Journal of Biogeography
SUPPORTING INFORMATION
A continent-wide study reveals clear relationships between regional abiotic
conditions and post-dispersal seed predation
John L. Orrock, Elizabeth T. Borer, Lars A. Brudvig, Jennifer Firn, Andrew S. MacDougall,
Brett A. Melbourne, Louie H. Yang, Dirk V. Baker, Avi Bar-Massada, Michael J. Crawley, Ellen
I. Damschen, Kendi F. Davies, Daniel S. Gruner, Adam D. Kay, Eric Lind, Rebecca L.
McCulley and Eric W. Seabloom
Appendix S1 Study data and additional details on study sites, methods, ancillary seed-removal
trials and supporting results.
Additional information about climatic data
Data used in this study are listed in Table S1. The strong significant relationships between mean
annual precipitation from WorldClim, from historical data used to calculate annual actual
evapotranspiration (AET) and from average site precipitation data from 2008 to 2010, suggested
that WorldClim and weather station data were excellent proxies for conditions at each site (all
r ≥ 0.72, P ≤ 0.018); sources of these data are described in the main text. Similarly, mean annual
precipitation and temperature from WorldClim were significantly correlated with weather station
data collected during 2009, the year of the study (all r ≥ 0.72, P ≤ 0.018), and AET calculated
using long-term data and 2009 data were in very close agreement (r = 0.96, P ≤ 0.001).
Pairwise correlations among climate variables across the 11 sites can be found in Table
S2. We also examined the relationship between total average aboveground biomass collected at
each site in 2007, 2008, 2009 (for details on biomass collection, see O’Halloran et al., 2013),
aboveground biomass collected during only 2009, and the climatic variables used in our
1
analyses, including AET (calculated using long-term data, as described in the main text); for the
Boulder (Colorado, USA) site, only 2008 and 2009 data were available. When data from all sites
were used, there was no significant relationship between biomass and long-term AET, biomass
and mean annual precipitation, or any of the other climatic variables that were calculated using
long-term data (Table S2). However, upon closer inspection of the data, we suspected that
herbaceous biomass from one of our sites (a longleaf pine savanna) might not be expected to
respond directly to AET and precipitation, as scattered mature longleaf pine trees (Pinus
palustris Miller) might intercept light energy and precipitation before it can be used by
understorey vegetation. Examination of the data without using that site (i.e. using 10 total sites)
supported this hypothesis, as there was a significant relationship between AET, mean annual
precipitation and biomass when that site was excluded (Table S3). Importantly, there was also a
significant relationship between local biomass production and seed removal in that analysis
(Table S3), further suggesting that the link between seed removal and large-scale climatic data in
grassland ecosystems is often mediated by local herbaceous plant production.
Additional details on seed-removal depots
Removal depots were based on a modified version of the protocol used in Moles et al. (2003).
During seed placement in removal depots, surface litter and vegetation (where necessary) were
gently displaced to allow placement of the seeds on the soil surface. When possible, seeds were
placed in natural depressions where they occurred, to minimize wind or rain removal (although
rain events were rarely observed during the study). Depots were marked with a toothpick within
2 cm of the depot and by placing a larger marker (e.g. a pin flag or flagging tape) within 30 cm
of the depot.
Viability of heat-treated seeds
Viability assays using tetrazolium (Hahn & Orrock, 2014) indicated that Avena sativa seeds
heated for 15 min at 150 °C were not viable compared with control seeds that were left at room
temperature (24 °C) for 15 min (five replicates of 10 seeds per temperature level). Zero seeds
were viable after being heated to 150 °C, whereas the proportion of viable seeds from the 24 °C
2
treatment was 0.92 ± 0.06 (binomial generalized linear model, deviance = 94.46, 1 d.f., P <
0.001); for additional information regarding methods, see Hahn & Orrock (2014).
Removal of heat-treated seeds and seeds of other species
We conducted three separate follow-up experiments to determine whether or not removal of
heat-treated A. sativa seeds was indicative of removal of other seed species, and to ensure that
heat-treated seeds were removed at similar rates compared with non-treated seeds.
At three of the study locations, removals of additional species were conducted at the
same time as A. sativa removals were conducted, using identical protocols. The sites (locations
and species used in parentheses) were: Cowichan (British Columbia, Canada; Danthonia
californica Bolander), Hall’s Prairie (Kentucky, USA; Andropogon gerardii Vitman) and
Spindletop (Kentucky, USA; Dactylis glomerata L.). Data were pooled within an experimental
plot and evaluated using a paired t-test to compare removal of Avena sativa and removal of the
additional species. There was no difference between removal of A. sativa and the other species
when all three sites were pooled (t = 1.13 P = 0.24, d.f. = 8). Similarly, there was no difference
between A. sativa removal and the other seed species when the data were analysed by site and
species: Danthonia californica (t = –1.00, P = 0.42, d.f. = 2), Andropogon gerardii (t = 1.66,
P = 0.24, d.f. = 2), Dactylis glomerata (t = 0.023, P = 0.84, d.f. = 2). These removals provided
strong evidence, from the same locations and times where Avena sativa removal was assessed,
that removal of A. sativa provides useful information regarding granivory pressure on other
species.
To determine whether heating affected removal of A. sativa, we established 10 paired
depots at one of our research areas where removal rates were relatively high (Savannah River
Site, South Carolina, USA). At each pair of depots, a depot of five unheated seeds was separated
by less than a metre from a depot of five heated seeds. Seeds were deployed for 2 days (2–4
November 2011), identical to the other seed-removal trials used in this experiment. This work
confirmed that the removal of heated seeds was highly correlated with the removal of unheated
seeds (r = 0.77, P < 0.01, d.f. = 8).
3
Removal of heat-treated seeds was also assessed in an ancillary seed-removal experiment
in a restored prairie in Madison (Wisconsin, USA); this experiment provided an opportunity to
assess how removal of A. sativa is related to the removal of other species. Fourteen seed-removal
depots consisting of translucent plastic buckets (for a description of depots, see Mattos et al.,
2013) were established with 1 L of sand and 10 seeds each of A. sativa and five additional
species that varied significantly in size: Andropogon gerardii (seed mass = 2.63 ± 0.08 mg, n = 5
samples of 10 seeds each), Asclepias syriaca L. (4.52 ± 0.19 mg), Schizachyrium scoparium
Michaux (1.03 ± 0.04 mg), Silphium terebinthinaceum Jacq. (16.72 ± 0.22 mg) and Sorghastrum
nutans Nash (1.80 ± 0.04 mg). Seven of the depots had Avena sativa seeds that were heat-treated
(37.17 ± 3.18 mg) while seven depots used A. sativa seeds that were not heat-treated
(32.46 ± 1.82 mg). Depots were placed in the field for 7 days, 1–8 November 2012. There was
no significant difference in removal of A. sativa seeds by heat treatment (t-test, t =1.04, P = 0.32,
d.f. = 12). Moreover, there was a significant positive relationship between the number of A.
sativa seeds removed and the number of seeds removed of Andropogon gerardii (r = 0.89, P <
0.001, d.f. = 12), Asclepias syriaca (r = 0.94, P < 0.001, d.f. = 12), Schizachyrium scoparium
(r = 0.73, P = 0.003, d.f. = 12), Silphium terebinthinaceum (r = 0.94, P < 0.001, d.f. = 12) and
Sorghastrum nutans (r = 0.65, P = 0.01, d.f. = 12). These results suggested that removal of
Avena sativa seeds was also indicative of the removal rates of seeds of native prairie species
across a substantial range of seed masses.
4
REFERENCES
Borer, E.T., Harpole, W.S., Adler, P.B., Lind, E.M., Orrock, J.L., Seabloom, E.W. & Smith,
M.D. (2014) Finding generality in ecology: a model for globally distributed experiments.
Methods in Ecology and Evolution, 5, 65–73.
Hahn, P.G. & Orrock, J.L. (2014) Effects of temperature on seed viability of six Ozark glade
species and eastern redcedar (Juniperus virginiana). American Midland Naturalist, 171,
147–152.
Mattos, K.J., Orrock, J.L. & Watling, J.I. (2013) Rodent granivores generate context-specific
seed removal in invaded and uninvaded habitats. American Midland Naturalist, 169, 168–
178.
Moles, A.T., Warton, D.I. & Westoby, M. (2003) Do small-seeded species have higher survival
through seed predation than large-seeded species? Ecology, 84, 3148–3161.
Nakagawa, S. & Schielzeth, H. (2013) A general and simple method for obtaining R2 from
generalized linear mixed-effects models. Methods in Ecology and Evolution, 4, 133–142.
O’Halloran, L.R., Borer, E.T., Seabloom, E.W. et al. (2013) Regional contingencies in the
relationship between aboveground biomass and litter in the world's grasslands. PLoS ONE,
8, e54988.
5
Table S1 Data used for a continent-wide study of the relationships between regional abiotic conditions and post-dispersal seed
predation. The site at Cedar Creek is a Long-term Ecological Research site (LTER). The study sites at McLaughlin (California, USA)
and Sagehen (California, USA) are part of the University of California Natural Reserve System (UCNRS). The sites at Hopland
(California, USA) and Sierra Foothills (California, USA) are University of California Research and Extension Centers (REC).
Site name
(location)
Latitude
Longitude
Elevation (m)
Mean annual
precipitation (mm)
Mean annual
temperature (°C)
Annual actual
evapotranspiration
(mm)
Annual temperature
range
Intra-annual variation
in mean annual
precipitation (CV)
Inter-annual variation
in mean annual
precipitation (CV)
Proportion of seeds
removed (logit
transformed)
Boulder
(Colorado,
USA)
Cedar Creek
LTER
(Minnesota,
USA)
Cowichan
(British
Columbia,
Canada)
Hall’s
Prairie
(Kentucky,
USA)
Hopland
REC
(California,
USA)
McLaughlin
UCNRS
(California,
USA)
Sagehen
UCNRS
(California,
USA)
Savannah
River
(South
Carolina,
USA)
Sierra
Foothills
REC
(California,
USA)
Spindletop
(Kentucky,
USA)
Tyson
(Missouri,
USA)
39.972
45.401
48.460
36.872
39.013
38.864
39.430
33.344
39.236
38.136
38.519
–105.234
–93.201
–123.380
–86.729
–123.067
–122.400
–120.240
–81.651
–121.280
–84.501
–90.565
1633
270
50
193.6
598
642
1920
71
197
271
169
425
750
764
1282
1127
867
882
1194
935
1140
997
9.7
6.3
9.8
13.6
12.3
13.5
5.7
17.3
15.6
12.5
12.5
506.22
602.06
497.03
759.84
473.36
405.37
345.631
910.24
440.42
728.05
716.40
38.5
45.6
18.5
36.3
31.0
32.2
35.4
32.8
32.6
35.6
39.4
42
51
64
14
87
88
69
19
84
15
18
1.663
2.130
3.304
2.167
2.445
1.965
2.558
4.018
2.718
3.119
3.102
–1.708
–0.523
–1.708
1.437
–1.708
–1.267
–1.041
0.998
–1.398
0.531
0.084
6
Site name
(location)
Mean aboveground
biomass 2009 (g m–2)
Mean aboveground
biomass 2007–2009
(g m–2)
Boulder
(Colorado,
USA)
Cedar Creek
LTER
(Minnesota,
USA)
Cowichan
(British
Columbia,
Canada)
Hall’s
Prairie
(Kentucky,
USA)
Hopland
REC
(California,
USA)
McLaughlin
UCNRS
(California,
USA)
Sagehen
UCNRS
(California,
USA)
Savannah
River
(South
Carolina,
USA)
Sierra
Foothills
REC
(California,
USA)
Spindletop
(Kentucky,
USA)
Tyson
(Missouri,
USA)
243
59
334
650
332
460
107
78
314
402
431
213
134
343
715
202
319
122
95
204
417
470
7
Table S2 Pearson’s correlation values among the climatic variables tested, the dependent variable proportion of seeds removed (logit
transformed) and local site biotic variables. The proportion (prop.) of seeds removed (rem.) was logit-transformed. Lat., latitude;
long., longitude; MAP, mean annual precipitation; MAT, mean annual temperature; AET, annual actual evapotranspiration; ann. temp.
range, annual temperature range; intra-annual MAP CV, coefficient of variation in within-year mean annual precipitation; inter-annual
MAP CV, between-year coefficient of variation in between-year mean annual precipitation; elev., elevation; mean AG biom. 2009,
mean aboveground biomass taken from each site in 2009; mean AG biom. 2007–2009, mean aboveground biomass taken from each
site from sampling areas in 2007, 2008 and 2009 (except Colorado, where data are from 2008 and 2009). Bold is used to highlight the
matrix diagonal.
Variables
Prop. seeds removed (logit)
Latitude
Longitude
MAP
MAT
AET
Ann. temp. range
Intra-annual MAP CV
Inter-annual MAP CV
Elevation
Mean AG biom. 2009
Mean AG biom. 2007–2009
Prop.
Seeds
rem.
(logit)
1.00
–0.56†
0.88***
0.70*
0.38
0.86***
0.34
–0.84**
–0.40
–0.43
0.24
0.49
Lat.
Long.
MAP
MAT
AET
Ann.
temp.
range
Intraannual
MAP CV
–
1.00
–0.430
–
–
1.00
0.39
0.24
–
–
–
1.00
0.57†
0.54†
–0.07
–0.33
–0.03
–
–
–
–
1.00
0.48
–0.24
–0.19
0.23
–
–
–
–
–
1.00
0.23
–
–
–
–
–
–
–
–
–
–
–
–
1.00
–0.52†
0.38
0.33
–0.57†
0.38
0.36
–0.58†
0.10
0.33
–0.58†
–0.65*
–0.45
–0.27
0.36
–0.10
–0.04
–0.15
–0.11
0.92***
0.53†
–0.94***
–0.54†
–0.36
0.03
0.27
–0.88***
–0.50
1.00
–0.35
–0.14
0.20
–0.19
–0.22
0.58†
0.27
–0.16
–0.40
Interannual
MAP
CV
–
–
–
–
–
–
–
–
1.00
0.11
–0.10
–0.44
Elev.
Mean AG
biom.
2009
–
–
–
–
–
–
–
–
–
1.00
–0.31
–0.49
–
–
–
–
–
–
–
–
–
–
1.00
0.91***
Mean AG
biom.
2007–
2009
–
–
–
–
–
–
–
–
–
–
–
1.00
†P  0.10, *P  0.05, **P  0.01, ***P  0.001.
8
Table S3 Pearson’s correlation values among the climatic variables tested, the dependent variable proportion of seeds removed (logit
transformed) and local site biotic variables. Data from a longleaf pine savannah grassland (Savannah River Site, SC) were not includes
in the analysis. The proportion (prop.) of seeds removed (rem.) was logit-transformed. Lat., latitude; long., longitude; MAP, mean
annual precipitation; MAT, mean annual temperature; AET, annual actual evapotranspiration; ann. temp. range, annual temperature
range; intra-annual MAP CV, coefficient of variation in within-year mean annual precipitation; inter-annual MAP CV, between-year
coefficient of variation in between-year mean annual precipitation; elev., elevation; mean AG biom. 2009, mean aboveground biomass
taken from each site in 2009; mean AG biom. 2007-2009, mean aboveground biomass taken from each site from sampling areas in
2007, 2008, and 2009 (except Colorado, where data are from 2008 and 2009). Bold is used to highlight the matrix diagonal.
Variables
Prop. seeds removed (logit)
Latitude
Longitude
MAP
MAT
AET
Ann. temp. range
Intra-annual MAP CV
Inter-annual MAP CV
Elevation
Mean AG biom. 2009
Mean AG biom. 2007–2009
Prop.
Seeds
rem.
(logit)
Lat.
Long.
MAP
MAT
AET
1.00
–0.43
0.86**
0.65*
0.20
0.83**
0.42
–0.82**
–0.39
–0.37
0.53
0.78**
–
1.00
–0.26
–0.51
–0.51
–0.19
–0.36
0.23
–0.20
–0.21
–0.49
–0.35
–
–
1.00
0.29
0.02
0.92***
0.63*
–0.93***
–0.54
–0.29
0.27
0.57†
–
–
–
1.00
0.49
0.45
–0.05
–0.25
–0.01
–0.48
0.61†
0.60†
–
–
–
–
1.00
0.24
–0.24
–0.01
0.35
–0.53
0.78**
0.52
–
–
–
–
–
1.00
0.35
–0.90***
–0.54
–0.56†
0.51
0.71*
Ann.
temp.
range
–
–
–
–
–
1.00
–0.41
–0.16
0.19
–0.25
–0.07
Intraannual
MAP
CV
Interannual
MAP
CV
Elev.
Mean
AG
biom.
2009
–
–
–
–
–
–
–
1.00
0.58†
0.21
–0.36
–0.68*
–
–
–
–
–
–
–
–
1.00
0.08
–0.17
–0.40
–
–
–
–
–
–
–
–
–
1.00
–0.48
–0.50
–
–
–
–
–
–
–
–
–
–
1.00
0.90***
Mean
AG
biom.
2007–
2009
–
–
–
–
–
–
–
–
–
–
–
1.00
†P  0.10, *P  0.05, **P  0.01, ***P  0.001.
9
Table S4 Summary of the top generalized linear mixed models (binomial distribution) for
describing the proportion of seeds removed [all models are within four AICc units of the best model]
when using climate variables calculated using 2009 data from each site. Models are ordered starting
with the best-fitting model (i.e. the lowest AICc). Values of R2C represent the conditional coefficient
of determination (which includes only fixed effects); values of R2M represent the marginal
coefficient of determination (which includes both fixed and random effects), as described in
Nakagawa & Schielzeth (2013).
Model parameters
Independent variables
Estimate
(± SE)
zvalue
P-value AICc
AICc weight
R2M
R2C
Annual precipitation
0.008 ± 0.001
5.85
< 0.001 64.82
0.64
0.64 0.78
Annual precipitation
0.008 ± 0.001
6.10
< 0.001 67.00
0.22
0.68 0.78
Annual temperature range
0.148 ± 0.091
1.62
0.106
Annual actual
evapotranspiration
0.017 ± 0.003
5.58
< 0.001 67.79
0.14
0.63 0.77
AICc, Akaike information criterion corrected for small sample size.
10
Figure S1 Diagram of a typical Nutrient Network site (Borer et al., 2014). (a) Sites contained
between two and four replicate blocks (Blk). (b) Each block contained ten 5 m × 5 m plots, with
each plot receiving a factorial combination of nutrient treatments and two plots receiving a fencing
treatment to manipulate consumer pressure. Only control plots were used in this experiment (i.e.
plots without any nutrient additions or fencing treatment). Seed-removal depots were placed within
each plot in the areas indicated by an ‘X’.
11
Figure S2 The 11 study sites across North America (see Fig. 1) captured a strong longitudinal
Ann. actual evapotrans. (mm)
gradient in annual actual evapotranspiration (F1,9 = 51.15, r2 = 0.85, P < 0.001).
900
800
700
600
500
400
-120
-110
-100
-90
-80
Longitude
12
Figure S3 Correlograms for the five top models as ranked by AICc; see Table 1. Open circles
indicate that the value was not significant (two-tailed test, alpha = 0.05) using a permutation test
with 1000 replications. Closed circles indicate a value that was significant.
13
14