以生態模式探討植物性狀與土壤微生物組成對植物土壤回饋之種間差異的影響

(1)

୯ҥᆵ᡼εᏢғނၗྍᄤၭᏢଣහ݅ᕉნᄤၗྍᏢس ᅺγፕЎ

School of Forestry and Resource Conservation College of Bioresources and Agriculture

National Taiwan University Master Thesis

ағᄊኳԄ௖૸෌ނ܄ރᆶβᝆ༾ғނಔԋჹ!

෌ނβᝆӣ㎸ϐᅿ໔ৡ౦ޑቹៜ!

Plant Trait and Microbial Composition Interactively Determine Species Variation in Plant Soil Feedback

Ϋa Modeling Approach

࢒࢙ӵ Po-Ju Ke

ࡰᏤ௲௤ǺΟЕ଼ റγ Takeshi Miki, Ph. D

΍ے᝵ റγ Tzung-Su Ding, Ph. D

ύ๮҇୯ 102 ԃ 6 Д

June 2013

(2)

i

Acknowledgement

གᖴךޑࡰᏤԴৣǴ΍ے᝵ԴৣᆶΟЕ଼ԴৣǶགᖴٿՏԴৣᕴࢂᡣךԾҗ Ӧ،ۓࣴزޑБӛǴᡣךԖᏢಞᐱҥࣴزޑᐒ཮Ƕ੝ձགᖴΟЕ଼ԴৣӧךεΟ ޑਔং஥ך຾Ε౛ፕғᄊޑሦୱǴаϷգ΋ၡаٰ׫ݙޑਔ໔ᆶЈΚǶགᖴٿՏ ԴৣᕴࢂόीӣൔӦගٮӚᅿၗྍǴᡣ΋ঁᅺγғԖᐒ཮ૈр୯ٗሶӭԛǶགᖴ ٿՏԴৣӧࡐӭಒ༾ޑӦБ๏ϒග࣪ᆶѕᓲǴਔதಒ༾ډᡣךࣁԾρޑόᐱҥག ډόӳཀࡘǶٿՏԴৣک๓ޑࡑΓೀ٣ǵيࣁࣽᏢৎޑᝄᙣᄊࡋǵаϷӧࣴزᆶ ғࢲ໔ޑКٯ৾ਅࢂךޑᏢಞڂጄǶ

གᖴα၂ہ঩ᖴדᇬԴৣǵ݅ەᓉԴৣǵ݅ߘቺԴৣǴགᖴգॺᜫཀ޸ਔ໔

᎙᠐ፕЎǴ٠வόӕޑय़ӛග࣪ךځύޑόىǶགᖴᜢޚےԴৣаϷ༫ξճೕԴ

ৣӧεᏢਔ௲ך಍ीၟ RǴᡣךӧࣴز܌όӆࣁԜགډ্܂ǶགᖴΦԯܻ࠹Դৣǵ ᗛ୯ޱԴৣӧӚБय़ޑڐշᆶႴᓰǶ

གᖴੇࢩ܌ޑεৎǶགᖴ೭ࢤਔ໔΋ଆ૸ፕޑ 403 Ꮲߏۆॺ:ृሺǵ੦ᆝǵ ጰࡹᑣаϷځԳᏢߏǶགᖴՙߙǵᒗǵCotǵOscarǵ܃։ǵ߷൛ǵ▲۸ǵ܎ކǵ

఍ಌǵ᝵҇۰ǵCirkǵEricǵᑉဖǵৱᒑǵࡿԋǵCarmen Ϸߪ଻೭ࢤਔ໔ޑྣ៝

аϷࣁғࢲ஥ٰޑ៿኷Ǵᡣӧੇࢩ܌อයۚ੮ޑךόԿܭϼၸېൂǶགᖴᖴדᇬ ԴৣᆶЦችྼԴৣǴགᖴգॺᜫཀᡣ΋Տډ౜ӧᗋࢂόϼᔉੇࢩޑΓୖу meetingǴ٠ᕴࢂࡐک๓Ӧ๏ϒࡐӳޑཀـǶ੝ձགᖴ TakeǴᕴࢂدΠ௲ךࡐӭ λמѯ٠๏ϒႴᓰǶ

གᖴහ݅سޑεৎǶགᖴࣴز܌೭ࢤਔ໔΋ଆոΚޑӕᏢ:೚ษǵ๥໦ǵے ՙǵ኷ჱǵਜ٥ǵەޱǵᛄ⧶Ƕགᖴ܌ԖӢࣁහ݅سԶ่᛽ޑգॺǺλܴǵᇳ৒ǵ

࿰໥ǵߓ൫ǵЦӹಅǵߓ٧ǵλඬǵԴҜǵߓቺǵ۸ችǵṓᗱǶ੝ձགᖴआ♻ǵ λΏǵλᆅаϷ livingǴགᖴգॺᕴࢂᡣךԖܤ࡜کӗვޑӦБǶᗨฅךவа߻

൩΋ޔӧ༙գॺаϷහ݅سޑᕴᕴǴՠහ݅سࢂৎǴ൩ᆉךࡕٰКၨதࡑӧੇࢩ

܌Ǵහ݅سϝࢂ୤΋΋ঁૈᡣךགډӼЈᆶܫ᚞ޑӦБǶ

(3)

ii

གᖴෞԑᣴǶाόࢂӧчεݓξ΢ޑգǴаϷךച໗ԲᓐξࡕӧѠчξНѯ ၶޑգǴך๊ჹό཮و΢౛ፕғᄊ೭చၡǶགᖴ݅◖৥Ǵکգޑፋ၉ᕴࢂ෧጗ך ӧੇࢩ܌ޑғࢲᓸΚǶགᖴλभᏢۆǴکգॺޑᇡ᛽੿ޑࢂጔϩǶ

གᖴ 401 ޑεৎǴ੝ձགᖴҖῑǵDevilǵElaineǵٵᆺǵߓᗪǵ҅ӹǵ֮ᆺǴ གᖴգॺᕴࢂᜫཀکך૸ፕᏢೌᆶΖړǴܤᄹךࡕٰӳϿѐ 401Ǵՠך΋ޔΜϩ གྷۺٗ॥Ӏܴ൜ޑӦБǶᖴᖴහӭཷޑշ௲ဂᆶᏢғǴᗨฅךୂᚷѝ஥Α΋ԃǴ ՠӢࣁգॺǴךωёаӧғᄊኳᔕޑӕਔǴϝฅόבჹεԾฅޑ዗௃Ƕӧۺࣴز

܌ޑ೭΋ٿԃ໔Ǵךว౜೭ࢂ΋ҹคКख़ाޑ٣Ƕ

གᖴጰܵ੿Ƕᖴᖴգ೭൳ԃٰޑഉՔǴགᖴգᕴࢂҔԾρޑх৒ඤڗךπբ ޑਔ໔Ǵךࡐࣔெ೭ளٰόܰޑጔϩǶգᕴࢂ኷ᢀӦӣᔈΓғύޑ਋௳Ǵ٠ӧཱུ

ࡋᕷԆޑВηύ٦ڙғࢲΔ೭ኬޑΓғᄊࡋࢂך҉ᇻޑᏢಞჹຝǶ

നࡕǴགᖴךޑৎΓǶᡣךவλ൩ёаᒧ᏷ԾρགྷۺޑࣽسǶᗨฅךᕴࢂࡐ Һ܄ӦᒧΑനᡣΓᏼЈǴԶЪ׫ၗൔၿ౗നեޑᒧ໨Ǵՠགᖴգॺϝฅ΋ၡЍ࡭

ډ౜ӧǶ

(4)

iii

Abstract

Interaction of plants with the nearby soil environment, a process termed

plant-soil feedback (PSF), is a structuring force for vegetation development.

Understanding how plant functional traits control PSF strength variation among

species is thus critical for plant community ecology. Studies have highlighted either

nutrient cycling (litter-mediated PSF) or soil biota (microbial-mediated PSF)

separately as two main drivers of PSF and thus focus on different sets of plant traits.

However, the two PSF drivers are not independent and their way of interaction

depends on the functional type of microbes (i.e. pathogens and mycorrhizas). An

ecosystem model coupling indirect interaction between litter and microbial feedback

is presented to identify which traits have strongest effect on PSF strength and, its

dependence on soil microbial community composition. This model shows that the

identity of the most influential plant functional traits alters when microbial-mediated

PSF is considered along with litter-mediated PSF. The relative importance of traits

depends on microbial composition. In particular, the importance of litter

decomposability increases with the relative abundance of mycorrhizas due to its

indirect positive effects on litter production. Plants with more easily-decomposable

litter and with more beneficial plant-mycorrhiza associations are more advantageous

than other plants species in pathogen-free soils. On the other hand, plants with better

(5)

iv

defense traits are expected to be dominant in pathogen-rich soils. The results can

provide useful insights into understanding the key determinants of successful plant

invasion in different soil environments.

Keywords

functional trait-based ecology, litter decomposition, mycorrhiza, soil pathogen,

indirect interaction, population stage structure, exotic plant invasion

(6)

v

ύ

ύЎᄔा

^!

෌ނᆶβᝆ໔ҬϕբҔ܌ౢғޑӣ㎸౜ຝ(෌ނβᝆӣ㎸)཮ׯᡂ෌ဂว৖Ƕ

ӢԜΔΑှ෌ނфૈ܄ރӵՖፓ௓෌ނβᝆӣ㎸மࡋϐᅿ໔ᡂ౦ࣁ෌ဂࣴزޑख़

ा᝼ᚒǶၸѐࣴزѝ௖૸࢏ပނᎦϩൻᕉ܈βᝆ༾ғނϐൂ΋Ӣη܌೷ԋϐӣ㎸

౜ຝǶฅԶ೭ٿ໨Ӣη٠ߚᐱҥΔϩ໒૸ፕ཮۹ౣ࢏ပނᆶ༾ғނ໔ޑ໔ௗҬϕ

բҔǶԜѦΔၸѐࣴز۹ౣΑόӕޑ༾ғނ཮೸ၸόӕޑᐒڋቹៜ࢏ပނ୏ᄊǶ

ҁࣴزࡌҥ΋ঁғᄊኳԄǴӕਔયΕ࢏ပނൻᕉᆶβᝆ༾ғނჹ෌ނβᝆӣ㎸ޑ

ቹៜΔаΑှ෌ނфૈ܄ރӵՖፓ௓෌ނβᝆӣ㎸மࡋΔ٠௖૸βᝆ༾ғނဂᆫ

ಔԋӵՖቹៜόӕфૈ܄ރޑ࣬ჹख़ा܄Ƕ่݀ᡉҢ྽עβᝆ༾ғނޑቹៜયΕ

࢏ပނӣ㎸ޑኳԄਔǴቹៜΚၨமޑфૈ܄ރৡ౦࣬྽εǶԜѦǴβᝆ༾ғނဂ

ᆫಔԋ཮ቹៜфૈ܄ރޑ࣬ჹख़ा܄Ǵځύ࢏ပނϩှೲ౗ޑख़ा܄཮ᒿ๱๵ਥ

๵ޑᙦࡋቚуԶᡉ๱΢ϲǶԜኳԄႣෳӧؒԖβ൞ੰ্ޑβᝆύΔܰϩှϷ཮ᆶ

๵ਥ๵׎ԋଯਏ੻ϕճӅғޑ෌ނ՞Ԗғߏᓬ༈ΙԶ྽β൞ੰ্ᙦࡋଯਔΔڀԖ

ၨ٫ٛᑇ฼ౣޑ෌ނ཮՞ᓬ༈Ƕҁࣴز่݀ёᔈҔܭΑှфૈ܄ރӵՖӧόӕβ

ᝆᕉნύቹៜѦٰᅿޑΕߟǶ!

!

ᜢᗖӷ!

фૈ܄ރғᄊᏢǵ࢏ပނϩှǵ๵ਥ๵ǵβ൞ੰ্ǵ໔ௗҬϕբҔǵ௼ဂቫભ่

ᄬǵѦٰ෌ނΕߟ!

(7)

vi

Introduction

Understanding the processes controlling community structure and the influence

of functional traits of organisms on community characteristics through these processes

is a central question for trait-based ecology (McGill et al., 2006). While many studies

have discussed the importance of plant functional traits (e.g. leaf economics, wood

density) in determining aboveground species interaction and ecosystem processes in

terrestrial ecosystems (Wright et al., 2004; Chave et al., 2009; de Bello et al., 2010;

Kunstler et al., 2012), the plant traits that affect the interaction between plants and

soil, a process termed plant-soil feedback (PSF), are also found important to plant

community structure (Bever et al., 2010; De Vries et al., 2012). Plants with different

traits are able to cultivate nearby soil environment differently, causing changes in the

physical, chemical and biological properties of the soil, and eventually influence

future performance of these same individuals or other individuals that grow nearby

(Bever et al., 1997; Ehrenfeld et al., 2005). The direction and strength of such PSF

can be determined by the relative growth rate of plants in soils with different

cultivation history. When a plant changes the soil environment in a direction such that

growth response of conspecific plants are larger (or smaller) compared to those

individuals that were grown in heterospecific-cultured soils, the feedback is defined as

positive (or negative) PSF (Bever et al., 1997; Brinkman et al., 2010).

(10)

2

Past empirical studies have revealed that PSF processes are ubiquitous in various

ecosystems, ranging from temperate grassland (Klironomos, 2002), temperate forest

(Packer and Clay, 2002), to tropical rainforest (Bell et al., 2006; Mangan et al., 2010).

Empirical studies have also showed that species vary greatly in realized PSF strength,

and suggested that such variation acts as a structuring force of plant community

(Bever et al., 2011; van der Putten et al., 2013 and reference therein). PSF can change

plant community composition by accelerating species replacement, which is the major

driving force of succession in sand dune and grasslands (van der Putten and Peters,

1997; Kardol et al., 2006; Kardol et al., 2007). Other studies in temperate grasslands

and tropical forests (Klironomos, 2002; Mangan et al., 2010) showed that rare species

suffer stronger negative PSF, implying that variation in PSF strength is a key to

explain rarity of some species and thus have the power to shape community relative

abundance. Moreover, there has been increasing recognition that PSF mediates the

success of exotic plant invasion (Reinhart & Callaway, 2006). Invasive species often

experience more positive PSF in their introduced range compared to that in their

native range by leaving their belowground natural enemies behind (i.e. enemy escape

hypothesis, Callaway et al., 2004). Native species in the invaded area, however, often

received stronger negative PSF after invasion processes due to altered nutrient cycling

(Eppinga et al., 2011) or accumulation of pathogens which have stronger negative

(11)

3

impact on natives (i.e. pathogen accumulation hypothesis, Eppinga et al., 2006).

These examples all show that PSF have the ability to shape vegetation properties such

as community composition and resilience against perturbations (Miki & Kondoh,

2002; Miki et al. 2010). Given the importance of PSF in vegetation development,

understanding what functional traits control species variation in PSF strength is a

critical, but remain unsolved issue, for plant community ecology (van der Putten et al.,

2013).

While many mechanisms can generate PSF, the importance of nutrient cycling

(i.e. litter-mediated PSF) and soil biota (i.e. microbial-mediated PSF) were most

highlighted in past studies. Litter-mediated PSF considers the indirect interaction

between plant and soil chemistry through litter dynamics, emphasizing the role of

species-specific litter traits in controlling local nutrient cycling process through litter

quantity (e.g. litter production rate) and quality (e.g. litter carbon: nitrogen (C:N) ratio,

secondary compound concentration, Binley & Giardina, 1998). Litter feedback studies

often suggest that the direction and strength of PSF depends on the litter

decomposability of nearby trees (Berendse, 1994; Miki & Kondoh, 2002; Eppinga et

al., 2011). For example, species may create positive PSF by enhancing nutrient

cycling through the production of quickly decomposing litter (Berendse, 1994; Miki

& Kondoh, 2002), while negative PSF may be realized due to soil nutrient depletion

(12)

4

by other individuals or release of phytotoxic compounds during decomposition

(Mazzoleni et al., 2007).

The other well-documented mechanism is termed microbial-mediated PSF,

which investigates the direct interactions between plants and soil organisms. Such

studies emphasize the well-established observation that plants differ in their locally

associated microbial community (e.g. due to difference in root exudation and

architecture), and their response to individual microbial species (Bever et al., 2011).

Bever et al (1997) summarized past evidences and synthesized the

microbial-mediated PSF model, which viewed the local soil community as a whole to

have either net positive or negative effects on local plant performance (Bever et al.,

1997). Follwoing this framework, microbial-mediated PSF studies concluded that the

sum of effect of each microbial group on plant performance is the most important

determinant of PSF strength. Positive PSF is thus believed to occur when the plant

facilitates population growth of beneficial microbes (e.g. mycorrhizal fungi and

nitrifying bacteria) more than detrimental microbes (e.g. soil-borne pathogens and

nematodes) during cultivation, while PSF will be negative if the impacts of

detrimental microbes overwhelm that of beneficial organisms (Bever et al., 1997;

Kulmatiski et al., 2011).

Although litter- and microbial-mediated PSF have been both widely documented,

(13)

5

some key properties of the PSF process are often neglected and thus hinder progress

towards characterizing the functional traits that act as major determinants of

interspecific variation in PSF strength. First, litter- and microbial-mediated PSF are

not independent, but have strong indirect interaction between litter and soil microbes

and thus affect plant performance simultaneously. Soil microbes can cause PSF

directly by affecting plant population dynamics (e.g. influencing mortality and

reproduction), and such process will also indirectly influence nutrient cycling through

controlling litter input and nutrient uptake of plants (van der Heijden et al., 2008).

Litter dynamics, similarly, will influence microbial-mediated PSF since their effects

on plant primary production will also influence plant-microbe interactions (Wardle,

2006). In addition, experimental studies often show that the direction and strength of

PSF caused by soil microbes will change with soil nutrient conditions (de Dyne et al.,

2004; Manning et al., 2008). This line of evidence also suggests that litter- and

microbial-mediated PSF are not independent. A single plant trait can thus influence

the direction and strength of PSF through both mechanisms simultaneously. Thus, the

relative importance of traits in controlling PSF strength remains unclear unless effects

of single trait on both litter- and microbial-mediated mechanisms are considered.

Moreover, viewing the soil community as a black box neglects the complex

nature of indirect interaction between litter and microbes. While it is often assumed

(14)

6

that the net effect of local soil community on plant growth is the sum of effect size of

each microbe when discussing plant-microbe interaction, it should be noted that

microbes from different functional groups (e.g. detrimental pathogens and beneficial

mycorrhizas) may indirectly influence litter dynamics through different processes.

Pathogens interact with litter dynamics via inducing additional plant mortality since

dead plant materials becomes litter, while mycorrhizal fungi operate through helping

the plant to deplete soil nutrient, which can increase both plant productivity and litter

production (Read & Perez-Moreno, 2003; Orwin et al., 2011). Litter-mediated

nutrient cycling also indirectly affects pathogens and mycorrhizas differently. An

increase in soil nutrient content and plant biomass due to faster decomposition may

support larger population size of soil-borne pathogens, but may alter plant-mycorrhiza

interactions due to shifts in nutrient limiting status of the microbes (Wallander, 1995;

Treseder, 2004; Johnson, 2009). With these lines of empirical evidence, studies need

(1) to incorporate indirect interaction between litter and microbes via combining the

two mechanisms, and (2) to separate microbial functional groups for better

understanding of the relative importance of traits in determining PSF strength under

different microbial composition.

When combining litter- and microbial-mediated PSF for different microbial

groups, it is important to note that individuals of different stage classes within the

(15)

7

plant population interact with PSF drivers differently and thus their effects on soil

properties are not identical. Seedlings and adults differ in their contribution to litter

production and therefore may play different roles in litter feedback. In particular,

adults can produce litter through tissue turnover (i.e. annual litter production) in

addition to mortality (Clark et al., 2001), while seedlings only contribute to the litter

pool when individuals die. Seedlings are thus passively influenced by soil nutrient

status which is mainly controlled by adults. Seedlings and adults also differ in their

interactions with soil microbes. Seedlings are highly vulnerable to detrimental

pathogens, while adults suffer less pathogen-induced mortality due to their better

protected roots (Augspurger & Kelly, 1984; Alvarez-Loayza & Terborgh, 2011). In the

case of beneficial mycorrhizal fungi, a larger percentage of fungal hyphae are

associated with adults since they can provide more resources to the carbon limited

microbes (Šmilauerová et al., 2012), while seedlings usually depend on mycorrhizal

network which is mainly supported by adults (Dickie et al., 2005).

Although empirical studies had detected large variation in PSF strength, it may

be difficult to experimentally quantify the individual contribution of each trait since

traits are often correlated and thus growth response measured in empirical studies

cannot reveal their relative importance. Modeling is useful to approach this problem.

Here, I present a stage-structured open ecosystem PSF model that couples both litter-

(16)

8

and microbial-mediated PSF. My main purpose is to identify which plant traits have

the strongest effects on PSF strength when both litter- and microbial- mediated

feedbacks are incorporated, and to elucidate how microbial community composition

and stage structure of plants influence the relative importance of traits. In the model, I

separate plant population as seedlings and adults, and model dynamics of litter and

soil nitrogen to incorporate litter-mediated PSF. I consider two distinct groups of

microbes: detrimental soil-borne pathogens and beneficial mycorrhizal fungi. I

highlight these two groups of soil microbes because they have intense direct

interactions with plant roots and may represent the positive and negative extremes of

plant-microbe interaction, also because these two microbial groups are associated

with most plant species and are abundant in soils of all ecosystems. I showed that the

identity of the most influential plant functional traits alters when microbial-mediated

PSF is considered along with litter-mediated PSF. The predicted relative importance

of traits is different for different microbial functional groups. In particular, the

importance of litter decomposability increases with the relative abundance of

mycorrhizas in the whole microbial community. The results provide insights into

understanding the key deternimant of success of exotic plant invasion and restoration,

and can give better predictions of plant growth response in different soil

environments.

(17)

9

(18)

10

Materials and Methods

Model Description

I developed a stage-structured model of PSF with six state variables in an open

ecosystem to explore the effect of different plant traits on PSF strength via coupling

litter and microbial dynamics. The model includes two stages of plants: seedlings and

adults (with densities indicated by S and

A

, respectively). I included litter (

L

) and

soil nitrogen content (

R

) for describing litter feedback mechanisms, while pathogen

(

P

) and mycorrhiza nitrogen content (

M

) are included to represent two distinct

functional groups for microbial feedback mechanisms. Major fluxes of the model are

shown in Fig. 1, model equations are shown in Table 1 and 2.

Seedling and adult demographic dynamics

I divided plant individuals into seedlings and adults to emphasize the differences

in litter contribution and interactions with microbes between the two stages. Density

of seedlings increases with reproduction from adults. The rate of reproduction per

adult is proportional to its soil nitrogen uptake rate rR (where r represents

species-specific reproduction rate per nitrogen uptake). The density of adults increases

with seedling maturation into the adult stage via nitrogen uptake for biomass growth (with biomass growth rate per nitrogen uptake g ). I consider a negative shading

(19)

11

effect of adults on plant demography. Adult reproduction rate and seedling maturation rate decreases linearly with increasing adult density, represented by

max

(1 A ) rR A and

max

(1 A )

gR A respectively, where A_max represents the maximum adult density in the

model ecosystem (Eqn.1 and 2 in Table 1). Plants are assumed to die with a density

independent mortality rate (m and _S m for seedlings and adults, respectively) or A

due to pathogen-induced mortality rate DSP and DAP (with infection efficiency DS and DA for seedlings and adults, respectively).

Mycorrhiza and mycorrhizal-enhancement of plant growth

I assume mycorrhizal biomass is homogeneously distributed belowground and

can thus be divided, proportionally to plant biomass, into those associated with seedlings or adults, _S ^S

S A

M M S B

S B A B

and _A ^A

S A

M M A B

S B A B

, where

BS and B_A are the individual carbon biomass for seedlings and adults, respectively (Table 2). I assume interaction between mycorrhizas and plants is mainly based on

negotiation for two currencies: plant photosynthetic carbon products and mycorrhizal

nitrogen uptake. I assume mycorrhizas uptake soil nitrogen at a rate of uR(with uptake coefficient u), and transfers a minimum proportion n_min to associated host

plant. Such a process is aimed to separate true mutualistic from parasitic relationships

in the model. Mycorrhizas request for carbon from the plant in order to maintain its

(20)

12

C:N ratio (JM), resulting in carbon demand as

¹ⁿ^min

^uRM ^J^M (Wallander, 1995;

Johnson, 2009).

Plants use carbon from the atmosphere for primary production. The amount of

carbon uptake when the plants are not associated with mycorrhizas is

^Nûptake^S ^N ûptakeÂ

^{J , where}^P ^J^P is the plant tissue C:N ratio. Nuptake_S

and N uptake_A within the parentheses represents soil nitrogen uptake by seedlings

and adults when unassociated with mycorrhizas, respectively (see Soil nitrogen

section for detailed formulation). When plant-mycorrhiza associations are formed, plants may adjust their carbon uptake to

^Nûptake^S ^NûptakeÂ

^J^P

¹ ^C^max

^,

where C_max represents the maximum proportion of primary production that plants

will use as root exudation for benefit exchange (Cowden & Peterson, 2009). The

maximum amount of carbon that the mycorrhiza can acquire from the plant after accounting for respiration loss is

^Nûptake^S ^N ûptakeÂ

^J^P^C^max ^e^M ^{, where}

eM is the mycorrhiza carbon assimilation ratio (Bryla & Eissenstat, 2005).

I assume plants and mycorrhizas compare the carbon demand and supply at

every time step in order to decide the amount of exchange between parties. If

mycorrhizal carbon demand is larger than the maximum supply offered by plants,

mycorrhizas are in a carbon-limited status. Under this situation, plants will transfer

maximum amount of carbon, whereas mycorrhiza will transfer excess nitrogen to the

(21)

13

plant after meeting its own metabolic demands (Johnson, 2009). The total amount of

transferred nitrogen is thus

^S ^A

^P ^max ^M

M

uRM ª¬ Nuptake N uptake J C º e ¼ J

1 .

On the other hand, if mycorrhizal nitrogen content is too low such that it is also

limited by nitrogen (i.e. mycorrhizal carbon demand is less than plant supply), the

microbe will keep nitrogen for its own metabolism (Johnson, 2009) and only transfer

minimum amount of nitrogen, n_minuRM . Plants adjust their carbon uptake and

transfer the amount of carbon just enough to meet mycorrhizal demand.

I assume plants allocate their nitrogen, both from root uptake and mycorrhizal

transfer, to reproduction, litter production, and growth. The quantity of mycorrhizal

nitrogen content and mycorrhizal-enhancement of plant demography thus depends on

the nutrient limitation status of mycorrhizas. I assume adults allocate nitrogen to reproduction and primary production with fixed proportion ^S

P

l r: B

J , whereas seedlings allocate all their nitrogen for biomass growth. Seedling and adult dynamics

are given by eqn. 1 and 2 in Table 1, whereas equations for mycorrhizal-enhancement of plant reproduction and growth is given as r_{S A M R}_{, ,} _, and G_{S A M R}_{, ,} _, in Table 2,

respectively.

Mycorrhizal nitrogen content is released back to the soil following natural mortality ( GM ). Turnover of mycorrhiza hyphae may also result from the

density-dependent negative effect of pathogens (EPM) during competition for root

(22)

14

colonization sites (Sikes et al., 2009). The dynamics of mycorrhiza is shown as eqn. 3

in Table 1, while its population growth under different nutrient limitation status is given as P_{S A M R}_{, ,} _, in Table 2.

Pathogens

I assume that pathogen infection causes the mortality of seedlings and adults

with per plant infection rate DSP and DAP, respectively (DS and DA represent

infection efficiency). I assume that the infection efficiency of adults (DA) is much

smaller compared to that of seedlings (DS) since adults have stronger physical

protection (Augspurger & Kelly, 1984; Reinhart et al., 2010). The nitrogen flux from

plants to soil due to pathogen-induced mortality is the product of the number of dead

individuals, their individual carbon biomass, and nitrogen:carbon ratio ( JP¹ )

( _S ^S

P

SPB

D J and _A ^A

P

APB

D J for seedlings and adults, respectively). A fraction of this

nitrogen flux will be incorporated into pathogen nitrogen content with assimilation

ratio (b ). Another fraction, _P f , will accumulate as litter and thus influencing litter P

dynamics. The remaining part, 1b_P f_P , is released back to the soil nitrogen pool.

I assume that nitrogen in pathogens is released back to the soil due to natural

mortality (GP) and density-dependent competition and inhibition by mycorrhizal

fungi (EMP) (Raaijmakers et al., 2008). The dynamics of pathogens are shown in eqn.

(23)

15

4 in Table 1.

Litter

Nitrogen in plants enters the litter nitrogen pool due to mortality (natural or

pathogen infection) and tissue turnover. The amount of litter caused by mortality is

determined by the number of dead plants and their nitrogen content. Tissue turnover is

assumed to be contributed by adults only. Litter production from seedlings is

neglected due to their relatively small size. Adults uptake nitrogen for primary production at a rate of

max

(1 A )

lR A , and release it at the same rate into litter pool in

order to remain fixed individual size. In addition, litter production can increase due to

mycorrhizal-enhancement of nitrogen uptake, where the increment depends on the nutrient limitation status of mycorrhiza (indicated by l_{S A M R}_{, ,} _, in Table 2). For litter

decomposition, I assume decomposition rate (dec) is mainly determined by litter

quality (Berendse, 1994; Kurokawa & Nakashizuka, 2008). Nitrogen in litter is also lost from the system due to leaching (

φ

), which is an important characteristic of open

ecosystems (Menge et al., 2009). The equation for litter dynamics is given as eqn. 5 in

Table 1.

Soil nitrogen

(24)

16

In this model, inorganic nitrogen is considered as the main limiting resource in

the ecosystem. Nitrogen dynamics consists of external supply and loss, litter

decomposition, release from dead material of plants and microbes, and biological

uptake of plants and mycorrhizas (eqn. 6 in Table 1). Soil nitrogen pool is supplied by

a constant deposition rate I and lost through leaching with the rate of Le. Nitrogen

is released from biological components through litter decomposition, nitrogen

released following pathogen-induced mortality of plants and those resulting from

microbial turnover. Biological uptake includes both mycorrhizal and plant nitrogen

uptake. The amount of plant nitrogen uptake (i.e. Nuptake_S and Nuptake_A for

seedlings and adults, respectively) is determined by plant carbon allocation, biomass

growth and plant tissue quality (Eppinga et al., 2011). Seedlings accumulate

photosynthetic carbon as its own biomass during maturation, the amount of carbon

biomass yield is the product of the number of matured seedlings and the difference between carbon biomass of the two stage classes:

max

(1 )

A S

gRA B B A

ª º A

¬ ¼ . Adults

allocate primary production to reproduction and litter production and thus the amount

of carbon uptake can be quantified by the sum of these two processes as

> @

max

(1 )

S P

rRA B lRA A

J A , where the first term within the bracket represents biomass of new recruits and the second term is the primary production that ended up

as litter. The total amount of nitrogen uptake is determined via dividing primary

(25)

17

production by its C:N ratio.

Model Analysis and Simulation Experiments

In this study, I focused on the plant and microbial traits which are related to plant demography processes ( r , g , m_A , m_S ), litter dynamic processes (l , dec),

plant-microbe interactions (JP, DS, DA, b_P, f_P, GP , u, C_max, e_M , _γ_M , GM),

and microbial interaction (EMP, EPM ) (Table 4). I examined the interactive effects of

litter- and microbial-mediated PSF and the relative importance of traits via the

following four scenarios: (1) a system only considering litter-mediated PSF without

any direct-interacting microbes (P M 0 in Fig. 1), (2) a system considering

litter-mediated PSF and pathogens as representative of microbial-mediated PSF

(P!0, 0M in Fig. 1), (3) a system with litter-mediated PSF and mycorrhizas

(P 0, 0M! in Fig. 1) and, (4) a system with litter-mediated PSF and both

pathogens and mycorrhizas (P M, 0! in Fig. 1). In order to address the effect of

microbial community composition on relative importance of traits, I set two

interaction scenarios between pathogens and mycorrhizas for the system with both

microbes: (4a) bidirectional interaction between the two microbes due to competition

for root colonization site and (4b) unidirectional impact of mycorrhiza on pathogen

(26)

18

via producing antimicrobial metabolites. The effect of microbial community

composition on PSF can thus be obtained via comparing the results of these two

scenarios.

Empirical experiments typically quantify PSF strength via comparing biomass

growth of the target species’ seedlings in self-cultivated soil in greenhouses to those

that grown in heterospecies-cultured soils. I perform a simulation experiment to

quantify PSF strength following a similar framework. I consider two hypothetical

plant species: a reference plant species (species ref) with trait values obtained from

empirical studies (detailed parameter values shown in Table 4 and Appendix S4) and a

target plant species (species tar). I also conducted basic mathematical analysis to

check the positivity of microbe-free equilibrium and invasibility of microbes into the

microbe-free equilibrium when choosing the plausible set of parameter values

(Appendix S3). I set the target plant species to have only one trait value deviated from

the reference plant species at a time in order to identify the relative importance of

each trait on the direction and strength of PSF. Deviation of trait values was set to

±50% for scenarios without mycorrhizas. For the scenarios with mycorrhizas,

deviation range was set to ±20% in order to remain a realistic range for e_M. I also

calculate the effect size of larger deviation for other traits while still remaining in a

realistic range (10% to 300% of the references plant species value, Appendix S1).

(27)

19

I first run the model numerically to equilibrium for each microbial composition

scenario with both reference and target plant species parameter settings. This is a step

for simulating plant trait-specific cultivation of nearby soil environment for a long

period of time. Numerical simulation was carried out by C language using the

fourth-order Runge-Kutta method with a fixed interval (see Appendix S5 for detailed numerical method). Equilibrium values of soil nitrogen (R ) and microbes (_k^* P_k^* and

*

M ) are recorded to represent properties of the soil cultivated by the specific plant k

species k (k = ref or tar, stage 1 in Fig. 2). A sub-model was then prepared to simulate

the dynamics of seedling growth of plant species i in the pots filled with cultivated

soils (eqns. shown in Table 2 and 3). The sub-model is similar to seedling and adult

dynamics in the full model in that it considers growth of seedlings, growth enhancement by mycorrhiza (represented as _* _* _* _*

( , , , ) i S A M R

g in Table 2), seedling natural

mortality and pathogen infection. However, the sub-model does not consider

reproduction and shading effect of adults and the equilibrium values recorded from the full model are used to represent specific soil properties (i.e. R_k^*, P_k^* and M_k^*).

These soil properties are assumed to be constant during the simulation of the

sub-model following the recognition that seedling growth is determined by historical

plant growth legacies (Kardol et al., 2007; Kulmatiski & Beard, 2011). I run the

sub-model to equilibrium (i.e. all seedlings are matured or dead) and denote the

(28)

20

equilibrium adult density as A_{i k}^**_, , which represents the growth response of plant

species i in soil cultivated by plant species k (i and k = ref or tar, stage 2 in Fig. 2).

The PSF strength for the target plant is determined by comparing its biomass growth

in the two different soils (i.e. A_{tar ref}^**_, and A_{tar tar}^**_, ) via the formulation following

Petermann et al. (2008):

**

,

**

,

log ^{tar tar}

tar

tar ref

PSF A

A

§ ·

¨ ¸

means that the growth of the target plant species is benefitted in soils cultivated by

conspecies due to deviation of that specific trait. However, if the PSF is negative, it

means that such deviation of trait from the reference plant corresponds to a net

disadvantage on home soils compared to the reference soils.

(29)

21

Results

Litter-mediated PSF without any direct-interacting microbes

Plant tissue and litter quality traits are most important traits in determining PSF

strength compared to other plant functional traits under the scenario with

litter-mediated PSF only (Fig. 3a, effect size of ±50% deviation from the reference

value). An increase in litter decomposition rate ( dec ) and plant tissue C:N ratio (J ) P

resulted in strongest positive PSF as such increases produced larger soil nitrogen

content in the target plant species cultivated soil compared to those of the reference

plant species (i.e. 'R^*> 0, Fig. 3b). An increase in adult mortality rate (m ) caused _A

the target plant to realize negative PSF due to decreased soil nitrogen. I found

qualitatively the same pattern for even larger deviation of the trait value (i.e. 10% -

300% of the reference plant species value, Appendix S1, Fig. S1). The identity of the

most influential traits remains unchanged when other trait values, in addition to the

target trait, were randomly assigned simultaneously (Appendix S2, Fig. S6). The

direction of PSF resulting from positive deviation of traits is summarized in Table 5.

Litter-mediated PSF and pathogens

The traits related to plant defense against pathogens are most influential in

determining PSF strength under the scenario with pathogens as the only representative

(30)

22

of microbial-mediated PSF (Fig. 4, effect size of ±50% deviation from the reference

value). An increase in plant tissue C:N ratio (J ), pathogen mortality rate (_P GP) and

seedling biomass growth rate per nitrogen (g) resulted in strongest positive PSF (Fig.

4a). This positive PSF is due to a combined effect of increased soil nitrogen (i.e. 'R^*

> 0, Fig. 4b) and decreased pathogen nitrogen content (i.e. 'P^*< 0, Fig. 4c) of the target plant species cultivated soil. An increase in pathogen assimilation ratio (b_P) of

plant tissue resulted in strongest negative PSF due to increased pathogen nitrogen

content despite the accompanied increase in soil nitrogen. Increased plant

reproduction (r) also generated negative PSF, which is due to synergic effects that are

opposite to those of increased biomass growth rate. Relative importance of litter

decomposability ( dec ) on PSF strength is low under this scenario because benefits of

increased soil nitrogen are offset by the slightly increased pathogen nitrogen content. I

found the same pattern for larger deviation of the trait value (Appendix S1, Fig. S2).

The identity of the most influential traits remains unchanged when other trait values

are simultaneously randomly assigned (Appendix S2, Fig. S7).

Litter-mediated PSF and mycorrhizas

The traits related to plant tissue and litter quality, as well as those related to

mycorrhiza nutrient acquisition ability are most influential in determining PSF

(31)

23

strength under the scenario with beneficial mycorrhizal fungi as the only

representative of microbial-mediated PSF (Fig. 5, effect size of ±20% deviation from

the reference value). An increase in trait value such as plant tissue C:N ratio (J ), _P

plant carbon transfer ratio (C_max) and mycorrhiza carbon assimilation ratio (e_M) (i.e.

traits related to plant-mycorrhiza carbon exchange) resulted in strong positive PSF

(Fig. 5a). This positive PSF is due to significantly increased mycorrhiza nitrogen

content (i.e. 'M > 0, Fig. 5c), despite soil nitrogen level was depleted (i.e. ^* 'R^*^{< 0,}

Fig. 5b) to support growth of both plant and microbes. Relative importance of litter

decomposability ( dec ) is high for this scenario as an increase of decomposition rate

resulted in synergic increase of both soil nitrogen and mycorrhiza. Increase in other

mycorrhiza traits such as mycorrhiza nitrogen uptake coefficient (u) and mycorrhiza

C:N ratio (JM ) resulted in large decrease of soil nitrogen and mycorrhiza nitrogen

content, respectively, mainly due to carbon starvation of the microbe and are thus

influential in determining negative PSF. The pattern remains qualitatively the same

even for larger deviation of the trait value (Appendix S1, Fig. S3). The identity of the

most influential traits remains unchanged when other trait values are simultaneously

randomly assigned (Appendix S2, Fig. S8).

(32)

24

Litter-mediated PSF and both pathogens and mycorrhizas

The traits related to plant tissue quality and nutrient acquisition ability of

microbes are most influential in determining PSF strength under the scenario with

both pathogen and mycorrhiza (Fig. 6 and 7, effect size of ±20% deviation for

bidirectional and unidirectional competition, respectively). An increase in plant tissue

C:N ratio (J ) and traits related to plant-mycorrhiza carbon exchange (i.e. _P Cmax and

eM) resulted in strong positive PSF for both competition scenarios (Fig. 6a and 7a).

The positive PSF resulting from positive deviation of these traits was due to increased mycorrhiza relative abundance (i.e. ^'

^M^*^M^*^P^*

> 0, Fig. 6c and 7c), despite such

increase is accompanied by decreased soil nitrogen (i.e. 'R^*< 0, Fig. 6b and 7b). An increase in mycorrhiza nitrogen uptake coefficient (u) and mycorrhiza C:N ratio (JM )

resulted in negative PSF due to similar reasons in the case only with mycorrhizas.

The relative importance of some traits is different among the two competition

scenario (bidirectional Fig. 6 vs. unidirectional Fig. 7). In particular, the effect size of

litter decomposability ( dec ) is low when negative effects are bidirectional (i.e.

, 0

MP PM

E E ! , Fig. 6a), while it acts as an important determinant for PSF strength

when only mycorrhizas have negative impact on pathogens (i.e. E_MP !^{0, 0}E_PM ,

Fig. 7a). This pattern is confirmed when quantifying the effect size of litter

decomposability under different competition scenarios via continuous changes in

(33)

25

EMP and E (Fig. 8a, effect size of ±20% deviation from the reference value), _PM

which can be explained by the difference in the relative abundance of microbes (Fig.

8b). When the competition tends to be symmetrical (i.e. E_MP |E_PM) or negative

impact from pathogen on mycorrhiza is stronger (i.e. E_MP E_PM ), the relative

importance of litter decomposability is low and is accompanied with lower relative

abundance of mycorrhiza. This result corresponds to the case only with pathogens

(Fig. 4a). On the other hand, effect size of litter decomposability is high when

negative impact of mycorrhiza on pathogen is stronger (i.e. E_MP !E_PM), since under

such asymmetric interaction scenario the relative abundance of mycorrhiza is higher

(Fig. 8b). The results under these scenarios thus correspond to the case only with

mycorrhizas (Fig. 5a). The same pattern was found for larger deviation of the trait

value (Appendix S1, Fig. S4 and S5). The identity of the most influential traits

remains unchanged when other traits, in addition to the target trait, are randomly

assigned (Appendix S2, Fig. S9 and S10).

(34)

26

Discussion

The present study is a modeling attempt to identify the relative importance of

traits on PSF strength when litter- and microbial-mediated PSF are coupled. The

inclusion of microbial-mediated PSF alters the identity of the most influential traits

predicted via litter-mediated PSF due to indirect interactions between litter and

microbes. Pathogens and mycorrhizas interact with litter dynamics differently, and

thus the relative importance of traits depends on the microbial community

composition of cultivated soils.

Effects of microbial community composition on relative importance of traits

While there has been increasing interest in understanding how plant functional

traits may determine community properties and ecosystem processes (de Bello et al.,

2010), this study shows that the effect size of different functional traits on PSF

direction and strength is context-dependent. In particular, identity of the most

influential trait in determining PSF strength depends on the relative abundance of

mycorrhiza and pathogen.

When the soil lacks of both pathogenic and beneficial microbes, litter

decomposability is most influential in determining PSF strength (Fig. 3a). Higher

litter decomposition rate can release nitrogen stored in the organic litter to the soil at a

(35)

27

faster rate, resulting in nutrient rich environments and positive PSF (Fig. 3b).Other

litter feedback models agree with the results by showing species that produce easily

decomposing litter can accelerate nutrient cycling, and may gain growth advantage

over other plants if it favors nutrient-rich environments (Berendse, 1994; Miki &

Kondoh, 2002). In addition, the model shows that an increase in adult mortality rate

resulted in strong negative PSF under this model setting. I speculate this is because

higher adult mortality will produce more open canopy, which can enhance plant

population growth but decrease soil nitrogen content due to larger uptake flux, and

thus in turn has negative impact on future seedling growth.

When the soil biota is dominated by pathogens, the plant defense traits against

pathogens are influential in determining PSF strength. Plant roots are able to exhibit a

large variety of defense strategies, including physical defense via producing lignified

roots, or chemical defense through excreting secondary metabolites (van Dam, 2009;

Rasmann et al., 2011). The model shows that an increase in plant tissue C:N ratio (e.g.

increase in wood density) can result in strong positive PSF (Fig. 4a). Plants with such

trait are low in nutrition quality and thus can significantly suppress pathogen level in

the target plant-cultivated soils (Fig. 4c). This prediction is supported by Augspurger

& Kelly (1984) as they show species with higher basic wood density suffer less

disease mortality. Other defense traits that can increase pathogen mortality (e.g.

(36)

28

excrete secondary metabolites that directly act against pathogens or attract predators

of the pathogen, van Dam, 2009), or decrease pathogen assimilation ratio of plant

tissue (e.g. more lignified roots) can also generate positive PSF in the target

plant-cultivated soils.

When mycorrhizas are included in the model, the traits that characterize

plant-mycorrhiza interactions are influential in determining the strength of PSF.

Moreover, when mycorrhizas are abundant, the model predicts that litter

decomposability is also equally important as litter- and microbial-mediated PSFs

interact strongly. Plant tissue C:N ratio remains an important determining factor for

positive PSF. Plants with higher carbon transfer ratio and/or cooperating with

mycorrhizas that have higher carbon assimilation ratio also realize strong positive

PSF (Fig. 5a). In the model, carbon is assumed to be unlimited for plants. Plants with

higher C:N ratio and higher carbon transfer ratio are able to supply more

photosynthetic product to the microbe per unit nitrogen uptake. Since both plants and

mycorrhizas rely on soil nitrogen for metabolism, species with such properties can

form more beneficial plant-mycorrhiza associations and increase the mycorrhiza level

in soils (Fig. 5c). In contrast, soils that are abundant in mycorrhizas with

characteristics such as higher nitrogen uptake coefficient and/or higher C:N ratio will

result in strong negative PSF. I speculate that this is due to stronger competition

(37)

29

between plants and mycorrhiza for soil nitrogen, and larger carbon demand of the

microbe resulting from higher mycorrhizal C:N ratio, which may easily lead to

microbial carbon starvation (Wallander, 1995; Johnson, 2009). Mycorrhiza abundance

is expected to be lower for such plant-mycorrhiza associations and thus growth of

seedlings is suppressed in self-cultivated soil.

Effects of microbial composition on the relative importance of litter-mediated PSF

When combining litter- and microbial-mediated PSF caused by different

microbes, the importance of litter decomposability dramatically changes with

microbial composition. Litter decomposition contributes largely to positive PSF when

the soil contains no direct-interacting microbes (Fig. 3a). However, the importance of

litter decomposability reduces when pathogens are included as the only

representatives of microbial-mediated PSF (Fig. 4a). This is because although higher

litter decomposability of the target plant can increase plant primary production due to

enhanced nutrient cycling, increased plant population will also support more

pathogens and thus cancelling out the beneficial effects of higher decomposability.

Compared to pathogen-dominated soils, litter decomposability has strong

positive effect on PSF strength when mycorrhiza is abundant (Fig. 5a and 7a). Such

difference is due to the different interaction of pathogens or mycorrhizas with litter

(38)

30

dynamics. The hyphae of mycorrhizas can grow into soil micropores to help the host

plant obtain nutrients that would otherwise be unavailable (Veresoglou et al., 2002).

This mycorrhiza-enhanced nutrient uptake process can increase plant productivity and

result in greater litter production (Orwin et al., 2011). A higher decomposition rate of

the target plant under such scenario will thus result in faster nutrient release of

mycorrhizal-enhanced litter production back to the low-nitrogen soil, causing strong

positive PSF due to its synergic positive effects on soil nitrogen content and

mycorrhizal nitrogen content. This result is consistent with other studies

demonstrating that mycorrhizas can increase plant litter input to the soil and thus

increase soil carbon accumulation (Read & Perez-Moreno, 2003; Orwin et al., 2011).

Effects of stage structure on PSF strength and relative importance of traits

Plant individuals from different stage classes within the population may

experience different interaction strength with PSF drivers. Such differential

interaction is especially important when discussing relative importance of traits if the

soil biota mainly consists of pathogens. In particular, higher biomass growth of

seedlings results in faster transition from the vulnerable seedling stage to a

better-defended adult stage. Such characteristic can shorten the period of time when

plants are most susceptible to disease and thus creating positive PSF via suppressing

(39)

31

pathogen population size in the target plant cultivated soils (Augspurger, 1990;

Reinhart et al., 2010). The pathogen level, however, will increase due to a larger

supply of susceptible seedlings and thus create negative PSF if the target plant has

higher reproduction rate per soil nitrogen. These results are consistent with empirical

studies, which showed that plants in gap environments grow faster and thus suffer less

pathogen-induced mortality (Augspurger, 1990; Reinhart et al., 2010). Other

experimental studies also revealed that negative density-dependent mortality is

pervasive and severity of damping-off disease is higher when seedlings grow in high

densities (Augspurger, 1983; Bell et al., 2006, Bagchi et al., 2010).

Insights for exotic plant invasion success

Interactions between plant and soil had long been recognized as an important

factor determining the success of exotic plant invasion (Reinhart & Callaway, 2006).

Past studies attribute the success of invasion to either modified nutrient cycling

caused by the invader through altering litter dynamics (Miki & Kondoh, 2002;

Eppinga et al., 2011), or due to novel plant-microbial interactions (Callaway et al.,

2003; Mitchell & Power, 2003; Eppinga et al., 2006) following the invasion process.

Understanding the linkage between plant traits and PSF allows us to predict which

(40)

32

invading species is most likely to successfully invade due to its specific combination

of functional traits.

One of the dominating hypotheses connecting plant invasion to local soil biota is

the enemy release hypothesis, which states that the release from host-specific soil

pathogens in their native range contributes to the success of plant invasion (Callaway

et al., 2003; Mitchell & Power, 2003). From the invader point of view, soils in the

invaded area may be characterized as having few direct-interacting microbes (i.e. both

pathogens and mycorrhizas are absent) or higher mycorrhiza relative abundance due

to the absence of host-specific pathogens. My model indicates that under all soil

environments without pathogens, decomposition rate is an important factor for the

success of plant invasion. When the soil biota mainly consists of mycorrhizas, I

suggest that traits of plant-mycorrhiza carbon exchange are also important factors in

addition to litter decomposability. Invaders that produce easily decomposing litter are

predicted to be most powerful in invading areas with little direct-interacting microbes,

while invaders with easily decomposing litter and produces more root exudation as

carbon supply to the mycorrhiza are best able to invade such enemy-released soils.

Empirical studies support the predictions by showing that litter from the invader often

decomposes at a faster rate (Allison & Vitousek, 2004), and litter decomposition rate

is usually higher in invaded areas (Liao et al., 2008). My model thus suggests that

(41)

33

indirect interaction between litter and mycorrhiza in enemy-released soils can explain

the trend of higher litter decomposability of invaders.

Multiple possible outcomes have been documented when soils of the invaded

area are also pathogen-rich for the invaders (e.g. pathogens are generalists or brought

by the invader along invasion). One possible outcome is that the invasion may fail as

the invader encounters pathogens in the invaded area which prevent the invasion, a

situation related to the biotic resistance hypothesis (Mitchell & Power, 2003). Another

possible scenario is termed pathogen accumulation hypothesis. Under this scenario

invasion can still succeed if the invader promotes growth of pathogens that have a

stronger negative effect on the surrounding native plants than on the invader itself

(Eppinga et al., 2006; Mangla et al., 2008). My model indicates that when the invaded

area contains high pathogen level, defense traits against pathogens plays a key role for

successful invasion. In particular, the model predict that plants that produce higher

wood density and/or exhibit other defense strategies are able to invade despite high

level of pathogen in the invaded area. This may also answer why in some cases higher

litter decomposition rate is not associated with invasion status (Drenovsky & Batten,

2007, Kurokawa et al., 2010) since litter decomposability is not an important

determinant of PSF in soils with high pathogen level. In conclusion, my model

suggests that the contrasting trait characteristic of invaders at different sites can be

(42)

34

explained by the difference in soil microbial community composition of different

invaded area.

Future work and Conclusion

To my best knowledge, this study is the first attempt to link plant functional

traits with PSF strength through a modeling approach. The model presented is novel

in its ability to combine litter-mediated and microbial-mediated PSF, also in its ability

to give predictions by simulating field experimental settings. I highlighted the

importance of indirect effects of microbes on litter dynamics and demonstrated that

separation of plant stage classes and microbial functional groups is necessary for

understanding the determinants of PSF strength. In particular, the effect of litter

decomposability increases with the relative abundance of mycorrhizas. Past

trait-based ecological studies tried to predict the shifts of ecological and ecosystem

functions along abiotic gradients through focusing of the changes in values of specific

traits (McGill et al., 2006). My model results would challenge this traditional research

framework by arguing that some traits may lose its impact on ecological and

ecosystem processes along the abiotic gradient since its importance is also determined

by indirect biotic interactions with other species. In the case of PSF, my results

suggest a closer look in the microbial community composition is necessary for

(43)

35

development of trait-based approach. While many studies had revealed a systematic

change of bacteria to fungi ratio across large geographic scale, I suggest that further

studies targeted to reveal the ratio between detrimental and beneficial microbes across

different ecosystems (e.g. from tropic to temperate regions) may provide knowledge

for understanding the frequency distribution of functional traits (Manning et al., 2008).

Experimental studies often show that the response of species’ PSF strength to nutrient

enrichment is species-specific, suggesting that further studies focusing on more

detailed characteristics of the plant, in addition to species identity, are needed. My

trait-based approach can thus provide insights and pinpoint some potential traits

which could be important PSF determinants.

Future works that extend the model to include multiple plant species interaction

may reveal the importance of species-specific combination of litter quality and

nutrient competition strategy (Berendse 1994, Miki & Kondoh 2002, Miki et al.,

2010). Results from such models may give valuable predictions about the role of PSF

in determining plant coexistence and the relative abundance patterns in plant

community. I believe that integration of this new model framework with modeling

aboveground and belowground interactions mediated through plant induced response

(Wardle et al., 2004; van der Putten et al., 2009) can also contribute to a solid and

(44)

36

improved theoretical framework for understanding the role of functional traits in

controlling plant community development.

(45)

37

References

Alvarez-Loayza P, Terborgh J. 2011. Fates of seedling carpets in an Amazonian floodplain forest: intra-cohort competition or attack by enemies? Journal of Ecology 99: 1045-1054.

Allison SD, Vitousek PM. 2004. Rapid nutrient cycling in leaf litter from invasive plants in Hawai’i. Oecologia 141: 612-619.

Augspurger CK. 1983. Seed Dispersal of the Tropical Tree, Platypodium Elegans, and the Escape of its Seedlings from Fungal Pathogens. The Journal of Ecology 71: 759-771.

Augspurger CK, Kelly CK. 1984. Pathogen mortality of tropical tree seedlings:

experimental studies of the effects of dispersal distance, seedling density, and light conditions. Oecologia 61: 211-217.

Augspurger CK. 1990. Spatial patterns of dampinf-off disease during seedling recruitment in tropical forests. In: Burdon JJ, Leather SR, eds. Pests, Pathogens and Plant Communities. Oxford, UK: Blackwell Scientific Publications, 131-144.

Bagchi R, Swinfield T, Gallery RE, Lewis OT, Gripenberg S, Narayan L, Freckleton RP. 2010. Testing the Janzen-Connell mechanism: pathogens cause overcompensating density dependence in a tropical tree. Ecology Letters 13:

1262-1269.

Bell T, Freckleton RP, Lewis OT. 2006. Plant pathogens drive density-dependent seedling mortality in a tropical tree. Ecology Letters 9: 569-574.

de Bello F, Lavorel S, Díaz S, Harrington R, Cornelissen JHC, Bardgett RD, Berg MP, Cipriotti P, Feld CK, Hering D, et al. 2010. Towards an assessment