Functional consequences of realistic extinction scenarios in Amazonian soil food webs

32 Pages·2017·4.66 MB·English

Functional consequences of realistic extinction
scenarios in Amazonian soil food webs
RAFAELE.CA(cid:1)RDENAS,1,2,3,(cid:1)DAVIDA.DONOSO,4ADRIANAARGOTI,1ANDOLIVIERDANGLES1,2,3
1PontificiaUniversidadCat(cid:1)olicadelEcuador,EscueladeCienciasBiol(cid:1)ogicas,MuseodeZoolog(cid:1)ıaQCAZ,
LaboratoriodeEntomolog(cid:1)ıa,Av.12deoctubre1076yRoca,Apdo.17-01-2184,Quito,Ecuador
2InstitutdeRecherchepourleD(cid:1)eveloppement(IRD),UR072,LEGS-CNRS,UPR9034,CNRS,Gif-sur-Yvette,Cedex91198France
3Universit(cid:1)eParis-Sud11,Orsay,Cedex91405France
4EscuelaPolit(cid:1)ecnicaNacional,InstitutodeCienciasBiol(cid:1)ogicas,Av.Ladr(cid:1)ondeGuevaraE11-253,Quito,Ecuador
Citation:C(cid:1)ardenas,R.E.,D.A.Donoso,A.Argoti,andO.Dangles.2017.Functionalconsequencesofrealisticextinction
scenariosinAmazoniansoilfoodwebs.Ecosphere8(2):e01692.10.1002/ecs2.1692
Abstract. Global biodiversity loss is creating a more urgent need to understand the role organisms
play in ecosystem functioning and mechanisms of control. Decomposition of dead organic matter is a
key ecological process that ensures soil formation, nutrient availability, and carbon sequestration. To
gain understanding of how biodiversity and ecosystems function together to control leaf-litter decom-
position processes in a tropical rain forest (Yasun(cid:1)ı National Park, Ecuador), we predicted the conse-
quences of the decomposition process using a protocol in which we systematically disassemble the
structural functionalityof the soil macrofauna communities. We (1) describe the structure and function
of the edaphic communities in detail and (2) explore the functional consequences of structural changes
in these communities using a non-random exclusion experiment to simulate body size-related extinc-
tions. To do this, we manipulated access of five size classes of soil invertebrates to eight types of plant
leaf-litter resources. After measuring and identifying about 4400 soil individuals belonging to 541 mor-
phospecies, 12 functional groups, and following the fate of about 2000 tree leaves in a 50-ha plot, we
showed that (1) soil invertebrate communities were composed of a few common and many rare mor-
phospecies that included mostly leaf-litter transformer groups, with the most morphospecies and the
greatest abundance coming from Hymenoptera, Collembola, and Coleoptera; (2) our survey captured
63–74%ofthetotalsoilbiodiversityofthestudyarea(meaningtheremaybeupto860 morphospecies);
(3)littertransformerscoveredthewidestrangeofbodyvolume,andallgroupswereevenlydistributed
at small and large spatial scales (i.e., we found no patterns of spatial aggregation); (4) changes in food
webstructuresignificantlyalteredbiomasslossforonlythreeoftheeightleaf-littertreatments,suggest-
ing the decomposition process was highly resistant to drastic changes such as size-biased biodiversity
loss independent of resource quality. We conclude organic matter decomposition may depend on all
non-additive effects that arise from multi-species interactions, including facilitation, interspecific
interference competition, and top-down control that predators exert over detritivores at all body size
ranges.
Key words: belowground; biodiversity and ecosystem functioning; brown food webs; decomposers; Ecuador;
extinctionorder;invertebrates;leaf-litterdecomposition;Yasun(cid:1)ıNationalPark.
Received4December2016;accepted12December2016.CorrespondingEditor:DebraP.C.Peters.
Copyright:©2017C(cid:1)ardenasetal.ThisisanopenaccessarticleunderthetermsoftheCreativeCommonsAttribution
License,whichpermitsuse,distributionandreproductioninanymedium,providedtheoriginalworkisproperlycited.
(cid:1)E-mail: [email protected]
❖ www.esajournals.org 1 February2017 ❖ Volume8(2) ❖ Articlee01692
(cid:1)
CARDENASETAL.
I
NTRODUCTION the role of invertebrate fauna (primarily detriti-
vores)onleaf-litterdecompositioninbothaquatic
Biodiversity loss directly disrupts ecosystem and terrestrial habitats (Lavelle et al. 2006, Gess-
functioning, undermining ecosystem servicesand neret al.2010,Garc(cid:1)ıa-Palacioset al.2016).Detriti-
ultimatelyaffectinghumanwell-being(D(cid:1)ıazet al. vore biodiversity has proven to be critical to the
2006, Cardinale et al. 2012). Biodiversity and biogeochemical and ecological functioning of ter-
ecosystem functioning research (BEF) seeks to restrial ecosystems, with consequences for fertil-
determine how species diversity is related to the ity, plant growth, environmental structure, and
magnitude and stability of ecosystem processes carbonstorage(Brussaard1998).
(Griffinet al.2009,Hooperet al.2012).Thereisa We know little about how loss of detritivore
great deal of ecological debate about three tenets biodiversity affects leaf-litter decomposition and
of BEF, which are that biodiversity is likely to (1) otherecosystemprocesses,particularlywhenspe-
improveproductivity(Cardinaleet al.2007,Hoo- cies in a focal community differ in key functional
per et al. 2012), (2) increase ecosystem stability attributes (Bardgett and Wardle 2010, Wall et al.
(Tilman et al. 2006), and (3) enhance the magni- 2010).Thereisampleevidencethatsoilcommuni-
tude of a variety of ecosystem processes (Hooper ties are adapted to specific environmental condi-
et al. 2005, Balvanera et al. 2006, Wagg et al. tions and resource types; therefore, any changes
2014).However,theeffectsofchangesinbiodiver- in either factor could negatively affect species
sity on ecosystem functioning may vary across richness,withpotentialtoseverelyimpairecosys-
ecosystem types and study groups, and depend tem functioning (Wall and Nielsen 2012, Lavelle
on the trophic relationships involved (Huston et al.2016).Arecentglobalreviewofexperiments
1997, Smith and Knapp 2003, Wardle et al. 2008, that have explored the relationship between car-
Schmid et al. 2009). This has raised the question: bon cycling and soil biodiversity concluded that
Does the relationship between biodiversity and although species richness, on average, led to bet-
ecosystem functioning apply in the real world ter ecosystem functioning—measured as greater
(acrosstrophiclevels),outsideofcontrolledexper- biomass, decomposition rates, and/or respiration,
iments? To answerthis, new studies mustlook at especially in species-poor communities—the rela-
theentirefoodwebwithinanecosystem(Tyliana- tionship was neither linear nor redundant (Niel-
kiset al.2008,Thompsonet al.2012).Futurecon- sen et al. 2011). However, Handa et al. (2014)
servation strategies require an understanding of conducted a first concerted set of experiments
the reciprocal nature of relationships between acrossfiveterrestriallocations(fromthesubarctic
foodwebstructureandthefunctioningofagiven tothetropics),andfoundthatfunctionalbiodiver-
ecosystem (Thompson et al. 2012, Poisot et al. sitylossofbothsoilfaunaandlittertypesslowed
2013). cyclingoflittercarbonandnitrogen.Theysuggest
Decomposition of dead organic matter (OM), that documented differences in the effects of bio-
categorized as a supporting service inthe Millen- diversitylossondecompositionacrossglobalspa-
nium Ecosystem Assessment (2005), is a key tial scales mayat least partlyarise from variation
ecosystem function that ensures soil formation, in experimental protocols, plant species, and
nutrientavailabilityforplants,andcarbonseques- types of decomposers studied in a given experi-
tration (Swiftet al. 1979, Chapin et al. 2002).OM ment. Moreover, in ecosystems with high natural
processesareinfluencedbyfactorssuchasclimate community evenness, stress or disturbances are
(Wall et al. 2008, Powers et al. 2009),the physical unlikely to impair all species’ contributions to an
and chemical properties of dead OM (Kaspari ecosystem service, making these communities
€
et al. 2008, Hattenschwiler et al. 2011), the more resistant and/or resilient to disturbances
(cid:1)
sequential action of soil invertebrates, fungi, and (Andrenet al.1995).Communitieswithloweven-
bacteria(Inghamet al.1985,Pramaniket al.2001, nessmaylosedominantspeciesduetostressora
Mulder2006),andthefunctionaldiversityofboth disturbance. Dominantspeciestendtoplaya sig-
dead OM (e.g., plant litter) and soil consumers nificant role in local ecosystem processes (the
€
(Hattenschwiler and Gasser 2005, Gessner et al. mass ratio hypothesis states that ecosystem pro-
2010, Dangles et al. 2012, Handa et al. 2014). A cesses are determined by the functional traits of
significant part of BEF research has focused on the dominant species; Grime 1998, Smith and
❖ www.esajournals.org 2 February2017 ❖ Volume8(2) ❖ Articlee01692
(cid:1)
CARDENASETAL.
Knapp 2003), and their loss impacts ecosystem loss on ecosystem stability and functioning. One
functioning(Dangleset al.2004,WallandNielsen potential surrogateisbodysize (aproxytobody
2012). Understanding detritivore community massandbodyvolume).Bodysizeisakeyfunc-
functionalstructureundernaturalsettingswillbe tionaltraitofspeciesthatiscorrelatedwithmany
key to understanding relationships between bio- life-history characteristics, and therefore, it is a
diversity and ecosystem functioning in complex, goodsurrogateforalargeamountofthebiologi-
real-worldsystems. cal information embedded within an ecological
(cid:1)
Cardenas and Dangles (2012) conducted an network (Emmerson and Raffaelli 2004, Wood-
experiment using mesh bags to exclude both ward et al. 2005, Reiss et al. 2011). Body size
macro- and meso-detritivoresfromleaflitter, and could reflect the mass-dependent metabolic
foundabouta50%declineinleaf-litterdecompo- needs of an individual or a species community,
sition rates in an Ecuadorian cloud forest. Using predictingtheimpactofadrasticchangeinnatu-
the similar methodology to exclude macro- ral abundance of either on a given ecosystem
detritivores, Coq et al. (2010) found a 17.4% functioning (Reiss et al. 2009). Size-dependent
decline in leaf-litter decomposition rates in consumption and processing rates appear to be
French Guiana and Yang and Chen (2009) found linked to energy flow (Dossena et al. 2012, Lang
a40%declineintropicalChina.Althoughmacro- et al. 2014) and may even compromise stability
and meso-detritivore fauna have been shown to of complex food webs (Otto et al. 2007). This is
playacrucialroleinfragmentingdeadOMinthe extremelyimportantgiventhatlargerspeciesare
tropics(Swiftet al.1979),studiesthathavemanip- especiallyvulnerableto environmental perturba-
ulated the effects of detritivore diversity loss on tions such as global warming and changes in
leaf-litter decomposition in real field conditions precipitation patterns (Salazar et al. 2007,
(e.g., non-random extinction scenarios; Gross and Smith et al. 2009, Sheridan and Bickford 2011),
Cardinale2005)arestillveryscarce(Schmidet al. habitat fragmentation (Klein 1989), or land use
2009).Furthermore,factorssuchaslanduse,nitro- (McCracken and Bignal 1998). Moreover, recent
gen enrichment, acidification, and climate change research has shown that smaller species are not
havebeenreportedtoaltersoilandstreamdetriti- simply miniature copies of larger ones (due in
vorediversity(Gessneret al.2010). part to mass-specific metabolic constraints, Reiss
Experiments that manipulate invertebrate et al. 2011), suggesting that a range of animal
diversity in natural ecosystems are extremely size classes is needed to maintain ecosystem
importantforunderstandingtheconsequencesof functioning(Dangleset al.2012).Inaddition,lar-
potential extinctions on decomposition and ger-bodied invertebrates can directly or indi-
nutrient cycling. However, one specific problem rectly influence the diversity of smaller-sized
in soil biology, and particularly in the tropics, is organisms by promoting dispersal and modify-
with how to deal with diverse and complex ing the soil habitat (Wardle 2006). For example,
groups such as soil fauna (Giller 1996)—espe- fragmentation by large detritivores could facili-
cially in these megadiverse systems where most tate the ingestion/colonization of OM particles
species (>80% of all invertebrates of tropical for- by smaller detritivores (including microflora;
ests) have yet to be described by science and Lavelleet al.1997,Jonssonet al.2002,Yanget al.
almostnothingisknownoftheremainder’secol- 2012). Or, larger detritivores, due to their needs
ogy (Primack and Corlett 2005, Wall et al. 2010). to grow and higher mobility in the forest floor
Ecosystem processes are a product of multiple (e.g., giant annelids or cockroaches), could mod-
biological and environmental variables (Petchey ify the accessibility to phosphorous, which may
et al.1999),underscoringtheneedformorereal- decreaselocallyandinthelongtermthebiomass
istic experiments under natural conditions that of small invertebrates relative to the larger ones
considernon-randomlossofassemblagesinlocal (Mulder and Elser 2009). Finally, documenting
extinctions (e.g., Zavaleta and Hulvey 2004, thedifferentialcontributionoffaunalsizeclasses
O’ConnorandCrowe2005). may help demonstrate how much redundancy
Community ecologists need tractable metrics andspecializationisactuallyfoundinsoildetriti-
€€
that can serve as surrogates of interaction vore communities (Setala et al. 2005, Bezemer
strength to better evaluate effects of biodiversity et al.2010).Size-biased,non-randomspeciesloss
❖ www.esajournals.org 3 February2017 ❖ Volume8(2) ❖ Articlee01692
(cid:1)
CARDENASETAL.
therefore has important implications for the risk 2003) to understand the potential implications of
of cascading secondary extinctions and the loss detritivore fauna extinction on the leaf-litter
(cid:1)
offunctionaldiversityfromecosystems(Soleand decompositionprocessinthisecosystem.
Montoya2002,Ebenmanet al.2004). Here, we evaluate six hypotheses: (1) Amazo-
Using the megadiverse invertebrate soil com- nian soil invertebrate fauna biodiversity is domi-
munities of Yasun(cid:1)ı National Park (northwestern nated by the leaf-litter transformer (detritivore)
Amazonia, Ecuador) as a model system, our functional group (in number of species, abun-
studyaimed toanswer two general questions: (1) dance, and biovolume); (2) leaf-litter transformer
Whatistheoverallstructuralandfunctionalorga- species encompass the largest range of body vol-
nizationofsoilinvertebratecommunitiesinterms umefoundintheforestfloor;(3)therearenopat-
of taxa diversity, relative abundance, functional terns of spatial aggregation, meaning that soil
traits (based on trophic guilds, and biovolume as functional diversity does not increase with
a proxy of biomass), and the diversity–area rela- increased area based on the fact that most plant
tionship? (2) How does non-random extinction species in Yasun(cid:1)ı follow the negative density
within the soil invertebrate food web affect the dependence hypothesis (Metz et al. 2010), and
leaf-litter decomposition process, and does such therefore, they show a “random” distribution
aneffectdependonresourcequality?Theseques- withinour50-haplot(Valenciaet al.2004),which
tions arise from (1) the “mass ratio hypothesis” should be reflected by a leaf-litter diversity (and
(cid:1)
that states that ecosystem processes are over- quality) homogeneity in this forest (Cardenas
whelminglydeterminedbythefunctionaltraitsof et al. 2014, not shown); (4) leaf litter exposed to
the dominant species. These dominant species thesmallerfoodwebcommunitymaydecompose
may contribute most of the biomass, which at significantly lower rates than that exposed to
actively controls fluxes of energy and matter larger soil fauna; (5) smaller species do not com-
through the ecosystem (Grime 1998, Smith and pensate for the loss of the larger ones in terms of
Knapp 2003). Therefore, we first need a detailed leaf-litter transformation (there is a positive rela-
assessment of the relationship between species tionshipbetweendecompositionratesanddetriti-
diversity,abundance,andbodymassofthediffer- vore diversity), suggesting that some size classes
ent functional groups present in the soil fauna of facilitatetheeffectivenessofothersandthatdetri-
ourstudyarea—anelementaryfirststeptounder- tivore extinction may disrupt the decomposition
standing ecosystem functioning and the role and process in this ecosystem; and (6) the effect of
impactofspeciesinthefoodchain(Jonssonet al. non-random soil fauna extinction on leaf-litter
2005); (2) larger soil species in the Amazon are decompositionisindependentofresourcequality.
more prone to extinction in the current climate
M M
changeconditions(Salazaret al.2007,Smithet al. ATERIALS AND ETHODS
2009,SheridanandBickford2011).
Given how little is known about soil biodiver- Studysite
sity and functional diversity in Amazonian tropi- Yasun(cid:1)ı National Park (YNP) and the adjacent
calecosystems(PrimackandCorlett2005,Moreira WaoraniIndigenousterritorycover1.6millionha
et al. 2008), we followed the following steps to offorest.Togethertheyformthelargestprotected
evaluate the realistic effect of biodiversity loss on area in Amazonian Ecuador (~17.7% of the
leaf-litter decomposition: (1) We describe the soil Ecuadorian Territory, Valencia et al. 2004) and
invertebrate communities in detail, to understand harbor the world’s most diverse tropical forest
howtheyareorganizedanddistributedatthetax- (Basset al.2010).YNPisawetevergreenlowland
onomic, functional (including body size classes), forest ranging in altitude from 200 m to 300 m
and spatial levels; (2) we used a non-random above sea level. It has a 15- to 30-m canopy with
exclusion experiment in which we manipulated some emergent trees reaching 50 m (Valencia
the accessibility of five size classes of soil detriti- et al. 2004). Rainfall and temperature are asea-
vore to eight types of plant leaf-litter resources sonal; there are no clear patterns of dry/rainfall
(analogous to a removal experiment where we and warm/cold seasons during the year. Over
tested the influence of the dominant species, as 53 months of records at the research station, the
explainedbythemassratiohypothesis;D(cid:1)ıazet al. longest rainless period was three weeks and the
❖ www.esajournals.org 4 February2017 ❖ Volume8(2) ❖ Articlee01692
(cid:1)
CARDENASETAL.
least rainy month was August. The mean annual Atotalof40Winklerextractions(from1 m2of
rainfall was 2826 mm, and no calendar months soil leaf litter) were analyzed. Twenty samples,
averaged less than 100 mm (Valencia et al. 2004). each 10 m apart along a 200-m transect, were
Over742 daysbetween2008and2011,meantem- analyzed using the ALL protocol (Ants of the
perature at the research station was 24.9°C (cid:1) 3.9 Leaf Litter, see Agosti and Alonso 2000 for
(ranging from 22° to 32°C; min: 16.9°C; max: details).Theremaining20weredistributedalong
38.9°C;warmestmonths:November–March;cold- 10, 20-m transects that followed the diagonal of
estmonths:April–October),andameanhumidity the 1000 9 500 m plot. Each transect was set in
was 88.4% (cid:1) 13.9 (data obtained from Yasun(cid:1)ı 10, 100 9 50 m subplots where two samples, 5–
Research Station meteorological station of the 20 m apart, were extracted and analyzed
Pontificia Universidad Cato(cid:1)lica del Ecuador, (AppendixS1:Fig.S1b).
http://www.yasuni.ec). The study area was in the
vicinityofYasun(cid:1)ıResearchStationinthe50 haof Speciesidentificationandallocationtofunctional
the Yasun(cid:1)ı Forest Dynamics Plot (see http:// groups
www.ctfs.si.edu/site/Yasuniforadetaileddescrip- Specimens were examined under the stereo-
tionofthearea).Soilsinthestudyareaaremostly scopeat0.689–509(LeicaM275,LeicaMicrosys-
clay-like,acidic,udultultisolswithanaveragepH tems AG, Wetzlar, Germany). They were
of 4.6 (John et al. 2007) and a texture dominated countedandidentifiedtothefinestpossibletaxo-
bysilt(Tuomistoet al.2003). nomic group using specialized literature (see
Appendix S2: Table S1 for a complete list of all
Soildetritivorebiodiversitysurvey identifiedspecimens).Whenamorphospecieswas
We sampled invertebrate communities within recognized for the first time, lateral, dorsal, and
the studyarea, including only those that may be ventralimagesweretakenusinganadaptabledig-
involved in the fragmentation of leaf-litter mate- italcamera(FutureOpticsScienceandTechnology,
rial from the humic leaf-litter layers using two 1.3MP,MEM1300model,Hangzhou,China).This
well-knownandcomplementarysamplingmeth- image served for comparison for all subsequent,
ods: pitfall traps and Winkler extractions (see similarspecimensinthecollection.Larvaeofholo-
Appendix S1). Our survey designs for both metabolous insects could not be associated with
methodologies were conceived to evaluate the any adult species, so they were classified into
effect of the spatial area on soil food web struc- differentmorphospecies.Nymphsofhemimetabo-
ture and functionality. We are conscious that our lous insects that differed structurally (but not
samplingmethodologyisnotsuitableforcollect- by color) from any adult morphospecies were
ing few important groups of invertebrates such assumed to be new morphospecies. This level of
asearthwormsornematodes,whicharenotcon- taxonomic resolution has been found to be suffi-
sidered as litter shredders (except probably for cient for detecting significant patterns of commu-
some omnivore nematodes) and are mainly col- nitycompositionintemperateandtropicalsystems
lected in the soil below the leaf-litter layer (see (Timmset al.2013,Lamarreet al.2016).Moreover,
Cares and Huang 2008 for nematodes soil sam- family-level identification has provided ecologi-
plingandextraction). cally adequate surrogates for species in studies of
For the pitfall traps, we set a nested rectangu- functional diversity (see also Cardoso et al. 2011
lar grid of six different spatial scales across the for both temperate and tropical systems). Length,
forest floor (smallest scale: 10 9 5 m; largest width,anddepthweremeasuredforupto10spec-
scale: 1000 9 500 m; each scale doubled the imensofthesamemorphospeciestoimproveaccu-
length and width of the previous). At each of racy of morphometric dimensions. Finally, one or
these scales, we sampled four plots (one in each morefunctionalgroupcategorieswereassignedto
corner of each scale) for a total of 69 plots eachmorphospeciesbasedontheclassificationby
(AppendixS1:Fig.S1a).Eachplotconsistedofat Moreira et al. (2008): herbivores, ecosystem engi-
leastthreepitfalltraps(uptofiveinsomecases), neers (Jones et al. 1994, Lavelle et al. 1997, Boze
which remained opened for 24 h. Pitfall traps et al.2012,Jones2012),littertransformers,decom-
consistedofplasticcupsthatwere5 cmindiam- posers, predators, microregulators, and soil-borne
eterand10 cmdeepandburiedtosoillevel. pests, diseases, and parasites (primary producers,
❖ www.esajournals.org 5 February2017 ❖ Volume8(2) ❖ Articlee01692
(cid:1)
CARDENASETAL.
microsymbionts,andprokaryotictransformercate- sub-adult trees of eight common angiosperm
gories were not included in our collection target). species belonging to eight different families:
An extra functional group category, so-called Matisiamalacocalyx(A.Robyns&S.Nilsson)W.S.
mesoregulators, was assigned to encompass the Alverson (Malvaceae), Inga capitata Desv. (Faba-
mesofauna that regulates nutrient cycles through ceae), Nectandra viburnoides Meisn. (Lauraceae),
grazing and other interactions with decomposer Miconia “purpono” [nomen nudum] (Melastom-
microorganisms (analogous to microregulators, ataceae), Siparuna decipiens (Tul.) A. DC. (Mon-
but at larger scale). Feeding habits were deter- imiaceae), Pseudolmedia laevis (Ruiz & Pav.) J.F.
mined using specialized literature and Internet Macbr (Moraceae), Neea “comun” [nomen
resources (e.g., Gillot 2005, Triplehorn and Jonson nudum] (Nyctaginaceae), and Leonia glycycarpa
~
2005,Brandaoet al.2011,http://soilbugs.massey. Ruiz&Pav.(Violaceae).Leaveswerecollectedby
ac.nz/index.php, http://www.collembola.org/). shaking two to seven trees from the trunk or
Scolytines (Coleoptera) were considered ecosys- branches 5–20 times. Leaves fell on white fabric
tem engineers because of their hole-digging sheets (cotton, 1.5 m 9 3.5 m) that were tied
behaviorthatphysicallychangesthesurrounding 1 mabovetheforestfloor.Leavesthatwerechar-
environment and creates access for subsequent acteristically young (presenting bright green or
decomposers (Muller et al. 2002). Although the reddishcolorsand/orsoftlamina),tooold(rotten
taxonomic order Acari represents an important or presenting large amounts of necrosis), or pre-
groupinthesoilfoodweb,wewereunabletodis- sented evident fungal infection or insects gal-
criminate specimens at the morphospecies level leries were discarded. For the experiment, we
and accurately assign them to any of their many collected only senescent leaves with herbivore
potential functional groups. Appendix S3, how- damage covering less than 30% of the leaf area
ever,showsindependentanalysesofthedistribu- (visuallycalculatedinsitu).
tionofbodysizeandbodywidthinthisrichand
complexgroup(AppendixS3:Fig.S1).Acariwere Soilfaunafoodwebexclusionanddecomposition
excluded from this and further analyses that experiments
include morphospecies identity. We recognize To assess whether a sub-group of species dis-
that the functional traits assigned in our study proportionately influences leaf-litter decomposi-
may still be a simplistic representation of insect tion, we designed an exclusion experiment in
ecological niches; however, we believe this is a terra firme using 20 cm diameter plastic and
novel and realistic approach based on what is polyester fabric leaf litterbags. As a general rule,
known in the literature. Appendix S4 shows metabolism depends on body size (Brown et al.
examplesofsoilinvertebratemorphospecies,sep- 2004), so this experiment aimed to simulate
arated by functional groups and body dimen- microcosms where the soil fauna that has access
sionsthatwerefoundinbothpitfallandWinkler to litter resources is filtered sequentially as a
samples. function of body size. So, at one extreme, we
excluded everything except microfauna (e.g.,
Leaf-littercollection micro-detritivores, predators, omnivores). At the
Leaves are the most important component of other extreme, we allowed all size groups of soil
litterfall in Amazonian forests (Chave et al. fauna. This is also in line with paleontological
2010). Most above-ground production comes analyses showing that burrowing invertebrates
from angiosperm trees (Kurokawa and Nakashi- such as beetles, bees, spiders, wasps, ants, and
zuka 2008). In a previous 11-month survey, cicadas of the Paleocene–Eocene Thermal Maxi-
C(cid:1)ardenaset al.(2014)foundthatforestfloorwas mum shrank in size by 50–75% (Smith et al.
mainly covered with leaf litter from 53 common 2009).Salazaret al.(2007)analyzingfuturemod-
tree species belonging to 21 plant families (rare els of the Amazonian forest climate suggest that
species were not considered in that survey). temperatures rising (2–6°C through year 2100)
Basedonthislist,speciesandfamilieschosenfor may induce larger evapotranspiration in tropical
the present study aimed to represent a wide regions. Over time, these ecological and other
range of intrinsic chemical traits (Appendix S5: factorsmayleadtoevolutionaryresponsesfavor-
Table S1). Leaves were collected from adult and ing smaller individuals (Sheridan and Bickford
❖ www.esajournals.org 6 February2017 ❖ Volume8(2) ❖ Articlee01692
(cid:1)
CARDENASETAL.
2011). Microcosms were constructed by slightly The experiment consisted of testing the leaf
folding and sewing together top and bottom decomposition rate of the eight species over the
meshes to form a bag. Theresulting oval-shaped same period of time and area, but in different
litterbags prevented the litter from flattening, litterbag types in an area covering about 4000 m2
and allowed it to retain a natural, three- oftheforestfloorin“ridge-slope”-typemicrohabi-
dimensional structure. Five different mesh sizes tat (see Valencia et al. 2004 for a detailed descrip-
that allowed the access/exit of different-sized tion of microhabitat designation). In total, we
invertebrate groups were used (see Swift et al. analyzed the decomposition process of 8 (plant
1979 for soil microflora and fauna size classifica- species) 9 5 (mesh treatments) 9 10 (replica-
tion): 268.8 mm2 (15.2 9 17.7 mm; micro-, tions) 9 2–5 (leaves per bag, with the number of
meso-, macro-, and megafauna), 118 mm2 (10 9 leaves depending on leaf size), for a total of 400
11.8 mm; micro-, meso-, and macrofauna), leaf litterbags and up to 2000 leaves placed ran-
16.1 mm2 (3.2 9 5.1 mm; micro-, meso-, and domly in the study area (Fig. 1). Each litterbag
macrofauna), 2.7 mm2 (1.1 9 2.5 mm; micro- wasfilledwith2.78 (cid:1) 1.04 gofleaf-littermaterial,
andmesofauna),and<0.01 mm2(~0.1 9 0.1 mm; excludingthepetioles.Thelargedifferenceinlitter
microfauna). Our non-random experimental mass added to the litterbags resulted from inter-
design was conceived to ensure the accessibility specific differences in leaf size (mean (cid:1) SD:
ofallspeciesofthesmallestsize class(those that 108.6 cm2 (cid:1)50.3;min.:55.6 cm2;max.:195.3 cm2)
weredominantintermsofabundance)atallrich- andleafvolume(leafsize 9thickness;1.84 cm3 (cid:1)
ness levels (microcosm treatments), emulating 0.8; min.: 0.8 cm3; max.: 3.1 cm3; data of specific
whatwouldbefoundinnature,wheredominant mean leaf size taken from Kraft and Ackerly
species are less likely to be lost from communi- 2010). We did not collect the largest leaves (e.g.,
tiesunlesstheyarevulnerabletoparticularcatas- from M. malacocalyx or N. viburnoides species), as
trophic events (Smith and Knapp 2003). This wewantedtoavoidhavinglitterbagswithdrama-
experimentaldesignalsoaimedtodetectfacilita- tically higher leaf-litter mass with a comparable
tion in litterbags. Previous research has shown numberofleaves(twotofive).Litterbagswereset
that the design does not confound access with in groups of five treatments, each with a random-
“leakiness,” has quantified the assumption that ized set of species, in a 50 9 80 m grid-type plot
increasing mesh size leads to greater soil fauna (see Appendix S6: Fig. S1 for details). Mesh bags
access, and has ruled out microclimate effects were collected for analyses after 104 days of
across different mesh sizes (Milton and Kaspari decomposition (mass loss of ~58% on average for
(cid:1)
2007). Besides,BokhorstandWardle(2013),after thesameeightspeciesafter103 days,seeCardenas
examining the microclimate effects of three lit- 2013). In the laboratory, leaves from each litterbag
terbagsofdifferentmeshsizesonlittermassloss were gently cleaned to remove soil particles,
and leaching in the absence of soil animals, con- adhering debris, and invertebrates; dried at 40°C
cluded that studies can use different mesh sizes for48–72 h;andweighedat0.001 gprecision.
to reliably quantify the role of soil animals in lit-
ter mass loss from litterbags. This methodology Dataanalyses
may furthermore allow us to decouple diversity Soil functional groupcluster analysis.—Groupsof
effects—niche complementarity—from those of organismsinthesoilfoodweboverlapintermsof
sampling effect (Huston 1997), because we con- feeding habits and body sizes, and therefore in
sidered naturally uneven species abundance, their functional role and extent of their impact on
includingthosefromdominantgroupsacrossthe soil ecological processes. A cluster analysis was
trophicweb(SmithandKnapp2003). first performed to statistically classify the different
Finally, we removed the peciolum of every leaf functional groups found in the forest floor. This
and placed the leaves to dry at 40°C for 48– classification was used for all subsequent analyses
72 hours in cotton fabric bags (containing no in our study. For this, we used morphospecies
more than 10 leaves per bag) and weighed to bodylength,width,anddepth,and0-1binarydata
0.001 g precision. The leaves were remoistened of the eight functional group categories. Twelve
using rainwater to make them pliant and were groupsofspeciesweredefinedusingGower’sdis-
enclosedinthelitterbags. tance(minimumspanningtree)andWard’slinkage
❖ www.esajournals.org 7 February2017 ❖ Volume8(2) ❖ Articlee01692
(cid:1)
CARDENASETAL.
Fig. 1. Experimentaldesignoftheleaf-litterdecompositiontestinahypotheticalextinctionscenariowherebig-
ger invertebrates may be more (and first) susceptible to extinction as a consequence of current climate change.
Bluegridsrepresentthefivedifferentmeshsizesusedtograduallyexcludetheaccessofgroupsofinvertebrates.
treat. = treatments;r = replicates;+referstotheadditionalaccessibilityoflargerinvertebratestocoarsermeshes.
Leavespackedinplasticmeshbagswererandomlydistributedontheforestfloorinanareathatcoveredabout
4000 m2,andletdecomposefor104 d.
(minimum variance). We chose Gower’s distance rank plots (Magurran 2004) to describe the com-
asitallowsmixed-scaletypesofdata(quantitative, munity structure of Yasun(cid:1)ı soil fauna in relation
interval,nominalorordinal,ratios,and/ormissing to abundance (Appendix S2) at the levels of
values)andhasproventoconsistentlyprovidethe Order and morphospecies. Morphospecies rank
most reliable results and to minimize tree dissimi- plots were also used to compare, in detail, the
larity (Mouchet et al. 2008). Ward’s method was relative abundance of the different functional
chosen as it has been shown to produce more groups of soil fauna that make up the inverte-
clearly defined clusters than does average linkage brate community. Additionally, following Pre-
(Pla et al. 2012). Cluster analysis was performed ston’s (1948) boundaries of octaves as a measure
using InfoStat v.2012 software (InfoStat Group, of the degree of commonness, we classified the
Córdoba, Argentina) with default program data number of species per functional group in rela-
standardization(DiRienzoet al.2012). tion to its abundance in eight categorical ranges.
Description of the structural and functional Finally, data were fitted to lognormal (equations
organizationofsoilinvertebratecommunities.—Using of the form y¼a(cid:3)eð(cid:4)0:5ðlnðx=bÞcÞ2Þ, where a is the
data from all samples combined, we first used intersection with the y-axis and b and c are
❖ www.esajournals.org 8 February2017 ❖ Volume8(2) ❖ Articlee01692
(cid:1)
CARDENASETAL.
parametersoftheslopeexpression)andlogarith- analyses of similarity indexes for the total number
mic(formesoregulatorsonly)distributionsusing of collected morphospecies, their relative abun-
Table Curve 2D software v.5.01 (Systat Software dance, biovolume, and functional groups identity
Inc.,SanJosé,California,USA). are detailed in Appendix S8. Analyses were per-
Rarefaction “sample-based” accumulation formedusingPastv.3.07(Hammeret al.2001).
curveswereusedtocomparetheefficacyofdiffer- Sørensen’s similarity coefficient was similarly
entsamplingmethodsusedinthesamearea(San- used to analyze the spatial heterogeneity in spe-
ders 1968, Longino et al. 2002, Ellison et al. 2007) cies composition of the leaf litter from 28 com-
and to assess whether pitfall traps and Winkler mon species found in the study area using data
(cid:1)
extractionsreachedanasymptoteforspeciesrich- fromCardenaset al.(2014).
nessandfunctionalgroups.Forthis,weusedPast Meshsize-dependentlitterdecomposition.—Shapiro-
v.3.07 software (Oslo, Norway; Hammer et al. Wilk normality tests for the residuals of each
2001) that implements the analytical solution in treatment showed normal distribution (with all
which standard errors are transformed in (cid:1)95% W’s > 0.96 and all P’s > 0.14) for four species:
confidence intervals. This is known as “Mao tau” I. capitata, L. glycycarpa, N. viburnoides, and S. de-
following Colwell et al. (2004). Sample-based rar- cipiens, and non-normal distribution (with all
efaction curves implicitly reflect empirical levels W’s < 0.94 and all P’s < 0.015) for the remaining
of within-species aggregation of individuals by four species: M. malacocalyx, M. “purpono,” N.
considering only incidence, providing a realistic “comun,” and P. laevis. The percentage of mass
estimateofthenumberofspeciesfoundinsetsof loss of the leaf-litter material of the eight plant
real-world samples (Gotelliand Colwell2001).To species in each of the five mesh-bag treatments
assess whether soil fauna diversity was aggre- was compared using analysis of variance
gated,wecomparedbothindividual-andsample- (ANOVA,forthosespeciespresentingnormaldis-
basedrarefactioncurvesbyplottingthemtogether tributions of their residuals) and Kruskal–Wallis
(Appendix S7:Fig. S1). GotelliandColwell(2001) tests (for those species presenting non-normal
explain that when the sample-based curve lies distributions of their residuals). Significance of
belowtheindividual-basedcurve,onecanassume the differences between treatments was assessed
there is species aggregation. An estimation of the using Tukey’s post hoc tests and Dunn’s z post
total soil fauna biodiversity was evaluated using hoc, respectively. ANOVAs, Kruskal–Wallis, and
Jackknife 1 and Jackknife 2 (employing incidence Shapiro-Wilk tests were performed using Past
data) and Chao 1 (employing abundance data) v.3.07(Hammeret al.2001).
species richness estimators (Gotelli and Colwell In order to analyze the factors controlling
2001, detailsinAppendixS7). decomposition of the pool of our studied species,
Biovolumedistributionoftheclusterizedfunc- the differences in the percentage of averaged leaf-
tionalgroupswasplottedtovisualizethespread littermasslossbetweenthemeshtreatmentswere
of mass volume and body width in the soil food analyzedusingageneralizedlinearmodel(GLM).
web community. Finally, radar charts were plot- For this analysis, mesh sizes (mm) and chemical
ted to compare the potential accessibility of the traitsofthedifferentspecieswereusedasindepen-
six functional groups of soil fauna to the five dent variables. Prior to GLM, a principal compo-
types ofmicrocosms (mesh bags T1–T5) in terms nents analysis (PCA) of nutrient content of all
of their accumulated biovolume and the relative species (22 chemical traits, see Appendix S5 for
numberofspecies. details)wasperformedinordertoevaluatewhich
Species–area relationship (soil fauna and leaf elements better explain the quality differences
litter).—We compared content of each collection between our resource treatments. A previous
unit (pitfall traps and Winkler extraction sites) to studyon thesesameplant specieshasshown that
oneanotherusingsimilarityindexesandEuclidean condensedtannins,lignin:Nratio,andMn:Curatio
distances to evaluate whether there is a patchy or significantlyexplainOMdecomposition(C(cid:1)ardenas
uniform distribution of soil fauna within the com- et al.2015);hence,wealsoincludedthesetraitsin
munitiesoffunctionalgroups.Sørensen’ssimilarity thePCAandGLM.PCAwasperformedusingPast
coefficient was used for the species–area analysis v. 3.07 (Hammer et al. 2001), and GLM was ana-
using binary presence–absence data. Further lyzedusingR(RDevelopmentCoreTeam2015).
❖ www.esajournals.org 9 February2017 ❖ Volume8(2) ❖ Articlee01692
(cid:1)
CARDENASETAL.
Table 1. Functionalgroup(FG)categories(Category1)andsub-categories(Category2)showingthetotalnum-
berofspecies(S),abundance(N),andbiovolume(B;N 9 mm3;calculatedfromaveragedvolumeperspecies).
Category1 FG Category2 S N B
Littertransformers LT1 48 127 75,322.9
LT2 Ecosystemengineersandmesoregulators 57 294 7798.3
LT3 Microregulators 57 347 155.2
Total 162 768 83,276.4
Omnivores O1 +mesoregulators 26 128 892.7
O2 +predators 85 398 4795.4
Total 111 526 5688.1
Mesoregulators Me 38 145 339.9
Total 38 145 339.9
Herbivores H1-Pd1 Soil-bornepests,diseases,andparasites 24 58 310.4
H2 32 95 4340
Total 56 153 4650.4
Predators P1 Ants 30 161 1743
P2-Pd2 Soil-bornepests,diseases,andparasites 26 79 36.3
P3 64 250 1308.3
Total 120 490 3087.6
Soil-bornepests,diseases, Pd3 29 68 27.9
andparasites Total 29 68 27.9
Total 516 2150 97,070.3
Notes: Omnivores(O1andO2)correspondtoherbivoresand/orlittertransformers,includingguildsinCategory2.Acariare
notconsideredinthisanalysis.
To compare the effect of thefive differentmesh waytovisuallycontrasttheeffectoflargerbiovol-
size treatments on the decomposition process, we umeaccessibilitywithresourcedecomposition.
calculated the weighted mean of the final:initial
R
ratio of leaf dry mass for the eight plant species. ESULTS
We used the diagonal size of the grid holes for
each of the five meshes (mm) as the weight (w) Characterizationoffunctionalgroupsofsoil
parameter. Because data elements with a high invertebrates
weight contribute more to the weighted mean Cluster analysis discriminated six main func-
thando elementswitha lowweight, weexpected tionalgroupclassesinthesoilfoodweb(cophe-
the coarser mesh bags to contribute more to netic correlation = 0.628). Some were mainly
decomposition. The averaged weighted means of related to dead plant resources (leaf litter and
the eight species were compared to the arithmetic wood), while others were related to fun-
means using a two-sample t-test with Past soft- gal resources: “litter transformers” (LT; mainly
ware v. 2.17 (Hammer et al. 2001, Shapiro-Wilk collembolans,diplopods,blattids,ants,staphyly-
normality test: W = 0.912, W = noids, and termites), “omnivores” (O; mainly
weighted arithmetic
0.869, P > 0.05 in both cases; details in gryllids,thysanopterans,staphylynoids,isopods,
AppendixS9). and ants), and “mesoregulators” (Me; mainly
Body width is probably the most important Diptera larvae, mycetophylids, ptiliids, endomy-
morphometricparameterthatdiscriminatesaccess chids, and staphylinids). Others were predators
to the leaf-litter resource in each mesh litterbag or parasites: “predators” (P; mainly ants, arach-
treatment (width:height average ratio (cid:1) SD of nids, phorids, and chilopods), “soil-borne pests,
the litter transformer + omnivore communities = diseases, and parasites” (Pd; mainly mymarids,
1.28 (cid:1) 0.89). We assessed the biovolume sapygids,scelionids,braconids,andtrychogram-
(N 9 mm3) of litter transformer and omnivore matids),and“herbivores”(H;mainlycicadellids,
functionalgroupsthatpotentiallyhadaccesstothe orthopterous, and curculionids). These six
different mesh treatments, and plotted along with groups were divided into 12 sub-categories
the leaf-litter mass loss of the pooled dataset as a (Table 1, Fig. 2). Of these, the most speciose and
❖ www.esajournals.org 10 February2017 ❖ Volume8(2) ❖ Articlee01692