Fp

J

D

RA

A

rbcsgdtrraw

©

K

I

t

pD

vS

3

h1

Basic and Applied Ecology 21 (2017) 94–100

unctional dependence underlies a positivelant-grasshopper richness relationship

esse E.D. Miller∗,1,2, Philip G. Hahn1,3, Ellen I. Damschen, John Brennan

epartment of Zoology, University of Wisconsin-Madison, 430 Lincoln Dr, Madison, WI 53706, United States

eceived 17 November 2016; accepted 15 June 2017vailable online 28 June 2017

bstract

A central focus of ecology is identifying the factors that shape spatial patterns of species diversity and this is particularlyelevant in an era of global change. Positive relationships between plant and consumer diversity are common, but could be driveny direct responses of each trophic level to underlying environmental gradients, or indirectly where changes in environmentalonditions propagate through food webs. Here we use structural equation modeling to examine the relative importance ofoil resource availability and disturbance (fire) in mediating relationships between plant and grasshopper richness in insularrasslands. We found a positive relationship between plant and consumer richness that became stronger after accounting foristurbance, despite unique responses of plants and consumers to the two environmental gradients. Plant richness respondedo an underlying gradient in soil resource availability. Time since the last fire had a direct positive effect on grasshopperichness but had no effect on plant richness. This work supports that plant and consumer richness are functionally linked,ather than having similar responses to environmental gradients. By disentangling the direct and indirect processes underlying

positive relationship between plant and consumer diversity in a natural system that spans multiple environmental gradients,

e demonstrate the importance of investigating biodiversity through explicit multivariate models.

2017 Gesellschaft fur Okologie. Published by Elsevier GmbH. All rights reserved.

(s

eywords: Ozarks; Glades; Indirect relationships; Herbivores

ntroduction

A central focus of ecology is determining how environmen-al factors underlie changes in species diversity across space

∗Corresponding author. Fax: +1 5307523350.E-mail addresses: jedmiller@ucdavis.edu (J.E.D. Miller),

hil.hahn@mso.umt.edu (P.G. Hahn), damschen@wisc.edu (E.I.amschen), brennanwisconsin@gmail.com (J. Brennan).1 Authors contributed equally.2 Present address: Department of Environmental Science and Policy, Uni-

ersity of California, Davis, 1 Shields Avenue, Davis, CA 95616, Unitedtates.

3 Present address: Division of Biological Sciences, University of Montana,2 Campus Dr., HS 104, Missoula, MT 59812, United States.

ulad2tftalil

ttp://dx.doi.org/10.1016/j.baae.2017.06.001439-1791/© 2017 Gesellschaft fur Okologie. Published by Elsevier GmbH. All rig

Whittaker 1960; Rosenzweig 1995). Drivers of diversity atingle trophic levels have been well studied, but factors thatnderpin relationships between diversity of multiple trophicevels remain poorly understood. Empirical studies often find

positive relationship between plant diversity and consumeriversity (Castagneyrol et al., 2012; Lewinsohn & Roslin008; Qian & Ricklefs 2008), but the mechanisms that drivehis pattern remain elusive and likely depend upon multipleactors acting in concert (Lewinsohn & Roslin 2008). Posi-ive relationships between plant and consumer diversity couldrise through direct responses of each trophic level to under-

ying environmental conditions (Castagneyrol et al., 2012) orndirectly where environmental conditions affect one trophicevel and then propagate through food webs. For instance,

hts reserved.

J.E.D. Miller et al. / Basic and Applied Ecology 21 (2017) 94–100 95

F eon Ho

ptAtcc

pfidih(tptib(ms

oliR&ricsS

ediici(tacm

tpia(w2agr(af1t(sf

ig. 1. Photograph of one of our dolomite glade study sites, Spurg

lant and consumer diversity could both respond directlyo disturbance (e.g., fire) or other environmental variables.lternatively, if consumers respond directly to plants but not

o an environmental variable, the environment may still affectonsumer diversity indirectly through changes in the plantommunity.

Highly controlled experiments often support that morelant species support more herbivore taxa—i.e., theunctional dependence hypothesis—suggesting that special-zation of consumers on their plant resources is a primaryriver of multitrophic diversity relationships in North Amer-can tallgrass prairies (Haddad et al. 2009). Similar findingsave been reported in experimental European grasslandsScherber et al. 2010), but there have been few attempts toest mechanisms underpinning positive relationships betweenlants and consumers in natural systems where environmen-al variation and disturbance could have strong confoundingnfluences on diversity patterns. Consumer diversity may alsoe related to plant productivity rather than diversity per seSrivastava & Lawton 1998; Joern & Laws 2013), althoughost previous research has focused on, and supported as

ummarized above, diversity-diversity relationships.Resource availability and fire are well-documented drivers

f both plant and arthropod diversity patterns in grass-ands, with both positive and negative—and direct andndirect—effects in different contexts (Cavender-Bares &eich 2012; Joern & Laws 2013; Kirkman, Mitchell, Helton,

Drew 2001; Swengel 2001). Moreover, plant and arthropodesponses to these environmental variables may be linked; fornstance, fire can indirectly influence arthropod diversity by

hanging plant species richness, vegetation structure, or tis-ue quality (Kim & Holt 2012; Van Der Plas et al. 2012).imultaneous direct and indirect effects are also possible; for

1dy

llow, in Shannon County, Missouri. Photo credit: J.E.D. Miller.

xample, fire could directly and negatively affect consumeriversity via increased mortality while simultaneously hav-ng an indirect and positive effect mediated through changesn the diversity or productivity of the plant community. Thus,onsidering only direct effects when examining how plant andnsect communities are structured can lead to spurious resultsGrace 2006). Understanding the mechanisms regulating pat-erns of diversity between multiple trophic levels necessitatesnalytical techniques that fully account for direct and indirectommunity drivers, since these drivers may otherwise maskultitrophic relationships.In this paper we use structural equation modeling to inves-

igate relationships between plant and consumer richness, andlant productivity and consumer abundance, after consider-ng direct and indirect influences of disturbance and resourcevailability. We conducted this study in dolomite gladesrocky grasslands) in the Missouri Ozarks, USA (Fig. 1),hich support a diverse flora and arthropod fauna (Nelson005). Glades are typically embedded in a woodland matrixnd range in size from <1 to >100 ha (Nelson 2005). Theserasslands are a fire-adapted ecosystem, with historic fireeturn intervals ranging from one year to several decadesBatek et al. 1999; Guyette & McGinnes 1982). Grasshoppersre generalist herbivores that often show strong preferencesor certain plant species or families (Chapman & Joern990). They are diverse and abundant in grassland ecosys-ems (Capinera, Scott, & Walker 2005) including gladesHill & Dakin 2011), and regulate many ecosystem processesuch as nutrient cycling (Belovsky & Slade 2000) and trans-erring energy to higher trophic levels (Chapman & Joern

990). Grasshopper abundance can respond to fire, usually byecreasing immediately following a fire and recovering in theears thereafter (Branson & Sword 2010; Joern 2005). Fire

96 J.E.D. Miller et al. / Basic and Applied Ecology 21 (2017) 94–100

Fig. 2. Structural equation model showing relationships between environmental variables and plant and grasshopper richness. Fire historyrepresents time since fire in years, and soil resource availability represents increasing soil organic matter and clay content. Solid lines represents esent nc ly coml

cp2asH

bpmdcrwarvH2gbfi

M

S

i

mssAsomt0tlgwstaoi

P

1Ptf

tatistically significant (P < 0.05) paths, while the dashed lines reproefficients are standardized values, meaning that they can be directines mean stronger relationships.

an kill active grasshoppers and may also destroy grasshop-er eggs laid on plants or in the soil (Branson & Vermeire013; Evans 1984; Swengel 2001). Similarly, grasshopperbundance can be limited by physical properties of the soil,uch as soil moisture, which can effect egg survival (Stauffer,atle, & Whitman 2011).We predicted that a positive relationship would emerge

etween herbaceous plant richness and consumer (grasshop-er) richness because higher herb richness would provideore host plants for herbivores (i.e., support for the functional

ependence hypothesis). Specifically, we hypothesized thatonsumer richness would be more strongly influenced by herbichness than by site productivity, and that consumer richnessould respond more strongly to herb richness than consumer

bundance does. We hypothesized that herb richness wouldespond to fire history and soil resource availability, as pre-ious research has shown in this system (Miller, Damschen,arrison, & Grace 2015) and others (Massad et al. 2013,015; Massad et al., 2013). Finally, we hypothesized thatrasshopper richness would also respond directly to distur-ance history, and indirectly to soil resource availability andre via their effects on plant richness (Fig. 2).

aterials and methods

tudy system and site selection

We selected 15 glades on public conservation landsn southern Missouri for study. We consulted with land

apca

on-significant paths that were removed from the final model. Pathpared. Line thickness is scaled to path coefficients such that thicker

anagers to locate glades that are managed with pre-cribed fire. Prescribed fires occurred during the dormanteason—usually in the later winter in February or March.ll of the glades we sampled have been managed with pre-

cribed fire on approximately 3–6 year intervals, and nonef the glades have experienced any wildfires since manage-ent began. We used management records to calculate the

ime since the most recent fire. Time since fire ranged from to 6 years (mean = 2.2). We also avoided sampling gladeshat are believed to have been intensively grazed by domesticivestock in the past, and none of the study glades had beenrazed by livestock in recent decades. All glades occurredithin mostly natural landscapes. Within our study glades,

oils ranged from clay soils with high organic matter contento less fertile, sandier soils. Our study was conducted during

relatively wet year (129% of average annual precipitationver the last 20 years at NOAA’s weather station on Hwy. 1n Eminence, Missouri).

lant surveys

We sampled herbaceous plant communities using one00 m2 (2 × 50 m) modified Whittaker plot within each glade.lot locations were randomly chosen within a glade, but had

o be more than 5 m from grassland edges and more than 50 mrom roads. Study glades were >500 m apart, in all cases,

nd usually much farther (median pairwise distance betweenlots = 23 km, Q1 = 15 km, Q3 = 107 km). Plant surveys wereonducted twice during the growing season, once in the springnd once in the late summer, to identify both early and late

Applie

bci1d

G

tmmJtope2T(PaWntRU

E

gssSboapoTi(tfyta8ltraa

Niv(

S

vadebvta

htihantemtafv2emrs(

acaWataggnnd

J.E.D. Miller et al. / Basic and

looming species. We combined data from these visits toonstruct a master species list for each glade. Plants weredentified according to the Flora of Missouri (Yatskievych999, 2006, 2013). See Miller et al. (2015) for additionaletails on glades and plant surveys.

rasshopper sampling

We sampled grasshoppers (families Acrididae and Tet-igoniidae) using sweep nets. Sweep net sampling is the

ost common method used to measure grasshopper com-unity composition (Belovsky & Slade 1995; Evans 1984;

oern 2005). We sampled each glade by vigorously sweepinghrough vegetation while walking a transect along the edgef the 2 × 50 m plot (hereafter “transect”) for a 10 minuteeriod. This resulted in approximately 3–4 passes alongach transect. Sampling was conducted from 12–15 August013, between 10:00–17:00 during sunny and calm weather.ime of day did not affect total grasshopper abundanceF1,12 = 0.8, P = 0.38) or grasshopper richness (F1,12 = 0.3,

= 0.62). Grasshoppers were stored on ice immediatelyfter collection and kept frozen until they were identified.e identified grasshoppers to species using regional taxo-

omic guides (Capinera et al. 2005). Nymphs were identifiedo genus. Voucher specimens were deposited in the Insectesearch Collection of the Department of Entomology at theniversity of Wisconsin-Madison.

nvironmental variables

We collected soil samples on the first (spring) visit to eachlade. We used a trowel to dig samples of the top 15 cm ofoil from six evenly spaced locations along the 50 m tran-ect which were combined and analyzed at the glade level.oils were analyzed for texture and organic matter contenty Brookside labs in New Knoxville, Ohio. We chose soilrganic matter and clay content to represent soil resourcevailability because these variables are known to be strongredictors of plant species richness in Ozark glades as well asther systems (Miller et al. 2015; Miller & Damschen 2017).hese variables are correlated with soil water-holding capac-

ty, which can be a strong predictor of plant diversity patternsBrady & Weil 2007; Kirkman et al. 2001). Soil organic mat-er ranged from 3.4 to 9.3%, and soil clay content rangedrom 2 to 33%. We performed a principal component anal-sis (PCA) of soil organic matter and clay content and usedhe first axis from this PCA to represent soil resource avail-bility in the model in the present study. This axis explained5% of variation in soil organic matter and clay content (axisoadings: 0.82 with organic matter and 0.57 with square-root

ransformed clay content). We used time since fire to rep-esent fire history because it has been shown to affect plantnd arthropod diversity in grasslands (Joern 2005) and it wasvailable for all of our study glades.

R

g

d Ecology 21 (2017) 94–100 97

To estimate glade productivity, we downloaded 4-bandational Agricultural Imagery Program (NAIP) landscape

magery from the Missouri Spatial Data Information Ser-ice. We calculated the Normalized Daily Vegetation IndexNDVI) in ArcGIS 10 (ESRI 2011).

tatistical analysis

We began analysis by examining correlations amongariables of interest (grasshopper richness, grasshopperbundance, plant richness, NDVI [representing glade pro-uctivity], time since fire, and soil resource availability). Wexamined variable distributional properties and linearity ofivariate relationships. To meet assumptions of homogenousariances and normal distributions, we log-10 transformedime since fire and square-root transformed soil clay contentnd grasshopper abundance.

We used structural equation modeling (SEM) to test ourypotheses about multivariate controls on multitrophic rela-ionships. We began by constructing a meta-model thatncluded all hypothesized relationships (Fig. 2); this modelad two exogenous variables (time since fire and soil resourcevailability) and two endogenous variables (herb species rich-ess and grasshopper species richness). We included onlywo variables—soil resource availability and fire history—asxternal influences on plant and grasshopper richness to keepodel complexity appropriate for the size of our dataset per

he recommendations of Grace (2006). Resource availabilitynd fire are known to be two of two of the most importantactors that affect both plant and grasshopper richness in aariety of systems including the Ozark glades (Massad et al.013; Miller et al. 2015). Following procedures for hypoth-sis testing with SEM described by Grace (2006), we thenodified the meta-model to create a parameterized model by

emoving non-significant paths. We performed SEM analy-es in the lavaan package (Rosseel 2012) in R version 3.1.2R Core Team 2016).

To compare the relative influence of plant productivitynd species richness on the grasshopper community, and toompare the relative responses of grasshopper abundancend richness, we performed three additional SEM analyses.e used the same structure as the meta model described

bove, except for the following changes: (1) Glade produc-ivity (NDVI) replaces herb species richness, (2) grasshopperbundance replaces grasshopper species richness, and (3)lade productivity (NDVI) replaces herb species richness andrasshopper abundance replaces grasshopper species rich-ess. We present the complete models without removingon-significant paths, though removing non-significant pathsoes not qualitatively change the results.

esults

We found 145 total herb species across all glades, withlade-level richness ranging from 14 to 62 species. We found

9 Applied Ecology 21 (2017) 94–100

arraGr

(eGfin4sisbwSrSphpda

D

pdsschtaaisamwotprtmt

pop

Fig. 3. Scatterplots showing (A) the relationship between fire his-tory and grasshopper richness after accounting for the effects ofplant species richness; (B) the relationship between plant speciesrichness and grasshopper richness after accounting for the effects offire; and (C) the relationship between soil resource availability andplant species richness. Grasshopper richness values are residualsfrom a model with only plant richness as a predictor (A) and onlyfire as a predictor (B). Time since fire and soil resource availabilityare in transformed units. Herb species richness is in untransformedunits.

8 J.E.D. Miller et al. / Basic and

total of 16 grasshopper species (Table A.1) and grasshopperichness at the glade level ranged from 3 to 15. Grasshopperichness and abundance were correlated (R = 0.48, P = 0.044),s were plant richness and NDVI (R = 0.71, P = 0.003).rasshopper richness and herb richness were marginally cor-

elated (R = 0.45, P = 0.089).Our final SEM provided an adequate fit to the data

χ2 = 1.574, model P = 0.455, model df = 2; Fig. 2) andxplained 56% of the variation in grasshopper richness.rasshopper richness increased with increasing time sincere (path coefficient = 1.33; Fig. 3A) and increasing herb rich-ess (path coefficient = 1.02; Fig. 3B). The SEM explained6% of variation in herb richness; herb richness increased asoil resource availability (clay and organic matter content)ncreased (path coefficient = 0.68; Fig. 3C). The hypothe-ized paths between fire history and herb species richness andetween soil resource availability and grasshopper richnessere not significant, so we removed them for the final model.oil resource availability indirectly affected grasshopperichness via plant richness (Fig. 2). The three supplementalEMs showed that glade productivity was not a significantredictor of grasshopper richness or abundance, and thaterb richness was not a significant predictor of grasshop-er abundance (Fig. A.1). These supplemental SEMs alsoid not explain as much variation in grasshopper richness orbundance as did the main model (10–42%; Fig. A.1).

iscussion

Our results show that a positive relationship betweenlant and consumer (grasshopper) diversity is driven throughirect effects of plant diversity and disturbance (fire) on con-umer diversity, as well as indirect effects of changes inoil resources, which affected consumer diversity throughhanges in plant diversity. Specifically, our model rejects theypothesis that both plants and consumers respond directlyo the same environmental variables (resource availabilitynd fire). Rather, the bottom-up influence of soil resourcevailability on plants affected grasshopper species richnessndirectly. Similarly, grasshopper richness responded to timeince fire whereas plant richness did not, indicating that plantsnd grasshoppers respond to different underlying environ-ental gradients. Although glade productivity was correlatedith herb species richness, it did not have a significant effectn either grasshopper richness or abundance, indicating thatotal plant biomass is likely not a limiting factor for grasshop-er diversity or abundance in glades. The importance of plantichness but not glade productivity for grasshoppers suggestshat the presence of particular plant species that occur only in

ore productive glades may drive the grasshopper diversityhere.

The mechanisms generating functional linkages betweenlant and grasshopper richness could arise through the effectsf host plant availability, effects of structural or habitat com-lexity provided by plant diversity, or both. For instance,

Applie

gaddcbi11patte

getdrmrclehoftht

iiievssehr(eyshf

A

ara

aHpafpaNotN

A

ch

R

B

B

B

C

C

C

D

D

E

E

J.E.D. Miller et al. / Basic and

reater plant diversity could support a more diverse insectssemblage (Dyer et al. 2007; Novotny et al. 2006) becauseifferent grasshopper species exhibit substantially differentiets (Joern 1979). Changes in structural complexity asso-iated with increased plant richness could alter grasshopperehavior because grasshoppers spend much of their time rest-ng, thermoregulating (Belovsky & Slade 1986; Joern et al.986), or avoiding predators (Schmitz, Beckerman, & Brien997). Therefore it is possible that increasing structural com-lexity could increase grasshopper richness (Joern 2005) byccommodating more behavioral niches. We were unableo distinguish between the mechanisms driving these func-ional linkages, and this remains an area that is ripe for futurexperimental tests.

In our study, the positive relationship between plant andrasshopper richness was partially masked before the influ-nce of fire was controlled for statistically. By identifyinghe major direct and indirect drivers of plant and consumeriversity, our results highlight the importance of studyingelationships between the plant and consumer richness in aultivariate context where environmental variables—such as

esource availability and disturbance history—are explicitlyonsidered. Future experiments should manipulate the under-ying gradients, in addition to plant diversity, and measure theffects on the diversity of higher trophic levels. Research ineterogeneous environments could elucidate other nuancesf diversity relationships across trophic levels. In addition,uture studies should work to disentangle reciprocal rela-ionships between richness at multiple trophic levels, sinceerbivory may feedback to affect plant community composi-ion (Deraison, Badenhausser, Börger, & Gross 2015).

Maintaining biodiversity is a primary conservation prior-ty in an era of rapid biodiversity loss and anthropogenicnfluences (e.g., grazing domestic animals) can weaken pos-tive multitrophic relationships (Klink et al. 2015; Manningt al. 2015). Management efforts often focus on restoringegetation structure and diversity; our results support thatuch management approaches may provide a foundation forupporting biodiversity of organisms at higher trophic lev-ls (Hahn & Orrock 2015; Panzer 2002). Previous researchas found that prescribed fire can be an important tool forestoring and maintaining glade plant and animal diversityTempleton, Robertson, Brisson, & Strasburg 2001). How-ver, because grasshopper communities appear to take severalears to recover after fire, providing unburned refuges for fire-ensitive organisms within managed glade landscapes mightelp maximize the benefits that glade communities receiverom prescribed fire.

cknowledgements

We thank the numerous Ozark ecologists and land man-gers who assisted with glade selection and provided fireecords. We also thank Justin Thomas for providing expertssistance with plant identifications, and numerous field

G

H

d Ecology 21 (2017) 94–100 99

ssistants for helping with surveys. Comments from Nickaddad greatly improved the manuscript. James Gracerovided guidance on structural equation modeling. Wecknowledge the IAN/UMCES Symbol and Image Librariesor providing clipart. This material is based upon work sup-orted by National Science Foundation (NSF) DEB-0947432nd the NSF Graduate Research Fellowship under Granto. 2012149884. Any opinion, findings, and conclusionsr recommendations expressed in this material are those ofhe authors and do not necessarily reflect the views of theational Science Foundation.

ppendix A. Supplementary data

Supplementary data associated with this arti-le can be found, in the online version, atttp://dx.doi.org/10.1016/j.baae.2017.06.001.

eferences

elovsky, G. E., & Slade, J. B. (1986). Time budgets of grass-land herbivores: Body size similarities. Oecologia, 70(1), 53–62.http://dx.doi.org/10.1007/BF00377110

elovsky, G. E., & Slade, J. B. (1995). Dynamics of two Montanagrasshopper populations: Relationships among weather, foodabundance and intraspecific competition. Oecologia, 101(3),383–396. http://dx.doi.org/10.1007/BF00328826(/DOI

rady, N. C., & Weil, R. R. (2007). The nature and properties ofsoils (14th ed.). Pearson.

apinera, J. L., Scott, R. D., & Walker, T. J. (2005). Field guide tograsshoppers, katydids, and crickets of the United States. NewYork, USA: Cornell University Press.

astagneyrol, B., Jactel, H., & Cardinale, B. J. (2012).Unraveling plant-animal diversity relationships: Ameta-regression analysis. Ecology, 93(9), 2115–2124.http://dx.doi.org/10.1890/11-1300.1

avender-Bares, J., & Reich, P. (2012). Shocks to the system:Community assembly of the oak savanna in a 40-year fire fre-quency experiment. Ecology, 93(8), 52–69. Retrieved from:http://www.esajournals.org/doi/full/10.1890/11-0502.1

eraison, H., Badenhausser, I., Börger, L., & Gross, N. (2015). Her-bivore effect traits and their impact on plant community biomass:An experimental test using grasshoppers. Functional Ecology,29(5), 650–661. http://dx.doi.org/10.1111/1365-2435.12362

yer, L. A., Singer, M. S., Lill, J. T., Stireman, J. O., Gentry, G. L.,Marquis, R. J., . . . & Coley, P. D. (2007). Host specificity of Lep-idoptera in tropical and temperate forests. Nature, 448(7154),696–699. http://dx.doi.org/10.1038/nature05884

SRI. (2011). ArcGIS desktop: Release 10. Redlands, CA: Environ-mental Systems Research Institute.

vans, E. W. (1984). Fire as a natural disturbance to grasshop-per assemblages of tallgrass prairie. Oikos, 43(1), 9–16.http://dx.doi.org/10.2307/3544239

race, J. B. (2006). Structural equation modeling and natural sys-tems. Cambridge: University Press.

ahn, P. G., & Orrock, J. L. (2015). Land-use history alters contem-porary insect herbivore community composition and decouples

1 Applie

J

J

J

J

K

K

M

M

M

M

N

P

R

R

R

S

S

T

V

W

Y

Y

00 J.E.D. Miller et al. / Basic and

plant-herbivore relationships. Journal of Animal Ecology, 84(3),745–754. http://dx.doi.org/10.1111/1365-2656.12311

oern, A. (1979). Feeding patterns in grasshoppers (Orthoptera:Acrididae): Factors influencing diet specialization. Oecologia,38(3), 325–347.

oern, A. (2005). Disturbance by fire frequency and bison grazingmodulate grasshopper assemblages in tallgrass prairie. Ecology,86(4), 861–873. http://dx.doi.org/10.1890/04-0135

oern, A., & Laws, A. N. (2013). Ecological mech-anisms underlying arthropod species diversity ingrasslands. Annual Review of Entomology, 58(1), 19–36.http://dx.doi.org/10.1146/annurev-ento-120811-153540

oern, A., Mitschler, R., & O’Leary, H. (1986). Activity andtime budgets of three grasshopper species (Orthoptera:Acrididae) from a sandhills grassland. Journal of theKansas Entomological, 59(1), 1–6. Retrieved from:http://www.jstor.org/stable/10.2307/25084728

im, T. N., & Holt, R. D. (2012). The direct and indirect effects offire on the assembly of insect herbivore communities: Examplesfrom the Florida scrub habitat. Oecologia, 168(4), 997–1012.http://dx.doi.org/10.1007/s00442-011-2130-x

irkman, L. K., Mitchell, R. J., Helton, R. C., & Drew, M. B. (2001).Productivity and species richness across an environmental gradi-ent in a fire-dependent ecosystem. American Journal of Botany,88(11), 2119–2128. http://dx.doi.org/10.2307/3558437

anning, P., Gossner, M. M., Bossdorf, O., Allan, E., Zhang,Y.-Y., Prati, D., . . . & Fischer, M. (2015). Grasslandmanagement intensification weakens the associationsamong the diversities of multiple plant and animal taxa.Ecology, 96(November), 1492–1501. Retrieved fromhttp://www.esajournals.org/doi/abs/10.1890/14-1307.1

assad, T. J., Balch, J. K., Davidson, E. A., Brando, P. M., Mews, C.L., Porto, P., . . . & Trumbore, S. E. (2013). Interactions betweenrepeated fire, nutrients, and insect herbivores affect the recov-ery of diversity in the southern Amazon. Oecologia, 172(1),219–229. http://dx.doi.org/10.1007/s00442-012-2482-x

iller, J. E. D., & Damschen, E. I. (2017). Biological soil crust coveris negatively related to vascular plant richness in Ozark sand-stone glades. The Journal of the Torrey Botanical Society, 144(2),170–178. http://dx.doi.org/10.3159/TORREY-D-15-00076

iller, J. E. D., Damschen, E. I., Harrison, S. P., & Grace, J.B. (2015). Landscape spatial structure affects specialists butnot generalists in naturally fragmented grasslands. Ecology, 96,3323–3331. http://dx.doi.org/10.1890/15-0245.1

Y

Available online at www.s

ScienceD

d Ecology 21 (2017) 94–100

ovotny, V., Drozd, P., Miller, S. E., Kulfan, M., Janda, M., Basset,Y., & Weiblen, G. D. (2006). Why are there so many species ofherbivorous insects in tropical rainforests? Science, 313(5790),1115–1118. http://dx.doi.org/10.1126/science.1129237

anzer, R. (2002). Compatibility of prescribed burning withthe conservation of insects in small, isolated prairiereserves. Conservation Biology, 16(5), 1296–1307.http://dx.doi.org/10.1046/j.1523-1739.2002.01077.x

Core Team. (2016). R: A language and environment for statis-tical computing. Vienna, Austria: R Foundation for StatisticalComputing. Retrieved from:. http://www.r-project.org/

osenzweig, M. L. (1995). Species diversity in space and time.Cambridge: University Press.

osseel, Y. (2012). lavaan: An R package for structural equationmodeling. Journal of Statistical Software, 48(2), 1–36.

chmitz, O. J., Beckerman, A. P., & Brien, K. M. O. (1997). Behav-iorally mediated trophic cascades: Effects of predation risk onfood web interactions. Ecology, 78(5), 1388–1399.

wengel, A. B. (2001). A literature review of insect responsesto fire, compared to other conservation managements of openhabitat. Biodiversity and Conservation, 10(7), 1141–1169.http://dx.doi.org/10.1023/A:1016683807033

empleton, A. R., Robertson, R. J., Brisson, J., & Stras-burg, J. (2001). Disrupting evolutionary processes: The effectof habitat fragmentation on collared lizards in the Mis-souri Ozarks. Proceedings of the National Academy ofSciences of the United States of America, 98(10), 5426–5432.http://dx.doi.org/10.1073/pnas.091093098

an Der Plas, F., Anderson, T. M., & Olff, H. (2012). Trait similaritypatterns within grass and grasshopper communities: Multi-trophic community assembly at work. Ecology, 93(4), 836–846.http://dx.doi.org/10.1890/11-0975.1

hittaker, R. (1960). Vegetation of the SiskiyouMountains, Oregon and California. EcologicalMonographs, 30(3), 279–338. Retrieved from:http://www.esajournals.org/doi/abs/10.2307/1943563

atskievych, G. (1999). (2nd ed.). Steyermark’s Flora of Missouri(Vol. 1) Jefferson City and St. Louis: Missouri Department ofConservation and Missouri Botanical Garden.

atskievych, G. (2006). (2nd ed.). Steyermark’s Flora of Missouri

(Vol. 2) St. Louis: Missouri Botanical Garden Press.

atskievych, G. (2013). (2nd ed.). Steyerkmark’s Flora of Missouri(Vol. 3) St. Louis: Missouri Botanical Garden Press.

ciencedirect.com

irect