PLANTING YEAR CONDITIONS SHAPE COMMUNITY ASSEMBLY 

IN RESTORED PRAIRIE COMMUNITIES 

By 

Anna Marjorie Groves 

 

 

 

 

 

 

 

 

A DISSERTATION 

Submitted to 

Michigan State University 

in partial fulfillment of the requirements 

for the degree of 

Plant Biology – Doctor of Philosophy 

Ecology, Evolutionary Biology and Behavior – Dual Major 

2018 

ABSTRACT 

PLANTING YEAR CONDITIONS SHAPE COMMUNITY ASSEMBLY 

IN RESTORED PRAIRIE COMMUNITIES 

By 

Anna Marjorie Groves 

This dissertation focuses on the influence of planting conditions on the restoration of prairie 

communities, both in the applied context of improving restoration outcomes, and as a test of our 

understanding of mechanisms of community assembly. Ecological restoration is incredibly important in 

this era of human-induced landscape degradation, but restoration projects do not always result in 

outcomes that match practitioners’ goals. A possible factor in this variation is the weather conditions 

that occur at the outset of a restoration project that are typically referred to as “year effects.” However, 

we lack strong tests of year effects that disentangle them from site-to-site variation and other 

confounding factors, such as management. Additionally, there has been no investigation of the 

mechanisms for year effects; current evidence for this phenomenon is correlational and lacks direct 

experimentation. Inter-annual variation in precipitation is often assumed to be a major factor causing 

year effects, though there are countless potential drivers that could vary inter-annually. In addition to 

implications for restoration, evidence for year effects would suggest assembly is driven by stochastic 

mechanisms (i.e. random variation due to weather) as opposed to deterministic mechanisms (i.e. drivers 

with consistent, repeatable effects such as site soil characteristics or prescribed fire management).  

To address these challenges, I investigated the causes and consequences of planting-year 

variation on community assembly in a tallgrass prairie restoration system using two main methods: (I) 

an experimental manipulation of precipitation as part of a long-term prairie restoration at Michigan 

State University’s W. K. Kellogg Biological Station (KBS), and (II) an analysis of differences in species 

composition among 83 restored tallgrass prairies established in different years across the Midwestern 

United States of America. I first explored how year effects on seedling emergence and survival. I created 

6 replicate prairies in three different years, using identical methods at a single large site. In each year, 

experimental plots within each restored prairie received water manipulation treatments for the first 

nine weeks after sowing. In these plots, rainfall was blocked with rain-out shelters and then watered low 

(1.39 cm/month), mean (9.91 cm/month), or high (17.81 cm/month) amounts compared to the 30-year 

average in the area. I found large differences in seedling emergence across the three planting years as 

well as across watering treatments. Initial community composition differed across years but not 

watering treatment. I continued to monitor these communities to determine if initial differences would 

persist. I found that some initial community differences dissipated while others persisted, suggesting the 

operation of both deterministic and stochastic assembly mechanisms in this system.  

In the second study, I looked for year effects in existing restored prairies across a wide 

geographic and temporal range. In summer 2016, colleagues and I visited 83 prairies in Illinois, Indiana, 

and Michigan that had been restored between 2000-2015 and determined their current plant 

community compositions. Using these data, I evaluated the importance of planting conditions using 

historical weather data on restoration outcomes, also accounting for different management histories 

and locations. I found that planting year precipitation had lasting effects on the richness and abundance 

of non-sown species. These signatures explained as much variation in the community as known drivers 

of assembly such as soil characteristics, fire frequency, and seed mix richness. These data provide 

additional support that planting year conditions can have profound effects on communities in sown 

restoration projects which can persist through assembly over time. Documentation of persistent year 

effects will be helpful for understanding the mechanisms that drive variation in restoration outcomes 

and could be used to inform tailored remedial management that addresses contingencies that arise 

after different first year conditions.

For Andy

 

iv 

ACKNOWLEDGMENTS 

 

Thank you, thank you, thank you to my committee members for helping this work take shape 

over the years: Kay Gross, Jen Lau, and Phoebe Zarnetske. To many field technicians and undergraduate 

researchers who directly contributed their sweat and patience, particularly Meg Kargul, Michelle 

Franklin, Katie Grantham, Logan Brissette, George Wheeler, and Ashley Knudson. To the post-docs and 

other colleagues always willing to share their sage wisdom: Nash Turley, Emily Grman, Tyler Bassett, and 

Becky Barak. To Steve Norris for not living up to his completely inaccurate reputation. To Mark 

Manuszak for ingenuity and vital morale support. To friends who helped me in the field for selfless 

reasons still unknown to me: Carina Baskett, Abby Busekrus, Matt Chansler, Andy Funk, and Lisa 

Stelzner. To Christopher Warneke for sharing genuine compassion, botanical enthusiasm, and occasional 

birding outings. To Chad Zirbel for vital statistical support, but much more importantly, for being a really 

good buddy. To my advisor Lars Brudvig; I would not have made it through without you. 

Meg Engelhardt showed me that plants are cool and field work is a delight. Tom Simpson 

showed me that prairies can be planted, some work out better than others, and that experimenting with 

management can lead to fascinating results. R. Given Harper taught me ecology, how to write like a 

scientist, and gave me the critical nudge to apply for graduate programs. It is the fault of these three 

ecologists that I ended up here, an ecologist.  

To the cute boy I met at the MSU Plant Sciences recruitment weekend: I can’t wait to marry you. 

 

v 

TABLE OF CONTENTS 

 

LIST OF TABLES ............................................................................................................................................ viii 
 
LIST OF FIGURES ............................................................................................................................................ix 
 
CHAPTER ONE ............................................................................................................................................... 1 
PLANTING YEAR EFFECTS IN RESTORATION .................................................................................................. 1 
INTRODUCTION ................................................................................................................................ 1 
Historically contingent assembly models ........................................................................... 2 
Testing year effects in restored communities .................................................................... 4 
Study system ....................................................................................................................... 6 
LITERATURE CITED ........................................................................................................................................ 8 
 
CHAPTER TWO ............................................................................................................................................ 13 
INTER-ANNUAL VARIATION IN PRECIPITATION AND OTHER PLANTING CONDITIONS  
IMPACTS SEEDLING ESTABLISHMENT IN SOWN PLANT COMMUNITIES ..................................................... 13 
ABSTRACT ....................................................................................................................................... 13 
INTRODUCTION .............................................................................................................................. 14 
METHODS ....................................................................................................................................... 17 
Study site and restoration methods ................................................................................. 17 
Experimental design ......................................................................................................... 18 
Precipitation treatments ................................................................................................... 19 
Seedling and plant community surveys ............................................................................ 20 
Statistical analysis ............................................................................................................. 21 
RESULTS ......................................................................................................................................... 22 
Ambient conditions ........................................................................................................... 22 
Precipitation treatments ................................................................................................... 23 
Species responses ............................................................................................................. 25 
DISCUSSION.................................................................................................................................... 26 
ACKNOWLEDGEMENTS .................................................................................................................. 29 
APPENDIX .................................................................................................................................................... 30 
Figures ............................................................................................................................................ 31 
Tables ............................................................................................................................................. 39 
LITERATURE CITED ...................................................................................................................................... 46 
 
CHAPTER THREE .......................................................................................................................................... 53 
EXPERIMENTAL MANIPULATION OF PLANTING YEAR CONDITIONS  
ALTERS PRAIRIE RESTORATION TRAJECTORIES ........................................................................................... 53 
ABSTRACT ....................................................................................................................................... 53 
INTRODUCTION .............................................................................................................................. 54 
METHODS ....................................................................................................................................... 58 
Study site and restoration methods ................................................................................. 58 
Experimental design ......................................................................................................... 58 
Plant surveys ..................................................................................................................... 60 
Statistical analyses ............................................................................................................ 60 

 

vi 

RESULTS ......................................................................................................................................... 61 
Community composition................................................................................................... 61 
Cover of sown species ....................................................................................................... 61 
Richness of sown species .................................................................................................. 62 
Non-sown species ............................................................................................................. 62 
DISCUSSION.................................................................................................................................... 63 
ACKNOWLEDGEMENTS .................................................................................................................. 66 
APPENDIX .................................................................................................................................................... 68 
Figures ............................................................................................................................................ 68 
Tables ............................................................................................................................................. 72 
LITERATURE CITED ...................................................................................................................................... 87 
 
CHAPTER FOUR ........................................................................................................................................... 95 
HISTORICAL CONTINGENCY IN ECOLOGICAL RESTORATION: EVIDENCE FROM 83 PRAIRIES ..................... 95 
ABSTRACT ....................................................................................................................................... 95 
INTRODUCTION .............................................................................................................................. 96 
METHODS ..................................................................................................................................... 100 
Study system ................................................................................................................... 100 
Field data collection ........................................................................................................ 100 
Management histories and historical weather data....................................................... 101 
Statistical analysis ........................................................................................................... 102 
RESULTS ....................................................................................................................................... 103 
Effects of planting year on community metrics .............................................................. 104 
Relative influence of weather compared to known assembly drivers ........................... 105 
DISCUSSION.................................................................................................................................. 105 
ACKNOWLEDGEMENTS ................................................................................................................ 109 
APPENDIX .................................................................................................................................................. 110 
Figures .......................................................................................................................................... 111 
Tables ........................................................................................................................................... 122 
LITERATURE CITED .................................................................................................................................... 128 
 
CHAPTER FIVE ........................................................................................................................................... 133 
CONCLUDING REMARKS ........................................................................................................................... 133 
 
 

 

 

vii 

LIST OF TABLES 

 

Table 2.1. Low richness and high richness seed mixes sown. ..................................................................... 39 

Table 2.2. Overview of results with Panicum capillare included. ............................................................... 41 

Table 2.3. Plot-level means for seedling survival measured across three planting years in ambient and 

watering treatment plots. .............................................................................................................. 42 

Table 2.4. Statistical results for planting restored prairie in three planting years (ambient conditions). . 43 

Table 2.5. Statistical results for planting restored prairie with three watering treatments in three 

planting years. ................................................................................................................................ 44 

Table 3.1. Weather conditions in three planting years. ............................................................................. 74 

Table 3.2. Statistical model results predicting similarity of community composition. ............................... 75 

Table 3.3. Statistical model results predicting the cover of sown species. ................................................ 77 

Table 3.4. Statistical model results predicting the richness of sown species. ............................................ 79 

Table 3.5. Statistical model results predicting the cover of non-sown species. ......................................... 82 

Table 3.6. Statistical model results predicting the richness of non-sown species. .................................... 85 

Table 4.1. Weather variable rotations for PC axes 1, 2, and 3.................................................................. 122 

Table 4.2. Six variables used as response metrics to evaluate effects of planting year weather conditions 
and other assembly drivers on prairie restoration outcomes. .................................................... 123 

Table 4.3. Soil variable rotations for PC axis 1. ......................................................................................... 124 

Table 4.4. Complete statistical results for models analyzing the relative importance of weather 

conditions and other factors on prairie restoration outcomes. .................................................. 125 

 

viii 

LIST OF FIGURES 

 

Figure 2.1. Methods schematic. .................................................................................................................. 31 

Figure 2.2. Mean seedling emergence across three planting years for sown and non-sown species under 
ambient conditions. ....................................................................................................................... 32 

Figure 2.3. Mean seedling emergence across three precipitation treatments for sown and non-sown 

species. ........................................................................................................................................... 33 

Figure 2.4. Mean seedling emergence across three planting years when precipitation was held constant 

for (A) sown and (B) non-sown species. ........................................................................................ 34 

Figure 2.5. (A) Interaction between sown richness and planting year on sown species emergence under 

ambient conditions. (B) Interaction between sown richness and precipitation treatment on sown 
species emergence. ........................................................................................................................ 35 

Figure 2.6. Nonmetric multidimensional scaling (NMDS) in three dimensions. ......................................... 36 

Figure 2.7. Seedling emergence across three planting years and precipitation treatments for the three 

most abundant non-sown (Panicum capillare, Digitaria ischaemum, and Centaurea stoebe) and 
three most abundant sown (Bouteloua curtipendula, Echinacea purpurea, and Dalea purpurea) 
species. ........................................................................................................................................... 37 

Figure 2.8. Rain-out shelter design. ............................................................................................................ 38 

Figure 3.1. Non-metric Multidimensional Scaling (NMDS) of plots planted in 2014, 2015, and 2016 over 

time. ............................................................................................................................................... 69 

Figure 3.2 Photos taken in 2017 of prairie plots planted in (A) 2014, (B) 2015, and (C) 2016. .................. 70 

Figure 3.3. Sown species cover over time in plots that were exposed to ambient conditions in their first 

year. ............................................................................................................................................... 71 

Figure 3.4. Sown species richness over time in plots that were exposed to ambient conditions in their 

first year. ........................................................................................................................................ 72 

Figure 3.5. Non-sown species cover over time in plots that received watering treatments in their first 

year. ............................................................................................................................................... 73 

Figure 4.1. Map of study sites. .................................................................................................................. 111 

Figure 4.2 Photos of restored prairies. ..................................................................................................... 112 

Figure 4.3. Scatter plots and pairwise correlations (r) between weather variables. ................................ 113 

 

ix 

Figure 4.4. Significant effects of planting year weather on restoration outcomes. ................................. 114 

Figure 4.5. The effect of planting year temperature (weather PC1) on 2016 non-sown species cover 

(mean 1 m2 plot) varied depending on site age. .......................................................................... 115 

Figure 4.6. The relative importance of weather conditions and other factors on different measures of 

restoration outcomes across 83 restored prairies in the Midwestern USA. ............................... 116 

Figure 4.7. Scatter plots of the three PC axis rotations of weather variables against each prairie assembly 
predictor. ..................................................................................................................................... 117 

Figure 4.8. Distributions of weather and site condition variables across sites. ....................................... 119 

 

 

 

x 

CHAPTER ONE 

PLANTING YEAR EFFECTS IN RESTORATION 

"If the science can't predict the outcome of a restoration effort with  

some confidence, it clearly has a long way to go." -John Cairns, 1987 

 

INTRODUCTION 

 

What’s in a planting year? Any farmer or restoration practitioner might tell you, a lot. Weather 

happens, and weather affects plants. But can weather events in one important year shape a plant 

community for the long term? How much do planting conditions influence efforts to restore ecological 

communities that have been degraded or destroyed? 

Ecological restoration has the potential to slow or reverse the loss of global biodiversity, whose 

greatest threat is the alteration of ecosystems for human land use (Foley et al. 2005; Cardinale et al. 

2012). Restored systems, like the natural systems they strive to emulate, can provide many functions 

and services that benefit humankind such as habitat for crop pollinators, storm water filtration and 

retention, or carbon sequestration (Dobson et al. 1997; Choi 2007; Bullock et al. 2011). But this recovery 

of biodiversity and provisioning of ecosystem services relies on the success of restoration practitioners 

in meeting their goals; for instance, establishing the intended community. In reality, restoration 

outcomes can vary widely and often differ from goals (Hobbs & Norton 1996; Matthews & Spyreas 2010; 

Suding 2011). A better understanding of the determinants of restoration outcomes is essential to 

achieve the expected benefits of restoring native ecosystems. 

Ecological restoration is often focused on re-establishing native plant communities to restore 

the functionality of the target system (Ruiz-Jaen & Mitchell Aide 2005; Brudvig 2011). Successful 

restoration therefore requires a clear understanding of the structure of the target community (e.g., 

 

1 

relative abundances of desired species) as well as methods for establishing this community in context of 

the current conditions at a particular site. Ecologists have been striving to understand the processes that 

drive the establishment and structure of communities for over a century (Cowles 1899; Clements 1916; 

Weiher et al. 2011); therefore, ecological theories on how communities assemble should be able to 

inform restoration practices aimed at reassembling native communities (Palmer et al. 1997; Young et al. 

2001, 2005). Restoration, in turn, provides opportunities to test theories of how ecological communities 

assemble, under different field settings (Bradshaw 1987). Multiple ecological theories, invoking a variety 

of mechanisms such as ecological filters (Belyea 2004), assembly rules (Diamond 1975, Weiher and 

Keddy 1995), and succession (Pickett et al. 2001; White & Jentsch 2004) have been advanced to explain 

patterns of community assembly in restoration (e.g., Temperton et al. 2004). However, the continued 

unpredictability of restoration outcomes implies that no currently implemented theory provides a fully 

workable understanding of how communities assemble during restoration. 

 

Historically contingent assembly models  

A missing consideration in these models of community assembly may be the way in which the 

development of a community, or its history, can affect the subsequent community state (e.g., Drake 

1991). These ideas can be traced back to early assembly models where “accidents” of seed dispersal 

were more important than the overall system (Gleason 1926) and where initial floristic composition 

shaped communities (Egler 1954); they also have strong parallels to theories of succession (e.g., Connell 

& Slatyer 1977). These historically contingent assembly models, in which communities are structured by 

events such as the arrival order of species during assembly, are now prevalent in the community 

assembly literature (Drake 1990; Law & Morton 1993; Chase 2003; Fukami et al. 2005), but have 

received relatively little attention in a restoration context (Brudvig 2011). Microcosm experiments and 

computer simulations modeling temporal events that could affect community assembly have indicated 

 

2 

that the identity of community members early in assembly can affect later community composition, 

even when early community members do not persist in the community (Lockwood & Samuels 2004). 

Species that become established early can also have a greater chance of remaining competitively 

dominant than later arrivers (i.e. “priority effects”, Alford and Wilbur 1985). It follows that different 

ecological communities can develop under otherwise identical environmental conditions due to 

differences in the order in which species colonize a space (Gleason 1926; Egler 1954; Sutherland 1974; 

Robinson & Edgemon 1988; Fukami et al. 2005). 

This case for historically contingent assembly seems particularly strong for restoration projects 

that are initiated through a single sowing (e.g., most tallgrass prairies), as many environmental factors 

that could impact seedling emergence and survival are highly variable from year to year. For these 

communities, temporally variable abiotic or biotic conditions such as temperature, precipitation, 

competitive dynamics, or predator pressures may control establishment order by favoring or disfavoring 

species at this critical early stage of the assembly process. Some evidence has shown that inter-annual 

environmental variation during the first growing season of a restored community can have persistent 

effects both in restoration projects (Bakker et al. 2003; MacDougall et al. 2008; Stuble et al. 2017) and 

ecological experiments in general (Johnson 2002, reviewed in Vaughn and Young 2010), but the full 

scope of these effects remains unclear because few studies of experimental restoration are replicated 

across years (Vaughn & Young 2010). As a result, much evidence for year effects on restoration 

outcomes is correlational and confounded with differences in management practices (e.g., seed mix 

diversity/composition) and site-specific conditions (e.g., edaphic conditions). To better understand 

variation in the composition of restored communities and increase our understanding of the importance 

of historically contingent assembly models, we need empirical evidence to determine how and when 

establishment history, and specifically year effects, influence community assembly and restoration 

outcomes. 

 

3 

Testing year effects in restored communities 

For my dissertation, I investigate the effects of planting year conditions on restored plant 

communities. Throughout, I evaluate the interaction between year effects and a major management 

decision for prairie restoration practitioners: the richness of the seed mix. In the chapters that follow, I 

examine: (I) mechanisms that shaped plant establishment in three planting years in an experimental 

restored community; (II) whether differences from the first growing season in these experimental 

communities persisted over time; and (III) the relative importance of planting year weather conditions 

for restored prairies of various ages and management histories across the Midwest. These three 

chapters provide three important lines of evidence that together make for a strong test of whether year 

effects, disentangled from site, management, and other factors, are important drivers of variation in 

restoration outcomes. 

In chapter 2 of this dissertation, I examined mechanisms of year effects. Evidence is needed to 

demonstrate that year effects occur without confounding factors and to elucidate the conditions and 

subsequent processes that could cause a year effect. For instance, warmer temperatures, especially in 

early mornings or springtime, can advance the phenology of some plant species (Forrest & James 2011), 

which can favor certain species over others (Wainwright & Cleland 2013). Precipitation, too, has known 

impacts on plant communities (Silvertown et al. 1999; Robinson & Gross 2010). Populations of rodents 

can fluctuate between years (e.g., Getz and Hofmann 1999), which can in turn impact the establishing 

plant community through post-dispersal seed predation (Hulme & Kollmann 2005; Notman & Villegas 

2005) or herbivory (Howe & Brown 1999; Howe 2008; MacDougall & Wilson 2007). Knowing the extent 

to which temporal variation in environmental factors impact community assembly during restoration 

has obvious practical significance, especially when coupled with an understanding of the actual process 

by which the plant community is affected (e.g., whether the condition affects germination, seedling 

survival, or adult growth). To address this in my second dissertation chapter, I established experimental 

 

4 

prairie plots in three planting years, using identical methods. In each year, I established three rainfall 

treatments, spanning the historical levels of precipitation in this area over the past 30 years (record low, 

1.39 cm/month; mean, 9.91 cm/month; and 93rd percentile, 17.81 cm/month). The goal was to 

disentangle to effects of inter-annual precipitation—my leading hypothesis— from other factors that 

may vary annually. I tracked seedling emergence and survival in the first growing season in each year to 

determine whether seedling emergence or survival is most affected by conditions at the time of 

planting, and finally compared the composition of the communities that developed in each first growing 

season. This investigation of the planting-year community provides a mechanistic explanation for how 

year effects could be introduced to a restored community.  

To determine what, if any, differences in the early community persist over time, I monitored the 

community composition of the experimental prairie plots outlined above for an additional 1-3 years 

(depending on planting year). These results are reported in Chapter 3. If the communities that 

developed in subsequent years remained different over time, then this provides evidence that assembly 

is historically contingent, and year effects can impact longer-term restoration outcomes. Otherwise, the 

communities will have converged (i.e. become more similar) through time regardless of initial 

differences, due to the shared conditions (e.g. soils, invasive species pressures, and management), 

supporting a deterministic assembly model in which year effects do not impact restoration outcomes. 

Because the above experiment was conducted at a single site in Michigan, a critical final 

consideration is whether there are detectable effects of initial planting conditions on real restored 

prairie sites across the Midwest. To address this question, in my fourth dissertation chapter, I report 

results from field surveys of 83 restored prairies located across the Midwest, ranging in age from 2-19 

years old. At each site I collected data on soils (e.g. water holding capacity, nutrient composition) and 

management histories (e.g. seed mix sown, prescribed fire history) and determined first growing season 

weather conditions using PRISM online database. This allowed me to evaluate the effect of first growing 

 

5 

season planting conditions on the present-day plant community, with interest in their effects relative to 

other factors known to structure assembling restoration communities. 

Finally, the impact of a year effect may vary depending on the specific methods used to 

establish a restoration in the first year, especially the species richness of the seed mix used. Sowing a 

higher richness seed mix may buffer against adverse year effects; that is, the impact of poor planting 

year conditions on restoration success may be less in high than in low diversity prairies. This buffering 

would occur when a greater selection of species increases the likelihood that a seed mix includes species 

that can establish given the specific environmental conditions (i.e., through a portfolio/insurance effect, 

Tilman and Downing 1994, Ives et al. 1999, Tilman et al. 2006, Hector et al. 2010). If the species richness 

of seed mix does not interact with a year effect, the community should more closely resemble targets 

(i.e., higher native cover and diversity—common metrics of restoration success) in high than in low 

diversity plantings regardless of initiation year. I address this question throughout my dissertation; by 

crossing a high and low richness seed mix throughout my experiment, and by considering the richness of 

the seed mixes sown throughout the sampled Midwestern prairies in their analysis. 

 

Study system 

I conducted the above restoration experiments and observational studies in a tallgrass prairie 

restoration system. Nearly 99.9% of native prairie has been lost since European settlement of North 

America, making this system a high priority for restoration (Samson & Knopf 1994). Decades to centuries 

of intensive agriculture on most former prairie sites has resulted in a complete loss of remnant prairie 

species in the seed bank or in the landscape that could naturally recolonize a site following agricultural 

abandonment, therefore prairie restoration projects require seed sowing to establish new communities 

(Thompson et al. 1998; Suding & Gross 2006; Cramer et al. 2008). This restoration by seed sowing allows 

for explicit manipulation of arrival order of species to a community. The fine-scale community size of a 

 

6 

prairie, plus the relatively fast maturation and manageable size of prairie plants, combine to make 

prairies a prime system for experimentation with community assembly mechanisms (Jordan 1997, e.g., 

Foster et al. 2007, Dickson et al. 2012, Martin and Wilsey 2012). Prairie restoration research can 

therefore be beneficial both to basic ecological knowledge as well as to active conservation practice. 

 

 

7 

LITERATURE CITED

 

8 

LITERATURE CITED 

 

Alford RA, Wilbur HM (1985) Priority effects in experimental pond communities: Competition between 

Bufo and Rana. Ecology 66:1097–1105 

Bakker JD, Wilson SD, Christian JM, Li X, Ambrose LG, Waddington J (2003) Contingency of grassland 

restoration on year, site, and competition from introduced grasses. Ecological Applications 
13:137–153 

Belyea LR (2004) Beyond ecological filters: Feedback networks in the assembly and restoration of 

community structure. In: Assembly rules and restoration ecology: Bridging the gap between 
science and practice. Temperton, VM, Hobbs, RJ, Nuttle, T, & Halle, S, editors. Island Press. 

Bradshaw AD (1987) Restoration: An acid test for ecology. In: Restoration Ecology: A synthetic approach 
to ecological research. Jordan, WRI, Gilpin, ME, & Aber, JD, editors. Cambridge University Press 
pp. 22–29. 

Brudvig LA (2011) The restoration of biodiversity: Where has research been and where does it need to 

go? American Journal of Botany 98:549–558 

Bullock JM, Aronson J, Newton AC, Pywell RF, Rey-Benayas JM (2011) Restoration of ecosystem services 

and biodiversity: Conflicts and opportunities. Trends in Ecology and Evolution 26:541–9 

Cardinale BJ, Duffy JE, Gonzalez A, Hooper DU, Perrings C, Venail P, Narwani A, Mace GM, Tilman D, 
Wardle DA, Kinzig AP, Daily GC, Loreau M, Grace JB, Larigauderie A, Srivastava DS, Naeem S 
(2012) Biodiversity loss and its impact on humanity. Nature 486:59–67 

Chase JM (2003) Community assembly: When should history matter? Oecologia 136:489–98 

Choi YD (2007) Restoration ecology to the future: A call for new paradigm. Restoration Ecology 15:351–

353 

Clements FE (1916) Plant Succession: An Analysis of the Development of Vegetation. Carnegie Institution 

of Washington, Washington 

Connell JH, Slatyer RO (1977) Mechanisms of succession in natural communities and their role in 

community stability and organization. The American Naturalist 111:1119–1144 

Cowles HC (1899) The Ecological Relations of the Vegetation on the Sand Dunes of Lake Michigan (Part 

III). Botanical Gazette 27:281–308 

Cramer VA, Hobbs RJ, Standish RJ (2008) What’s new about old fields? Land abandonment and 

ecosystem assembly. Trends in Ecology and Evolution 23:104–12 

Diamond JM (1975) Assembly of species communities. In: Ecology and evolution of communities. Cody, 

M & Diamond, JM, editors. Belknap Press of Harvard University Press, Cambridge, 
Massachusetts pp. 342–444. 

 

9 

Dickson TL, Hopwood JL, Wilsey BJ (2012) Do priority effects benefit invasive plants more than native 

plants? An experiment with six grassland species. Biological Invasions 14:2617–2624 

Dobson AP, Bradshaw AD, Baker AJM (1997) Hopes for the future: Restoration ecology and conservation 

biology. Science 277:515–522 

Drake J a. (1990) The mechanics of community assembly and succession. Journal of Theoretical Biology 

147:213–233 

Drake JA (1991) Community-Assembly Mechanics and the Structure of an Experimental Species 

Ensemble. The American Naturalist 137:1–26 

Egler FE (1954) Vegetation Science Concepts I. Initial Floristic Composition, a Factor in Old-Field 

Vegetation Development. Vegetatio 4:412–417 

Foley JA, DeFries R, Asner GP, Barford C, Bonan G, Carpenter SR, Chapin FS, Coe MT, Daily GC, Gibbs HK, 
Helkowski JH, Holloway T, Howard EA, Kucharik CJ, Monfreda C, Patz JA, Prentice IC, Ramankutty 
N, Snyder PK (2005) Global consequences of land use. Science 309:570–574 

Forrest JRK, James DT (2011) An examination of synchrony between insect emergence and flowering in 

Rocky Mountain meadows. Ecological Monographs 81:469–491 

Foster BL, Murphy C a., Keller KR, Aschenbach T a., Questad EJ, Kindscher K (2007) Restoration of prairie 

community structure and ecosystem function in an abandoned hayfield: A sowing experiment. 
Restoration Ecology 15:652–661 

Fukami T, Martijn Bezemer T, Mortimer SR, Putten WH (2005) Species divergence and trait convergence 

in experimental plant community assembly. Ecology Letters 8:1283–1290 

Getz LL, Hofmann JE (1999) Diversity and stability of small mammals in tallgrass prairie habitat in central 

Illinois, USA. Oikos 85:356–363 

Gleason HA (1926) The individualistic concept of the plant association. Bulletin of the Torrey Botanical 

Club 53:7–26 

Hector S, Hautier Y, Saner P, Wacker L, Bagchi R, Joshi J, Scherer-Lorenzen M, Spehn EM, Bazeley-White 

E, Weilenmann M, Caldeira MC, Dimitrakopoulos PG, Finn JA, Huss-Danell K, Jumpponen A, 
Mulder CPH, Palmborg C, Pereira JS, Siamantziouras ASD, et al. (2010) General stabilizing effects 
of plant diversity on grassland productivity through population asynchrony and overyielding. 
Ecology 91:2213–2220 

Hobbs RJ, Norton DA (1996) Towards a conceptual framework for restoration ecology. Restoration 

Ecology 4:93–110 

Howe HF (2008) Reversal of fortune: Plant suppression and recovery after vole herbivory. Oecologia 

157:279–86 

Howe HF, Brown JS (1999) Effects of Birds and Rodents on Synthetic Tallgrass Communities. Ecology 

80:1776–1781 

 

10 

Hulme PE, Kollmann J (2005) Seed predator guilds, spatial variation in post-dispersal seed predation and 

potential effects on plant demography: A temperate perspective. In: Seed Fate. Predation, 
dispersal and seedling establishment. pp. 9–30. 

Ives AR, Gross K, Klug JL (1999) Stability and variability in competitive communities. Science (New York, 

N.Y.) 286:542–544 

Johnson DH (2002) The importance of replication in wildlife research. The Journal of Wildlife 

Management 66:919–932 

Jordan III WR (1997) Forward. In: The Tallgrass Restoration Handbook. Packard, S & Mutel, CF, editors. 

Island Press pp. xiii–xviii. 

Law R, Morton RD (1993) Alternative permanent states of ecological communities. Ecology 74:1347–

1361 

Lockwood JL, Samuels CL (2004) Assembly models and the practice of restoration. In: Assembly Rules 

and Restoration Ecology. Temperton, VM, Hobbs, RJ, Nuttle, T, & Halle, S, editors. Island Press, 
Washington, USA pp. 55–70. 

MacDougall AS, Wilson SD (2007) Herbivory limits recruitment in an old-field seed edition experiment. 

Ecology 88:1105–1111 

MacDougall AS, Wilson SD, Bakker JD (2008) Climatic variability alters the outcome of long-term 

community assembly. Journal of Ecology 96:346–354 

Martin LM, Wilsey BJ (2012) Assembly history alters alpha and beta diversity, exotic-native proportions 

and functioning of restored prairie plant communities Cadotte, M, editor. Journal of Applied 
Ecology 

Matthews JW, Spyreas G (2010) Convergence and divergence in plant community trajectories as a 

framework for monitoring wetland restoration progress. Journal of Applied Ecology 47:1128–
1136 

Notman EM, Villegas AC (2005) Patterns of seed predation by vertebrate versus invertebrate seed 
predators among different plant species, seasons and spatial distributions. In: Seed Fate: 
Predation, Dispersal and Seedling Establishment. pp. 55–75. 

Palmer MA, Ambrose RF, Poff NL (1997) Ecological Theory and Community Restoration Ecology. 

Restoration Ecology 5:291–300 

Pickett STA, Cadenasso ML, Bartha S (2001) Implications from the Buell-Small Succession Study for 

vegetation restoration. Applied Vegetation Science 4:41–52 

Robinson J V, Edgemon MA (1988) An Experimental Evaluation of the Effect of Invasion History on 

Community Structure. Ecology 69:1410–1417 

Robinson TMP, Gross KL (2010) The impact of altered precipitation variability on annual weed species. 

American journal of botany 97:1625–9 

Ruiz-Jaen MC, Mitchell Aide T (2005) Restoration success: How is it being measured? Restoration 

Ecology 13:569–577 

 

11 

Samson F, Knopf F (1994) Prairie conservation in North America. BioScience 44:418–421 

Silvertown J, Dodd M, Gowing DJ, Mountford JO (1999) Hydrologically defined niches reveal a basis for 

species richness in plant communities. Nature 400:61–64 

Stuble KL, Fick SE, Young TP (2017) Every restoration is unique: Testing year effects and site effects as 

drivers of initial restoration trajectories. Journal of Applied Ecology 54:1051–1057 

Suding KN (2011) Toward an era of restoration in ecology: Successes, failures, and opportunities ahead. 

Annual Review of Ecology, Evolution, and Systematics 42:465–487 

Suding KN, Gross KL (2006) Modifying native and exotic species richness correlations: The influence of 

fire and seed addition. Ecological Applications 16:1319–1326 

Sutherland JP (1974) Multiple Stable Points in Natural Communities. The American Naturalist 108:859–

873 

Temperton VM, Hobbs RJ, Nuttle T, Halle S, eds. (2004) Assembly rules and restoration ecology: Bridging 

the gap between theory and practice. Island Press 

Thompson K, Bakker JP, Bekker RM, Hodgson JG (1998) Ecological correlates of seed persistence in soil 

in the north-west European flora. Journal of Ecology 86:163–169 

Tilman D, Downing J a. (1994) Biodiversity and stability in grasslands. Nature 367:363–365 

Tilman D, Reich PB, Knops JMH (2006) Biodiversity and ecosystem stability in a decade-long grassland 

experiment. Nature 441:629–32 

Vaughn KJ, Young TP (2010) Contingent conclusions: Year of initiation influences ecological field 

experiments, but temporal replication is rare. Restoration Ecology 18:59–64 

Wainwright CE, Cleland EE (2013) Exotic species display greater germination plasticity and higher 

germination rates than native species across multiple cues. Biological Invasions 15:2253–2264 

Weiher E, Freund D, Bunton T, Stefanski A, Lee T, Bentivenga S (2011) Advances, challenges and a 

developing synthesis of ecological community assembly theory. Philosophical Transactions of 
the Royal Society B: Biological Sciences 366:2403–13 

Weiher E, Keddy PA (1995) The assembly of experimental wetland plant communities. Oikos 73:323–335 

White P, Jentsch A (2004) Disturbance, succession, and community assembly in terrestrial plant 

communities. In: Assembly rules and restoration ecology: Bridging the gap between science and 
practice. Temperton, VM, Hobbs, RJ, Nuttle, T, & Halle, S, editors. Island Press. 

Young TP, Chase JM, Huddleston RT (2001) Community succession and assembly. Ecological Restoration 

19:5–18 

Young TP, Petersen DA, Clary JJ (2005) The ecology of restoration: historical links, emerging issues and 

unexplored realms. Ecology Letters 8:662–673  

 

 

 

12 

CHAPTER TWO 

INTER-ANNUAL VARIATION IN PRECIPITATION AND OTHER PLANTING CONDITIONS  

IMPACTS SEEDLING ESTABLISHMENT IN SOWN PLANT COMMUNITIES 

 

ABSTRACT 

Ecological restoration can reverse biodiversity loss worldwide, yet restoration goals and 

outcomes vary widely, which limits this potential. Divergent restoration outcomes may stem from 

variation in conditions at the outset of restoration, but empirical evidence is lacking and typically 

confounded with site differences. Additionally, precipitation is usually cited as the source of this 

variation, although a wide range of conditions can vary annually. We tested for effects of planting year 

on seedling establishment by installing identical restorations in three different years. Within those years, 

we manipulated rainfall with rain-out shelters to disentangle the effects of precipitation from other 

annually variable conditions. Additionally, we tested whether increasing seed mix richness buffers 

against adverse planting conditions. For the first growing season after planting, we followed emergence 

and survival of sown prairie species and non-sown weed species to determine how planting year 

conditions influence an establishing plant community, if at all. We found that seedling establishment 

differed across planting years and precipitation treatments, and that varying emergence patterns by 

species led to differences in the composition of the first-year community. We also found significant 

variation in sown species establishment across years when precipitation was held constant, illustrating 

the previously overlooked role of non-precipitation drivers on planting year effects. Higher seed mix 

richness did not consistently improve establishment of sown species under different planting conditions. 

This research provides important experimental evidence for effects of inter-annual variation in planting 

conditions on first-year establishment. Future work will examine how these initial changes affect longer-

term assembly dynamics.  

 

13 

INTRODUCTION 

 

The greatest threat to global biodiversity is the alteration of ecosystems for human land uses 

(Foley et al. 2005; Cardinale et al. 2012). Ecological restoration has the potential to slow or reverse 

biodiversity loss in human-impacted ecosystems and landscapes (Dobson et al. 1997; Choi 2007; Bullock 

et al. 2011); however, this recovery is reliant on restoration efforts achieving these goals (Rowe 2010). 

For instance, ecological restoration often focuses on rebuilding plant communities to promote 

ecosystem functionality (Ruiz-Jaen & Mitchell Aide 2005; Brudvig 2011), yet restoration outcomes vary 

widely and often differ from goals (Hobbs & Norton 1996; Matthews & Spyreas 2010; Suding 2011). A 

critical step toward development of more predictably successful restoration practices is understanding 

the factors determining restoration outcomes (Brudvig et al. 2017). 

Variation in restoration outcomes may result from the numerous ways that restored sites differ 

from one another (e.g., soils, landscape context, etc.; e.g., Grman et al. 2013). However, a critical 

realization is that outcomes may differ among otherwise similar or even identical sites due to 

differences in environmental conditions at the outset of a restoration effort. These “year effects” relate 

to historically contingent theories of community assembly (Fukami 2015). For instance, microcosm 

experiments and computer simulations modeling temporal events that could affect community 

assembly have indicated that the identity of community members early in assembly can affect later 

community composition, either by remaining competitively dominant over later arrivers (i.e., “priority 

effects,” Shulman et al. 1983) or for other reasons even when early establishers do not persist in the 

community (Lockwood & Samuels 2004).  The implication for restoration is that conditions favoring the 

establishment of particular species early in restoration may subsequently influence the composition of 

the restored community (Young et al. 2001; Brudvig 2011; Miller et al. 2016). 

 

14 

In practice, understanding the influence of year effects on restoration is limited for several 

reasons. First, although agronomic studies are commonly replicated across years (Vaughn & Young 

2010), there is limited evidence for year effects in restoration and ecological experiments in general 

(Johnson 2002; Bakker et al. 2003; MacDougall et al. 2008; Young et al. 2015; Wilson 2015; Stuble et al. 

2017), because it is rare that restoration projects or ecological experiments are replicated across years 

(Vaughn & Young 2010). Second, most studies have evaluated year effects in restoration by comparing 

sites established in different years, resulting in confounding of years with differences in management 

practices (e.g., seed mix richness/composition) and site-specific conditions (e.g., edaphic factors). As 

such, results interpreted as year effects may in fact be management or site effects. Third, we currently 

have little understanding of the mechanisms underlying year effects. Previous studies have correlatively 

linked year effects with variable rainfall patterns (Bakker et al. 2003; MacDougall et al. 2008; Stuble et 

al. 2017); however, year effects could result from numerous factors spanning variation in rainfall, 

temperature, consumer abundances and identities, the abundances of competitors of focal restoration 

species, and many others. For instance, MacDougall et al. (2008) planted grassland species across 

multiple years in a restoration experiment, and found that 10 years later, plots sown in two normal 

precipitation years were dominated by natives, while plots seeded in a single dry year reverted to 

dominance by exotics. Although this relationship with planting-year precipitation is plausible, conditions 

in addition to precipitation could have varied across those years or sites.  

We tested for year effects in an experiment that controlled for site effects and manipulated the 

leading putative cause of year effects— inter-annual variation in precipitation. Specifically, we 

established sown prairie plots at a single site, using identical methods in each of three years. Within 

these plots, we manipulated precipitation during the planting year, resulting in plots established under 

identical amounts of precipitation, but in different years. This approach provided a strong test of 

precipitation as a determinant of year effects and, at the same time, considered the influences of non-

 

15 

precipitation causes of year effects (i.e., inter-annual variation not explained by the precipitation 

treatment). 

Our work also considered whether year effects might interact with aspects of management. In 

some restorations, all target species are sown at the same time, with a goal of maximizing sown species 

establishment while suppressing non-sown community members. In these systems, the plant 

community that establishes will be determined by the practitioner’s methods as well as the 

environmental conditions after sowing. Increasing the species richness of the seed mix is expected to 

increase the number of species suited to respond favorably to the environmental conditions of any 

given year (Tilman et al. 2014; Isbell et al. 2015) and therefore increase the overall establishment of 

target species. This should be particularly important under harsh conditions such as insufficient 

precipitation (Chase 2003), which may preclude establishment of many species. We therefore 

considered the role of propagule addition in our study, investigating whether a greater diversity of seeds 

used to establish a plant community would buffer against adverse year effects and allow for greater 

establishment of sown species. 

Finally, there is contradictory evidence for whether emergence versus seedling survival is most 

likely to limit establishment in sown systems and determine a species’ presence in the plant community. 

For example, in an arid grassland system, sown plant recruitment was limited by a bottleneck between 

germination and emergence (James et al. 2011), while Turnbull et al. (2000) found in a meta-analysis 

across systems that most establishment was limited at the transition from seedling to adult. 

Additionally, the ecologically relevant conditions that break seed dormancy and cue germination are 

highly dependent on climate (Walck et al. 2011), and sometimes the conditions favoring germination do 

not in turn favor the survival of seedlings (Lloret et al. 2009). Understanding the responses of target 

(sown) and non-target (non-sown) plants in restoration at these demographic transitions has important 

ramifications for management. For instance, if emergence rates and seedling mortality are high under 

 

16 

certain conditions, this would deplete the pool of sown propagules and managers should prepare to re-

sow certain restoration projects. In this way, responses of individual species to different planting 

conditions could determine the composition of the first-year community. This, in turn, may influence the 

composition of subsequent years, even when the first-year community does not itself resemble the later 

community.  

Empirical evidence is needed to clearly disentangle the mechanisms by which precipitation and 

other factors that vary inter-annually influence the establishment of a sown restoration. This research 

addressed four main questions: (1) Do emergence and/or seedling survival vary in different planting 

years? We predicted significant variation in the emergence and survival of seedlings across three 

planting years. (2) Do these processes vary due to precipitation, or non-precipitation conditions? We 

expected both seedling emergence and survival to increase with increasing precipitation after planting. 

(3) What is the influence of sown species richness on the above processes? We hypothesized a higher 

seed mix richness to buffer against adverse planting conditions, leading to higher seedling 

establishment. Finally, (4) do different planting conditions result in different first-year communities? We 

predicted variation in species responses to planting year conditions and precipitation treatments would 

result in different composition of first-year communities. For each of these questions, we considered 

sown and non-sown species separately, owing to their different impacts on restoration outcomes and 

potential for these two groups to respond differently to planting year effects. 

 

METHODS 

 

Study site and restoration methods 

This experiment was conducted at Michigan State University’s Kellogg Biological Station, at the 

Lux Arbor Reserve near Prairieville, MI. Tallgrass prairie was a historical ecosystem type in this region 

 

17 

and restored prairies are prevalent in this region (Grman et al. 2014). The site was historically mixed-oak 

savanna (Comer et al. 1995), was cleared for cultivation prior to 1946, used as a private air-strip in the 

1960s and 1970s, abandoned sometime before 1980, and has been mowed approximately annually to 

prevent woody encroachment. Before this experiment, vegetation at the site closely resembled other 

abandoned fields in the area, and was dominated by Bromus inermis (Smooth brome), Centaurea stoebe 

(Spotted knapweed), and other exotic and native weedy perennials. 

We converted a subset of this abandoned field to prairie in each year (2014-2016) of the 

experiment. Each May, we mowed plots to stimulate growth, sprayed vegetation with glyphosate 

herbicide, tilled, hand broadcasted native prairie seeds, and packed with a cultipacker to increase soil-

seed contact and probability of germination. We cultipacked on 19-20 May each planting year. 

 

Experimental design 

To test how restorations are impacted by planting year conditions, precipitation treatment, and 

seed mix richness, we utilized a randomized complete block design (n=6 for each year by precipitation 

by seed mix treatment, 24 combinations total, Figure 2.1). We installed six replicate prairie blocks (15.2 

m x 36.6 m) in three planting years (2014-2016); each sown with a baseline low richness prairie mix (8 

species, 379 seeds/m2) to serve as a buffer around experimental plots. Within each block, we installed 

four experimental plots (1.0 m x 0.5 m), to which four whole-plot factor treatments (ambient conditions; 

low, mean, and high precipitation) were randomly assigned.  

Finally, each experimental plot was split, with each half (0.5 m x 0.5 m) receiving additional 

seeds of either a high or low richness seed mix. The low richness plot received an additional 379 

seeds/m2 of the 8 species mix already sown across the entire block (total sown density, 758 seeds/m2). 

The high richness plot received 379 seeds/m2 of a 22 species mix, bringing total sown richness to 30 

species (total sown density, 758 seeds/m2). We chose species for the seed mixes that were typical of 

 

18 

low- and high-diversity plantings in this region. The high and low richness seed mixes contained the 

same proportions of seeds from different life-form groups (i.e., forb, legume, C3 grass, C4 grass). Due to 

a seed supplying error, 2016 plots were sown with Echinacea pallida instead of E. purpurea; these two 

species have been combined for analysis, though no Echinacea emerged in 2016. Each species was 

tested for viability each year by the seed supplier and mixes were adjusted such that each planting 

received the same quantity of live seed (Table 2.1). 

 

Precipitation treatments 

In order to simulate drought, normal, and wet year conditions relevant for this field site, we 

used the 30-year monthly averages for May-July from the nearest weather station (Gull Lake, MI, 10 km 

from field site) to determine frequency (8 days/month) and quantity of watering. The three treatments 

were based on: the record low (1.39 cm = 13.8 L/m2 monthly), the 30-year mean (9.91 cm = 99.1 L/m2 

monthly), and the 93rd percentile (17.81 cm = 178.1 L/m2 monthly). Precipitation treatments were 

rounded to the nearest gallon to facilitate compatibility with the watering cans used.  

To implement precipitation treatments, we diverted and collected all rainfall from the 

experimental plots for the first nine weeks after sowing using rain-out shelters (Figure 2.8). To meet 

precipitation treatments, rainwater was supplemented with well water. Frequency of water application 

was held constant throughout the treatment period (two times/week; every 2-5 days); all treatments 

were watered on the same day. We installed rubber gardening edging around experimental plots to 

prevent overland flow of water in the precipitation treatment plots and to permanently mark the plots.  

Sediment collecting on rain-out shelters reduced light availability by 19 + 1% by the end of July when 

shelters were taken down each year. Although rain-out shelters may have had impacts on plot-level 

microclimate (e.g. slightly increasing temperature through greenhouse effects), this is not a concern for 

 

19 

interpreting precipitation treatment results since all plots would have been similarly affected and plots 

with and without rain-out shelters are treated separately in all statistical analyses. 

 

Seedling and plant community surveys 

We conducted plant surveys in the plots that were sown each year (2014-2016) to examine how 

seedling emergence, seedling survival, and the first-year community varied in different planting years, 

precipitation treatments, and levels of seed mix richness. Within each experimental split-plot, three 10 x 

10-cm sub-sampling plots were permanently marked and individual seedlings were marked with color-

coded sewing pins for week-to-week tracking. In 2014, seedling surveys were conducted weekly for the 

first nine weeks after sowing (30 May-24 July) to measure seedling emergence and seedling death rates 

of all sown and non-sown species in each sub-sampling plot. We used these data to determine the 

minimum sampling required to account for 87.5% of all seedlings, and thereafter surveyed 2015 and 

2016 seedlings in weeks four (mid-June) and nine (late July) and subset 2014 data accordingly for all 

analyses. At peak growing season (late July) each planting year, we visually estimated percent cover of 

all species in each 0.5 x 0.5-m split-plot. 

Finally, although all plots were mowed in late July of their planting year to prevent recruitment 

between planted areas and neighboring unplanted areas, a non-sown annual grass Panicum capillare set 

seed a second time after mowing and before the growing season was over. P. capillare therefore 

showed a marked increase in emergence with each subsequent planting year, making up 11.1% of all 

seedlings in 2014, 29.9% of all seedlings in 2015, and 71.2% of all seedlings in 2016. We believe this 

increase reflected increasing propagule pressure—an artifact of the experimental design—rather than a 

true year effect. We therefore conducted all analyses both with and without P. capillare included. We 

found that key results were consistent across the two methods; results presented here exclude P. 

 

20 

capillare (leaving 1966 total seedlings) in order to better represent true year effects, with differences 

noted in table legends. Results including P. capillare are available in Table 2.2.  

 

Statistical analyses 

We used generalized linear mixed models to examine the fixed effects of first-year precipitation 

treatment, planting year, and their interactions with sown richness on the emergence and survival of all 

sown species, all non-sown species, and the three most abundant species from each of those two 

groups (separate models for each of these groupings; package “lme4”, Bates et al. 2015). For D. 

ischaemum, B. curtipendula, and E. purpurea, species models did not include interaction effects to allow 

for model convergence. Ambient plots were analyzed separately from rain-out shelter plots to test for 

naturally occurring planting year effects and to avoid confounding effects of rain-out shelter design. 

Models of sown species emergence used a binomial distribution to fit the number of seedlings and the 

number of non-emerged sown seeds. Models of seedling survival used a binomial distribution to fit the 

number of survivors and the number of deaths. All other emergence models used a Poisson distribution 

to fit the counts of seedlings. In models using ambient (no rain-out shelter) data, a random effect of 

block was added to control for heterogeneity across the experimental site; for rain-out data, a nested 

random effect of Block/Treatment was used. We compared within groups using Tukey adjusted post-hoc 

tests. 

To test for community-level differences among planting year, precipitation treatment, and sown 

richness, we used permutational multivariate analysis of variance (PERMANOVA, Anderson 2001) with 

Bray-Curtis dissimilarity and stratified by planting block (package “vegan,” Oksanen et al. 2017). We 

used nonmetric multidimensional scaling (NMDS) in three dimensions also with Bray-Curtis dissimilarity 

to visualize differences in plant communities across planting years and precipitation treatments 

 

21 

(stress=0.19). All analyses were conducted in R version 3.3.2 (Fox and Weisberg 2011; Bates et al. 2015; 

R Core Team 2016). 

 

RESULTS 

 

A total of 3771 seedlings spanning 57 species were surveyed, of which 2860 were non-sown 

volunteers (38 species) and 461 were sown in this experiment (15 species). We were unable to identify 

246 seedlings (6.5%) to species and therefore they were not classified as “sown” or “non-sown.” An 

additional 204 individuals (5.4%) belonged to the genus Elymus which we dropped from the “sown” and 

“non-sown” analyses because we could not confidently distinguish between the sown Elymus 

canadensis and non-sown Elymus repens as small seedlings. 

 

Ambient conditions 

There was a year effect on seedling emergence and survival in plots exposed to ambient 

planting conditions. Sowing in three different years resulted in different emergence rates for both sown 

(χ2

(2,25)=37.4, p<0.0001) and non-sown species (χ2

(2,28)=37.5, p<0.0001). Sown prairie seedlings had 2.3 

times higher emergence rates in 2015 (32.6% of seeds sown) than in 2014 (11.6% of seeds sown, z=-5.8, 

p<0.0001) or 2016 (16.7% of seeds sown, z=4.5, p<0.0001) (Figure 2.2). Similarly, non-sown species had 

2.0 times emergence in 2015 than in 2014 (z=-4.5, p<0.0001) or 2016 (z=5.4, p<0.0001) (Figure 2.2). 

There was not a significant difference in emergence between 2014 and 2016 for sown (z=-1.4, p=0.35; 

however, when P. capillare included, z=-12.7, p<0.0001) or non-sown (z=1.0, p=0.55) species. 

Seedling survival rates were high (93.6% of 608 seedlings survived in their first year in ambient 

plots (Table 2.3). Only 5 (of 168) sown seedlings died across all ambient plots in this experiment; 

therefore, statistical results are not presented. For non-sown species, total first year survival was 

 

22 

significantly different across years (χ2

(2,28)=20.7, p<0.0001; however, when P. capillare included, 

χ2

(2,28)=2.0, p=0.37). There was 2.6 times more seedling mortality in 2016 than in 2014 (z=-3.2, p=0.004) 

or 2015 (z=-4.1, p=0.0001). There was not a significant difference in non-sown seedling mortality 

between 2014 and 2015 (z=0.67, p=0.78).  

There was a significant interaction between seed mix richness and planting year in ambient plots 

(χ2

(2,25)=6.4, p=0.04, Figure 2.5a). The high richness mix led to 2.3 times more sown seedlings in 2016 

(z=2.87, p=0.0041), but no significant difference between the two seed mixes in 2014 (z=-0.76, p=0.45) 

or 2015 (z=1.04, p=0.30). Full statistical results for the above are presented in Table 2.4. 

Finally, there was a significant effect of planting year on community composition under ambient 

conditions (PERMANOVA: F(2,17)=1.9, p=0.004), because species varied in their emergence and 

abundance in different years (see Species Responses below). 

 

Precipitation treatments 

Precipitation treatment had a significant effect on seedling emergence of sown (χ2

(2,83)=25.6, 

p<0.0001) and non-sown species (χ2

(2,84)=19.0, p<0.0001) in rain-out shelter plots (Figure 2.3). As 

expected, more seedlings emerged as water increased; however, both groups only responded 

significantly to the increase from drought to average water levels (sown species, 2.3 times more 

emergence in mean than low treatment, z=-3.6, p=0.001; non-sown species, 1.9 times, z=-3.0, p=0.009). 

The strength of these effects on non-sown species varied with year (χ2

(4,84)=39.4, p<0.0001), with 

significant effects of watering on emergence in 2015 and 2016 but not 2014 (Table 2.5). Seedling 

survival was not affected by precipitation treatment for sown (χ2

(2,80)=1.17, p=0.56) or non-sown species 

(χ2

(2,80)=5.91, p=0.05, Table 2.3).  

Rain-out plots also exhibited a significant effect of planting year on sown seedling emergence, 

meaning there was still a year effect on sown species when precipitation was held constant across years 

 

23 

(i.e., differences existed among plots receiving the same rainfall treatment, but in different years; 

hereafter 'non-precipitation year effects') (χ2

(2,83)=51.9, p<0.0001, Figure 2.4). Sown species emergence 

was again 2.2 times higher in 2015 than in 2014 (z=-5.7, p<0.0001) or 2016 (z=5, p<0.0001) under 

constant precipitation (Figure 2.4a). There was not a significant interaction effect between planting year 

and precipitation treatment on sown species (χ2

(4,83)=3.21, p=0.52). Seedling survival of sown species did 

not vary across planting years (χ2

(2,80)=1.85, p=0.40). The emergence of non-sown species was not 

affected by these non-precipitation year effects (χ2

(2,84)=5.39, p=0.07, Figure 2.4b; however, when P. 

capillare included, χ2

(2,84)=288.8, p<0.0001); however, the survival of these seedings varied across years 

(χ2

(2,80)=12.6, p=0.002). The death rate of non-sown species was 3.1 times higher in 2014 than in 2015 

(z=-3.3, p=0.003). Finally, there was a significant interaction between planting year and precipitation 

treatment on non-sown species emergence (χ2

(4,84)=39.4, p<0.0001, Figure 2.4b). Within the low 

precipitation treatment, non-sown species emerged significantly more in 2014 than the other two years; 

within the mean treatment, 2016 had highest emergence; and for the high water treatment, 2015 had 

the highest emergence. 

There was a significant interaction effect between seed mix richness and precipitation 

treatment on sown species emergence (χ2

(2,83)=7.97, p=0.02, Figure 2.5b). The high richness mix led to 

slightly higher emergence of sown species under low and mean precipitation treatments; however, 

emergence was significantly lower for the high richness mix under the high precipitation treatment. 

There was not a significant interaction between seed mix richness and year in the rain-out plots 

(χ2

(2,83)=2.55, p=0.28), nor was there a main effect of mix richness on sown species emergence 

(χ2

(1,83)=0.18, p=0.67). Full statistical results for the above are presented in Table 2.5. 

Community composition varied across three planting years with precipitation held constant 

(F(2,53)=5.5, p<0.0001, Figure 2.6), again because species varied in their emergence and abundance in 

different planting years (see Species Responses below). However, composition did not differ under the 

 

24 

three precipitation treatments (F(2,53)=1.3, p=0.10) when P. capillare was excluded from analysis. When 

P. capillare was included in the analysis, the three precipitation treatments had different first-year 

communities as well (F(2,53)=2.4, p=0.007). 

 

Species responses 

The three most abundant species in this experiment were Panicum capillare (Witchgrass, 1805 

seedlings; significant effect of planting year, χ2

(2,83)=342.67, p<0.0001; precipitation treatment, 

χ2

(2,83)=27.3, p<0.0001; and interaction χ2

(4,83)=72.63, p<0.0001), Digitaria ischaemum (Smooth crab grass, 

188 seedlings; significant effect of planting year, χ2

(2,83)=342.67, p<0.0001; precipitation treatment, 

χ2

(2,83)=27.3, p<0.0001), and Centaurea stoebe (Spotted knapweed, 186 seedlings, significant effect of 

planting year, χ2

(2,83)=8.45, p=0.015; no effect of precipitation treatment, χ2

(2,83)=3.16, p=0.21; and 

significant interaction χ2

(4,83)=32.13, p<0.0001). The three most abundant sown species were Bouteloua 

curtipendula (Side-oats grama, 122 seedlings, significant effect of planting year, χ2

(2,83)=41.56, p<0.0001; 

precipitation treatment, χ2

(2,83)=7.30, p=0.026), Echinacea purpurea (Purple coneflower, 86 seedlings, 

significant effect of planting year, χ2

(2,83)=32.33, p<0.0001; precipitation treatment, χ2

(2,83)=9.93, p=0.007), 

and Dalea purpurea (Purple prairie clover, 72 seedlings; significant effect of planting year, χ2

(2,83)=8.61, 

p=0.014; no effect of precipitation treatment, χ2

(2,83)=1.73, p=0.42; and no interaction χ2

(4,83)=2.97, 

p=0.56). These species showed markedly variable responses to planting conditions (Figure 2.7). For 

instance, E. purpurea had 8.5 times higher emergence in 2015 than in 2014; while D. purpurea had 2.9 

times higher emergence in 2016 than the previous two years.   

There was an increase in emergence for these three sown species with decreasing seed mix 

richness; likely a reflection of the 50% decrease in seeding density for each individual species in the 

higher diversity mix. E. purpurea emerged 1.8 times more in low richness than high richness plots 

(χ2

(1,83)=4.30, p=0.038), D. purpurea emerged 2.9 times more in low richness plots (χ2

(1,83)=11.14, 

 

25 

p=0.00084), and B. curtipendula emerged 1.4 times more, though this increase was not statistically 

significant (χ2

(1,83)=2.17, p=0.14). 

 

DISCUSSION 

 

Because of the extensive suite of variables controlled, this study provides experimental evidence 

for year effects during restoration, driven by planting condition effects on seedling emergence. By 

conducting this experiment within a single site and with identical management techniques, we were 

able to isolate the effects of planting year and infer a causal relationship between planting conditions 

and seedling establishment (Vaughn & Young 2010). Along with Stuble et al. (2017), who recently 

documented year effects in California grasslands in an experiment replicated across years and sites, our 

results are among the first to document year effects independently of otherwise confounding influences 

of site and management variation. 

  

Additionally, our precipitation experiment allowed us to directly confront the leading putative 

hypothesis for year effects: rainfall in the planting year (Bakker et al. 2003; MacDougall et al. 2008). By 

replicating our precipitation manipulations in each of the three planting years, we were able to 

disentangle precipitation effects from a suite of other planting conditions that may have varied among 

years. We found that both sown and non-sown species emerged more with increased water availability, 

showing that precipitation can cause year effects. However, we also found that sown species emergence 

varied across years within precipitation treatments, meaning non-precipitation factors can also be 

important drivers of year effects. Future work might evaluate the roles of non-precipitation factors, such 

as temperature, consumer pressure, pathogen loads, and others to further mechanistic understanding 

of the causes of year effects. Understanding these drivers will help managers take appropriate rectifying 

measures; for instance, if temperature prohibits seeds from breaking dormancy in the first growing 

 

26 

season, practitioners may use plugs or simply wait for future seasons; but if consumers and pathogens 

destroy seeds after sowing, inter-seeding will be necessary for a successful restoration. 

We observed substantial variation in the responses of individual species among years and 

treatments in our experiment. For instance, we found that 2015 was the most conducive year for 

emergence of sown prairie species, even with precipitation held constant across the years. This pattern 

is largely driven by the increased emergence of Echinacea purpurea (mean 1.3 out of 4.2 emerged 

individuals/300 cm2 plot) and Bouteloua curtipendula (mean 1.6) that year. There was congruence of 

many conditions that promoted germination for those species. For example, although 2015 was, on 

average, the coolest year of the three, it was the only year to have 5+ consecutive days in May-June that 

reached a high of 23.8 C (75°F), which has been shown to be an important germination requirement for 

E. purpurea (Hassell et al. 2004a; Hassell et al. 2004b). Other species exhibited variable patterns, with 

some emerging in two (e.g. Coreopsis lanceolata in 2014 and 2016, Oenothera biennis in 2015 and 2016) 

or one (e.g. Astragalus canadensis in 2014) of the three planting years. Across precipitation treatments, 

species patterns were far less variable, with only one otherwise present sown species completely absent 

from drought plots (O. biennis) and most showing increases in emergence rates with increased water, 

suggesting that it is environmental attributes besides rainfall varying between years that drove many 

species-level patterns. This variation in responses between species results in a different first-year 

community composition after different first-year planting conditions. It is this initial community 

composition upon which all future assembly must build. Without these early community differences, 

historically contingent community assembly is not possible (Fukami 2015). The first year community 

assemblage may facilitate, inhibit, or merely tolerate future community members (Connell & Slatyer 

1977) and future work will be required to understand the consequences of our emergence results on 

community assembly. 

 

27 

There was no evidence that increasing seed mix richness improved establishment of sown 

species in one growing season across planting conditions. We found an interaction between seed mix 

richness and planting year in ambient (but not rain-out) plots, and an interaction between seed mix 

richness and precipitation treatment in rain-out plots. Increased seed mix richness did not consistently 

improve establishment in the first year of restoration. These patterns likely again reflect the variable 

responses to planting conditions across species, and suggest that including a higher abundance of 

reliable species (i.e., species that emerge under a wide range of conditions, such as Dalea purpurea) can 

be as valuable to managers as diversifying across species with different emergence requirements when 

the goal is to improve emergence rates of sown species.  

Prairie restoration practitioners have long known the importance of site preparation and weed 

suppression for restoration success (Rowe 2010), and our results emphasize the ability of weeds to 

establish at higher rates under wet conditions. Additional mowing prior to seed set should reduce the 

propagule pressure of weeds after planting, especially in wet years. Watering, although usually 

logistically infeasible, could improve sown species emergence, but at the cost of increased non-sown 

emergence. Finally, at a constant seeding rate, including a higher species richness in a seed mix can 

come at the cost of reducing the abundance of well-performing species (e.g., D. purpurea, E. purpurea) 

and therefore reduce overall initial establishment at a site. Managers should therefore consider this 

balance between increasing sown species richness (to increase possible diversity at a site) and increasing 

sowing rates of well-establishing species (to increase probability of sown species establishment) when 

deciding how to allocate resources to their seed mixes (see also Packard & Mutel 1997 for discussion of 

seed mix design for prairie restoration).  

Ecological restoration can be an important tool for promoting biodiversity, but outcomes can 

vary. Our study shows how planting year conditions can be important in an establishing community. 

Future work will build on these first-year findings to see whether planting year differences persist over 

 

28 

years to influence community assembly. Lasting year effects will likely vary by species (Werner et al. 

2016), and in some cases, first-year communities may leave their mark on communities assembling into 

the future due to processes of historically contingent assembly (e.g. Lockwood and Samuels 2004, 

Fukami 2015). In the end, it will be important to examine how year effects matter alongside other major 

causes of variation in restored communities, e.g. soil edaphic conditions (Grman & Brudvig 2014) and 

fire frequency (Bowles & Jones 2013).  

 

ACKNOWLEDGEMENTS 

 

Many thanks to field assistants Meg Kargul, Michelle Franklin, and Logan Brissette; and to Mike 

Bassett, Maddy Cleary, George Wheeler, Evan Johnson, and Andy Funk for assistance building rain-out 

shelters. George Groves helped engineer the rain-out shelter design. Nash Turley, Kristen Nolting, and 

Elise Zipkin gave helpful statistical advice. Mark Manuszak did the heavy lifting at Lux Arbor and 

maintained a friendly, safe working environment for which Anna is forever grateful. Jared Foster at 

Native Connections provided critical seed mix and planting advice. This study was funded by two 

Michigan State University Graduate School Fellowships and two Porter Research Awards, all graciously 

awarded by the W.K. Kellogg Biological Station. This material is based upon work supported by the 

National Science Foundation under Grant No. DEB-1552197 (L.A.B.) and Graduate Research Fellowship 

DGE-1424871 (A.M.G.). This is KBS contribution #2012.

 

29 

 

 

APPENDIX 

30 

APPENDIX 

Figures 

 

 

Figure 2.1. Methods schematic. Six large plots were restored in each of three study years. Each large 

plot contained four precipitation treatments: ambient conditions (no rain-out shelter), low precipitation, 

mean precipitation, or high precipitation. Each precipitation treatment plot contained a high (30 

species) and low (8 species) sown richness split plot. Finally, each split-plot contained three randomly 

distributed sub-sample plots in which seedling emergence and death were monitored. 

 

 

 

 

31 

 

 

 

 

 

 

 

 

Figure 2.2. Mean seedling emergence across three planting years for sown and non-sown species 

under ambient conditions. Mean values are per three pooled 10 x 10-cm sub-sample plots; a total area 

of 300 cm2. All error bars indicate ±1 SE. Letters represent Tukey post-hoc differences within each 

species group (sown, non-sown) and are corrected for multiple comparisons. 

 

 

32 

 

 

 

 

 

 

 Figure 2.3. Mean seedling emergence across three precipitation treatments for sown and non-sown 

species. Mean values are per three pooled 10 x 10-cm sub-sample plots; a total area of 300 cm2. All 

error bars indicate ±1 SE. Letters represent Tukey post-hoc differences within each species group (sown, 

non-sown) and are corrected for multiple comparisons. 

 

 

33 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 2.4. Mean seedling emergence across three planting years when precipitation was held 

constant for (A) sown and (B) non-sown species. Mean values are per three pooled 10 x 10-cm sub-

sample plots; a total area of 300 cm2. All error bars indicate ±1 SE. Letters represent Tukey post-hoc 

differences for planting year within each precipitation treatment and are corrected for multiple 

comparisons.  

 

 

34 

2014201 201 Low Mean HighWatering TreatmentLow Mean HighWatering Treatment2014201 201 Planting  earPlanting  earaaaaaabaABBAAA                                                                          ABAb 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 2.5. (A) Interaction between sown richness and planting year on sown species emergence 

under ambient conditions. (B) Interaction between sown richness and precipitation treatment on 

sown species emergence. Mean values are per three pooled 10 x 10-cm sub-sample plots; a total area 

of 300 cm2. All error bars indicate ±1 SE. Letters represent Tukey post-hoc differences for planting year 

within each precipitation treatment and are corrected for multiple comparisons. 

 

 

35 

LowHigh2014201  201 Planting  earSeed Mix Richness                                                                                    aaaaabLowHighLowMeanHighWatering TreatmentSeed Mix Richnessaaaab                                                                a 

 

Figure 2.6. Nonmetric multidimensional scaling (NMDS) in three dimensions (stress=0.19) with Bray-

Curtis dissimilarity to visualize differences in plant communities among three planting years (2014, 2015, 

2016; R2=0.18) and three precipitation treatments (low, mean, high; R2=0.04) held constant across the 

years. Ellipses are 95% confidence intervals.  

 

 

36 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 2.7. Seedling emergence across three planting years and precipitation treatments for the three 

most abundant non-sown (Panicum capillare, Digitaria ischaemum, and Centaurea stoebe) and three 

most abundant sown (Bouteloua curtipendula, Echinacea purpurea, and Dalea purpurea) species. 

Because E. purpurea was mistakenly replaced with E. pallida in 2016, results are not shown for E. 

purpurea in that year.  Mean values are per three pooled 10 x 10-cm sub-sample plots; a total area of 

300 cm2. All error bars indicate ±1 SE. Photos by John Hilty, Illinois Wildflowers. 

 

 

37 

 

 

 

 

 

 

Figure 2.8. Rain-out shelter design. Rain-out shelters consisted of a clear vinyl roof, t-posts, and a gutter 

that directed water into a kiddie pool for use in the watering treatments. Each sampled area under the 

shelter was surrounded by rubber gardening edging to prevent overland flow of water. Numbers near 

each corner denote the approximate height at that location 

 

 

 

 

 

. 

 

 

38 

Tables 

Table 2.1. Low richness and high richness seed mixes sown. Seed mixes were updated each planting 

year with fresh seeds from Native Connections (Three Rivers, MI) with updated % pure live seed 

measurements. During sowing, first each large area was sown with the low diversity mix at a rate of 

378.9 seeds/m2 (35.2 seeds/ft2) divided evenly among the 8 species in the low richness mix (47.4 

seeds/m2 each). Then, low richness experimental plots were sown with an additional 378.9 seeds/m2 

divided evenly among the 8 species (for a total of 94.7 seeds/m2 each); while high richness plots were 

sown with 378.9 seeds/m2 divided evenly among 22 different species (17.2 seeds/m2 each). 

 

 

 

Seed mixes sown (seeding rate in seeds/m2) 

Species name 

Common name 

Type 

Low 

High 

Andropogon gerardii 

Big Bluestem 

C4 grass  94.7 

47.4 

Bouteloua curtipendula 

Side Oats Grama 

C4 grass  94.7 

47.4 

Bromus kalmii 

Prairie Brome 

C3 Grass  — 

17.2 

Calamagrostis canadensis 

Bluejoint Grass 

C3 Grass  — 

17.2 

Elymus canadensis 

Canada Wild Rye 

C3 Grass  94.7 

47.4 

Koeleria macrantha 

Junegrass 

C3 Grass  — 

17.2 

Panicum virgatum 

Switchgrass 

C4 grass  — 

17.2 

Schizachyrium scoparium 

Little Bluestem 

C4 grass  — 

17.2 

Sorghastrum nutans 

Indian Grass 

C4 grass  — 

17.2 

Asclepias tuberosa 

Butterflyweed 

Forb 

— 

17.2 

Astragalus canadensis 

Canadian Milkvetch 

Legume  — 

17.2 

Baptisia alba 

White Wild Indigo 

Legume  — 

17.2 

Chamaechrista fasciculata 

Partridge Pea 

Legume 

94.7 

47.4 

Coreopsis lanceolata 

Lance-leaf Coreopsis 

Forb 

— 

17.2 

Dalea purpurea 

Purple Prairie Clover 

Legume 

94.7 

47.4 

39 

Table 2.1      ’   
 

Desmodium canadense 

Showy Tick Trefoil 

Legume  — 

17.2 

Echinacea purpurea/pallida 

Purple/Pale Coneflower 

Forb 

94.7 

47.4 

Eryngium yuccifolium 

Rattlesnake Master 

Heliopsis helianthoides 

False Sunflower 

Oenothera biennis 

Evening Primrose 

Oligoneuron rigidum 

Stiff Goldenrod 

Forb 

Forb 

Forb 

Forb 

94.7 

47.4 

— 

— 

— 

17.2 

17.2 

17.2 

Penstemon digitalis 

Foxglove Beardtongue 

Forb 

94.7 

47.4 

Potentilla arguta 

Prairie Cinquefoil 

Forb 

Ratibida pinnata 

Gray-headed Coneflower  Forb 

Rudbeckia hirta 

Black-eyed Susan 

Scrophularia lanceolata 

Lanceleaf Figwort 

Symphyotrichum urophyllum  Arrow-leaf Aster 

Forb 

Forb 

Forb 

— 

— 

— 

— 

— 

17.2 

17.2 

17.2 

17.2 

17.2 

Tephrosia virginiana 

Goat’s Rue 

Legume  — 

17.2 

Verbena stricta 

Hoary Vervain 

Verbesina alternifolia 

Wingstem 

Forb 

Forb 

— 

— 

17.2 

17.2 

                                                                                                         Total 

757.8  757.8 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

 

40 

Table 2.2. Overview of results with Panicum capillare included. †Denotes result’s significance is 

different when P. capillare excluded. 

 
Predictor 
Non-sown seedling emergence in ambient conditions 

p-value 

χ2 

df 

Planting year 

167.1 

2, 28  <0.0001*** 

Non-sown seedling death in ambient conditions 
Planting year 
Non-sown seedling emergence in watering treatments 

2, 28  0.37† 

2.0 

Planting year 

288.8 

2, 84  <0.0001***† 

Watering treatment  42.7 

2, 84  <0.0001*** 

Post-hoc contrasts 

2014-2015, z=-8.3, p<0.0001*** 
2014-2016, z=-12.7, p<0.0001***† 
2015-2016, z=5.6, p<0.0001*** 

NA† 

2014-2015, z=-9.1, p<0.0001***† 
2014-2016, z=-15.2, p<0.0001***† 
2015-2016, z=-6.3, p<0.0001***† 
low-mean: z=-3.5, p=0.002** 
low-high: z=5.9, p<0.001*** 
mean-high: z=2.5, p=0.04† 

Year * watering 
Non-sown seedling survival in watering treatments 

4, 84  <0.0001***  NA 

108.8 

Planting year 

30.1 

2, 80  <0.0001*** 

Watering treatment  2.2 
First year plant community in ambient conditions 

2, 80  0.33 

2014-2015: z=5.2, p<0.0001*** 
2014-2016: z=0.4, p=0.91 
2015-2016: z=-5.3, p<0.0001***† 
NA 

Planting year 

F=1.9 
R2=0.20 

2, 17  0.04* 

NA 

First year plant community in watering treatments 

Planting year 

Watering treatment 

Year * watering 

F=6.9 
R2=0.20 
F=2.4 
R2=0.07 
F=1.0 
R2=0.06 

2, 53  <0.0001***  NA 

2, 53  0.007**† 

4, 53  0.32 

NA 

NA 

 
 

 

 

41 

Table 2.3. Plot-level means for seedling survival measured across three planting years in ambient and 

watering treatment plots. “All species” category includes sown, non-sown, and unknown species. Mean 

values (n=6) are per 3 pooled 10 cm x 10 cm sub-sample plots; a total area of 0.03 m2. 

Ambient & rain-out plots 
Plot level means 

2014 

2015 

2016 

Ambient Plots 
(no rain-out shelter) 

Treatment 1 
(low precipitation) 

Treatment 2 
(mean precipitation) 

Treatment 3 
(high precipitation) 

Mean # died 

Sown 
Non-sown 
All 

0.2 
0.2 
0.7 

0.0 
0.3 
0.8 

0.3 
0.9 
2.4 

% Survived 

Sown 
Non-sown 
All 

94.9% 
97.8% 
96.1% 

100.0% 
98.9% 
97.8% 

93.5% 
97.4% 
94.2% 

Mean # died 

Sown 
Non-sown 
All 

0.0 
0.3 
0.3 

0.4 
0.0 
0.3 

0.0 
0.8 
1.2 

% Survived 

Sown 
Non-sown 
All 

100.0% 
96.4% 
96.4% 

79.5% 
100.0% 
97.6% 

100.0% 
92.8% 
91.4% 

Mean # died 

Sown 
Non-sown 
All 

0.0 
0.3 
0.9 

0.0 
0.0 
0.3 

0.3 
2.2 
2.8 

% Survived 

Sown 
Non-sown 
All 

100.0% 
97.1% 
94.6% 

100.0% 
100.0% 
98.9% 

89.3% 
93.6% 
93.1% 

Mean # died 

Sown 
Non-sown 
All 

0.1 
1.4 
2.0 

0.2 
0.5 
0.9 

0.3 
2.1 
3.5 

% Survived 

Sown 
Non-sown 
All 

95.5% 
87.5% 
87.7% 

96.9% 
98.8% 
98.1% 

92.0% 
94.3% 
92.1% 

 

42 

 
 

 

Table 2.4. Statistical results for planting restored prairie in three planting years (ambient conditions). 

Post hoc tests on statistically significant main effects (α=0.05) used Tukey contrasts. †Result p>0.05 or 

††result p<0.05 when P. capillare included, see Table 2.2. 

 
A. Seedling emergence in ambient conditions 
Predictor 
Sown species 
Planting year 
   (See Figure 2a) 

37.4  2, 25  <0.0001*** 

p-value 

χ2 

df 

Seed mix richness  3.5 
Year * richness 
6.4 
   (See Figure 5a) 

1, 25  0.06 
2, 25  0.04* 

Non-sown species 
Planting year 
   (See Figure 2b) 

37.5  2, 28  <0.0001*** 

Post hoc contrasts 

2014-2015, z=-5.8, p<0.0001*** 
2014-2016, z=-1.4, p=0.35†† 
2015-2016, z=4.5, p<0.0001*** 
NA 
Year effect dependent on richness: 
   High mix richness: 
      2014-2015, z=-4.82, p<0.0001*** 
      2014-2016, z=-2.87, p=0.011* 
      2015-2016, z=2.27, p=0.060 
   Low mix richness: 
      2014-2015, z=-3.35, p=0.0024** 
      2014-2016, z=0.76, p=0.73 
      2015-2016, z=3.96, p=0.0002*** 
Richness effect dependent on year: 
    2014, z=-0.76, p=0.45 
    2015, z=1.04, p=0.30 
    2016, z=2.87, p=0.0041** 

2014-2015, z=-4.5, p<0.0001*** 
2014-2016, z=1.0, p=0.55 
2015-2016, z=5.4, p<0.0001*** 

df 

χ2 

B. Seedling death in ambient conditions 
Predictor 
Sown species 
NA— of 168 seedlings in ambient plots, only 5 died 
Non-sown species 
Planting year 

p-value 

contrasts 

20.7  2, 28  <0.0001***†  2014-2015, z=0.67, p=0.78 

 
 

 

2014-2016, z=-3.2, p=0.004**† 
2015-2016, z=-4.1, p=0.0001***† 

 

43 

Table 2.5. Statistical results for planting restored prairie with three watering treatments in three 

planting years. Tukey contrast p-values adjusted for comparing families of three estimates. †Result 

p>0.05 or ††result p<0.05 when P. capillare included, see Table 2.2. 

A. Seedling emergence in watering treatments 
Predictor 
Sown species 
Planting year 

p-value 

χ2 

df 

Post hoc contrasts 

51.9  2, 83  <0.0001***  2014-2015: z=-5.7, p<0.0001*** 

Watering treatment  25.6  2, 83  <0.0001*** 

Seed mix richness 
Year * watering 
Year * richness 
Watering * richness 
  (See figure 5b) 

0.18  1, 83  0.67 
3.21  4, 83  0.52   
2.55  2, 83  0.28 
7.97  2, 83  0.02* 

2014-2016: z=-1.3, p=0.40 
2015-2016: z=4.5, p<0.0001*** 
low-mean: z=-3.6, p=0.001** 
low-high: z=4.7, p<0.0001*** 
mean-high: z=1.2, p=0.47 
NA 
NA 
NA 
Watering effect dependent on richness: 
   Low richness:  low-mean, z=-3.12, p=0.0052** 
          low-high, z=5.07, p<0.001** 
          mean-high, z=2.25, p=0.064 
   High richness:  low-mean, z=-2.64, p=0.023* 
          low-high, z=2.41, p=0.043* 
          mean-high, z=-0.24, p=0.97 
Richness effect dependent on watering: 
    low, z=1.20, p=0.23 
    mean, z=0.67, p=0.51 
    high, z=-2.52, p=0.012* 

 

 

 

 

 

 

 

 

 

 

 

 

44 

      2.5      ’   

Non-sown species 
Planting year 
Watering treatment  19.0  2, 84  <0.0001*** 

5.39  2, 84  0.07† 

NA 
low-mean: z=-3.0, p=0.009** 
low-high: z=5.1, p<0.001*** 
mean-high: z=2.2, p=0.07†† 

39.4  4, 84  <0.0001***  Year effect dependent on watering: 

Year * watering 
   (See figure 5c) 

   Low: 2014-2015, z=4.24, p=0.0001*** 
        2014-2016, z=4.87, p<0.0001*** 
        2015-2016, z=0.84, p=0.68 
   Mean: 2014-2015, z=1.15, p=0.48 
        2014-2016, z=0.00, p=1.00 
        2015-2016, z=-1.15, p=0.48 
   High:   2014-2015, z=-2.66, p=0.022* 
        2014-2016, z=0.00, p=1.00 
        2015-2016, z=2.66, p=0.022*   
Watering effect dependent on year: 
   2014:  low-mean, z=-0.59, p=0.83 
        low-high, z=1.96, p=0.11 
        mean-high, z=1.39, p=0.35 
   2015:  low-mean, z=-2.76, p=0.016* 
        low-high, z=5.73, p<0.0001*** 
        mean-high, z=3.22, p=0.0037** 
   2016:  low-mean, z=-3.92, p=0.0003*** 
        low-high, z=5.16, p<0.0001*** 
        mean-high, z=1.39, p=0.35 

contrasts 

NA 
NA 

2014-2015: z=-3.3, p=0.003** 
2014-2016: z=-2.0, p=0.11 
2015-2016: z=0.89, p=0.65†† 
NA 

df 

p-value 

χ2 

B. Seedling survival in watering treatments 
Predictor 
Sown species 
Planting year 
1.85  2, 80  0.40 
Watering treatment  1.17  2, 80  0.56 
Non-sown species 
Planting year 

12.6  2, 80  0.002** 

Watering treatment  5.91  2, 80  0.05 
 
 

 

 

 

 

 

 

 

45 

LITERATURE CITED

 

46 

LITERATURE CITED 

 

Aarssen LW, Laird RA, Pither J (2003) Is the productivity of regetation plots higher or lower when there 
are more species? Variable predictions from interaction of the ‘sampling effect’ and ‘competitive 
dominance effect’ on the habitat templet. Oikos 102:427–432 

Alford RA, Wilbur HM (1985) Priority effects in experimental pond communities: Competition between 

Bufo and Rana. Ecology 66:1097–1105 

Anderson MJ (2001) A new method for non-parametric multivariate analysis of variance. Austral Ecology 

26:32–46 

Averett JM, Klips RA, Nave LE, Frey SD, Curtis PS (2004) Effects of soil carbon amendment on nitrogen 
availability and plant growth in an experimental tallgrass prairie restoration. Restoration Ecology 
12:568–574 

Baer SG, Engle DM, Knops JMH, Langeland K a, Maxwell BD, Menalled FD, Symstad AJ (2009) 

Vulnerability of rehabilitated agricultural production systems to invasion by nontarget plant 
species. Environmental management 43:189–96 

Bakker JD, Wilson SD, Christian JM, Li X, Ambrose LG, Waddington J (2003) Contingency of grassland 

restoration on year, site, and competition from introduced grasses. Ecological Applications 13:137–
153 

Bakker JP, Poschlod P, Strykstra RJ, Bekker RM, Thompson K (1996) Seed banks and seed dispersal: 

Important topics in restoration ecology. Acta Botanica Neerlandica 45:461–490 

Baskin CC, Baskin JM (1988) Germination ecophysiology of herbaceous plant species in a temperate 

region. American Journal of Botany 75:286–305 

Bates D, Maechler M, Bolker B, Walker S (2015) Fitting linear mixed-effects models using lme4. Journal 

of Statistical Software 67:1–48 

Belyea LR (2004) Beyond ecological filters: Feedback networks in the assembly and restoration of 
community structure. In: Assembly rules and restoration ecology: Bridging the gap between 
science and practice. Temperton, VM, Hobbs, RJ, Nuttle, T, & Halle, S, editors. Island Press. 

Belyea LR, Lancaster J (1999) Assembly within a contingent ecology. Oikos 86:402–416 

Blumenthal DM, Jordan NR, Russelle MP (2003) Soil carbon addition controls weeds and facilitates 

prairie restoration. Ecological Applications 13:605–615 

Bowles ML, Jones MD (2013) Repeated burning of eastern tallgrass prairie increases richness and 

diversity, stabilizing late successional vegetation. Ecological Applications 23:464–478 

Cardinale BJ, Duffy JE, Gonzalez A, Hooper DU, Perrings C, Venail P, Narwani A, Mace GM, Tilman D, 

Wardle DA, Kinzig AP, Daily GC, Loreau M, Grace JB, Larigauderie A, Srivastava DS, Naeem S (2012) 
Biodiversity loss and its impact on humanity. Nature 486:59–67 

 

47 

Carter DL, Blair JM (2012) High richness and dense seeding enhance grassland restoration establishment 

but have little effect on drought response. Ecological Applications 22:1308–19 

Chase JM (2003) Community assembly: When should history matter? Oecologia 136:489–98 

Comer PJ, Albert DA, Wells H., Hart BL, Raab JB, Price DL, Kashian DM, Corner RA, Schuen DW, Leibfreid 
TR, Austin MB, DeLain CJ, Prange-Gregory L, Scrimger LJ, Korroch KM, Spitzley JG (199 ) Michigan’s 
presettlement vegetation, as interpreted from the General Land Office surveys 1816-1856. Lansing, 
MI 

Connell JH, Slatyer RO (1977) Mechanisms of succession in natural communities and their role in 

community stability and organization. The American Naturalist 111:1119–1144 

Cramer VA, Hobbs RJ, Standish RJ (2008) What’s new about old fields? Land abandonment and 

ecosystem assembly. Trends in Ecology and Evolution 23:104–12 

Deering RH, Young TP (2006) Germination speeds of exotic annual and native perennial grasses in 

California and the potential benefits of seed priming for grassland restoration. Grasslands XVI:14–
16 

Dickson TL, Hopwood JL, Wilsey BJ (2012) Do priority effects benefit invasive plants more than native 

plants? An experiment with six grassland species. Biological Invasions 14:2617–2624 

Dobson AP, Bradshaw AD, Baker AJM (1997) Hopes for the future: Restoration ecology and conservation 

biology. Science 277:515–522 

Doll JE, Haubensak K a., Bouressa EL, Jackson RD (2011) Testing disturbance, seeding time, and soil 
amendments for establishing native warm-season grasses in non-native cool-season pasture. 
Restoration Ecology 19:1–8 

Drake JA (1990) The mechanics of community assembly and succession. Journal of Theoretical Biology 

147:213–233 

Foley JA, DeFries R, Asner GP, Barford C, Bonan G, Carpenter SR, Chapin FS, Coe MT, Daily GC, Gibbs HK, 
Helkowski JH, Holloway T, Howard EA, Kucharik CJ, Monfreda C, Patz JA, Prentice IC, Ramankutty N, 
Snyder PK (2005) Global consequences of land use. Science 309:570–574 

Forrest JRK, James DT (2011) An examination of synchrony between insect emergence and flowering in 

Rocky Mountain meadows. Ecological Monographs 81:469–491 

Fox J, Weisberg S (2011) An {R} companion to applied regression. Second Ed. Sage, Thousand Oaks, CA 

Fukami T (2015) Historical contingency in community assembly: Integrating niches, species pools, and 

priority effects. Annual Review of Ecology, Evolution, and Systematics 46:1–23 

Fukami T, Mordecai EA, Ostling A (2016) A framework for priority effects. Journal of Vegetation Science 

27:655–657 

Getz LL, Hofmann JE (1999) Diversity and stability of small mammals in tallgrass prairie habitat in central 

Illinois, USA. Oikos 85:356–363 

 

 

48 

Grant K, Kreyling J, Heilmeier H, Beierkuhnlein C, Jentsch A (2014) Extreme weather events and plant–

plant interactions: shifts between competition and facilitation among grassland species in the face 
of drought and heavy rainfall. Ecological Research 29:991–1001 

Grman E, Bassett T, Brudvig L (2013) Confronting contingency in restoration: Management and site 

history determine outcomes of assembling prairies, but site characteristics and landscape context 
have little effect. Journal of Applied Ecology 50:1234–1243 

Grman E, Bassett T, Brudvig LA (2014) A prairie plant community data set for addressing questions in 

community assembly and restoration. Ecology 95:2363 

Grman E, Bassett T, Zirbel CR, Brudvig LA (2015) Dispersal and establishment filters influence the 

assembly of restored prairie plant communities. Restoration Ecology 23:892–899 

Grman E, Brudvig LA (2014) Beta diversity among prairie restorations increases with species pool size, 

but not through enhanced species sorting. Journal of Ecology 102:1017–1024 

Group PC (2018) PRISM Climate Data. 

Groves AM, Brudvig LA (2018) Inter-annual variation in precipitation and other planting conditions 

impacts seedling establishment in sown plant communities. Restoration Ecology 

Harper JL (1977) Population Biology of Plants. Academic Press, London 

Harper JL (1987) The heuristic value of ecological restoration. In: Restoration Ecology: A synthetic 

approach to ecological research. Jordan, WRI, Gilpin, ME, & Aber, JD, editors. University Press, 
Cambridge, U.K. pp. 35–46. 

Hassell RL, Dufault R, Phillips T (2004) Relationship among seed size, source and temperature on 

germination of Echinacea angustifolia, pallida and purpurea. Acta Horticulturae 629:239–243 

Hassell RL, Dufault RJ, Phillips T, Hale TA (2004) Influence of temperature gradients on Pale and Purple 

Coneflower, Feverfew, and Valerian germination. HortTechnology 14:368–371 

Heisler-White JL, Blair JM, Kelly EF, Harmoney K, Knapp AK (2009) Contingent productivity responses to 

more extreme rainfall regimes across a grassland biome. Global Change Biology 15:2894–2904 

Hobbs RJ, Harris JA (2001) Restoration Ecology: Repairing the Earth’s Ecosystems in the New Millennium. 

Restoration Ecology 9:239–246 

Hooper DU (2005) Effects of Biodiversity on Ecosystem Functioning A Consensus of Current Knowledge. 

Ecological Monographs 75:3–35 

Howe HF (2008) Reversal of fortune: Plant suppression and recovery after vole herbivory. Oecologia 

157:279–86 

Howe HF, Brown JS (1999) Effects of Birds and Rodents on Synthetic Tallgrass Communities. Ecology 

80:1776–1781 

Hulme PE, Kollmann J (2005) Seed predator guilds, spatial variation in post-dispersal seed predation and 

potential effects on plant demography: A temperate perspective. In: Seed Fate. Predation, 
dispersal and seedling establishment. pp. 9–30. 

 

49 

Johnson DH (2002) The importance of replication in wildlife research. The Journal of Wildlife 

Management 66:919–932 

Jones HP, Jones PC, Barbier EB, Blackburn RC, Benayas JMR, Holl KD, McCrackin M, Meli P, Montoya D, 
Moreno-Mateos D (2018) Restoration and repair of Earth’s damaged ecosystems. Proceedings of 
the Royal Society B 285 

Keddy PA (1992) Assembly and response rules: Two goals for predictive community ecology. Journal of 

Vegetation Science 3:157–164 

Larson JE, Sheley RL, Hardegree SP, Doescher PS, James JJ (2015) Seed and seedling traits affecting 
critical life stage transitions and recruitment outcomes in dryland grasses. Journal of Applied 
Ecology 52:199–209 

Lenth R V (2016) Least-squares means: The {R} package {lsmeans}. Journal of Statistical Software 69:1–

33 

Lockwood JL, Samuels CL (2004) Assembly models and the practice of restoration. In: Assembly Rules 
and Restoration Ecology. Temperton, VM, Hobbs, RJ, Nuttle, T, & Halle, S, editors. Island Press, 
Washington, USA pp. 55–70. 

MacDougall AS, Wilson SD (2007) Herbivory limits recruitment in an old-field seed edition experiment. 

Ecology 88:1105–1111 

MacDougall AS, Wilson SD, Bakker JD (2008) Climatic variability alters the outcome of long-term 

community assembly. Journal of Ecology 96:346–354 

Manning GC, Baer SG (2018) Interannual variability in climate effects on community assembly and 

ecosystem functioning in restored prairie. Ecosphere 9:e02327 

Martin LM, Wilsey BJ (2012) Assembly history alters alpha and beta diversity, exotic-native proportions 

and functioning of restored prairie plant communities Cadotte, M, editor. Journal of Applied 
Ecology 

Mason CH, Perreault Jr. WD (1991) Collinearity, power, and interpretation of multiple regression 

analysis. Journal of Marketing Research 28:268–280 

Mousseau T, Fox C (1998) The adaptive significance of maternal effects. Trends in Ecology & Evolution 

13:403–407 

Myers JA, Harms KE (2009) Seed arrival, ecological filters, and plant species richness: A meta-analysis. 

Ecology letters 12:1250–1260 

Nemec KT, Allen CR, Helzer CJ, Wedin DA (2013) Influence of Richness and Seeding Density on Invasion 

Resistance in Experimental Tallgrass Prairie Restorations. Ecological Restoration 31:168–185 

Notman EM, Villegas AC (2005) Patterns of seed predation by vertebrate versus invertebrate seed 

predators among different plant species, seasons and spatial distributions. In: Seed Fate: 
Predation, Dispersal and Seedling Establishment. pp. 55–75. 

 

 

50 

Oksanen J, Blanchet FG, Friendly M, Kindt R, Legendre P, McGlinn D, Minchin PR, O’Hara RB, Simpson GL, 

Solymos P, Stevens MHH, Szoecs E, Wagner H (2017) Vegan: Community ecology package. R 
package version 2.4-2 https://CRAN.R-project.org/package=vegan 

Packard S, Mutel CF, eds. (1997) The tallgrass restoration handbook: For prairies, savannas, and 

woodlands. Island Press, Washington, D.C. 

Palmer MA, Ambrose RF, Poff NL (1997) Ecological Theory and Community Restoration Ecology. 

Restoration Ecology 5:291–300 

Piper JK, Schmidt ES, Janzen AJ (2007) Effects of species richness on resident and target species 

components in a prairie restoration. Restoration Ecology 15:189–198 

Roach DA, Wulff RD (1987) Maternal effects in plants. Annual Review of Ecology and Systematics 

18:209–235 

Rowe HI (2010) Tricks of the trade: Techniques and opinions from 38 experts in tallgrass prairie 

restoration. Restoration Ecology 18:253–262 

Ruiz-Jaen MC, Mitchell Aide T (2005) Restoration success: How is it being measured? Restoration 

Ecology 13:569–577 

Samson F, Knopf F (1994) Prairie conservation in North America. BioScience 44:418–421 

Schindler DE, Armstrong JB, Reed TE (2015) The portfolio concept in ecology and evolution. Frontiers in 

Ecology and the Environment 13:257–263 

Seabloom EW, Borer ET, Boucher VL, Burton RS, Cottingham KL, Goldwasser L, Gram WK, Kendall BE, 

Micheli F (2003) Competition, seed limitation, disturbance, and reestablishment of California 
native annual forbs. Ecological Applications 13:575–592 

Stuble KL, Fick SE, Young TP (2017) Every restoration is unique: Testing year effects and site effects as 

drivers of initial restoration trajectories. Journal of Applied Ecology 54:1051–1057 

Suding KN, Collins SL, Gough L, Clark C, Cleland EE, Gross KL, Milchunas DG, Pennings S (2005) 

Functional- and abundance-based mechanisms explain diversity loss due to N fertilization. 
Proceedings of the National Academy of Sciences of the United States of America 102:4387–92 

Suding KN (2011) Toward an era of restoration in ecology: Successes, failures, and opportunities ahead. 

Annual Review of Ecology, Evolution, and Systematics 42:465–487 

Team RC (2016) R: A language and environment for statistical computing. R Foundation for Statistical 

Computing, Vienna, Austria 

Thompson K, Grime JP (1983) A Comparative Study of Germination Responses to Diurnally-Fluctuating 

Temperatures. Journal of Applied Ecology 20:141–156 

Tilman D, Reich PB, Knops JMH (2006) Biodiversity and ecosystem stability in a decade-long grassland 

experiment. Nature 441:629–32 

Trudgill DL, Honek A, Li D, Van Straalen NM (2005) Thermal time - Concepts and utility. Annals of Applied 

Biology 146:1–14 

 

51 

Turnbull LA, Rahm S, Baudois O, Eichenberger-Glinz S, Wacker L, Schmid B (2005) Experimental invasion 

by legumes reveals non-random assembly rules in grassland communities. Journal of Ecology 
93:1062–1070 

Vaughn KJ, Young TP (2010) Contingent conclusions: Year of initiation influences ecological field 

experiments, but temporal replication is rare. Restoration Ecology 18:59–64 

Vaughn KJ, Young TP (2015) Short-term priority over exotic annuals increases the initial density and 

longer-term cover of native perennial grasses. Ecological Applications 25:791–799 

Wainwright CE, Staples TL, Charles LS, Flanagan TC, Lai HR, Loy X, Reynolds VA, Mayfield MM (2018) 
Links between community ecology theory and ecological restoration are on the rise. Journal of 
Applied Ecology 55:570–581 

Wainwright CE, Cleland EE (2013) Exotic species display greater germination plasticity and higher 

germination rates than native species across multiple cues. Biological Invasions 15:2253–2264 

Wang Y, Frei M (2011) Stressed food - The impact of abiotic environmental stresses on crop quality. 

Agriculture, Ecosystems and Environment 141:271–286 

Weiher E, Freund D, Bunton T, Stefanski A, Lee T, Bentivenga S (2011) Advances, challenges and a 

developing synthesis of ecological community assembly theory. Philosophical Transactions of the 
Royal Society B: Biological Sciences 366:2403–13 

Weiher E, Clarke GDP, Keddy PA (1998) Community assembly rules, morphological dispersion, and the 

coexistence of plant species. Oikos 81:309–322 

Werner CM, Vaughn KJ, Stuble KL, Wolf K, Young TP (2016) Persistent asymmetrical priority effects in a 

California grassland restoration experiment. Ecological Applications 26:1624–1632 

Wilsey BJ, Barber K, Martin LM (2014) Exotic grassland species have stronger priority effects than natives 

regardless of whether they are cultivated or wild genotypes. The New phytologist 

Wilsey BJ, Daneshgar PP, Polley HW (2011) Biodiversity, phenology and temporal niche differences 

between native- and novel exotic-dominated grasslands. Perspectives in Plant Ecology, Evolution 
and Systematics 13:265–276 

Wilson SD (2015) Managing contingency in semiarid grassland restoration through repeated planting. 

Restoration Ecology 23:385–392 

Wuebbles DJ, Hayhoe K (2004) Climate change projections for the United States Midwest. Mitigation 

and Adaptation Strategies for Global Change 9:335–363 

Young TP, Chase JM, Huddleston RT (2001) Community succession and assembly. Ecological Restoration 

19:5–18 

Young TP, Stuble KL, Balachowski JA, Werner CM (2016) Using priority effects to manipulate competitive 

relationships in restoration. Restoration Ecology 25:S114–S123 

 

52 

CHAPTER THREE 

EXPERIMENTAL MANIPULATION OF PLANTING YEAR CONDITIONS  

ALTERS PRAIRIE RESTORATION TRAJECTORIES 

 

ABSTRACT 

 

Variation in outcomes of ecological restoration projects is commonplace, and often these 

outcomes do not match practitioners’ goals. This mismatch can suggest a gap in our ecological 

knowledge of how communities assemble: if we cannot successfully rebuild a community, we have yet 

to learn something about its assembly. A better understanding of how restored ecological communities 

assemble will thus inform both ecology and restoration. Interannual variation in environmental 

conditions at the outset of restoration projects, or “year effects,” may contribute to variation among 

restoration outcomes. However, a lack of experimental tests limits understanding of the mechanisms 

driving year effects and if these effects persist over time. We conducted an experiment to evaluate year 

effects on the assembly of plant communities in tallgrass prairie restorations, testing a leading putative 

mechanism: variation in first year rainfall. We established replicate restored prairie plots at the same 

site in SW Michigan using identical methods in three different years. During each planting year, we 

manipulated precipitation with rain-out shelters, and tracked community assembly trajectories for up to 

four years. We asked how plant community assembly varied across: (i) planting years, (ii) first year 

precipitation treatments, and (iii) planting years within precipitation treatments (i.e. non-precipitation 

planting year effects). We also investigated whether the richness of the restoration seed mix used to 

restore these prairies interacted with planting year conditions to influence assembly trajectories. We 

found strong evidence that planting years, even when watering was held constant, resulted in different 

community compositions that lasted into future years. One planting year resulted in significantly higher 

 

53 

cover of non-sown species, while sown species cover and richness fluctuated with each survey year, and 

the richness of non-sown species was stable across planting years, survey years, and watering 

treatments.  We found no evidence that more diverse seed mixes mitigated any of these year effects. 

Our results illustrate that planting year effects can vary, demonstrating a complex interplay of 

deterministic and historically contingent assembly mechanisms, depending on the context. Embracing 

this unpredictability in restoration outcomes suggests increased focus on post-planting adaptive 

management that can mediate initial differences in community composition that are not consistent with 

restoration goals. 

 

INTRODUCTION 

 

 

Ecological restoration has the potential to slow or reverse biodiversity losses globally (Dobson et 

al. 1997; Hobbs & Harris 2001; Ruiz-Jaen & Mitchell Aide 2005). However, restoration efforts are prone 

to unpredictable outcomes, which can be an impediment to their success (Suding 2011; Jones et al. 

2018). Understanding why this variation occurs is an important question for restoration research and 

practice, while also serving as a valuable framework for testing community assembly theories within 

restored ecosystems (Palmer et al. 1997; Young et al. 2001; Wainwright et al. 2018). 

 

Restoration ecology has traditionally adopted a deterministic view of community assembly, 

demonstrated by its common use of the watchmaker analogy: just as an expert watchmaker can 

assemble a working watch from its components, restorationists should be able to reassemble working  

ecosystems from their parts (Harper 1987). This analogy suggests that restoration outcomes should be 

predictable and repeatable, given sufficient knowledge of how the systems work that we are trying to 

restore. However, community ecology theory acknowledges both deterministic and stochastic 

(unpredictable) drivers of assembly (e.g., Drake 1990; Belyea 2004; Weiher et al. 2011; Fukami 2015) 

 

54 

and the literature shows increasing support for stochastic assembly mechanisms in restoration (e.g., 

Wilsey et al. 2014; Vaughn & Young 2015; Werner et al. 2016; Young et al. 2016). For instance, 

experiments altering the assembly histories of restored prairie sites have shown that the timing of 

arrival of native and exotic species has significant effects on the resulting communities via priority 

effects (Dickson et al. 2012; Martin & Wilsey 2012). This suggests that the watchmaker analogy may be 

inadequate for restoration. Instead, the disassembly and reassembly of an ecosystem may not 

repeatedly render the same outcome, even when conducted in the same way each time. Although this 

suggests variation may be inevitable, identifying drivers of stochasticity may help us understand 

variation in restoration outcomes.  

In practice, the causes and consequences of variation in assembly history for restoration 

outcomes remain largely unexplored. There is evidence that weather conditions can exert a selective 

force on an establishing community (Bakker et al. 2003; Seabloom et al. 2003; MacDougall et al. 2008; 

Vaughn & Young 2010; Stuble et al. 2017). Despite being an importance source of stochasticity in 

restoration, differences in planting year weather, often called “year effects,” are not well understood  by 

ecologists because the vast majority of restoration projects and ecological studies are not replicated in 

multiple years (Johnson 2002; Vaughn & Young 2010). Moreover, when planting year effects have been 

considered in restoration, it has often taken place by comparing sites planted in different years and 

locations, confounding planting year effects with site characteristics and/or management approaches; 

few studies have worked to disentangle these effects (though see Stuble et al. 2017; Groves & Brudvig 

2018). Restorations initiated by seed sowing should be particularly sensitive to year effects, because of 

the way species vary in their requirements for germination (Harper 1977; Baskin & Baskin 1988; Larson 

et al. 2015). As a result, the subset of species emerging from a sown seed mix (as well as the non-sown 

weed species emerging from the seed bank) is likely to depend on the weather conditions at the time of 

 

55 

or soon after sowing, and this can result in different initial plant communities (Groves and Brudvig 

2018).   

 

Most of the evidence for year effects points to variation in precipitation as a key driving 

mechanism; however, this has rarely been tested (but see Chapter 2; Groves & Brudvig 2018). Most, if 

not all, studies documenting year effects show correlations with precipitation in the planting year 

(Bakker et al. 2003; MacDougall et al. 2008; Stuble et al. 2017), but cannot demonstrate causality. There 

is also literature documenting a correlation between precipitation variability with outcomes of, for 

instance, competitive interactions between native and exotic species (e.g., Seabloom et al. 2003). That 

said, there are countless abiotic and biotic variables that could vary inter-annually and have an impact 

on plant community assembly; for instance, fluctuations in population sizes of rodents (e.g., Getz and 

Hofmann 1999) could impact a sown community through post-dispersal seed predation (Hulme & 

Kollmann 2005; Notman & Villegas 2005) or herbivory (Howe & Brown 1999; MacDougall & Wilson 

2007; Howe 2008). Many of these variables might also correlate with patterns of precipitation. 

Alternatively, aspects of weather besides precipitation, like springtime warming patterns that dictate 

the phenology of germinating plant species (Forrest & James 2011) could in turn alter the identities of 

early community members (Wainwright & Cleland 2013). Causal evidence is therefore needed to 

support the hypothesis that precipitation is the leading cause of year effects. 

Creating a compelling case for year effects will require understanding of how differences caused 

by planting year conditions play out over time in an assembling community. Early differences in a 

community can have significant impact on future community composition because of the way later 

arrivers interact with the early community, even when the early community members do not persist 

(Connell & Slatyer 1977; Belyea & Lancaster 1999). However, if deterministic mechanisms drive 

assembly, differences in initial community composition may dissipate over time. 

 

56 

If year effects are pervasive, managers may seek preventative actions, for instance by sowing a 

more diverse seed mix that spans a greater range of germination requirements or contains more species 

that can germinate under a wide range of conditions. A higher seed mix richness should improve total 

establishment (Myers & Harms 2009) but also might increase variation possible under diverse conditions 

(Chase 2003; Grman & Brudvig 2014). Additionally, understanding year effects may allow managers to 

better inform remedial management, for instance the need to sow additional seeds or intensify weed 

control after certain planting conditions. At present, however, we know of only one study that has 

considered whether aspects of management can mitigate the influence of year effects for plant 

communities undergoing restoration (Deering & Young 2006), and none that consider seed mix richness. 

We tested the effects of planting year conditions on community assembly in a tallgrass prairie 

restoration experiment. We planted replicate prairies using identical methods and experimental 

treatments, in three planting years and followed the communities over time. In a previous paper 

(Groves & Brudvig 2018), we report results from the first year following establishment of the restoration 

and found evidence of planting year effects; we now ask whether these lead to longer-term differences 

in community composition. We follow the communities over time (1-3 years following establishment) to 

determine if initial differences in communities planted (I) in different years, (II) under different 

precipitation regimes, and (III) in different years under the same precipitation regimes (i.e. non-

precipitation year effects) persist over time; and (IV) if seed mix richness at planting mediates these year 

effects and improves the establishment and cover of more sown native species. 

 

 

 

 

 

 

57 

METHODS 

 

Study site and restoration methods 

We conducted this experiment at the Lux Arbor Reserve in Prairieville, Michigan; a part of Michigan 

State University’s W. K. Kellogg Biological Station. Prior to restoration, the one-hectare field site 

contained old field vegetation typical to the area (e.g. dominated by Bromus inermis, Centaurea steobe, 

and other native and exotic weedy perennials) and was mowed as needed to prevent woody 

encroachment. The site was a mixed-oak savanna before being cleared for cultivation sometime before 

1946 and was used as a private air-strip in the 1960s and 1970s before its abandonment sometime 

before 1980.  

 

Experimental design 

 

We used a randomized complete split-split plot design to test the effects of two seed mixes 

nested within three precipitation treatments within three planting years (n=6 for each treatment 

combination). In each of three planting years (2014, 2015, 2016), we randomly converted six of eighteen 

15.2 m x 36.6 m blocks from old field to tallgrass prairie. For each planting, in early May, as close to the 

same calendar date as possible, we prepared and planted the blocks by mowing, spraying herbicide 

(41% glyphosate, 2.5 L/ha), tilling (15.2 cm depth), hand broadcasting native prairie seeds, and 

cultipacking. Planting was completed on 19-20 May of each planting year, with each block sown with a 

baseline prairie mix (8 species, detailed below, 379 seeds/m2). Within each of the six blocks (15.2 x 36.6 

m) planted each year, four smaller experimental plots (1.0 x 0.5 m) were randomly assigned one of three 

precipitation treatments or exposed to ambient weather conditions. Precipitation treatments were 

controlled with rain-out shelters that diverted and collected all rainfall from the plots; the collected rain 

water was used to water the plots at a steady rate (twice/week) that reflected (a) the 30-year record 

 

58 

low for the area (1.39 cm/month), (b) the 30-year average rainfall (9.91 cm/mo), or (c) above average 

rainfall (mean + 1.5 standard deviations, 17.81 cm/mo; Table 3.1). The experimental plots were 

surrounded by rubber gardening edging to prevent the overland flow of water in or out of the watering 

treatment plots. Additional details for this experimental setup can be found in Groves & Brudvig (2018).  

The precipitation treatment plots were split again, with one half (0.5 x 0.5 m) randomly assigned 

to a high species richness treatment that received seeds of 22 additional species (379 seeds/m2) or a low 

species richness treatment that received seeds of the same 8 species planted in the background (379 

seeds/m2). This created two seed mix richness treatments of 8 or 30 species (total sown density, 758 

seeds/m2). Species were chosen to represent a typical restoration planting in our area, as well as to 

ensure the low and high richness mixes had comparable proportions of seeds from different life-form 

groups (forb, legume, C3 grass, C4 grass; Table 2.1). Seed mixes were adjusted each year to account for 

annual differences in viability of seed, so that each planting received the same amount of live seed for 

each species. Seeds were local genotypes acquired from the same seed distributor each year.  

In 2016, a seed supplying error resulted in sowing Echinacea pallida instead of E. purpurea; as a 

result, the relative abundance (percent cover) of these plants was used as a predictor in statistical 

analyses and not included in any response variables. Because these two species typically differ in 

establishment success in restorations (E. purpurea is known to establish readily while E. pallida is 

notoriously difficult; this was confirmed in our field studies), we felt this was a fair way to ensure these 

plants did not confound our results. Additionally, one of the non-sown species prevalent in the first-year 

community, Panicum capillare, set seed again after being mowed to prevent propagule pressure from 

the planted prairies to affect the neighboring plots. We saw an increase in abundance of this species in 

each subsequent planting year, and believe this is an artifact of the experimental design, and not a true 

year effect. Therefore, we have also used the abundance of this species in the planting year as a 

predictor throughout each statistical model, and excluded it from all response variables. 

 

59 

Plant surveys 

We conducted surveys of the plant community in each 0.5 x 0.5 m experimental plot in late July 

each year (2014-2017) by visually identifying each species and quantifying the percent cover of the plot 

it occupied.  

 

Statistical analyses 

We used linear mixed models to examine how sown and non-sown species richness and cover 

responded to fixed effects of planting year, seed mix richness, and watering treatment (where relevant), 

random effects of block and plot the abundance of Echinacea sp. and the abundance of P. capillare in 

the planting year (detailed above; package “lme4,” Bates et al. 2015). We considered ambient (no rain-

out shelter) and precipitation treatment (rain-out shelter) plots in separate analyses in case of possible 

side effects of the shelters, e.g., greenhouse effects. We analyzed the response of community 

compositions (averaged across split-plots/seed mix richness treatments) to planting year and watering 

treatment using permutational multivariate analysis of variance stratified by planting block 

(PERMANOVA, Anderson 2001; package “vegan,” Oksanen et al. 2017). For all analyses, we compared 

plots by age (1st year, 2nd year, 3rd year) and sampling year (2017), to assess their individual effect on 

variability within the communities. We visualized differences in plant communities using nonmetric 

multidimensional scaling (NMDS) in three dimensions using Bray-Curtis dissimilarity. Post-hoc 

comparisons for significant predictors were conducted using Tukey adjustments for multiple 

comparisons (package “lsmeans,” Lenth 2016), except for PERMANOVA post-hoc tests, for which we 

manually conducted pairwise comparisons. All analyses were conducted in R version 3.3.2 (Fox and 

Weisberg 2011; Bates et al. 2015; R Core Team 2016). 

 

 

 

60 

RESULTS 

 

Of the 30 species sown, we observed a total of 26 as adults within the experimental plots. Over the 4 

years of this experiment, we also observed an additional 66 species that were not sown but emerged 

within the experimental plots. Weather observed during the 2014, 2015, and 2016 planting years is 

reported in Table 3.1. 

 

Community composition 

 

We found evidence for year effects on the composition of the restored prairie communities. 

Communities in plots that were exposed to ambient conditions, as well as those that received watering 

treatments, were dissimilar across the three planting years, and these differences remained over time 

(Figure 3.1; photos, Figure 3.2). Initial effects of watering treatments on community composition in the 

first year dissipated over time, and there was not a significant effect of first-year watering treatment on 

community composition in later years (Table 3.2).  

 

Cover of sown species 

 

We found an effect of planting year on the cover of sown species; however, this appeared to be 

a result of how age related to survey year (Figure 3.3). A comparison of prairie plots in their respective 

third growing seasons showed 3.8 times higher cover in plots planted in 2015 compared to those 

planted in 2014 (for ambient first-year plots as well as watering treatment plots, Table 3.3). However, 

this is likely an artifact of the different survey years (2017 and 2016, respectively), because the 2016 

growing season resulted in low sown cover for all prairie plots, regardless of planting age (mean 6%), 

while the 2017 growing season resulted in high sown cover for all prairie plots, regardless of age or 

initial conditions (mean 24%). Additionally, effects of watering treatment, seed mix richness, or 

 

61 

interactions therein on the cover of sown species did not persist into the third growing season (or 2017, 

Table 3.3). 

 

 

Richness of sown species 

 

The richness of sown prairie species likewise was more dependent on the conditions in the year 

of the survey than those in the year of planting (Figure 3.4). Prairie plots planted in 2014, though they 

had lower sown species richness in their first year when compared to 2015 and 2016 plots (Figure 3.4), 

no longer had significantly lower richness by 2017 (Table 3.4). Although the mean and high watering 

treatments had significantly higher richness in the first growing season, these effects did not persist into 

the third growing season (or 2017, Table 3.4). Finally, increased seed mix richness did increase the 

richness of sown species observed across all years; however, we found no significant interactions 

between seed mix richness and planting year or watering treatment, suggesting that mix richness does 

not alter the impact of planting year effects. 

 

Non-sown species 

 

Year effects were apparent in the cover of non-sown species, driven by differences in the 2016 

prairie plots (Figure 3.5). Non-sown cover in the planting year was negatively correlated with the 

abundance of P. capillare (Table 3.5). P. capillare, an annual, was not present in later growing seasons, 

and non-sown cover in all plots increased over time (Figure 3.5). However, the 2016 plantings, which 

had the most P. capillare and the least cover of all other non-sown species in their first year, had the 

most non-sown cover of all the prairies in their second year, 2017—especially in plots that had been 

watered in their first year—and surpassed the steady increase in non-sown cover seen in the 2014 and 

2015 plantings. We found a significant effect of planting year on non-sown cover across second year 

plots and across 2017 plots because of this spike in non-sown cover for these 2016 plots in their second 

 

62 

year. We found no other significant predictors for non-sown cover that persisted into their third growing 

season (or 2017, Table 3.5). We found no effect of planting year, watering treatment, or seed mix 

richness on the richness of non-sown species during any year after planting (Table 3.6). 

 

DISCUSSION 

 

 

We provide experimental evidence that planting year effects alter community assembly 

trajectories in restoration. Specifically, we found that after holding all site and management variables 

constant—even watering the same amount— plant communities differed across planting years and 

these differences persisted. These differences were not caused by watering, because although the 

watering treatments resulted in different first-year communities, these differences did not persist over 

time. Instead, these results suggest that a factor besides water availability drive year effects on these 

communities. Additionally, although high seed mix richness increased the richness of sown species, we 

found no evidence that that increased richness mitigated planting year effects to improve establishment 

and cover of sown native species. Finally, we found that key metrics of restoration success, namely sown 

and non-sown cover and sown and non-sown species richness, were generally not affected by year 

effects. Sown cover and sown richness both varied in tandem with favorable and unfavorable growing 

seasons, regardless of their planting year, and non-sown richness showed no response to planting year, 

survey year, or watering treatment. Only the cover of non-sown species was different across plots 

planted in different years, with the 2016 plantings having significantly higher weed cover in their second 

year than any other plantings then or in any prior year. 

 Our direct test of planting year, along with manipulation of precipitation— the leading 

hypothesized mechanism for planting year effects— allows us to infer causality for these effects without 

confounding with site- and management-related variables. We did not find evidence that precipitation 

 

63 

caused year effects; on the contrary, we found that year-to-year variation was important even across 

plots where precipitation had been held constant in different years. This mechanistic backing suggests 

that some other factor—likely correlated with precipitation—may have been at play in previous studies 

that have documented year effects that correlated with rainfall (Bakker et al. 2003; MacDougall et al. 

2008; Vaughn & Young 2010; Stuble et al. 2017).  

The planting-year differences we saw in community composition persisted into later growing 

seasons. This is an important consideration, given that under deterministic assembly, communities could 

converge in response to weather or other conditions (e.g. soils) in later years despite initial differences 

(e.g., Weiher et al. 1998; Turnbull et al. 2005). Tracking these communities over time allowed us to 

disentangle planting year effects from effects of survey years and site age (Stuble et al. 2017). Further 

monitoring of this site will allow us to evaluate whether or not these communities will eventually 

converge and, if so, over what time scale.  

The plantings that resulted in the most non-sown species cover were planted in the year with 

the driest summer (2016, Table 3.1). Other studies documenting year effects have demonstrated similar 

patterns, e.g. Bakker et al. (2003) and MacDougall et al. (2008) saw a strong hinderance to restoration 

outcomes after a particularly dry planting year, while two other normal planting years had similar 

outcomes to one another. Because we also manipulated rainfall availability directly through watering 

each year, and found no effect on non-sown species cover, we can infer that it is not the lack of rainfall 

alone that causes these differences in assembly in these restored communities. Instead, there may be 

other factors that correlate with a dry summer—such as variation in heat, humidity, or cloud cover—

that influence the phenology of the sown and non-sown seeds, or that biotic considerations like 

granivores and herbivores are at play (e.g., Hulme & Kollmann 2005; Howe 2008; Wainwright & Cleland 

2013). More research is needed to disentangle the influence, and relative impact, of these drivers. 

 

64 

Increased seed mix richness was correlated with increased sown species richness across our 

experiment. Increasing mix richness has many proven benefits in restoration (e.g. Nemec et al. 2013; 

reviewed in Myers & Harms 2009). We did not, however, find evidence that increasing mix richness 

mitigated planting year effects (see also Carter & Blair 2012). We predicted to find interactions between 

mix richness and planting year conditions, driven by an increased likelihood that the high richness mix 

would contain species suited for any given conditions (Chase 2003; Tilman et al. 2006). We found no 

such interactions except one transient (i.e. seen in 2nd-year prairies only) in which the low richness seed 

mix correlated with an increase in relative abundance of sown species for prairies planted in 2016. 

Instead of seeing the “portfolio effects” (Hooper 2005; Schindler et al. 2015) that we expected, in which 

diversity decreases risk of establishment failure and increases total success, this seemingly counter-

intuitive result highlights the potential that a less diverse mix, though theoretically riskier, will 

occasionally be particularly suited for the conditions and out-perform the high richness mix in terms of 

total establishment (e.g., Aarssen et al. 2003). This outcome might be particularly common in 

restoration, where less diverse seed mixes are typically compositional subsets of the most reliable 

species found in more diverse seed mixes (e.g., Packard & Mutel 1997; see Groves & Brudvig 2018). 

When this is the case, post-first-year management may be necessary to mitigate year effects. For 

instance, additional seed additions or increased weed control following for restorations planted during 

poor conditions may be very important (Rowe 2010; Wilson 2015). 

One possible unintended source of variation in this experiment is the introduction of year 

effects through the seed mix. Our seeds were sourced from the same supplier each year and adjusted 

based on annually updated tests of seed viability; however, these seed “populations” could have been 

genetically distinct across years in a way that impacted how they would respond to the site or weather 

conditions. Seeds collected in different years could also come from maternal plants that have faced 

 

65 

different conditions as they produced seed, leading to different seed quality between batches, 

regardless of general viability (Roach & Wulff 1987; Mousseau & Fox 1998; Wang & Frei 2011). 

Our work highlights the important role of stochasticity in community assembly. Strong effects of 

planting conditions on community composition in the first year persisted into subsequent growing 

seasons. However, basic metrics of restoration success often responded more to weather or other 

conditions in later survey years. We sought to make broad generalizations about assembly trajectories in 

this system by controlling as many factors as possible (site, management, and sown species), and the 

data tell a compelling story about variation in composition across similarly “good” outcomes. Returning 

to the watch maker analogy, we learned that ensuring that watch parts are the same does not ensure 

recreation of the same watch; quite the contrary—restoration trajectories are diverse in ways that 

remain unexplained. This variation can be a good thing, as long as it stays within the bounds of what 

we'd consider acceptable outcomes for restoration. In other words, we don't want to increase variation 

among outcomes by including failures, but variation in general can increase beta diversity across sites 

and, thus, landscape-level diversity. Managers working in restoration may benefit from expecting and 

embracing this variation, sharing their attention between trusted restoration methods and how to 

respond to divergent outcomes from those same methods.  

 

ACKNOWLEDGEMENTS 

 

We thank Meg Kargul, Michelle Franklin, Katie Grantham, and Logan Brissette for their 

assistance in the field, Native Connections for supplying seed, Jared Foster for mix and planting advice, 

Mark Manuszak for field logistics, and the W.K. Kellogg Biological Station for the use of this field site. 

George Groves helped engineer the rain-out shelter design; Mike Bassett, Maddy Cleary, George 

Wheeler, Evan Johnson, and Andy Funk assisted with building and tear-down of shelters. This study was 

 

66 

funded by two Michigan State University Graduate School Fellowships and two Porter Research Awards, 

all graciously awarded by the W.K. Kellogg Biological Station. This material is based upon work 

supported by the National Science Foundation Graduate Research Fellowship under Grant No. DGE-

1424871 (AMG) and DEB-1552197 (LAB).

 

67 

 

 

APPENDIX 

68 

APPENDIX 

Figures 

 

 

Figure 3.1. Non-metric Multidimensional Scaling (NMDS) of plots planted in 2014, 2015, and 2016 over 

time. Plots shown received watering treatments in their first growing season. (A) shows the progression 

of individual plots. Points of each color group (red, purple, green) represent plots planted in the three 

planting years (2014, 2015, 2016, respectively) and become darker with age. (B) shows the centroid of 

each group of plots. For both, points closer together represent communities that are more similar in 

their species composition.  

 

 

69 

(A) Watering treatment plotsover time(A) 2014 

(B) 2015 

(C) 2016 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 3.2 Photos taken in 2017 of prairie plots planted in (A) 2014, (B) 2015, and (C) 2016. Phots were 

taken July 25-August 1, 2017, at the W.K. Kellogg Biological Station’s Lux Arbor Reserve in Prairieville, 

Michigan. 

 

 

70 

 

Figure 3.3. Sown species cover over time in plots that were exposed to ambient conditions in their 

first year. 

 

 

 

71 

 

Figure 3.4. Sown species richness over time in plots that were exposed to ambient conditions in their 

first year. 

 

 

 

72 

 

Figure 3.5. Non-sown species cover over time in plots that received watering treatments in their first 

year. 

 

 

 

73 

Tables 

 

 

Table 3.1. Weather conditions in three planting years. 1981-2010 averages from Michigan State 

Climatologist’s Office (Gull Lake Weather Station, https://climate.geo.msu.edu, accessed 14 May 2018). 

2014-2016 records from Weather Underground (Battle Creek Weather Station, 

https://www.wunderground.com, accessed 14 May 2018). 

 
 

 

Spring weather 

(March 1 – May 31) 

Summer weather 
(June 1 – July 31) 

 
)
C
°
(
 
e
r
u
t
a
r
e
p
m
e
t

 

w
o

l
 

e
g
a
r
e
v
A

 
)
C
°
(
 
e
r
u
t
a
r
e
p
m
e
t

 
)

m
c
(
 
n
o
i
t
a
t
i
p
i
c
e
r
p

 
l

a
t
o
T

 
)
C
°
(
 
e
r
u
t
a
r
e
p
m
e
t

 

w
o

l
 

e
g
a
r
e
v
A

 
)
C
°
(
 
e
r
u
t
a
r
e
p
m
e
t

 
)

m
c
(
 
n
o
i
t
a
t
i
p
i
c
e
r
p

 
l

a
t
o
T

 

 

i

h
g
h
e
g
a
r
e
v
A

 

 

i

h
g
h
e
g
a
r
e
v
A

1981-2010 

2.9 

15.8 

25.5 

15.1 

28.5 

19.9 

2014 

1.1 

12.8 

16.3 

15.6 

25.6 

24.8 

2015 

3.3 

15.0 

18.4 

15.6 

26.1 

32.5 

2016 

3.9 

15.0 

21.2 

15.0 

27.2 

13.4 

 

74 

 

 

Table 3.2. Statistical model results predicting similarity of community composition. Subplots are 

averaged across high and low seed mix richness. PERMANOVA models for watered first-year plots are 

stratified by block. P. capillare abundance included as a model predictor refers to the abundance of that 

species in the planting year. Echinacea sp. abundance included as a model predictor refers to abundance 

of that species in the survey year. Pairwise comparisons were conducted manually using subsetted data 

in the PERMANOVA models. 

A. In plots exposed to ambient weather conditions in their first year 
Community dissimilarity in the first year 
Predictor 
Year planted 

F 
2.52  0.0002*** 

R2 
0.13 

p-value 

Echinacea sp. abundance  0.90  0.56 
P. capillare abundance 
2.16  0.0078** 
Community dissimilarity in the second year 
Predictor 
p-value 
Year planted 

F 
4.21  0.0002*** 

0.023 
0.055 

R2 
0.20 

0.013 
0.028 

R2 
0.14 
0.022 
0.12 

R2 

Model R2=0.21 
2014-2015 F=1.66, p=0.093 
2014-2016 F=2.62, p=0.00030*** 
2015-2016 F=3.12, p=0.00020*** 
 
 

Model R2=0.25 
2014-2015 F=5.79, p<0.0001*** 
2014-2016 F=4.49, p=0.00020*** 
2015-2016 F=2.41, p=0.020* 
 
 

Model R2=0.29 
 
 
 

Model R2=0.25 
2014-2015 F=2.85, p=0.0099** 
2014-2016 F=4.70, p<0.0001*** 
2015-2016 F=4.49, p<0.0001*** 
 
 

Echinacea sp. abundance  0.52  0.90 
P. capillare abundance 
1.17  0.29 
Community dissimilarity in the third year 
Predictor 
p-value 
Year planted 
Echinacea sp. abundance  0.61  0.72 
P. capillare abundance 
Community dissimilarity in 2017 
Predictor 
Year planted 

F 
4.03  0.0021** 

3.50  0.0035** 

p-value 

F 
4.10  <0.0001***  0.20 

1.70  0.083 

Echinacea sp. abundance  0.62  0.80 
P. capillare abundance 
 
 
 
 
 
 
 

 

0.015 
0.041 

75 

p-value 

R2 

Watering treatment 

1.56  0.019* 

0.027 

0.012 
Echinacea sp. abundance  1.42  0.12 
3.34  <0.0001***  0.028 
P. capillare abundance 
Year planted * Watering 
1.51  0.0040** 
0.051 
Community dissimilarity in the second year 
Predictor 
p-value 
Year planted 

F 
8.22  <0.001*** 

R2 
0.13 

F 
5.33  <0.0001***  0.091 

      3.2      ’   
 
B. In plots receiving watering treatments in their first year 
 
Community dissimilarity in the year planted 
Model R2=0.21 
Predictor 
2014-2015 F=5.38, p<0.0001*** 
Year planted 
2014-2016 F=6.86, p<0.0001*** 
2015-2016 F=3.84, p<0.0001*** 
Low-mean F=1.58, p=0.029* 
Low-high F=1.58, p=0.024* 
Mean-high F=1.58, p=0.026* 
 
 
 
 
Model R2=0.25 
2014-2015 F=7.44, p<0.0001*** 
2014-2016 F=7.50, p<0.0001*** 
2015-2016 F=6.70, p<0.0001*** 
 
 
 
 
 
Model R2=0.18 
 
 
 
 
 
 
Model R2=0.25 
2014-2015 F=6.15, p<0.0001*** 
2014-2016 F=13.76, p<0.0001*** 
2015-2016 F=6.63, p<0.0001*** 
 
 
 
 

Watering treatment 
Echinacea sp. abundance  3.26  0.0012** 
P. capillare abundance 
Year planted * Watering 
Community dissimilarity in the third year 
F  
Predictor 
p-value 
Year planted 
6.56  <0.001*** 
Watering treatment 
1.19  0.22 
Echinacea sp. abundance  1.42  0.35 
1.28  0.14 
P. capillare abundance 
Year planted * Watering 
1.44  0.10 
Community dissimilarity in 2017 
Predictor 
Year planted 

Watering treatment 
1.30  0.15 
Echinacea sp. abundance  1.48  0.15 
P. capillare abundance 
0.71  0.65 
Year planted * Watering 
1.76  0.0033** 
 

R2 
0.084 
0.030 
0.018 
0.016 
0.037 

p-value 

F 
9.84  <0.001*** 

2.06  0.018* 
1.55  0.0086** 

0.023 
0.026 
0.017 
0.050 

R2 
0.16 

0.021 
0.012 
0.0056 
0.056 

1.43  0.069 

 

 

76 

Table 3.3. Statistical model results predicting the cover of sown species. Linear mixed effect models for 

ambient first-year plots included a random effect of plot; models for watered first-year plots included 

random effects of plot and block. Cover of sown species was log(x+0.05) transformed to improve model 

fit. P. capillare abundance included as a model predictor refers to the abundance of that species in the 

planting year. Echinacea sp. abundance included as a model predictor refers to abundance of that 

species in the survey year. 

R2 
 
0.098 
0.0050 
0.0070 
0.032 
0.000 

R2 
 
0.14 
0.0070 
0.014 
0.023 

p-value 
0.00094 
0.12 
0.54 
0.46 
0.47 
0.0075**  0.19 

A. In plots exposed to ambient weather conditions in their first year 
Cover of sown species in the year planted 
χ2 
p-value 
Predictor 
0.0079 
7.07 
(Intercept) 
0.11 
4.43 
Year planted 
0.61 
Seed mix richness 
0.26 
0.62 
Echinacea sp. abundance  0.24 
0.38 
0.78 
P. capillare abundance 
Seed mix * Year planted 
0.79 
0.48 
Cover of sown species in the second year 
χ2 
Predictor 
10.94 
(Intercept) 
4.21 
Year planted 
0.38 
Seed mix richness 
Echinacea sp. abundance  0.55 
0.53 
P. capillare abundance 
Seed mix * Year planted 
9.80 
Cover of sown species in the third year 
p-value 
Predictor 
0.0069 
(Intercept) 
0.0091**  0.25 
Year planted 
0.12 
Seed mix richness 
0.10 
0.00 
Echinacea sp. abundance  0.0011  0.97 
0.20 
P. capillare abundance 
Seed mix * Year planted 
0.055 
Cover of sown species in 2017 
χ2 
Predictor 
24.98 
(Intercept) 
Year planted 
0.46 
Seed mix richness 
3.60 
Echinacea sp. abundance  1.12 
P. capillare abundance 
1.40 
Seed mix * Year planted 
3.30 
 

p-value 
<0.0001 
0.79 
0.058 
0.29 
0.24 
0.19 

R2 
 
0.017 
0.081 
0.034 
0.051 
0.039 

χ2 
7.29 
6.80 
2.68 

4.91 
1.16 

0.027* 
0.28 

Model R2=0.21 
 
 
 
 
 
 

Model R2=0.29 
 
 
 
 
 
 

Model R2=0.61 
 
 
 
 
 
 

Model R2=0.33 
 
 
 
 
 
 

R2 
 

 

77 

4.76 
12.34 
3.51 
0.13 

p-value 
0.0023 
0.12 
0.10 
0.010** 
0.17 
0.72 
0.69 
0.098 
0.051 

      3.3      ’   
 
B. In plots receiving watering treatments in their first year 
 
Cover of sown species in the year planted 
χ2 
Model R2=0.28 
p-value 
Predictor 
0.014 
5.98 
 
(Intercept) 
0.20 
3.24 
 
Year planted 
0.28 
2.57 
Watering treatment 
 
0.0052  0.94 
 
Seed mix richness 
0.78 
 
Echinacea sp. abundance  0.077 
0.029* 
 
P. capillare abundance 
0.015* 
Year planted * Watering 
 
 
0.17 
Seed mix * Year planted 
 
Seed mix * Watering 
0.94 
 
Cover of sown species in the second year 
χ2 
Model R2=0.33 
Predictor 
9.33 
 
(Intercept) 
4.30 
 
Year planted 
4.55 
Watering treatment 
 
 
Seed mix richness 
6.72 
 
Echinacea sp. abundance  1.85 
 
0.13 
P. capillare abundance 
Year planted * Watering 
2.24 
 
 
4.64 
Seed mix * Year planted 
 
Seed mix * Watering 
5.97 
 
Cover of sown species in the third year 
χ2 
Model R2=0.47 
Predictor 
8.60 
 
(Intercept) 
3.40 
 
Year planted 
5.24 
Watering treatment 
 
 
Seed mix richness 
2.48 
 
Echinacea sp. abundance  0.94 
 
0.67 
P. capillare abundance 
Year planted * Watering 
1.09 
 
 
1.33 
Seed mix * Year planted 
 
Seed mix * Watering 
0.83 
 
Cover of sown species in 2017 
χ2 
Model R2=0.32 
Predictor 
 
31.41 
(Intercept) 
 
2.32 
Year planted 
Watering treatment 
3.48 
 
 
Seed mix richness 
0.95 
 
Echinacea sp. abundance  0.55 
 
3.42 
P. capillare abundance 
Year planted * Watering 
3.03 
 
 
4.09 
Seed mix * Year planted 
Seed mix * Watering 
2.27 
 

p-value 
0.0034 
0.065 
0.073 
0.12 
0.33 
0.41 
0.58 
0.25 
0.66 

p-value 
<0.001 
0.31 
0.18 
0.33 
0.46 
0.064 
0.55 
0.13 
0.32 

R2 
 
0.046 
0.014 
0.000 
0.0010 
0.043 
0.11 
0.045 
0.0010 

R2 
 
0.061 
0.027 
0.041 
0.018 
0.0020 
0.037 
0.041 
0.043 

R2 
 
0.067 
0.056 
0.020 
0.014 
0.012 
0.014 
0.010 
0.0070 

R2 
 
0.020 
0.018 
0.0070 
0.0050 
0.037 
0.026 
0.014 
0.020 

 

78 

Table 3.4. Statistical model results predicting the richness of sown species. Negative binomial mixed 

effect models for ambient first-year plots included a random effect of plot; models for watered first-year 

plots included random effects of plot and block. P. capillare abundance included as a model predictor 

refers to the abundance of that species in the planting year. Echinacea sp. abundance included as a 

model predictor refers to abundance of that species in the survey year. 

0.58 
0.19 
0.058 
0.29 
0.48 

R2 
 
0.027 
0.067 
0.11 
0.035 
0.038 

A. In plots exposed to ambient weather conditions in their first year 
Richness of sown species in the year planted 
χ2 
Predictor 
p-value 
32.91  <0.001 
(Intercept) 
Year planted 
1.11 
Seed mix richness 
1.70 
Echinacea sp. abundance  3.60 
1.14 
P. capillare abundance 
Seed mix * Year planted 
1.47 
Richness of sown species in the second year 
χ2 
Predictor 
p-value 
(Intercept) 
27.45  <0.001 
0.20 
3.25 
Year planted 
0.49 
Seed mix richness 
0.48 
0.68 
Echinacea sp. abundance  0.17 
0.014  0.90 
P. capillare abundance 
Seed mix * Year planted 
0.16 
3.62 
Richness of sown species in the third year 
χ2 
p-value 
Predictor 
31.32  <0.001 
(Intercept) 
0.050* 
3.86 
Year planted 
Seed mix richness 
2.04 
0.15 
Echinacea sp. abundance  0.063  0.80 
0.39 
0.73 
P. capillare abundance 
Seed mix * Year planted 
0.54 
0.46 
Richness of sown species in 2017 
χ2 
p-value 
Predictor 
62.01  <0.001 
(Intercept) 
0.84 
Year planted 
Seed mix richness 
4.59 
Echinacea sp. abundance  0.66 
0.10 
P. capillare abundance 
Seed mix * Year planted 
0.77 
 
 
 

R2 
 
0.045 
0.10 
0.00 
0.0020 
0.042 

R2 
 
0.14 
0.015 
0.00 
0.001 
0.14 

R2 
 
0.20 
0.10 
0.017 
0.028 
0.028 

0.66 
0.032 
0.42 
0.75 
0.68 

Model R2=0.40 
 
 
 
 
 
 

Model R2=0.38 
 
 
 
 
 
 

Model R2=0.45 
 
 
 
 
 
 

Model R2=0.44 
 
 
 
 
 
 

 

79 

      3.4      ’   
 
B. In plots receiving watering treatments in their first year 
Richness of sown species in the year planted 
χ2 
Predictor 
p-value 
36.19  <0.001 
(Intercept) 
Year planted 
7.58 
0.023* 

R2 
 
0.16 

Watering treatment 

11.60  0.0030** 

0.092 

0.27 
0.66 

6.29 

0.012* 

Seed mix richness 
Echinacea sp. abundance  0.089  0.77 
0.033  0.86 
P. capillare abundance 
0.11 
7.53 
Year planted * Watering 
Seed mix * Year planted 
3.64 
0.16 
Seed mix * Watering 
0.70 
0.72 
Richness of sown species in the second year 
χ2 
Predictor 
p-value 
(Intercept) 
50.93  <0.0001 
2.59 
Year planted 
0.84 
Watering treatment 
Seed mix richness 
15.87  <0.0001***  0.17 
0.053 
Echinacea sp. abundance  3.74 
0.66 
0.19 
P. capillare abundance 
0.91 
0.97 
Year planted * Watering 
0.37 
Seed mix * Year planted 
2.00 
Seed mix * Watering 
0.13 
4.08 
Richness of sown species in the third year 
χ2 
p-value 
Predictor 
50.49  <0.0001 
(Intercept) 
0.40 
Year planted 
4.97 
Watering treatment 
Seed mix richness 
7.19 
Echinacea sp. abundance  3.08 
P. capillare abundance 
Year planted * Watering 
Seed mix * Year planted 
Seed mix * Watering 
 
 
 
 
 

0.041  0.84 
0.56 
0.76 
0.040  0.84 
0.29 
2.46 

0.53 
0.083 
0.0073** 
0.079 

0.099 
0.0010 
0.00 
0.16 
0.078 
0.013 

R2 
 
0.026 
0.010 

0.014 
0.0020 
0.0070 
0.024 
0.066 

R2 
 
0.014 
0.050 
0.13 
0.031 
0.0020 
0.021 
0.00 
0.031 

 

80 

 
Model R2=0.62 
 
Tukey post-hoc tests: 
2014-2015 z=-0.49, p=0.87 
2014-2016 z=-1.00, p=0.58 
2015-2016 z=-0.49, p=0.88 
Tukey post-hoc tests: 
High-low z=3.13, p=0.0049** 
High-mean z=-0.85, p=0.67 
Low-mean z=-3.91, p=0.0003** 
 
 
 
 
 
 
 
Model R2=0.48 
 
 
 
 
 
 
 
 
 
 
Model R2=0.39 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Table 3.4 (cont’d) 
 
Richness of sown species in 2017 
χ2 
p-value 
Predictor 
30.16  <0.001 
(Intercept) 
3.28 
Year planted 
4.32 
Watering treatment 
4.50 
Seed mix richness 
Echinacea sp. abundance  3.90 
0.45 
P. capillare abundance 
1.25 
Year planted * Watering 
Seed mix * Year planted 
3.65 
1.04 
Seed mix * Watering 
 
 

0.19 
0.12 
0.034* 
0.048* 
0.50 
0.87 
0.16 
0.60 

 
 
 
Model R2=0.46 
 
 
 
 
 
 
 
 
 

R2 
 
0.050 
0.032 
0.058 
0.033 
0.0050 
0.029 
0.048 
0.0070 

 

81 

Table 3.5. Statistical model results predicting the cover of non-sown species. Linear mixed effect 

models for ambient first-year plots included a random effect of plot; models for watered first-year plots 

included random effects of plot and block. Cover of non-sown species was square root transformed to 

improve model fit. P. capillare abundance included as a model predictor refers to the abundance of that 

species in the planting year. Echinacea sp. abundance included as a model predictor refers to abundance 

of that species in the survey year. 

15.32 
1.26 

R2 
 
0.046 
0.012 
0.028 

0.043 

R2 
 
0.0080 
0.00 
0.0060 
0.10 
0.059 

R2 
 
0.23 
0.023 
0.0070 
0.26 
0.022 

p-value 
<0.0001 
0.30 
0.48 
0.33 
<0.0001***  0.37 
0.53 

A. In plots exposed to ambient weather conditions in their first year 
Cover of non-sown species in the year planted 
χ2 
Predictor 
36.34 
(Intercept) 
2.44 
Year planted 
Seed mix richness 
0.51 
Echinacea sp. abundance  0.95 
P. capillare abundance 
Seed mix * Year planted 
Cover of non-sown species in the second year 
χ2 
p-value 
Predictor 
<0.0001 
(Intercept) 
72.03 
0.14 
Year planted 
0.93 
Seed mix richness 
0.0046  0.95 
0.59 
Echinacea sp. abundance  0.30 
0.14 
2.19 
P. capillare abundance 
Seed mix * Year planted 
4.73 
0.094 
Cover of non-sown species in the third year 
χ2 
p-value 
Predictor 
150.18  <0.0001 
(Intercept) 
4.02 
Year planted 
Seed mix richness 
1.15 
Echinacea sp. abundance  0.16 
3.46 
P. capillare abundance 
Seed mix * Year planted 
1.23 
Cover of non-sown species in 2017 
χ2 
Predictor 
90.81 
(Intercept) 
0.46 
Year planted 
Seed mix richness 
0.35 
Echinacea sp. abundance  0.028 
P. capillare abundance 
0.88 
Seed mix * Year planted 
4.33 
 

p-value 
<0.0001 
0.80 
0.55 
0.87 
0.35 
0.11 

R2 
 
0.020 
0.0060 
0.0010 
0.041 
0.046 

0.045* 
0.28 
0.69 
0.063 
0.27 

Model R2=0.54 
 
 
 
 
 
 

Model R2=0.25 
 
 
 
 
 
 

Model R2=0.41 
 
 
 
 
 
 

Model R2=0.29 
 
 
 
 
 
 

 

82 

      3.5      ’   
 
B. In plots receiving watering treatments in their first year 
Cover of non-sown species in the year planted 
Predictor 
(Intercept) 
Year planted 
Watering treatment 

p-value 
<0.0001 
0.40 
0.042* 

R2 
 
0.034 
0.086 

χ2 
80.45 
1.82 
6.35 

 
Model R2=0.39 
 
 
Tukey post-hoc tests: 
High-low t=1.63, p=0.24 
High-mean t=-0.20, p=0.98 
Low-mean t=-1.83, p=0.17 
 
 
 
 
 
 
 
Model R2=0.41 
 
Tukey post-hoc tests: 
2014-2015 t=0.029, p=0.96 
2014-2016 t=-3.97, p=0.0005*** 
2015-2016 t=-5.00, p<0.0001*** 
 
 
 
 
 
 
 
 
Model R2=0.30 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

R2 
 
0.063 

0.058 
0.013 
0.0050 
0.029 
0.061 
0.058 
0.026 

R2 
 
0.058 
0.014 
0.001 
0.005 
0.025 
0.011 
0.00 
0.032 

83 

0.022 
0.044* 
0.91 
0.00 
<0.0001***  0.21 
0.25 
0.50 
0.26 

0.076 
0.0050 
0.021 

0.0016** 

4.08 

Seed mix richness 
Echinacea sp. abundance  0.012 
22.48 
P. capillare abundance 
5.34 
Year planted * Watering 
Seed mix * Year planted 
1.40 
Seed mix * Watering 
2.68 
Cover of non-sown species in the second year 
χ2 
Predictor 
p-value 
152.08  <0.0001 
(Intercept) 
Year planted 
12.92 

0.088 
0.14 
0.47 
0.11 
0.31 
0.031* 
0.017* 

4.86 
Watering treatment 
Seed mix richness 
2.19 
Echinacea sp. abundance  0.52 
2.60 
P. capillare abundance 
4.76 
Year planted * Watering 
Seed mix * Year planted 
6.92 
Seed mix * Watering 
8.14 
Cover of sown species in the third year 
χ2 
p-value 
Predictor 
189.85  <0.0001 
(Intercept) 
3.11 
Year planted 
0.74 
Watering treatment 
Seed mix richness 
0.14 
Echinacea sp. abundance  0.33 
1.41 
P. capillare abundance 
0.89 
Year planted * Watering 
Seed mix * Year planted 
0.029 
Seed mix * Watering 
3.04 
 
 
 
 
 

0.078 
0.69 
0.71 
0.57 
0.24 
0.64 
0.87 
0.22 

 

 
 
 
Model R2=0.36 
 
Tukey post-hoc tests: 
2014-2015 t=2.43, p=0.049* 
2014-2016 t=-2.44, p=0.046* 
2015-2016 t=-4.85, p<0.0001*** 
 
 
 
 
 
 
 

Table 3.5 (cont’d) 
 
Cover of sown species in 2017 
χ2 
Predictor 
p-value 
182.65  <0.0001 
(Intercept) 
13.92 
Year planted 

R2 
 

0.00095***  0.073 

4.03 
0.26 

Watering treatment 
Seed mix richness 
Echinacea sp. abundance  0.014 
P. capillare abundance 
Year planted * Watering 
Seed mix * Year planted 
Seed mix * Watering 
 

3.36 
5.80 
0.40 
5.30 
 

0.13 
0.61 
0.90 
0.067 
0.21 
0.82 
0.071 

0.026 
0.0020 
0.00 
0.038 
0.039 
0.0010 
0.035 

 

84 

Table 3.6. Statistical model results predicting the richness of non-sown species. Linear mixed effect 

models for ambient first-year plots included a random effect of plot; models for watered first-year plots 

included random effects of plot and block. P. capillare abundance included as a model predictor refers 

to the abundance of that species in the planting year. Echinacea sp. abundance included as a model 

predictor refers to abundance of that species in the survey year. 

0.075 
0.69 
0.90 
0.68 
0.92 

0.039* 
0.79 
0.11 
0.99 
0.37 

p-value  R2 

0.17 
0.17 

0.20 
0.58 

Model R2=0.39 
 
 
 
 
 
 

Model R2=0.067 
 
 
 
 
 
 

 
0.14 
0.0050 
0.0030 
0.010 
0.0060 

p-value  R2 

χ2 
114.28  <0.0001 
0.59 
0.39 

χ2 
130.57  <0.0001 
5.19 
0.16 

A. In plots exposed to ambient weather conditions in their first year 
Richness of non-sown species in the year planted 
p-value  R2 
Predictor 
(Intercept) 
Year planted 
Seed mix richness 
Echinacea sp. abundance  0.017 
P. capillare abundance 
Seed mix * Year planted 
Richness of non-sown species in the second year 
Predictor 
(Intercept) 
0.75 
Year planted 
Seed mix richness 
0.53 
Echinacea sp. abundance  0.0010  0.98 
P. capillare abundance 
0.65 
Seed mix * Year planted 
0.75 
Richness of non-sown species in the third year 
Predictor 
(Intercept) 
Year planted 
Seed mix richness 
Echinacea sp. abundance  2.52 
0.00 
P. capillare abundance 
Seed mix * Year planted 
0.80 
Richness of non-sown species in 2017 
Predictor 
(Intercept) 
Year planted 
Seed mix richness 
Echinacea sp. abundance  2.96 
1.48 
P. capillare abundance 
Seed mix * Year planted 
0.41 
 
 
 

χ2 
105.55  <0.0001 
2.19 
0.34 
0.0001  0.99 
0.086 
0.22 
0.82 

p-value  R2 

χ2 
151.32  <0.0001 
4.25 
0.068 

 
0.022 
0.019 
0.00 
0.015 
0.015 

 
0.29 
0.011 
0.18 
0.037 
0.062 

 
0.15 
0.00 
0.16 
0.11 
0.047 

Model R2=0.36 
 
 
 
 
 
 

Model R2=0.41 
 
 
 
 
 
 

 

85 

p-value  R2 

1.20 
0.78 
0.74 
0.43 

p-value  R2 

 
0.017 
0.016 
0.012 
0.00 
0.022 
0.11 
0.001 
0.007 

 
0.012 
0.016 
0.0020 
0.0010 
0.018 
0.0060 
0.0060 
0.0060 

 
0.041 
0.021 
0.013 
0.0070 
0.0040 
0.0080 
0.00 
0.015 

χ2 
112.49  <0.0001 
1.83 
0.82 
0.16 

      3.6      ’   
 
B. In plots receiving watering treatments in their first year 
Richness of non-sown species in the year planted 
χ2 
p-value  R2 
Predictor 
71.00 
<0.0001 
(Intercept) 
0.81 
0.67 
Year planted 
2.69 
Watering treatment 
0.26 
0.32 
Seed mix richness 
0.99 
0.68 
Echinacea sp. abundance  0.17 
0.20 
1.63 
P. capillare abundance 
Year planted * Watering 
6.55 
0.16 
0.95 
0.10 
Seed mix * Year planted 
Seed mix * Watering 
0.57 
0.75 
Richness of non-sown species in the second year 
Predictor 
(Intercept) 
0.40 
Year planted 
0.66 
Watering treatment 
Seed mix richness 
0.69 
Echinacea sp. abundance  0.0077  0.93 
0.27 
P. capillare abundance 
Year planted * Watering 
0.94 
0.69 
Seed mix * Year planted 
Seed mix * Watering 
0.81 
Richness of sown species in the third year 
χ2 
Predictor 
203.84  <0.0001 
(Intercept) 
0.14 
2.22 
Year planted 
0.53 
1.27 
Watering treatment 
0.38 
Seed mix richness 
0.77 
0.56 
Echinacea sp. abundance  0.34 
0.89 
0.019 
P. capillare abundance 
0.69 
0.75 
Year planted * Watering 
0.0098  0.92 
Seed mix * Year planted 
Seed mix * Watering 
0.64 
0.73 
Richness of sown species in 2017 
χ2 
Predictor 
143.81  <0.0001 
(Intercept) 
0.37 
Year planted 
1.92 
Watering treatment 
Seed mix richness 
0.011 
Echinacea sp. abundance  0.18 
0.23 
P. capillare abundance 
Year planted * Watering 
3.70 
0.67 
Seed mix * Year planted 
Seed mix * Watering 
0.10 

 
0.0070 
0.017 
0.00 
0.0030 
0.0020 
0.049 
0.0070 
0.0030 

 
Model R2=0.19 
 
 
 
 
 
 
 
 
 
 
Model R2=0.19 
 
 
 
 
 
 
 
 
 
 
Model R2=0.19 
 
 
 
 
 
 
 
 
 
 
Model R2=0.18 
 
 
 
 
 
 
 
 
 

p-value  R2 

0.83 
0.38 
0.92 
0.67 
0.63 
0.45 
0.71 
0.95 

 

86 

LITERATURE CITED

 

87 

LITERATURE CITED 

 

Aarssen LW, Laird RA, Pither J (2003) Is the productivity of regetation plots higher or lower when there 
are more species? Variable predictions from interaction of the ‘sampling effect’ and ‘competitive 
dominance effect’ on the habitat templet. Oikos 102:427–432 

Alford RA, Wilbur HM (1985) Priority effects in experimental pond communities: Competition between 

Bufo and Rana. Ecology 66:1097–1105 

Anderson MJ (2001) A new method for non-parametric multivariate analysis of variance. Austral Ecology 

26:32–46 

Averett JM, Klips RA, Nave LE, Frey SD, Curtis PS (2004) Effects of soil carbon amendment on nitrogen 
availability and plant growth in an experimental tallgrass prairie restoration. Restoration Ecology 
12:568–574 

Baer SG, Engle DM, Knops JMH, Langeland K a, Maxwell BD, Menalled FD, Symstad AJ (2009) 

Vulnerability of rehabilitated agricultural production systems to invasion by nontarget plant 
species. Environmental management 43:189–96 

Bakker JD, Wilson SD, Christian JM, Li X, Ambrose LG, Waddington J (2003) Contingency of grassland 

restoration on year, site, and competition from introduced grasses. Ecological Applications 13:137–
153 

Bakker JP, Poschlod P, Strykstra RJ, Bekker RM, Thompson K (1996) Seed banks and seed dispersal: 

Important topics in restoration ecology. Acta Botanica Neerlandica 45:461–490 

Baskin CC, Baskin JM (1988) Germination ecophysiology of herbaceous plant species in a temperate 

region. American Journal of Botany 75:286–305 

Bates D, Maechler M, Bolker B, Walker S (2015) Fitting linear mixed-effects models using lme4. Journal 

of Statistical Software 67:1–48 

Belyea LR (2004) Beyond ecological filters: Feedback networks in the assembly and restoration of 
community structure. In: Assembly rules and restoration ecology: Bridging the gap between 
science and practice. Temperton, VM, Hobbs, RJ, Nuttle, T, & Halle, S, editors. Island Press. 

Belyea LR, Lancaster J (1999) Assembly within a contingent ecology. Oikos 86:402–416 

Blumenthal DM, Jordan NR, Russelle MP (2003) Soil carbon addition controls weeds and facilitates 

prairie restoration. Ecological Applications 13:605–615 

Bowles ML, Jones MD (2013) Repeated burning of eastern tallgrass prairie increases richness and 

diversity, stabilizing late successional vegetation. Ecological Applications 23:464–478 

Cardinale BJ, Duffy JE, Gonzalez A, Hooper DU, Perrings C, Venail P, Narwani A, Mace GM, Tilman D, 

Wardle DA, Kinzig AP, Daily GC, Loreau M, Grace JB, Larigauderie A, Srivastava DS, Naeem S (2012) 
Biodiversity loss and its impact on humanity. Nature 486:59–67 

 

88 

Carter DL, Blair JM (2012) High richness and dense seeding enhance grassland restoration establishment 

but have little effect on drought response. Ecological Applications 22:1308–19 

Chase JM (2003) Community assembly: When should history matter? Oecologia 136:489–98 

Comer PJ, Albert DA, Wells H., Hart BL, Raab JB, Price DL, Kashian DM, Corner RA, Schuen DW, Leibfreid 
TR, Austin MB, DeLain CJ, Prange-Gregory L, Scrimger LJ, Korroch KM, Spitzley JG (199 ) Michigan’s 
presettlement vegetation, as interpreted from the General Land Office surveys 1816-1856. Lansing, 
MI 

Connell JH, Slatyer RO (1977) Mechanisms of succession in natural communities and their role in 

community stability and organization. The American Naturalist 111:1119–1144 

Cramer VA, Hobbs RJ, Standish RJ (2008) What’s new about old fields? Land abandonment and 

ecosystem assembly. Trends in Ecology and Evolution 23:104–12 

Deering RH, Young TP (2006) Germination speeds of exotic annual and native perennial grasses in 

California and the potential benefits of seed priming for grassland restoration. Grasslands XVI:14–
16 

Dickson TL, Hopwood JL, Wilsey BJ (2012) Do priority effects benefit invasive plants more than native 

plants? An experiment with six grassland species. Biological Invasions 14:2617–2624 

Dobson AP, Bradshaw AD, Baker AJM (1997) Hopes for the future: Restoration ecology and conservation 

biology. Science 277:515–522 

Doll JE, Haubensak K a., Bouressa EL, Jackson RD (2011) Testing disturbance, seeding time, and soil 
amendments for establishing native warm-season grasses in non-native cool-season pasture. 
Restoration Ecology 19:1–8 

Drake JA (1990) The mechanics of community assembly and succession. Journal of Theoretical Biology 

147:213–233 

Foley JA, DeFries R, Asner GP, Barford C, Bonan G, Carpenter SR, Chapin FS, Coe MT, Daily GC, Gibbs HK, 
Helkowski JH, Holloway T, Howard EA, Kucharik CJ, Monfreda C, Patz JA, Prentice IC, Ramankutty N, 
Snyder PK (2005) Global consequences of land use. Science 309:570–574 

Forrest JRK, James DT (2011) An examination of synchrony between insect emergence and flowering in 

Rocky Mountain meadows. Ecological Monographs 81:469–491 

Fox J, Weisberg S (2011) An {R} companion to applied regression. Second Ed. Sage, Thousand Oaks, CA 

Fukami T (2015) Historical contingency in community assembly: Integrating niches, species pools, and 

priority effects. Annual Review of Ecology, Evolution, and Systematics 46:1–23 

Fukami T, Mordecai EA, Ostling A (2016) A framework for priority effects. Journal of Vegetation Science 

27:655–657 

Getz LL, Hofmann JE (1999) Diversity and stability of small mammals in tallgrass prairie habitat in central 

Illinois, USA. Oikos 85:356–363 

 

89 

Grant K, Kreyling J, Heilmeier H, Beierkuhnlein C, Jentsch A (2014) Extreme weather events and plant–

plant interactions: shifts between competition and facilitation among grassland species in the face 
of drought and heavy rainfall. Ecological Research 29:991–1001 

Grman E, Bassett T, Brudvig L (2013) Confronting contingency in restoration: Management and site 

history determine outcomes of assembling prairies, but site characteristics and landscape context 
have little effect. Journal of Applied Ecology 50:1234–1243 

Grman E, Bassett T, Brudvig LA (2014) A prairie plant community data set for addressing questions in 

community assembly and restoration. Ecology 95:2363 

Grman E, Bassett T, Zirbel CR, Brudvig LA (2015) Dispersal and establishment filters influence the 

assembly of restored prairie plant communities. Restoration Ecology 23:892–899 

Grman E, Brudvig LA (2014) Beta diversity among prairie restorations increases with species pool size, 

but not through enhanced species sorting. Journal of Ecology 102:1017–1024 

Group PC (2018) PRISM Climate Data. 

Groves AM, Brudvig LA (2018) Inter-annual variation in precipitation and other planting conditions 

impacts seedling establishment in sown plant communities. Restoration Ecology 

Harper JL (1977) Population Biology of Plants. Academic Press, London 

Harper JL (1987) The heuristic value of ecological restoration. In: Restoration Ecology: A synthetic 

approach to ecological research. Jordan, WRI, Gilpin, ME, & Aber, JD, editors. University Press, 
Cambridge, U.K. pp. 35–46. 

Hassell RL, Dufault R, Phillips T (2004) Relationship among seed size, source and temperature on 

germination of Echinacea angustifolia, pallida and purpurea. Acta Horticulturae 629:239–243 

Hassell RL, Dufault RJ, Phillips T, Hale TA (2004) Influence of temperature gradients on Pale and Purple 

Coneflower, Feverfew, and Valerian germination. HortTechnology 14:368–371 

Heisler-White JL, Blair JM, Kelly EF, Harmoney K, Knapp AK (2009) Contingent productivity responses to 

more extreme rainfall regimes across a grassland biome. Global Change Biology 15:2894–2904 

Hobbs RJ, Harris JA (2001) Restoration Ecology: Repairing the Earth’s Ecosystems in the New Millennium. 

Restoration Ecology 9:239–246 

Hooper DU (2005) Effects of Biodiversity on Ecosystem Functioning A Consensus of Current Knowledge. 

Ecological Monographs 75:3–35 

Howe HF (2008) Reversal of fortune: Plant suppression and recovery after vole herbivory. Oecologia 

157:279–86 

Howe HF, Brown JS (1999) Effects of Birds and Rodents on Synthetic Tallgrass Communities. Ecology 

80:1776–1781 

 

 

90 

Hulme PE, Kollmann J (2005) Seed predator guilds, spatial variation in post-dispersal seed predation and 

potential effects on plant demography: A temperate perspective. In: Seed Fate. Predation, 
dispersal and seedling establishment. pp. 9–30. 

Johnson DH (2002) The importance of replication in wildlife research. The Journal of Wildlife 

Management 66:919–932 

Jones HP, Jones PC, Barbier EB, Blackburn RC, Benayas JMR, Holl KD, McCrackin M, Meli P, Montoya D, 
Moreno-Mateos D (2018) Restoration and repair of Earth’s damaged ecosystems. Proceedings of 
the Royal Society B 285 

Keddy PA (1992) Assembly and response rules: Two goals for predictive community ecology. Journal of 

Vegetation Science 3:157–164 

Larson JE, Sheley RL, Hardegree SP, Doescher PS, James JJ (2015) Seed and seedling traits affecting 
critical life stage transitions and recruitment outcomes in dryland grasses. Journal of Applied 
Ecology 52:199–209 

Lenth R V (2016) Least-squares means: The {R} package {lsmeans}. Journal of Statistical Software 69:1–

33 

Lockwood JL, Samuels CL (2004) Assembly models and the practice of restoration. In: Assembly Rules 
and Restoration Ecology. Temperton, VM, Hobbs, RJ, Nuttle, T, & Halle, S, editors. Island Press, 
Washington, USA pp. 55–70. 

MacDougall AS, Wilson SD (2007) Herbivory limits recruitment in an old-field seed edition experiment. 

Ecology 88:1105–1111 

MacDougall AS, Wilson SD, Bakker JD (2008) Climatic variability alters the outcome of long-term 

community assembly. Journal of Ecology 96:346–354 

Manning GC, Baer SG (2018) Interannual variability in climate effects on community assembly and 

ecosystem functioning in restored prairie. Ecosphere 9:e02327 

Martin LM, Wilsey BJ (2012) Assembly history alters alpha and beta diversity, exotic-native proportions 

and functioning of restored prairie plant communities Cadotte, M, editor. Journal of Applied 
Ecology 

Mason CH, Perreault Jr. WD (1991) Collinearity, power, and interpretation of multiple regression 

analysis. Journal of Marketing Research 28:268–280 

Mousseau T, Fox C (1998) The adaptive significance of maternal effects. Trends in Ecology & Evolution 

13:403–407 

Myers JA, Harms KE (2009) Seed arrival, ecological filters, and plant species richness: A meta-analysis. 

Ecology letters 12:1250–1260 

Nemec KT, Allen CR, Helzer CJ, Wedin DA (2013) Influence of Richness and Seeding Density on Invasion 

Resistance in Experimental Tallgrass Prairie Restorations. Ecological Restoration 31:168–185 

 

91 

Notman EM, Villegas AC (2005) Patterns of seed predation by vertebrate versus invertebrate seed 

predators among different plant species, seasons and spatial distributions. In: Seed Fate: 
Predation, Dispersal and Seedling Establishment. pp. 55–75. 

Oksanen J, Blanchet FG, Friendly M, Kindt R, Legendre P, McGlinn D, Minchin PR, O’Hara RB, Simpson GL, 

Solymos P, Stevens MHH, Szoecs E, Wagner H (2017) Vegan: Community ecology package. R 
package version 2.4-2 https://CRAN.R-project.org/package=vegan 

Packard S, Mutel CF, eds. (1997) The tallgrass restoration handbook: For prairies, savannas, and 

woodlands. Island Press, Washington, D.C. 

Palmer MA, Ambrose RF, Poff NL (1997) Ecological Theory and Community Restoration Ecology. 

Restoration Ecology 5:291–300 

Piper JK, Schmidt ES, Janzen AJ (2007) Effects of species richness on resident and target species 

components in a prairie restoration. Restoration Ecology 15:189–198 

Roach DA, Wulff RD (1987) Maternal effects in plants. Annual Review of Ecology and Systematics 

18:209–235 

Rowe HI (2010) Tricks of the trade: Techniques and opinions from 38 experts in tallgrass prairie 

restoration. Restoration Ecology 18:253–262 

Ruiz-Jaen MC, Mitchell Aide T (2005) Restoration success: How is it being measured? Restoration 

Ecology 13:569–577 

Samson F, Knopf F (1994) Prairie conservation in North America. BioScience 44:418–421 

Schindler DE, Armstrong JB, Reed TE (2015) The portfolio concept in ecology and evolution. Frontiers in 

Ecology and the Environment 13:257–263 

Seabloom EW, Borer ET, Boucher VL, Burton RS, Cottingham KL, Goldwasser L, Gram WK, Kendall BE, 

Micheli F (2003) Competition, seed limitation, disturbance, and reestablishment of California 
native annual forbs. Ecological Applications 13:575–592 

Stuble KL, Fick SE, Young TP (2017) Every restoration is unique: Testing year effects and site effects as 

drivers of initial restoration trajectories. Journal of Applied Ecology 54:1051–1057 

Suding KN, Collins SL, Gough L, Clark C, Cleland EE, Gross KL, Milchunas DG, Pennings S (2005) 

Functional- and abundance-based mechanisms explain diversity loss due to N fertilization. 
Proceedings of the National Academy of Sciences of the United States of America 102:4387–92 

Suding KN (2011) Toward an era of restoration in ecology: Successes, failures, and opportunities ahead. 

Annual Review of Ecology, Evolution, and Systematics 42:465–487 

Team RC (2016) R: A language and environment for statistical computing. R Foundation for Statistical 

Computing, Vienna, Austria 

Thompson K, Grime JP (1983) A Comparative Study of Germination Responses to Diurnally-Fluctuating 

Temperatures. Journal of Applied Ecology 20:141–156 

 

92 

Tilman D, Reich PB, Knops JMH (2006) Biodiversity and ecosystem stability in a decade-long grassland 

experiment. Nature 441:629–32 

Trudgill DL, Honek A, Li D, Van Straalen NM (2005) Thermal time - Concepts and utility. Annals of Applied 

Biology 146:1–14 

Turnbull LA, Rahm S, Baudois O, Eichenberger-Glinz S, Wacker L, Schmid B (2005) Experimental invasion 

by legumes reveals non-random assembly rules in grassland communities. Journal of Ecology 
93:1062–1070 

Vaughn KJ, Young TP (2010) Contingent conclusions: Year of initiation influences ecological field 

experiments, but temporal replication is rare. Restoration Ecology 18:59–64 

Vaughn KJ, Young TP (2015) Short-term priority over exotic annuals increases the initial density and 

longer-term cover of native perennial grasses. Ecological Applications 25:791–799 

Wainwright CE, Staples TL, Charles LS, Flanagan TC, Lai HR, Loy X, Reynolds VA, Mayfield MM (2018) 
Links between community ecology theory and ecological restoration are on the rise. Journal of 
Applied Ecology 55:570–581 

Wainwright CE, Cleland EE (2013) Exotic species display greater germination plasticity and higher 

germination rates than native species across multiple cues. Biological Invasions 15:2253–2264 

Wang Y, Frei M (2011) Stressed food - The impact of abiotic environmental stresses on crop quality. 

Agriculture, Ecosystems and Environment 141:271–286 

Weiher E, Freund D, Bunton T, Stefanski A, Lee T, Bentivenga S (2011) Advances, challenges and a 

developing synthesis of ecological community assembly theory. Philosophical Transactions of the 
Royal Society B: Biological Sciences 366:2403–13 

Weiher E, Clarke GDP, Keddy PA (1998) Community assembly rules, morphological dispersion, and the 

coexistence of plant species. Oikos 81:309–322 

Werner CM, Vaughn KJ, Stuble KL, Wolf K, Young TP (2016) Persistent asymmetrical priority effects in a 

California grassland restoration experiment. Ecological Applications 26:1624–1632 

Wilsey BJ, Barber K, Martin LM (2014) Exotic grassland species have stronger priority effects than natives 

regardless of whether they are cultivated or wild genotypes. The New phytologist 

Wilsey BJ, Daneshgar PP, Polley HW (2011) Biodiversity, phenology and temporal niche differences 

between native- and novel exotic-dominated grasslands. Perspectives in Plant Ecology, Evolution 
and Systematics 13:265–276 

Wilson SD (2015) Managing contingency in semiarid grassland restoration through repeated planting. 

Restoration Ecology 23:385–392 

Wuebbles DJ, Hayhoe K (2004) Climate change projections for the United States Midwest. Mitigation 

and Adaptation Strategies for Global Change 9:335–363 

 

 

93 

Young TP, Chase JM, Huddleston RT (2001) Community succession and assembly. Ecological Restoration 

19:5–18 

Young TP, Stuble KL, Balachowski JA, Werner CM (2016) Using priority effects to manipulate competitive 

relationships in restoration. Restoration Ecology 25:S114–S123 

 

 

 

94 

HISTORICAL CONTINGENCY IN ECOLOGICAL RESTORATION: EVIDENCE FROM 83 PRAIRIES 

CHAPTER FOUR 

 

ABSTRACT 

 

Reliable ecological restoration outcomes are of paramount importance to reverse or reduce 

losses in native biodiversity caused by extensive land use by humans. Yet, restoration outcomes are 

notoriously variable, reducing public confidence in our ability to restore native biodiversity and its 

related functions. One commonly assumed source of this variation is weather conditions during the 

planting year, though empirical evidence for these effects are scarce. Finding evidence for “year effects” 

in prairie restorations would provide support for historically contingent assembly models, in which early 

community dynamics impact longer-term community trajectories. It might also suggest ways that 

subsequent management might mediate these effects and promote higher establishment by native 

species. Despite experimental evidence for short-term year effects in restoration, their long-term 

persistence and importance compared to other factors is unclear. The research we report here based on 

surveys of 83 sites distributed across Illinois, Indiana, and Michigan sheds light on the implications of 

year effects across a broad geographic gradient. It also allows us to evaluate the persistence of these 

effects because these restorations had been established for 2-19 years. In 2016, we surveyed plant 

community composition in the prairies, and considered the interplay between management histories, 

current community compositions, and the temperature and precipitation in each planting year to ask 

whether there is a measurable signature of planting-year conditions on present-day plant communities, 

and what the importance of these effects are compared to other known drivers of restored communities 

(e.g. seed mix richness, fire frequency). We found that weather conditions in the first growing season 

explained more variation in the 2016 richness of non-sown species than did site age or fire frequency 

 

95 

across the 83 sites. Sites associated with more precipitation in the planting year had higher richness of 

non-sown species, community compositions more dissimilar than their seed mixes, and marginally 

higher non-sown cover. Warmer planting years were associated with increased cover of non-sown 

species in young sites and decreased cover in older sites. Conversely, prairies planted in cooler years had 

less non-sown cover in younger sites, and higher non-sown cover in older sites. By using this extensive 

dataset, we provide evidence that weather conditions during planting years can influence communities 

that assemble during ecological restoration. 

 

INTRODUCTION 

 

Human impacts on natural ecosystems are a major driver of biodiversity loss worldwide (Foley 

et al. 2005; Cardinale et al. 2012), and ecological restoration is an important tool for combating this loss 

(Dobson et al. 1997; Hobbs & Harris 2001; Ruiz-Jaen & Mitchell Aide 2005; Jones et al. 2018). However, 

outcomes of restoration efforts can vary greatly, with a vast array of possible reasons for this variation 

including edaphic conditions, invasive species pressure, seed mixes sown, landscape context, land-use 

history, management decisions, and others (Blumenthal et al. 2003; Averett et al. 2004; Baer et al. 2009; 

Rowe 2010; Suding 2011; Grman et al. 2015). These drivers represent the suite of deterministic 

environmental filters that govern community assembly by dictating which species are able to establish 

and persist at a site given the current conditions (Weiher et al. 2011). However, the establishment 

history of a community can also affect the outcome of assembly (Fukami 2015). For instance, early 

arrivers to a community can influence later arrivers through processes such as priority effects (Alford & 

Wilbur 1985; Fukami et al. 2016; Young et al. 2016) or facilitation (Connell & Slatyer 1977; Grant et al. 

2014), leading to disparate communities even when environmental conditions are otherwise the same. 

In restored systems, the composition of these early arrivers is shaped by the interactions among 

 

96 

restoration methods and site conditions. Much research has assessed the influence of the former, in 

hopes to quantify best practices for restoration (e.g., Rowe 2010). However, less is known about the 

effects of different site conditions. In addition, conditions during the planting year, and particularly 

weather conditions after planting, are likely to have strong effects on restoration outcomes. Though a 

few token studies and experiments provide evidence that year effects can exist, large-scale assessment 

across numerous restoration projects is lacking. 

These planting year effects (hereafter, year effects) on restored communities make logical 

sense, but evidence for year effects in the literature is rare, since most restoration projects and 

ecological studies are not replicated across planting years (Vaughn & Young 2010). In a recent study, we 

found that different planting years and experimental precipitation regimes significantly altered the 

seedling emergence patterns of different species, leading to different plant communities in the first 

growing season of a restored grassland community (Groves & Brudvig 2018). Important work in Kansas 

and California grasslands has also shown the importance of both planting year (Manning & Baer 2018) 

and planting year and site conditions (Stuble et al. 2017) on restoration success. This evidence has 

demonstrated the potential for influence of planting conditions on a community; however, it is unclear 

how long this influence will last. Evidence that demonstrates year effects over longer time scales is even 

rarer, with one notable exception finding stark effects: in a grassland restoration experiment in the 

North American Great Plains, herbicide application on an invasive grass followed by sowing of a native 

grass was highly successful even a decade later when planted in average or above-average rainfall years, 

but reverted to dominance by the invasive grass when planted in an exceptionally dry year (Bakker et al. 

2003; MacDougall et al. 2008).  

Additionally, although these studies support the conclusion that year effects can exist, their 

relative importance is difficult to study due to the myriad other factors that shape a plant community. 

Current emphasis on grasslands in the year effect literature likely stems from the accentuated 

 

97 

importance of planting year conditions in highly invaded, highly degraded sites that are restored by 

intensive vegetation removal efforts followed by sowing seeds of target species, usually within a single 

season. There is already evidence that restoration outcomes in these systems can be highly influenced 

by deterministic factors such as soil characteristics, site land use history, management (e.g. prescribed 

fire frequency), and the richness and density of the seed mix sown (Suding et al. 2005; Piper et al. 2007; 

Doll et al. 2011; Grman et al. 2013). It therefore remains unknown how important year effects are in 

context of these other factors known to influence assembly in restored systems. Moreover, there may 

be an interaction between weather conditions during planting and these other factors; for instance, 

with site age because of the possibility of decreasing importance of planting year conditions over time, 

or with soil water holding capacity, because of different effects of rainfall depending on the soil 

moisture retention at the site.   

Finally, it remains unclear exactly what aspects of planting year weather conditions might be 

most important for subsequent restoration trajectories. Past studies have explored correlations with a 

limited suite of environmental variables: annual mean temperature, total rainy days (Stuble et al. 2017); 

June temperature, June-July precipitation (Bakker et al. 2003); and deviations from normal growing 

season (March-October) precipitation and temperature (MacDougall et al. 2008). Stuble et al. (2017) 

and MacDougall et al. (2008) notably found no long-term effects of temperature metrics, only 

precipitation metrics, on plant communities sown in different years. However, the experimental work by 

Groves & Brudvig (2018) suggested that precipitation, though important, is not the only driver of sown 

community differences after different planting years. Additional factors are known to be important 

determinants of seeds breaking dormancy, germinating, and establishing. For instance, warmer 

temperatures, especially in early mornings or springtime, can advance the phenology of some plant 

species (Forrest & James 2011), which can favor certain species over others (Wainwright & Cleland 

2013). In agronomic systems, temperature-based units such as growing degree days are used to predict 

 

98 

development of plants and other species (Trudgill et al. 2005). Better understanding of which first year 

weather conditions influence restoration plantings would clarify drivers of variation in restoration 

outcomes and benefit restoration planning by enabling targeted post-planting restoration actions (i.e., 

to sites planted in years when such actions would be most beneficial). 

The tallgrass prairie is nearly extirpated in North America, with only 0.1% of its original extent 

remaining (Samson & Knopf 1994). After decades of degradation from intensive agriculture and other 

human activities, prairie species no longer occur in many landscapes, nor do their seeds persist in the 

seed bank (Bakker et al. 1996). Therefore, restoration often takes place through seed sowing onto 

abandoned agricultural lands (Cramer et al. 2008), which could make the establishment of these 

restored communities particularly susceptible to abiotic and biotic conditions present during their first 

growing season.  

In this study, we surveyed 83 tallgrass prairie restorations to investigate (I) whether planting 

year conditions related to temperature and precipitation influence the resulting plant community 2-19 

years after planting, across a variety of metrics (e.g. species richness and cover of sown and non-sown 

species), (II) what the relative influence of planting year conditions are compared to other known drivers 

of variation in restored prairie plant communities (i.e. fire frequency, edaphic conditions, site age, and 

the richness of the seed mix sown), and (III) whether the influence of planting conditions changes with 

soil productivity or time since planting.  We show that weather conditions have lasting effects on prairie 

plant communities, particularly through effects on the non-sown species, and that these effects can be 

of greater magnitude than those of other important drivers of restoration outcomes in this system. 

 

 

 

 

 

99 

METHODS 

 

Study sites 

We surveyed 83 restored prairie sites in Illinois, Indiana, and Michigan July-September 2016 

(Figure 4.1). Eighteen of the restored prairie plantings were managed by the McHenry County 

Conservation District (MCCD) in Illinois (Figure 4.2a). Another 32 sites were managed by The Nature 

Conservancy’s Kankakee Sands Preserve in Indiana (Figure 4.2b). The final 33 sites were scattered across 

southwest Michigan, installed by the restoration firm Native Connections (Three Rivers, MI) onto 

properties under various ownership and management (Figure 4.2c). Prairies ranged in size from 0.2-118 

hectares and were established with different seeding techniques (broadcast or drill-seeded), and in 

different planting seasons (dormant or growing season). Sites also differed in previous land use (soy, 

corn, winter wheat, hay; some converted directly from production, some left fallow), landscape context 

(forest, agriculture, or other restored prairie sites), soil characteristics (soil water holding capacity 

ranged from 0.33-0.80, mean 0.50 percent of total soil weight), and planting year (1998-2014). The 

majority of sites were managed with prescribed fire, with 77% of sites burned within the past 4 years; 

20% of sites had not been burned. Seed mixes sown ranged from a richness of 8 to 239 species (mean 

57.4 species) and a sowing density of 1.6 to 54.1 kg/ha (mean 9.7 kg/ha). 

 

Field data collection 

We visited the 83 restored prairies between 18 July-16 September 2016. In each site, we 

established a 50 m transect in a random orientation. For Michigan and Illinois sites, transects were 

centered on the approximate centroid of the site; for Indiana sites, which were much larger, transects 

were placed inside the boundary of visible edge effects. At 5- or 10- m intervals (for Michigan and 

Illinois/Indiana sites, respectively), we placed a 1 m x 1 m quadrat frame on the ground and visually 

 

100 

quantified the percent cover of each species present. We surveyed 5 plots per site in Indiana and Illinois, 

and 10 plots in Michigan (for compatibility with another project outside the scope of this manuscript). In 

addition, we systematically walked a 20 m x 50 m area centered on the transect and recorded additional 

species observed. We collected two soil cores at 5 m intervals along each transect using an Oakfield soil 

probe (20 total per site, 2 cm × 20 cm soil cores). Soil cores were homogenized at each site. We 

determined soil water holding capacity by saturating field samples and then drying for 72 hours at 105°C 

and calculating the proportional difference in saturated wet weight and oven‐dried weight. Air-dried 

samples were also analyzed for a variety of soil properties including pH, % organic matter, % 

clay/silt/sand, and soil nutrients (S, P, Ca, Mg, K, Na, B, Fe, Mn, Cu, Zn, and Al) by the Brookside 

Laboratories (New Knoxville, OH, USA). To reduce the potential number of predictor variables, we 

conducted a principal components analysis on the soils data using prcomp in R with each variable scaled 

due to large variation the concentrations of soil nutrients. For later analyses, we used the first principal 

component, which accounted for 40.5% of the variation in the soils data and was associated with higher 

percent organic matter, higher water holding capacity, higher clay content, higher silt content, higher 

nutrients (except Zn, Fe, and P), and less sand (Table 4.3). Hereafter this first PC axis for soils is referred 

to as “soil productivity.”  

 

Management histories and historical weather data 

 

We worked with the land managers at each site to compile information about the restored 

prairies, including the date of planting, prescribed fire history, and the relative abundance (by weight) of 

each species sown in the initial planting. We used 30-arcsec (approximately 800 m) spatially gridded 

PRISM Climate Data to compile interpolated information on the daily precipitation accumulation and 

minimum and maximum temperature in the first growing season for each site based on its latitude and 

longitude (PRISM Climate Group 2018). We used these data to calculate cumulative growing degree days 

 

101 

(base 10°C), cumulative precipitation, and averages related to temperature and precipitation at various 

stages throughout the growing season. 

 

Statistical analysis 

To characterize weather conditions at the time of planting we focused on eleven variables that 

we hypothesized to influence the germination and establishment of plant communities during the first 

growing season of prairie restoration: spring temperatures (degree day accumulation March 1 – June 1), 

spring rainfall (precipitation accumulation March 1 – June 1), summer temperatures (degree day 

accumulation June 1 – September 1), summer rainfall (precipitation accumulation June 1 – September 

1), full growing season temperatures (degree day accumulation March 1 – September 1), full growing 

season rainfall (precipitation accumulation March 1 – September 1), the hottest month (maximum 

degree days accumulated in a 30-day period), the longest drought (maximum days without a 

precipitation event), the driest month (minimum precipitation in a 30-day period, in mm rainfall), 

average low temperature (March 1 – September 1, °C), and the average monthly rainfall (March 1 – 

September 1, mm). We focused on this set of variables owing to the reported roles of temperature and 

precipitation for plant establishment dynamics (e.g., Thompson & Grime 1983; Groves & Brudvig 2018) 

and for year effects broadly (Bakker et al. 2003; MacDougall et al. 2008; Stuble et al. 2017). 

Because many of these variables were correlated (Figure 4.3), we conducted a Principal 

Components Analysis to determine composite metrics with which to test our hypotheses (Table 4.1). We 

focused on the first three axes (out of eleven total) that together accounted for 74.2% of the variation in 

weather data. The first axis, PC1, accounted for 37.6% of the variation in the weather dataset and 

positively correlated with all metrics that related to higher temperatures. The second axis, PC2 (22.0% of 

the variation), was positively correlated with all metrics that related to more rainfall. The third axis, PC3, 

accounted for 14.7% of the variation in the weather dataset and was negatively associated with summer 

 

102 

precipitation, spring temperatures, and the maximum number of days without a precipitation event, and 

positively associated with minimum precipitation in a 30-day period and the mean number of rainy days 

in a 30-day period. We interpret this axis as mild droughts: low overall precipitation that is consistently 

distributed throughout the growing season, with the worst dry spell relatively wet and short. Plots of the 

three weather PC axes against the other environmental variables (soils, site age, seed mix richness; also, 

state) are provided in Figure 4.7. The distribution of each variable across sites is provided in Figure 4.8. 

We used a series of linear models to test for an effect of planting year weather conditions (PC1, 

PC2, and PC3) on the present-day community. In addition to the three weather PCs, models included the 

richness of the seed mix, the seeding density (kg/ha), the age of the site in 2016, the frequency of fire 

(characterized as the years since the last prescribed burn), and soil productivity. None of these variables 

were colinear, with each correlation coefficient (r) < 0.4 (Mason & Perreault Jr. 1991). Each model also 

included interactions between PC1, PC2, and PC3 and site age and soil productivity. 

Finally, we considered response variables representing typical prairie restoration goals: namely, 

improving sown species richness and cover and minimizing non-sown species richness and cover. We 

therefore ran six models, predicting the richness of sown, non-sown, and total species at the site level 

(20 m x 50 m survey area); the cover of sown and non-sown species at the mean 1 m2 plot level; and the 

Bray-Curtis dissimilarity between the composition of the seed mix sown and the vegetation at the site in 

2016 (Table 4.2). We compared R2 values of predictors in the models to compare their relative influence 

on planting outcomes. 

 

RESULTS 

 

Total observed species richness in the 83 prairie sites ranged from 23-73 (mean 50); at the plot 

(1 m2) scale, species richness ranged from 14-48 (mean 28). Sown species richness ranged from 3-35 

 

103 

(mean 18) species at the field scale and from 0-28 (mean 11) at the 1 m2 scale. The average percent 

cover of sown species per plot ranged from 0-106% (mean 51%), while mean non-sown species cover 

ranged from 2-102% (mean 33%); overlapping individuals allow for sums greater than 100% per plot. 

The average dissimilarity between the composition of the seed mix and the standing community was 

0.69 (range 0.34-1.00); between 0 and 88% of species sown were present at each site.  

 

Effects of planting year on community metrics 

We found a significant effect of planting year conditions on plant community metrics (Figure 

4.4). The strongest responses were related to the non-sown species: we found that increased 

precipitation during the first growing season (PC2) was significantly correlated with increased richness of 

non-sown species (β=2.22, F=7.70, p=0.0072, Figure 4.4A) and marginally correlated with increased 

cover of non-sown species (β=0.14, F=3.77, p=0.0  , Figure 4.4B). Increased planting year precipitation 

(PC2) was also marginally correlated with decreased richness of sown species (β=-1.30, F=3.40, p=0.070, 

Figure 4.4C) and was correlated with greater Bray-Curtis dissimilarity between the composition of the 

seed mix and the 2016 plant community, meaning the communities in prairies planted in wetter years 

did not resemble their seed mixes as much as prairies planted in drier years (β=0.02 , F= .  , p=0.012, 

Figure 4.4D).  

We also found evidence for a weather by age interaction (Figure 4.5), in which warmer planting 

years (PC1) were associated with higher cover of non-sown species in young sites and less cover in older 

sites, while prairies planted in cooler years (PC1) had less non-sown cover in younger sites and higher 

non-sown cover in older sites. 

We found no effects of PC3 (mild droughts in planting year) on any 2016 community metrics 

(Table 4.4). We found no significant relationships between weather conditions in the first growing 

season and sown species cover, sown species richness, or total species richness (Table 4.4). 

 

104 

Relative influence of weather compared to known assembly drivers 

The variation in non-sown richness that was explained by planting-year precipitation (PC2, 

R2=0.10, Figure 4.5) was greater than the variation explained by fire frequency (R2=0.0011), site age 

(R2=0.00016), soil productivity (R2=0.0016), seed mix richness (R2=0.036), or seeding density (R2=0.052).  

For Bray-Curtis dissimilarity between the seed mix and 2016 community composition, the effect 

of planting year precipitation (PC2), though significant, was weaker than other predictors. This 

dissimilarity was better explained by increased seed mix richness (R2=0.20, β=0.0  , F=1 .78, 

p=0.00012) and older site age (R2=0.13, β=0.0 0, F=10.09, p=0.0023) than increased planting year 

precipitation (PC2, R2=0.092, β=0.02 , F= .  , p=0.012).  

Non-sown species cover was best predicted by, listed in decreasing importance: soil productivity 

(R2=0.17, β=0.19, F=13.45, p=0.00049), the interaction between temperature in the first growing season 

(PC1) and site age (R2=0.13, β=0.18, F=9.73, p=0.0027), site age (R2=0.084, β=0.27, F= .02, p=0.017), 

seed mix richness (R2=0.066, β=0.24, F=4. 4, p=0.03 ), and precipitation in the first growing season 

(PC2, R2=0.054, β=0.14, F=3.77, p=0.0  ).  

The cover of sown species was not well predicted by our models (total model R2=0.19), and was 

best predicted by soil productivity (β=-3.27, F=5.13, p=0.027, R2=0.072). Similarly, sown species richness 

(total model R2=0.22) was best predicted by the richness of the seed mix (β=3.38, F=8.7 , p=0.0043, 

R2=0.12; total model R2=0.22). We found no significant predictors for overall species richness. 

 

DISCUSSION 

 

Our results show clear evidence for year effects on restored prairie community composition 

across 83 sites through their effects on the non-sown species. This result is important because the sites 

surveyed span an extensive gradient of geography (3 states), age (2 to 19 years old), and soil conditions. 

 

105 

These year effects were predominantly driven by wetter planting years resulting in higher non-sown 

species richness and marginally higher non-sown cover. The planting-year effects we detected on non-

sown richness were greater than the effects of other known drivers of community establishment in 

restored prairies (i.e., fire frequency, sown species richness, and soil characteristics).  

Sown species richness was marginally decreased following wet planting years (PC2). It is unclear 

whether this is a result of suppression by the increase in non-sown species that also coincided with wet 

planting years, or if drier conditions are in fact better for the sown species. We suspect the former, as 

competition seems a more likely suppressant of sown species establishment than increased water 

availability. For instance, Groves & Brudvig (2018) found that increased planting-year rainfall increased 

both sown and non-sown species emergence 3-fold— evidence that sown species do not perform better 

under drought stress, and that non-sown species increase dramatically under wet conditions.  

The second axis of our weather PCA corresponded with increased precipitation accumulation 

during spring (March 1 – June 1) in addition to increases in the other precipitation-related metrics (Table 

4.1) A mechanism by which early spring rainfall in particular might increase non-sown while decreasing 

sown richness is that non-sown species in this system are predominantly exotic and have been 

demonstrated to exhibit advanced phenology compared with sown native species in this system (Wilsey 

et al. 2011), and show strong priority effects when competing with native species (Dickson et al. 2012). 

Further study and direct experimentation will be needed to conclude whether increased precipitation 

during a prairie’s first growing season truly increases weed abundance, and whether the resulting 

competitive pressures stifle development of the sown community for the long term. 

We found that the magnitude of the effect of planting year precipitation on non-sown richness 

was greater than the effects of known major drivers of community assembly in this system: fire 

frequency, soil productivity, seed mix richness and density, and site age. Though there was some prior 

evidence for planting year effects, it was unclear whether these would have minor or major influence on 

 

106 

restoration outcomes. We show that year effects can be large, but are not consistent across all 

community metrics (e.g. we found no year effects on total species richness or sown species cover).  

We predicted that effects of planting year weather conditions might dampen over time as 

plantings matured and were subjected to ongoing management (i.e. deterministic assembly dependent 

on environmental filters, Keddy 1992); however, we found no evidence of this, instead finding support 

for lasting year effects (Vaughn & Young 2010). Additionally, we predicted that effects of planting year 

conditions, specifically related to precipitation, might be stronger in sites with less productive soils 

(Heisler-White et al. 2009) because of their decreased ability to maintain water balance during times of 

water stress; however, we found no evidence for an interaction between year effects and soils, which 

again emphasizes the importance of planting year conditions across a diversity of site conditions (Stuble 

et al. 2017). 

Because of the increased non-sown and decreased sown richness, wet planting years also 

correlated with a higher Bray-Curtis dissimilarity between the seed mix composition and the standing 

vegetation. These results suggest the need for stronger weed control measures following wet planting 

years and, if possible, support for planting during drier years— a suggestion that previously would have 

seemed counter-intuitive. Our work contrasts work by Bakker et al. (2003) and MacDougall et al. (2008) 

which found that a non-sown dominant invader prevailed when restoration occurred in a particularly 

dry year, which might be a result of climactic differences between our Midwestern sites and the Great 

Plains, where water may be more limiting, such that drought years more dramatically prevent 

establishment of sown species. 

The impact of planting-year weather patterns on restoration outcomes can also be examined 

through the lens of future climate projections. In the Midwestern United States, projections are for 

warmer average temperatures, increased spring rainfall, and decreased summer rainfall (Wuebbles & 

Hayhoe 2004). A decrease in summer rainfall may not be as much of a burden to newly planted prairie 

 

107 

restorations as an increase in spring rainfall, which in our study was correlated with higher weed 

pressures that persisted over time, lower richness of sown species, and a vegetation community more 

dissimilar than the sown seed mix. Our results would therefore suggest an increasing weed burden for 

prairie restoration practitioners in sites planted under future climate scenarios. 

Evidence for lasting planting year effects lends support to historically contingent assembly 

models (e.g. Fukami 2015), demonstrating that past events can influence the composition of a current 

community, regardless of deterministic factors like soils or disturbance. Past work has demonstrated the 

impact of assembly history on experimental communities, including some work in grasslands (e.g., Chase 

2003; Martin & Wilsey 2012; Plückers et al. 2013; Helsen et al. 2016; Werner et al. 2016; Young et al. 

2016), and a handful of studies have specifically demonstrated year effects in restored grasslands 

(Bakker & Wilson 2004; MacDougall et al. 2008; Stuble et al. 2017; Groves & Brudvig 2018; Manning & 

Baer 2018). Our work builds on this body of research by examining the impact of historical contingency 

across 83 sites that have been restored across time and space, in context of other factors, and in the 

pragmatic setting of prairies restored by practitioners.   

Our work also contributes important understanding to the question of why restoration 

outcomes vary, showing that both stochastic (weather) and deterministic factors (mix richness, soil 

productivity) factors have roles to play. The marrying of these two principles has been important to 

community ecology over the past decade (Adler et al. 2007; Chase 2007; Chase & Myers 2011; Myers & 

Harms 2011); we show how they apply to restoration with important implications for interpreting 

restoration outcomes. 

 

 

 

 

 

108 

ACKNOWLEDGEMENTS 

 

We acknowledge George Wheeler for assistance collecting field data. Jared Foster from Native 

Connections, Tom Simpson from the McHenry County Conservation District, and Stuart Orr, Ted Anchor, 

Alyssa Nyberg, and Fiona Becker with the Indiana Chapter of The Nature Conservancy were instrumental 

in providing information about the prairies. Chad Zirbel helped extensively with statistical analyses, and 

helped us utilize code from “sapsuckRz,” available on GitHub, to interpolate PRISM weather data and 

calculate degree days. This material is based upon work supported by the Agriculture and Food Research 

Initiative Competitive Grant No. 2016-67012-24680 (J.T.B.) from the USDA National Institute of Food and 

Agriculture and the National Science Foundation under Grant No. DEB-1552197 (L.A.B.) and Graduate 

Research Fellowship DGE-1424871 (A.M.G.). 

 

 

109 

APPENDIX 

 

110 

Figures 

 

APPENDIX 

Figure 4.1 Map of study sites.  

 

 

111 

 

(A) 

 (B)  

(C)  

 

 

  

Figure 4.2. Photos of restored prairies. (A) Marengo Ridge (planted 2009), McHenry County, Illinois on 

September 13, 2016. (B) Kankakee Sands Preserve unit J (planted 2009), Newton County, Indiana on 

September 8, 2016, and (C) Ed Lowe Foundation (planted 2011), Cass County, Michigan on July 27, 2016. 

 

112 

 

Figure 4.3. Scatter plots and pairwise correlations (r) between weather variables. 

 

 

 

113 

 

 

 

Figure 4.4. Significant effects of planting year weather on restoration outcomes. We found an effect of 

planting year weather (PC2: rainfall) on (A) the richness of non-sown species (1000 m2 site), (B) the 

cover of non-sown species (mean 1 m2 plot), (C) the richness of sown species (1000 m2 site), and (D) on 

the Bray-Curtis dissimilarity between the seed mix and the 2016 vegetation composition.  

 

114 

 

Figure 4.5. The effect of planting year temperature (weather PC1) on 2016 non-sown species cover 

(mean 1 m2 plot) varied depending on site age. For plotting, sites within one standard deviation from 

the average PC1 value were assigned the “average” category (58 sites). Sites more than one standard 

deviation above average were assigned “warmest: (11 sites); sites more than one standard deviation 

below average were assigned “coolest” (14 sites). 

 

 

115 

 

Figure 4.6. The relative importance of weather conditions and other factors on different measures of 

restoration outcomes across 83 restored prairies in the Midwestern USA. Each panel shows the 

standardized regression coefficients with 95% confidence intervals for each continuous predictor. 

Models included interactions between weather PC1, PC2, PC3 and both site age and soil productivity. 

Interaction effects and the rest of the full statistical results are reported in Table 4.4. 

 

 

116 

                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   

 

 

 

 

 

 

 

 

 

 

Figure 4.7. Scatter plots of the three PC axis rotations of weather variables against each prairie 

assembly predictor. PC axis 1 (temperature) is plotted against (A) site age, (B) seed mix richness, (C) 

state, (D), soil productivity, (E) seeding rate, and (F) years since last prescribed fire; PC axis 2 (rainfall) is 

likewise plotted against (G) site age, (H) seed mix richness, (I) state, (J), soil productivity, (K) seeding 

rate, and (L) years since last prescribed fire; and PC axis 3 (mild droughts) is plotted against (M) site age, 

(N) seed mix richness, (O) state, (P), soil productivity, (Q) seeding rate, and (R) years since last prescribed 

fire. 

 

 

117 

Figure 4.7      ’   

 

 

 

 

 

 

 

 

 

 

 

118 

 

 

 

 

 

 

 

 

 

Figure 4.8. Distributions of weather and site condition variables across sites. 

 

 

119 

Figure 4.8      ’   

 

 

 

 

 

 

 

 

120 

  

  

F      4.8      ’   

 

 

 

 

 

 

121 

Tables 

 

 

 

 

 

 

Table 4.1. Weather variable rotations for PC axes 1, 2, and 3. Values >|0.3| are denoted in bold for 

emphasis. 

 

Spring weather (March 1 – June 1)  
Degree day accumulation 
Precipitation accumulation 
Summer weather (June 1 - September 1) 
Degree day accumulation 
Precipitation accumulation 
Total growing season weather (March 1 – September 1) 
Degree day accumulation 
Precipitation accumulation 
Mean low temp 
Growing season extremes 
Maximum degree days in a 30-day period 
Minimum precipitation in a 30-day period 
Maximum number of days without a precipitation event 
Mean number of rainy days per 30-day period 
 

PC1 
(temperature) 

PC2 
(rainfall) 

PC3 
(mild  
droughts) 

0.28 
-0.20 

0.44 
-0.19 

0.47 
-0.25 
0.39 

0.45 
-0.10 
-0.036 
-0.12 

0.28 
0.37 

-0.37 
-0.056 

0.058 
0.35 

0.17 
0.49 
0.33 

0.022 
0.32 
-0.26 
0.34 

0.17 
-0.33 

-0.054 
-0.24 
-0.15 

0.13 
0.52 
-0.48 
0.35 

 

 

 

 

 

 

 

 

122 

Table 4.2. Six variables used as response metrics to evaluate effects of planting year weather 

conditions and other assembly drivers on prairie restoration outcomes. 

Restored prairie response variables 
Site level (1000 m2 survey area) 
Richness of sown species 
Richness of non-sown species 
Total species richness 
Plot level (mean 1 m2) 
Cover of sown species 
Cover of non-sown species 
Bray-Curtis dissimilarity between the composition of  
       the seed mix sown and the vegetation at the site 

 

123 

Table 4.3. Soil variable rotations for PC axis 1. PC axis 1 accounted for 40.5% of the variation in the soils 

data. PC axis 2 accounted for 15.4%, PC axis 3 accounted for 12.0%. 

 
pH 
Soil organic matter 
S (ppm) 
P (mg/kg) 
Ca (mg/kg) 
Mg (mg/kg) 
K (mg/kg) 
Na (mg/kg) 
B (mg/kg half detection) 
Fe (mg/kg) 
Mn (mg/kg) 
Cu (mg/kg half detection) 
Zn (mg/kg) 
Al (mg/kg) 
Water holding capacity 
% Clay 
% Silt 
% Sand 
 

PC1 
0.15 
0.28 
0.081 
-0.14 
0.31 
0.33 
0.23 
0.17 
0.28 
-0.13 
0.22 
0.29 
-0.027 
0.045 
0.29 
0.31 
0.28 
-0.31 

 

 

124 

Table 4.4. Complete statistical results for models analyzing the relative importance of weather 

conditions and other factors on prairie restoration outcomes. Asterisks denote statistical significance at 

p<0.05. 

F 
0.019 
0.86 
0.0020 
0.72 
2.65 
0.71 
0.0022 
0.011 
0.27 
1.18 
0.0001 
2.71 
0.82 
0.0001 

p-value 
0.89 
0.36 
0.96 
0.40 
0.11 
0.40 
0.96 
0.92 
0.61 
0.28 
0.99 
0.10 
0.37 
0.99 

R2 
0.0075 
0.0020 
0.048 
0.00070 
0.032 
0.017 
0.00030 
0.0022 
0.0010 
0.025 
0.029 
0.097 
0.011 
0.023 

A. Species richness (1000 m2 site)                                                                                  Model R2=0.24 
Predictor 
Weather PC1 (temperature) 
Weather PC2 (precipitation) 
Weather PC3  
Seed mix richness 
Seeding density 
Years since last fire 
Soil productivity 
Site age 
Weather PC1 (temperature) X Site Age 
Weather PC2 (precipitation) X Site Age 
Weather PC3 X Site Age 
Weather PC1 (temperature) X Soils 
Weather PC2 (precipitation) X Soils 
Weather PC3 X Soils 
B. Sown species richness (1000 m2 site)                                                                        Model R2=0.22 
R2 
Predictor 
0.0026 
Weather PC1 (temperature) 
0.049 
Weather PC2 (precipitation) 
0.020 
Weather PC3  
0.12 
Seed mix richness 
0.00028 
Seeding density 
0.011 
Years since last fire 
0.0029 
Soil productivity 
0.0010 
Site age 
0.0034 
Weather PC1 (temperature) X Site Age 
0.0028 
Weather PC2 (precipitation) X Site Age 
0.0045 
Weather PC3 X Site Age 
0.0042 
Weather PC1 (temperature) X Soils 
0.013 
Weather PC2 (precipitation) X Soils 
Weather PC3 X Soils 
0.00028 
C. Non-sown species richness (1000 m2 site)                                                               Model R2=0.24 
Predictor 
Weather PC1 (temperature) 
Weather PC2 (precipitation) 
Weather PC3  
Seed mix richness 
Seeding density 
Years since last fire 
Soil productivity 

p-value 
0.68 
0.070 
0.25 
0.0043** 
0.89 
0.39 
0.66 
0.80 
0.64 
0.67 
0.59 
0.60 
0.35 
0.89 

F 
0.17 
3.40 
1.35 
8.75 
0.018 
0.76 
0.19 
0.067 
0.22 
0.18 
0.30 
0.28 
0.88 
0.018 

F 
0.038 
7.70 
0.94 
2.44 
3.59 
0.073 
0.11 

β 
-0.11 
0.92 
0.056 
1.36 
3.40 
-1.23 
-0.035 
0.16 
0.42 
-1.36 
0.016 
0.39 
0.33 
0.0071 

β 
-0.23 
-1.30 
1.04 
3.38 
0.20 
-0.91 
-0.23 
0.29 
-0.28 
0.38 
-0.53 
0.090 
0.24 
0.057 

β 
0.12 
2.22 
-0.98 
-2.02 
3.20 
-0.32 
0.20 

p-value 
0.85 
0.0072** 
0.34 
0.12 
0.062 
0.79 
0.74 

R2 
0.00058 
0.10 
0.014 
0.036 
0.052 
0.0011 
0.0016 

 

125 

0.92 
0.29 
0.089 
0.62 
0.12 
0.77 
0.92 

-0.13 
0.70 
-1.74 
0.54 
0.30 
0.085 
-0.050 

β 
-2.11 
-2.16 
2.13 
-1.62 
0.065 
-2.77 
-3.27 
-2.10 
-2.56 
1.54 
1.29 
-0.15 
0.55 
-0.28 

p-value 
0.16 
0.27 
0.39 
0.61 
0.99 
0.34 
0.027* 
0.50 
0.11 
0.53 
0.63 
0.76 
0.44 
0.81 

      4.4      ’   
 
0.00016 
Site age 
0.017 
Weather PC1 (temperature) X Site Age 
0.043 
Weather PC2 (precipitation) X Site Age 
0.0037 
Weather PC3 X Site Age 
0.036 
Weather PC1 (temperature) X Soils 
0.0013 
Weather PC2 (precipitation) X Soils 
Weather PC3 X Soils 
0.00017 
D. Cover of sown species (1 m2 plots)                                                                           Model R2=0.19 
R2 
Predictor 
0.029 
Weather PC1 (temperature) 
0.018 
Weather PC2 (precipitation) 
0.011 
Weather PC3  
0.0040 
Seed mix richness 
<0.0001 
Seeding density 
0.014 
Years since last fire 
0.072 
Soil productivity 
0.0069 
Site age 
0.037 
Weather PC1 (temperature) X Site Age 
0.0060 
Weather PC2 (precipitation) X Site Age 
0.0035 
Weather PC3 X Site Age 
0.0015 
Weather PC1 (temperature) X Soils 
0.0090 
Weather PC2 (precipitation) X Soils 
Weather PC3 X Soils 
0.00087 
E. Cover of non-sown species (1 m2 plots) (log transformed)                                   Model R2=0.40 
R2 
Predictor 
0.037 
Weather PC1 (temperature) 
0.054 
Weather PC2 (precipitation) 
0.0020 
Weather PC3  
0.066 
Seed mix richness 
0.0027 
Seeding density 
0.0084 
Years since last fire 
0.17 
Soil productivity 
0.084 
Site age 
0.13 
Weather PC1 (temperature) X Site Age 
0.0037 
Weather PC2 (precipitation) X Site Age 
0.0011 
Weather PC3 X Site Age 
0.0080 
Weather PC1 (temperature) X Soils 
0.0090 
Weather PC2 (precipitation) X Soils 
Weather PC3 X Soils 
<0.0001 
F. Bray-Curtis dissimilarity between seed mix and 2016 vegetation                       Model R2=0.54 
Predictor 
Weather PC1 (temperature) 
Weather PC2 (precipitation) 
Weather PC3  
Seed mix richness 

p-value 
0.12 
0.056 
0.72 
0.035* 
0.67 
0.46 
0.00049*** 
0.017* 
0.0027** 
0.62 
0.79 
0.47 
0.44 
0.98 

β 
0.085 
0.14 
0.032 
0.24 
-0.062 
0.077 
0.19 
0.27 
0.18 
-0.044 
-0.026 
0.012 
-0.020 
0.0012 

0.011 
1.12 
2.98 
0.25 
2.48 
0.083 
0.011 

F 
1.99 
1.23 
0.75 
0.27 
0.0003 
0.94 
5.13 
0.46 
2.57 
0.40 
0.23 
0.097 
0.60 
0.058 

F 
2.50 
3.77 
0.13 
4.64 
0.18 
0.56 
13.45 
6.02 
9.73 
0.24 
0.072 
0.53 
0.60 
0.0008 

F 
2.40 
6.66 
0.36 
16.78 

β 
0.012 
0.026 
-0.0075 
0.066 

p-value 
0.13 
0.012* 
0.55 
0.00012*** 

R2 
0.035 
0.092 
0.0055 
0.20 

 

126 

      4.4      ’   
 
Seeding density 
Years since last fire 
Soil productivity 
Site age 
Weather PC1 (temperature) X Site Age 
Weather PC2 (precipitation) X Site Age 
Weather PC3 X Site Age 
Weather PC1 (temperature) X Soils 
Weather PC2 (precipitation) X Soils 
Weather PC3 X Soils 
 

-0.012 
0.0082 
0.011 
0.050 
0.0068 
-0.0054 
0.00041 
0.0029 
-0.0048 
-0.0014 

0.35 
0.31 
0.024 
10.09 
0.70 
0.19 
0.0009 
1.47 
1.69 
0.060 

0.55 
0.58 
0.88 
0.0023** 
0.41 
0.67 
0.98 
0.23 
0.20 
0.81 

0.0053 
0.0048 
0.0036 
0.13 
0.022 
0.025 
0.00090 
0.022 
0.025 
0.00090 

 

127 

LITERATURE CITED 

 

128 

LITERATURE CITED 

 

Adler PB, Hille Ris Lambers J, Levine JM (2007) A niche for neutrality. Ecology letters 10:95–104 

Alford RA, Wilbur HM (1985) Priority effects in experimental pond communities: Competition between 

Bufo and Rana. Ecology 66:1097–1105 

Averett JM, Klips RA, Nave LE, Frey SD, Curtis PS (2004) Effects of soil carbon amendment on nitrogen 
availability and plant growth in an experimental tallgrass prairie restoration. Restoration Ecology 
12:568–574 

Baer SG, Engle DM, Knops JMH, Langeland K a, Maxwell BD, Menalled FD, Symstad AJ (2009) 

Vulnerability of rehabilitated agricultural production systems to invasion by nontarget plant 
species. Environmental management 43:189–96 

Bakker JD, Wilson SD, Christian JM, Li X, Ambrose LG, Waddington J (2003) Contingency of grassland 

restoration on year, site, and competition from introduced grasses. Ecological Applications 13:137–
153 

Bakker JD, Wilson SD (2004) Using ecological restoration to constrain biological invasion. Journal of 

Applied Ecology 41:1058–1064 

Bakker JP, Poschlod P, Strykstra RJ, Bekker RM, Thompson K (1996) Seed banks and seed dispersal: 

Important topics in restoration ecology. Acta Botanica Neerlandica 45:461–490 

Blumenthal DM, Jordan NR, Russelle MP (2003) Soil carbon addition controls weeds and facilitates 

prairie restoration. Ecological Applications 13:605–615 

Cardinale BJ, Duffy JE, Gonzalez A, Hooper DU, Perrings C, Venail P, Narwani A, Mace GM, Tilman D, 

Wardle DA, Kinzig AP, Daily GC, Loreau M, Grace JB, Larigauderie A, Srivastava DS, Naeem S (2012) 
Biodiversity loss and its impact on humanity. Nature 486:59–67 

Chase JM (2003) Community assembly: When should history matter? Oecologia 136:489–98 

Chase JM (2007) Drought mediates the importance of stochastic community assembly. Proceedings of 

the National Academy of Sciences of the United States of America 104:17430–4 

Chase JM, Myers JA (2011) Disentangling the importance of ecological niches from stochastic processes 

across scales. Philosophical transactions of the Royal Society of London. Series B, Biological 
sciences 366:2351–2363 

Connell JH, Slatyer RO (1977) Mechanisms of succession in natural communities and their role in 

community stability and organization. The American Naturalist 111:1119–1144 

Cramer VA, Hobbs RJ, Standish RJ (2008) What’s new about old fields? Land abandonment and 

ecosystem assembly. Trends in Ecology and Evolution 23:104–12 

 

129 

Dickson TL, Hopwood JL, Wilsey BJ (2012) Do priority effects benefit invasive plants more than native 

plants? An experiment with six grassland species. Biological Invasions 14:2617–2624 

Dobson AP, Bradshaw AD, Baker AJM (1997) Hopes for the future: Restoration ecology and conservation 

biology. Science 277:515–522 

Doll JE, Haubensak KA, Bouressa EL, Jackson RD (2011) Testing disturbance, seeding time, and soil 
amendments for establishing native warm-season grasses in non-native cool-season pasture. 
Restoration Ecology 19:1–8 

Foley JA, DeFries R, Asner GP, Barford C, Bonan G, Carpenter SR, Chapin FS, Coe MT, Daily GC, Gibbs HK, 
Helkowski JH, Holloway T, Howard EA, Kucharik CJ, Monfreda C, Patz JA, Prentice IC, Ramankutty N, 
Snyder PK (2005) Global consequences of land use. Science 309:570–574 

Forrest JRK, James DT (2011) An examination of synchrony between insect emergence and flowering in 

Rocky Mountain meadows. Ecological Monographs 81:469–491 

Fukami T (2015) Historical contingency in community assembly: Integrating niches, species pools, and 

priority effects. Annual Review of Ecology, Evolution, and Systematics 46:1–23 

Fukami T, Mordecai EA, Ostling A (2016) A framework for priority effects. Journal of Vegetation Science 

27:655–657 

Grant K, Kreyling J, Heilmeier H, Beierkuhnlein C, Jentsch A (2014) Extreme weather events and plant–

plant interactions: shifts between competition and facilitation among grassland species in the face 
of drought and heavy rainfall. Ecological Research 29:991–1001 

Grman E, Bassett T, Brudvig L (2013) Confronting contingency in restoration: Management and site 

history determine outcomes of assembling prairies, but site characteristics and landscape context 
have little effect. Journal of Applied Ecology 50:1234–1243 

Grman E, Bassett T, Zirbel CR, Brudvig LA (2015) Dispersal and establishment filters influence the 

assembly of restored prairie plant communities. Restoration Ecology 23:892–899 

Groves AM, Brudvig LA (2018) Inter-annual variation in precipitation and other planting conditions 

impacts seedling establishment in sown plant communities. Restoration Ecology 

Heisler-White JL, Blair JM, Kelly EF, Harmoney K, Knapp AK (2009) Contingent productivity responses to 

more extreme rainfall regimes across a grassland biome. Global Change Biology 15:2894–2904 

Helsen K, Hermy M, Honnay O (2016) A test of priority effect persistence in semi-natural grasslands 

through the removal of plant functional groups during community assembly. BMC Ecology 16:22 

Hobbs RJ, Harris JA (2001) Restoration Ecology: Repairing the Earth’s Ecosystems in the New Millennium. 

Restoration Ecology 9:239–246 

Jones HP, Jones PC, Barbier EB, Blackburn RC, Benayas JMR, Holl KD, McCrackin M, Meli P, Montoya D, 
Moreno-Mateos D (2018) Restoration and repair of Earth’s damaged ecosystems. Proceedings of 
the Royal Society B 285 

 

130 

Keddy PA (1992) Assembly and response rules: Two goals for predictive community ecology. Journal of 

Vegetation Science 3:157–164 

MacDougall AS, Wilson SD, Bakker JD (2008) Climatic variability alters the outcome of long-term 

community assembly. Journal of Ecology 96:346–354 

Manning GC, Baer SG (2018) Interannual variability in climate effects on community assembly and 

ecosystem functioning in restored prairie. Ecosphere 9:e02327 

Martin LM, Wilsey BJ (2012) Assembly history alters alpha and beta diversity, exotic-native proportions 

and functioning of restored prairie plant communities Cadotte, M, editor. Journal of Applied 
Ecology 

Mason CH, Perreault Jr. WD (1991) Collinearity, power, and interpretation of multiple regression 

analysis. Journal of Marketing Research 28:268–280 

Myers JA, Harms KE (2011) Seed arrival and ecological filters interact to assemble high-diversity plant 

communities. Ecology 92:676–86 

Piper JK, Schmidt ES, Janzen AJ (2007) Effects of species richness on resident and target species 

components in a prairie restoration. Restoration Ecology 15:189–198 

Plückers C, Rascher U, Scharr H, von Gillhaussen P, Beierkuhnlein C, Temperton VM (2013) Sowing 

different mixtures in dry acidic grassland produced priority effects of varying strength. Acta 
Oecologica 53:110–116 

Prism Climate Group (2018) PRISM Climate Data. http://prism.oregonstate.edu 

Rowe HI (2010) Tricks of the trade: Techniques and opinions from 38 experts in tallgrass prairie 

restoration. Restoration Ecology 18:253–262 

Ruiz-Jaen MC, Mitchell Aide T (2005) Restoration success: How is it being measured? Restoration 

Ecology 13:569–577 

Samson F, Knopf F (1994) Prairie conservation in North America. BioScience 44:418–421 

Stuble KL, Fick SE, Young TP (2017) Every restoration is unique: Testing year effects and site effects as 

drivers of initial restoration trajectories. Journal of Applied Ecology 54:1051–1057 

Suding KN, Collins SL, Gough L, Clark C, Cleland EE, Gross KL, Milchunas DG, Pennings S (2005) 

Functional- and abundance-based mechanisms explain diversity loss due to N fertilization. 
Proceedings of the National Academy of Sciences of the United States of America 102:4387–92 

Suding KN (2011) Toward an era of restoration in ecology: Successes, failures, and opportunities ahead. 

Annual Review of Ecology, Evolution, and Systematics 42:465–487 

Thompson K, Grime JP (1983) A Comparative Study of Germination Responses to Diurnally-Fluctuating 

Temperatures. Journal of Applied Ecology 20:141–156 

 

131 

Trudgill DL, Honek A, Li D, Van Straalen NM (2005) Thermal time - Concepts and utility. Annals of Applied 

Biology 146:1–14 

Vaughn KJ, Young TP (2010) Contingent conclusions: Year of initiation influences ecological field 

experiments, but temporal replication is rare. Restoration Ecology 18:59–64 

Wainwright CE, Cleland EE (2013) Exotic species display greater germination plasticity and higher 

germination rates than native species across multiple cues. Biological Invasions 15:2253–2264 

Weiher E, Freund D, Bunton T, Stefanski A, Lee T, Bentivenga S (2011) Advances, challenges and a 

developing synthesis of ecological community assembly theory. Philosophical Transactions of the 
Royal Society B: Biological Sciences 366:2403–13 

Werner CM, Vaughn KJ, Stuble KL, Wolf K, Young TP (2016) Persistent asymmetrical priority effects in a 

California grassland restoration experiment. Ecological Applications 26:1624–1632 

Wilsey BJ, Daneshgar PP, Polley HW (2011) Biodiversity, phenology and temporal niche differences 

between native- and novel exotic-dominated grasslands. Perspectives in Plant Ecology, Evolution 
and Systematics 13:265–276 

Wuebbles DJ, Hayhoe K (2004) Climate change projections for the United States Midwest. Mitigation 

and Adaptation Strategies for Global Change 9:335–363 

Young TP, Stuble KL, Balachowski JA, Werner CM (2016) Using priority effects to manipulate competitive 

relationships in restoration. Restoration Ecology 25:S114–S123 

 

 

 

 

132 

CHAPTER FIVE 

CONCLUDING REMARKS 

 

This dissertation has sought to strike a balance between research that could inform restoration 

and research that could inform ecological theory. The study of planting year effects on community 

assembly provides a framework upon which tests of historically contingent assembly can be merged 

with practical, mechanistic questions that could be of more immediate interest to restoration 

practitioners. 

Through a four-year field experiment at the Kellogg Biological Station’s Lux Arbor Reserve, we 

were able to find evidence to support the year effect concept. We were also able to disentangle the 

mechanistic considerations of the influence of precipitation versus other weather conditions, and the 

impacts of weather on the germination of seeds versus the survival of seedlings. We learned, by directly 

watering experimental plots, that increased rainfall increased the number of seeds that emerged from 

both the sown seed mix and the weed seed bank, and that this increase was particularly dramatic for 

weed species. Watering did not, however, affect the identities of the species that emerged. Instead, 

different communities arose in three different planting years that had different weather conditions—

even when comparing among plots that received the same watering treatment in each planting year. 

This suggests that conditions besides rainfall, such as temperature, dictate which sown and non-sown 

species will establish in the first year of a prairie restoration. Understanding these planting-year 

differences were an important first step toward understanding how year effects operate in this system. 

Without this evidence for community differences in the first growing season, there would be no basis 

from which different community assembly trajectories could be catalyzed by planting year conditions, 

and no biological mechanism by which year effects could occur. 

 

133 

Differences among prairies in their first growing season did not alone constitute a year effect, 

however, which is why we followed these prairies into their second, third, or fourth growing seasons. 

We saw the cover and richness of sown species fluctuate each year, seemingly responding to current 

conditions regardless of the prairie’s age or original planting conditions. But the cover of non-sown 

species remained significantly different across the prairies planted in different years, as did the overall 

community composition, despite the prairies being initially planted using the same mix and the same 

methods at the same site. This provides evidence that differences in first-year emergence patterns can 

indeed lead to continuing year effects. 

Throughout this experiment, we found no evidence that increased seed mix richness mitigated 

any adverse impacts of weather conditions. We predicted that sowing a greater number of species, 

much like diversifying a portfolio, would increase the total number of species that would be successful 

under any given conditions, and therefore increase the overall success of a planting (i.e. increasing 

species richness and cover). Although we found increases in the number of sown species with increased 

mix richness, a higher richness mix did not lead to an increased quantity of sown species under any 

planting conditions. In addition, it seems that if species in a high richness mix are not chosen carefully, 

an increase in richness (if seeding density is held constant) could possibly take away individuals from 

species that are more consistent in their emergence across a variety of weather conditions. Practitioners 

likely already deal with this tradeoff between rarity and likelihood of establishing when considering 

species for inclusion in a high diversity mix. 

Our final endeavor was to learn whether the results from this relatively small scale, highly 

controlled field experiment would be relevant to restored prairies of different ages, larger sizes, 

different geographic locations, and diverse prairie management strategies. We surveyed 83 restored 

prairies that ranged from 2-19 years old in Illinois (McHenry County Conservation District), Indiana 

(Nature Conservancy’s Kankakee Sands Reserve), and Michigan (private plantings restored by Native 

 

134 

Connections) to look for signatures of planting year conditions on the present-day plant communities. 

Despite the variation in landscapes, soils, seed mixes, ages, and management decisions across the 

prairies, we found that the best predictor of the richness of non-sown species was precipitation in the 

planting year— even when compared to predictors like prescribed fire frequency, soil productivity, or 

site age. Increased planting year rainfall was also marginally correlated with decreased sown species 

richness and increased non-sown species cover. Prairies planted in rainier years were also found to be 

less similar in composition to their seed mixes than those planted in drier years. In general, wet planting 

years led to worse prairie restoration outcomes. 

There are at least two ways to synthesize these results. In the first scenario, increased planting 

year precipitation acts through increases in the abundance of weeds during the first growing season. 

The dominant weeds in a young prairie are typically ruderal species that drop out in future years, so 

their identities in this scenario (which vary with non-precipitation metrics such as temperature) are 

irrelevant. The increased density of a flush of ruderal weeds, despite its transience, suppresses the 

establishment of target species in a way that leaves the prairie vulnerable to a future of continually 

elevated non-sown species pressures. In some cases, aggressive perennials are also present in a young 

prairie, and increasing their numbers increases their ability to exert priority and maintain their foothold 

into the future, directly adding to the future non-sown community. 

An alternate scenario is that different first-year community compositions are indeed the catalyst 

for lasting year effects, but that the quantity of water hitting the ground does not capture the influence 

of a rainy season on the identities of emerging plants. Increased precipitation across a growing season 

could be associated with any number of other weather conditions or biotic influences besides rainfall 

per se: greater cloud cover, increased humidity, increased fungal pathogen loads, different behavioral 

patterns of granivores or herbivores, and much more. Therefore, it is possible that the year effects we 

demonstrated in our field experiment (i.e., different community compositions across three planting 

 

135 

years when watering treatment was held constant) were still caused by something linked to the 

raininess of that season, despite maintaining a constant amount of watering in each year. In this 

scenario, a more traditional historically contingent assembly model can be invoked, in which the 

different species identities early in community assembly dictate which species are able to establish and 

persist over time. In this case, the species that establish in a rainy planting year are more conducive to 

the long-term establishment of non-sown species than are the species that establish in a drier year.   

Although the first scenario is mechanistically more appealing, evidence from our experiment 

suggests the second scenario is more likely. This is because we did not find any impacts of first year 

watering treatment on the communities that we followed over time, which we would have expected if 

the increase in first year weed abundance was the catalyst for different community assembly 

trajectories.  

Regardless of the mechanistic explanation for how year effects occur, finding a correlation 

between any planting year weather condition and the present-day community (up to 19 years later) 

across such a diverse spectrum of sites is noteworthy, and gives concrete support for historically 

contingent models of community assembly. Under these models, the outcome of community assembly 

is altered depending on what happened earlier in assembly—like the establishment of different species, 

or a flush of weeds at the outset of a restoration project. Although it increasingly seems that any 

ecological conclusion other than “it depends” is rather suspect, our results do provide strong evidence 

that historically contingent processes can exert a real force on community assembly in this system. 

For restoration practitioners, we may not have unlocked any great secrets of restoration. I 

would surmise that most Midwestern prairie restoration managers are already operating at full weed 

suppression capacity, and learning that a wet planting year makes weedier prairies will not be terribly 

helpful until we can predict annual weather patterns. Even then, only the most fortunate practitioners 

will likely have the resources and time necessary to stall restoration activities for an entire year or more 

 

136 

until conditions improve. That said, understanding that non-sown species pressures could be particularly 

and continuously high in prairies restored under wetter conditions could prepare practitioners to 

allocate more resources toward aggressive weed control measures for those sites, such as mowing or 

herbicide application. 

In the end, it is this author’s hope that these six years of graduate study have contributed some 

incremental step forward in understanding historically contingent assembly models as well as variation 

in tallgrass prairie restoration outcomes. Future research should continue to explore the mechanisms by 

which year effects occur. Specifically, a direct test of the effects of different temperatures on first year 

community establishment, plus a direct test of the effects of increased weed abundance (compared to 

different weed identities) on sown species and general community establishment over time would add 

clarity to the ideas presented here. 

Thanks for reading. 

 

 

 

137