Predicting the effects of landuse change on water and salt balance—a case study of a catchment affected by dryland salinity in NSW, Australia

Journal of Hydrology 283 (2003) 67–90 www.elsevier.com/locate/jhydrol

Predicting the effects of landuse change on water and salt balance—a case study of a catchment affected by dryland salinity in NSW, Australia Narendra Kumar Tutejaa,*, Geoffrey Bealeb, Warrick Dawesc, Jai Vazea, Brian Murphyd, Paul Barnetta, Aleksandra Rancicb, Ray Evanse, Guy Geevesd, Daud W. Rassamf, Michelle Millerb a

DIPNR Centre for Natural Resources Queanbeyan, NSW Department of Infrastructure, Planning and Natural Resources, P.O. Box 189, Queanbeyan, NSW 2620, Australia b DIPNR Centre for Natural Resources Wagga Wagga, P.O. Box 5336, Wagga Wagga, NSW 2650, Australia c CSIRO Land and Water, Black Mountain Laboratories, GPO Box 1666, Canberra, ACT 2601, Australia d Centre for Natural Resources Cowra, P.O. Box 445, Cowra, NSW 2794, Australia e Salient Solutions P/L, 30 Carolyn Jackson Dr., Jerrabomberra, NSW 2619, Australia f Department of Natural Resources and Mines, 80 Meiers Rd., Indooroopilly, Qld 4068, Australia Received 5 September 2002; accepted 23 June 2003

Abstract An integrated and comprehensive framework for the assessment of water and salt balance for large catchments is presented. The framework is applied to the Mandagery Creek catchment (1688 km2), located in the south-eastern part of Australia. The catchment is affected by dryland salinity and the effects of landuse, climate, topography, soils and geology on water and salt balance are examined. Landuse change scenarios designed to: (a) increase the perennial content of the pastures and crop rotations and (b) increase the current remnant native woody vegetation with additional tree cover are investigated to determine the level of intervention required to develop ameliorative strategies. Likely downstream impacts of the reduction in water flow and salt export are also estimated. q 2003 Elsevier B.V. All rights reserved. Keywords: Dryland salinity; Landuse change; Topographic indices; CATSALT; FLOWTUBE; HYDRUS-2D

1. Introduction Market based solutions for the management of integrated problems of water quality, river regulation * Corresponding author. Tel.: þ61-2-6298-4021; fax: þ 61-26299-6619. E-mail address: [email protected] (N.K. Tuteja). 0022-1694/$ - see front matter q 2003 Elsevier B.V. All rights reserved. doi:10.1016/S0022-1694(03)00236-1

and water resources allocation to primary industry and the environment are currently under development in river basins across Australia. A system of water quality targets, benchmarked at major basin outlets and strategic locations within each river basin network has been adopted (DLWC, 2000; MDBMC, 2001). Targets enable policy to systematically

68

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

prioritise public spending and facilitate the introduction of trading in environmental credits. These include salinity, carbon and biodiversity credits. A prerequisite for developing policy initiatives for controlling salinity is modelling of the impacts of landuse change on water yields, salt export, and aquifer response times, to assess the biophysical capacity to change. This paper describes the results from a study of the effects of landuse change on sub-catchment scale, daily water and salt balances for Mandagery Creek (1688 km2) in the Lachlan River basin, New South Wales (NSW), Australia. It is the last major tributary to join the Lachlan River above the town of Forbes, which is the location of the primary salinity target for the main catchment. The Forbes target is set to protect downstream environmental, community and primary industry assets. These include major irrigation assets, town water supplies and an internationally significant RAMSAR listed wetland (The Great Cumbung Swamp) in which the system generally terminates. Flow in the Lachlan River is regulated by a series of weirs and major dams in the upper catchment. Water sharing plans are being developed across the state of NSW. The plans will establish water-sharing rules that will determine the availability of water to licensed water users and will include provision of flow for basic riparian rights as well as environmental flows. The contribution of flows from unregulated streams such as Mandagery Creek is an important component of the plan. Salinity abatement strategies involving large-scale landuse change in the unregulated tributary catchments aimed at achieving the catchment salinity targets must be balanced against availability of water by the water sharing plans. A comprehensive modelling framework that combines modelling at different scales using some new and innovative techniques has been developed and implemented to support these policy initiatives. It is generally accepted that the spread of dryland salinity in the upper parts of the Murray-Darling Basin has resulted from the clearing of native vegetation for European-style agriculture (Walker et al., 1992; Williamson et al., 1997). The issue of vegetation effects on catchment water balance has been the subject of extensive observation and modelling across the world for many years (Horton, 1919; Penman,

1963; Bosch and Hewlett, 1982; Cornish, 1993; Vertessy et al., 1996; Zhang et al., 1999). Complex distributed parameter process models such as SHE (Abbott et al., 1986), TOPOG_IRM (Dawes and Hatton, 1993) and CATPRO (Ruprecht and Sivapalan, 1991) require comprehensive calibration data sets to parameterise the many micro-scale processes incorporated in their procedures. As such, they are used primarily as research rather than management tools. One or two dimensional process water balance models such as APSIM (McCown et al., 1996), HYDRUS-2D (Sˇimunek et al., 1999) and PERFECT (Littleboy et al., 1992), used in a GIS framework, are useful for assessing the relative efficiencies of runoff generation and leakage from various combinations of landuse, soil type, slope class and land management practices (Paydar et al., 1999; Ringrose-Voase and Cresswell, 2000). However, there is often a mismatch between the calculated recharge rates and base flows observed in streams due to scale effects and inadequate accounting for lateral fluxes in a purely vertical analysis. Considerable attention has been given in the literature to the effect of scaling issues and spatial variations in soil moisture (Rodrı´guez-Iturbe and Gupta, 1983; Dooge, 1986; Blo¨schl and Sivapalan, 1995; Grayson et al., 1997; Western and Blo¨schl, 1999). In recent times, complex process modelling approaches have given way to more simplistic approaches due to issues of scale and data availability. Many authors (Holmes and Sinclair, 1986; Turner, 1991; Zhang et al., 2001) have developed relationships between vegetation type and average annual evapotranspiration from a small number of readily available parameters. Although useful, these coarse average annual relationships provide insufficient information on the temporal effects of tributary flow and salinity required for water management purposes. Reporting on a steady state analysis using a groundwater model, MODFLOW (McDonald and Harbaugh, 1988) and a contaminant transport model MT3DMS (Zheng and Wang, 1999), Gates et al. (2002) highlight the need for unsteady analysis to assess temporal stream impacts of strategies to lower water tables and soil salinity in an irrigation district in Colorado USA. Topography is the dominant physical driver of flow in a watershed and is the primary determinant of

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

catchment response to rainfall (Beven and Kirkby, 1979; O’Laughlin, 1986; Scanlon et al., 2000). Several topographic indices relating position in the landscape to hydrologic behaviour have gained acceptance as predictors of surface, sub-surface and groundwater flow (Quinn et al., 1995; Roberts et al., 1997). Moore et al. (1991) reviewed the application of digital terrain modelling in hydrology, geomorphology and biology. Tarboton (1997) proposed robust, precise and useful multiple flow path algorithms that minimise dispersion and avoid grid bias in determining the up-stream contributing area. Gessler et al. (1995) demonstrated that subclasses within the distribution of the compound topographic index (CTI) were able to predict the boundaries of pedologic differentiation of soils forming a catena. Using similar techniques and alternative indices with additional salinity information from soil profiles and salt outbreak mapping, an acceptable surface salt store map can be produced from soil landscapes mapped according to the methodology described by McDonald et al. (1990), Abraham and Abraham (1996) and Chapman and Atkinson (2000). Lumped parameter conceptual rainfall-runoff models such as SMAR (O’Connell et al., 1970; Kachroo, 1992) and the SACRAMENTO Model (Burnash and Ferral, 1972) have been used around the world for many years in river flow forecasting. They are particularly useful in estimating the major components of the water balance at a daily time step (evapotranspiration, surface runoff, baseflow and interflow) from rainfall, pan evaporation and gauged total stream flow (Franchini and Pacciani, 1991). In this study, lumped conceptual rainfall runoff modelling techniques are combined with landuse efficiency indices, obtained from process modelling, topographic modelling, salinity hazard mapping and salt outbreak mapping, to investigate the effects of landuse on water and salt balance. The study draws its strength from a new and innovative approach of combining these techniques at the appropriate scale, while allowing for heterogeneity within the catchment. Unlike conventional salinity studies that focus on groundwater alone, this study explores surface and groundwater interactions with salt stores and the stream.

69

2. Modelling methodology The modelling methodology consists of application of three different types of models—CATSALT (see Fig. 1; Tuteja et al., 2002; Beale et al., 2000), HYDRUS-2D (Sˇimunek et al., 1999) and FLOWTUBE (Dawes et al., 2000). CATSALT is a quasi-physical model and is developed to couple landscape salinisation and stream salinity (Sections 2.1– 2.3). The model operates on a daily time step. It includes three modules: (a) a lumped conceptual rainfall runoff model SMAR (O’Connell et al., 1970; Kachroo, 1992; Tuteja and Cunnane, 1999; see Fig. 2 and Section 2.1), (b) a runoff distribution component based on landuse and topography (Section 2.2), and (c) a salt mobilisation and washoff component (Section 2.3). To incorporate the effects of landuse change, the distribution methodology requires information on leakage rates for all soil types and landuse combinations within the catchment. These are obtained from the Richard’s equation based process model HYDRUS-2D and also from published data. Additionally, by combining the results from HYDRUS-2D with those from SMAR, the water balance can be confidently closed. FLOWTUBE is a Darcian concept based groundwater flow model. The model provides information on long term shifts in groundwater flux and the associated time constants corresponding to the changes in landuse/recharge regime. 2.1. Lumped rainfall runoff modelling The SMAR model consists of two components in sequence, namely, a water balance component and a flow routing component (see Fig. 2). The water balance component divides the soil column into horizontal layers, which contain a prescribed amount of water at field capacity. Evaporation from the soil layers is assumed to vary as a function of the form Ci21 ; where C is a parameter (less than 1) and i ¼ 1; 2; 3… refers to the successive soil layers. When treated in this way, a constant potential evaporation would reduce the soil moisture storage in approximately an exponential manner. The routing component transforms the runoff generated from the water balance component into discharge at the catchment

70

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

Fig. 1. Schematic diagram of the CATSALT model.

outlet by a Gamma Function Model form (Nash, 1960), a parametric solution of the differential equation relating the input to the output with the shape and the scale factors as n and nk; respectively,
n21 1 t hðtÞ ¼ e2t=k ð1Þ kGðnÞ k where hðtÞ is the unit impulse response function ðd21 Þ; n is the number of linear reservoirs (dimensionless) in the cascade with equal storage coefficients kðdÞ; and Ð 2t n21 GðnÞ ¼ 1 dt is the incomplete Gamma 0 e t function. The water balance component has five parameters ðC; Z; H; Y&TÞ and the routing component has four parameters (n and nk; G and Kg ). A brief description of the water balance parameters is as follows—the parameter C (dimensionless) regulates evaporation from the soil layers, the parameter Z (mm) represents the effective soil moisture storage capacity, the parameter H (dimensionless) is used to estimate the variable H 0 (a fraction of rainfall

excess that contributes to the generated runoff), the parameter Y (mm d21) represents the infiltration capacity of the soil and the parameter T (dimensionless) is used to estimate the potential evaporation from pan evaporation. The variable H 0 is estimated as H £ zðtÞ=125; where H is a parameter, zðtÞ is available soil moisture in mm at time tðdÞ and 125 mm represents the maximum effective soil moisture content of the first five layers (Khan, 1986). A proportion G of the generated runoff resulting from moisture in excess of effective soil storage capacity is routed through a linear reservoir (Eq. (1) with n and GðnÞ equal to 1). This represents the ground water contribution to the stream and the remaining amount is added to the generated surface runoff, which is routed by a Gamma Function representing the surface runoff (Kachroo and Liang, 1992). All nine parameters, comprising five water balance parameters and four routing parameters in SMAR, are optimised according to the Nash and Sutcliffe (1970) efficiency criterion.

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

71

Fig. 2. Schematic diagram of the SMAR model (from Kachroo (1992)).

2.2. Distributing runoff to source areas From SMAR simulations, the estimates of total ^ daily-simulated streamflow QðtÞ; the daily-simulated r ^ surface runoff Q ðtÞ; and the daily-simulated

^ g ðtÞ are available. Runoff groundwater discharge Q is distributed on the basis of landuse and topography using the following methodology. Five dryland landuses are considered: trees, perennial pastures, annual pastures, cropping and other (e.g.

72

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

urban, wetland etc.). Surface runoff from the area ^ ri ðtÞ (mm d21) depends on area under each landuse Q and the relative landuse efficiency of runoff generation. 0 1 ATþdT p P5 i · P5 i TþdT B C r j¼1 pj j¼1 Aj CQ ^ ^ ri ðtÞ ¼ B ð2Þ Q @ P5 A ðtÞ ATk pk · P5 k¼1 P5 T j¼1 pj j¼1 Aj where i; j and k refers to the landuse, ATi ; ATj and ATk refers to the area under each current landuse in dT dT the catchment (km2), ATþ and ATþ refers to the i j area under each future landuse change scenario (km2), pi ; pj and pk refer to the runoff efficiency of landuses i; j and k relative to annual pasture (dimensionless). The runoff efficiency pi is defined as the ratio of long-term mean annual runoff from areas under landuse i compared with thatP from areas under annual pasture. The ratio pi = 5j¼1 pj incorporates the scaled effect of landuse i on runoff generation P dT P5 TþdT and the ratios ATþ = A and ATi = 5j¼1 ATj j¼1 j i incorporates the scaled effect of the area under a specified future landuse change scenario at time T þ dT and current landuse scenario T: When dT dT equals zero, the areas ATþ collapse to the i respective current landuse areas ATi : Under such conditions, runoff volume conservation P5 ^ r from^ r the catchment is guaranteed—i.e. j¼1 Qi ðtÞ ¼ Q ðtÞ: However, under changed landuse scenario runoff volume will not be conserved (e.g. if the area under annual pasture is changed to trees, the result is higher evapotranspiration and therefore a reduced runoff efficiency of trees relative to annual pasture; see Section 4.2). The effect of landuse on the groundwater component can be estimated by substituting for pi in Eq. (2), the efficiency of the landuse i to groundwater recharge with respect to annual pasture denoted by mi (Section 4.2). The effect of topography is introduced using the TOPMODEL wetness index (Beven and Kirkby, 1979) to disaggregate surface runoff within a given landuse:
 a y ¼ ln ð3Þ tan b where y is wetness index at a given pixel location, a is upstream contributing area at the given location per

unit contour length, and tan b is slope of the landscape at the given location. Many schemes are available for calculating the upstream contributing area (O’Callaghan and Mark, 1984; Quinn et al., 1991; Lea, 1992; Costa-Cabral and Burges, 1994; Tarboton, 1997). Tarboton (1997) proposed a new multiple flow path algorithm, called the D1 method, that performed better than most other methods. This method was used for estimating the upstream contributing area in this study. The wetness index from Eq. (3) generally varies in the range 0– 20 and can be divided into lmax number of classes with a predefined uniform class interval. The wetness index for all pixel locations is disaggregated on the basis of landuse—i.e. a probability distribution function (pdf) of the topographic wetness index is constructed for each landuse by masking ^ ri ðtÞ obtained from Eq. other landuses. Surface runoff Q (2) for each landuse i is then used to obtain surface runoff from each wetness index class l for the ^ ril ðtÞ (mm d21) as in Eq. (4). specified landuse area Q pX max

^ ril ðtÞ Q

¼

yil ðpÞ p¼1 max lX max pX

^ ri ðtÞ ¼ wil Q ^ ri ðtÞ Q

ð4Þ

yil ðpÞ

l¼1

p¼1

where yil ðpÞ is wetness index of each pixel p within landuse i and wetness index class l; p max is number of pixels within each landuse i and wetness index classes l; wil is scaled wetness index of the wetness index class l and landuse i (dimensionless), l max refers to maximum number of the wetness index classes. 2.3. Salt mobilisation and washoff The salt mobilisation and wash-off model links salt sources from within the catchment with the distributed surface runoff obtained using the procedure described above. Water entering the river as surface runoff undergoes sorptive exchange with salts (cations and anions) adsorbed to the soil matrix. The general form of the non-linear Freundlich isotherm is used to describe the exchange process. The relationship is linear ða ¼ 1Þ when abundance of salts in water does not affect the dissolution of salts from the soil. However, the process is non-linear ða , 1Þ when high

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

salt concentration in water affects dissolution of salts from soil in water (see Schwarzenbach et al. (1993) for details on sorptive exchange). Cs ¼ KF Cwa

ð5Þ 21

where Cs is soil salinity (g kg ), Cw is water salinity (g m23), KF is Freundlich constant (m3 kg21), and a is measure of non-linearity of the sorptive exchange process (dimensionless). Using Eqs. (4) and (5), salts entering the river with surface runoff can be estimated from all landuses and wetness index classes as in Eq. (6). A constant rainfall salinity value is added to the surface runoff component, which in the Australian context represents cyclic salt derived primarily from oceanic sources (Blackburn and McLeod, 1983). " #
 5 lX max X Cs;il 1=a ^ r r r ^ ^ Qsalt ðtÞ ¼ g ð6Þ þQil ðtÞCrain Qil ðtÞ KF i¼1 l¼1 where Cs;il is soil salinity of saline or non-saline soils for each landuse i and wetness index class l (g kg21), Crain is salt concentration in rainfall (g m23), g ¼ 0:0864 is the conversion factor that converts the expression on right hand side of Eq. (6) from g s21 to ^ rsalt ðtÞ is salts entering the river with surface t d21, Q runoff (t d21) and i ¼ 1; 2…5 refers to the 5 landuses. Cs;il in Eq. (6) represents the depth averaged soil salinity, i.e. depth through which surface runoff occurs including the shallow sub-surface runoff. Surface runoff from each landuse and wetness index ^ ril ðtÞ when used in Eq. (6) is converted from class Q 21 mm d to m3 s21. In implementing Eq. (6), a separate accounting was done for saline and nonsaline areas within the same wetness index class. The residence time of groundwater within catchments at this scale is quite large. Salts entering the river with base flow can be expressed as in Eq. (7). ^ g ðtÞ ¼ gbCg Q salt

5 lX max X

^ g ðtÞ Q il

ð7Þ

i¼1 l¼l1

where Cg is salt concentration in groundwater (g m23), b is parameter to allow hydraulic contact between aquifer river exchange and mixing in the shallow fresh ^ g ðtÞ is mean daily water lenses (dimensionless), Q il groundwater flow contribution to the river from different landuse and wetness index class combinations (m3 s21).

73

It is assumed that groundwater drains into the stream and not vice versa i.e. effluent stream. Primary interest is in estimating how much salt enters the stream from the groundwater system through hydraulically connected pathways. These connections are mainly from shallow groundwater lenses usually with low salt concentration. Therefore, it is reasonable to assume that considerable dilution of the groundwater flow occurs when water from the groundwater system enters the stream. The parameter b represents the net effect of these processes and it can take a value between 0 and 1. The remaining amount contributes to soil salinisation in the alluvium from where it is picked up by the near surface transport processes. Daily total simulated in-stream salt load is obtained by adding surface runoff contribution from Eq. (6) and groundwater contribution from Eq. (7). Three additional parameters, KF ; a and b; are introduced over and above nine parameters in the water balance model. These are calibrated against the daily observed in-stream salt load. There are many catchments in NSW where long-term salt export rates are required but the available salinity data are either for a short period or are patchy. In such instances, a daily salt load time series can be created using an objective methodology based on stochastic hydrology concepts (see Beale et al. (2000) and Tuteja et al. (2000a) for details). 2.4. Long term groundwater response to vegetation changes The CATSALT model is well suited to describing a system that is in dynamic equilibrium, where none of the fitted parameters change over time (e.g. 50 years or more). A suitably long record of stream flow and salt load with stable land management allows CATSALT to be calibrated and reproduce historical flow and salt behaviour. What is more difficult is to estimate the catchment behaviour under a very different vegetation pattern. Some, or all, of the water balance and routing parameters, or the area of saline land, may change when a catchment is fully cleared or fully forested for example, as the groundwater system equilibrates over time with the new recharge regime. The shift in water balance is incorporated using process modelling with HYDRUS2D and (2). However, long-term issues relating to

74

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

timing and extent of groundwater response need to be assessed. For this reason, a one-dimensional numerical groundwater flow model FLOWTUBE (Dawes et al., 1997, 2000, 2001) is used. It is a mass-balance model that solves for a change in hydraulic head induced by recharge and discharge fluxes, and lateral transfers in the direction of flow described by Darcy’s Law. The results from FLOWTUBE include a transect of hydraulic heads along an aquifer. FLOWTUBE considers a one- or two-layer system. In the case of a single layer system the aquifer is assumed to be unconfined with variable transmissivity dependent on the saturated thickness of the aquifer. In the case of a two-layer system, the lower layer is assumed to contain any lateral transmission of water while the upper layer contributes to storage capacity only. In this case, the lower layer is confined or semi-confined, and this controls the simulated mechanism for groundwater recharge and discharge. FLOWTUBE is ideally suited to homogeneous uniform isotropic media, such as sand and gravel aquifers, and massive clay deposits without preferred pathways or barriers. The Mandagery Creek catchment aquifers contain both alluvial material and fractured rock that are geologically complex with faulting and folding. The fractured rock parts are the least applicable situation for FLOWTUBE, and the parameterisation of the aquifer is a critical issue for both fitting of current heads and confidence in simulated future scenarios. In fractured rock environments it is also difficult to estimate the extent of aquifers, which in many cases are simply assumed to be coincident with topography. Although difficult, fractured rock aquifers can be parameterised by selecting representative volumes large enough to make Darcian flow estimation valid (Freeze and Cherry, 1979). In the absence of measured parameters, the fitted values will be representative at the scale of the FLOWTUBE segments. The FLOWTUBE model requires three types of input: aquifer geometry, aquifer state parameters, and boundary conditions. The aquifer geometry is inferred from geological maps, drilling logs and topography, to determine the areal extent of an aquifer or unit, its thickness, and how deep it is buried if at all. The aquifer state parameters are hydraulic conductivity in the direction of flow and available porosity. These

values are most often determined from pump tests of the aquifer, or inferred from examining similar geological types and situations. The final type of input is in many ways the most difficult. At the top end of the FLOWTUBE, a flux of water enters the aquifer that may be zero, and at the bottom end there is a flux calculated from the simulated head at the bottom and a fixed head outside the catchment that may also be zero. Along the aquifer itself there is diffuse recharge, which is the water that escapes the surface soil layers and percolates to the aquifer being simulated. This recharge value is an input to FLOWTUBE, and may vary spatially, for each segment, and temporally, at each time step.

3. Catchment description and data used in the study The modelling methodology described above was applied to the Mandagery Creek catchment located in Central West of New South Wales (NSW), Australia (see Fig. 3). The total catchment area of Mandagery Creek upstream of Eugowra (Gauging station number 412030) is 1688 km2. The study area is divided into four sub-catchments: Mandagery Creek at Manildra: 489 km2 (412075), Boree Creek at Cudal: 253 km2 (412090), and Bourimbla Creek at Cudal: 172 km2 (412076). The fourth sub-catchment, called the residual catchment, comprises the area downstream of the above three sub-catchments and upstream of the outlet at Eugowra: 774 km2 (412030). 3.1. Climate Climate surfaces are available for the Australian continent using the widely accepted methodology of Jeffrey et al. (2001). Climate surfaces were drilled at a 5 km grid for each sub-catchment and spatially averaged for modelling purposes. Climate, runoff and slope statistics for each sub-catchment are shown in Table 1. 3.2. Landuse Landuse data for the Mandagery Creek catchment was obtained from the following sources: DIPNR landuse maps for the catchment, and the Australian

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

75

Fig. 3. Location map of the Mandagery Creek catchment, NSW, Australia Gauging stations: 412075 – Mandagery Creek at Manildra, 412090 – Boree Creek at Cudal no. 2, 412076 – Bourimbla Creek at Cudal and 412030 – Mandagery Creek at Eugowra.

Table 1 Climate, runoff and slope statistics for the mandagery catchment Catchment

Mandagery at Eugowra

Boree Ck. at Cudal no. 2

Mandagery at Manildra

Bourimbla Ck. at Cudal

Gauging station Area (km2) Rainfall (mm yr21)a Pan evap (mm yr21)a Observed runoff (mm yr21) Avg. slope (deg.) Slope range (deg.)

412030 1688 689 1529 40.2 (1995–99) 5.4 0–42.9

412090 253 751 1489 68.3 (1971–89) 4.8 0– 33.4

412075 489 715 1525 61.1 (1967–81) 4.2 0–29.3

412076 172 788 1480 54 (1975–99) 5.4 0 –33.3

a

Rainfall and pan evaporation data refers to the period 1975–99.

76

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

Landuse Classification System—ALCC (Barson et al., 2000; Stewart et al., 2001). These data sets comprise six landuse categories (cropping, trees, pasture, urban, bare, and water see Sections 4.2 and 4.3). Water bodies, bare and urban areas constitute a very small proportion of the total catchment and therefore have been classified as ‘others’. Pastures are further sub divided into perennial and annual components, based on land capability class and expert local knowledge. Tree cover in the catchment occurs mainly as open woodlands, comprising remnant native vegetation. The landuse change scenarios were developed based on the land capability mapping and soil conservation guidelines of the DIPNR. The developed landuse change scenarios were spatially distributed by linking with the wetness index. † Scenario 1. All area under annual pasture was changed to perennial pasture and 30% of the area under cropping was changed to perennial pasture. This scenario is designed to increase the perennial content of the vegetation. † Scenario 2. Increase the tree cover to 20% (Scenario 2.1), 30% (Scenario 2.2) and 40% (Scenario 2.3) of the sub-catchment over and above the landuse changes in Scenario 1. Scenarios 2.1 –2.3 are based on incorporating more stringent recharge control measures to mitigate landscape salinisation in the discharge areas: upland areas of the landscape units modelled in Scenario 1 as perennial pasture were changed to tree cover. Typically, these areas have relatively higher leakage rates and are associated with low values of the wetness index. 3.3. Description of geology The catchment consists of a complex combination of geological units that can be grouped into nine distinct groundwater flow systems (see Fig. 4) (Evans, 2001). All of the flow systems (except the youngest alluvium) are hard-lithified units that have fracturing as their main form of permeability. In general, it is assumed that groundwater flow in these systems follows the general topographic gradient. However, there may be some local scale exceptions to this assumption. As well, the flow systems have short flow

paths (1 – 2 km) and are considered to be generally low in hydraulic conductivity. Upper Devonian Sediments are generally forested and are associated with the highest relief areas of the catchment with slopes in the range 10 – 308. These sediments have low hydraulic conductivity and extremely high hydraulic gradients, and often act as a barrier to other groundwater systems. Granites are found in the northern and western parts of the catchment in areas with moderate relief. The flow system is best developed where there has been substantial weathering with groundwater flow occurring in the colluvial formations. Mesozoic Sediments are found in the northern part of the catchment and are a subset of a more expansive set of sediments further north of the catchment. The geology exhibits substantial layering of the aquifer system that creates perched watertable conditions and a resultant prevalence of lateral groundwater flow, sometimes independent of topographic slope. Dulladerry Volcanic forms massive hills with rock outcrops in the north-western part of the catchment. The flow system is developed in extensive shallow colluvial deposits that act as a main conduit of groundwater and salt fluxes. Cudal, Mumbil and Goonigal Group Volcanics and Sediments form a broad belt with north-south bedding planes through the middle part of the catchment. The rocks are composed of inter-bedded sequences of volcanics and sediments that produce substantial hydraulic conductivity contrasts. The hydraulic gradient from the eastern and north-eastern parts of the catchment is orthogonal to this geology and there is sufficient opportunity for groundwater and salt to discharge to streams. Ordovician Sediment and Volcanic formations in the southern and eastern parts of the catchment are somewhat uniformly fractured, highly folded and faulted. Groundwater generally occurs under phreatic conditions in this geology upstream of sub-catchment boundaries, but tends to be semi-confined in low slope areas. Tertiary Basalts comprise highland areas around the Mt Canobolas volcano remnant to the east and former valley infills that now occur as ridge capping deposits in the central parts of the catchment around Cudal. These deposits are gently dipping layered bodies of lithified ash and basalts. Flow paths are

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

77

Fig. 4. Map of the geological units in the Mandagery catchment (from Evans (2001)). Note that the soil distribution of soils is strongly linked to the geology as discussed in Section 3.4.

generally short due to high relief and deep incision of the bounding streams. Gregra Group Sediments, Volcanics and Limestones on the eastern side of the catchment are similar to the Cudal group, with somewhat lesser contrast in hydraulic conductivity. This geological unit is uniformly fractured and flow paths correspond to the hillslopes. This geology contains many limestone units that have well-developed karst with high hydraulic conductivities. It is considered that although the limestone layers are thick and exhibit karstic features, they are not interconnected enough to be considered as a regional aquifer. Cainozoic Alluvium constitutes small areas distributed across the catchment in the major drainage

lines. For the purpose of groundwater analyses, the flow system across this geology is insignificant and has been ignored. 3.4. Description of soils The soils in the catchment are strongly influenced by geology and have been mapped as soil landscapes at 1:250 000 scale (Kovac et al., 1989; King, 1998). The soil landscape units are topographic sequences of soils, with different soil types occupying different landform elements. The usual sequence consists of shallower, well-drained soils on crests and upper slopes, grading to deeper, poorly drained soils on lower slopes and depressions. Alluvial soils are

78

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

associated with definable floodplains and terraces. For the modelling exercise, terrain analysis was used to sub-divide the catchment into four slope classes (low, low to mid, mid to upper and upper slopes) to separate areas of individual soil types within soil landscapes as in McDonald et al. (1990). Soils hydraulic properties affect the flow of water through the landscape by controlling the partitioning of water between runoff and storage for evapotranspiration and deep drainage. In order to predict the soil hydraulic properties with a sufficient degree of confidence it was necessary to identify individual soil types down to the soil family or soil series level (Chapman and Atkinson, 2000). In general, identification of soil types down to this level enabled the texture and structure of individual soil horizons to be identified. The assumption was made that each Great Soil Group (Stace et al., 1968) within a soil landscape was equivalent to a soil family or soil series. For example, the non-calcic brown soils within the Nangar Soil landscape are assumed to form the soil family ‘Nangar non calcic brown soils’. Some soil landscapes that only covered small areas (, 2%) were assigned the soil properties of soil landscapes with similar soil types. Relationships between texture and structure and soil hydraulic properties that had been developed in the literature and from other data sets were then used to predict the soil hydraulic properties of the individual soil types (see Section 4.2). The soil landscapes within the Mandagery Creek catchment are related to the geology (see Fig. 4) and the soils are described according to the Great Soil Group system (Stace et al., 1968), but the links to Soil Taxonomy (USDA United States Department of Agriculture, 1975) are also shown (see Appendix A). 3.5. Stream salinity data Streamflow and salinity data at gauging stations Manildra (412075), Cudal (412090) and Eugowra (412030) shows relatively higher salt concentration from the sub-catchment at Manildra (412075) (see Fig. 6). Salinity is measured as electrical conductivity (EC) where 1 EC unit ¼ 1 mS cm21 or 0.64 mg l21 at 25 8C. Additionally, a rapid streamflow and salinity survey (RSS) was conducted at 14 strategic locations on June 20 – 21, 2001 during low

flow conditions (comprising baseflow from groundwater) to assess salt contributions from different geological units. A water sample was taken at each site for laboratory testing to determine ion composition and the total dissolved salts. During the RSS, salinity of the groundwater flux draining through the northern part of the catchment at Manildra (412075) was about four times that from eastern parts of the catchment draining at gauging stations 412090 and 412076. The primary reason for this is that more saline groundwater flux from the Dulladerry Volcanics is intercepted by the Cudal Group Volcanic (see Fig. 4) with short flow paths and bedding planes transverse to the hydraulic gradient. Long term mean annual flow (1975 – 99) at Manildra (412075) is about 48% of the combined flow at Eugowra (412030). Therefore, the sub-catchment at Manildra (412075) accounts for bulk of the salt load exported from Mandagery Creek at Eugowra (412030). The average ion composition of the total mass of dissolved salts was: Chloride 32%, Sodium 15%, Calcium 13%, Magnesium 9%, Sulphate 10% and Potassium 1%. The concentration of other anions, such as nitrate and carbonate, was not measured. During the RSS, streamflow contributions at Manildra (412075), Boree Creek (412090) and Bourimbla Creek (412076) and the residual area expressed as percentages of the Mandagery Creek at Eugowra (412030) were 13, 49, 3 and 35%, respectively. The respective percentage salt load contributions and electrical conductivity were 24% and 2679 mS cm21, 26% and 748 mS cm21, 2% and 824 mS cm21, 48% and 167 mS cm21. These data represent in-stream salinity during low flow conditions. In-stream salinity data for the gauging stations at 412075, 412090 and 412030 were used for model calibration. The underlying geologies of the Boree Creek (412090) and Bourimbla Creek (412076) subcatchments are mainly Tertiary basalt and Ordovician sediments (see Fig. 5). The missing salinity data for 412076 were therefore constructed using the flow and EC/salt load relationships for 412090. Long-term salt loads from Mandagery Creek at Manildra (412075), Boree Creek (412090) and Bourimbla Creek (412076) are 13565, 4710 and 3476 t yr21, respectively.

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

Fig. 5. Flow and electrical conductivity (EC) data for the gauging stations in the Mandagery catchment: 412030 (EC_412030), 412075 (EC_412075), 412090 (EC_412090) and data from the rapid stream survey (EC_RSS).

4. Methodology implementation and results 4.1. Water balance components and runoff routing Observed streamflow data for the three subcatchments (412075, 412090 and 412076) and the complete catchment (412030) were available for different periods (Table 1). Many short periods of streamflow at 412075 were missing during the period 1967 –81. These were filled up on the basis of rainfall, potential evaporative demand and trends in streamflow. The SMAR model was first applied to each of the three sub-catchments and also to the complete catchment to estimate the partitioned water balance components consistent for the period 1975 – 99. Model performance for each case was evaluated using the R2 efficiency criterion of Nash and Sutcliffe (1970). To further evaluate the model performance, streamflow from each sub-catchment was routed using the multiple input-single output system methodology of Kachroo and Liang (1992). O’Connell et al. (1970) and Kachroo (1992) and others found that very simple forms of the SMAR model, of only one or two water balance parameters, were often sufficient to adequately explain the catchment scale water balance. A calibration procedure based on explained variance was implemented, where one or more water balance parameters were optimised together with the routing parameters. The remaining water balance parameters were set at values such that there is no effect on model performance (inoperative level). The above procedure, using inoperative levels of parameters is commonly used

79

during investigations of a catchment with conceptual models (see Kachroo, 1992). To identify the most suitable model form, we proceeded from the simplest to more complex forms, judging whether the improvement in R2 justified the more complex model form. With the exception of the infiltration capacity of the soil ðYÞ; all the remaining parameters were found to explain a reasonable proportion of the total variance for all sub-catchments. Therefore, Y was fixed at an inoperative level (100 mm d21) by assigning a value larger than the maximum daily rainfall and the remaining parameters were optimised. The model was first implemented using a split sample approach, wherein two thirds of the data were used for calibrating the model and the remaining data were used for model verification. After obtaining a satisfactory model performance, the complete data set was pooled and used for estimating parameter values and water balance components for each sub-catchment (see Table 2). The observed and Table 2 Water balance components and parameters from the SMAR model Catchment

412030

412090

412075

412076

C (2) Z (mm) Y (mm d21) H (2) T (2) G (2) n (2) nK (d) Kg (d) R2 (%) Rain (mm yr21) ^ r (mm yr21) Q ^ g (mm yr21) Q ETa (mm yr21) Qrech (mm yr21) Error (mm yr21) KF (m3 kg21) a (2) b (2)

0.900 191 100 0.100 0.860 0.570 1.00 1.350 895 79 689 26.2 9.8 658.8 17.2 -0.65 0.032a 0.88a 0.46a

0.761 166 100 0.123 0.855 0.775 3.86 1.210 463 64 751 41.6 36.4 676.9 47.0 -0.54 0.00256 0.72 0.25

0.700 166 100 0.123 0.855 0.774 3.86 1.210 547 40 715 33.0 23.0 662.6 29.7 -0.48 0.00576 0.79 0.80

0.966 274 100 0.147 0.886 0.502 1.346 1.148 406 65 788 48.0 8.0 745.0 15.9 0.9 0.00128 0.71 0.58

C; Z; Y; H and T are the SMAR water balance parameters; ^ r —surface n; nK; G and Kg are the SMAR routing parameters; Q ^ g —groundwater discharge, ETa—actual evapotranspirarunoff, Q tion, Qrech —leakage from the soil profile, error represents the mass balance error and KF ; a; and b represent the salt balance parameters. a Parameters for salt contribution from the residual catchment. Simulation periods—412030 (1975 – 99), 412090 (1971 – 99), 412075 (1967– 99) and 412076 (1975–95).

80

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

Fig. 6. (a) Observed and estimated daily streamflow data at 412090 (1971–89), (b) observed and estimated annual streamflow data at 412090 (1971–89).

the estimated daily and annual streamflow data are shown in Fig. 6 for the sub-catchment 412090. The proportion of areas under different landuse is broadly similar in the sub-catchments at Manildra (412075) and Cudal (412090) (see Figs. 7a and 8a). In order to avoid the effect of gap filling on the parameter values for the sub-catchment at Manildra (412075), the parameters relating to surface runoff (H; Z; n; nK and T) for Manildra were adopted from the Boree Creek sub-catchment and were constrained during the optimisation process. The remaining parameters were optimised during the calibrations (G; Kg ; C and Y; see Table 2). This resulted in lower values of R2 for Manildra (412075) than the case of unconstrained optimisation due mainly to the fitting process. The observed streamflow data at 412030 are available for 1995 – 99. Therefore this period was

used to implement the multiple input (412090, 412075, 412090)-single output (412030) routing model based on constrained least squares theory (Kachroo and Liang, 1992). Total volume gain at 412030 equal to 12.6% over and above the combined flow of 412090, 412075 and 412076 was estimated for 1995– 99. The gain factors 1.038, 1.065 and 1.023 were obtained for 412090, 412075 and 412076, respectively. The gain factors when added together represent 12.6% volume gain at 412030. A comparison of the routed flow at 412030 with the observed flow for 1995 – 99 using the above methodology results in R2 of 0.62. However, when comparison of the simulated flow for the lumped catchment at 412030 is made with the routed flow, R2 improves substantially from 0.62 to 0.78. This is because the non-linear effects are somewhat diffused when using

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

81

Fig. 7. (a) Current landuse map of the Mandagery catchment, (b–d) mean annual fluxes (1975–95) for each landuse and soil type combinations obtained from HYDRUS-2D simulations; flux across the atmospheric boundary represents the sum of soil evaporation and runoff. (Note: Landuse change scenarios 1,2.1-3 are simulated only if target area is greater than the respective area under current land use.

simulated flow for the lumped catchment 412030. Given the scale of the problem and the objectives of assessing long-term impacts of landuse change, this is considered a reasonable representation of the water balance at the sub-catchment scale.

4.2. Leakage rate for soil type and landuse combinations Distribution of the water balance components within the sub-catchments is required to incorporate

82

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

Fig. 8. (a) Areas under current landuse (km2), (b) mean annual runoff from areas under current landuse (ML yr21), (c) mean annual salt load from areas under current landuse (t yr21), (d) mean annual leakage with landuse change (mm yr21), (e) mean annual surface runoff with landuse change (mm yr21), (f) mean annual salt load with surface runoff with landuse change (mm yr21), (g) mean annual groundwater flow with landuse change (ML yr21), and (h) Mean annual groundwater salt load with landuse change (t yr21). (Note: Landuse change scenarios 1, 2.1-3 are respective area under current landuse).

the effects of landuse change on export rates from the catchment (parameters pi in Eq. (2) and mi substituted for pi Eq. (2), Section 2.2). Additionally, areas most suitable for implementing landuse change are also to be identified. Therefore, HYDRUS-2D was implemented for all combinations of soil types and landuse combinations in the Mandagery Creek catchment. In this study, HYDRUS was implemented

in one-dimensional mode ie. along the vertical direction. Soil hydraulic properties (SHP). These include values of the parameters required to describe variation of the soil moisture and hydraulic conductivity with pressure head (van Genuchten, 1980). Given the scale of the study area and large variability in the SHP within a soil landscape unit, a reasonable description

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

of the SHP is a difficult task. Two primary data sources, the Australian and the published overseas soils data, were used to develop Pedotransfer Functions (PTF). The Australian data sets are more relevant to the study area for describing SHP for each soil type. These include Geeves et al. (1995), NSW Soil and Land Information System SALIS (database) and Williams et al. (1983). The overseas data set and a hierarchical neural network technique of Schaap and Leij (1998) (included in HYDRUS) was used, along with the Australian data sets for determining the SHP. Soil physical properties used to develop the PTF were based on the following data: soil moisture at 10 and 1500 kPa, bulk density, saturated hydraulic conductivity, soil carbon and particle size distribution. The PTF were applied to the descriptions of the main soil types in Section 3.4. Landuse, plant water uptake and domain size. Areas under current landuse are shown in Figs. 7a and 8a. Three representative landuses were modelled for all soil types in each of the four slope classes (low, low to mid, mid to upper and upper slopes). These are trees, crops and a composite mix of annual and perennial pasture. In the case of cropping, a wheatcanola cropping rotation was used. The plant water uptake function of Feddes et al. (1978) was used in the study. The adopted plant water uptake parameters for the composite pasture are sourced from Tuteja et al. (2000b). For cropping and trees, the plant water uptake parameters were taken from published data included in the HYDRUS-2D database. The rooting depth for each landuse was adopted on the basis of local knowledge. The rooting depths adopted for composite pasture, cropping and trees were 1.25, 0.9 and 6 m, respectively. In the case of cropping and composite pasture, a 2.3 m domain of soil depth was modelled, whereas in the case of trees a 6m domain was used. Three soil horizons were used for all soil types. Leaf Area Index (LAI) and evaporative demand. HYDRUS-2D does not simulate plant growth and therefore plant growth was simulated using PERFECT (Littleboy et al., 1992). The LAI time series from PERFECT simulations was used with Belmans et al. (1983) method to split the potential evapotranspiration demand into potential soil evaporation and potential plant transpiration. Potential transpiration

83

was used with the plant water uptake function and transient soil moisture conditions in HYDRUS to estimate the actual plant water use. Boundary conditions. The atmospheric boundary condition was used where the potential fluid flux (rainfall minus potential soil evaporation) is imposed at the soil surface. The actual fluid flux is computed internally depending on the transient soil moisture condition and is less than or equal to the potential fluid flux. McCord (1991) has shown that the specified unit gradient boundary condition is far superior in the vicinity of the lower boundary than, for instance, the specified head or specified flux boundary conditions. Numerically, this is the only feasible option, as both flux and pressure heads at the lower boundary are unknown. Using daily climate data with the above information, one-dimensional simulations using HYDRUS were carried out. Mean annual leakage from the soil profile, plant water use and flux across the atmospheric boundary (sum of actual soil evaporation and runoff) were obtained and are shown in Fig. 7b –d. The area weighted leakage rates for all sub-catchments obtained from HYDRUS are compared with those from SMAR lumped over the sub-catchment (see Table 3). Leakage from the soil column in SMAR is lumped over the sub-catchment and depends on atmospheric forcing and partitioning of runoff into surface runoff and groundwater discharge. However, leakage in a hill slope context in the case of HYDRUS, depends on interaction between the atmospheric forcing, SHP and vegetation. In view of the scale effects and the difference in modelling approaches, the leakage rates by the two methods are comparable for the subcatchment at Manildra (412075) and the residual

Table 3 Comparison of the mean annual leakage rates (mm yr21) obtained from SMAR and HYDRUS (1975–95) Sub-catchment

Rain

Leakage (SMAR)

Leakage (HYDRUS)

412075 412090 412076 RC

715 751 788 633

29.7 47.0 15.9 5.0

17.4 22.0 34.7 11.7

All values in mm yr21; RC denotes the residual catchment.

84

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

catchment. High mean annual rainfall for the Bourimbla Creek sub-catchment (412076) results in relatively high leakage rate of 34.7 mm yr21 from HYDRUS. However, by incorporating feedback from streamflow where a lower volume of groundwater runoff is seen, a lower leakage rate of 15.9 mm yr21 is estimated from SMAR. Additionally, the underlying geology contains north-south bedding planes with an east-west hydraulic gradient. Combining all the information, it is estimated that a bypass flow in the range 15– 25 mm yr21 from the Bourimbla Creek sub-catchment (412076) to the Boree Creek subcatchment (412090) very likely accounts for the difference in the leakage rate from the two approaches. The landuse efficiency parameters for runoff and recharge were averaged for each landuse in each subcatchment (see Table 4). Since the pastures in HYDRUS were modelled as a composite mix, a partitioning coefficient for annual and perennial pasture equal to 0.6 was assumed for runoff. Similarly, the partitioning coefficient for groundwater recharge equal to 0.5 was assumed. These assumed values were based on published data from Johnston et al. (1999)

and a few HYDRUS-2D runs for the annual pasture system. 4.3. Effect of landuse change on water and salt balance Areas under current landuse in the catchment are shown in Fig. 8a. Estimated runoff from each landuse for each sub-catchment was obtained using the landuse efficiency parameters pi and mi from Table 4 with Eq. (2) and the respective areas under current landuse (see Fig. 8b). Runoff contribution from areas under perennial pasture is more than runoff from areas under annual pasture because of larger areas, even though they are less efficient in runoff generation. The simulated daily salt export rates at each subcatchment outlet were calibrated with the respective time series data. The calibrated parameters KF ; a and b for each sub-catchment are shown in Table 2. Salt load from each landuse for each sub-catchment was obtained using Eq. (6), with the respective data (see Fig. 8c). Highest runoff (26590 ML yr21) and salt load contributions (13670 t yr21) were obtained from the Manildra sub-catchment (412075). Total flow and

Table 4 Landuse efficiency parameters for runoff ðpi Þ and recharge ðmi Þ Catchment

Trees/woodland

Annual pasture

Perennial pasture

Crop

Source/reference

pi Residual 412090 412075 412076

0.56 0.67 0.50 0.58

1.0 1.0 1.0 1.0 1.0

0.60a 0.60a 0.60a 0.60a 0.71L, 0.56P, 0.44N

2.31 1.22 1.28 1.19 2.4Y2, 1.6Y3

0–0.2b 0.5–0.7c 0.2

1.0 1.0

0.6a 0.6–0.8

0.8–1.3c 1.2–1.5

HYDRUS HYDRUS HYDRUS HYDRUS Johnston et al. (1999) Vertessy and Bessard (1999) Ringrose-Voase and Cresswell (2000)

0.0 0.0 0.0 0.0 0–0.3c 0–0.1

1.0 1.0 1.0 1.0 1.0 1.0

0.5a 0.5a 0.5a 0.5a 0.5a 0.5–0.7

2.8 2.1 2.4 1.5 1.0–2.6c 1.0–1.5

Adopted mi Residual 412090 412075 412076 Adopted a b c

HYDRUS HYDRUS HYDRUS HYDRUS Ringrose-Voase and Cresswell (2000)

L: Lucerne, P: Phalaris, N: Native, Y2/3-2/3 year crop rotation (from Table 2; Johnston et al., 1999). Assumed efficiency ratio for annual and perennial pasture. Obtained from Eqs. (1) and (2) of Vertessy and Bessard (1999); all HYDRUS results were obtained from this study. Simulated data from Ringrose-Voase and Cresswell (2000) averaged across 20 soil types and three climate zones.

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

salt export from the Mandagery Creek catchment at Eugowra (412030) were estimated as 55280 ML yr21 and 23530 t yr21, respectively. By replacing annual pasture with more water use efficient perennial pasture and trees for different landuse change scenarios, reduced sub-catchment mean annual leakage rates are obtained as shown in Fig. 8d. Simulations show that at least 30% tree cover is required to bring leakage rates for current landuse from 12 – 35 mm yr21 to a new level of 9 – 27 mm yr21 for different sub-catchments (Scenario 2.2). The corresponding sub-catchment leakage rates for 40% tree cover are in the range 8– 24 mm yr21 (Scenario 2.3). The respective surface runoff reductions for different scenarios are shown in Fig. 8e (Scenario 2.2: 13– 22%; Scenario 2.3: 22– 29%). Long term gains in reducing leakage rates and hence stabilising the groundwater discharge are somewhat offset by the reduced catchment yield. Overall, this is still a better option that will mitigate the rate of expansion of the saline discharge areas from its current extent of about 9 km2. The respective salt load reductions with surface runoff for different scenarios are shown in Fig. 8f (Scenario 2.2: 32 –70%; Scenario 2.3: 46– 71%). The reductions in groundwater flow and salts in groundwater flow are shown in Fig. 8g and h, respectively. A very small proportion of the area is currently saline (9 km2 or 0.5%) with typical soil salinity in the range 25 – 44 dS m21 (16 –28 g kg21). Soil salinity for non-saline areas in the catchment typically varies in the range 0 –0.7 dS m21. The reductions in salinity from landuse change scenarios are likely to be highest in the Manildra (412075) and Boree Creek (412076) sub-catchments. No appreciable reductions in salinity are expected from landuse change scenarios in the residual catchment. This is because the bulk of recharge and runoff generation areas are located in upland areas most influenced by landuse changes. For the complete Mandagery Creek catchment at Eugowra (412030), estimated reductions from current mean annual salinity of 665 mS cm21 are 13, 84 and 164 mS cm 21, corresponding, respectively to the landuse change Scenarios 1 (cropping and annual pasture to perennial), 2.1 (20% tree cover) and 2.2 (30% tree cover). Therefore, increasing perennial content of the crops and pastures does not produce substantial benefits with regard to salinity

85

(Scenario 1). Tree plantation over and above 30% tree cover (Scenario 2.2) does not produce any significant additional benefit with regards to stream salinity but further reduces runoff from the catchment. Downstream impact assessment of the landuse change scenarios in the Lachlan River after the confluence of the Mandagery Creek indicate a 1% reduction in flow and 4% reduction in salt load (Scenario 2.2). Estimated reductions in salinity in the Lachlan River at Forbes are 2, 6 and 10 mS cm21, corresponding, respectively, to landuse change Scenarios 1 (cropping and annual pasture to perennial), 2.1 (20% tree cover) and 2.2 (30% tree cover). 4.4. Effect of landuse change on long-term groundwater response The primary concern of groundwater modelling with FLOWTUBE was to estimate the time taken for groundwater in Mandagery Creek to come to a new equilibrium under a changed landuse pattern, and therefore recharge regime. Three different FLOWTUBE simulations were made: the first considered the entire catchment, the second modelled only the eastern catchments (Boree Creek and the Bourimbla Creek); and the third considered individual hillslopes. Given that the upslope edge of these groups were all catchment divides, the upstream water contribution was assumed to be zero, and at the downstream end the head in the river or nearest confluence was used to control the flux. Over the surface diffuse recharge was considered to be 1 mm yr21 for native pre-clearing conditions, and a step change was made to ‘current’ landuse pattern, with the area weighted Qrech from Tables 2 and 3 equal to 25 mm yr21. As such, the time estimates from FLOWTUBE must be the shortest possible, since broad-scale landuse change may occur over decades, thus prolonging the change to a new recharge regime. To estimate the initial groundwater head distribution, the catchment of interest was analysed with all heads at the ground surface and 1 mm yr21 of diffuse recharge, and allowed to equilibrate over 1000 years. The groundwater catchment boundaries were assumed to be the topographic catchment boundaries, for direct comparison with CATSALT results, and initial estimates of hydraulic conductivity (1 m d21) and available porosity (0.01) were obtained from Evans (2001).

86

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

The gauging station at Eugowra (412030) coincides with a major groundwater feature. To the north and south are forested hills on Upper Devonian Sediments. These hills have very low hydraulic conductivity and act as a barrier to groundwater flow, thus producing a bottleneck along the creek at this point. The pre-clearing simulation of the whole Mandagery Creek catchment indicated shallow water levels and groundwater discharge along an 8 km stretch of the creek upslope of the bottleneck; this is downstream of the confluence of the Mandagery and the Bourimbla Creeks (see Fig. 3). The simulated step change in recharge to 25 mm yr21 over the catchment, extended shallow water levels and groundwater discharge downstream of the pre-clearing discharge location to the topographic catchment outlet at Eugowra (412030), and 18 km further upstream of Manildra (412075), and also into Boree Creek (412090) and Bourimbla Creek (412076) (see Fig. 3). Most of the head changes resulting in discharge occurred within 20 years, but some storage changes in the upper catchment were increasing heads locally for up to 60 years. Modelling a hillslope in the Bourimbla Creek (412076) we can see how far the shallow water levels have extended upslope under increased recharge. Under pre-clearing conditions there was no simulated surface discharge, instead the aquifer carried the recharge directly to the stream. With the increase to current recharge levels the local aquifer running orthogonal to the stream was capable of transmitting this amount, so there was no expansion of shallow water levels there. This short 3.6 km hillslope that exhibited no surface discharge equilibrated to a new head distribution in 15 years. Finally, the eastern catchments Boree Creek (412076) the Bourimbla Creek (412090) and were modelled with FLOWTUBE. These were selected to run together because of the uncertainty from the SMAR recharge modelling as to whether they are actually independent below ground. The initial condition simulation with 1 mm yr21 recharge produce a very small area of shallow water levels at the confluence of the streams; this area is currently mapped as having shallow water levels. When increased recharge was applied the discharge zone extended upstream by 8 km (see Fig. 3), the same as the whole catchment model, with most head changes

taking place in 15 –20 years, but with storage changes occurring up to 45 years.

5. Conclusion An integrated framework for the assessment of water and salt balance for large catchments is presented. The framework was implemented for the Mandagery Creek catchment. Understanding of the salinity processes operating in the Mandagery Creek catchment was conceptualised using reasonably sound descriptions of geology and soils and was supported by modelling and expert knowledge. The framework comprised three models each operating on a different scale. Sub-catchment scale water and salt balances were first established using a quasi-physical semi-distributed model CATSALT. Hillslope scale effects on water balance were examined using HYDRUS-2D for each slope class, soil type and landuse combination. These were assembled under a GIS framework and then used to estimate the effects of landuse change and passed to CATSALT to estimate the effects on water and salt fluxes out of the catchment. A groundwater model FLOWTUBE was used to estimate the long-term effects of landuse change on groundwater discharge. The study is unique in that a process model is implemented at the hillslope scale to create useful information for the subcatchment scale model. Some new algorithms that enable effective transfer of information across the scale warrant consideration. Mean annual leakage rates under current landuse for the sub-catchments at Manildra (412075), Boree Creek (412090), Bourimbla Creek (412076) and the residual catchment were estimated as 29.7, 47, 15.9 and 5 mm yr21, respectively. The corresponding total runoff from the sub-catchments were 26590, 17442, 8617 and 2629 ML yr21, respectively, and salt exports were 13670, 4690, 3520 and 1660 t yr21, respectively. The breakup of surface and groundwater runoff and salt load differed between the subcatchments. Investigation of various landuse change scenarios indicates that changing annual pastures and cropping areas to perennial pastures is not likely to result in substantial improvement of water quality of

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

the Mandagery Creek. Landuse changes of 25– 30% tree-cover would be needed in the upland subcatchments to decrease mean annual stream salinity by 140 –170 mS cm21 (21 – 25%) from the current level. Maximum benefits in water quality improvement from landuse changes are likely to be from the sub-catchments 412075 and 412076. Estimated stream salinity reductions in the main Lachlan River downstream of its confluence with the Mandagery Creek are about 10 mS cm21, corresponding to the 30% tree cover scenario. The FLOWTUBE modelling within Mandagery Creek catchment indicated that discharge areas under filling conditions, i.e. increased recharge, could re-equilibrate in around 20 years for a catchment, and around 15 years for individual hillslopes. The small mapped area of saline land, together with the hillslope modelling, indicates that shallow water levels are found locally adjacent to the discharging stream reaches, rather than extensively below large areas of land. If the area of saline land is linearly related to the length of discharging stream, then under pre-clearing conditions, it is likely that around 1 km2 of land within Mandagery Creek had shallow water levels. Depending on the behaviour of the native vegetation, this land may have been only seasonally waterlogged and not developing surface salinity.

Appendix A. Soil landscapes in the Mandagery catchment. Geology

Slope class

Upper Devonian sediments

Steep

Upper to mid Lower Floodplains Gregra and Crests/upper Cudal Group volcanics

Mid

Lower Acknowledgements Rodger Grayson, Frantisek Dolezal and Glen Walker reviewed the CATSALT modelling framework. Dugald Black and Greg Bowman reviewed this manuscript at DIPNR. The work was done under the supervision of Ross Williams, Dugald Black and Peter Barker. The authors gratefully acknowledge valuable contributions of the following DIPNR staff: Allan Nicholson, Andrew Wooldridge, Gregory Summerell, Natasha Herron and Rachel Gilmore. Contributions of many staff members from the Central West Region of DIPNR to this study are acknowledged.The authors wish to acknowledge the NSW Department of Infrastructure, Planning and Natural Resources and the New South Wales Government for providing financial support for this work under the NSW Salinity Strategy.

87

Floodplains Ordovician Upper sediments Mid

Lower

Depressions

Soil classification according to Stace et al. (1968) and the links to Soil Taxonomy (USDA United States Department of Agriculture, 1975) are shown in brackets Lithosols (Lithic Xerorthents) Non-calcic brown soils (Ultic Palexeralfs) Yellow podzolic soils (Ultic Paleudalfs) Yellow solodic soils (Typic Natrudalfs) Red earths (Typic Haploxeralfs /Typic Rhodoxeralfs) and red podzolic soils (Typic Palexeralfs /Typic Rhodoxeralfs) Non-calcic brown soils and red podzolic soils (Typic Palexeralfs /Typic Rhodoxeralfs) Yellow podzolic soils (Typic Paleudalfs/ Paleustalfs) and yellow solodic soils (Typic Natrudalfs) Alluvial soils Euchrozems and noncalcic brown soils (Typic Rhodoxeralfs) Terra rossa soils (Typic Rhodoxeralfs) in association with limestones Non-calcic brown soils (Typic Palexeralfs / Typic Rhodoxeralfs) Yellow podzolic (Typic Paleudalfs /Paleustalfs) (continued on next page)

88

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

Table 4 (Continued)

Tertiary Basalts

Upper to mid Krasnozems (Eutrorthox /Oxic Rhodustalfs) Lower Yellow podzolic soils (Typic Paleudalfs/ Paleustalfs) Flat terraces Euchrozems (Typic Ustochrept)/dark clays (Vertisols) Dulladerry Steep/upper Shallow soils (Lithic Rhyolite Xerorthents) Mid Red podzolic soils (Ultic Palexeralfs) and red solodic (Typic Natrixeralfs) Lower Yellow solodic (Aquic Natrixeralfs) and red solodic soils (Typic Natrixeralfs); pockets of salinised soils (Aquic Natrustalfs with salic horizons) Depressions Yellow solodic soils (Aquic Natrixeralfs)

References Abbott, M.B., Bathurst, J.C., Cunge, J.A., O’Connell, P.E., Rasmussen, J., 1986. An introduction to the European hydrological system—Systeme Hydrologique Europeen SHE, 1. History and philosophy of a physically-based, distributed modeling system. J. Hydrol. 87, 45–59. Abraham, S.M., Abraham, N.A., 1996. Soil Data System—Site and Profile Information Handbook, second ed., NSW Department of Land and Water Conservation, ISBN 0 7310 5013 4. Barson, M.M., Randall, L.A., Bordas, V., 2000. Land Cover Change in Australia. Results of the Collaborative Bureau of Rural Sciences—State Agencies’ Project on Remote Sensing of Land Cover Change, Bureau of Rural Sciences, Canberra. Beale, G.T.H., Beecham, R., Harris, K., O’Neill, D., Schroo, H., Tuteja, N.K., Williams, R.M., 2000. Salinity predictions for NSW rivers within the Murray-Darling Basin, Centre for Natural Resources, NSW Department of Land and Water Conservation, ISBN 0 7347 51117. Belmans, C., Wesseling, J.G., Feddes, R.A., 1983. Simulation Model of the Water Balance of a Cropped Soil: SWATRE. J. Hydrol. 63, 271–286. Beven, K.J., Kirkby, M.J., 1979. A physically-based variable contributing area model of basin hydrology. Hydrol. Sci. Bull. 24 (1–3), 43 –69.

Blackburn, G., McLeod, S., 1983. Salinity of Atmospheric Precipitation in the Murray-Darling Drainage Division, Australia. Aust. J. Soil Res. 21, 411 –434. Blo¨schl, G., Sivapalan, M., 1995. Scale issues in hydrological modelling: a review. Hydrol. Process. 9, 251 –290. Bosch, J.M., Hewlett, J.D., 1982. A review of catchment experiments to determine the effects of vegetation changes on water yield and evapotranspiration. J. Hydrol. 55, 3 –23. Burnash, R.J.C., Ferral, R.L., 1972. Generalized hydrologic modeling, a key to drought analysis. Second International Symposium in Hydrology, Fort Collins, Colorado. Chapman, G.A., Atkinson, G., 2000. Soil survey and mapping. In: Charman, P.E.V., Murphy, B.W. (Eds.), Soils—Their Properties and Management, Oxford University Press, Melbourne. Cornish, P.M., 1993. The effects of logging and forest regeneration on water yields in a moist eucalypt forest in New South Wales, Australia. J. Hydrology 150, 301 –322. Costa-Cabral, M., Burges, S.J., 1994. Digital elevation model networks (DEMON): a model for flow over hillslope for computation of contributing and dispersal areas. Water Resour. Res. 30 (6), 1681– 1692. Dawes, W., Hatton, T., 1993. TOPOG_IRM 1. Model Description, Technical Memorandum 93/5, CSIRO Division of Water Resources, Australia. Dawes, W.R., Stauffacher, M., Walker, G.R., 2000. Calibration and modeling of groundwater processes in the Liverpool Plains. CSIRO Land and Water Technical Report 5/, Canberra, Australia, 41 pp. Dawes, W.R., Walker, G.R., Stauffacher, M., 1997. Model building: process and practicality. In: McDonald, A., McAleer, M. (Eds.), Proceedings of MODSIM97, 8–11 December, Hobart, Tasmania, vol. 1., pp. 317 –322. Dawes, W.R., Walker, G.R., Stauffacher, M., 2001. Practical modeling for management in data-limited catchments. Math. Comput. Model. 33, 625 –633. DLWC, 2000. The New South Wales Salinity Strategy—Taking on the challenge, NSW Department of Land and Water Conservation, ISBN 0 7347 5146 X. Dooge, J.C.I., 1986. Looking for hydrologic laws. Water Resour. Res. 22, 46 –58. Evans, R., 2001. Mandagery catchment conceptual groundwater flow model, unpublished report. Feddes, R.A., Kowalik, P.J., Zaradny, H., 1978. Simualtion of Field Water Use and Crop Yield, Wiley, New York. Franchini, M., Pacciani, M., 1991. Comparative analysis of several conceptual rainfall-runoff models. J. Hydrol. 122, 161–219. Freeze, R.A., Cherry, J.A., 1979. Groundwater, Prentice Hall, Englewood Cliffs, ISBN 0-13-365312-9. Gates, T.K., Burkhalter, J.P., Labadie, J.W., Valliant, J.C., Broner, I., 2002. Monitoring and modeling flow and salt transport in a salinity-threatened irrigated valley. J. Irrigation Drainage Engng, ASCE 128 (2), 87– 99. Geeves, G.W., Cresswell, H.P., Murphy, B.W., Gessler, P.E., Chartres, C.J., Little, I.P., Bowman, G.M., 1995. The physical, chemical, morphological properties of soils in the wheat belt of southern NSW and northern Victoria. NSW Department of

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90 Conservation and Land Management/CSIRO Aust. Division of Soils Occasional report. Gessler, P.E., Moore, I.D., McKenzie, N.J., Ryan, P.J., 1995. Soil landscape modeling and spatial prediction of soil attributes. Int. J. Geographical Information Syst.. Grayson, R.B., Western, A.W., Chiew, H.S., 1997. Preferred states in spatial soil moisture patterns: local and nonlocal controls. Water Resour. Res. 33 (12), 2897–2908. Holmes, J.W., Sinclair, J.A., 1986. Water yield from some afforested catchments in Victoria. Paper presented at Hydrology and Water Resources Symposium, Institute of Engineering, Brisbane, Queensland, Australia, November 25 –27 1986. Horton, R.E., 1919. Rainfall interception. Monthly Weather Rev. 47, 603 –623. Jeffrey, S.J., Carter, J.O., Moodie, K.B., Beswick, A.R., 2001. Using spatial interpolation to construct a comprehensive archive of Australian climate data. Environ. Model. Software 16, 309–330. Johnston, W.H., Clifton, C.A., Cole, I.A., Koen, T.B., Mitchell, M.L., Waterhouse, D.B., 1999. Low input grasses useful in limiting environments. Aust. J. Agric. Res. 50, 29–53. Kachroo, R.K., 1992. River flow forecasting. Part 5. Applications of a conceptual model. J. Hydrol. 133, 141–178. Kachroo, R.K., Liang, G.C., 1992. River flow forecasting. Part 2 Algebraic development of linear modelling techniques. J. Hydrol. 133, 17–40. Khan, H., 1986. Conceptual modelling of rainfall-runoff systems, M.Eng. Thesis, National University of Ireland. King, D.P., 1998. Soil Landscapes of the Forbes 1:250 000 Sheet, NSW Department of Land and Water Conservation, Sydney. Kovac, M., Lawrie, J.A., Murphy, B.W., 1989. Soil Landscapes of the Bathurst 1:250 000 Sheet, Soil Conservation Service of NSW, NSW Department of Land and Water Conservation, Sydney. Lea, N.L., 1992. An aspect driven kinematic routing algorithm. In: Parsons, A.J., Abrahams, A.D. (Eds.), Overland Flow: Hydraulics and Erosion Mechanics, Chapman and Hall, New York. Littleboy, M., Silburn, D.M., Freebairn, D.M., Woodruff, D.R., Hammer, G.L., Leslie, J.K., 1992. Impact of erosion on production in cropping systems. I. Development and validation of a simulation model. Aust. J. Soil Res. 30, 757–774. McCord, J.T., 1991. Application of second-type boundaries in unsaturated flow modeling. Water Resour. Res. 27 (12), 3257–3260. McCown, R.L., Hammer, G.L., Hargraves, J.N.G., Holzworth, D.L., Freebairn, D.M., 1996. APSIM—a novel software system for model development, model testing, simulation in agricultural systems research. Agric. Syst. 50, 255 –271. McDonald, M.G., Harbaugh, A.W., 1988. A Modular ThreeDimensional Finite-Difference Ground-Water Flow Model, Techniques of Water Resources Investigations of the United States Geological Survey, USGS, Denver, CO, Chapter A1, Book 6. McDonald, R.C., Isbell, R.F., Speight, J.G., Walker, J., Hopkins, M.S., 1990. Australian Soil and Land Survey Field Handbook, second ed., Inkata Press, Melbourne.

89

MDBMC, 2001. Basin Salinity Management Strategy 2001–2015. Murray-Darling Basin Ministerial Council. Ref. No I&D 6719. August 2001. ISBN 1 876830 17 4. (http://www.mdbc. gov.au). Moore, I.D., Grayson, R.B., Ladson, A.R., 1991. Digital terrain modeling: a review of hydrological, geomorphological, biological applications. Hydrol. Processess 5 (1), 3–30. Nash, J.E., Sutcliffe, J., 1970. River flow forecasting through conceptual models. Part 1, A discussion of principles. J. Hydrol. 10, 282–290. Nash, J.E., 1960. A unit hydrograph study with particular reference to British catchments. Proc. Inst. Civ. Eng. 17, 249–282. O’Callaghan, J.F., Mark, D.M., 1984. The extraction of drainage networks from digital elevation data. Comput. Vision, Graph. Image Process. 28, 328–344. O’Connell, P.E., Nash, J.E., Farrel, J.P., 1970. Riverflow forecasting through conceptual models, Part 2, The Brosna catchment at Ferbane. J. Hydrol. 10, 317 –329. O’Laughlin, E.M., 1986. Prediction of surface saturation zones in natural catchments by topographic analysis. Water Resour. Res. 22 (5), 794–804. Paydar, Z., Huth, N., Ringrose-Voase, A., Young, R., Bernardi, A., Keating, B., Cresswell, H., Holland, J., Daniels, I., 1999. Modeling Deep Drainage Under Different Land Use Systems. 1. Verification and Systems Comparison, Proceedings of MODSIM 99. International Congress on Modeling and Simulation, University of Waikato, New Zealand. 6-9 December 1999, vol. 1. The modeling and Simulation Society of Australia and New Zealand, ISBN 0-86422-948-8. Penman, H.L., 1963. Vegetation and Hydrology. Technical Communication No 53 Commonwealth Bureau of Soils Hardpenden England. Quinn, P.F., Beven, K.J., Lamb, R., 1995. The ln(a/tanb)Index: how to calculate it and how to use it within the topmodel framework. Hydrol. Process. 9, 161– 182. Quinn, P., Beven, K., Chevallier, P., Planchon, 1991. The prediction of hillslope flow paths for distributed hydrological modeling using digital terrain models. Hydrol. Process. 5, 59– 80. Ringrose-Voase, A., Cresswell, H., 2000. Measurement and prediction of deep drainage under current and alternative farming practice. Final Report to the Land and Water Resources Research and Development Corporation Project CDS16, CSIRO Land and Water. Roberts, D.W., Dowling, T.I., Walker, J., 1997. FLAG: a fuzzy landscape analysis gis method for dryland salinity assessment. CSIRO Land and Water, Technical Report No 8/97. Rodrı´guez-Iturbe, I., Gupta, V.K., 1983. Scale problems in hydrology. J. Hydrol. 65 (Special issue). Ruprecht, J.K., Sivapalan, M., 1991. Salinity modeling of experimental catchments. In Proceedings of the International Hydrology and Water Resources Symposium, Perth. Australia 2 –4 October 1991. Scanlon, T.M., Raffensperger, J.P., Hornberger, G.M., Clapp, R.B., 2000. Shallow subsurface storm flow in a forested headwater catchment: observations and modeling using a modified TOPMODEL. Water Resour. Res. 36 (9), 2575–2586.

90

N.K. Tuteja et al. / Journal of Hydrology 283 (2003) 67–90

Schaap, M.G., Leij, J., 1998. Database-related accuracy and uncertainty of pedotransfer functions. Soil Sci. 163 (10), 765–779. Schwarzenbach, R.P., Gschwend, P.M., Imboden, D.M., 1993. Environmental Chemistry, Wiley, New York, ISBN 0-47183941-8. Sˇimunek, J., Sejna, M., van Genuchten, M.Th., 1999. HYDRUS2D/MESHGEN-2D, Simulating Water Flow and Solute Transport in Two-Dimensional Variably Saturated Media, Colorado School of Mines, USA. Soil and Land Information System, SALIS (database), NSW Department of Land and Water Conservation, Vendor: Oracle, daily updating, http://www.dlwc.nsw.gov.au/care/soil/salis/ index.html Stace, H.C.T., Hubble, G.D., Brewer, R., Northcote, K.H., Sleeman, J.R., Mulcahy, M.J., Hallsworth, E.G., 1968. A Handbook of Australian Soils, Rellim Technical Publications, Glenside, South Australia. Stewart, J.B., Smart, R.V., Barry, S.C., Veitch, S.M., 2001. 1996/97 Land Use of Australia—Final Report for Project BRR5. National Land and Water Resources Audit, Canberra. Tarboton, D.G., 1997. A new method for the determination of flow directions and upslope areas in grid digital elevation models. Water Resour. Res. 33, 309–319. Turner, K.M., 1991. Annual evapotranspiration of native vegetation in a Mediterranean-type climate. Water Resour. Bull. 27, 1–6. Tuteja, N.K., Cunnane, C., 1999. A quasi physical snowmelt runoff modeling system for small catchments. Hydrol. Processes. 13 (12/13), 1961–1975. Tuteja, N.K., Beale, G., Gilmore, R., Beecham, R., Woolley, D., Williams, M., 2000a. Current and predicted future salt loads and stream salinity for the Murrumbidgee river from dryland salinisation processes. Proceedings of the third International Hydrology and Water Resources Symposium, Perth, Australia. Tuteja, N.K., Beale, G.T.H., Summerell, G.K., Johnston, W.H., 2002. Development and validation of the catchment scale salt balance model CATSALT version 1. NSW Department of Land and Water Conservation, CNR 2001.031, ISBN 0 7347 5244 X. Tuteja, N.K., Rancic, A., Vaze, J., Johnston, W., Beale, G., 2000b. Modeling report in Johnston, W., Management of grazed key native grass communities in the Murray-Darling Basin’, NRMS

D7015, Final report submitted to Murray-Darling Basin Commission. USDA United States Department of Agriculture, 1975. Agricultural Handbook No 436, Washington DC, USA. van Genuchten, M.Th, 1980. A closed form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 48, 703–708. Vertessy, R.A., Bessard, Y., 1999. Anticipating the negative hydrologic effects of plantation expansion: results from a GISbased analysis of the Murrumbidgee Basin. In Proceedings of the Second Forest Erosion Workshop, Report 99/6, CRC for Catchment Hydrology, Australia. Vertessy, R.A., Hatton, T.J., Benyon, R.G., Dawes, W.R.J., 1996. Long term growth and water balance for a mountain ash (Eucalyptus Regnans) forest catchment subject to clear felling and regeneration. Tree Physiol. 16, 221 –232. Walker, J., Bullen, F., Williams, J., 1992. Hydroecological changes in the Murray Darling Basin. 1: The number of trees cleared over two centuries. J. Appl. Ecol. 30 (265), 73. Western, A.W., Blo¨schl, G., 1999. On the spatial scaling of soil moisture. J. Hydrol. 217, 203–224. Williams, J., Prebble, R.E., Williams, W.T., Hignett, C.T., 1983. The influence of texture, structure and clay minerology on the soil moisture characteristic. Aust. J. Soil Res. 21, 15– 32. Williamson, D., Gates, G., Robinson, G., Linke, G., Seker, M., Evans, W., 1997. Salt trends: historic trend in salt concentration and saltload of stream flow in the Murray-Darling Drainage Basin. Dryland Technical Report No. 1, Murray-Darling Basin Commission. Zhang, L., Dawes, W.R., Walker, G.R., 2001. Response of mean annual evapotranspiration to vegetation changes at a catchment scale. Water Res. Res. 37 (3), 701 –708. Zhang, L., Dawes, W.R., Hatton, T.J., Reece, P.H., Beale, G.T.H., Packer, I., 1999. Estimation of soil moisture and groundwater recharge using the TOPOG_IRM model. Water Resour. Res. 35 (1), 149– 161. Zheng, C., Wang, P.P., 1999. MT3DMS: a modular threedimensional multispecies transport model for simulation of advection, dispersion and chemical reactions of contaminants in groundwater systems; documentation and user’s guide. Contract Report SERDP-99-1, US Army Engineer Research and Development Center, Vicksburg, MS.