dryland salinity management - csiro · the original case-study was funded through nrms-lwrrdc grant...

KN

OW

LE

DG

E

Land

scap

es &

Indu

strie

s

Dryland SalinityManagement: CAN SIMPLE CATCHMENT-SCALE MODELS PROVIDERELIABLE ANSWERS? AN AUSTRALIAN CASE STUDY

Mirko Stauffacher, Glen Walker, Warrick Dawes, Lu Zhang, Peter Dyce

Dryland SalinityManagement: CAN SIMPLE CATCHMENT-SCALE MODELS PROVIDERELIABLE ANSWERS? AN AUSTRALIAN CASE STUDY

Mirko Stauffacher, Glen Walker, Warrick Dawes, Lu Zhang, Peter Dyce

Authors: Mirko Stauffacher1, Glen Walker2, Warrick Dawes1, Lu Zhang1, Peter Dyce1

1. CSIRO Land and Water, GPO Box 1666, Canberra ACT 26012. CSIRO Land and Water, PMB 2, Glen Osmond SA, 5064

CSIRO Land and Water Technical Report 27/03MDBC Publication 13/03

Published by: Murray-Darling Basin Commission

Level 5, 15 Moore Street

Canberra ACT 2600

Telephone: (02) 6279 0100

from overseas + 61 2 6279 0100

Facsimile: (02) 6248 8053

from overseas + 61 2 6248 8053

Email: [email protected]

Internet: http://www.mdbc.gov.au

ISBN: 1 876830 53 0

Cover photo: Arthur Mostead Margin photo: Mal Brown

© 2003, Murray-Darling Basin Commission and CSIRO

This work is copyright. Photographs, cover artwork and logos are not to be reproduced, copied or stored by any process without the written

permission of the copyright holders or owners. All commercial rights are reserved and no part of this publication covered by copyright may be

reproduced, copied or stored in any form or by any means for the purpose of acquiring profit or generating monies through commercially exploiting

(including but not limited to sales) any part of or the whole of this publication except with the written permission of the copyright holders.

However, the copyright holders permit any person to reproduce or copy the text and other graphics in this publication or any part of it for the

purposes of research, scientific advancement, academic discussion, record-keeping, free distribution, educational use or for any other public use or

benefit provided that any such reproduction or copy (in part or in whole) acknowledges the permission of the copyright holders and its source

(‘Dryland salinity management: can simple catchment-scale models provide reliable answers? An Australian case study’) is clearly acknowledged.

To the extent permitted by law, the copyright holders (including its employees and consultants) exclude all liability to any person for any

consequences, including but not limited to all losses, damages, costs, expenses and any other compensation, arising directly or indirectly from using

this report (in part or in whole) and any information or material contained in it.

The contents of this publication do not purport to represent the position of the Murray-Darling Basin Commission or CSIRO in any way and are

presented for the purpose of informing and stimulating discussion for improved management of Basin's natural resources.

The writing of this report was funded through Murray-Darling Basin Commission StrategicInvestigations and Education Program Grant Number D9004: ‘Catchment characterisation andhydrogeological modelling to assess salinisation risk and effectiveness of management options’.

The original case-study was funded through NRMS-LWRRDC Grant D6026 ‘Improving DrylandSalinity Management through Integrated Catchment Scale Management’. We thank all the teaminvolved in this complex project for their valuable input; and the Liverpool Plains community fortheir help and support.

Acknowledgements

DRYLAND SALINITY MANAGEMENT: CAN SIMPLE CATCHMENT-SCALE MODELS PROVIDE RELIABLE ANSWERS?

i

Executive summary

DRYLAND SALINITY MANAGEMENT: CAN SIMPLE CATCHMENT-SCALE MODELS PROVIDE RELIABLE ANSWERS?ii

Introduction

Prior to this study, there was no clearunderstanding of the salinisation processes,their cause and extent in the Liverpool Plainscatchment. For this work, we did a first-cutwhole catchment analysis and then, basedon the outcomes of this first appraisal,focused the research down to the areas thatare extensively affected by salinity. For thesetwo salt affected subcatchments, wedeveloped a conceptualisation of thesalinisation processes; developed somesimple models that capture thisconceptualisation and illustrated how thesemodels were calibrated and applied fordifferent scenarios. From the scenariomodelling, we suggest the impacts ofdifferent options for this catchment withsome assessment of confidence.

The Liverpool Plains catchment is on thewestern edge of the Great Dividing Range innorthern New South Wales. The dominantaquifer in the catchment is a semi-confinedintermediate to regional scale alluvial aquifer,with recharge beds located at the upstreamend at the foot of hills and more ruggedcountry. This report represents a broadscalescoping study with the following aims:

1. to define the processes leading tohigh watertables and land salinisationin the Liverpool Plains

2. to develop a conceptual model of thewater and salt sources and sinkswithin the catchment

3. to apply a range of simple analyticaltools to both surface and groundwaterdata

4. to simulate future groundwater trendsin the catchment with a variety ofrecharge reduction targets.

Site description

The Liverpool Plains catchment lies within theNamoi River catchment in northern NewSouth Wales. The study area covers 11,728 km2 south of the town of Gunnedah.Ephemeral drainage occurs within the

subcatchments of the Liverpool Plains, withephemeral streams of the Mooki River andCox’s Creek draining to the Namoi River, andinto the Murray-Darling River system. Meanannual rainfall varies from just over 1000mm/yr at the top of the Liverpool Ranges inthe south, to around 550 mm/yr on theplains proper. Rainfall is summer dominantand highly variable, with high intensitythunderstorms causing flooding on theplains. Sheep and cattle grazing were thedominant land use in the mid-1800s; thischanged to cropping on the red soil of thefoot-slopes of the ranges early in the centuryuntil the 1970s. In the early 1960s, advancesin technology allowed the heavy-textured clayplains to become the main agricultural area,supporting a range of dryland annual crops.The steeper hills and ridges of the LiverpoolPlains have never been cleared and thusremain covered by native eucalypts.

Groundwater system andsalinity

The Liverpool Plains aquifer is anIntermediate to Regional flow system inAlluvial Deposits (Cainozoic/MesozoicVolcanics) (Coram et al. 2000). The hydraulicgradient of groundwater flow follows thegeneral topography, from the foothills of theLiverpool Ranges across the plains towardsthe Mooki River. The main aquifer is made upof gravel and sand and is locally called theGunnedah Formation, while the semi-confining heavy clay layer over the top iscalled the Narrabri Formation. The LiverpoolPlains catchment can be conceptualised asfive almost independent groundwatersystems, separated by bedrock highs.

There are two main complicating factors togroundwater flow in the aquifer. The first isthat the permeable sediments of theGunnedah Formation that provide lateralgroundwater flow are contained in a muchless permeable basalt bedrock structure.This results in the aquifer becoming boththinner and narrower in the direction ofgroundwater flow. The result is a greatlyreduced capacity to transmit water and thus


iii

the pressure increases and the water levelrises. The second factor is the presence ofshoe-string aquifers of sand in the NarrabriFormation, laid down as the bed of a priorwater course. These provide partial hydraulicconnection with the Gunnedah Formation,and may provide a conduit for upwardmovement of pressured groundwater.

Recharge to the Gunnedah Formation is atthe southern upstream end of the aquifer,where extensive alluvial fans have beendeposited by the upland creeks. As thesecreeks flow and flood they cut through thesoil layers on the hills and allow water topercolate through the alluvial fans to theGunnedah Formation directly. There isprobably no diffuse recharge over the plainthat contributes to the Gunnedah Formation,but recharge below the root-zone of crops inthe Narrabri Formation will contribute to localwatertables pushed up by the highpotentiometric heads below. Since flooding ispresent in the catchment, it provides apathway for a large amount of water to exitthe catchment, bypassing both soil andgroundwater. The actual amount of rechargeto the Gunnedah Formation annually is onlyaround 5 mm/yr.

There are few signs of surface salinity in thesubcatchments of the Liverpool Plains. Onlytwo subcatchments are showing signs ofdryland salinity, Pine Ridge and Lake Goranand very salty creeks drain the areas, andgroundwater is as high as 35 dS/m.

The conceptualisation of the salinisationprocesses on the Liverpool Plains identifiedtwo subcatchments (Pine Ridge, Lake Goran)as being affected by salinity and therefore,this study focused on those twosubcatchments.

Surface and groundwateranalyses

Groundwater trends of the Pine Ridge andLake Goran catchments were fitted using theFLOWTUBE model. It is a simplegroundwater model based on Darcy’s Lawand water mass balance. Comparison ofavailable water and the capacity of theaquifer to transmit these volumes suggestthat, even prior to land clearing andconsequent recharge increases, someshallow water levels may have been typical

for the catchment. The native grass wasadapted to ephemeral water logging, and the swampy floodplain may haveexhibited some signs of salinity even withnative vegetation cover.

Modelling a no-change scenario, there aresteady but small increases in groundwaterheads mainly in the upper parts of thecatchments, and certainly only minor risesnear the outlet. This latter result is consistentwith data collected by Timms (1996, 1998).Recharge reduction to the GunnedahFormation is likely to be difficult given thelocation and size of the potential rechargebeds. No recharge reduction scenario was modelled.

Stream flow and salinity in the nearby BigJacks Creek subcatchment was analysed.Results indicate that stream flow is lowunless flooding occurs, when a substantialvolume of water can be mobilized. In theyears with average or less rainfall, salt isstored in the catchment in the NarrabriFormation, and when a flooding event occurs this salt is remobilised and appears in the stream.

Portability of conceptualmodel, tools and results

The alluvial Groundwater Flows System ofthe Pine Ridge and Lake Goran aquifers iscommon in much of the western rim of theGreat Dividing Range through New SouthWales and Victoria. Similarly, the aquiferconfiguration is found in many parts of thesetwo States. The management of recharge tothe aquifer is likely best solved withengineering controls of flood events, while onthe plains it is best to minimize water lossbeneath the root-zone of crops to controllocal water levels using deep-rootedperennials. The buffer zone (unsaturated)generated by deep-rooted perennials wouldprovide an effective store for floodwaters. Inother parts of the Basin water is pumpedfrom the relatively fresh lower aquifer andused as a water resource. Pumping in othersubcatchments of the Liverpool Plainscauses water levels to decline in anunsustainable manner, although salinitymanagement in this way could berecommended in the Pine Ridge and LakeGoran subcatchments. In those


iv

subcatchments, the groundwater of theGunnedah Formation is fresh and could beused or disposed of easily. The pumpingrates should be established as not to drawthe highly saline shallow watertables of theNarrabri Formation down into the GunnedahFormation. Pumping from the NarrabriFormation or the shoe-string sands presentsa serious disposal problem due to highsalinity. Coram et al. (2000) suggests thatregional scale alluvial aquifers present majorproblems to biological recharge control duemainly to scale issues (magnitude ofintervention needed). Pumping isrecommended for sufficiently fresh water,along with living with salt if saline areas aresmall. Modelling and observations in the PineRidge and Lake Goran subcatchmentssupport these recommendations.

From the results of other studies it is clear thatmanagement options are highly transferablebetween catchments of this type. Locally,groundwater and surface water salinity willdictate the mix of biological and pumpingmanagement of recharge and discharge areas.

Conclusions

• The Groundwater flow systems in thePine Ridge and Lake Goran catchmentsare similar to many along the westernrim of the Great Dividing Range, and arelikely to share similar managementstrategies for saline outbreaks.

• In the Pine Ridge and Lake Gorancatchments, recharge to theGunnedah Formation is quite differentfrom water loss below the root-zoneof crops to the Narrabri Formation,and must be managed separately.

• Localised recharge to the GunnedahFormation through the lower hillslopealluvial fans can be controlled byreafforestation and erosion control toprevent the perennial creeks fromdeeply incising the fans. On the plainsthemselves (Narrabri Formation), usingappropriate deep-rooted perennials,the deepest possible unsaturatedbuffer in the root-zone must bedeveloped and maintained so that thesize of flood events can be minimised(increased water storage), and saltdischarge to streams reduced.

Acknowledgements i

Executive summary ii

1. Introduction 1

2. Site description 3

3. Methods 53.1 Unique Mapping Areas 53.2 Conceptual understanding of the groundwater system 53.3 Recharge estimation 7

3.3.1 Localised recharge 73.3.2 Estimation of diffuse recharge 8

3.4 14C dating 83.5 Groundwater modelling 8

3.5.1 The FLOWTUBE model 83.5.2 Calibration 11

4. Results 154.1 Runoff-interflow 154.2 14C groundwater dating 154.3 Scenario modelling results 16

5. Discussion 185.1 Extent of land salinity 185.2 Salt loads to streams 185.3 Importance of localised recharge 195.4 Biological recharge reduction 205.5 Groundwater pumping 205.6 Overall methodology 20

6. Conclusions 23

References 24

Table of contents


v


vi


In Australia, dryland salinity causesdegradation to infrastructure, land andstreams estimated to be worth $350 millionannually. It also causes the loss of invaluablebiodiversity. European style agriculturalpractices brought upon by the early settlersin the 1800s profoundly altered Australia’scatchment water balances. This has resultedin the rising groundwater tables and saltremobilisation that lead to dryland andstream salinisation. If changes are notundertaken soon, the cost of salinity isestimated to increase at least five-fold in thecoming decades.

The changes required for managing salinityfall into three broad categories. Since drylandsalinity is caused by increased rechargeunder current agronomic practices, it is oftenassumed that biological recharge reductionwill reverse the situation. Biological rechargereduction includes the use of perennials suchas lucerne (alfalfa), agro-forestry or forestryplantation or incorporation of perennials incropping rotations. Where these may not beeffective on a reasonable time-scale,groundwater pumping or drainage may beused to protect important assets. In manycases, it is expected that neither of theformer two options will be appropriate andthat there could be a large increase in thearea of saline land. We need to adapt to thisor find opportunities for saline water usage.To implement changes, decisions will need tobe made about which of the above optionsare appropriate, the degree of interventionrequired the spatial targeting of these optionsand incentives for the adoption of theseoptions.

The Australian scientific community has thusbeen given an unprecedented challenge inproviding the technical information to supportthese decisions over large areas of Australia.There is unlikely to be the time or money toconduct large-scale field investigationseverywhere. It is therefore likely that suchdecisions will be made in a framework goingfrom scoping to design stages. Each designstage is conducted at more detail, usuallyrequiring more data and methods of greater

complexity and sophistication. However,because of the constraints on time andmoney, it is likely that at each stage, muchless area will be covered so that optionsbecome highly targeted. At the scopingstage, decisions need to be over large areaswith only sparse data. We need to be able topredict with sufficient confidence andaccuracy to distinguish the different optionsas to where they are clearly nonsensical;where they are clearly appropriate; andwhere further work is required (and hencemove to the next stage). In this paper, we concentrate on this broadscale scopingstage and tools to support this process.

At this stage, it is important to understandand communicate the salinisation processesthat are operating in a given catchment andpredict the future costs of salinity for a rangeof scenarios, including the status quo. Thedata will often need to be appropriate foreconomic analysis. The basis of anytechnical analysis is the conceptualisation ofthe salinisation processes. While increasedrecharge is the cause of dryland salinity,there is a large variation in the waygroundwater systems respond to increasedrecharge and hence how they will respond tomanagement changes. Prediction of impactsunder different scenarios implies that thisconceptualisation needs to be captured inprocess-based models. There are a numberof sophisticated groundwater and catchmentmodels that can predict catchment responseunder different scenarios e.g. MIKE-SHE,TOPOG, MODFLOW and AQUIFEM-N.However, the general lack of available datamakes it difficult to properly calibrate thesemodels. Under these circumstances, themodels can not provide the confidence inoutputs as would normally be expected forsuch sophisticated models. The morecomplex the model, the greater the time toinitiate and calibrate the model, the moretraining required to run the model and themore difficult to follow the information trailthat leads from data and assumptions to theconclusions. Therefore, at the scopingstages, there would be advantages in usingsimpler models more appropriate to the

1

1. Introduction


2

availability of data. Top-down modellingapproaches usually have the simplicity beingsought, but often are stochastic in natureand hence difficult to use in a predictivefashion. They can be used in a predictivefashion provided they capture the keyprocesses associated with dryland salinity orare calibrated on data that incorporateappropriate processes.

We illustrate a suite of top-down analysesthat provide the predictive capability. This isapplied to assist salinity management in animportant agricultural catchment in easternAustralia, relying on commonly available data.The study area is the Liverpool Plainscatchment (1,172,800 ha) in northern NewSouth Wales, Australia. It is a highlyproductive farming region, for which it waspredicted in the early 1990s that 195,000 hamay become salt affected in the next tenyears. Land use changes such as treeclearing, agricultural expansion and increasedinfrastructure have modified the catchment’shydrology over the last 30 years, leading insome areas to rising groundwater levels andconsequently soil salinisation.

Prior to this study, there was no clearunderstanding of the salinisation processes,their cause and extent in the Liverpool Plainscatchment. For this work, we did a first-cutwhole catchment analysis and then, basedon the outcomes of this first appraisal,focused the research down to the areas thatare extensively affected by salinity. For thesetwo salt affected subcatchments, wedeveloped a conceptualisation of thesalinisation processes; developed somesimple models that capture thisconceptualisation and illustrated how thesemodels were calibrated and applied fordifferent scenarios. From the scenariomodelling, we suggest the impacts ofdifferent options for this catchment withsome assessment of confidence. Somefieldwork is required as part of developing theconceptualisation, but does not involve muchtime or cost.


3

The study area is situated in the upper partsof the Liverpool Plains catchment, in easternAustralia (Figure 1). These plains encompassan area of 11,728 km2 and are limited to thesouth by the Liverpool Ranges (which formpart of the Great Dividing Range), to the eastby the Melville Ranges and to the west bythe Warrumbungle Range and Pilliga Scrub.Two rivers, Mooki River and Cox’s Creek,drain northwards into the Namoi River, whichis part of the Murray-Darling Basin.

The geological history of the catchment ischaracterised by a succession of erosionaland depositional episodes (Gates 1980,Timms 1996), resulting in a geologicallycomplex pre-Tertiary bedrock surface. ThePermian basalts and shale basement isoverlain by patches of Triassic conglomeratesand remnants of Jurassic basalts (Garawilla)and sandstones (Pilliga). During the Eocene,the northern Liverpool Plains were coveredby extensive basaltic flows of the LiverpoolShield volcano, which was heavily erodedduring the Pliocene, receding to its currentposition, the Liverpool Ranges.

These form the basement bedrock. LateTertiary and Quaternary thinning-upwardalluvial deposits filled the landscape’s palaeo-valleys, with thicknesses of up to 110 m. Thedeeper layer of alluvium, the GunnedahFormation, contains gravels and sands, whilethe overlying layer, the Narrabri Formation,contains mostly clays and silts. These twoformations are in partial hydraulic contact,the Narrabri Formation acting as a semi-confining layer. Poorly connected sand andgravel palaeo-channels described as ‘shoe-string aquifers’ (Broughton 1994a) can befound throughout the Narrabri Formation.Over the lower half of the salt affectedsubcatchments, the Narrabri groundwatersystem is saline with electrical conductivity(EC) values up to 35 dS/m, while theGunnedah is uniformly fresh (EC<2 dS/m).

The hill slopes of the upper parts of thosecatchments expressing salinity, Pine Ridge andLake Goran, are made out of basalts overlainby shallow colluvium and alluvium. On the flats,the alluvium uncomfortably overlays different

lithologies windowed out by the erosion of theTertiary basaltic cap. In these subcatchments,ephemeral creeks lose their channels on thechange of surface gradient in going from theupland areas to the Plains and form extensivealluvial fans. In the last 50 years (R Banks pers.comm.), run-off on the ranges has increasedand these creeks have deeply incised the soillayer on the hill-slopes, reactivating buriedpalaeo-channels. The groundwater flow in thealluvial aquifers is constricted at the outlet ofthese catchments by basement highsconsisting of highly cemented Triassicsandstone and conglomerate outcrops. Theconstrictions and the semi-confining NarrabriFormation choke the groundwater system,resulting in high potentiometric heads that arecurrently two metres below ground level in thelower parts of the catchment.

The underlying deeper basement aquifers(Tertiary basalts, Triassic conglomerates,Permian basalts and limestone) have saturatedhydraulic conductivity values comprisedbetween 10-4 m/d and 0.8 m/d with EC valuesranging from 1 to 2 dS/m. According to theUniversity of New South Wales (UNSW) waterquality studies (Broughton 1994a) and CSIROLand and Water (Andrew Herczeg pers.comm. June 1997), there is only minor mixingbetween the basement and alluvial aquifers.Bore records indicate deeply weathered frontson the basement, decreasing transmissivity(Gates 1980).

2. Site description

Brisbane

Condobolin

Orange

Sydney

Melbourne

0 100 200 km

Canberra

Cobar

Wagga

Wagga

Charleville

Hamilton

Mildura

V I C T O R I A

N E W

S O U T H

W A L E S

Q U E E N S L A N DMurray-Darling

Basin

AUSTRALIA

S O U T H

A US T R A L I ADarli

ngRive

r

Murray

River

Adelaide

Liverpool Plains

TamworthBroken Hill

Figure 1. Liverpool Plains catchment.


4

The rainfall decreases from over 1000 mm/yrat the top of the Liverpool Ranges (elevationup to 1000 m) in the south east to 550mm/yr on the flats near Pine Ridge (elevation:310 m). It is slightly summer dominated,when it is often short duration, high intensityrain or thunderstorms. Rainfall is extremelyvariable between years and seasons,resulting in drought, low river flows or floodconditions. The monthly maximumtemperature in the study area is highestbetween November and March (up to 43°C)and lowest during July and August (0°C).Temperature varies mostly with altitude andshows some large diurnal variation. Frostsare common, and occasional snowfalls areexperienced the ranges. Annual averagepotential evaporation of the area is 1900 mm, with a maximum of 275 mm inDecember and a minimum of 65 mm inJune. Gates (1980) reported a predominance of potential evaporation overrainfall, varying from 3.1/1 to 1.2/1 forsummer and winter respectively.

European settlement began in the 1830s andthe land was predominantly used for sheepand cattle grazing until the 1880s. Thencropping became an important land use onthe lighter textured red soils on the foot-slopes, resulting in extensive tree clearing. In the early 1950s, the heavy clays of the lowlying alluvial flats became the mainagricultural areas, replacing the deep rootednative plains grass (Stipa aristiglumis) with arange of annual crops (barley, sorghum,wheat). Cropping on the foot-slopes wasprogressively abandoned and replaced withpastures used for grazing. The steep ridgesof the ranges are covered by various speciesof eucalypts.

3.1 Unique Mapping Areas

The land surface has been divided into anumber of so-called Unique Mapping Areas(UMAs) based on slope classed derived from a digital elevation model, as shown inFigure 2. These areas represent geologicallyand geomorphologically consistentlandscape units used as a framework forresearch and management within thecatchment. Initially, eleven UMAs weredefined for the Liverpool Plains (Johnston etal. 1995). For the purpose of this study, theywere redefined according to their relevanceto catchment-scale water balance processesand some were consequently lumped,resulting in three units. UMA 1 comprises theLiverpool Ranges and Hills, encompassingthe non-alluvial component of the landsurface. These are the higher rainfall zonesand the soils are mainly shallow red-brownearths. Because of the low conductivity ofthe underlying basement groundwatersystems, any water not evaporated ortranspired is conceptualised as movinglaterally as surface run-off or subsurface flow.UMA 2 comprises the colluvial/alluvial rims,defined as those Quaternary alluvial systemswith slopes greater than one per cent. Theseare conceptualised as the recharge area tothe semi-confined Gunnedah Formation.UMA 3 is the component of the Tertiary andQuaternary alluvial system with slope lessthan one per cent. These are considered tobe the areas of diffuse recharge for theNarrabri Formation. The soil type consistsmainly of Black Earths. A summary of theircharacteristics is presented in Table 1.

3.2 Conceptual understandingof the groundwater system

The conceptual model for the groundwatersystem is detailed in Stauffacher et al. (1997),and follows an extensive review of availablereports on the Liverpool Plains hydrogeology(Bradd et al. 1994, Broughton 1994a, 1994b,1994c, Debashish et al. 1996, Gates 1980,Hamilton 1992, Tadros 1993) and analysis ofborehole data. The main features of themodel are shown in Figure 3. The lateTertiary-Quaternary unconsolidated alluvialdeposits overlying the bedrock have a hightransmissivity relative to the bedrock and aretherefore considered to be the majorcontributor to the salinity process.

3. Methods


5

Figure 2. UMAs in the Pine Ridge and Lake Goransubcatchments.

Range and Hills (UMA 1)Colluvial/alluvial rim (UMA 2)Alluvial Plains (UMA 3)

20 0 20 40 km

Gunnedah

Pine Ridgesubcatchment

Lake Goransubcatchment

TABLE 1. Unique Mapping Areas description.

UMA Description Hydrological significance

UMA 1 Ranges and hills, thin soils, low permeability High run-off generation areabasement aquifer, high rainfall

UMA 2 Colluvial/alluvial rims and fans on the lower Localised recharge to the Gunnedah aquiferslopes. Ephemeral creek beds deeply incising through the deeply incised creek bedsthe alluvium/colluvium.

UMA 3 Alluvial plains Diffuse recharge into the Narrabri Formation


6

The conceptual model consists of seven key points:

1. The alluvial groundwater system haspoor hydraulic connection to the othergroundwater systems of the LiverpoolPlains. Based on work by Broughton(1994a, 1994b, 1994c) who found thebedrock system to have very lowpermeabilities and on groundwaterdating undertaken for this study, it hasbeen assumed that the bedrockaquifers (mainly basalts in the saltaffected areas) do not transport muchwater compared to the unconsolidatedalluvial systems (gravels, sands). Thesubcatchment groundwaterboundaries are therefore defined bybedrock highs and outcrops.

2. The Liverpool Plains catchment canbe conceptualised as five almostindependent groundwater systems.The five subcatchments are separatedby bedrock highs as described above(Figure 4).

3. The most transmissive groundwatersystem is the semi-confinedGunnedah Formation (gravels andsands). This assumption is justified bythe hydraulic conductivity in theGunnedah Formation being up to1000 times higher and the specificyield being up to 100 times lower thanthe Narrabri Formation (clays and silts)(Broughton 1994a). Flow to theNarrabri occurs through verticalleakage from the pressurisedGunnedah aquifer.

4. Only two subcatchments are affectedby dryland salinity. There arecharacterised by bedrock highsconstricting the alluvial groundwaterflow at the outlets of thesubcatchments. This, together withlow topographic and potentiometricgradients and the lower hydraulicconductivity (higher clay content) atcatchment outlets, limit thegroundwater flow. These flowrestrictions in the Gunnedah aquifercause groundwater to dischargethrough the overlying NarrabriFormation to the soil surface. It is thisdischarge that results in landsalinisation.

5. Groundwater recharge to theGunnedah Formation consists of twocomponents: localised rechargewhere the Gunnedah Formationoutcrops in the colluvial/alluvial fanson the lower hill-slopes of theLiverpool Ranges; and downwardleakage from the Narrabri Formation.The latter can only occur in the upperparts of the salt affected catchments,where groundwater gradients aredownward. In the lower parts of thecatchments, vertical gradients reverseand there will be some discharge fromthe Gunnedah Formation to theNarrabri Formation.

6. Groundwater recharge to the Narrabrifrom the land surface can only occurin the upper parts of the salt affectedcatchments. In the lower parts of thecatchment, there is no storage in theaquifer for recharge to occur.Evidence of this is the higher salinitiesin shallow groundwater. Rainfall andfloods can recharge the aquifer in theshort-term but recharged water isevaporated or transpired afterwards.Floods on the land surface occur witha recurrence interval of about twoyears and probably dominate theephemeral recharge. Removal of thedeeper-rooted perennial vegetation onthe Plains may have led to moreshallow watertables and decreasedavailable storage in the NarrabriFormation. This may have contributedto the increased land salinity.

7. Much of the salt discharge from thecatchments occurs during the floods.Recharge from flood water causesdischarge of saline water.

The elements of the conceptual model forthe salinised subcatchments areschematically represented in Figure 3. Thisstudy will focus on the two catchmentsaffected by dryland salinity, the Pine Ridgeand Lake Goran catchments highlighted inFigure 4. These catchments arecharacterised by poor surface drainage andtheir groundwater outlets are laterally andvertically constricted by bedrock highs,leading to groundwater discharge andevaporative concentration of salt in the lowerreaches.


7

3.3 Recharge estimation

3.3.1 Localised recharge

Much of the localised recharge into theGunnedah Formation (Stauffacher et al.1997) is thought to occur in the transitionzone (UMA 2: alluvial fans on the lower hill-slopes) between the Ranges and Plains, bydirect run-off interflow from UMA 1 (Ranges& Hills, Figure 2, Table 1). Our approachinvolves two steps, the first to provideestimates of run-off interflow from UMA 1and the second to estimate, throughgroundwater modelling, the fraction of this‘lost’ water from the uplands that effectivelyrecharges the alluvial aquifer (GunnedahFormation) locally through the incised streambead on the colluvial-alluvial lower slopes.

Given high rainfall in the Ranges (over 1000mm/yr), the localised recharge may becomesignificant part of the overall catchment waterbalance. The major land use change thatcould impact on the upper catchment wateryield would be either clearance of forests orreafforestation. In the choice of approachthere were three considerations:

1. The methodology has to be based uponavailable data at a similar temporal (long-term groundwater processes, years) andspatial scale (catchment) as required forthis project

2. Since the key driving variable is rainfalland in the lack of other daily or monthlymeteorological information in the Ranges,the methodology should be based onmean annual rainfall

3. The methodology must be able to provideanswers for different levels of afforestationon the Ranges.

A relationship derived by Holmes and Sinclair(1986) is of the appropriate form. They foundthat, for 17 catchments within Victoria, therewere clear differences betweenevapotranspiration rates for forested andgrassland catchments. Zhang et al. (2001)demonstrated with data from over 250catchments worldwide that the Holmes-Sinclair relationship can be modified toprovide a robust relationship between long-term average evapotranspiration and rainfallfor a given forest cover (Figure 5). Hence,the modified Holmes-Sinclair relationshipprovides us with an appropriate predictor ofthe impacts on water yield of reafforestationor clearance of trees from the upland areas.

Zhang et al. (1997) used this relationship toestimate the amount of run-off interflow waterthat contributes to the water balance for thePine Ridge and Lake Goran catchments. Onefraction of the run-off interflow can contributeto localised recharge to the Gunnedah

Figure 3. Block diagram presenting the conceptualmodel for the hydrogeological processes on the saltaffected catchments on the Liverpool Plains: rechargeoccurs at the break of slope and due to the bedrockhighs at the outlet of the groundwater catchment,discharge to the surface occurs resulting in a saline area.

Runoff

Infilt

ratio

n

Basalt

Conglomerates

Geo

logi

cal c

onst

rictio

n

Salinearea

Evapotranspiration

Sand /gravel Clay

10 km

1km

Figure 4. Subcatchments of the Liverpool Plains. Thecatchments highlighted in grey are salt affected.

3.

Pine Ridge

Breeza5.Cox’sCreek 2.

LakeGoran

1.LowerMooki

4.PineRidge

Upper Mooki

Cox’s

Ck

Nam

oi R

Quirindi Ck

MookiR

0 20 40 km

Figure 5. Relationship between rainfall and runoffadapted from Holmes and Sinclair (1986) (—);obtained from the WAVES model (•).

500 1000 1500 2000

0

10

20

30

40

50

60

70

Rainfall (mm)

Runoff

(%)

A

Cleared catchment

Fully forested catchment

Figure 6. Bore numbers, locations and lithology in the Pine Ridge catchment.

Basalt

Basalt, shale, coal, conglomerate, sandstone

30 [km]

1[k

m]

16458 (AG)248110/3 (BB)248108/1(AG), 2(AG),3(BB)

40823/1(AC),2(AG),3(BS) 30060/2 (AG)CSIRO 3(AC),4(AC)30061(AG)40822/1(AC),2(AC),3(AG),4(BB)

Narrabri form

Gunnedah form.

AC: alluvial clay (Narrabri formation)AG: alluvial gravel (Gunnedah formation)BB: bedrock basaltBS: bedrock sandstone


8

Formation, where it outcrops in UMA 2. Asecond fraction of the latter exits thecatchment mainly in floods. A third fractionevaporates directly or from the shallowgroundwater. A fourth fraction recharges theNarrabri Formation (diffuse recharge).

3.3.2 Estimation of diffuse recharge

Steady state deep drainage estimates weremade on the Liverpool Plains by Kalma andGordon using a soil properties model (SaLF)(Shaw & Thorburn 1985). The calculationsindicate that 55% of the 92 sampled siteswould have recharge values less that 20mm/yr (Table 3). This is consistent with rainfall-fed recharge estimates from crop simulationmodels (Greiner 1997, Abbs & Littleboy 1998).Based on these results and as well as generalexperience on similar kind of soils and landuse, an estimate of 20 mm/yr seemsreasonable. For simplicity, we have chosen aconstant value in spite of spatial land use andsoils characteristics variability to provide someinsight into the relativity of diffuse to localisedrecharge. If the diffuse recharge component inthe overall recharge becomes an importantdriver for the groundwater modelling, we wouldthen refine the estimates.

3.4 14C dating14C dating (using a direct absorption methodand measurement made by liquid scintillationcounting) were undertaken to test the

conceptual model used as framework for thegroundwater modelling. The percentages ofmodern carbon are expected to give someinsight into the relativity of the groundwatertravel times in the bedrock and alluviumaquifers. Groundwater was sampled in thePine Ridge catchment at the top, middle andbottom of catchments, in three differentaquifers: the bedrock, the lower alluvium(Gunnedah Formation) and the upperalluvium (Narrabri Formation) (Figure 6).

Due to the scarcity of bores in the Lake Gorancatchment, no groundwater samples werecollected. Nevertheless, due to its geographicalproximity to the Pine Ridge catchment andtheir very similar geological, geomorphologicaland hydrogeological characteristics (Dyce andRichardson 1997), the results are expected tobe applicable to both catchments.

3.5 Groundwater modelling

3.5.1 The FLOWTUBE model

The FLOWTUBE groundwater model wasdeveloped for the Liverpool Plains work to:

• be as simple as possible

• require the fewest number of parameters

• require the least amount of input data

• incorporate all the key processes ofgroundwater movement in alluvialsystems.


9

Figure 7. Plan view and cross section of Big JacksCreek catchment, as discretised from available borelithology information presented in Dyce andRichardson (1997). Note that aquifer thickness doesnot indicate the depth below surface of the aquifer,only the average thickness of material.

Land SurfaceGroundwater

Aquifer Thickness

0 5km

0

30 m

Horizontal Scale

VerticalScale

Top of Catchment

Bottom of Catchment

Figure 8. Plan view and cross section ofYarramanbah Creek catchment, as discretised fromavailable bore lithology information presented in Dyceand Richardson (1997). Note that aquifer thicknessdoes not indicate the depth below surface of theaquifer, only the average thickness of material.

Top of Catchment


Aquifer Thickness

0 5 km

30 m

0

Horizontal Scale

ScaleVertical

Bottom of Catchment

Figure 9. Plan view and cross section of PumpStation Creek catchment, as discretised fromavailable bore lithology information presented in Dyceand Richardson (1997). Note that aquifer thicknessdoes not indicate the depth below surface of theaquifer, only the average thickness of material.

Top of Catchment

Land Surface

Groundwater

Aquifer Thickness

Bottom of Catchment

0 5km

Horizontal Scale

0

30 mVerticalScale

Figure 10. Plan view for three arms and main trunk ofthe Lake Goran catchment, as discretised from borelithology information presented in Dyce andRichardson (1997).

GoranWest

Central East

0 10 km

Horizontal Scale

Top Bottom

The basis of the method is a groundwaterbudget of the catchment. The catchment isfirst described as a segmented tube and agroundwater balance is calculated for eachcell. Inflows to each cell are vertical rechargeover the cell, and lateral movement of waterfrom higher up in the catchment. Outflowsfrom each cell are surface discharge out ofthe cell, and lateral movement ofgroundwater to parts lower down in thecatchment. The difference between inputsand output will cause a rising or lowering ofthe groundwater. All lateral fluxes arecalculated using Darcy’s Law, which can bewritten as:

q=k d i w (1)

where q is the flux (L3/T), K is hydraulicconductivity of the aquifer (L/T), d issaturated depth of flow (L), i is hydraulicgradient (L/L), and w is saturated width offlow (L).

Figures 7 to 9 show the segmentation forthe three subcatchments of the Pine Ridgecatchment, and Figures 10 and 11 show thelong and cross sections for the three armsand trunk section of the Lake Gorancatchment. The West arm feeds into thewestern end of the main trunk, the Centralarm feeds in at the next cross-section downgradient, and the East arm feeds in at thesecond last cross-section down gradient.


10

Figure 11. Cross sections for three arms and maintrunk of the Lake Goran catchment, as discretisedfrom bore lithology information presented in Dyce andRichardson (1997). Note that aquifer thickness doesnot indicate the depth below surface of the aquifer,only the average thickness of material.

EasternCentral

West Goran

Land Surface

Groundwater

Aquifer Thickness

0

50 mVertical

Scale

Examining the conceptual model ofStauffacher et al. (1997) summarised inSection 3.2, all the factors affectinggroundwater flow and rising watertables inthe Liverpool Plains are present in Darcy’sLaw. Firstly the material making up theaquifer grades from boulders, to gravel, tosand, to sand and clay beds. The order ofthese materials means that the hydraulicconductivity of the aquifer material isdecreasing from the hills-plains interface tothe catchment outlet, and the flow of watercan be expected to be restricted. Secondly,the slope of the land and the groundwatersurfaces decrease moving from the hills andranges to the catchment outlet, which againwill slow down any water movement. Finallythe saturated thickness and width of theaquifer, due to the intruded sandstone hillsand bedrock topography, is reduced towardthe catchment outlet further restricting flow.Application of Darcy’s Law should be usefulin various analyses of the water-balance andparameter estimation in the subcatchmentsof the Liverpool Plains. It forms the basis ofthe numerical model, and is used in theestimation of aquifer physical properties,such as hydraulic conductivity.

The FLOWTUBE groundwater model is asolution of 1D Darcy’s Law for saturated flow,with variable properties along a tube. Thereare special conditions however for theconceptual model we are using that affectthe equations. Reiterating, they are:

• The groundwater system consists ofthree layers (the semi-confiningNarrabri Formation the highlyconductive Gunnedah Formation, andsome bedrock material).

• Underlying the region of interest issome basement material which playsno role in groundwater movement orstorage (in effect this simply providesa lower limit to the extent of theaquifer).

• The middle layer is a highly conductiveaquifer, and the water in this layer isassumed to always be underpressure, i.e. the heads are above thetop of the aquifer, and so this layercontributes to water movement only.

• At the surface is a semi-confining layerwith low conductivity; this layercontributes to storage of water underpressure but not to any lateralmovement of water.

The mass in the tube at any time t is:

(2)

where V is volume of water per unit length(m3/m), � is porosity (m3/m3), A is saturatedcross-section area (equal to d*i; m2), t is thetime coordinate (days), and the subscripts 1and 2 refer to the conducting (GunnedahFormation) and confining (Narrabri Formation)layers respectively.

The flux within the tube at any point x asdescribed by Darcy’s Law is:

(3)

where q is flux (m3/d), K is hydraulicconductivity of the conducting layer (m/d), h is hydraulic head (m), and x is the spacecoordinate (m).

Mass balance demands that the rate ofchange of volume in the tube (�V/�t) is equalto the rate of change of water flux along thetube (�q/�x). Differentiating (2) with respect totime (and storage in the conducting layer isconstant), (3) with respect to distance (andno flux carried by the confining layer),equating them and dropping the time andspace ordinates for clarity, we get:

(4)

where R is the diffuse recharge per unitlength of tube (m3/d/m).

Rx

hKA

xt

A1

2

2 +∂∂−

∂∂=

∂∂ρ

x

h)x(K)x(A)x(q 1 ∂

∂−=

)t(A)t(A)t(V 2211 ρ+ρ=


11

Equation (3) can be expressed in finite-difference form between nodes i and i+1 as:

(5)

where �xi is the distance between node iand i+1 (m).

Equation (4) can be rearranged andexpressed as a fully-explicit finite-differencesolution at node i and time j as:

(6)

where �tj is the length of time step j (d).

The head at each node can be updated by:

(7)

where wi is width of the aquifer at node i (m).

The numerical solution of this problem isanalogous to a diffusion equation, which hasbeen extensively studied and is wellunderstood. According to Crank (1975), theforward-difference solution is stable if thefollowing condition is met:

(8)

where D is the diffusion coefficient (m2/d),which in our case is the product of hydraulicconductivity and aquifer width, divided by thespecific yield. Equation (8) can be rearrangedto give a �t for any desired aquifer properties.

1x

tD2

<∆

∆

i

j

i,2

1j

i,2j

i

1j

iw

AAhh

−+=

++

j

i,2

i,2

jj

i

i

j

i

j

1i1j

i,2 At

Rx

qqA +

ρ∆+

∆−

= −+

i

j

1i

j

i1i1i,1ii,1j

ix

hh

2

KAKAq

∆−⋅

+= +++

This model has been implemented with atree structure of tubes, reflecting the networkof streams, which form the alluvialgroundwater system. Each branch of the treefollows the above equations, while at thejunctions of the branch, continuity of headand fluxes is maintained. This only requirestrivial modifications to the conceptual andnumerical model.

In the discharge areas, the amount ofdischarge can be estimated in two ways. Inthe first, discharge occurs when the aquifercan no longer transmit all of the water. Theamount of discharge is equated fromdifferences in the amount of water carried bythe aquifer at different distances down theflowtube. In the second mode, a maximumdischarge rate is specified.

3.5.2 Calibration

Within the alluvial aquifers, there is a need toestimate the distribution of hydraulicconductivity, porosity, and groundwater headalong the flowtube. Table 4 shows how thebore data is distributed across the catchments,and includes the number of bores andfrequency of measurement. The simpleconclusion is that there is insufficientgroundwater head data to calibrate simulationswith transient inputs over any reasonable timeframe. For calibration purposes, only long-termsteady state simulations will be performed, byusing constant input conditions and runningthe model until calculated groundwater headsdo not change over a single year.

TABLE 4. Available bore, lithology, hydrograph, and physical data for dryland salinity affectedcatchments in the Liverpool Plains. The section for long-term Hydrographs indicates the number ofpiezometer nests and the length of time regular reading have been taken for, and the Pump Test sectionindicates the number of pumping tests performed along with the estimated hydraulic conductivity.

Yarramanbah Big Jacks Creek Pump Station Lake GoranCreek Creek catchment

Number of bores 70 122 53 674

Valid depth to water 60 90 30 431

Lithology & >5 readings 11 14 10 23

Long-term hydrograph 1, 8 years 1, 10 years 0 7, 24 years

Pump tests and results 0 1, 10-20 m/d 0 0

Hills/ranges area, km2 102 213 118 354

Contact area, km2 33 54 42 138

Plains area, km2 55 110 35 605


12

Hydraulic conductivity is the most criticalparameter and was estimated by calibration.A range of hydraulic conductivity waschosen to constrain the calibration. Freezeand Cherry (1979) suggests that for gravel,the conductivity is between ten and 1000m/d. Alluvial systems in eastern Australiasimilar to the Gunnedah Formation are in therange of ten to 100 m/d (WR Evans pers.comm. 1997). Salotti (1997) fitted bothhydraulic conductivity and porosity to anaquifer system in the adjacent BorambilCreek catchment, where there was 17 yearsworth of good quality bore hydrographs,with irrigation and pumping data for the lastten years. The results suggested a range ofconductivity from one to 30 m/d and arange of porosity of 0.05 to 0.3, for a singlelayer, combined sand and clay aquifer, ofsimilar dimensions to those in Pine Ridge.Given that the modelled aquifer containedgravel, sand, and clay, the cleaner aquifersystems modelled in the Liverpool Plainscatchments would have a higherconductivity than fitted by Salotti. The NewSouth Wales Department of Land and WaterConservation (NSW DLWC) keeps anextensive database of bore lithological logsand hydraulic properties, and havedeveloped a system where these propertiescan be estimated for any site where only alithology log is available. Using profilestypical of those used in the cross-sections to estimate aquifer size in Dyce andRichardson (1997), the NSW DLWCprocedure gave a conductivity range of fiveto 100 m/d, and a porosity range of 0.1 to0.3. There has been a single pumping testcarried out in the Pine Ridge catchment,near the outlet of the Yarramanbah Creeksubcatchment (W Timms, pers. comm.1997). This test indicated that the hydraulicconductivity within the most constricted partof the catchment was between ten and 20 m/d. The most consistent overlappingrange of conductivity values from all thedifferent data sources for the LiverpoolPlains sites is ten to 100 m/d over thecatchments; the data and selected range is shown in Figure 12.

Due to lack of transient data, it is difficult tocalibrate the porosity. Hence, the porositywas assigned a constant value over all thesubcatchments of 0.2. It is also notable thatthe various estimated porosity ranges allbracket the constant value chosen.

Since the physical shape and extent of theaquifer and confining layer, the physicalproperties of the aquifer and the currentwater levels were defined by borelogs (Dyce& Richardson 1997); and diffuse rechargewas held as constant at 20 mm/yr, the onlyparameters that require fitting are thehydraulic conductivity of the aquifer and theproportion of run-off interflow that becomeslocalised recharge to the aquifer. Theseparameters are co-dependent and thecombination must fit within all the constraintsimplicit in the observed data. The proportionof run-off that becomes localised rechargeconstant is constrained to be the same foreach of the three subcatchments of the PineRidge catchment, as we would expect theerosion history of each of these to be similar,and therefore the processes and rates offlows to be similar. While this condition doesnot necessarily apply to the Lake Gorancatchment, we assumed it did.

The FLOWTUBE model is a combination ofDarcy’s Law and a mass balance equation. It is possible to establish a relationshipbetween the amount of localised rechargeand the hydraulic conductivity at thehills/plains interface, thus:

(9)

where RO is the average rate of run-off fromthe hills and ranges (m3/d), and f is theproportion of that run-off that becomeslocalised recharge. In calibrating theFLOWTUBE model in steady state mode, theratio of conductivity and f is actually fitted. Atthis stage, we fix f and then calibrate the

wid

RO

f

K =

Figure 12. The ranges of hydraulic conductivityestimated from various sources, along with theadopted range for the study. F&C is Freeze andCherry (1979), EP are the expert partners in the work,SBC is the Salotti (1997) study in Borambil Creek,and DLWC are estimates from NSW Department ofLand and Water Conservation database.

DLWCSBCEPF&C

0.1

1

10

100

1000

Hydra

ulic

conductivity

(m/d

)

Ad

op

ted

ran

ge


13

conductivity. In doing so, the amount oflocalised recharge must satisfy severalconstraints. First it should be an amount, thatwhen taken on an annual average basis,does balance reasonable recharge rates inthe alluvial fans of the catchments anddischarge rates in the lower parts of thecatchment and produce a conductivityestimate from (9) in the recharge area at thetop of the catchment that is consistent withthe a priori range determined in Figure 12,i.e. between ten and 100 m/d. On the basisof these limiting factors, we propose that fiveper cent of the run-off interflow from the hillsbecomes localised recharge in eachsubcatchment.

Using the area of alluvial fans in Table 4 andthe volumes of run-off interflow water fromTable 5, an estimate of the annual averagerecharge rate in the alluvial fans can bemade. In the subcatchments of the PineRidge catchment, the average rate variesbetween 0.06 and 0.07 mm/d, and in theLake Goran catchment the average rate is0.05 mm/d. Since the average rate for eachof the three subcatchments in Pine Ridge arealmost the same, it gives credence to theassumption that these subcatchmentsevolved together, and have similarhydrogeological behaviour. The fact that therate is also very similar in the Lake Gorancatchment, with different annual rainfallamounts, adds weight to the assumption of afixed proportion of run-off interflow waterbecoming localised recharge.

Steady state calibration

With a fixed aquifer geometry and any givenamount of localised recharge, we cancalculate what the conductivity will be withany distribution of hydraulic heads usingEquation (9). An inferred conductivitydistribution that is relatively constant, or thatis monotonically increasing or decreasingalong the tube, will provide furtherconfidence in the physical structure andconceptual model. If the distribution israndom or chaotic however, then withoutsignificant geological changes along the tubethis would indicate a poorly constructedmodel. Additional to calculating conductivitywithin the flow tube, using the results of thepump test at the catchment outlet from Table 4, we can infer the hydraulic gradient

at the outlet and fix this as a boundarycondition for the model.

Big Jacks Creek, a small area five kilometresfrom the outlet, showed high conductivityestimates outside the prescribed range.It is a simple conclusion that this areais likely to suffer from rising water levels (sothat the steady state assumptions may nolonger be valid) or is part of a discharge area.In the model runs, this area was set to havethe maximum hydraulic conductivity (100 m/d)within the prescribed range, and showedthe greatest rates of rise with transientsimulations. Yarramanbah Creek had wellbehaved conductivity estimates throughoutexcept right at the outlet, where salinity isexpected to be expressed. Pump StationCreek required a high conductivity across theentire catchment due to its very thin lowersection. This also inferred a high hydraulicgradient at the outlet of five per cent. A similar large drop in water level wasreported by Timms (1998) who measuredgroundwater depth along transects passingthrough the outlet of the Pine Ridgecatchment, so this model inference can beaccepted as real. In the Lake Gorancatchment, conductivity estimates wereuniformly low, near the bottom of the range,except in the main trunk. Here a significantarea required much higher conductivity totransmit the input water. Since this areacontains the lake this was not seen asunusual, but conductivity was lowered to be consistent with the other fitted values;water levels here were modelled at theground surface.

Conductivity estimates were rounded tomultiples of five and some manualmodifications performed to allow for thisrounding. Figures 13 to 16 show the fittedsteady-state hydraulic heads for Big JacksCreek, Yarramanbah Creek, Pump StationCreek and Lake Goran, respectively. The fitsare excellent, as would be expected fromcalculating conductivity based on the headsto be fitted. All the catchments showedconductivity distributions that were wellbehaved, and thus we can be moreconfident in the structure of the aquifers andthe conceptual model.


14

Figure 13. Measured and fitted steady stategroundwater heads in the Gunnedah Formation forBig Jacks Creek catchment. The root mean squareerror (RMSE) between the observed and calculatedheads is 0.69 m.

0 5 10 15 20 25

Distance (km)

GroundwaterSurface Elevation

Fitted Heads

310

320

330

340

350

360

Ele

va

tio

n(m

AH

D)

Figure 14. Measured and fitted steady stategroundwater heads in the Gunnedah Formation forYarramanbah Creek catchment. The RMSE betweenthe observed and calculated heads is 0.70 m.

300

310

320

330

340

350

360

0 3 6 9 12 15 18

Distance(km)


Fitted Heads

Ele

vation

(mA

HD

)

Figure 15. Measured and fitted steady stategroundwater heads in the Gunnedah Formation forPump Station Creek catchment. The RMSE betweenthe observed and calculated heads is 0.85 m.

310

320

330

340

350

360

0.0 2.5 5.0 7.5 10.0 12.5

Distance (km)


Fitted Heads

Ele

va

tio

n(m

AH

D)

Figure 16. Measured and fitted steady stategroundwater heads in the Gunnedah Formation forthe Lake Goran catchment. The RMSE between theobserved and calculated heads for each of the threearms and trunk varied between 1.17 and 2.71 m, andis 1.78 m overall.

280

300

320

340

360

380

0 15 30 45 60 75 90

Distance (km)


Fitted Heads

Ele

vation

(mA

HD

)

Lake Goran

WesternArm

CentralArm

EasternArm


15

4. Results

4.1 Run-off interflow

The run-off interflow for the Liverpool Rangesand hill slopes was calculated using themethod described in Section 3.1. Table 5lists the run-off interflow and diffuse rechargevalues from Zhang et al. (1997).

In the Big Jacks Creek subcatchment, agauging station was installed and has goodrecords for 1996 to 1998. The data from thisthree year period indicates that betweenthree per cent and five per cent of the rainfalling in the hills and ranges leaves thecatchment during floods.

4.2 14C groundwater dating

The 14C dating results are summarised inTable 6.

In the alluvium, the trend is an ageing of thewater down the flow-path. On the top of thecatchment (dark shade), the water contains96% modern carbon, in the middle (lightshade), 50% and at the outlet 18.5% (noshade). For the bedrock, the ageing trend isthe same, but due to a much lowerpermeability, the percentage of moderncarbon is respectively 55, 11 and less thantwo from the top to the bottom of thecatchment. At the outlet of the catchments,the deepest bores in the alluvium (just abovethe bedrock) also contain less than two percent modern carbon, probably due to thevery low flux of groundwater in the deepestalluvium layers. The alluvium aquifers areclearly the more dynamic systems on theLiverpool Plains and have therefore thelargest water carrying capacity.

TABLE 5. Annual water inputs for dryland salinity affected catchments in the Liverpool Plains,estimated by Zhang et al. (1997). Run-off interflow is the volume of water coming from the hillsand ranges above each catchment annually that may become Localised Recharge, and the DiffuseRecharge is a volume of water over the whole plains based on an annual leakage rate below theroot-zone of 20 mm/yr. The later in considered as an upper boundary.

Catchment Run-off interflow 106m3/yr Diffuse recharge 106m3/yr

Big Jacks 27.8 1.1

Yarramanbah 15.5 0.6

Pump Station 21.4 0.35

Lake Goran 29.9 6

TABLE 6. 14C dating results.

Sample Lithology �13C 14C ‰ PDB pMC±1�

16458 Alluvium: gravel, sand -11.7 96.5 ± 1.5

248108/3 Bedrock: basalt -13.7 55.3 ± 1.9

40823/2 Alluvium: gravel/sand -10.4 49.9 ± 1.1

40823/3 Bedrock: sandstone -11.6 11.7 ± 0.9

30061 Alluvium: gravel/sand -13.7 BG < 2PMC

40822/3 Alluvium: gravel/sand -14.1 BG < 2PMC

40822/4 Bedrock: basalt -10.6 BG < 2PMC

30060/2 Alluvium: gravel, sand -12.2 18.5 ± 1.0


16

4.3 Scenario modelling results

Scenarios to be considered are historicalconditions, and future conditions undercurrent and poor management practices.Problems with estimating prior conditions inthese and other catchments are a generallack of historical water levels, difficultyestimating the spatial distributions ofconductivity and porosity for groundwatermodels, and estimating temporal waterinputs. Of these, the latter is often the mostdifficult, and particularly in the LiverpoolPlains. The current modified conceptualmodel using a fixed proportion of run-offinterflow water as recharge to the GunnedahFormation may not apply uniformly from yearto year, and may also depend on the erosionhistory of the catchments. It is likely thatinfiltration to the Gunnedah Formation islimited by available storage in the upper partsof the catchment, so size and timing ofindividual events will be important to theactual amount.

Zhang et al. (1997) have estimated theaverage annual run-off interflow volume forthe catchments of the Liverpool Plains, asreported in Table 5, but how these valuesare modified to become Localised Rechargeby historical circumstances is unknown. Forthe purposes of estimating a pre-Europeansettlement situation, we can use the run-offinterflow values from Zhang et al. for fullyforested hills and uplands and assume thatthe same proportion as today infiltrates. Thisis run until a steady state is reached thenused as the starting condition for scenariomodelling. From 1950 to 2000 the run-offinterflow for the current amount of clearing isused, then from 2000 to 2050 the run-offinterflow value for fully cleared land is used.Figures 17 to 22 show the historical initialgroundwater level and modelled water levelrise for Big Jacks Creek, YarramanbahCreek, Pump Station Creek, and Lake Gorancatchments. It is these simulations that willtest the assumed value of porosity. If waterlevels rise much faster or slower thanmeasured then the value used must bewrong, as well as the assumption of auniform distribution. Bore hydrographsreported by Timms (1998) indicate waterlevel rises of between one and fivecentimetres per year near the catchmentoutlets, which are consistent with the trendspresented in Figures 17 to 22.

Figure 17. Transient water level simulation forestimated historical and future conditions in the BigJacks Creek catchment.

300

310

320

330

340

350

0 5 10 15 20 25

Distance (km)


Year 2050Year 2000Year 1950

Discharge2000 to 20501.2 -> 3.2 mm

Ele

vation

(mA

HD

)Figure 18. Transient water level simulation forestimated historical and future conditions in theYarramanbah Creek catchment.

300

310

320

330

340

350

0 3 6 9 12 15 18Distance (km)

Discharge2000 to 20501.1 -> 1.4 mm

Ele

vation

(mA

HD

)



Figure 19. Transient water level simulation forestimated historical and future in the Pump StationCreek catchment

300

310

320

330

340

350

0 2 4 6 8 10 12



Ele

vation

(mA

HD

)

Distance (km)

Discharge

2000 to 205012 -> 53 mm

Figure 20. Transient water level simulation forestimated historical and future conditions in theEastern valley within the Lake Goran catchment.

275

300

325

350

375

0 5 10 15 20 25 30 35

Distance (km)

Ele

vation

(mA

HD

)




17

Figure 21. Transient water level simulation forestimated historical and future conditions in theWestern valley within the Lake Goran catchment.

275

300

325

350

375

0 5 10 15 20 25 30 35Distance (km)

Ele

va

tio

n(m

AH

D)



Figure 22. Transient water level simulation forestimated historical and future conditions in the LakeGoran basin.

275

300

325

350

375

0 5 10 15 20 25 30 35Distance (km)

Discharge2000 to 20504.4 -> 6.6 mm

Ele

va

tio

n(m

AH

D)



5. Discussion

5.1 Extent of land salinity

We estimate that over the next 20 years, thearea at risk of salinity in the two salt affectedcatchments will show only a minor increase.The area at risk corresponds roughly tothose areas with shallow saline groundwaterin the lowest parts of the subcatchments,close to the outlets. The highly salineshallow watertables themselves areevidences that these areas have beenhistorically saline and would have expressedsome salinity in the past.

5.2 Salt loads to streams

Rainfall, daily stream flow and EC data fromBig Jacks Creek for 1996 to 1998 have beenanalysed. The raw data of stream flow andstream electrical conductivity are shown inFigures 23a and 23b. A clear observationfrom Figure 23a is that for any significantflows, above say 500 ML/d, the stream ECis approximately constant. This stronglyindicates a system that is not limited by saltavailability, but rather the capacity to mobiliseand shift the salt store. Figure 23b reinforces


18

this observation, with EC varying within aclear band during the major event in 1998.

Figure 23a also shows a power curve fittedto the flow-EC data. This appears tooverestimate EC under the highest flows,which would have a corresponding effect oninfilling EC from flow data alone. Figures 24aand 24b show the observed and predictedsaltload (flow multiplied by EC and convertedto tonnes) on a daily and annual basis usingthe curve fitted in Figure 23a. Theobservation of over-prediction is borne outwith the best-fit linear regression passingthrough the origin having a slope greater thanone in both Figures 24a and 24b. It is clear,however, that several points dominate thedaily saltload graph where the scatter isgreatest. The annual saltload appears tohave enough compensating higher and lower flows such that the observed andpredicted results show a much strongerlinear relationship.

Figures 23a and 23b. Daily flow versus daily meanEC, and the time series of flow and EC for Big JacksCreek, for the three year period 1996 to 1998.

0

500

1000

1500

2000

2500

Jan-96 Jul-96 Jan-97 Jul-97 Jan-98 Jul-98 Jan-99

Date

Mean Daily Flow (ML/d)

Mean EC (uS/cm)

Flo

w(M

L/d

)o

rE

C(

S/c

m)

�

y = 1175.2x-0.161

R2

= 0.53

0

500

1000

1500

2000

0 500 1000 1500 2000 2500

Daily Flow (ML/day)

Mean

EC

(S

/cm

)�

Figures 24a and 24b. Observed versus predicted (a)daily and (b) annual saltload from Big Jacks Creek,based on a simple power regression.

0

2000

4000

6000

8000

10000

0 2000 4000 6000 8000 10000

Observed annual salt load

Pre

dic

ted

annualsalt

load

y = 1.105xR =0.99

2

0

100

200

300

400

500

0 100 200 300 400 500

Observed daily salt load

Pre

dic

ted

da

ilysa

ltlo

ad y = 1.217x

R =0.832


19

To compensate the annual saltload figures, a second regression for flow-EC wasconceived where the multiplier in theequation was reduced by ten per cent, thisbeing the approximate annual overestimatebased on Figure 24b. The effect of thischange is shown by the dotted line in Figure 23a. The changed daily and annualsaltload figures are shown in Figures 25a and 25b. On a daily basis the best-fit slopethrough the origin is now closer to one, andthe annual load fits a 1:1 line as required.

Moving on to annual salt load versus rainfallinput, Table 7 includes all the relevant figures for the three years 1996 to 1998. The annual saltfall is based on a catchmentarea of 380 km2 and an average rainfall saltconcentration of 3.3 mg/L as measured at Gunnedah.

There are a number of inferences that can bedrawn from the figures in Table 7. The first isthat the catchment does not yield a largeamount of stream flow, except under floodconditions, when there is a much larger butshort-lived response. Another interpretation isthat during average and below-averagerainfall years, the catchment will store saltthat is not removed by surface water flows.Then during large flood events the nearsurface salt store is mobilised and a greatamount of salt is flushed from the system.

The average ratio of salt output to salt input(SO:SI) for the three years shown is 2.32,while total salt output divided by salt input is2.77. From Jolly et al. (1997), the averageSO:SI ratio is 1.03 for Cox’s Creek atBoggabri which is a stream to the west(4040 km2). The Mooki River is an adjacentcatchment to the east, and has averageSO:SI of 3.26 at Caroona (area 2540 km2),and 2.75 at Breeza (area 3630 km2). Thelatter values are similar to those found for BigJacks Creek, and appear in the upperreaches of the Mooki River system. Furtherdownstream in the Mooki there isgroundwater pumping for irrigation and fallingwater levels, which may artificially alter anysimple salt and water balance analysis.

5.3 Importance of localisedrecharge

Unlike many of the other reported areas ofAustralia, the recharge to the main aquiferrelevant to salinity (Gunnedah Formation) hasnot increased by one or two orders ofmagnitude, but more likely about 1.5 to twotimes. This is because of the importance oflocalised recharge in the colluvial fan areasboth prior to and after changing from nativevegetation to agriculture. This localisedrecharge results from the lateral flow both

Figures 25a and 25b. Observed versus predicted (a)daily and (b) annual saltload from Big Jacks Creek,based on a modified power regression.

0

2000

4000

6000

8000

10000

0 2000 4000 6000 8000 10000

Observed annual salt load

Pre

dic

ted

annualsalt

load

y = 0.995xR =0.99

2

0

100

200

300

400

500

0 100 200 300 400 500

Observed daily salt load

Pre

dic

ted

da

ilysa

ltlo

ad y = 1.095x

R =0.832

TABLE 7. Annual salt and water balance components for Big Jacks Creek.

Year Rainfall Saltfall Streamflow Observed Modelled(mm) (t) (mm) Saltload (t) Saltload (t)

1996 879 1102 15 1478 1618

1997 704 883 3 435 445

1998 1225 1536 77 7863 7794


20

from run-off and subsurface flow from thehigher rainfall upland areas. We estimate thatthe lateral flow may have increased by about40% following clearing of trees for cropping.It is however more difficult to estimate howmuch this means as an increase in effectivelocalised recharge as it is not clear whatlimits it. However, the increase in thiscomponent is unlikely to be greaterproportionately.

The flooding may be expected to be a majorcontributor to recharge to the less permeableNarrabri Formation, but over a large fractionof this, there is simply no storage in theaquifer to accept recharge. Rather, it isexpected that much of this recharge is lostby evapotranspiration afterwards. The nativevegetation on the plains was not trees, butso-called Plains Grass (Stipa aristiglumis),adapted to the swampy floodplain areas thatcharacterised the region prior to clearing. Thearea of shallow watertables is likely to haveincreased (means less or no buffer to storeflood water) and this situation coupled withthe deforestation of the upper catchmentarea has probably led to more flooding in thecatchments. For the upper part of thecatchment where watertables are deeper, thegradients are also greater and flooding maybe expected to dominate the recharge.

We estimate that the localised recharge tothe Gunnedah Formation is comparable tothe combined net long-term diffuse rechargeand flood recharge to the Narrabri. It is likelythat the diffuse recharge has increased in thearea of shallow watertables and there issome ability through agronomic practices tolimit this.

5.4 Biological rechargereduction

The two key issues associated with usingbiological recharge reduction as a means ofcontrolling salinity are the degree ofintervention required and the time responseto recharge control. As above, the minimumrecharge rate that is likely is about ten percent of the current recharge. This could onlybe achieved with a major land use change inthe catchment. Perhaps 70% of thecatchment would need to be done toachieve ten per cent change in saline areasand a ten per cent decrease in stream

salinity. Clearly, this could only be effective ifthese land uses were nearly profitable in theirown right. Replanting of the colluvial fans andchanged land use on the lower parts of thesemay achieve a partial recharge reduction.Again 70% of the catchment may reduce theamount of saline area to ten per cent.

5.5 Groundwater pumping

Given the difficulties of biological rechargecontrol, engineering options may be feasibleto control salinity in the shorter-term. Themost likely of these is groundwater pumpingfrom the Gunnedah Formation. Thegroundwater model used here is not suitablefor estimating drawdown cones, but it doessuggest the recharge volume that can beharvested. While the transmissivity is not ashigh within the salt affected catchments as inthe other subcatchments where groundwaterpumping already occurs, it is still reasonable.One of the difficulties that can occur is thedrawdown of saline water from the NarrabriFormation. Despite the low permeability ofthe Narrabri Formation, it is feasible to pumpfrom the ‘shoe-string aquifers’ but thegroundwater will generally be saline anddisposal will be an issue.

It is possible that biological recharge optionsand engineering options may not beeconomically feasible. In any case, there islikely to be a large area of saline land that isbetter with some vegetative cover. Becauseof the saline groundwater, vegetation in theshallow watertable areas is unlikely to usegroundwater to any great extent.

5.6 Overall methodology

The methodology consisted of a number ofimportant steps:

1. Disaggregation of the study area into‘unique mapping areas’

2. Development of the conceptual model

3. Conduct of a salt balance study of thecatchment

4. Estimation of flows and recharge

5. Use of a simple groundwater model

6. Testing of assumptions usinggeochemistry

7. Investigation of the role of flooding.

All of the steps were important in testing theoptions. The development and the testing ofthe conceptual model for the salinisationprocesses on the Liverpool Plains catchmentwas an iterative process. The first step wasto identify the areas affected by salinity, thenidentify the main processes involved andfinally build up a framework for the numericalmodelling. The resulting conceptual modelproved robust and a good platform for thenumerical analysis. The groundwater datinggave some insight into the relativity of thetiming of the groundwater processes in thedifferent aquifers. It allowed us to eliminatethe bedrock as a major player in theobserved salinity processes and we couldtherefore confidently focus the work on theprocesses in the alluvium.

Since the run-off interflow estimates basedon the Holmes and Sinclair (1986)relationship is based on and constrained byobservations, we expect the relationship tobe both robust and scientifically justified. Themodel has advantages over more traditionalprocess-based models, requiring datagenerally available at regional scales andbeing very easy to apply either to anindividual catchment or in a spatial modellingframework. The model developed isconsistent with previous theoretical work andshowed good agreement with over 250catchment-scale measurements from aroundthe world (Zhang et al. 2001). It is a practicaltool that can be readily used to predict thelong-term consequences of reforestation,and has potential uses in many catchment-scale vegetation management studies. Forthe Liverpool Plains study, it provided us withreliable estimates of run-off interflow.Coupled to the results of the groundwatermodelling and the data from the gaugingstation, it gave some insight in the broadprocesses controlling waterlogging andsalinisation and emphasised the importanceof floods.

One of the most critical steps was theestimation of the hydraulic conductivity. Thisis highly correlated with the fraction of therun-off interflow that recharged. Severalconstraints needed to placed on thecalibration process. If the hydraulicconductivity was towards the highest rangeof conductivities for gravels, much of thefloodwater needed to become recharge and

discharge rates needed to be very high in thedischarge area. If the conductivity wastowards the lowest end of the range forsands, then the only very little run-offinterflow recharged the groundwater systemand diffuse recharge would have beenrelatively more important and discharge rateswould be too low to cause any problemswith salinity.

We have reasonably asserted that aproportion of five per cent enters thegroundwater system, and measured aboutfive per cent leaving during flood events, sowhere does the other 90% go? The onlypossible sinks are the basement materialunderlying the Gunnedah Formation, and thestorage and evaporation of flood water as itsits on the surface of the plains. While waterdoes appear in the bedrock, 14C resultsindicate that this is very slow moving and isprobably not a significant sink in the system.

This leaves storage on the plains as the onlycandidate to close the run-off interflowbudget. The weather systems in theLiverpool Plains consist of steady frontal rainin the winter months, and more violentconvective thunderstorms in the summermonths. Zhang et al. (1997) and Ringrose-Voase (pers. comm. 1999) have providedestimates of the amount of available storagein the cracking clays of the Liverpool Plains.In the top 2 m of soil up to 400 mm ofstorage is available. Anecdotal evidencesuggests that the clay plains initially absorblarge run-off events, but if followed closely bya second large event, flooding occurs on thesaturated plains. If the heavy summer rainscause most flooding events, then it ispossible to evaporate a large amount ofsurface water when the potential evaporationrates are highest to empty this store readyfor the next event. The depth to which plantscan empty the soil store will thereforecontribute to more or less flood events, so the practice of removing deep-rootedperennials for shallow-rooted crops may be a factor contributing to increased flooding.

Thus far the Gunnedah Formation has beenmodelled as receiving no recharge throughthe Narrabri Formation due to its high claycontent and thickness. Poor managementcan result in a local impact given that currentagricultural systems are leaking water. Under


21


22

this regime, no management practice on theplains itself can affect the water levels in theGunnedah Formation, although these levelsare currently rising. The productivity of theplains must be protected by managing croprotations to minimise water leaving the root-zone and prevent local salinity, while theGunnedah Formation water levels need adifferent approach.

The modelling done in this work has notconsidered engineering options, such asgroundwater pumping. If the hydraulic headsin the Gunnedah Formation could be loweredwith suitable pumping regimes, then the localwatertable would slowly move downwards ifsurface recharge was controlled. Withsuitable vegetation management, it may bepossible to allow periodic recharge to leachthe salt from the new de-watered zone. Thiswould create additional potential root-zonefor crops and grasses, along with a storagebuffer when large episodic events causerecharge that cannot be controlled byvegetation alone. Careful economic analyseswould be required to compare the increasedreturn from cropping enterprises against thecost of pumping and disposal ofgroundwater. A secondary effect would bethe possible contamination of the relativelyfresh water in the Gunnedah Formation withsaline groundwater currently found nearer tothe surface, due to a reversal in headgradient down from the Narrabri to theGunnedah Formation.


23

6. Conclusions

• Importance of flood management: Floodmanagement is likely to be the singlemost important issue in recharge andsalinity control. The native vegetation onthe Plains was not trees, but so-calledPlains Grass (Stipa aristiglumis), adaptedto the swampy floodplain areas thatcharacterised the region prior to clearing.The area of shallow watertables is likely tohave increased (means less or no bufferto store flood water) and this situationcoupled with the deforestation of theupper catchment area has probably led tomore flooding in the catchments. For theupper part of the catchment wherewatertables are deeper, the gradients arealso greater and flooding may beexpected to dominate the recharge.

• Model complexity: Simple models aregood enough to get the processesunderstanding and the relativities right,some more detailed work is needed formanagement options implementation(groundwater pumping, localised rechargecontrol, cropping rotations on the flats).

References

Abbs, K & Littleboy, M 1998, Recharge estimation for the Liverpool Plains, Australian Journal of Soil Research, 36(2), 335-357.

Bradd, JM, Waite, D & Turner, J 1994, Determination of Recharge/Discharge Areas and Water/Salt Distribution inAquifers of the Liverpool Plains, Department of Water Engineering, University of New South Wales, Sydney.

Broughton, A 1994a, Mooki River Catchment Hydrogeological Investigation and Dryland Salinity Studies, Vols 1 & 2.Report TS 94.026, Department of Land & Water Conservation, Sydney.

Broughton, A 1994b, Coxs Creek Catchment Hydrogeological Investigation and Dryland Salinity Studies, Vols 1 & 2.Report TS 94.082, Department of Land & Water Conservation, Sydney.

Broughton, A 1994c, Liverpool Plains Catchment Hydrogeological Map (1:250 000), Department of Land & WaterConservation, Sydney.

Coram, JE, Dyson, PR, Houlder, PA & Evans, WR 2000, Australian Groundwater Flow Systems contributing toDryland Salinity, Bureau of Rural Sciences for the National Land and Water Resources Audit, Canberra.

Crank, J 1975, The Mathematics of Diffusion, 2nd edn, Oxford University Press, New York.

Debashish, P, Demetriou, C & Punthakey, JF 1996, Gunnedah Groundwater Model in the Upper Namoi Valley, NSW.Report TS 96.019, Department of Land & Water Conservation, Sydney.

Dyce, P & Richardson DP 1997, Characterisation of subcatchment aquifers in the Liverpool Plains for the purposeof groundwater modelling. Technical Report 16/97, CSIRO Land and Water, Canberra.

Freeze, RA & Cherry JA 1979, Groundwater, Prentice-Hall, New Jersey.

Gates, GWB 1980, The Hydrogeology of the Unconsolidated Sediments in the Mooki River Valley, New SouthWales. M.Sc. Thesis, University of New South Wales, Sydney.

Greiner, R 1997, Optimal farm management responses to emerging soil salinisation in a dryland catchment inEastern Australia. Journal of Land Degradation and Development, 8, 71-93.

Hamilton, S 1992, Lake Goran Catchment Groundwater Study. Report TS 92.009, Department of Land & WaterConservation, Sydney.

Holmes, JW & Sinclair, JA 1986, Water Yield from some afforested catchments in Victoria. Hydrology and WaterResources Symposium, Griffith University, Brisbane, 25-27 November 1986,National Conference Publication86/13, Institution of Engineers, Australia, Canberra.

Johnston, R, Abbs, K, Banks, R, Donaldson, S & Greiner, R 1995, Integrating biophysical and economic models forthe Liverpool Plains using unique mapping area, in P Binning, H Bridman, H and B Williams(eds.)International Congress on Modelling and Simulation, vol. 1, pp. 150-154.

Jolly, ID, Dowling, T.I, Zhang, L, Williamson, DR & Walker, GR 1997, Water and salt balances of the catchments ofthe Murray-Darling Basin. Technical Report 37/97, CSIRO Land and Water, Adelaide.

Salotti, D 1997, Borambil Creek groundwater model. Report CNR97.014, Technical Services Division, NSWDepartment of Land and Water Conservation.

Shaw, RJ & Thorburn, PJ 1985, Prediction of Leaching Fraction from Soil Properties, Irrigation water and Rainfall.Irrigation Science, 6, 73-83.

Stauffacher, M, Walker, GR & Evans, WR 1997, Salt and water movement in the Liverpool Plains—What’s going on?LWRRDC Occasional Paper No. 14/97, LWRRDC, Canberra, 10p.

Tadros, NZ 1993, The Gunnedah Basin, New South Wales. Memoir 12, DMR Coal and Petroleum Geology BranchGeol. Surveys of NSW, Sydney.

Timms, W 1998, Hydraulic linkages between shallow saline groundwaters and pressurised alluvial aquifers on theLiverpool Plains, NSW. Centre for Natural Resources Research Report T2816, NSW Department of Landand Water Conservation, Sydney.


24


25

Timms, WA 1996, Salinity on the Liverpool Plains: A Hydrogeologic Investigation of the Lower Yarramanbah Valley.BSc Thesis, Australian National University, Canberra.

Zhang, L, Stauffacher, M, Walker, GR & Dyce, P 1997, Recharge estimation in the Liverpool Plains (NSW) for inputto groundwater models. Technical Report 10/97, CSIRO Land and Water, Canberra.

Zhang, L, Dawes, WR & Walker, GR 2001, The response of mean annual evapotranspiration to vegetation changesat catchment scale. Water Resources Research, 37, 701-708.


26

dryland salinity management - csiro · the original case-study was funded through nrms-lwrrdc grant...

Documents