the effects of large-scale afforestation and climate change on water allocation in the macquarie...

Journal of Environmental Management (2002) 65, 369±381doi:10.1006/jema.2002.0562, available online at http://www.idealibrary.com on

1

The effects of large-scale afforestation andclimate change on water allocation in theMacquarie River catchment, NSW, Australia

Natasha Herron*², Richard Davis³ and Roger Jones§

² NSW Department of Land and Water Conservation, Suite U101, Level 1, 131±139 Monaro St,PO Box 189, Queanbeyan NSW 2620, Australia³ CSIRO Land and Water, GPO Box 1666, ACT 2601, Australia§ CSIRO Atmospheric Research, Private Bag No.1, Aspendale Victoria 3195 Australia

Received 27 April 2001; accepted 29 January 2002

Widespread afforestation has been proposed as one means of addressing the increasing dryland and stream salinityproblem in Australia. However, modelling results presented here suggest that large-scale tree planting will substantiallyreduce river ¯ows and impose costs on downstream water users if planted in areas of high runoff yield. Stream¯owreductions in the Macquarie River, NSW, Australia are estimated for a number of tree planting scenarios and globalwarming forecasts. The modelling framework includes the Sacramento rainfall-runoff model and IQQM, a stream¯owrouting tool, as well as various global climate model outputs from which daily rainfall and potential evaporation data ®leshave been generated in OzClim, a climate scenario generator. For a 10% increase in tree cover in the headwaters of theMacquarie, we estimate a 17% reduction in in¯ows to Burrendong Dam. The drying trend for a mid-range scenario ofregional rainfall and potential evaporation caused by a global warming of 0�5�C may cause an additional 5% reduction in2030. These ¯ow reductions will decrease the frequency of bird-breeding events in Macquarie Marshes (a RAMSARprotected wetland) and reduce the security of supply to irrigation areas downstream. Inter-decadal climate variability ispredicted to have a very signi®cant in¯uence on catchment hydrologic behaviour. A further 20% reduction in ¯ows fromthe long-term historical mean is possible, should we move into an extended period of below average rainfall years, suchas occurred in eastern Australia between 1890 and 1948. Because current consumptive water use is largely adapted tothe wetter conditions of post 1949, a return to prolonged dry periods would cause signi®cant environmental stress giventhe agricultural and domestic water developments that have been instituted.

Crown Copyright # 2002 Published by Elsevier Science Ltd. All rights reserved.

Keywords: salinity management, water availability, afforestation, integrated assessment.

Introduction

The natural resource base of Australia has beenplaced under considerable pressure since Europeansettlement, 210 years ago. Large tracts of forestand woodland have been cleared for cropping andgrazing, rivers have been impounded to supply

0301±4797/02/$ ± see front matter Crown

* Corresponding author. Email: nherron@dlwc.nsw.gov.au. Thiswork was undertaken while the ®rst author was working at theBureau of Rural Sciences, Canberra.

water for downstream irrigation and urban use,and nutrient and sediment loads have increased asa result of erosion of upland catchments. In recentyears, State and Commonwealth governments haveworked together to institute national policies, aimedat ameliorating the worst of this environmentaldegradation. Many of these policies have identi®eda role for the re-establishment of trees within thelandscape. Incentives to encourage greater treereplanting are currently being evaluated, andinclude, in addition to the obvious timber produc-tion value, carbon trading, biodiversity and salinitymanagement schemes.

Copyright # 2002 Published by Elsevier Science Ltd. All rights reserved.

370 N. Herron et al.

In 1997, Commonwealth and state governmentscommitted themselves to a tripling of plantationforests by 2020 (Plantation Vision 2020 Implemen-tation Committee, 1997). Although the 2020 policywas primarily designed to protect remaining areasof native forest and determine opportunities forfuture commercial use, it is also expected to con-tribute towards biodiversity, salinity managementand carbon sequestration objectives.

Australia has also argued for the use of increasesin vegetation biomass as an allowed mechanism tomeet CO2 emission targets (1990 levels plus 8%) setby the Kyoto Protocol (Australian GreenhouseOf®ce, 2000). The vegetation will sequester carbon,thereby offsetting emissions. Efforts are currentlyunderway in Australia to develop trading mechan-isms to make tree planting economically viable insub-commercial areas.

Most recently, natural resource managementpolicy has been directed towards the managementof dryland salinity. The area affected by drylandsalinisation in the Murray-Darling Basin (MDB),Australia's most productive agricultural region,is estimated to rise from about 300 000 hectares(1996 estimate) to as much as 9 million hectares(Murray±Darling Basin Commission, 1999), withrelated increases in stream salinities and salt loads(Jolly et al., 2001). In response to this threat,the Murray±Darling Basin Commission1 (MDBC),the NSW and South Australian Governmentsand the Federal Government have released salinitystrategies for their areas of governance (MDBC,2000; Department of Land and Water Conservation,2000; Primary Industries and Resources, SouthAustralia, 2000a,b). In Victoria, 21 regional salinityplans have been implemented since the release ofSalt Action: Joint Action in 1988 (Departmentof Natural Resources and Environment, 1988).Tree planting is one of the major responses proposedwithin these strategies.

While the re-introduction of trees into the land-scape will address many of the environmentalproblems that their removal caused, somenegative impacts are also anticipated. These includeeconomic losses arising from the loss of agriculturalproduction, possible social impacts (e.g. ruraldecline) and reductions in stream¯ow volumes,which could place further stresses upon waterallocations within catchments. There is suf®cient

1 The MDBC is an autonomous organisation charged with man-aging the River Murray and the Menindee Lakes system of thelower Darling River, and advising the Ministerial Council onmatters related to the use of the water, land and otherenvironmental resources of the MDB.

scienti®c evidence to indicate that a change invegetation cover from grasses (or crops) to treesleads to a reduction in mean annual runoff (Boschand Hewlett, 1982; Vertessy, 1999; Zhang et al.,2001). In a country, where the allocation ofwater resources in many catchments is alreadyover-committed or at capacity, the bene®ts arisingfrom tree re-planting must be evaluated againstany dis-bene®ts, particularly those arising froma reduced water supply.

In this paper, we investigate the impact ofa number of tree planting scenarios on stream¯owgeneration in the Macquarie River catchment inNew South Wales, Australia. We also model theprobable impacts of global warming on water avail-ability in this catchment, as climate change is likelyto compound the drying effects of tree establish-ment in this area, and must be considered in anylong term catchment management planning pro-cess. The impact of these ¯ow reductions is assessedin terms of their impact upon environmental ¯owsto a RAMSAR listed waterbird breeding area andto irrigation users in the Macquarie River catch-ment in New South Wales, Australia.

The Macquarie River catchment

The Macquarie-Bogan catchment covers an areaof about 75 000 km2. It extends from the GreatAustralian Divide (east of Sydney) to its junctionwith the Barwon River, some 560 km to the north-west (Figure 1). Mean annual rainfall varies fromaround 1000 mm in the headwaters to less than400 mm at the end of the catchment.

Virtually all the Macquarie River ¯ow is gener-ated in the headwater catchments upstream ofNarromine, and is referred to here as the contribut-ing area. Below Narromine, there is a net loss tostream¯ow due to abstractions for irrigation, evap-oration and losses to groundwater. BurrendongDam, located upstream of Wellington, is the mainstorage in the catchment supplying irrigationdemands downstream to Oxley, as well as supple-menting ¯ows to the Macquarie Marshes.

The Macquarie Marshes, located in the lowerreaches of the catchment (Figure 1), are a largeand diverse system of wetlands. They have particu-lar signi®cance as a refuge and breeding area forwaterbirds, and the Macquarie Marshes NatureReserve is listed under the Ramsar Convention ofWetlands of International Importance.

Water allocations are classed as either highsecurity or general security. High security water is

Macquarie Rive r

Bog an River

Figure 1. The Macquarie-Bogan catchment. InsetÐThe Murray-Darling Basin (light grey) and the Macquarie catchment(dark grey).

Climate change on water allocation in NSW, Australia 371

a ®xed allocation amount guaranteed in all but thelowest ¯ow years. Town water supply is a highsecurity water allocation. General security water isallocated on the basis of water availability andmost irrigation water is of this type. Every irrigatoris licensed for a ®xed volume of water, but theproportion of their entitlement that they actuallyreceive in any year is calculated once dead storageand high security water allocations have beendeducted from the dam storage volume for thatyear. As such, irrigators are more vulnerable toreductions in catchment runoff than recipients ofhigh security water, and they often receive less than100% of their licensed allocation in a water year.

Under the Water Management Plan for theMacquarie Marshes 1996 (Department of Landand Water Conservation and National Parks and

Wildlife Service, 1996), 50 GL of high security waterand 75 GL of general security water have beenallocated to the Marshes to ensure their ecologicalsustainability. These allocations supplement ¯owsfrom tributaries downstream of Burrendong Damand dam spills during periods of high runoff. TheMarshes are very sensitive to changes in ¯owregime; they have shrunk by more than 40% sincethe construction of Burrendong Dam in 1969 andthe regulation of river ¯ows for irrigation(Kingsford and Thomas, 1995).

Beef cattle, sheep, wool, wheat and cotton are themajor contributors to the agricultural economy.Wheat dominates the dryland cropping sector; cot-ton production comprises the majority of revenuefrom irrigated summer cropping, and accounts forapproximately 60% of consumptive water use in the

372 N. Herron et al.

Macquarie (Foreman et al., 1998). The area plantedto cotton more than tripled between 1985 and 1995,and has continued to expand since.

Modelling framework

A modelling system linking a climate scenario gen-erator for Australia with the rainfall-runoff andstream¯ow management models for the Macquariecatchment, currently used by the river manager,has been developed to carry out a risk assessment ofclimate change on water resources (Page and Jones,2001). To assess the added impacts of tree planting,a tree growth model which had been run for NewSouth Wales to produce a state-wide forest capabil-ity map (ABARE and BRS, 2001) was used to deter-mine the areas suitable for commercial forestrywithin the Macquarie catchment, and to guide thetree replanting scenarios.

The stream¯ow routing model

The Integrated Quantity±Quality Model (IQQM) isa hydrologic modelling tool developed by the NewSouth Wales Department of Land and WaterConversation (DLWC) (1995a) for planning andevaluating water resource management policies.The model, incorporating the Sacramento rainfall-runoff sub-model, is currently applied to rivers inNSW to investigate and resolve water-sharingissues between competing groups, including envir-onmental requirements. Thus it is well known toNSW water resource managers, and its use hereincreases the trustworthiness of the results of thepresent study in their eyes.

IQQM operates at a daily time step and can beused to simulate river system ¯ows for periods up tohundreds of years. The key component of the modelis the quantity module, which routes water downa river system subject to tributary in¯ows, losses toirrigation and other extractions. Water allocationrules are built into the model and are catchmentspeci®c. IQQM also uses the Sacramento rainfall-runoff model to generate daily stream¯ows in eachof the tributary subcatchments.

The rainfall-runoff model

The Sacramento model (Burnash et al., 1984) is alumped parameter rainfall-runoff model. Modelled

runoff for each subcatchment in the Macquarieis well calibrated with actual stream¯ow data fromgauged stream segments (DLWC, 1995b). The totalcontribution to stream¯ow from ungauged streamsections is estimated at a downstream calibra-tion point and apportioned to the ungauged sectionsaccording to contributing area and the rainfall-runoff pattern of adjacent gauged catchments.

The Sacramento model requires daily rainfalland potential or A-Class pan evaporation data.Monthly evaporation coef®cients are speci®ed foreach IQQM sub-catchment to convert point poten-tial evaporation data contained in the input data ®leto areal potential evaporation. In the Macquariecatchment, evaporation coef®cients between 0�7 and0�8 have been used for current conditions (DLWC,1995b). We change these coef®cients to mimic theeffect of changing tree cover for the scenariosmodelled.

The climate model

To model the impact of global warming inducedclimate changes on stream¯ow, new rainfall andpotential evaporation ®les were generated forinput into Sacramento model by coupling theAustralian climate scenario generator, OzClim(CSIRO, 1996) to IQQM (Page and Jones, 2001).This linking of a climate forecasting model witha water resource management tool represents anew and powerful development in water resourcemanagement. OzClim contains ten model pat-terns that can be used to explore a comprehen-sive range of climate scenarios, and is the ®rstmodel used for down-scaling potential evapor-ation sequences.

Precipitation and potential evaporation weregenerated from eight global circulation model(GCM) simulations of the enhanced greenhouseeffect. These eight models were evaluated for theirability in simulating climate in the Australianregion by the Commonwealth Scienti®c andIndustrial Research Organisation (CSIRO, 2001).Three are reported in this project:

� Max-Planck ECHAM4/OPYC3 1860-2099 simu-lation forced by the IS92a emission scenarioafter 1990 (DKRZ Model User Support Group,1992; Oberhuber, 1992);

� CSIRO DARLAM 125 km 1960±2100 simulation,IS92a emissions from 1990 (McGregor andKatzfey, 1997); and

� Hadley Centre HADCM3 1861±2100 simulation1% CO2 p.a. from 1990 (Johns et al., 1997).

Climate change on water allocation in NSW, Australia 373

All models simulate historical CO2 before 1990.The Max Planck GCM output simulates the wettestregional climate change for the Macquarie catch-ment, while the Hadley Centre GCM simulates thedriest (Jones et al., 2001). A thirdÐthe DARLAMregional climate modelÐwas also included to rep-resent a mid-range climate projection.

Climate patterns for each GCM were calculatedby regressing local change against annual averageglobal warming for the whole run to producea measure of local change per degree of globalwarming. This is done by linearly regressing localmonthly seasonal mean rainfall (or potential evap-oration) against global average temperature andtaking the slope of the relationship as the estimatedlocal response to global warming (e.g. Hennessyet al., 1998). A standardised pattern of regionalclimate change is produced that can be re-scaled bynew values of global warming, derived from simpleclimate models, to generate a range of climatechange scenarios for regional impact analysis(IPCC-TGCIA, 1999). Using this method, OzClimproduces monthly values of change for regionalrainfall and evaporation for different levels of global

Figure 2. Forest capability map of the Macquarie catchmen

warming based on the regional patterns of a givenGCM. These values are used to alter the historicalrainfall and potential evaporation data input®les providing climate change scenarios forinput into the Sacramento Model (Jones et al.,2001). Improvements in both GCMs and scalingtechniques have produced a more robust range ofprojections for seasonal changes in rainfall andevaporation than were available for an earlierstudy of the Macquarie catchment produced byWhetton et al. (1998). The regional patterns ofchange which tend towards rainfall increases insummer-autumn period and towards decreasesin the winter-spring are consistent throughoutsouthern Australia in most of the climate models(CSIRO, 2001).

The tree growth model

Tree planting scenarios were constructed froma forest capability map of the Macquarie catchment(Figure 2), generated using a spatial version ofthe Physiological Principles in Predicting Growth

t, produced using the 3-PG model.

374 N. Herron et al.

(3-PG) model (Landsberg and Waring, 1997). 3PG-SPATIAL has been modi®ed to work in a GISframework for assessing forest growth potentialover varying spatial scales (Tickle et al., 2000).Gross primary production is calculated from cli-mate, site factors and initial conditions and a setof parameters characterising the performance ofthe trees to be grown. This model had previouslybeen run for New South Wales to produce a state-wide forest capability map (ABARE and BRS, 2001)and these data were used simply to determine theareas suitable for commercial forestry within theMacquarie catchment, and to guide the tree replant-ing scenarios. Because 3-PG had been used forthis earlier government policy work, its use hereincreased the acceptability of the current project towater resource managers.

Figure 2 shows forest capability in terms ofbiomass production potential for a generic hard-wood timber. Commercially viable productionpotential (12 m3/ha/yr) is associated with the higherrainfall areas in the southeast of the catchment.Yields become increasingly marginal towards thenorthwest. Tree planting in areas of lower pro-duction potential might be pursued for salinitymanagement, carbon sequestration or biodiversityreasons should markets develop for these productsor subsidies be made available.

Scenarios and forecasts

The tree planting scenarios

In the work described here, we choose three scenar-ios that re¯ect high, medium and low levels of treeplanting, including both commercially viable andsub-commercial plantings.

Scenario 1Ðhigh-level adoptionÐassumes that allland that is currently classed as commerciallycapable will be planted to trees by 2030. In theheadwater areas around Oberon, Bathurst andOrange, this means relatively high proportions ofthe sub-catchments will be forested. There are lowto no increases in tree cover in sub-catchmentsfurther downstream. Overall, this scenario corre-sponds to a further 10% of the contributing areaplanted to trees.

Scenario 2Ðlow-level adoptionÐassumes thatonly 20% of the commercially viable land isplanted. Again plantings are focussed in the uppercatchment. This represents only a 2% increase inthe area planted to trees.

Scenario 3Ðmoderate level adoptionÐincludesthe 20% of commercially viable land in Scenario 2,plus 10% of the more marginal 8±12 m3/ha/ybiomass production class being planted. Theinclusion of trees in the more marginal productionareas represents tree planting for salinity man-agement, carbon sequestration targets and/orbiodiversity reasons. The 10% level of adoption inthese marginal areas is consistent with observa-tions over the last 7 years in the mid-Macquarie ofthe willingness of farmers to shift from cropping/grazing into tree production (Alan Nicholson,Department of Land and Water Conservation,pers. comm.). Under this scenario, 5% of thecontributing area is planted to trees.

Vegetation-runoff relationships

The impact of tree planting on runoff was cal-culated using the empirical relationship of Zhanget al. (2001) which relates mean annual evapo-transpiration (ET) to mean annual precipitation(P) and the proportion of forest and grass coverwithin a catchment using a world-wide, 295 catch-ment data set.

ET � P1�w�Ez=P�

1�w�Ez=P�� P=Ez

� ��1�

where w is the plant available water coef®cient andEz is the potential evaporation parameter. Forforested areas, Zhang et al. (2001) estimate w to be2 and Ez to be 1410; for grassed areas, w is 0�5 andEz is 1100. The relationship assumes that at anannual time-step there is no change in soil waterstorage. Recharge is also assumed to be negligible.

The Zhang curves (Figure 3) predict that for thesame mean annual precipitation, more runoff isgenerated from a grassland than a forest. Thedifference in runoff increases with increasingmean annual precipitation, such that increasingtree cover in a high rainfall area will cause a biggerloss in runoff than planting in a low rainfall area.There is a signi®cant difference between the twocurves, as shown by the 95% con®dence bands inFigure 3.

The Zhang curves were incorporated into a GISto estimate mean annual evapotranspiration forsub-catchments in the Macquarie catchment,using the approach of Vertessy and Bessard(1999), but excluding their adjustment for elevation.Areas not covered by trees were assumed to be

0 500 1000 1500 2000 2500

Mean Annual Precipitation (mm)

Mea

n A

nnua

l S

trea

mflo

w (

mm

)

Grass

Trees

0

200

400

600

800

1000

1200

1400

1600

1800

Figure 3. Stream¯ow as predicted from Zhang et al.(2001), with 95% con®dence limits shown for each curve.These curves describe the empirical relationship betweenstream¯ow and mean annual precipitation for forested andgrassed catchments. It is assumed that on an annual time-step, there is no change in soil water storage, such that thecatchment water balance can be described by Q�PÿET.

Climate change on water allocation in NSW, Australia 375

grassland. For the current condition, the proportionof each sub-catchment area under trees was deter-mined from the current land use map. The addi-tional area of each sub-catchment under trees foreach of the three scenarios was based upon theforest capability for that sub-catchment as deter-mined by the 3-PG model. Mean ET for each gridcell within a sub-catchment could then be calculatedfor each scenario using Eq (1): for forested cells, 1ÿfis zero and the second term on the RHS of theequation drops out, while for grassed cells, f is 0and the ®rst term drops out.

Given the assumptions of no deep recharge andnegligible soil water storage on an annual basis,mean annual runoff could be calculated for eachgrid cell from

Q � Pÿ ET �2�

A total mean annual runoff for each sub-catchment was determined for current conditions,and for each of the 3 scenarios, by summing therunoff from all grid cells within each sub-catchment. When compared to their correspondingcalibrated IQQM mean annual stream¯ow volumes,runoff estimates for current conditions based on theZhang formula were mostly over-estimated, and

became less reliable with decreasing annual rainfall.Since the IQQM-Sacramento model is calibrated forthe Macquarie catchment under current conditions(DLWC, 1995b) and this calibration needs to beretained, the IQQM stream¯ow volumes undercurrent condition were scaled by the ratio of theZhang runoff estimate for each tree planting scen-ario to the Zhang runoff estimate under currentconditions, to obtain predicted IQQM stream¯ow,as follows:

QIQQM�predicted�

� QZhang(predicted)

QZhang�current� �QIQQM�current� �3�

The Zhang approach is based on long-term meanannual catchment responses and is, therefore, notappropriate for analysis of intra-annual behaviour.Analysis of the IQQM ¯ow data at daily, monthlyand annual time-steps indicates signi®cant differ-ences in the ¯ow distributions for tree and grasscover at the daily and, to a lesser extent, monthlytime-steps. These differences are attributed to thein¯uence of antecedent soil moisture on runoffgeneration over relatively short time-scales. At theannual time-step, ¯ow distributions were notfound to differ signi®cantly between forested andgrassed catchments. Therefore, despite the use of adaily time-step model (IQQM) to quantify changesin stream¯ow with changes in vegetation cover, onlyannual stream¯ow totals and inter-annual compar-isons are reported in the results.

The monthly evaporation coef®cients within eachsub-catchment systems ®le in the Sacramentomodel were changed until the ¯ow generated bythe model yielded runoff volumes approximatingthe predicted stream¯ows for each scenario in thatsub-catchment. IQQM was run with these newtributary in¯ows for the upper catchment to routethe water through Burrendong Dam and throughthe lower part of the catchment.

Climate scenarios

Global warming was simulated by modifying theinput climate ®les to the Sacramento model.Global temperatures were assumed to rise by0�5�C by 2030, the median estimate from theInter-governmental Panel on Climate Change(IPCC, 1996) and the low estimate from IPCC(2001). Three climate scenarios were produced byscaling changes per degree of global warming from

No C lim ateCh ang e M ax P lan ck DARL AM

Had le yCe n t re

% C

han

ge

in

Str

ea

mfl

ow

No tree replanting

A ll commerc ially capable land

20% commerc ially capable land

20% commerc ially capable + 10%marginal

–50

–40

–30

–20

–10

0

10

Figure 4. % Change from current long term annualaverage in¯ow to Burrendong Dam for tree planting andglobal warming scenarios.

376 N. Herron et al.

the Max-Planck, Hadley Centre and DARLAMGCMs described above by 0�5�C.

To evaluate the in¯uence of episodic shifts inclimate regime on stream¯ow in the Macquariecatchment, the 107 year stream¯ow record wasdivided into two periodsÐa relatively dry period be-tween 1890 and 1948 and a wetter regime between1949 and 1996, roughly re¯ecting the droughtdominated regime (DDR) and the ¯ood dominatedregime (FDR) as de®ned by Warner (1987).

Results

UpstreamÐdam in¯ows

Figure 4 shows the estimated reductions in in¯owsto Burrendong Dam for each of the tree plantingscenarios and climate forecasts. Under currentconditions, the mean annual in¯ow to the dam isapproximately 1080 GL/year. The ®rst column ofpoints represents the impact of the tree plantingscenarios with no global warming. Tree plantingleads to reductions in mean annual in¯ows of 17%(182 GL) for Scenario 1, 4% (45 GL) for Scenario 2and 7% (75 GL) for Scenario 3. Changes in in¯ows toBurrendong Dam arising from global warming withno tree planting (̂ ) range from �1 and ÿ16% ofcurrent ¯ow, depending on the GCM used. Whenboth tree planting and global warming are com-bined, ¯ows in the upper Macquarie River are

Drought Dominated Period

-60

-50

-40

-30

-20

-10

0

10

20

30

NoClimateChange Max Planck DARLAM

HadleyCentre

%C

hang

e fr

om C

urre

nt L

ongt

erm

A

vera

ge

No tree re plantin g

All commer cially capable land

Figure 5. % Change from current long term annual average i¯ood and drought dominated regimes.

predicted to decrease by between 3% (40 GL) and30% (390 GL), depending on the planting scenarioand the GCM used.

The results change signi®cantly when thedrought and ¯ood dominated periods (DDR andFDR) are considered separately (Figure 5). Forconditions resembling the FDR, ¯ows are likely tobe greater than the long-term average in all but the

HadleyCentre

Flood Dominated Period

-60

-50

-40

-30

-20

-10

0

10

20

30

No ClimateChange Max Planck DARLAM

20% commer cially capable land

20% commer cially capable + 10% margina l

n¯ow to Burrendong Dam when ¯ow record is separated into

Climate change on water allocation in NSW, Australia 377

largest tree-planting scenario modelled (i.e.Scenario 1). That is, the increase in ¯ows fromincreased precipitation more than outweighsthe reduced runoff from the trees and global warm-ing. With no tree planting and the wettest GCMprojection, ¯ows will increase by as much as 29%over the current long-term average.

If DDR conditions were to develop again, reduc-tions in stream¯ow are likely to be considerablygreater than suggested from the 107-year record inFigure 4. In the absence of global warming and treeplanting and using 1890 to 1948 rainfall records,average ¯ows were 22% lower than the longer 107-year average. When global warming and tree plant-ings are included in the predictions, reductions in¯ow of as much as 46% are estimated.

DownstreamÐirrigation water

These reductions in stream¯ow into BurrendongDam have signi®cant effects on downstream waterusers, particularly those that rely on general secur-ity water. Except for very low ¯ow years when theallocation of general security water is less than orequal to 10%, town water supply will receive its fullallocation. However, the frequency of low ¯ow yearsis expected to increase, such that town water supplymay be at risk 5% more often.

Irrigators in the catchment are expected toexperience more signi®cant impacts. Figure 6shows allocation reliability for the wettest GCMprojection with no tree planting (the most optimisticforecast), the driest GCM projection with treeplanting scenario 1 (the most pessimistic forecast)

% Time Flow Exceeds

% o

f A

lloc

atio

n

Current

(i)

(ii)

(iii)0

10

20

30

40

50

60

70

80

90

100

0 10 20 30 40 50 60 70 80 90 100

Figure 6. Allocation reliability for (i) the wettest GCMprediction with no tree planting, (ii) the driest GCMprediction with tree planting Scenario 1 and (iii) the middleGCM prediction with tree planting Scenario 3. Currentallocation reliability is also shown.

and the middle GCM projection with tree plantingScenario 3 (mid-range forecast). Allocation reliabil-ity varies signi®cantly with scenario.

The most pessimistic forecast for irrigators isthat the frequency with which they can expecttheir full allocation will drop from the current 43%of the time to about 21% of the time. Under thesedrier conditions, irrigators are predicted to havean allocation of less than 42% of their currenttotal entitlement 50% of the time. On average,allocations will tend to be 27% less than undercurrent conditions under this pessimistic forecast,with the greatest differences in allocation reliabilityoccurring in the 30±60% probability range.

The most optimistic outcome, the wettest GCMand no tree planting, indicates a very minor reduc-tion in the frequency of full allocations, with theoverall distribution not differing signi®cantly fromcurrent allocation reliability. The mid-range out-come (middle GCM and tree planting Scenario 3)indicates that irrigators can expect their full alloca-tion 32% of the time (compared to 43% of the time atpresent), and greater than 65% of their allocation50% of the time (490% at present).

When DDR and FDR periods are separated out(Figure 7), a return to an extended below-averageprecipitation period would compromise irrigationenterprises further. For the most optimistic out-come, an 18% reduction from the long-term meanannual allocation is predicted; for the most pessim-istic outcome the reduction from the current long-term average is close to 30%. Under a wetter regime,a 30% increase on the current long-term averageallocation is estimated for the most optimisticscenario, and a 13% reduction for the most pessim-istic outcome.

% Time Allocation Exceeds

% o

f A

lloca

tion

Longterm - Current(i) fdr(i) ddr(ii) fdr(ii) ddr

0

10

20

30

40

50

60

70

80

90

100

0 10 20 30 40 50 60 70 80 90 100

Figure 7. Allocation reliability when ¯ow record isseparated into ¯ood and drought dominated regimes for(i) the wettest GCM prediction with no tree planting and(ii) the driest GCM prediction with tree planting Scenario 1.Current allocation reliability is also shown.

378 N. Herron et al.

DownstreamÐMacquarie Marshes

Large ¯ows of water in the Macquarie are the key tothe extensive breeding of ibis, herons, egrets andspoonbills in the Marshes (Kingsford and Johnson,1998; Johnson, 1998). The quantity of water deter-mines the extent of the habitat created for breeding;breeding populations are proportional to the area¯ooded. For these colonially nesting water birds, aminimum annual ¯ow of about 200 GL is requiredbefore bird breeding will occur, although greaterreliability of breeding comes with annual ¯ows of4250 GL. For large-scale breeding (450 000 pairs),annual ¯ows exceeding 350 GL are needed(Johnson, 1998). The high security and generalsecurity water allocated to the Marshes under theMacquarie Marshes Water Management Plan is stillwell below the quantity required for bird breeding.

% Time Flow Exceeds

Flo

w t

o M

acqu

arie

Mar

shes

(M

L)

Current

(i)(ii)

(iii)

No bird breeding

Large scale bird breeding: >350 GL/y

Small to moderate bird breeding

10000

100000

1000000

10000000

0 10 20 30 40 50 60 70 80 90 100

Figure 8. Impact of ¯ow reductions on bird breedingevents in the Macquarie Marshes for (i) the wettest GCMprediction with no tree planting, (ii) the driest GCMprediction with tree planting Scenario 1 and (iii) the middleGCM prediction with tree planting Scenario 3. Currentallocation reliability is also shown.

–50 –40 –30 –20 –10 0 10 20 30Current

Trees

Global warming

Trees + Glob al warming

Floo d-dominated period

Drought-dominated

% Change in streamflow from current conditions

Figure 9. Summary of results, showing relative signi-®cance of tree planting, global warming and climatevariability scenarios on ¯ow reductions in the Macquariecatchment.

Unregulated ¯ows from tributaries downstreamof Burrendong Dam are needed to bring total ¯owsup to threshold volumes for bird breeding.

The impacts of ¯ow reductions to the MacquarieMarshes are illustrated in Figure 8. The probabilityof receiving 200 GL decreases from 0�84 to 0�66 and0�8 for the driest and mid-range predictions,respectively. For large-scale bird breeding events,reductions of a similar magnitude occur, with theprobability of ¯ows 4350 GL decreasing from 0�54to 0�32 and 0�44, respectively. Overall there isa 18% reduction in mean annual ¯ows to theMarshes for the driest scenario and an 12% reduc-tion for the mid-range predictions, in spite of theMarshes receiving some of their allocation as highsecurity water.

The results presented above are summarised inFigure 9. Reductions of between 4 and 17% of thecurrent long-term average in¯ows to BurrendongDam were simulated for the three tree plantingscenarios modelled here. Since the trees are plantedpreferentially in the highest forest capability land,the effect of planting is `magni®ed' in the river¯ows. Thus, for example, a 10% increase in trees inthe upper catchment will result in approximately17% less river ¯ow.

Reductions in stream¯ow arising from globalwarming, including the driest outcome, werefound to be comparable to the reductions arisingfrom tree planting Scenario 1. If we ignore second-ary effects of higher temperatures and CO2 levelson forest productivity, the impacts of bothchanged land use and global warming are additive.Stream¯ow is predicted to be reduced by 30% ofthe current long term mean for Scenario 1, usingthe Hadley Centre climate projections. Jones et al.(2001) estimate changes in mean annual ¯owbased on the entire range of climate change in2030 taken from IPCC (1996) of 0�4±0�8�C as ran-ging from �1 to ÿ21%. The recent IPCC projectionsof 0�5±1�3�C in 2030 (IPCC, 2001) increase thisrange to �1 to ÿ30%, although the most likelyrange is 0 to ÿ15% (Jones and Page, 2001). Whilethere is considerable uncertainty about the exactmagnitude of the stream¯ow changes likely toarise from global warming and extensive tree plant-ing, the combined effects will certainly reducestream¯ow.

Discussion

With only general security allocations, the irrigatorsare particularly sensitive to long term stream¯ow

Climate change on water allocation in NSW, Australia 379

reductions in the Macquarie River. A reduction inmean annual stream¯ow translates to a reductionin the fraction of their entitlement received byirrigators. This, in turn, translates to a reductionin the sustainable crop area (de®ned as the areaharvested) and/or a change in crop mix. We havenot modelled the economic implications of thesereduced allocations because we do not have infor-mation on the likely changes in crop mix. However,we estimate that, in the most extreme scenario,irrigators will receive their full entitlements asmuch as 50% less often than at present.

The high security water allocation means thatthe Marshes are less vulnerable than irrigators tolong term ¯ow reductions. Nevertheless, the eco-logical functioning of the Marshes will be threa-tened should there be a long-term reduction instream¯ow. Impacts on the frequency of bird breed-ing events may be even larger than indicated byFigure 8 because the duration and magnitude of¯ows over shorter periods within the year are morecritical determinants of bird breeding events thanthe total annual ¯ow (Department of Land andWater Conservation and National Parks andWildlife Sevice, 1996). It is likely that bird breedingis determined by the ¯ow during a 5±7 month periodsurrounding the median breeding period inOctober.

Water availability in the Macquarie River ismore vulnerable to shifts in the rainfall regimeover periods lasting several decades than it is toeither tree planting or global warming by 2030. By2070, however, Jones et al. (2001) show that theimpacts of global warming on water availabilitycould outweigh the in¯uence of decadal shifts inrainfall regime. Decadal variability can decreasestream¯ow by as much as 20% or increase it by asmuch as 25% from the long-term historical average.Given that the water supply infrastructure hasbeen developed under the relatively favourablerainfall of an FDR, a return to below averagerainfall would signi®cantly stress the system basedon its current operating rules. Thus, all but themost optimistic water resources managers need toplan for a period of greater competition for morelimited water resources over the next 30 years.

Of the three changes modelled here, one (globalwarming) will almost certainly occur (IPCC, 2001),one (inter-decadal climate shifts) is a stochasticevent whose occurrence cannot be predicted andthe third (tree planting) is a societal choice andmay or may not occur. The decision whether topermit tree planting depends on the balancebetween costs and bene®ts to both private andpublic consumers. A parallel study (Heaney et al.,

2000; Heaney and Beare, 2000) predicts thatwidespread forestation in the headwater areaswould provide only modest, if any, long termreductions in stream salinities, yet impose a largenet cost from lower returns from irrigated agricul-ture and forest based land uses. The potentialbene®ts from carbon sequestration, local salinitycontrol and improved biodiversity were not inclu-ded in these calculations.

This paper illustrates some of the complexitiesassociated with implementing government policies.The promotion of salinity management, forestrysupply and improved biodiversity through treeplanting will be at the expense of reduced river¯ows. If the objectives of the environmental ¯owpolicies of the Australian Commonwealth and Stategovernments are also to be met, this implies lesswater for irrigation and other consumptive uses.The analysis in this paper needs to be supplementedwith an assessment of dryland agriculture, carbonsequestration, biodiversity improvements andopportunities for enhancing water use ef®cienciesin irrigated areas to gain an improved understand-ing of the costs and bene®ts from tree planting inthis catchment.

Acknowledgments

The authors wish to thank Cher Page at CSIROAtmospheric Research for running the combined globalwarming and tree planting scenarios; Rob O'Neill fromthe Department of Land and Water Conservation inNSW who guided and advised us in running IQQM andsupplied much needed information about water allocationprocedures in the Macquarie catchment; Adrian Buggfrom the Bureau of Rural Sciences who provided the forestcapability assessment for the Macquarie catchment; PeterHairsine from CSIRO Land and Water and Tom Aldredfrom the Natural Resource Management Policy Divisionin the Department of Agriculture, Fisheries and Forestry,Australia, for their incisive reviews of this paper; and theRural Industry Research and Development Corporationwho provided funding for the development of the OzClim-IQQM modelling system.

References

ABARE and BRS (2001). An Assessment of thePotential for Plantation Development in New SouthWales. 102 pp. Report for NSW Plantations Taskforceprepared by the Australian Bureau of Agricultural andResource Economics (ABARE) and Bureau of RuralSciences (BRS), Canberra.

380 N. Herron et al.

Australian Greenhouse Of®ce (2000). Greenhouse Sinksand the Kyoto ProtocolÐAn Issues Paper, 108 pp.Australian Greenhouse Of®ce, Canberra.

Bosch, J. M. and Hewlett, J. D. (1982). A review ofcatchment experiments to determine the effect ofvegetation changes on water yield and evapotranspira-tion. Journal of Hydrology 55, 3±23.

Burnash, R. J. E., Ferral, R. L. and McGuire, R. A.(1984). A Generalised Stream¯ow Simulation System,Joint Federal-State River Forecast Centre, Sacramen-to, California.

Commonwealth Scienti®c and Industrial ResearchOrganisation (2001). Climate Change Projections forAustralia, 8 pp. CSIRO Atmospheric Research, Mel-bourne. (http://www.dar.csiro.au/publications/projec-tions2001.pdf)

Commonwealth Scienti®c and Industrial ResearchOrganisation (1996). OzClim: A Climate ScenarioGenerator and Impacts Package for Australia. CSIROAtmospheric Research, Melbourne. (http://www.dar.csiro.au/res/cm/ozclim.htm)

Department of Land and Water Conservation(2000). New South Wales Salinity Strategy,Report prepared for NSW Government, NSW DLWC,Sydney.

Department of Land and Water Conservation (1995a).Integrated Quantity-Quality Model (IQQM) ReferenceManual, NSW DLWC, Sydney.

Department of Land and Water Conservation (1995b).IQQMÐMacquarie River System Calibration Report,TS94.041, NSW DLWC, Sydney.

Department of Land and Water Conservation andNational Parks and Wildlife Service (1996). TheMacquarie Marshes Water Management Plan 1996,40 pp. DLWC and NPWS, Sydney.

Department of Natural Resources and Environment(1988). Salt Action: Joint Action, DNRE, Melbourne.

DKRZ-Model User Support Group (eds) (1992).ECHAM3ÐAtmospheric General Circulation Model.Report No. 6, Deutsches Klimarechenzentrum,Hamburg. (http://www.dkrz.de/forschung/reports/ReportNo.6.ps)

Foreman, I., Fritsch, S. and Woldring, H. (1998).Consequences for the agricultural economy, In Hassalland Associates, Climate Change Scenarios and Mana-ging the Scarce Water Resources of the MacquarieRiver, pp. 46±60. Report prepared for the AustralianGreenhouse Of®ce, Canberra.

Heaney, A. and Beare, S. (2000). Basin Modeling ofIntervention Practices and Policy Options for SalinityManagement, Report on Preliminary Results for theGoulburn-Broken and Macquarie-Bogan Catchments,Murray Darling Basin Commission and ABAREProject, Canberra.

Heaney, A., Beare, S. and Bell, R. (2000). Targetingreforestation for salinity management, AustralianCommodoties 7, 511±518.

Hennessy, K. J., Whetton, P. H., Katzfey, J. J., McGregor,J. L., Jones, R. N., Page, C. M. and Nguyen, K.C.(1998). Fine resolution climate change scenarios forNew South Wales: annual report 1997±1998. iii, 48 pp.NSW Environment Protection Authority, Chatswood,N.S.W.

Intergovernmental Panel on Climate Change (1996).Technical Summary, in Houghton, J. T., Meira Filho,

L. G., Callander, B. A., Harris, N., Kattenberg, A.and Maskell, K. (eds), Climate Change 1995: TheScience of Climate Change, Contribution of WorkingGroup I to the second assessment report of the IPCC,pp. 9±49. Cambridge University Press, Cambridge.

Intergovernmental Panel on Climate Change (2001).Summary for Policymakers. In Climate Change 2001:The Scienti®c Basis (J. T. Houghton, Y. Ding, D. J.Griggs, M. Noguer, P. van der Linden and D. Xiaosu,eds.) pp. 1±20. Cambridge. Cambridge UniversityPress (http://www.unep.ch/ipcc/pub/spm22-01.pdf)

IPCC-TGCIA (1999). Guidelines on the Use of ScenarioData for Climate Impact and Adaptation Assessment.Version 1. Prepared by Carter, T. R., Hulme, M. andLal, M., 69 pp. IPCC Task Group on Scenarios forClimate Impact Assessment.

Johns, T. C., Carnell, R. E., Crossley, J. F., Gregory, J. M.,Mitchell, J. F. B., Senior, C. A., Tett, S. F. B. andWood, R. A. (1997). The second Hadley Centre coupledocean-atmosphere GCM: model description, spinupand validation, Climate Dynamics 13, 103±134.

Johnson, W. (1998). Consequences for the MacquarieMarshes. In Climate Change Scenarios and Managingthe Scarce Water Resources of the Macquarie River(Hassall and Associates, ed.), pp. 61±75. Report pre-pared for the Australian Greenhouse Of®ce, Canberra.

Jolly, I. D., Williamson, D R., Gilfedder, M., Walker, G. R.,Robinson, G., Jones, H., Zhang, L., Dowling, T. I.,Dyce, P., Nathan, R. J., Nandakumar, N., Clarke, R.and McNeill, V. (2001). Historical stream salinitytrends and catchment salt balances in the Murray-Darling Basin, Australia, Marine and FreshwaterResearch 52, 53±63.

Jones, R. N., Page, C. M., Herron, N. F., Davis, R. J.,O'Neill, R., Bennett, D. and McClintock, D. (2001).Climate change and the risk to long-term water supplyin the Murray-Darling Basin, AWA 19th Con-ventionÐA Water Odyssey, Canberra, 2±4 April,2001. Australian Water Australia, Paper 198.

Kingsford, R. T. and Johnson, W. (1998). Impact ofwater diversions on colonially-nesting waterbirds inthe Macquarie Marshes of arid Australia, ColonialWaterbirds 21, 159±170.

Kingsford, R. T. and Thomas, R. F. (1995). TheMacquarie Marshes in arid Australia and their water-birds: a 50 year history of decline. EnvironmentalManagement 19, 867±878.

Landsberg, J. J. and Waring, R. H. (1997). A generalisedmodel of forest productivity using simpli®ed conceptsof radiation-use ef®ciency, carbon balance, and parti-tioning. Forest Ecology and Management 95, 209±228.

McGregor, J. L. and Katzfey, J. J. (1997). Climatemodelling for the Australian region using DARLAM,Proceedings of the International Workshop on RegionalModeling of the `General Monsoon System' in Asia,Beijing, China (TEACOM Report, 4), pp. 13±18.START Regional Centre for Temperate East Asia.

Murray-Darling Basin Commission (2000). Draft BasinSalinity Management Strategy 2001±2015, 36 pp. MDBMinisterial Council, September 2000, Canberra.

Murray-Darling Basin Commission (1999). The SalinityAudit of the Murray-Darling BasinÐA 100 YearPerspective, 39 pp. MDBC, Canberra.

Oberhuber, J. M. (1992). The OPYC Ocean GeneralCirculation Model, Report No. 7, Deutsches

Climate change on water allocation in NSW, Australia 381

Klimarechenzentrum, Hamburg. (http://www.dkrz.de/forschung/reports/ReportNo.7.ps)

Page, C. M. and Jones, R. N. (2001). OzClim: Thedevelopment of a climate scenario generator forAustralia. In MODSIM 2001: International Congresson Modelling and Simulation: Proceedings, 10±13December 2001, Canberra, ACT (F. Ghassemi,H. Whetton, R. Little and M. Littleboy, eds.) pp. 667±671. Canberra: Modelling and Simulation Society ofAustralia and New Zealand.

Plantation Vision 2020 Implementation Committee(1997). Plantations for Australia: The 2020 Vision,Department of Primary Industries and Energy,Canberra ACT. (http://www.affa.gov.au/docs/forestry/plantations/vision.html)

Primary Industries and Resources South Australia(2000a). Draft State Dryland Salinity Strategy, 43 pp.Prepared by the Primary Industries and ResourcesSouth Australia for the Soil Conservation Council ofSouth Australia, October 2000.

Primary Industries and Resources South Australia(2000b). Draft State South Australian RiverMurray Salinity Strategy, 27 pp. Prepared by thePrimary Industries and Resources South Australia forthe Soil Conservation Council of South Australia,October 2000.

Tickle, P. K., Coops, N. C., Hafner, S. D. and The BagoScience Team (2000). Assessing forest productivity atlocal scales across a native eucalypt forest using a

process model, 3PG-SPATIAL, Forest Ecology andManagement 5366, 1±17.

Vertessy, R. A. (1999). The impacts of forestry onstream¯ows: A review, In Forest Management forWater Quality and Quantity (J. Croke and P. Lane,eds), pp. 93±109. Research Centre for CatchmentHydrology, Report 99/6, Canberra.

Vertessy, R. A. and Bessard, Y. (1999). Anticipatingthe negative hydrologic effects of plantation expan-sion: Results from a GIS-based analysis on theMurrumbidgee Basin. In Forest Management for WaterQuality and Quantity (J. Croke and P. Lane, eds),pp. 69±73. Cooperative Research Centre for CatchmentHydrology, Report 99/6, Canberra.

Warner, R. F. (1987). The impacts of alternating¯ood-and drought-dominated regimes on channelmorphology at Penrith, New South Wales, Australia,International Association of Hydrological SciencesPublication 68, 327±338.

Whetton, P., Jones, R. and Pittock, B. (1998). ClimateChange Methodology and Results. In Climate ChangeScenarios and Managing the Scarce Water Resourcesof the Macquarie River (Hassall and Associates, eds),pp. 18±26. Report prepared for the AustralianGreenhouse Of®ce, Canberra.

Zhang, L., Dawes, W.R. and Walker, G.R. (2001). Theresponse of mean annual evapotranspiration to vegeta-tion changes at catchment scale. Water ResourcesResearch 37, 701±708.