H

Ja

b

c

d

a

AA

KEWCVSC

1

t(tGm(zauw(i

0h

Ecological Modelling 249 (2013) 3– 18

Contents lists available at SciVerse ScienceDirect

Ecological Modelling

jo ur n al homep ag e: www.elsev ier .com/ locate /eco lmodel

ydrological controls on carbon metabolism in wetlands

anaine Z. Coletti a,∗, Christoph Hinza,b, Ryan Vogwill a,c, Matthew R. Hipseya,d

School of Earth and Environment, The University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, AustraliaHydrology and Water Resources Management, Brandenburg University of Technology, Konrad-Wachsmann-Allee 6, D-03046 Cottbus, GermanyDepartment of Environment and Conservation, 17 Dick Perry Avenue, Kensington, WA 6983, AustraliaCentre for Ecohydrology, The University of Western Australia, Crawley, WA 6009, Australia

r t i c l e i n f o

rticle history:vailable online 17 August 2012

eywords:cohydrological modeletland metabolism

arbon storageegetationemi-aridlimate change

a b s t r a c t

Governed by a series of non-linear feedback mechanisms among water, vegetation and decomposers,carbon storage within wetlands is important on a global scale. However, the effect that climatic fluc-tuations have on those mechanisms is not well documented. In this study, we introduce a mechanisticmodel connecting hydrology, vegetation and microbial biomass to investigate how changes in the cli-mate signal propagate through wetland ecosystems, via vegetation and microbial dynamics, and attemptto quantify how net rates of wetland carbon metabolism change in response to a changing climate. Ourparticular focus is the dryland–wetland systems found in south-west Western Australia (SWWA), as theyare expected to be sensitive to projected climatic changes due to their close linkage to the seasonal waterdelivery pattern. The model simulations investigate wetland carbon retention under different hydro-climatological conditions ranging across a regional gradient in the dryness index. The results indicate thatshort term and long term vegetation responses may be counter-intuitive due to adaptability in the wateruptake strategy of the vegetation community partially decoupling biomass from water availability. Fur-thermore, changes in water delivery are not a good indicator for overall changes in wetland metabolism,defined as the net rate of carbon assimilation, due to the dominance of the soil carbon storages and their

sensitivity to heightened bacterial metabolism rates with increasing temperatures. The results highlightthat an optimum combination of water supply and vegetation leads to a higher percentage of carbonbeing stored in soils, therefore increasing the resistance of the carbon storage to changes in precipitation.The model presented here provides a first step to explain how changing patterns of rainfall, temperatureand evapotranspiration can change carbon cycling characteristics and the carbon retention efficiency of dryland–wetlands.

. Introduction

In wetlands, high carbon accumulation rates are related toheir typically high vegetation productivity and low organic carbonC) decomposition (Bridgham et al., 2006). Low C decomposi-ion rates are promoted by waterlogged conditions (Dean andorham, 1998) that limit oxygen (O2) diffusion into the sedi-ent profiles, decreasing the efficiency of decomposition pathways

Reddy and DeLaune, 2008). In addition to a standing waterone, wetlands are also characterised by an intermittently floodedrea (saturated soil zone) and a terrestrial fringe area (unsat-rated soil zone). The boundaries of these zones partition theetland environment, thereby shaping the vegetation zonation

Coletti et al., submitted for publication), since species with sim-lar water uptake strategy tend to assemble spatially in accordance

∗ Corresponding author. Tel.: +61 8 6488 6867; fax: +61 8 6488 1037.E-mail address: coletj01@student.uwa.edu.au (J.Z. Coletti).

304-3800/$ – see front matter © 2012 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.ecolmodel.2012.07.010

© 2012 Elsevier B.V. All rights reserved.

to zones of distinct hydrological function (Mallik et al., 2001; Capon,2003).

The environment partitioning is dynamically changing accord-ing to seasonal and episodic changes in the boundaries ofhydrological zones that result from the interactions betweenhydrological pulses and vegetation responses, as mediated by soilproperties (Coletti et al., submitted for publication). Wetland veg-etation is adapted to the characteristic ‘hydroperiod’, defined asthe temporal extent of the maximum water level fluctuation, thata climate in dynamic equilibrium creates (Carter et al., 2005; Leighet al., 2010). Indeed, the hydroperiod of a wetland represents animportant seasonal hydrological pulse that shapes variability insoil moisture and drives wetland vegetation diversity and produc-tivity. Accordingly, the hydrological pulses also govern wetlandmetabolism, defined as the difference in the rate of C input andoutput (Tuttle et al., 2008). For example, when a system is accumu-

lating C with an input greater than output (i.e., net autotrophic), themetabolism is positive, and conversely when the system is releasingmore C than it is retaining (i.e., net heterotrophic), the metabolismis negative.

4 al Mo

rataci(e1aofa2mBcv

rvCteatu(la(sscMcuwhiatets

ch1cceAcaeoC(Taww

t

J.Z. Coletti et al. / Ecologic

If the climate is non-stationary, the dynamic (seasonal) equilib-ium that exists between water availability, vegetation productivitynd rates of decomposition can be disrupted, potentially leadingo shifts in ecosystem state. Although wetland vegetation toler-tes high fluctuations in soil aerobic and moisture levels, persistentonditions outside of the natural range that characterises a systemn equilibrium (Mitchell et al., 2009), such as long term floodingCarter et al., 2005; Reddy and DeLaune, 2008) or drought (Gillist al., 2009), can cause vegetation mortality (Nemani and Running,989; Horner et al., 2009). The shift may in fact lead to an undesir-ble but resilient state, such as when mortality of mature vegetationccurs and the loss of its niche is persistent (Coletti et al., submittedor publication). Since both vegetation mortality (Stephens, 2005)nd changes in hydrology (Reddy and DeLaune, 2008; Tuttle et al.,008; Page and Dalal, 2011) affect C budgets, a non-stationary cli-ate can potentially change the wetland metabolism as a whole.

esides changes in metabolism and C storage, changes in climatean affect other ecological services provided by wetlands, such asegetation assemblage and abundance.

Wetlands in dryland regions are exposed to high evaporationates, and therefore have a low ability to store water. Thus, theiregetation productivity and net metabolism dynamics (Nielsen andhick, 1997) closely mirror the randomness of the climate and/orhe flows that are responsible for their maintenance (Puckridget al., 1998; Tooth and McCarthy, 2007). As a result, the metabolismnd the vegetation dynamics in dryland–wetlands are expectedo be particularly sensitive to changes in climate and flow reg-lation that will lead to changes in the hydrological signatureKingsford, 2006). It therefore follows that the sensitivity of wet-and metabolism to changes in climate state will vary along anridity gradient, based on a metric such as the dryness indexDI = Ep/P). Whilst there has been an increase in the number oftudies related to wetland metabolism (Bridgham et al., 2006),ystem-scale quantification of the mechanisms that control Cycling in wetlands has not been well documented (Rodríguez-urillo et al., 2011; Page and Dalal, 2011). This is especially the

ase for dryland regions (Humphries et al., 2011) and it remainsnclear how changes in climate and water delivery can affectetland C cycling (Davidson and Janssens, 2006). For example,ow sensitive is the carbon cycle to different scales of variability

n hydro-climatological conditions? How important is vegetationdaptability in mediating the overall wetland metabolism? It ishe aim of this study to characterise how wetland metabolism, andcosystem services such as carbon storage and vegetation produc-ivity, respond to climate variability along an aridity gradient in theouth-west region of Western Australia (SWWA).

Wetlands in SWWA can potentially experience a significanthange in their C stocks and net C metabolism. The region alreadyas experienced a significant decline in rainfall and runoff since the970s (Petrone et al., 2010), and the already highly variable hydro-limatology is expected to be intensified, due to further predictedhanges in rainfall volume, an intensification of extreme rainfallvents and an increase in temperature (CSIRO, 2009; Nicholls, 2004;lexander et al., 2007). Such an intensification in the hydrologicycle has the potential to increase the severity of both droughtnd waterlogging extremes (e.g., Ojima and Lackett, 2002; Johnsont al., 2004), and both have been reported to cause significant lossf wetland vegetation in the Australian context (Gillis et al., 2009;arter et al., 2005). Temperature increase and longer dry periodsBurkett and Kusler, 2000) are likely to enhance C soil release.his phenomenon has already been observed in drained wetlandsround Australia (Page and Dalal, 2011). Given the high C stocks in

etland soils, this fact could serve as a positive feedback to globalarming.

In a previous study, an ecohydrological model that resolveshe dynamic partitioning of a wetland into different hydrological

delling 249 (2013) 3– 18

functional zones was developed for the SWWA region to char-acterise vegetation dynamics in dryland–wetlands (Coletti et al.,submitted for publication). Here, we extend this model to accountfor carbon cycling to test the hypothesis that water and carbonstorage are not necessarily positively related, since the variousbiotic and abiotic feedback mechanisms that ultimately define wet-land metabolism are highly complex and non-linear. Additionally,we test the hypothesis that changes to wetland C storage andmetabolism will vary in response to changes in precipitation alongan aridity gradient.

Our approach is to use the model in different contexts to pro-gressively develop our understanding of how the signal imposed byclimate propagates through vegetation and decomposers and howthe feedbacks between vegetation and hydrology can modulate theclimate signal. Firstly, the manifestation of climate on environmentpartitioning, and consequently C dynamics and metabolism, isinvestigated using meteorological data from three weather stationsfound across the SWWA climate gradient. Secondly, we drive themodel using synthetic rainfall created by a stochastic rainfall gener-ator to unravel the ecohydrological conditions that benefit carbonstorage in wetlands. Conducting this analysis under a controlled(synthetic) climate helps us to clarify the feedback mechanismamong biotic (vegetation and decomposers) and abiotic (water)variables that define C cycling and consequent metabolism in wet-lands. Thirdly, to observe the general trend in wetland metabolismand C storage given a decline in precipitation, we consecutivelydecrease the total annual rainfall depth of the synthetic climaterealizations for an extended period. In this test we systematicallyreduce rainfall by 5% for every 10 years, so that after a 40 yearsperiod the total annual precipitation was 20% less than in the begin-ning of the simulation, increasing the DI accordingly. This is a majorrainfall decline, but representative of the 15–20% rainfall reductionalready experienced in the region (Petrone et al., 2010). Finally, weassess the combined effect of an increase in the mean annual tem-perature and a decrease in rainfall depth on wetland metabolism.Here the response of the vegetation community and decompositionrates is assessed under a likely future climate scenario, assuming atemperature increase of 0.5 ◦C and a precipitation decline of 5% foreach decade.

2. Model description

The C cycling model implemented for this study was adaptedfrom Porporato et al. (2003) and coupled to an ecohydrologi-cal model simulating the wetland water balance and vegetationdynamics (Coletti et al., submitted for publication; Appendix A).Together the coupled model simulates surface water level, watertable dynamics and soil moisture, in conjunction with the carboncycling processes including vegetation productivity, plant litter-fall and respiration, and metabolism of detrital C by soil microbes.The water and carbon balance are driven by daily climate forcingrepresented by relative humidity, precipitation, temperature, solarradiation and wind speed. Three plant functional groups directlyrespond to climate via solar radiation, relative humidity and airtemperature, and indirectly through water availability. Plant litter-fall supplies C to the soil as detritus and which is consumed by themicrobial population, which responds to soil water content (usedas a proxy for O2 concentration), and soil temperature. No competi-tion between plants and microbes for nutrients is considered in thisinvestigation. The constitutive equations for the model form a setof ordinary and partial differential equations. As the length of the

simulations is much longer than the time-step, a simple 1st orderexplicit finite difference method was used as a suitable approxi-mation to the final solution, and the model was implemented inMATLAB (Mathworks Inc.).

al Mo

2

rwp(wdaamwsatl

wtrotfatlt(ciad

ufsdAcRp

2

2

gpAidnc

J.Z. Coletti et al. / Ecologic

.1. Water budget

The extent of the areas of similar hydrological function (the envi-onment partitioning) is conceptually defined by the position of theater table and lake level, and constrained by the wetland mor-hology. They are defined as the surface area of unsaturated soilAU), the surface area of saturated soil (AS) and the area of standingater/lake (AL) (Fig. 1). The water table and lake level are resolvedaily, as driven by climate conditions and losses to deep rechargend vegetation transpiration, which in turn is a function of watervailability. The internal hydrological fluxes are represented in theodel through linking three conceptual water storages: the openater/lake (L) and the water contained in the unsaturated (U) and

aturated (S) zones of the soil. Water fluxes in the model are defineds length of water per time (m d−1) and multiplied by relevant areaso convert to volume (m3). dL/dt (m3 d−1), which is the variation ofake volume with time, is calculated as:

dL

dt= PL + Qc + Qw ± Qs − EL − Qout (1)

here PL is the volume of precipitation that enters the lake, EL ishe lake evaporation rate integrated over the lake area, Qc is theiver inflow from the catchment area, Qout is the volume in excessf the lake maximum which exits the lake as outflow and QS ishe seepage that flows through the area of lake base (ASL) and theree saturated area (AS), which can be an inflow to the saturatedrea (i.e., a positive term) if the lake level is higher than the waterable level or an outflow if the water table is higher than the lakeevel (i.e., a negative term). Qw is surface runoff generated withinhe wetland domain and parameterised as the sum of infiltrationQie) and saturation (Qse) excess mechanisms. Infiltration excess isalculated as the amount of precipitation reaching the ground thats greater than the capacity of the soil to infiltrate. The water bal-nce in the unsaturated and in the saturated portion of the wetlandomain is defined respectively as:

dSUSdt

= 1 − EU − Eb U − Qp (2)

dSSATdt

= Qp ± Qs − ES − EbS − Qse − Qss (3)

In (2), I is the infiltration rate, EU is the transpiration from thensaturated zone, EbU is the bare soil evaporation via capillarityrom AU and Qp is the percolation from the unsaturated zone to theaturated soil zone. In (3), ES is the transpiration from AS by theifferent vegetation groups, EbS is the bare soil evaporation fromS, Qse is the saturation excess (the volume that surpass the soilapacity) and Qss is the groundwater recharge and Qs is the seepage.efer to Appendix A for a detailed summary of the water balancearameterisations.

.2. Carbon dynamics

.2.1. Vegetation biomassBiomass (B, kg C) dynamics are simulated for different functional

roups of vegetation, each governed by a group specific rate ofhotosynthesis, loss due to litterfall, root death and respiration.lthough the biomass of each vegetation type i, Bi, changes in time,

t is spatially stationary and experiences different hydrological con-itions depending on the distribution of zones at any given time,

(n = U, S or L). As the spatial extent of each hydrological areahanges, the amount of any vegetation type, i, present within each

delling 249 (2013) 3– 18 5

environment must be redistributed accordingly. The balance equa-tion for each vegetation type is therefore:

dBi,ndt

=

⎧⎨⎩

(˘Ai,n− Lli,n − Ri,n − Rdi,n )An + Di,n−1

dAndt

ifdAndt

> 0

(˘Ai,n− Lli,n − Ri,n − Rdi,n )An + Di,n

dAndt

ifdAndt

< 0

(4)

where Ll is the loss due to litterfall (kg C m−2 d−1), Rd is rootdeath (kg C m−2 d−1) and ˘A is the gross assimilation of carbon(kg C m−2 d−1), defined as in Running and Coughlan (1988). Thelatter is defined as a function of the non-dimensional uptakeefficiency, �˘ , the potential uptake rate, ˘0 (m d−1), �CO2, thecarbon dioxide air – leaf diffusion gradient (kg C m−3) and LAI(m2 leaf (m land)−2) (Lohammar et al., 1980). Plant respiration, R(kg C m−2 d−1), is configured as a function of temperature accord-ing to Running and Coughlan (1988). D is the carbon density perunit area (=B/A) and An is the area of the nth wetland zone (m2),and the last term in Eq. (4) represents the redistribution of vegeta-tion into the different hydrological zones as the area of inundationvaries.

Whether a specific vegetation functional group can uptakewater from below or above the phreatic level (or from both regions)depends on its uptake strategy. As a result, each plant type can sur-vive best in a particular hydrological environment (U, S or L) thatmatches its preferred uptake strategy. Therefore, different patternsof environment partitioning result in the evolution of the vegeta-tion community distribution and consequent wetland metabolism.The plant groups considered in this study are trees (V2), grasses(V3) and aquatic vegetation (V1). Grasses are exclusively adaptedto unsaturated soils, aquatic vegetation requires standing water,and trees are adapted to take water from both saturated andunsaturated areas of the soil profile (Coletti et al., submitted forpublication).

2.2.2. Carbon decompositionCarbon lost from vegetation via litterfall (Ll) and root turnover

(Rd) becomes part of either the C in the soil, or the lake pool ifthe trees are situated within the zone of standing water. The totalamount of carbon in the soil and in the lake that is not vegetationbiomass is termed belowground storage (CB). As the C present in thelake is generally low compared to the sediments and wetland soil,CB also accounts for the lake detrital carbon pool. It is partitionedbetween particulate (POC) and dissolved (DOC) organic fractions,and considers the fraction within the bacterial biomass (POCb).

The same spatial allocation approach used to partition the veg-etation biomass is applied to CB and its components, so that itis resolved in each hydrological zone and dynamically varies inresponse to changes in the water table and lake level (Fig. 1). Assuch, the fraction of CB that is within the unsaturated conditions istermed CU, and similarly for CS and CL, for carbon allocated to thesaturated or standing water environments, respectively. Within thewetland domain the amount of C is conserved and CB = CU + CS + CL.

As a simplification, the total volume where organic C can befound in the soil (Vsed) is constant and equal to the total wetlandarea multiplied by a sedimentation depth, hC, arbitrarily assumedin this study as 1 m. Vsed comprises an unsaturated and a satu-rated zone. The volume of sediments under unsaturated conditions(VU) is defined as the unsaturated area (AU) multiplied by hC. Con-versely, the volume of sediments under saturated conditions (VS)is defined as the area of saturated soil (AS) plus the lake area (AL),both multiplied by hC. The lake itself also maintains a carbon poolwithin its volume (L). The concentration of carbon is averaged overeach one of these areas, i.e., the minimum spatial resolution for C

decomposition is VU, VS and L.

Conceptually, contribution of carbon via Ll and Rd from theterrestrial zone, AU, is added to CU and undergoes decomposi-tion under oxic conditions. Roots that are lost below water table

6 J.Z. Coletti et al. / Ecological Modelling 249 (2013) 3– 18

F d sysr saturl

ldbfatClt

pcpifp

ut(ta

wb

�

ig. 1. Representation of the major fluxes that connect carbon pools in the wetlanepresented by the white area, the unsaturated sediment (VU) in light grey and theine. The arrows represent the major carbon fluxes.

evel and litterfall within the area AS contribute to CS and undergoecomposition under anoxic conditions. This includes all the rootiomass that is lost from plants belonging to the lake area and araction (=fsed) of the leaves that are lost within the lake area thatre assumed to reach the lake sediment. The leaves that remain inhe water column (1 – fsed) are retained within the lake carbon pool,L. In this conceptual model, organic carbon is decomposed under

ower rates when exposed to saturated conditions as it is assumedhis is a proxy for anaerobic decomposition pathways.

The carbon that reaches the soil pools, CU and CS, and the lakeool, CL (g C m−3), are subject to decomposition by bacteria, witharbon released back to the atmosphere as CO2 due to bacterial res-iration, Rb (g C m−3 d−1). In the soil, production of DOC occurs and

s mobilised via percolation (PC) or leaching (LC), or can be trans-erred to or from the lake via seepage (CQ). As such, this term can beositive or negative, depending on the lake and water table level.

Also, as the water table level changes, POC components from thensaturated and saturated pools are reallocated between them. Inhe lake, part of the total carbon pool is lost via lake overflow, Cout

Fig. 1). The differential equations that describe the carbon pool inhe lake and in the unsaturated and saturated soil zones are defineds:

dCLdt

= (1 − fsed)3∑i=1

L|i,LALL

± CQ − Cout − Rb (5)

dCUdt

=3∑i=1

(L|i,U + Rdi,U )AUVU

+ �U − LC − Rb (6)

dCSdt

= (fsed)3∑i=1

L|i,LALVS

+3∑i=1

(L|i,S + Rdi,S )ASVS

+ �S − PC − Rb ± CQ

(7)

In the above equations, � represents the distribution function,hich is the reallocation of C due to changes in water table level

etween CU and CS. �U and �S are formulated as follow:

U = CS(VtU − Vt−1U )

VtU(8)

tem. The three C pools (CU , CS and CL) are depicted in brackets. Lake volume (L) isated sediment (VS) in dark grey. Water table level (hS) is depicted as a dashed dot

�S = CS(VtS − Vt−1S )

VtS(9)

The organic C pool from any hydrological zone n = U, S or L, Cn,comprises both the DOC and POC components with POC furtherdivided into three pools: labile, refractory and microbial biomass,addressed by the subindexes l, r and b, respectively, such that thecarbon in each zone is Cn = POCl,n + POCr,n + POCb,n + DOCn.

The labile and refractory POC pools are assumed to be char-acterised by unique rates of microbial decomposition (Porporatoet al., 2003), with the labile, refractory and dissolved C set to break-down at rates of �l, �r and �d, respectively. Conceptually, a fractionof the labile carbon becomes refractory, fr, and a fraction of therefractory portion enters the dissolved pool, fd, to account for theprocess of humification. The rest is consumed by bacteria and ulti-mately respired (Fig. 2).

The differential equation that describes the dynamics of labilePOC (g C m−3 d−1) within the CL, CU and CS pool is as follows:

dPOClLdt

= (1 − fsed)3∑i=1

L|i,LALL

− �lL − POClout (10)

dPOClUdt

=3∑i=1

(Lli,U + Rdi,U )AUVU

− �lL + �UPOClnCn

(11)

dPOClSdt

= (fsed)3∑i=1

Lli,LALVS

3∑i=1

(Lli,S + Rdi,S )ASVS

− �lL + �SPOClnCn

(12)

and the dynamics of refractory POC (g C m−3 d−1) within the CL, CU

and CS pool are similarly defined as:

dPOCrLdt

= frL�lL − �rL − POCrout (13)

dPOCrU POCrn

dt

= frU �lU − �rU − �U Cn(14)

dPOCrSdt

= frS �lS − �rS − �SPOCrnCn

(15)

J.Z. Coletti et al. / Ecological Modelling 249 (2013) 3– 18 7

F ent, na

p

aa

ltr

C

�

�

�

lbt(cfte

f

f

wso

ig. 2. Representation of the organic carbon cycling in each hydrological environmnd dissolved C.

Dissolved organic carbon (g C m−3 d−1) within the CL, CU and CSools are described as:

dDOCLdt

= fdL�rL − �dL − DOCout (16)

dDOCUdt

= fdU �rU − �dU + �UDOCnCn

(17)

dDOCSdt

= fdS �rS − �dS + �SDOCnCn

(18)

nd the dynamics of the bacterial population (g C m−3 d−1) defineds:

dPOCbLdt

= (1 − frL )�lL + (1 − fdL )�rL + �dL − RbU − POCbout (19)

dPOCbUdt

= (1 − frU )�lU + (1 − fdU )�rU + �dU − RbU + �UPOCbnCn

(20)

dPOCbSdt

= (1 − frS )�lS + (1 − fdS )�rS + �dS − RbS + �SPOCbnCn

(21)

As a simplification, bacterial mortality is not explicitly formu-ated, and the respiration rate, Rb, is used as a generic biomass losserm, with the rate of microbial respiration (g C m−3 d−1) linearlyelated to POCb.

The decomposition rates for the labile, refractory and dissolved are parameterised respectively as:

ln = klPOCln

[POCbnkbn+POCbn

]f�fT (22)

rn = krPOCrn

[POCbnkbn+POCbn

]f�fT (23)

dn = kdDOCn

[POCbnkbn+POCbn

]f�fT (24)

where the term in brackets accounts for bacterial populationimitation, kl, kr and kd represent decay coefficients, with kl set toe greater than kr to reflect the characteristic higher resistanceo decay of refractory organic matter if compared to labile OMD’Odorico et al., 2003), and kb (g C m−3) is the maximum bacterialoncentration under unlimited OM supply. In the above equations,

� and fT represent the sensitivity to OM decomposition in responseo varying soil moisture (�) and temperature (T) conditions (Yangt al., 2002):

� = 1 −(

1

�2b

) (�

�b

)2

(25)

T = 0.157e−0.0512T (26)

here soil moisture is given in volume of water per volume of bulkoil (m3 m−3), temperature in ◦C and �b (m3 m−3) represents theptimum soil moisture for bacterial decomposition.

(n = U, S or L); �l , �r and �d represent the decomposition rates of labile, refractory

3. Model application

The hydrogeomorphic and climatic context of the simulatedwetland is typical of SWWA conditions. It is schematically rep-resented by a low slope, relatively deep soil base with low claycontent, and intermittently receiving river inflow from the catch-ment as hydrological pulses (Merz, 2000). Evapotranspiration andoverflow are the main simulated loss terms, and loss to the regionalgroundwater is considered to be less important than the otherfluxes.

Vegetation parameters were chosen to adjust values for pho-tosynthesis, respiration and litterfall of plants found in wetlandsin SWWA (Finlayson, 2005; Froend and McComb, 1994) or otherfloodplain wetlands found in Australia (Robertson et al., 1999).Also, the relation between photosynthesis and respiration fromunderstorey and overstorey vegetation was kept as reported insavannas in Australia (Kanniah et al., 2010; Hutley and O’Grady,2001). Appendix B provides a parameter summary and justificationand initial conditions for the model setup used here.

In addition to the direct rainfall contribution to the wetlanddomain, a simplified rainfall-runoff model (Farmer et al., 2003)driven by the same meteorological data was applied in the sur-rounding catchment to generate inflow pulses to the wetland.In this case, the resultant one-dimensional runoff rate (includingsurface and baseflow components) from the catchment, Rc, wasmultiplied by the total surrounding catchment area, Ac to generateQc.

A summary of parameters for the carbon decomposition modelused in all simulations can be found in Table 1. The objective of thisstudy is not to validate the model against field data and its purposeis purely to further our understanding of the system’s behaviourrather than quantitative prediction. Nonetheless, the results forcarbon accumulation in the soil are comparable to values foundin other wetlands in Australia, which has been reported between6.5 kg C m−2 and 28.3 kg C m−2 (e.g., Robertson et al., 1999; Page andDalal, 2011) and in Mediterranean climate wetlands (Rodríguez-Murillo et al., 2011); the results presented in Fig. 3 are within thisrange.

4. Pathways of carbon metabolism along an aridity gradient

To verify the manifestation of the climate gradient on carbonallocation, primary production and system metabolism, daily fielddata from three indicative stations within SWWA were used to runthe model for 19 years: C1 – Cape Leeuwin, C2 – Lake Toolibin andC3 – Northern Cross. The three locations were specifically chosen tofall into a geographical transect from northeast to southwest thatencompass the range of DI found in SWWA (Table 2). Daily rain-

fall, air temperature, wind speed, cloud cover and relative humidityobtained from these weather stations were used to drive the modeland each were assigned the same morphology similar to Lake Tooli-bin at site C2 (refer to Coletti et al., submitted for publication) in

8 J.Z. Coletti et al. / Ecological Modelling 249 (2013) 3– 18

Table 1Summary of the non-dimensional parameters for the carbon decomposition model.

Parameter description Symbol and value Reference/remarks

Rate of mortality of bacterial biomass kbU = 0.05; kbS = 0.055; kbL = 0.055a Mortality is assumed to be higher under low oxygenenvironment

Fraction of the litterfall that goes to VS fsed = 0.9 90% of the leaves that fall on the lake surface undergosedimentation

Rate of POCl decomposition klU = 0.025 × 10−3; klS = 0.009 × 10−3;klL = 0.008 × 10−3

In the U zone, bacterial decomposition is more efficient, asit is assumed the presence of oxygen

Rate of POCr decomposition krU = 1.6 × 10−6; krS = 0.6 × 10−6; krL = 0.7 × 10−6 Lower values than kl to reflect the higher resistance todecay of POCr in comparison to POCl

Rate of DOC decomposition kdU = 1.6 × 10−6; kdS = 0.6 × 10−6;kdL = 0.7 × 10−6

Fraction of POCl that goes to POCh frU = 0.1; frS = 0.1; frL = 0.1Soil moisture for optimum bacterial

decomposition�b = 0.6 m3 m−3 Yang et al. (2002)

Fraction of POCh that goes to DOC fdU = 0.01; fdS = 0.01; fdL = 0.01Fraction of microbial respiration that reaches

the atmospherefaU = 1; faS = 1; faL = 1 As a simplification, no “trapping” of gaseous carbon is

assumed

a U, S and L represent the environment of relevance.

Fig. 3. Water availability was not a proxy for C soil storage in the simulated wetlands: (a) fraction of inundated area (grey bars) and relative water table (black line), (b) theLAI of trees, grasses and aquatic plants, depicted in dark grey, light grey and grey bars, respectively, (c) carbon storage in the lake and soil (grey bars) and microbial respiration(black line), and (d) wetland metabolism (grey bars) and vegetation metabolism (black line).

J.Z. Coletti et al. / Ecological Modelling 249 (2013) 3– 18 9

Table 2Climate characteristic of three catchment locations in SWWA.

Scenario DI Indicative location Bureau of meteorology(BoM) station

Pannual (mm) EPannual (mm)

C1 1.2 Cape Leeuwin 009518 991.8 1200Long. 34◦20′44.92′′S/Lat. 115◦08′16.82′′E

C2 3.8 Cape Leeuwin 010614 478.7 1800◦ ′ ′′ ◦ ′ ′′

oc

lspoatp

fsmof0rwt7simdrotop

paiwawcAcrtadbtms

wAmapt

was not related to C soil density or linearly related to DI. Althougha higher primary production led to a higher input into soil C stocks,factors such as water availability and temperature defined C soil

0.1

0.2

0.3

f T * f θ

C1C2C3

Long. 32 55 15.17 S/Lat. 117 36 29.15 EC3 6.2 Northern Cross

Long. 31◦13′56.71′′S/Lat. 119◦19′54.57′′E

rder that we could isolate the effects of climate on vegetation andarbon dynamics.

The results showed no positive correlation between water tableevel and soil carbon storage in a long term simulation (for the entireimulation period). However, over short term periods (over a cou-le of years), an increase in water storage resulted in an increaser a decrease in C stocks (Fig. 3a and c), depending on hydrologicalnd biological (vegetation and microbial biomass) conditions prioro the water storage rise and the magnitude of the hydrologicalulse, discussed further below.

Overall, the three locations presented a positive net metabolismor the period of simulation, i.e., more C entering than leaving theystem during most of the simulation time (Fig. 3d). The positiveetabolism is caused by different reasons. In the particular case

f Northern Cross (C3), this overall positive metabolism resultsrom a general increase in daily average precipitation from 0.76 to.96 mm, over the first to last 5 year period, respectively. This 20%ainfall increase throughout 19 years of simulation increased theater table level, the soil moisture and the vegetation cover, with

he latter rising by approximately 50%. CB, however, decreased by% in the period (Fig. 3c). Indeed, the C3 scenario was the only tohow a consistent trend of soil C loss, since the microbial activityncreased with the improved hydrological conditions (higher soil

oisture) and by the increase in C soil input by the increasinglyense vegetation. However, as vegetation dynamics were closelyelated to the whole system metabolism, explaining more than 99%f the total C variation, scenario C3 still had a net C accumulation forhe period considered. Metabolism in the C3 wetland was negativenly when dry periods caused plant stress and respiration overtookhotosynthesis, as occurred during the summer peak in most years.

In the case of the Cape Leeuwin climate (C1), large hydrologicalulses caused significant changes in the water table level, whichffected vegetation and microbial biomass, causing major variabil-ty in metabolism. The first hydrological pulse identified in Fig. 3a

as responsible for a mortality of ∼20% of the plant population in period of approximately 1.5 years. During this event, the totaletland domain was waterlogged, generating persistent anoxic

onditions in areas previously occupied by no tolerant species.lthough aquatic plants increased their population under floodedonditions, trees and grasses decreased their LAI by 25 and 45%,espectively (Fig. 3b). This led to a greater input of litterfall, howeverhe microbial biomass did not respond positively to the increase in Cvailability, as bacterial respiration was reduced by the anoxic con-itions (Fig. 3c). The combined effect of vegetation mortality andacterial respiration inhibition allowed the system to increase theotal C soil stock. However, anoxic stress caused plants to respire

ore than the overall C uptake, and the net rate of metabolism wastrongly negative for the period.

Following water table depletion, the C uptake by plants, ˘A,as enhanced, and the population of grasses and trees grew faster.lthough bacterial respiration increased for the period, the overall

etabolism remained positive. The following hydrological pulses

ffected bacterial performance. However, as vegetation reached alateau, controlled by the maximum carrying capacity, minor dis-urbances in the C uptake by plants were sufficient to bring the

012320 323.4 2000

metabolism below zero. At the end of the simulation period, thewetland situated in Cape Leeuwin (C1) exhibited approximatelythe same C density in its soil and lake stores as at the beginning ofthe simulation of around 6 kg C m2. Soil C lost via groundwater werenot included as part of the net metabolism estimate, but reached amaximum of 0.003 g C m−2 d−1.

In the Lake Toolibin climate (C2), a major precipitation eventwas responsible for waterlogging 70% of the total wetland domain.This event took place 3.5 years after the beginning of the simulationand is identified in Fig. 3. In this case, the hydrological pulse causeda 40% enhancement in bacterial activity but resulted in a negativeeffect on net vegetation productivity. The combined effect was netC release to the atmosphere under these conditions. The followingmajor event increased microbial activity to ∼80%. However, thenegative effect on vegetation was not as evident as in the previousevent, so that the net metabolism remained positive overall. Fol-lowing this period, a high bacterial biomass was able to deplete Csoil stocks. The reduction in water availability, however, affectedthe bacterial population that was again stimulated by hydrologi-cal pulses and eventual litterfall inputs in the following years. Theresulting C budget was therefore observed to be less determined byvegetation than it was in the C3 scenario, but vegetation was stillthe primary driver of carbon dynamics relative to bacterial activity.

The C2 simulation demonstrated an accumulation of C that was,on average, 36% higher than the wetter C1 scenario, and 47% higherthan the dryer C3 scenario. Besides a vegetation cover almost 60%denser than C3, C2 was characterised by a more efficient rate ofmicrobial decomposition, ∼8% higher than C3 and 40% lower thanC1. This low efficiency in organic carbon decomposition is related tothe combined effect of soil moisture and temperature, representedby low values for the functions f� and fT, which govern the micro-bial activity performance (Fig. 4). However, the combined effect ofvegetation biomass increase and improvement in water availabil-ity caused by the hydrological pulses observed around year 5 and 9(from the beginning of the simulation), enhanced microbial activ-ity and depleted C soil stocks by ∼5%. At the end of the simulationperiod, however, C stocks were about the same as showed in year1, above 9 kg C m2.

For the three hypothetical scenarios tested, vegetation density

0 4 8 12 16Time (years)

Fig. 4. Locations C1 and C3 presented higher decomposition efficiency because ofideal soil moisture and temperature conditions, assessed by fT*f0.

1 al Modelling 249 (2013) 3– 18

semmrelatu

5hs

syrttdrotctb

a2bTe(saaws

ntt2uspfc

ttIso0mTaRmdal

R1 R2 R3

0.2

0.4

AL/A

W

9

11

13

15

CW

(kg

Cm

−2 )

0 J.Z. Coletti et al. / Ecologic

tocks, given their importance in defining microbial activity. How-ver, vegetation was indicative of the absolute rate of wetlandetabolism, particularly in cases where C storages were low andicrobial activity was less efficient, based on the closer apparent

elation between wetland metabolism and vegetation metabolismxhibited in the C3 simulation, relative to C1 and C2. Neverthe-ess, to better observe the exclusive effects of microbial efficiencynd vegetation cover on C storage and metabolism, an analysishat excludes temperature variations and weather fluctuations cannravel the relative mechanisms, as described next.

. Response of the carbon cycle to wetland partitioning:ow does the environment partitioning control carbontorage and vegetation assemblage in wetlands?

As a means to relate environment partitioning to carbontorage and vegetation assemblage, a synthetically generated one-ear rainfall time-series was repeated to allow the system toeach dynamic equilibrium (quasi-steady state). This same rainfallime-series was scaled, producing three synthetic rainfall realisa-ions with the same intra-annual distribution, but distinguishableepths, equal to 965, 580 and 350 mm (termed R1, R2 and R3,espectively). The aim was to reproduce the mean annual rainfallbserved in the representative catchments that were previouslyermed C1, C2 and C3, but by using a controlled rainfall statisti-al distribution, to enable us to reach dynamic equilibrium so thathe metabolism of each wetland was the result of vegetation andacterial biomass response to water availability.

For this exercise, precipitation was synthetically manipulatedpplying a stochastic rainfall model (as described in Hipsey et al.,003) to generate an annual realisation with an intra-annual distri-ution that is typical of SWWA, with wet winters and dry summers.he same intra-annual distribution was used because rainfall deliv-ry pattern is known to significantly affect vegetation assemblageColetti et al., submitted for publication). A one year long time-eries of weather conditions (cloud cover, wind speed, temperaturend relative humidity) was adopted for each precipitation scenario,s above for the C2 scenario. The key output of these scenariosas three distinct hydroperiods that were used to elucidate the

ensitivity of carbon cycling to the hydrological regime.As rainfall decreased from the wettest (R1) to the driest (R3) sce-

ario, the annual average soil moisture decreased from 0.13 m3 m−3

o 0.07 m3 m−3 and the annual average lake area from 39 to 15% ofhe total wetland domain. The water table level moved from 5 to5% of the total soil depth, on average. Similar to the simulationssing observed rainfall data from C1, C2 and C3, the total wetland Ctorage, CW, did not follow the same trend as water storage. Whenrecipitation decreased 40%, from R1 to R2, CW increased 3.6%. Aurther 40% decrease in rainfall from R2 to R3 scenario caused thearbon to deplete 31% (Fig. 5).

The reason for distinctive features in soil C storage is related tohe microbial decomposition efficiency, which is a non-linear func-ion of water availability, temperature and C input from vegetation.n this particular simulation, temperature was kept the same in allcenarios, so that the C storage was solely a response to hydrol-gy and to C inputs from plants. The reported litterfall was 0.78,.67 and 0.33 g C m−2 d−1 while f� , the sensitivity of bacteria to soiloisture, was 0.40, 0.28 and 0.23 for R1, R2 and R3, respectively.

hus, besides a 14% lower C input, the soil moisture conditions werelso 30% less favourable to decomposition in scenario R2 relative to1. As a result, the final decomposition efficiency, evaluated as the

icrobial respiration, Rb, divided by the C available, CB, was also

isproportional to the water storage (Fig. 7b). Other factors suchs groundwater loss also must be taken in account. Although veryow, around 0.03 g C m−2 d−1, the carbon lost via ground water in

Fig. 5. A decrease in water storage along the DI gradient did not present a decreasein C storage as highlighted by the relationship between the fraction of inundatedarea over the wetland domain (bars), and the total C storage, CW = CB + CV (circles).

R1 scenario was also greater than R2 and R3, which were almostnegligible.

Overall, the main C storage was located in the soil and lake (CB),accounting for about 69% of the total wetland carbon, CW (=CV + CB),in scenario R1 and 70 and 78% in scenarios R2 and R3, respectively(Fig. 6b). Thus, the relative fraction of carbon in the soil, CB, to thetotal wetland, CW, remained fairly constant. This was not observedwhen using the observed rainfall data in Section 4. The fact thatC1, C2 and C3 had different temperatures does not explain thediscrepancy in CB, as C2 presented a temperature average that ismore suitable to C decomposition. In fact the results show that thevariability in water delivery and vegetation dynamics acceleratesthe wetland metabolism, not just through pulses of water, but alsothrough litterfall, stimulating bacterial growth. In the case of C3, thelack of C inputs is believed to contribute to the low C soils stocks,as bacteria find the conditions in this scenario better than in C2.

Therefore, water availability, and C from litter, stimulated themetabolism of both plants and decomposers, and it was observedwith more clarity using the synthetic rainfall data. Although allsimulations (R1, R2 and R3) predict a neutral metabolism, i.e., Cinput equals to C output, the fluxes of C that define the final steadystate are different (Fig. 6a). For the highest mean average rainfallscenario, R1, plant C uptake was high, as grasses and aquaticplants, characterised by a rapid metabolism, were abundant(Fig. 7a). However, decomposers also benefited from high levelsof soil moisture. The equilibrium between input and output wasfound when C storage was lower than the storage found in R2,although the over all fluxes in R2 were lower. Observing the wateravailability and the C flux trends for R1, R2 and R3, we observedthat although water availability could not be used as a proxyfor C storage, it could be used as a reference to determine themagnitude of the flux of C transfer with the atmosphere. How eachof these steady state scenarios responds to non-stationary climaticconditions is evaluated next.

6. Wetland resilience to climate disturbance: how does thevegetation assemblage and C storage change in response toprojected shifts in precipitation and temperature?

6.1. Rainfall decrease

In an attempt to evaluate the C storage response to changes inclimate forcing, the system was stressed with a progressive decline

J.Z. Coletti et al. / Ecological Modelling 249 (2013) 3– 18 11

R1 R2 R30

7

14

CW

(kg

Cm

−2 )

CB

CV

0.2

0.4

AL/A

W

(b)

1

3

5

C f

luxe

s (g

C m

−2 d−

1 )

1

3

5

1

3

5

(a)

ΠA

RR

b

F area,

w ) the t

i5Cwepairtb

wswatsal

dwrfdtidR

The total vegetation biomass reached a peak for a precipitation of

ig. 6. A decrease in water availability (represented by the fraction of waterloggedetland as indicated by (a) the C fluxes that define the wetland metabolism, and (b

n rainfall to each one of the time series previously tested in Section. The aim was to evaluate which system, each with a particular

allocation (proportion between CV and CB) and C flux regime,ould be more resistant to a decline in water availability than oth-

rs, as assessed in terms of C storage. Specifically, after each 10 yeareriod, rainfall was reduced by 5% of the original value, so that after

40 year period, the total annual precipitation was 20% less thann the beginning of the simulation. For all periods, the pattern ofainfall timing was kept the same (same intra-annual distribution),hus, rainfall depth was altered by scaling the original time-seriesy a factor less than one.

The final annual average precipitation for scenario R1 and R2as very similar to the annual average in the beginning of the

imulations R2 and R3, respectively, and so was the fraction ofaterlogged area, AL/AW (Fig. 8a). The vegetation assemblage and

bundance responded accordingly to the water availability, so thathe vegetation status at the end of simulation R1 and R2 was alsoimilar to the beginning of simulation R2 and R3 (Fig. 8b), showing

close relation and relatively quick response to changing hydro-ogical conditions.

The vegetation response to the progressive water availabilityecline, however, was distinct in each scenario, suggesting that theater scarcity led to higher plant biomass sensitivity to changes in

ainfall depth. Indeed, although the precipitation fell around 20%rom the beginning to the end of each run in all scenarios, plantsecreased 47% in scenario R2 and almost 75% in scenario R3 forhe entire period. Conversely, vegetation biomass in scenario R1

ncreased by ∼4%. Therefore, as water became scarcer, the rate ofecline in vegetation biomass was steeper, especially for scenario3 (Fig. 8c).

R1 R2 R30

0.5

1

LA

I (m

2 m−

2 )

(a)

VVV

Fig. 7. Vegetation assemblage (represented by LAI) and t

AL/AW) decreased C transfer with the atmosphere, but not the total C stored by theotal carbon stored by the system.

The tight connection between water availability and vegetationabundance was not reflected in the C stored in the soil. The stor-age by the end of the R2 simulation did not reach the level foundin the beginning of R3. As the soil became drier, the ideal hydro-logical condition for carbon decomposition by microbial biomasswas lost. As litterfall decreases with the vegetation reduction, thecarbon available switched from being more labile to being morerefractory, so more energy was necessary to break the carbon down,slowing the process further. The result was a greater predomi-nance of carbon being stored in the soil than in the beginning of thesimulation.

Other reasons for the gap in C storage between the end of R2and the beginning of R3 relate to the many steady states that canbe attainable for C and water storage, or it is possible the initialcondition for C soil storage was not chosen correctly. Nonetheless,the steep decline in carbon stocks from 30 to 40 years and then50 years in the R2 scenario suggests that a 10 year time frame isnot enough for C soil storage to stabilize given changes in vege-tation and rainfall; if hydrological and vegetation conditions wereto be kept constant, eventually the levels of C stocks would reachthe same values as in the beginning of simulation R1, suggestinga unique maximum C storage that a certain amount of water cansupport.

From year 20 to 40 in simulation R1, CB increased due to acombination of higher carbon inputs from the vegetation and adecline in microbial efficiency in response to lower soil moisture.

about 800 and 600 mm. Although aquatic plants (V1) decreased,trees and grasses (V2 and V3, respectively) had the advan-tage of a smaller area under temporary inundation conditions.

231

R1 R2 R3

0.04

0.06

0.08

Rb /

CB (

g C

kgC

−1 )

(b)

he decomposition efficiency by bacterial biomass.

12 J.Z. Coletti et al. / Ecological Modelling 249 (2013) 3– 18

F as dems ariabi

WCrfo

ttdsgasuat(Rwfapicacac

ig. 8. Carbon storage changes caused by successive decline in water availability

aturated zone, (b) the resultant vegetation adaptation to the changes, and (c) the v

hen the vegetation declined during the last 10 year period, storage in the soil stopped increasing. This conditionepresented a similar state to that which simulation R2 beganrom; a lower vegetation and microbial metabolism, with anverall higher storage.

The disparity between the below to above ground C alloca-ion that happens as water availability becomes limiting reinforceshe idea that vegetation responds quicker to the hydrological con-itions than detrital carbon (Fig. 9a). This suggests that wetlandystems with their carbon allocation situated predominantly aboveround are less resistant to climatic changes, and this is amplifieds we move to areas of higher DI, as vegetation showed a higherensitivity to the same decline in precipitation under higher DI val-es. Further, the maximum C accumulation belowground occursfter a period of low microbial respiration efficiency comparedo the soil C input (represented by the litterfall and root death)Fig. 9b). During the 30–40 year period of the simulation, scenario1 reached around 700 mm of annual precipitation. Although thisater availability culminated in the maximum vegetation biomass

or the whole simulation, the decomposition rate did not respondccording to the increase in available detrital carbon for decom-osition. The result was an increase in CB that could be noticed

n the final period of the simulation, from 30 to 40 years, whenompared to the previous decade. The prediction that microbial

ctivity efficiency increased almost monotonically with rainfall, inontrast with vegetation biomass, within the range of water avail-bility tested, reinforces the idea that a maximum accumulationan be attained.

onstrated by (a) the decrease in precipitation depth and consequent drop in thelity in carbon allocated to belowground (CB) and vegetation (CV) pools.

It is expected that microbial efficiency would also reach a max-imum efficiency and decline when the system experiences anoxicconditions within a greater relative area of the wetland domain.However, our results suggest that the maximum flux of carbonfrom vegetation to soil and the maximum microbial decompositionefficiency do not occur for the same water availability.

Although vegetation experienced an overall decline for scenar-ios R2 and R3 its net metabolism was positive for most of thesimulated period (Fig. 10a). Negative metabolism for vegetationwas simulated at the beginning of each consecutively drier decade,when vegetation responded to the water stress. As the vegetationand water reached a new dynamic equilibrium, the overall carbonuptake was slighter greater than the respiration. This equilibriumwas possible as the total litterfall and root turnover balanced theC budget of plants. As precipitation declined, the range of valuesfor metabolism given the same annual precipitation depth gradu-ally increased as it experienced a steep drop in vegetation in theearly years which was reduced as the vegetation adapted to thenew conditions.

As the DI increased, vegetation became more sensitive tochanges in rainfall (Fig. 8), and the net metabolism became morevariable, reflecting the differences between the initial and finalstate of each 10 year period. Since the wetland metabolism canbe largely explained by the vegetation metabolism, the same fea-

ture could be observed when accounting for the total C respiredby decomposers (Fig. 10b). The overall wetland metabolism, how-ever, was only positive when C soil or C above ground was beingaccumulated. Carbon accumulation was possible only when low

J.Z. Coletti et al. / Ecological Modelling 249 (2013) 3– 18 13

200 400 600 800 1000

4

6

8

10

12

CB /

CV

P (mm)

(a)

R1R2R3

200 400 600 800 1000

0.02

0.04

0.06

0.08

P (mm)

10−

3 d−1

Rb / C B

0.2

0.4

0.6

0.8

g C

m−

2 d−1

(b)L

l + R d

F can cr biome

dt

tpans

6

r2p(tsw2ami

Fm

ig. 9. Distinct behaviour from vegetation and bacteria regarding water availabilityelation between the total C storage in the lake and soil (CB) and the total vegetationnter the soil pool (Ll + Rd) and the microbial respiration efficiency (Rd/CB).

ecomposition efficiency compared to the flux of carbon enteringhe soil pool happened, as showed in Fig. 9b.

When comparing the relation between wetland and vegeta-ion metabolism (Fig. 10c), the R3 simulation followed a differentattern than R1 and R2, so that the total wetland metabolismppeared to be less negative than when compared to wetter sce-arios. It is also an indication that is possible to attain several stabletates between vegetation and wetland metabolism.

.2. Rainfall decrease and temperature increase

In Australia, droughts have become hotter since about 1973, asegistered temperatures have been progressively higher (Nicholls,004). To evaluate the possible combined effect of an antici-ated increase in temperature (IOCI, 2010) and rainfall depletionNicholls, 2004) could have on wetlands, we repeated the simula-ions used in Section 6.1 with an increase of 0.50 ◦C per decade,o that after a 40 years period, the total temperature increaseas 2.0 ◦C, over the period where the total rainfall decline was

0%. The scenarios including the temperature increase are referreds R1*, R2* and R3* to distinguish from the time series of sameean annual rainfall depth that were not subject to temperature

ncrease.

200 400 600 800 1000

−0.5

0

0.5

ΠA

− R

(g

C m

−2 )

P (mm)

(a)

200 400 600 800 1000

−0.5

0

0.5

P (mm)

ΠA

− R

− R

b (g

C m

−2 )

(b)

−

ΠA

− R

− R

b (g

C m

−2 )

ig. 10. The (a) vegetation and (b) wetland metabolism response to the decrease in metabolism for each scenario.

ause wetlands to find a point of maximum C storage in soils, as indicated by (a) theass (CV); and (b) the relation between the mean annual precipitation and total C to

With the rise in temperature, the loss in vegetation biomassreached around 250 g C m−2, relative to scenarios where temper-ature was kept constant (Fig. 11a). This value reached as much as400 g of C in scenario R1, as R1* does not experience any increase invegetation biomass that was featured in R1 around years 30–40. Bythe end of the 50 year-long simulation, the total vegetation biomasswas around 5, 8 and 38% less, respectively, than the biomass foundin the constant temperature scenarios. This reduction in vegeta-tion cover was caused by an intensified rate of plant respiration(Fig. 11d). In a warmer climate, plants experienced around 15%higher respiration per carbon accumulated as biomass, as temper-ature directly affects respiration efficiency, according to our modelformulation.

Surprisingly, the hydroperiod was not affected greatly by thewarmer climate. Indeed, the total waterlogged area was around 1,0.2 and 0.1% greater in R1, R2 and R3 compared to R1*, R2* and R3*,respectively. Thus, we observed that the depletion in vegetationbiomass and lower water transpiration, buffered the effect of higherevaporation and transpiration rates.

The extra carbon loss from the increased temperature wasmore pronounced in the belowground pool, with a maximum of1.5 kg C m−2 (Fig. 11a). A depletion in carbon supply, together withhigher temperatures caused C soil storage to be around 15, 10 and

−0.5 0 0.5 1−1

0.5

0

0.5

ΠA

− R (g C m−2)

(c)

R1R2R3

ean annual rainfall depth, and (c) the relation between vegetation and wetland

14 J.Z. Coletti et al. / Ecological Modelling 249 (2013) 3– 18

10 20 30 40 50−1.5

−1

−0.5

0

CB* −

CB (

kg C

m−

2 )

Time (years)

(a)R1*− R1R2*− R2R3*− R3

10 20 30 40 50−0.4

−0.3

−0.2

−0.1

0

0.1

CV*

− C

V (

kg C

m−

2 )

Time (years)

(b)

10 20 30 40 501

1.1

1.2

1.3

(Rb* / C

B*)

/ (R

b / C

B) (c)

Time (years)

R1*/ R1

R2*/ R2

R3*/ R3

10 20 30 40 501

1.1

1.2

(R* / C

V*)

/ (R

/ C

V)

Time (years)

(d)

F getatid

7sw

7

wmppbct“ssain

ntbdttstid

aotawcs

ig. 11. An increase in temperature stimulated both (a) microbial activity and (d) veepleted. Time series with temperature increase are represented with an asterisk.

% less than in stationary temperature scenarios, by the end of theimulation period. In this case the microbial respiration efficiencyas elevated by up to 20% (Fig. 11c).

. Conclusions

These results highlight that the climate signal that drives aetland ecosystem can be strongly mediated by vegetation, andicrobial dynamics. Although water availability can be used as a

roxy for vegetation abundance, greater water storage was not pro-ortionally correlated to greater carbon storage in wetland soilsased on the simulations tested in this study. However, we con-lude that in water limited areas, wetland metabolism is closelyuned to water delivery, so that water availability defines theintensity” of the fluxes of C between the terrestrial and atmo-pheric environments. Also, it was shown that for higher DI’s, theensitivity of carbon stocks to changes in precipitation decline waslso higher; that is, the C depletion was greater for the same declinen precipitation when comparing a drier scenario to a wetter sce-ario.

Nevertheless, we highlight that the resultant C storage pools andet metabolism are related to the chosen parameters for respira-ion, photosynthesis and microbial respiration. The results may alsoe sensitive to the set of functional types of vegetation with theiristinct water uptake strategies and further modulate the responseo climate forcing. Therefore further work on the sensitivity ofhe overall metabolism and carbon storage to plant type selectionhould be conducted. Further, to improve our understanding onhe role of vegetation types in wetland metabolism, we suggest thenclusion of differentiation of the decomposition rate for litter fromifferent vegetation types.

In the context of management, although vegetation biomassnd biodiversity can be a proxy to evaluate the general “health”f an ecosystem, these results indicate that they did not representhe C storage efficiency of a wetland as vegetation biomass was

lso not proportionally related to C soil storage in the hypotheticaletlands analysed here. Management actions that alter upstream

atchment systems and delivery of flow to wetland environmentshould consider the potential implications on carbon storage.

on respiration. As a consequence, C soil storage (b) and vegetation biomass (c) were

Appendix A. The ecohydrological model described byColetti et al. (submitted for publication)

The relation between lake storage (L) and lake height (hL) isgiven by the paraboloid equation:

hL =√

2Lb�

(A.1)

AL is as given by:

AL = �

b

√2Lb�

(A.2)

The variation of lake volume, dL/dt (m3 d−1), is calculated as:

dL

dt= PL + Qc + Qw − QS − EL − Qout (A.3)

Evaporation in the lake is given by:

EL = cE0AL (A.4)

where E0 (m d−1) is the potential evaporation, and c is a pan-to-lakecorrection factor.

Qs is the seepage that flows through the area of lake base, ASL,and the free saturated area, AS:

Qs = 2Ks(hL + hB − hS)(ASL + AS)

rw(A.5)

where the area of the lake base, ASL, can be defined by the surfaceof the paraboloid:

ASL = �rL6hL

+ (r2L + 4h2L )

3/2 − r3L (A.6)

Qout is the flux of water exiting the wetland system:

Qout = Qin − (Lmax − (L − EL)) (A.7)

Flow to the area of standing water from the surrounding terres-

trial component of the wetland domain, Qw, is generated throughinfiltration (Qie) and saturation excess (Qse) mechanisms:

Qw = Qse − Qie (A.8)

al Mo

wSi

Q

be

Q

zwlSwre(

S

a

a

w

hfi

I

pscitsa

E

a

E

un

E

w�

J.Z. Coletti et al. / Ecologic

here Qse is the volume that exceeds the maximum soil storage,max, and Qie (m3 d−1) is the amount of effective precipitation thats greater than the capacity of the soil to infiltrate, I (m3 d−1):

ie = (PAU − 1) + PAS (A.9)

St, the total volume present in the soil is calculated as:

dStdt

= I ± QS − Qss − Esoil − Qse (A.10)

In the above equation, Esoil is the sum of transpiration (E) andare-soil evaporation (Eb) from both unsaturated and saturatednvironments, such that Esoil = EbU + EbS + EU + ES.

The baseflow is defined as:

ss = (˛GhS)AW (A.11)

The maximum net capacity of water storage in the unsaturatedone, Uc, is the difference Smax − Ssat, where Ssat is the volume ofater that is stored in the saturated pool, S, below the water table

evel, hS, such that Ssat = S�. When the soil is totally saturated,t = Ssat = Smax and Sus = Uc = 0, where Sus is the effective volume ofater present in the U zone. Under such a condition, the infiltration

ate, I, is equal to zero. For all other times, the total water store ofach soil column is comprised of a saturated (S) and an unsaturatedU) region, such that:

t = Ssat + Sus (A.12)

The water balance of sub-region U and S is respectively defineds:

dSusdt

= I − EU − EbU − Qp (A.13)

nd

dSsatdt

= Qp ± Qs − ES − EbS − Qss (A.14)

here Qp is the percolation rate.Soil type affects the infiltration rate through the saturated

ydraulic conductivity, Ks (m d−1), and the arbitrary recession coef-cient for infiltration, kI:

=

⎧⎪⎨⎪⎩

−KS(� − 1)k1AU if I < PAU

PAU if I ≥ PAU

Uc if I > Uc

(A.15)

The evaporation from bare soil, Eb, is calculated based on theotential evaporation, E0. If the soil is not saturated, the ratio ofoil water content, �, to water content at field capacity, fc, is alsoonsidered as a scaling factor. Further, evaporation from bare soils adjusted based on total LAI (sum of all vegetation types) to reflecthe cover that vegetation causes due to shading of the exposed soilurface. Therefore, the evaporative rate applied over the relevantreas, AS and AU, respectively, are:

bS = E0

(1 − 0.9

LAISLAImax

)AS (A.5′)

nd

bU = E0

(1 − 0.9

LAIULAImax

)(�

fc

)AS (A.6′)

Transpiration, E, is modeled as a function of the plant waterptake rate, W (m d−1), integrated over the relevant environment,, such that:

n = WnAn (A.7′)

here W is a function of the normalized potential water uptake, and the potential evapotranspiration, E0 (Skaggs et al., 2006). �

delling 249 (2013) 3– 18 15

depends on the plant functional type, i, and their associated wateruptake strategy, and normalized depending on the wetland zonemaximum LAIn, (n = L, U, S), and associated soil moisture conditionsexperienced by the roots at a given time. As a result, the total wateruptake rate is the sum of all vegetation groups, i, coexisting in aparticular environment, n, such that:

Wi,n =∑i

i,nE0LAIi,nLAImax,n

(A.19)

Percolation is calculated in a time step after losses from evapo-transpiration and surface runoff are computed, such that:

Qp ={S∗US − Ucfc if S∗

US > Ucfc

0 otherwise(A.20)

and

S∗US = St−1

US + I − EUS − EbU (A.21)

The total carbon amount accumulated as vegetation biomass, B(kg C), is governed by the rate of carbon uptake via photosynthesis,˘A (kg C d−1 m−2), and losses due to litterfall (Ll), root death (Rd)and respiration (R), with all the loss terms given in kg C d−1 m−2:

dBi,ndt

=

⎧⎨⎩

(˘Ai,n− Lli,n − Ri,n − Rdi,n )An+Di,n−1

dAndt

ifdAndt

> 0

(˘Ai,n− Lli,n − Ri,n − Rdi,n )An+Di,n

dAndt

ifdAndt

< 0

(A.22)

where An is the area of the nth wetland zone (m2), and D is thecarbon density per unit area (=B/A). It is constrained by Dmax, themaximum carrying capacity that the system can hold given a soilwater-holding capacity and climate, when in hydrological equilib-rium (Nemani and Running, 1989). Mass conservation is assured ifportion of biomass previously belonging to another area, Bn−1, isincorporated into An when dAn/dt > 0 such that:

Di,n−1 = Bi,n−1

Ai,n−1(A.23)

Conversely, biomass that no longer belongs to An (in casedAn/dt < 0) must be removed. Therefore:

Di,n = Bi,nAi,n

(A.24)

Litterfall (Ll) and root turnover (Rd) are linearly related to biomassaccording XLl and XRd, respectively (Friend et al., 1997). Plant res-piration is configured as a function of temperature, such that:

R = kReuT (KrBr + KlBl)B (A.25)

where kR (m2 d−1) is a coefficient that adjusts the respiration to thehydrological environment such that it is lower when vegetation isexposed to its most suited hydrological conditions. KR, KL and arescaling factors.

The gross assimilation of carbon, ˘A (kg C m−2 d−1), is a func-tion of the uptake efficiency �˘ (kg C kg CO2

−1), the potentialuptake rate, ˘0 (m d−1), �CO2, the carbon dioxide air – leafdiffusion gradient (kg CO2 m−3) (Lohammar et al., 1980) and LAI(m2 leaf m−2 land):

˘A = �˘˘0�CO2LAI (A.26)

where,

˘0i =CC CM

CC + CMdl (A.27)

Here, CM is based on a maximum mesophyll conductance

modified by normalizations of temperature and solar radiationdependencies, summarized as:

CM = CMmaxCM˚CMt (A.28)

1 al Mo

w

C

a

C

b

C

wu

evsmu

˛

6 J.Z. Coletti et al. / Ecologic

here,

M˚ = ˚C − ˚0

˚C + ˚0.5(A.29)

nd

Mt = Tmax − T

T − Tmin(A.30)

Similar to CM, CC is based on a maximum conductance modifiedy the normalized potential water uptake, � such that:

C = CCmax (A.31)

here � is a function of the normalized water availability for plantptake, ̨ and the root density in contact with the water table, ˇ:

= co˛U(1 − ˇ) + ca˛Sˇ (A.32)

Water availability under saturated conditions, ˛S, is alwaysqual to 1 since there is no water limitation. Conversely, in theadose zone, water availability for plant uptake, ˛U, depends onoil moisture (�) and also on plant characteristics such as the soiloisture at wilting point (�W) and optimal soil moisture for plant

ptake (�0) as follows:

U

⎧⎪⎪⎪⎨⎪⎪

(� − �0

�0 − �w

)+ 1, if � < �0

1, if �0 ≤ � < 1(A.33)

⎪⎩

0, � = 1

Parameter description Symbol Units

Catchment parametersSoil porosity �C –

Total soil depth hSmaxC mm

Soil moisture at field capacity fcC –

Recession coefficient for percolation kp –Recession coefficient for infiltration ki –

Recession coefficient for baseflow kg –

Wetland parametersPan-to-lake evaporation factor c –

Recession coefficient for baseflow ˛G –

Recession coefficient for infiltration kI –

Vegetation parametersMaximum LAI for carbon uptake LAIm m2 mScaling factor that relates temperature to respiration KR –

Scaling factor that relates temperature to respiration KL –

Scaling factor that relates temperature to respiration –

Ratio respiration efficiency to hydrological environment kR m2 d−

Carbon dioxide air – leaf diffusion gradient �CO2 kg COCarbon uptake efficiency �˘ kg C kMaximum mesophyll conductance CMmax m s−1

Photosynthesis light compensation point and ˚0 and kJ m−

Radiation level to normalize solar radiation ˚0.5 kJ m−

Maximum and minimum photosynthesis temperature Tmax , Tmin◦C

Maximum canopy conductance CCmax m s−1

Soil moisture at wilting point �W –

Soil moisture for optimal vegetation uptake (V3 and V2) �0 –

Geometric parameters Symbol Un

Maximum radius rW mmMaximum depth hW mmMaximum area AW mmRelation between hL and rL in the paraboloid equation b –

Maximum volume of the wetland (soil + lake volume) Wmax mmMaximum volume of the lake Lmax mm

delling 249 (2013) 3– 18

Whether the plants effectively take water from below or abovethe phreatic surface (or from both regions) depends on the uptakestrategy of the relevant functional group, and for each it is con-ceptually defined by the parameters co and ca, which represent theplant compatibility to take water from saturated or unsaturatedconditions. As a result, each plant group (described below) can suc-cessfully obtain water from the wetland environments that matchits hydrological requirements. In summary:

Whether the plants effectively take water from below or abovethe phreatic surface (or from both regions) depends on the uptakestrategy of the relevant functional group, and for each it is con-ceptually defined by the parameters co and ca, which represent theplant compatibility to take water from saturated or unsaturatedconditions. As a result, each plant group (described below) can suc-cessfully obtain water from the wetland environments that matchits hydrological requirements. In summary:

Vegetation Type 1 (V1) – aquatic vegetation: plants requirestanding water conditions.

Vegetation Type 2 (V2) – facultative vegetation (trees): plants cantake water from the unsaturated zone but are also able to take waterfrom below the water table due to the presence of deep roots. Bydefinition, a small fraction of roots must be aerated, and thereforewater uptake is assumed to stop as soon as the water table reachesthe ground (i.e., water-logged conditions).

Vegetation Type 3 (V3) – mesophyte vegetation (grasses): thewater uptake occurs just in the unsaturated portion of soil.

Appendix B. Summary of the soil, vegetation and geometricparameters for the ecohydrological model and the initialcondition for the simulation presented in Section 4

Value Reference/remarks

0.4 Merz (2000)5500 Merz (2000)0.3 Merz (2000) and Barrett-Lennard (2008)0.4 Farmer et al. (2003)3 Farmer et al. (2003)0.003 Farmer et al. (2003)

0.8 Dogramaci et al. (1996)0.0001 Farmer et al. (2003)3 Farmer et al. (2003)

−2 2 Nemani and Running (1989)0.002 Running and Coughlan (1988)2e−4 Running and Coughlan (1988)0.09 Running and Coughlan (1988)

1 2.3–50 This study2 m−3 0.0007 Lohammar et al. (1980)g CO2

−1 0.8 This study0.0008 Running and Coughlan (1988)

2 d−1 432 Running and Coughlan (1988)2 d−1 9730 Running and Coughlan (1988)

0, 37 Running and Coughlan (1988)0.0016 Running and Coughlan (1988)0.9fc Guswa (2005)0.12fc and 0.14fc Guswa (2005)

its Values Reference/remarks

1.06e6 (used to adjust the topography to Lake Toolibin) 4000 (used to adjust the topography to Lake Toolibin)2 3.55e12 (used to adjust the topography to Lake Toolibin)

1.77e−9 (used to adjust the topography to Lake Toolibin)3 1.42e16 (used to adjust the topography to Lake Toolibin)3 3.55e15 (used to adjust the topography to Lake Toolibin)

al Mo

R

A

B

B

B

C

C

C

C

D

D

D

D

F

F

F

F

G

G

G

H

H

H

J.Z. Coletti et al. / Ecologic

Variable Symbol

Lake level (C1, C2, C3, respectively) hL

Water table level (C1, C2, C3, respectively) hS

Soil moisture (C1, C2, C3, respectively) �

Leaf Area Index (V2, V3 and V1) (C1, C2, C3, respectively) LAI

eferences

lexander, L., Hope, P., Collins, D., Trewin, B., Lynch, A., Nicholls, N., 2007. Trends inAustralia’s climate means and extremes: a global context. Australian Meteoro-logical Magazine 56, 1–18.

arrett-Lennard, A., 2008. Surface Water Ponding on Low-Lying Valley Floors inSouth-West Western Australia: Interactions with Groundwater and its Rolein Secondary Dryland Salinity, Honours Dissertation. School of EnvironmentalSystems Engineering, The University of Western Australia, Crawley, Australia.

ridgham, S.D., Megonigal, J.P., Keller, J.K., Bliss, N.B., Trettin, C., 2006. The carbonbalance of North American Wetlands. Wetlands 26, 889–916.

urkett, V., Kusler, J., 2000. Climate change: potential impacts and interactions inWetlands of the United States. Journal of the American Water Resources Asso-ciation 36, 313–320, http://dx.doi.org/10.1111/j.1752-1688.2000.tb04270.x.

apon, S.J., 2003. Plant community responses to wetting and drying ina large arid floodplain. River Research and Applications 19, 509–520,http://dx.doi.org/10.1002/RRA.730.

arter, J.L., Colmer, T.D., Veneklaas, E.J., 2005. Variable tolerance of wetland treespecies to combined salinity and waterlogging is related to regulation of ionuptake and production of organic solutes. New Phytologist 169, 123–134,http://dx.doi.org/10.1111/j.1469-8137.2005.01552.x.

oletti, J., Hinz, C., Vogwill, R., Hipsey, M. Testing wetland vegetation responses tochanges in hydrological pulses in a semi-arid region. Ecohydrology, submittedfor publication.

SIRO, 2009. Water yields and demands in South-West Western Australia. In: Sum-mary of a Report to the Australian Government from the CSIRO South-WestWestern Australia Sustainable Yields Project.

’Odorico, P., Laio, F., Porporato, A., Rodriguez-Iturbe, I., 2003. Hydrologic controlson soil carbon and nitrogen cycles. II. A case study. Advances in Water Resources26, 59–70.

avidson, E.A., Janssens, I.A., 2006. Temperature sensitivity of soil carbon decompo-sition and feedbacks to climate change. Nature 440, 165–173.

ean, W., Gorham, E., 1998. Magnitude and significance of carbon burial in lakes,reservoirs, and peatlands. Geology 26, 535–538.

ogramaci, S., George, R., Mauger, G., Ruprecht, J., 1996. Water balance and salinitytrend, Toolibin catchment. Department of Conservation and Land ManagementReport, Western Australia.

armer, D., Sivapalan, M., Jothityangkoon, C., 2003. Climate, soil, and vegetationcontrols upon the variability of water balance in temperate and semiaridlandscapes: downward approach to water balance analysis. Water ResourcesResearch 39, 1035, http://dx.doi.org/10.1029/2001WR000328.

inlayson, C.M., 2005. Plant ecology of Australia’s tropical floodplain Wetlands: areview. Annals of Botany 96, 541–555.

riend, A.D., Stevens, A.K., Knox, R.G., Cannell, M.G.R., 1997. A process-based,terrestrial biosphere model of ecosystem dynamics (Hybrid v3.0). EcologicalModelling 95, 249–287.

roend, R.H., McComb, A.J., 1994. Distribution, productivity and reproductive phe-nology of emergent macrophytes in relation to water regimes at Wetlands ofSouth-western Australia distribution, productivity and reproductive phenology.Australian Journal of Marine & Freshwater Research 45, 1491–1508.

eorge, R., McFarlane, D., Nulsen, B., 1997. Salinity threatens the viability of agricul-ture and ecosystems in Western Australia. Hydrogeology Journal 5, 6–21.

illis, J.H., Baker, P.J., Mac Nally, R., Cunningham, S.C., Thomson, J.R., Hamilton,F., 2009. Mortality of developing floodplain forests subjected to a dry-ing climate and water extraction. Global Change Biology 15, 2176–2186,http://dx.doi.org/10.1111/j.1365-2486.2009.01915.x.

uswa, A.J., 2005. Soil-moisture limits on plant uptake: an upscaled relationship forwater-limited ecosystems. Advances in Water Resources 28, 543–552.

ipsey, M.R., Sivapalan, M., Menabde, M., 2003. A risk-based approach to thedesign of rural water supply catchments across Western Australia. HydrologicalSciences-Journal-des Sciences Hydrologiques 48, 5, 709–727.

umphries, M.S., Kindness, A., Ellery, W.N., Hughes, J.C., Bond, J.K., Barnes, K.B., 2011.Vegetation influences on groundwater salinity and chemical heterogeneity in a

freshwater, recharge floodplain wetland. South Africa, Journal of Hydrology 411,130–139.

utley, L.B., O’Grady, A.P., Eamus, D., 2001. Monsoonal influences on evapo-transpiration of savanna vegetation of Northern Australia. Oecologia 126,434–443.

delling 249 (2013) 3– 18 17

Units Initial value (Section3.3)

Reference/remarks

cm 50, 30, 20 George et al. (1997) (C2)Fraction 0.35 hW , 0.3 hW , 0.1 hW Merz (2000) and

Barrett-Lennard (2008)(C2)

– 0.14, 0.12, 0.09 Jolly et al. (2002) andBarrett-Lennard (2008)(C2)

m2 m−2 1.1, 0.9 and 0.3, 0.9, 0.2and 0.1, 0.6, 0.05 and0.05

Based on directobservations and satelliteimagery (C2)

Horner, G., BakerMac, P., Nally, R., Cunningham, S., Thomson, J., Hamilton, F., 2009.Mortality of developing floodplain forests subjected to a drying climate andwater extraction. Global Change Biology 15, 2176–2186.

Indian Ocean Climate Initiative, 2010. Stage 3 Report. Prepared by the Cli-mate Change Unit. Department of Environment and Conservation, Perth,Western Australia. Available at: http://www.ioci.org.au/pdf/2010 SymposiumReport.pdf.

Johnson, W.C., Boettcher, S.E., Poiani, K.A., Guntenspergen, G.R., 2004. Influence ofweather extremes on the water levels of glaciated prairie wetlands. Wetlands24, 385–398.

Jolly, I., McEwan,K., Cox, J., Walker, G., Holland, K., 2002. Managing Groundwater andSurface Water for Native Terrestrial Vegetation Health in Saline Areas, CSIROLand and Water, Technical Report 23/02.

Kanniah, K.D., Beringer, J., Hutley, L.B., 2010. The comparative role of key environ-mental factors in determining savanna productivity and carbon fluxes: a review,with special reference to northern Australia. Progress in Physical Geography 34,459–490.

Kingsford, R.T., 2006. In: Kingsford, R.T. (Ed.), Ecology of Desert Rivers. CambridgeUniversity Press, Cambridge, UK, pp. 336–345.

Leigh, C., Sheldon, F., Kingsford, R.T., Arthington, A.H., 2010. Sequential floods drive‘booms’ and wetland persistence in dryland rivers: a synthesis. Marine andFreshwater Research 61, 896–908.

Lohammar, T., Larsson, S., Linder, S., Falk, S., 1980. FAST: simulation models ofgaseous exchange in scots pine. Ecological Bulletins 32, 505–523.

Mallik, A.U., Lamb, E.G., Rasid, H., 2001. Vegetation zonation among the micro-habitats in a lacustrine environment: analysis and application of belowgroundspecies trait patterns. Ecological Engineering 18, 135–146.

Merz, S.K., 2000. Lake Toolibin Numerical Groundwater Model. Project Report –Final. Conservation and Land Management WA, Armandale, Australia.

Mitchell, P.J., Veneklaas, E., Lambers, H., Burgess, S., 2009. Partitioning of evap-otranspiration in a semi-arid eucalypt woodland in south-western Australia.Agricultural and Forest Meteorology 149, 25–37.

Nemani, R., Running, S., 1989. Testing a theoretical climate-soil-leaf area hydrologicequilibrium of forests using satellite data and ecosystem simulation. Agriculturaland Forest Meteorology 44, 245–260.

Nicholls, N., 2004. The changing nature of Australian droughts. Climatic Change 63,323–336.

Nielsen, D.L., Chick, A.J., 1997. Flood mediated changes in aquatic micro-phyte community structure. Marine and Freshwater Research 48, 153–157.

Ojima, D.S., Lackett, J.M., 2002. Preparing for a Changing Climate: The Potential Con-sequences of Climate Variability and Change-Central Great Plains. Central GreatPlains Steering Committee and Assessment Team. Colorado State University.

Page, K.L., Dalal, R.C., 2011. Contribution of natural and drained wetland systems tocarbon stocks, CO2, N2O, and CH4 fluxes: an Australian perspective. Soil Research49, 377–388.

Petrone, K.C., Hughes, J.D., Van Niel, T.G., Silberstein, R.P., 2010. Stream flow declinein southwestern Australia, 1950–2008. Geophysical Research Letters 37, L11401.

Porporato, A., D’Odorico, P., Laio, F., Rodriguez-Iturbe, I., 2003. Hydrologic controlson soil carbon and nitrogen cycles. I. Modeling scheme. Advances in WaterResources 26, 45–58.

Puckridge, J.T., Sheldon, F., Walker, K.F., Boulton, A.J., 1998. Flow variability and theecology of large rivers. Marine and Freshwater Research 49, 55–72.

Reddy, K.R., DeLaune, R., 2008. Biogeochemistry of Wetlands: Science and Applica-tions. CRC Press, p. 774.

Robertson, A.I., Bunn, S.E., Boonand, P.I., Walker, K.F., 1999. Sources, sinks andtransformations of organic carbon in Australian floodplain rivers. Marine &Freshwater Research 50, 813–829.

Rodríguez-Murillo, J.C., Almendros, G., Knicker, H., 2011. Wetland soil organic mat-ter composition in a Mediterranean semiarid wetland (Las Tablas de Daimiel,Central Spain): insight into different carbon sequestration pathways. OrganicGeochemistry 42 (7), 762–773.

Running, S.W., Coughlan, J.C., 1988. A general model of forest ecosystem processesfor regional applications. I. Hydrologic balance, canopy gas exchange and pri-

mary production processes. Ecological Modelling 42, 125–154.

Skaggs, T., van Genuchten, M., Shouse, P., Poss, J., 2006. Macroscopic approaches toroot water uptake as a function of water and salinity stress. Agricultural WaterManagement 86, 140–149.

1 al Mo

S

T

8 J.Z. Coletti et al. / Ecologic

tephens, S.L., 2005. Forest fire causes and extent on United States Forest Servicelands. International Journal of Wildland Fire 14, 213–222.

ooth, S., McCarthy, T.S., 2007. Wetlands in drylands: geomorphological and sedi-mentological characteristics, with emphasis on examples from southern Africa.Progress in Physical Geography 31 (1), 3–41.

delling 249 (2013) 3– 18

Tuttle, C., Zhang, L., Mitsch, L.W.J., 2008. Aquatic metabolism as an indicator of the

ecological effects of hydrologic pulsing in flow-through wetlands. EcologicalIndicators 8, 795–806.

Yang, X., Wang, M., Huang, Y., Wang, Y., 2002. A one-compartment model to studysoil carbon decomposition rate at equilibrium situation. Ecological Modelling151, 63–73.