Study Sites
We sampled 23 sites on three independent latitudinal gradients; each
of which spanned ~10 degrees of latitude and represented a different
major ecoregion of the Great Plains. The West gradient was
characterized by shortgrass prairie and desert grassland, the Middle
gradient was largely mixed-grass prairie, and the East gradient was
tallgrass prairie. Our latitudinal range was comparable to that of
similar studies (e.g., Andrew and Huhges 2004; Adams and Zhang 2009;
Kim 2014) and encompassed substantial climate variation, with mean
annual temperature ranging ~10 C north-south and precipitation varying
~1000 mm east-west (Shafer et al. 2014). Specifically, our study
spanned a gradient of 197 mm to 1001 mm in 30-year normal mean annual
precipitation (MAP) and 2341C to 5837C in mean growing degree days
(GDD). We sampled a total of 23 sites: 11 in the West, and six each in
the Middle and East (see Fig. 1). At the sites CPR, HAR, HPG, KNZ, and
SEV, we sampled a second location within the same landscape from
controls plots of an ongoing experiment (EDGE;
http://edge.biology.colostate.edu/index.html; Appendix Table S1). Most
sites occurred in national or local preserves that had not been grazed
by large vertebrate herbivores. However, sites CAD and DMT were likely
grazed by cattle, although we sampled from locations that had no
evidence of recent grazing so that no herbivory estimates included
cattle damage. Some damage observed in our survey may have been caused
by small rodents or other small vertebrates.
Focal Plant Species
We sampled five perennial C4 grasses: blue grama (Boutelou gracilis)
and buffalograss (B. dactyloides; formerly genus Buchloë), both
ubiquitous in shortgrass prairie; black grama (B. eriopoda), which
dominates desert grasslands; big bluestem (Andropogon gerardii), an
abundant species in tallgrass prairie; and little bluestem
(Schizachyrium scoparium), which is common in both tallgrass and
mixed-grass prairies. B. gracilis and S. scoparium were the most
widely sampled.
Latitudinal Survey
During summer 2015, at each site, we sampled twelve individual plants
per species. We closely examined two haphazardly chosen live, fully
expanded leaves per individual. Because all species were not present
at every site, the number of sites sampled varied among plant species
(see Table S1, Appendix). Individuals were selected as the nearest
plant every 10 m along five transects spaced at 10 m intervals (within
a sampling area of ~50 m x 50 m). For most sites, we used a separate
sampling grid for each grass species due to non-overlapping species
distributions at the local scale. Following standard methods for
herbivory assessment (Pennings et al. 2007), we visually estimated the
percentage of leaf area missing from each of two randomly selected
leaves per plant and we focused our sampling attention on chewing
insect damage. Instead of binning damage estimates into categories
(e.g., 11-25%), as in Pennings et al. (2007), values were recorded as
continuous variation from 0-50% (generally scored to the nearest 5%)
or scored as 75% damage for all leaves damaged by >50%. While this
latter category may have slightly inflated our estimates, only 18
leaves of 1618 were scored as >50% damaged. Maximum damage observed
was ~100%. A consequence of sampling on a large geographic scale at
similar phenology was that multiple observers were required for data
collection. Prior to sampling, all observers calibrated their
estimates of herbivore damage in the field to maintain consistency.
For analysis, we averaged herbivore damage between the two leaves per
individual plant.
To help control for phenological differences among plants at different
latitudes, we sampled all sites at similar growing degree days (GDD)
based on the 30-year climate average (2680 ± 418 s.d. degree days,
using a 0 C base). This ensured that leaves from different sites were
sampled at the same relative age. Sample dates appear in Table S1 in
the Appendix. Per field observations, grasshoppers were a dominant
component of the insect herbivore community in our system.
Grasshoppers can experience periodic outbreaks and vary greatly in
population size over time (Tscharntke and Greiler 1995); thus, it is
possible that results obtained during another year might differ from
our study. We focused sampling effort on coverage of a large
geographic area at the expense of collecting data over multiple
timepoints. However, we had no indications that herbivore abundance
was anomalous in 2015. For the two sites for which we had grasshopper
count data, abundance during 2015 was within 12-13% of the long-term
mean. At the Sevilleta Long Term Ecological Research (LTER) black
grama site in 2015, average grasshoppers per ha was 309 ± 23.2 s.e.,
and the long-term (1992-2015) average was 276 ± 7.5 s.e. At the
Sevilleta LTER blue grama site, mean grasshopper density per ha in
2015 was 357 ± 8.9 s.e. and the long-term (2002-2015) mean was 411 = ±
28.7 s.e.
Abiotic Factors
We examined six abiotic factors as possible correlates of herbivory.
Two were climatic: growing season precipitation and cumulative GDD.
The other four were edaphic: soil nitrogen (as nitrate), phosphorous,
pH, and organic matter (SOM). For precipitation and GDD, we defined
the growing season as March through October. We used a baseline
temperature of 0 C for GDD, as is typical for perennial grasses
(Henebry 2013). We created climate windows for each factor over three
separate time series, allowing us to determine whether variation in
herbivory was best explained by current, short-, or long-term climate
data. We used the year of field sampling (2015), the average of the
three most recent years (2013-2015), or the 30-year average (including
2015). We extracted climate data at the 800m spatial resolution using
the PRISM database (PRISM Climate Group 2016). The other four abiotic
factors were related to edaphic conditions: nitrogen (as nitrate),
phosphorous, pH, and soil organic matter (SOM). We collected soil
samples in situ, taking 10-20g from beneath each plant. Samples were
combined to obtain a single value per edaphic factor for each species
x site combination. Soil phosphorous and pH were determined using
protocols in Robertson et al. (1999). SOM was determined using the
loss on ignition method by Zhang and Wang (2014). Soil ammonium and
nitrate were determined calorimetrically using the Lachat Autoanalyzer
QuikChem method 12-107-06-1-A and 12-107-04-1-F (Loveland, CO).
Plant Traits
We assessed specific leaf area (SLA) and specific root length (SRL) as
possible correlates of herbivory. SLA and SRL are above- and
belowground indicators of resource acquisition trade-offs
(Pérez-Harguindeguy et al. 2013) but are not typically examined in
studies of latitudinal variation in foliar herbivory; high SLA and SRL
indicate high resource acquisition investment and low tissue longevity
(Reich et al. 1992). Traits were measured on the same individuals
sampled in the latitudinal survey following published protocols
(Pérez-Harguindeguy et al. 2013). Whole live plants were pressed in
the field immediately after herbivore damage assessments and two
leaves were used for measurement per individual. For SLA, we
rehydrated the dried leaves that had been stored in a plant press by
placing individual leaf samples in separate, sealed petri dishes with
~100 mL of water. Samples were stored at room temperature during
rehydration period (~48 hours). Rehydrated leaves were scanned and
digitized for total area (cm²) using WinFOLIA (Regent Instruments
Inc., Canada). After measuring leaf area, leaves were oven-dried at 65
C for ~48 h, then weighed for mass. SLA was calculated as rehydrated
leaf area divided by leaf mass (cm²/g). Literature suggests that
measuring SLA for live plants is preferred (Tomaszewski and Górzkowska
2016), but this was not feasible in our study due to the sampling
schedule required to collect data at numerous sites over a large
geographic area and control for phenology. For SRL, a subsample of
fine roots (ca. 10) from each individual (12 in total) from the field
collections were dug up then stored in 50% ethanol. For imaging, roots
were submerged in a small amount of DI water in a clear plastic tray
with individual roots teased apart. Units of total root length were
determined using WinRHIZO (Regent Instruments Inc., Canada). After
imaging, roots were oven-dried at 65 C for ~48 h, then weighed for
specific mass. SRL was calculated as total root length divided by mass
(cm/g).
