Historical contingency in ecology and restoration: year effects, priority effects, and climate change in California grasslands

INTRODUCTION
There are two opposing ideas about how natural communities develop after disturbance: 1) communities will tend to return to their pre-disturbance state, regardless of the vagaries of early establishment; and 2) species that make up communities are largely determined by the order of arrival of colonizers. This study seeks to address these ideas while asking whether year-to-year variation affects the communities in the long term. In this multi-year research program we examine the plant community ecology of California grasslands, controlling for colonization order of plants in the context year-to-year variation in rainfall. This research will also test whether rare climatic events are more important for plant communities than gradual changes in average climate, with implications for how we respond to climate change. We are carrying out a series of replicated experiments that test central questions shared by both conceptual ecology and ecological restoration, in the context of climate change, and we will implement these experiments identically for at least five consecutive years. At the heart of these experiments is the hypothesis that the year of establishment is a major determinant of long-term community structure.
There are two sets of core experiments being carried out at three sites in California grasslands. The Grass/Forb Priority Experiment asks whether short-term stochastic effects (arrival order across years) can override longer-term deterministic effects (succession) in perennial grasslands undergoing restoration. A similar Native/Weed Priority Experiment asks how much the competitive suppression of native perennial grasses by exotic annual grasses could be overcome by a single year of weed control. For each of these two core experiments we have established five replicate plots at each of three grassland sites in northern California. Treatments were initiated in 2011, and we will do this repeatedly for (at least) five consecutive years total, in a formal test of the importance of Year Effects in ecological systems. Each year we will also initiate five replicate plots of watering experiments at each of the three sites. These watering experiments initially will be based on our a priori understanding of the limiting factors in these systems, but as the experiments progress, we will use these targeted experiments to test hypotheses arising from inter-annual differences observed in our core experiments. These experiments also allow us to test the degree to which short-term variation in current climate can be used to predict the long-term responses of species and communities to climate change.

HYPOTHESES
H1) Simple Competition: Guilds grown in monocultures will perform better than when grown in competition with other guilds, and this competition will differ in years with different rainfall patterns (see below). Even replicating this simple experiment over several years at several sites will be both unique and revealing.
H2) Priority (Grass/Forb): Guilds planted a year earlier with have greater initial success in competition with guilds planted later, but this effect will be greater with grass priority over forbs than with forb priority over grasses.
H3) Priority (Grass/Forb): Over several years, priority effects will become muted, and different treatments will become more similar in species composition (convergence). However, the priority advantage of grasses will last longer than the priority advantage of forbs, and exclusion of forbs may occur (divergence).
H4) Priority (Grass/Forb): There will be distinct "grass years" and "forb years", expressed even in plots where guilds are planted alone, and these will correspond to different rainfall patterns (c.f. Lulow 2004). Variation in the success of forb and grass plantings will be more strongly correlated with patterns of rainfall than with total rainfall (c.f. Pitt & Heady 1978).
H5) Priority (Grass/Forb): Supplemental watering will alter the community structure in these plantings. For example, based on the patterns reported by Pitt & Heady (1978), adding water to plots in a dry fall may change a forb year into a grass year. Conversely, adding water during a mid-winter drought may change a forb year into a grass year. Both of these experimental treatments will alter the results of the priority experiments in ways similar to natural interannual variation in rainfall.
H6) Priority (Native/Weed): When exotic annual grasses are planted one year later than native perennial grasses their competitive advantage will be significantly reduced.
H7) Priority (Native/Weed): This competitive change will initially be asymmetrical: the effect of one year of priority will be more beneficial to the perennial grasses than it will be detrimental to the annual grasses. Later, as perennial grass biomass increases, the priority advantage will translate into significantly reduced weed biomass.
H8) Priority (Native/Weed): There will be distinct "native years" and "weed years", expressed even in plots where guilds are planted alone, and these will correspond to different rainfall patterns (c.f. Bakker et al. 2004).
H9) Priority (Native/Weed): Supplemental watering will alter the community structure in these plantings. For example, based on the patterns reported by Hamilton et al. (1999), adding water to plots during a mid-winter drought may favor perennial grasses more than exotic annual grasses. Conversely, adding water to plots in a dry fall may favor exotic annual grasses more than perennial grasses. Both of these experimental treatments will alter the results of the priority experiments in ways similar to natural interannual variation in rainfall.
H10) Year effects: Success of individual plantings will vary significantly among years, and our watering experiments will be able to reverse these year effects.
H11) Year Effects: The relative advantage gained by site priority will differ significantly from year to year, with those years favoring a particular guild in monoculture also producing stronger priority effects for that guild.
H12) Climate: Sites differing in mean historical climate will produce communities that differ significantly, averaged across all years.
H13) Climate: Population and community variation associated with climate in the year(s) of restoration will exceed variation associated with site differences.

METHODS
Within each site (Davis Research Fields 38° 32.61' N, 121° 47.19' W; Hopland Research and Extension Center 38° 58.99' N, 121° 5.18' W; and McLaughlin Natural Reserve 38° 52.17' N, 122° 25.31' W), we selected areas of uniform topography on similar soils (relatively fertile clay loams) to minimize within-site heterogeneity. We chose species mixes that represent appropriate native perennials and common exotic annual grasses, already present at these sites, and known to germinate and establish well in restoration settings. Species planted were as follows (with those sites in parentheses indicating particular species that were not planted at all three sites, in order to make adjustments at the species level to match local sites):

Native Perennial Grasses - Stipa (Nassella) pulchra, Hordeum brachyantherum, Elymus glaucus, Bromus carinatus
Native Perennial Forbs - Achillea millefolium, Artemisia douglasiana, Eschscholzia californica, Croton setigerus
Exotic Annual Grasses - Avena fatua (Davis), Avena barbata (Hopland and McLaughlin), Vulpia myuros (Davis and McLaughlin), Vulpia bromoides (Hopland), Bromus hordeaceus, Hordeum murinum

One experiment will use the native perennial grasses and forbs (Grass/Forb Priority Experiment), and the other will use the native perennial grasses and the exotic annual grasses (Native Perennial/Annual Weed Priority Experiment). All seed was purchased from native plant growers who have documented local provenances, or when available, collected locally from remnant populations. We will aggressively control all unplanted species in unused plots and buffer areas between the plots, as well as within the planted native grass/forb experiment.
For the Grass/Forb Priority Experiment there will be eight treatment combinations, representing all combinations of the two guilds planted over a two-year period. Each of the eight treatments are described as follows, giving with species planted in year 1 (2011) and species planted in year 2 (2012) if any (if no planting for that year, listed as "none"):

Treatment Planting Year 1, Year 2
1 Native grasses, none
2 Native forbs, none
3 Native grasses and forbs, none
4 Native grasses, Native forbs
5 Native forbs, Native grasses
6 none, Native grasses
7 none, Native forbs
8 none, Native grasses and forbs

While treatments 6, 7 and 8 from the Year One design are identical to treatments 1, 2 and 3 in the Year Two design, they must be independently replicated in order to test for interactions between priority and year effects. The entire design will be replicated in each of at least five years. Each year, the appropriate plot will be sown with a mix of native perennial grasses, native forbs, or both according to the assigned treatment. Prior to sowing, each plot will be cleared of all vegetation by herbicide, if needed, and shallow tilling. We will sow both grass and forb mixes at a rate of 800 live seeds/m2/guild in all treatments. We will plant at the onset of the rainy season each year, and rake seeds into the soil. Plots will be hand-weeded for obvious non-sown species.
For the Perennial/Annual Priority Experiment we repeated the above design, replacing the native perennial forbs with a mix of common exotic annual grasses. Five main treatments were applied as follows:

Treatment Planting
1 Native grasses sown alone
2 Native and exotic grasses sown together
3 Native grasses sown, followed by exotic grasses two weeks later (2-week priority)
4 Exotic grasses sown alone, two weeks after other initial plantings
5 Natives grasses sown alone, followed by exotic grasses one year later (1-year priority)

All of these exotic grasses are already common at each of the experimental sites, so there will be no risk of releasing new invasions from our study plots into the surrounding land. Rather than relying on natural invasion, seeding the exotic grasses will give control in a way that will simulate real-world conditions, where exotic seed often saturates many grassland sites.
Both of the above experiments will be also include a set of replicates that will receive a supplemental irrigation treatment; each block is split in half and receive watering. These replicates will be the subject of a different watering experiment each year, for which the remaining intact replicates ("core experiment") will serve as controls. For both experiments, core treatments were randomly assigned within each half block. Each experimental unit (plot) was 1.25 m on a side, and each was separated from adjacent plots by 1 m.
Each year's watering experiment will initially be based on what we can infer from previous descriptive studies as likely climatic drivers of community structure. Later, we will use data from our own replicate core experiments to generate hypotheses about which aspect of rainfall may be drivers of community structure. In each given year, the particular rainfall pattern for that year will determine which modifications are likely to be experimental tests of weather drivers. For example, if in a given year the fall rains are late, or there is a sustained midwinter drought (both common events, and putative drivers of community structure), we will add supplemental aerial applications of water by hand to the targeted replicates in amounts representative of normal rainfall years (~10 cm). By their nature, this watering will be reactive (with a lag) to actual conditions, but we believe they can nonetheless provide tests of previously described patterns based on natural climatic variation, with strong covariance among several rainfall and temperature variables. After each year's targeted treatment, these plots will be left intact and monitored for the duration of the experiments (at least 5 years).
We have carried out an analysis of the last 54 years of local rainfall (unpublished data). In fully half of those years there was a mid-winter drought of at least three weeks, while in nearly 40% of years a substantial fall rain (2.5 cm) did not occur until after December 1st. Therefore, we can expect that one (or both) of these drought events will occur in a majority of years. Our analysis further shows that in years with either of these drought types, total yearly rainfall is approximately 10 cm less, and that the effects of the two types of drought on total rainfall are simply additive. That is, years with both late onset rain and a sustained midwinter drought average 20 cm less rainfall than years with neither. Therefore, our supplemental watering during a natural drought will both eliminate the drought effect directly and increase overall rainfall in a way that mimics non-drought years.

Every three months, we will measure cover using a ten-pin point frame, placed 20 times in each plot, counting all hits by species and by plant part (living and dead). We will estimate plant density by counting all individuals (to species) in four 25 cm x 25 cm subplots in each plot.

STATISTICAL ANALYSES
Each hypothesis will be tested with a mixed model (Type III) ANOVA, with experimental treatments and guild as fixed effects, and site and (when applicable) year of initiation as random effects. Multiple measurements will be incorporated as repeated measures. We will test for differences among sites in general (H12), but do not have enough sites to test specific climatic or other correlates of differences across study sites. After five years, we believe that we will have enough power to test Hypothesis H13. The number of replicates in these experiments represents a compromise between the power to answer individual questions in a given year, and the power to ask questions across and between years. Based on the preliminary study, we believe that these replicates will be sufficient for both tasks, especially with the potential added power of multiple sites and (for some analyses) repeated measures.
We will compare similarity indices based on Euclidean distance in the first four principal components of PCA to test hypotheses about convergence and divergence, and to test whether variation in community structure associated with climate in the year(s) of species establishment exceeds variation associated with site differences.

CAVEAT
In addition to priority effects and year effects, we recognize that there are other drivers of ecosystem function in these grasslands, including herbivory, soil disturbance, fire, microbial interactions, and other forms of competition. We are not seeking at this time a comprehensive description (or literature review) of how all of these factors drive community structure, but rather how these particular historical factors do.