ensure even sampling of plants, one individual of each species was marked in every subplot where that species was present. If fewer than five individuals were marked for either species, we added replicates where possible outside of the subplots. Each marked seedling was encircled at the base with a colored twist tie, held in place by a colored plastic cocktail fork; the combination of loop and fork colors identified species and time of marking. We also marked C. purpurea seedlings where present, but few emerged except from planted seed.

In subsequent germination censuses, we checked all previously marked individuals for survival. In early January 2009 and December 2010 and 2011, we also replaced tagged individuals that had died by marking additional plants from the early germination cohort, when present. New seedlings were distinguished from plants that emerged earlier in the fall by the presence of cotyledons and paucity of true leaves. In 2009, there were two germination counts and so all individuals were classified as either early (October) or late (November–December) germinants (Fig. 1). We followed this same protocol in 2010 and 2011, although there were more than two germination counts in both those years. This facilitated comparison between years and avoided potential problems with classifying seedlings among germination cohorts on a finer time scale.

We also tracked survival of individuals germinating from planted seed for all three focal species. In 2009, we revisited all focal native seedlings associated with the planting points at subsequent censuses to score survival, using the plexiglass grid. Our 2009–2010 data showed no difference between survival of individuals from planted seed and those we tagged from background emergence. During the 2010 and 2011 growing seasons, we individually marked seedlings that emerged at planting points to increase sample size in survival estimates.

We checked marked seedling survival again during spring and recorded final size and flowering status (dates in Appendix S1: Table S1). Size was quantified as canopy volume, by measuring plant height and two diameters, one on the longest axis and the second perpendicular to the longest axis. For 2010 and 2011, we also recorded the number of inflorescences and the length of up to five inflorescences per plant, to characterize the relationship between size and reproductive success.

Data analysis

All analyses were carried out in R version 3.1.2 (R Core Team 2013).

Planted seed germination

Germination rates from planted seed were corrected for background emergence. For each focal species, we calculated the mean number of seedlings per plot recorded at planting points where the species was not sowed. Germination rate was determined by subtracting the estimated number of background seedlings from the total number of seedlings observed, and dividing by the number of seeds planted. We evaluated effects of clipping treatments on the fraction of planted seeds that germinated for each species and year with Fisher's exact tests (FET). Formal tests were not performed when fewer than 10 total seeds germinated in a given species and year combination.

Seedling emergence

We tested for effects of timing (early or late), year, and exotic grass clipping on background seedling emergence rates; these analyses excluded individuals germinating from planted seed. We found no evidence that plots clipped in the previous year differed from controls, so combined them in the final analyses. In cases where there were multiple counts for either the early or late category within a year (2010 early and late, 2011 late), we summed the abundance of seedlings scored as new over both censuses combined.

Total numbers of each species per plot at a census were determined by first assigning abundance classifications a value consisting of the category midpoint (e.g., for the 11–25 category, 17.5). We then averaged subsamples to estimate density per m2 and total plot abundance. Where whole plot counts were recorded, we used them directly. The distributions of estimated abundance were characterized by a high frequency of zero values and long tails; negative binomial and zero-inflated Poisson models did not produce good fits. We therefore separated the final analysis into log-linear models that predicted presence vs. absence, followed by general linear models (GLMs) testing for effects on seedling abundance only for plots where they were present. Log-linear models were generated for presence data using the R package MASS. Due to clear year by timing and clipping interactions, we separated analyses by both species and year. Saturated models including all effects of interest (presence by clipping by timing) were first created and then compared to alternative models using the backwards step procedure. Tests of effects are reported as Δ Akaike's information criteria (ΔAICs). For GLMs, we first recategorized all abundances (0 = no seedlings; 1 = 1 seedling/plot; 2 = 2–5; 3 = 6–25; 4 = 26–100; 5 = 101–200; 6 = 201–500; 7 = >500). The abundance category data were best fit by models with Gaussian error and log transformation of the response variable, which met assumptions. The effects of year, clipping, emergence timing, and year by emergence timing interactions on abundance were tested. We found no evidence of clipping by year interactions and did not include them in the final models.

Seedling survival

We separated seedling survival into two time periods. Early survival extended between initial seedling emergence of early germinators and that of late germinators; mortality during this period can be thought of as a cost to early germination. Spring survival encompassed December until flowering, when both early and late germinators were present. Only in 2009, spring data were analyzed separately for December and then from January on, because new individuals for both early and late germinators were marked at the January census to increase sample size.

We tested effects of year, exotic grass clipping, and their interactions on early survivorship of A. intermedia and Pdistans; timing by definition was not a factor, because only early-emerging seedlings experience early mortality. Saturated log-linear models with all effects of interest were first created and then compared to alternative models using the backwards step procedure. Tests of effects are reported as ΔAICs.

All years were combined in the spring survival analyses; we found no evidence of year by clipping or timing interactions, and sample sizes were small for individual years. We included A. intermedia and P. distans individuals that were tagged after emerging from planted seed in 2010–2011 and 2011–2012. These individuals from planted seed represented only 6% (A. intermedia) and 5.3% (P. distans) of tagged plants in the spring survival analyses, and removing them did not change any of our results. We first tested for an interaction between grass clipping and emergence timing effects on spring survival using log-linear models. If model comparison supported an interaction, we then used Fisher's exact tests for both early and late germinators to assess the effect of grass clipping on survivorship for each. Finally, we combined early and spring data into estimates of net survival to flowering, comparing early- and late-emerging A. intermedia and Pdistans in control and clipped plots. Bootstrap estimates of survival were generated by randomly resampling the data for each census in a given year with replacement, then calculating cumulative survival (= 10,000 replicates).

Low sample sizes limited our analysis of survival data for C. purpurea. We observed only three early C. purpurea germinants in all years combined, so did not estimate early survival or compare spring survivorship between cohorts. In 2009, 106 C. purpurea emerged from planted seed and we evaluated grass clipping effects on their spring survival with Fisher's exact tests. Only five total C. purpurea plants survived in 2010, and in 2011, just a single seedling emerged; we did not analyze survival in those two years.

Plant size and flowering

We tested whether grass clipping, emergence timing, or year affected final plant size for A. intermedia and Pdistans with GLMs. Data were rank-transformed because of skew. Interaction effects were not significant and removed from final models. We also evaluated the relationship between size and inflorescence data with a correlation test for the two years (2010 and 2011) where both measures were collected.


Planted seed germination

Average germination rates of planted seeds for all three natives exceeded 30% in 2009, but fell below 10% in 2010 and 2% in 2011 (Table 1). We found inconsistent effects of exotic grass clipping on germination. In 2009, clipping significantly increased Amsinckia intermedia germination, but marginally reduced germination for Phacelia distans and had no effect on Clarkia purpurea (Table 1). In contrast, 2010 clipping treatments significantly increased germination for C. purpurea but not A. intermedia, while only 10 total P. distans emerged. None of the study species germinated more than 10 total individuals from planted seed in 2011 (Appendix S2: Table S3).

Table 1. Proportion of planted seeds germinating for the focal native species in each year, separated by treatments (control and reduction of exotic grasses through clipping)
Grass treatment 2009 2010 2011
Amsinckia intermedia
Control 0.208 0.044 0.000
Clipped 0.375 0.061 0.019
P 0.040 0.640
Clarkia purpurea
Control 0.368 0.022 0.014
Clipped 0.368 0.119 0.000
P 1.000 0.006
Phacelia distans
Control 0.389 0.022 0.000
Clipped 0.222 0.024 0.028
P 0.046


  • The P values are from Fisher's exact tests comparing clipped to control for each year where at least 10 total seeds germinated. Bolded P values indicate statistical significance at a P < 0.05 level. Dashes indicate species and year combinations where small sample sizes prevented formal tests.

Seedling emergence

Amsinckia intermedia emerged in substantial numbers both early and late, but tended to germinate more after early storms. The probability of observing A. intermedia emergence did not differ between early and late censuses, except that late germination occurred in a significantly lower proportion of plots during 2010 (ΔAIC = 40.4, Fig. 2). However, in plots with germination A. intermedia seedlings were marginally more abundant for early counts than for late ones (Table 2, Fig. 3) and strongly so in 2010 (year by timing interaction, Table 2). Exotic grass clipping did not affect either the probability of finding A. intermedia seedlings in a plot (ΔAIC = −2) or the abundance of seedlings (Table 2).

Details are in the caption following the image
The proportion of plots where seedlings emerged each year, separated by emergence timing (early or late) and removal treatment (control or clipped), for (A) Amsinckia intermedia, (B) Phacelia distans, and (C) Clarkia purpurea.
Table 2. Results of GLM analyses testing effects of year, census timing (early or late), and exotic grass clipping on abundance of seedlings in plots where at least some individuals emerged
Factor Estimate SE t P
Amsinckia intermedia
(Intercept) 2.15 0.06 34.59 <0.0001
2010 −0.09 0.07 −1.33 0.183
2011 −0.58 0.07 −7.94 <0.0001
Timing (late) −0.17 0.08 −2.05 0.042
Clipping (clipped) −0.01 0.05 −0.26 0.792
2010 by timing (late) −0.71 0.13 −5.55 <0.0001
2011 by timing (late) 0.19 0.10 1.85 0.066
Phacelia distans
(Intercept) 2.21 0.08 28.46 <0.0001
2010 −0.23 0.09 −2.70 0.008
2011 −0.58 0.09 −6.77 <0.0001
Timing (late) −0.36 0.12 −2.92 0.004
Clipping (clipped) −0.09 0.05 −1.80 0.073
2010 by timing (late) −0.32 0.16 −2.04 0.042
2011 by timing (late) 0.31 0.13 2.30 0.022


  • SE, standard error. Values in boldface indicate statistical significance at a <0.05 level. The levels of factors to which estimates apply are given in parentheses.
Details are in the caption following the image
Relative seedling abundance in plots where emergence was observed, separated by timing (early or late) and removal treatment (control or clipped) for (A) Amsinckia intermedia and (B) Phacelia distans. Columns represent back-transformed values of the mean log abundance category, and error bars plus or minus one standard error. Abundance categories were assigned as (by number seedlings/plot): 0 = 0; 1 = 1; 2 = 2–5; 3 = 6–25; 4 = 26–100; 5 = 101–200; 6 = 201–500; 7 = >500.

Phacelia distans showed an even stronger tendency to germinate early more than late. In 2010, plots were more likely to contain P. distans seedlings early (ΔAIC = 49.9), and 2009 showed a similar but much weaker pattern (ΔAIC = 1.16, Fig. 2). Only in 2011 were there no differences in emergence probability (ΔAIC = −0.78). Across all years, P. distans seedlings were more abundant in early counts, although with some year by timing interactions (Table 2, Fig. 3). Responses to exotic grass reduction were neither strong nor consistent. In 2010, the probability of P. distans emergence was higher in clipped plots than in controls (ΔAIC = 5.1). Nevertheless, seedling abundance across all years combined was marginally lower in clipped plots, irrespective of emergence time (Table 2).

Alone of the three native focal species, Cpurpurea germinated almost exclusively later in the fall (Fig. 2). For 2010–2011 combined, C. purpurea were observed in just 2% of plots for early censuses compared to 25% for late ones (ΔAIC = 20.33). Clipping had no effect on Cpurpurea emergence (ΔAIC = −1.99). Clarkia purpurea did not germinate in enough plots to permit analyses of seedling abundance (= 13 across all years).

Early survival

Early germination entailed a large initial mortality cost in some years. For A. intermedia, early survival of the first germination cohort between October and December ranged from 40% in 2009 to only 8% in 2010 (Fig. 4, leftmost set of bars). Survivorship in P. distans followed a similar pattern, with very high mortality in 2010 (Fig. 4). We found no significant effect of clipping or year by clipping interactions on early survival of A. intermedia (Appendix S2: Fig. S2; ΔAIC = −1.52). In contrast, grass clipping significantly reduced survival of early-germinating P. distans, from 17% to 2.3% (Appendix S2: Fig. S2; ΔAIC = 28.94).

Details are in the caption following the image
Early, spring, and net survival for Amsinckia intermedia (A–C) and Phacelia distans (D–F) in each year (row), separated by emergence timing (early or late) and removal treatment (control or clipped). Early survival (leftmost bars in each panel) represents the proportion of early germinators that survived until late germinators emerged. Spring survival (middle bars) shows the proportion of seedlings present at the time of late emergence that survived until flowering. Net survival (rightmost bars) represents the cumulative survival from emergence to flowering, estimated over 10,000 bootstrap replicates with error bars showing 95% confidence bounds.

Spring survival

Effects of emergence timing on A. intermedia spring survival depended on the exotic grass treatment (Fig. 4, middle set of bars; ΔAIC = 2.1). Early germinators survived from December to flowering at double the rate of late germinators in control plots (Appendix S2: Fig. S3; FET, = 0.0004; early: 36.9.9%, late: 18.3%). Yet this spring survival advantage for early germinators was absent in clipped plots (FET, = 0.57; early: 37.8%, late: 33.4%). In December 2009, early germinators likewise survived better than late ones (ΔAIC = 5.89, early = 97.8%, late = 83.3%). However, survival of early and late germinators did not differ with exotic grass clipping for this specific window of time (ΔAIC = −0.02).

Phacelia distans followed a similar but weaker pattern (Fig. 4; Appendix S2: Fig. S3). Early germinators showed a marginally significant spring survival increase relative to late germinators in control plots (FET, = 0.085; early: 12.4%, late: 5.1%), but no difference in plots with exotic grass clipping (FET, = 0.62; early: 30.8%, late: 23.9%). Sample sizes were too low for P. distans in December 2009 to test for any differences in survival due to clipping treatment or emergence timing.

Clarkia purpurea showed the strongest response to clipping of any focal species; spring survival in 2009 was more than three times higher with grass clipping (= 0.004; control = 17.0%, clipped = 56.6%). These individuals were all late emerging and from planted seed, as we did not observe enough background C. purpurea germination to test for cohort effects on survival.

Net survival to flowering

Low numbers of seedlings surviving to spring contributed to uncertainty for integrated estimates of net survival, generating wide bootstrapped confidence intervals (Fig. 4, rightmost set of bars). Nevertheless, some interesting potential patterns emerged. For both A. intermedia and P. distans, net survival was highest for plants in clipped plots that emerged late. With grass clipping, early germination conferred no consistent benefits for spring survival but came with additional early mortality costs; as a result, later emergence led to higher net survival (Fig. 4). In contrast, for control plots, net survival differences between early and late germinators varied between years. In two of three years, A. intermedia that germinated early experienced about the same net mortality as those that germinated late, because the costs of early mortality in control plots were compensated by gains in spring survival. In 2010, however, early emergence was associated with >90% early mortality, and so this emergence strategy resulted in reduced net survival compared to late germinators. Similarly, early germination did not change (2010), potentially reduced (2009), or may have improved (2011) survival for P. distans in control plots, depending on the balance between early and spring mortality effects (Fig. 4).

Plant size and flowering

Plant canopy volume strongly correlated with total inflorescence length, indicating that size was a good proxy for flower production (Pearson's correlation r = 0.89, = 13.6, degrees of freedom [df] = 49, < 0.001). Neither A. intermedia nor P. distans final size differed with either exotic grass clipping or germination timing (Appendix S2: Table S4; clipping: A. intermedia t = −0.10, df = 252, = 0.92; P. distans t = −0.68, df = 99, = 0.50; timing: A. intermedia t = −1.15, df = 252, = 0.25; P. distans t = 0.50, df = 99, = 0.62).


Previous work suggests that native forbs germinate later than exotic grasses and experience a corresponding fitness cost due to priority effects on competition. Several authors have asked: If this is so, why do native forbs not evolve earlier germination phenology (Cleland et al. 2015)? Interestingly, our results show that some native annuals already tend to emerge early, and this strategy does ameliorate the effects of competition with exotic grasses. Nevertheless, the fitness consequences of earlier germination are not unambiguously positive, as high mortality in some years shifts the costs and benefits relative to later emergence. Our results indicate that native forbs may be between a rock and a hard place, experiencing strong competitive costs when germinating late but high mortality risk if germinating early.

Germination and emergence

Exotic species can exhibit greater germination plasticity than natives in growth chamber trials, emerging at higher rates and under a wider range of temperature and moisture conditions (Wainwright and Cleland 2013). Similarly, warm early rainstorms simulated through experimental watering in August and September cued exotic but not native germination in a southern California grassland (Wainwright et al. 2012). Levine et al. (2008) found that rare native annual abundance in a California Channel Islands grassland correlated positively with colder temperatures at the first major rainstorm. One interpretation of these results is that California native forbs exhibit little flexibility in germination timing, with emergence limited to colder storms later in fall. Yet our most abundant focal native species (Amsinckia intermedia and Phacelia distans) showed substantial variation in germination timing, and most seedlings emerged after the same first fall rainstorms that stimulated exotic grass germination (Figs. 2, 3).

One potential explanation is that fall rains began during October in all three years, when temperatures were not much higher than for later storms in November and December. We calculated mean minimum temperature for days with precipitation over 5 mm, also including the day before and the day after rainstorms occurred. Minimum temperatures for early storms were below 12°C in all years (2009, 11.8°C; 2010, 10.7°C; 2011, 9.6°C). Late storms were cooler on average by only 5–6°C (2009, 3.8°C; 2010, 6.0°C; 2011, 4.6°C). Previous growth chamber experiments simulating warm, early storms used higher temperatures, to test summer watering as a potential strategy for exotic control (Wainwright and Cleland 2013, Funk et al. 2015). In contrast, Levine et al. (2011) found that minimum temperatures during fall storms on the California Channel Islands ranged from 5°C to 10°C, with germination insensitive to variation below 15°C for two of three rare native annuals. The idea that native forbs require cold temperatures to germinate has been challenged by other recent research; trials for multiple populations of 42 California annuals showed no evidence for cold cuing in more than 80% of cases, including some Amsinckia and Phacelia species (Mayfield et al. 2014).

In contrast, late emergence by Clarkia purpurea is consistent with research demonstrating cold cuing in multiple populations for this species (Mayfield et al. 2014). Still, we have germinated local C. purpurea seeds at rates above 90% in a growth chamber, for minimum temperatures representative of both October (11.3°C) and December (5.1°C; R. King, unpublished data). This discrepancy between strong and flexible germination responses in the growth chamber and field data raises interesting questions. The absence of grass removal effects on C. purpurea germination suggests suppression by exotic thatch is not the explanation. Another possibility is differences in the degree of seed wetting for growth chamber compared to field conditions (Funk et al. 2015).

Temperatures associated with rainstorms did not vary between years and so are unlikely to explain annual differences in native emergence rates. Previous work supports a connection between the timing or size of initial fall rainstorms and germination behavior of native annuals (Levine et al. 2011). Yet fall precipitation and seedling emergence started between 5 October and 12 October in all years of our study, after storms with comparable total rainfall (2009, 34.03 mm; 2010, 32.52 mm; 2011, 35.81 mm). In 2010, a series of heavy storms led to substantially higher late rainfall than for other years, but did not result in greater emergence (Fig. 1; 2009, 33.77 mm; 2010, 456.14 mm; 2011, 32.76 mm).

Cover estimates from the preceding springs suggest that seed rain exerted the strongest influence on annual variation in emergence. In spring 2009, native cover reached much higher levels than are typical for the study site (Thomson et al. 2016). Native cover fell slightly in spring 2010 and dropped to very low values in spring 2011 (mean proportion of cover ± one SE for A. intermedia, 2009 = 0.30 ± 0.05, 2010 = 0.17 ± 0.05, 2011 = 0.002 ± 0.004; for P. distans, 2009 = 0.14 ± 0.03, 2010 = 0.12 ± 0.04, 2011 = 0.002 ± 0.002). Still, we also observed reduced germination rates in 2010 and 2011, suggesting that seed abundance alone did not drive lower background emergence in those years. Effects of parental physiological stress could also have played a role in reducing germination and emergence for 2010 and 2011, compared to more favorable conditions that might have promoted higher-quality seeds in spring 2009.

Emergence timing and competition

Lagged emergence relative to exotic annual grasses could affect native forbs on several time scales. First, native seedlings may emerge after the same rainstorm, but days later than exotic grasses; multiple growth chamber trials (Wainwright and Cleland 2013, Waterton and Cleland 2016) and some field studies (Wainwright and Cleland 2013) show this pattern. Natives might also respond to different germination cues than exotics, emerging after rainstorms occurring weeks apart; many previous studies of priority effects assumed differences in phenology within this range. Our data were not collected on a fine enough time scale to assess whether early germinators lagged exotic grasses in emergence by several days after October rainstorms. We also cannot distinguish what proportion of exotic grasses germinated in October, although Bromus spp. clearly reached substantial densities by the first emergence survey (from ~700/m2 in 2009 to >3000/m2 in 2010). Still, any short emergence delay for early germinators relative to exotic grasses did not affect their fitness. Similarly, Cleland et al. (2015) showed that California coastal scrub natives with a range of life histories experienced little to no negative effect of germinating a week later than exotic competitors, although some reached higher biomass when planted earlier.

In contrast, for all three of our focal species, the month or greater lag associated with late germination led to lower survival in control than in clipped plots (Fig. 4). Previous studies similarly show significant gains in biomass due to priority effects, when competing species are planted from 3 to 5 weeks apart (Grman and Suding 2010, Dickson et al. 2012, Stuble and Souza 2016). One important limitation is that our data likely underestimate priority effects. Clipping reduced exotic grasses but did not completely remove them, and the exotic forb Erodium cicutarium responded positively to clipped treatments for two of the three years (proportion of cover [mean ± one SE] 2009, clipped: 0.18 ± 0.05, control: 0.10 ± 0.03; 2010, clipped: 0.06 ± 0.02, control: 0.02 ± 0.01). Likewise, a long history of dominance by exotic grasses may have created soil legacy effects at our site; even short-term soil effects of exotic grasses would not have been removed by clipping treatments (Corbin and D'Antonio 2012).

Tradeoffs of germination timing: survival and growth

Early germination appears to confer significant benefits in the form of greater ability to compete with exotic grasses; this raises the question of why a substantial number of native seedlings emerged late. One might hypothesize that priority effects would promote earlier native germination (Cleland et al. 2015). Part of the answer is likely that an early emergence strategy comes with substantial survival tradeoffs (Gioria et al. 2016). Early germinators in our study experienced an additional mortality cost of approximately 40–95% depending on the species and year. Although our data set is too short for testing correlations with weather conditions, there was no obvious relationship between mortality and total growing year rainfall (Figs. 1, 4). Survivorship dropped in 2010 relative to 2009, even as total precipitation went up by nearly 30%. Much higher January precipitation may have increased spring survival in 2009 (2009, 19.5 cm; 2010, 1.4 cm; 2011, 2.8 cm). Hot conditions soon after the first cohort emerged could explain reduced early survival in 2010; maximum temperatures exceeded 29°C for five consecutive days beginning on 1 November, peaking at 36.7°C. Stochastic events such as storms or frosts can be major sources of seedling mortality, leading to high temporal variation in survival of different cohorts (Donohue et al. 2010).

The risks of early emergence could be ameliorated if natives possess strong predictive ability to time germination when high survivorship is likely (Levine and Rees 2004). Our results do not support a positive correlation between native emergence and survival; for example, early-season survival was greatest in 2011, but both overall emergence rates and the proportion of early germinators lowest of the three years. If abiotic conditions early in the growing season are highly variable and native species cannot reliably cue their germination to match, maintenance of multiple germination strategies could represent adaptive bet hedging (Rice 1987, Simons 2009, Jimenez et al. 2016).

Another important tradeoff influencing fitness consequences of germination phenology is growing season length. Our results contrast with previous work supporting the importance of growing season duration to size and reproductive success of annuals (Verdu and Traveset 2005). The absence of significant size benefits for early-germinating A. intermedia and P. distans may reflect low emergence and survival, resulting in small sample sizes at flowering. In a subsequent year, early-emerging P. distans reached final canopy volumes more than four times greater than late-emerging individuals (J. Kwok, unpublished data). Better estimates of how growing season length affects native annual fitness are particularly critical to predicting the consequences of climate change, as earlier onset of summer drought may impose additional costs on late emergence (Wolkovich and Cleland 2011).

For P. distans, another mechanism that influences fitness implications of emergence timing is the facilitative effect of exotic grasses on early-season survival (Fig. 4). We previously observed similar reductions in P. distans abundance with grass clipping during a dry year (Thomson et al. 2016). Interactions between plants can be highly context dependent and switch from competitive to facilitative, contingent upon abiotic conditions. Similar to our results, both native and exotic California annuals have been shown to survive better when planted at higher density, although with significant reductions in biomass due to competition (Leger and Espeland 2010). This pattern in our data likely results not from any specific benefit of exotic grasses per se, but changes in seedling exposure and water retention from clipping back species that can constitute more than 90% of the community cover. Uninvaded CSS provides greater shade and higher levels of litter than our clipped plots, both of which are likely to promote early forb survival.

Net survival and summary

Our net survival analyses suggest an interesting potential pattern and hypothesis for further exploration. When grasses were clipped, late emergence consistently led to higher estimated net survival. In contrast, for control plots, the net survival payoffs of early and late germination differed between years. Given that our data likely underestimate both the size benefits of early germination and the competitive costs of late germination, caution should be taken in interpreting them. Still, this result raises the interesting possibility that invasion by exotic grasses may have resulted in more variable fitness outcomes for different native forb emergence times, rather than a consistent advantage of earlier emergence. Some predictions about the consequences of climate change for California grasslands have focused on the fitness benefits of early germination for exotics (Wolkovich and Cleland 2011). If increasing variability in climate instead were to favor diverse germination strategies, this may have important implications for future interactions between exotic and native annuals (Jimenez et al. 2016).

Our results also help clarify how demographic responses shape relative abundance of the three most common native species in this grassland, which contribute on average 75.7% of the total native cover (2009–2012). The most common native (A. intermedia) generally survived at higher rates than the other two focal species, regardless of treatment or cohort. Interestingly, exotic grasses also suppressed A. intermedia least and the rarest native (C. purpurea) most. These patterns reinforce earlier findings that priority effects often act in concert with other mechanisms and species traits in structuring invaded grassland communities (Perkins and Hatfield 2014).

California grasslands are an important model system for research on both exotic plant invasions and priority effects (Abraham et al. 2009, Grman and Suding 2010, Mordecai et al. 2015), but with few field studies of native plant demography. Our experiment helps contextualize this previous work, illustrating the importance of field data generally and in particular realistic mortality estimates to a complete picture of native and exotic plant interactions. Still, not all exotic species germinate earlier than native competitors, raising the question of how generalizable these results might be. Interestingly, winter rainfall patterns and risk of early mortality have been linked to germination behavior in the widely invasive annual Centaurea solstitialis, with emergence strategies shifted in parts of the introduced range relative to the native one (Hierro et al. 2009). Results like these illustrate that not just native but introduced species respond to demographic tradeoffs between competitive advantage and early mortality risk. Research approaches that explicitly quantify relative costs and benefits of emergence timing can prove a useful, general framework for better understanding how priority effects influence both individual species and communities (Verdu and Traveset 2005, De Luis et al. 2008).

In summary, our results support that some native annuals could be limited in their response to exotic grass invasion by low germination plasticity, but others show substantial flexibility in emergence timing. Later germination led to greater native competitive suppression by exotic grasses, supporting the importance of priority effects; still, the most common native species tolerated grass competition better and survived at higher rates regardless of emergence timing. Early germination ameliorated negative effects of grass competition, but at the expense of reduced survival in some years. The degree to which germination cuing behavior is heritable or has changed in exotic grasslands relative to uninvaded habitat is not clear. One important hypothesis to explore further is that exotic invasion has increased fluctuating selection on native emergence times, potentially promoting maintenance of diversity in germination phenology and with important implications for future responses to climate change.


We thank Lauren Cole for help in collecting data in 2011–2012, as well as the editor and several anonymous reviewers, for helpful comments on the manuscript. This work was supported partially by NSF Award no. 0950106.