The impact of revegetation strategies.
Mechanisms of dryland salinity development
Dryland salinity is a major problem for the agricultural industry of the southwest of Western Australia. It was estimated to affect 9.4% of agricultural land in 1994 and was predicted to affect about 17% by 2011-2021, and about 32% when the new hydrological balance is achieved some time in the next century (Ferdowsian et al. 1996).
The increase in recharge caused by the clearing of perennial, deep-rooted, native vegetation, and its replacement with shallow-rooted, lower water-use, annual pastures and crops (McFarlane et al. 1993), has caused the widespread development of salt-affected land in the wheatbelt of Western Australia (Ferdowsian et al. 1996). The resulting rise in groundwater levels has also caused the salinisation of stream water (Peck 1978). This connection between land and stream salinisation, rising saline groundwater, and the clearing or destruction of native vegetation was first suggested by Wood (1924), and confirmed by subsequent research (Peck and Williamson 1987).
Wood (1924) described 2 aquifers, one in the shallow soil layer and the other in the underlying deeply weathered material (now known as the pallid zone, McArthur 1991), and believed that the destruction of the native vegetation and the consequent death of its root systems connected the 2 layers allowing fresh water to flow, down macropores, into the deeper, salty aquifer. George (1992) showed that this lower aquifer occurred in the saprolite (McArthur 1991), between the pallid zone and the fresh rock, not in the entire pallid zone as believed by Wood (1924), and concluded that this aquifer was the principal cause of dryland salinity.
Reversing, or halting, dryland salinity development
Several studies show that if the water use by the vegetation in a catchment is increased again after clearing, either by natural regeneration or revegetation, rising groundwater levels can be reversed. Borg et al. (1988) and Bari et al. (1996) have shown that groundwater levels will rise in response to the clearing of native vegetation, but they will recede if the native vegetation is allowed to regenerate. Bell et al. (1990) and Bari and Schofield (1992) showed that replacing the native vegetation with various reforestation strategies would also reduce groundwater levels. Bari and Schofield (1992) reported that extensive plantations of trees resulted in rapidly declining groundwater levels, but there was a slight rise in groundwater levels under strip planting. Schofield (1990) showed that for a number of experimental sites in the southwest of Western Australia, reforestation with 11% crown cover was required to maintain groundwater levels, and to reduce them at a rate of 200 mm/year would require a crown cover of 46%.
None of the successful strategies for controlling groundwater levels described above would allow for the continuation of conventional agricultural practices on the revegetated land, with some agricultural income forgone were they to be implemented. Engineering solutions such as deep drainage or pumping, both of which involve the difficulty of disposing of the saline water particularly in the subdued relief of much of the wheatbelt, may be part of the solution (McFarlane et al. 1993). It is clear, however, that the best way to reduce groundwater levels will be to increase the water use of the vegetation in the landscape.
The water use of the existing vegetation can be improved through rehabilitation and protection of the remnants of native vegetation (Hobbs 1994), and by changing the annual cropping and pasture cycle to favour a greater emphasis on cropping. For example, Nulsen and Baxter (1982) implied that a barley-lupin (Hordeum vulgare-Lupinus albus) rotation reduced annual recharge by 50% compared with a wheat-clover (Triticum aestivum-Trifolium subterraneum) rotation. However, the results of Nulsen and Baxter (1982) showed that, at their 2 sites, recharge under the higher water-use barley-lupin rotation was still 44 mm/year and 72 mm/year (27-28% of annual rainfall).
For higher water-use revegetation strategies to be successful, plant species will need to be both perennial and deep-rooted. This will be necessary in order to use, during summer, water stored from winter rainfall in the upper part of the pallid zone, as described by Johnson (1987), and in order to make use of summer rainfall. The current annum pastures and crops that have replaced the native vegetation will not perform either of these functions (Scott and Sudmeyer 1993). Deep-rooted trees would satisfy both of these criteria, as would deep-rooted perennial pastures such as lucerne (Medicago sativa). Other perennials such as phalaris (Phalaris aquatica) are shallow-rooted, although deeper than the current annual pastures. They behave more like extended-season annual plants (Scott and Sudmeyer 1993), since if there is no summer rainfall they will become dormant and are unable, because of their relatively shallow roots, to access the deep groundwater.
One farming system where partial revegetation is compatible with traditional agriculture is alley farming in which crops and pastures are grown between parallel belts of trees (Lefroy and Scott 1994). Alley farming has other advantages in addition to its higher water use: it provides shelter which will substantially reduce wind erosion and can lead to increased pasture, livestock, and crop production (Bird et al. 1992).
Lefroy (1994) stated that there had been very little research into alley farming systems and their effects, despite their acceptance and implementation by farmers, at least at a paddock scale, if not at a catchment or landscape scale. Lefroy and Scott (1994) described several alley farming systems in various parts of the south-west of Western Australia which have successfully lowered local groundwater levels, but also reported that competition with inter-row pastures is apparent with some tree species. An alternative to an alley farming system is perennial pasture planted over the entire cleared area, or it could be combined with alley farming.
Clearly, a wide range of revegetation strategies exist, but their efficacy for controlling salinity remains largely untested and it is not practical to test the variety of combinations of revegetation strategies. Computer modelling does, however, have the potential to undertake such testing.
Computer modelling of dryland salinity occurrence and rehabilitation
Dawes and Hatton (1991) concluded that despite the disadvantages of modelling there was perhaps no other way than a physically based, distributed-parameter catchment model `to describe and predict the complex interactions between land, vegetation and climate'. This is principally because of the absence of studies into revegetation strategies that include agriculture, the long lead time for trees to reach maturity, the many possible combinations of revegetation strategies, and because the aquifer that is its principal cause is at the base of a deep regolith (George 1992). Whereas the behaviour of the deep aquifer can only be studied using usually widely spaced, deep piezometers, computer models can interpolate between piezometers and can be used, with the appropriate validation and precautions, where there are no piezometers. Models can also substitute for time-series data where none are available.
The characteristics required of a computer program to model where, and how much, saline seepage will occur, and the effects that various revegetation treatments will have on these parameters, are determined by the mechanisms that control seepage which have been discussed above. This discussion can be reduced to a few key features of the processes involved since any model will itself only be simulating a few key processes.
Since increased recharge to the deep aquifer, and the consequent rise in its piezometric surface, is the principal cause of the salinity problem in south-western Australia, the model will require a layer representing this aquifer in addition to one representing the soil layer. Recharge can occur anywhere that is not actively discharging and is widespread in the Western Australian landscape (Engel et al. 1989; McFarlane et al. 1989; George et al. 1991; Salama et al. 1991). It is not controlled by differences in permeability in the soil layer. As has been discussed above, the most appropriate way to reduce the recharge to the deep aquifer is to replace some of the existing plants with higher water-use vegetation. The ideal model would therefore need to simulate vegetation water use in a realistic, physically based, and computationally efficient way.
There are many different programs in use in Australia to model salinity risk and the effect of revegetation strategies and it is not appropriate to review all of them here. Many of the models commonly used are single-layer (essentially the soil layer) models which exclude interaction with groundwater below this layer, for example, TOPOGIRM (Dawes and Hatton 1993) and THALES (Grayson et al. 1992). The single-layer nature of such models reduces their suitability for investigation of the Western Australian wheatbelt. In addition to being single layer, these models use topography to control water flow and incorporate permeability variation in their single soil layer.
One commonly used model which can account for more than one layer is the 3-dimensional, finite difference, groundwater model MODFLOW described by MacDonald and Harbaugh (1988). MODFLOW requires the recharge to be estimated externally and provided as an input. This is usually done by estimating, or measuring, the permeability of the soil. Discharge from the watertable is estimated within the program provided that various parameters concerning capillary upflow are given as inputs. It is therefore difficult to account for land, crop, and water-management practices which determine vertical fluxes to the watertable (Prathapar et al. 1994). Also, MODFLOW does not reproduce the water-use mechanisms of trees (Bolger 1991), so they need to be treated as drains or groundwater pumps within the model and the water extracted by these drains or pumps needs to be converted, externally to the program, into the equivalent evapotranspiration by trees. This method of modelling vegetation water use is neither physically based not computationally efficient.
Richardson and Narayan (1995) described an example of the use of MODFLOW as a tool to assess the effectiveness of typical South Australian treatments for dryland salinity control at Wanilla in South Australia, which has a regolith similar to the wheatbelt of Western Australia. The model was set up as a single-layer, unconfined (non-linear) aquifer and recharge was provided to the model on the basis of soil permeability which is the equivalent of allowing spatial variability in a shallow layer. Six zones of different permeability were imposed on the single layer modelled (equivalent to the pallid zone aquifer), based on slug and pump tests of boreholes. However, the values actually used in the model were varied from the measured values by up to 50% as part of the calibration process, and location of the zones was controlled by the pattern of the values.
Thus, although MODFLOW can simulate both the soil and lower aquifer layers necessary to model satisfactorily the Western Australian wheatbelt, it has a number of disadvantages. These include necessity to calculate a number of parameters externally to the program, its non-physically based simulation of vegetation, the fact that it may not converge unambiguously, and its computational and data storage intensive nature.
A program that has some useful properties is the Western Australian Water and Rivers Commission's (WRC) in-house, groundwater hydrology, modelling program, MAGIC. This is a physically based, distributed parameter, mathematical model, which the WRC has used to design and subsequently implement extensive and expensive revegetation strategies to reduce salinisation of several major water-supply catchments (Arumugasamy and Mauger 1994; Mauger 1994a, 1996a; Anon. 1996). Details of the operation of the MAGIC program are given in Mauger (1994b, 1996b), and only a summary of the most important points will be repeated in this section and the next.
It is a 2-layer, cellular, steady-state model with lateral flows assumed to be driven by a hydraulic gradient not greater in magnitude and direction than the topographic gradient. Groundwater flow within the model is Darcyan. The model simulates water use by remnant native vegetation, annual pasture, revegetation with trees, shallow-rooted perennial pasture such as phalaris, and deep-rooted perennial pasture such as lucerne (cf. Background to the MAGIC model in Methods). Running the model for 3 annual cycles usually gives results that have adequately converged, results at the end of the third annual cycle being similar to those at the end of the second. This simulation process represents a steady-state analysis, under average rainfall, with vegetation conditions fixed.
MAGIC involves some simplifications in the modelling process compared with MODFLOW but it gives similar results (Mauger 1996b; see also Model calibration in Results). As a finite-difference model, MODFLOW works by assuming the pressure heads and computing the corresponding groundwater flows. It then iteratively checks the continuity of flow and corrects the pressure-head differences to minimise discrepancies in the water balance. In MODFLOW, convergence to a minimum discrepancy in the water balance is often difficult to achieve; after reducing to a minimum (not necessarily sufficiently small) the discrepancy can then increase again if the number of iterations is increased. The model is also computationally intensive and requires large data storage facilities.
MAGIC is like a finite-difference model in that it assumes the pressure heads and calculates the flows, but it does not reiterate the calculation of the pressure heads. The assumption that hydraulic gradient will not exceed topographic slope is generally good for uniform layers, but can be inaccurate locally when ground surface gradients are very low, where there are mid-slope remnants of native vegetation, or where geological features reduce the uniformity of the layers (Mauger 1996b).
MAGIC is much less computationally intensive and handles the effect of vegetation in a simpler and physically based, more realistic way than MODFLOW, since MAGIC treats vegetation as transpirational demand on a cell not as a pump or a drain, as does MODFLOW. Thus, although MAGIC may not model the movement of groundwater in as detailed a way as MODFLOW, it is much faster and easier to operate, and gives an unambiguous result. Also, the input data required by MAGIC (Landsat Thematic Mapper, topography, rainfall, and evaporation data) are readily available. It was therefore decided to use MAGIC for the present study.
This paper describes the use of MAGIC as a tool to investigate the impact of a suite of revegetation strategies on the development of saline seepage from the deep saprolite aquifer in 3 adjacent catchments in the western wheatbelt of Western Australia. The second paper (Clarke et al. 1998) deals with the effect of the major fault that underlies 1 of the 3 catchments (Clarke et al. in press) on the impact of the modelled revegetation strategies on land and water salinisation.
The study area, which was described in detail by Clarke et al. (in press), consists of 3 contiguous, geomorphologically similar, third-order catchments (Strahler 1964) located on the divide between the Blackwood River and Collie River Catchments, some 185 km south-southeast of Perth, Western Australia (Fig. 1). The catchments are underlain mainly by granites and gneisses of Archaean age, with some later dolerite, microdiorite, and gabbro dykes. The crystalline basement is overlain by an approximately 10-40-m-thick layer of weathered material, which is clay rich because of the nature of the substrate (Clarke et al. in press), which in turn is overlain by about 1-2 m of sandy and/or gravelly soil, except in the valley bottoms where the soils are loams to clayey sands. All groundwater in the study area is saline (C. J. Clarke unpubl, data), as indeed is most of the groundwater in the wheatbelt; thus, all seepage of deep groundwater was assumed to be saline.
[Figure 1 ILLUSTRATION OMITTED]
Background to the MAGIC model
Contour, rainfall, evaporation, and Landsat Thematic Mapper data were imported into the program to create a MAGIC project for the study area using the techniques described by Mauger (1988). Spatial data inputs and outputs use the Intergraph graphics program MicroStation. Alphanumeric data are output as Microsoft DOS files. The principal outputs are the area and volume of seepage of groundwater from the deeper layer into the shallower layer, and the streamflow, as both maps and numerical data.
The model divides the study area into 25 m by 25 m square cells which have a 1.5-m-thick surface layer with a hydraulic conductivity 72 times higher than the 20-m-thick bottom layer. The thicknesses of the upper and lower layers correspond well with the thickness of the soil and the regolith in the study area (Clarke et al. in press; C. J. Clarke unpubl, data).
In the model, lateral flow of groundwater in cells is firstly based on a water-balance calculation in which the direction of flow from each cell is predetermined to be in accordance with surface topography. Flow may be distributed from a particular cell to [is less than or equal to] 5 downslope cells depending on the aspect and plan curvature of the topography at the site. Calculation of flows commences at the most upstream of cells and cascades downstream. This approximates to the direction of hydraulic gradients in the groundwater assuming that the layer being analysed is thin compared with topographic relief and that the relatively impermeable base of the layer generally conforms to surface topography.
The magnitude of lateral flow is calculated using one of 2 modes: a transient mode or a steady-state mode. The transient mode is usually used for simulating flows and water storage in the shallow layer of more permeable soil at the ground surface, through a series of monthly time steps. The steady-state mode is used to simulate average annual flow rates in the deep layer of low permeability pallid zone and saprolite above bedrock.
To simulate the surface soil layer in the transient mode, a volume of water stored in the cell is determined by algebraically adding inflows and outflows (excluding lateral outflows) to an initial stored water volume. Inflows consist of rainfall less interception, lateral inflows from upstream cells, and possibly discharge of groundwater from the lower soil layer. Outflows are evapotranspiration and possibly recharge to the lower layer. This initial volume of water in the cell is interpreted as a saturated depth which is limited to the depth of the soil layer. The lateral outflow is then calculated as the product of permeability, flow path width, depth of flow, and topographic gradient (Darcy's Law). After subtracting the lateral outflow from the total volume, any volume in excess of full saturation is assumed to become surface flow in the catchment. The final water volume stored in the cell becomes the initial volume for the next time step.
When simulating the deeper pallid zone and saprolite layer using the steady-state mode, the rate of lateral outflow equals the algebraic sum of other inflows and outflows, unless limited by the capacity for flow in the cell determined by Darcy's Law using the topographic gradient and transmissivity at the site. Inflows consist of lateral inflows from upstream cells, and possibly recharge of groundwater from the surface soil layer. Outflows are initially only evapotranspiration from deep-rooted vegetation. After calculation of the cell's flow capacity, if lateral inflows exceed that capacity then the balance is interpreted as discharge of groundwater from the cell to the surface soil layer at that cell. The recharge from the surface soil layer is limited to the difference between the cell's flow capacity and lateral inflows. If the lateral outflows from a cell are less than the cell's flow capacity then the deeper clay layer is less than fully saturated and/or the hydraulic gradient is less than the topographic gradient.
Transpiration may be from remnant native vegetation, annual pasture, revegetation with trees, shallow-rooted perennial pasture such as phalaris (Phalaris aquatica), and deep-rooted perennial pasture such as lucerne (Medicago sativa). The condition of the native vegetation was derived from a Landsat Thematic Mapper (TM) scene taken in January 1991.
Revegetation by trees is handled by expressing their density as a proportion of the natural density of native vegetation, termed the Greenness factor. When deep-rooted plants need to draw water from the bottom layer, the transpiration rate is reduced to 60%. This is based on the stomatal resistance of jarrah (Eucalyptus marginata) for summer and winter from Sharma (1984) and the Penman-Monteith equation (Monteith 1965), which were used to calculate the ratio stressed:unstressed evapotranspiration (0.61). For annual pasture, ratio peak leaf area index (LAI) is obtained by calibrating the modelled streamflow against the measured streamflow (see Table 1). This peak LAI is then modified with an annual growth cycle which sets it to its highest value in September and October, to zero in March and April, and with intermediate values for the other months. When the top layer of a cell dries, the transpiration is stopped. The modified LAI is then used to factor pan evaporation for the month. The LAI for shallow-rooted perennial pastures such as phalaris was obtained by scaling the results obtained by Scott and Sudmeyer (1993). This is maintained constant throughout the year, but transpiration is again stopped when the top layer becomes dry. Deep-rooted perennial pastures such as lucerne behave similarly to phalaris, except that they are allowed to access water in the bottom layer to a nominated depth. When the top layer becomes dry, water is drawn from the lower layer at 60% of the potential rate until the available water is used, when transpiration ceases until more water becomes available.
Table 1. MAGIC model outputs of catchment area, forest area, seepage area, seepage volume, and streamflow for the native vegetation as it was in the 1991 Landsat TM image (the base case)
Catchment Area Forest Cleared Seepage (ha) area area area (ha) (%) (ha) 1 383 46 88 128 2 382 72 81 124 3 187 66 65 47 Catchment Seepage area Seepage Streamflow (% cleared volume ([m.sup.3]/year) area) ([m.sup.3]/year) 1 38 129 046 167 153 2 40 131 050 219 496 3 39 50 819 75 086
The model requires the soil water content to be given for each cell at the beginning of the simulation year. Iteration is used to convert an initial assumption that all cells are saturated at the start of September into soil water contents that are almost the same at the end of the simulation year as at the beginning. After the first 12-month surface-layer run, simulation of the bottom layer gives an estimate of rates of discharge from the bottom layer into the surface layer (positive value), or the capacity of the bottom layer to accept recharge (negative value), at each cell. This is applied in the next 12-month surface-layer simulation as a monthly rate. If the water volume in the surface layer of a cell exceeds its capacity, the surplus is transferred to the surface as potential streamflow. If applicable, evaporation from a free water surface may be subtracted before accumulating the volume as streamflow. A second simulation of the bottom layer then refines the rates. A third cycle of top- and bottom-layer simulation usually satisfies the requirement that the soil water content is almost the same at the end of the simulation year as it is at the beginning: the model has adequately converged. The degree of convergence is estimated by examining seepage volume and streamflow for a catchment, since these elements of the water balance are integrated over the entire catchment, whereas the cell soil water content would need to be compared on a cell-by-cell basis. For the 3 catchments of the study area, relative differences in seepage volume at the end of the third iteration compared with those at the end of the second range from +0.2% to -0.9%, and for streamflow the range is from -2.8% to -6.7%.
The model has been calibrated against streamflow data collected for a number of years for 2 WRC experimental catchments just over the divide from the study area in the Collie River Basin. One catchment (Lemon) is the same size as the catchments in the study area and the other (Batalling) is 4-5 times larger than 2 of those catchments. The stream flow in these catchments was monitored with a V-notch weir with continuous recording of stage height and
frequent sampling of the water salinity (weekly-monthly). Both catchments were monitored for 6 years and were at, or approaching, hydrological equilibrium. The model predictions of deep groundwater seepage have been compared with piezometric data from the study area (Clarke et al. in press) and the catchment has been modelled in its pristine state to see if it is virtually free of saline seepage as would be expected. The model estimates of the area of salt-affected land also have been compared with those of Ferdowsian et al. (1996).
Background to revegetation strategies modelled
Several different revegetation strategies were modelled and compared with the base case (B) which represented the vegetation in the 1991 Landsat image on which the model of the catchments is based. Most of the remnant native vegetation has been grazed, which has lead to the dramatic deterioration of the understorey and has changed soil structure and composition, and introduced weeds have affected vegetation composition and regeneration (Saunders and Hobbs 1992). The remnant native vegetation is restored to its pristine condition in the model by applying the appropriate Greenness factor to the polygons classified as being forest within the model (not scattered trees). Lefroy et al. (1992, 1993) and Kubicki et al. (1993) describe many of the following treatment types including the conservation treatment, changing annual pasture to a perennial pasture, and alleys systems. The locations of blocks are imported into the model as polygons and linear belts of trees as line strings in MicroStation design files.
The block treatments were as follows: (i) replacing the remnant native vegetation with annual pasture (All AP); (ii) restoring the remnant vegetation to its pristine (pregrazed) condition (P); (iii) placing pristine native vegetation over the entire study area (All P); (iv) a conservation treatment that filled in re-entrants in blocks of remnant vegetation, and joined them together (C) occupying 14% of the cleared area of Catchment 1, 21% of Catchment 2, and 12% of Catchment 3; and (v) a treatment similar to that installed in the Maringee and Batalling WRC experimental catchments in the Collie River, which consists of block plantings in the lower parts of catchments (MB) occupying 22% of the cleared area of Catchment 1, 9% of Catchment 2, and 12% of Catchment 3.
Alleys are defined by 3 rows of trees 2 m apart with these groups of rows at various separations. MAGIC determines the Greenness of the trees in the rows and this is manually checked to see that the Greenness values required are within the limits possible for plantation trees in such climates. Treatments were 120-m spacing, 90-m spacing, 60-m spacing, and 30-m spacing.
The pasture treatments were as follows: (i) replacing the annual pasture with shallow-rooted perennials plants such as phalaris (SP); and (ii) replacing the annual pasture with deep-rooted perennial plants such as lucerne (DP).
All treatments were modelled with the remnant vegetation as it was in 1991 and in a pristine condition as follows: (i) 60-m alleys with deep-rooted perennial plants in mid-slope bays; and (ii) 60-m alleys with the conservation treatment.
For the base case representing the native vegetation in 1991 (B), the areas of saline seepage at the new equilibrium, generated directly from the MAGIC model, are shown in Fig. 1 and Table 1. The remnants of native vegetation are concentrated predominantly in the upper and to a lesser extent in the lower parts of the landscape. Most of the seepage is located in the main and tributary valleys and the adjacent slopes, but there are occurrences higher in the landscape located around convergence in landform and near breaks in slope. The results of the MAGIC modelling of the seepage area, seepage volume, and the streamflow for various treatments in the 3 catchments of the study area are shown in Appendix 1.
The MAGIC outputs for the remnant native vegetation in the condition it was in when the 1991 Landsat TM image was taken (the base case, B) are shown in Table 1. Catchments 1 and 2 are both about 380 ha in area, and are between 80% and 90% cleared for agriculture. Catchment 3 is smaller, and is about 65% cleared. For the purpose of determining the cleared area of the catchments, areas of scattered trees are included in the cleared area; however, the modelling process takes their hydrological effect into account. Modelling of the base case shows that, at the new equilibrium, about 40% of the cleared area of the catchments will become salt-affected; the seepage volume is about 130 000 [m.sup.3]/year for Catchments 1 and 2 and about 50 000 [m.sup.3]/year for Catchment 3, and the streamflow is about 170 000 [m.sup.3]/year for Catchment 1, 220 000 [m.sup.3]/year for Catchment 2, and 75 000 [m.sup.3]/year for Catchment 3 (Table 1).
The effects of the condition of the remnant native vegetation, and various block treatments, on seepage area, seepage volume, and streamflow are shown in Table 2. The effects of the treatments are similar for all 3 parameters. If the remnant native vegetation is returned to its pristine condition, this reduces the seepage area to 87% of the base case, the seepage volume to 84%, and the streamflow to 87%. If in addition to this change the annum pasture is replaced by pristine native vegetation (expected to be similar to the condition of the catchments before European settlement) the seepage area is reduced to 0.9% of the base case, the seepage volume to 5%, and streamflow to 31%. If the remnant native vegetation is allowed to continue its slow decline until it disappears, the area of salt-affected land will increase to about 134% of the base case, seepage volume to 134%, and streamflow to 137%.
Table 2. Effect of revegetation treatments on seepage area, seepage volume, and streamflow expressed as a percentage of the values of the parameters for the native vegetation as it was in the 1991 Landsat TM image (the base case)
Pr, remnant native vegetation pristine; B, base case; values are expressed as mean percentage of B [+ or -] s.e.
Treatment Seepage area Block treatments Pr 87 [+ or -] 2 Pr+annual pasture replaced with 0.9 [+ or -] 0.4 pristine native vegetation Remnant native vegetation 134 [+ or -] 13 replaced by annual pasture B+Conservation 86 [+ or -] 3 Pr+Conservation 75 [+ or -] 1 B+Maringee/Batalling 67 [+ or -] 7 Pr+Maringee/Batalling 54 [+ or -] 6 Alley treatments B+120-m-spaced alleys 73 [+ or -] 2 B+90-m-spaced alleys 68 [+ or -] 5 B+60-m-spaced alleys 53 [+ or -] 4 B+30-m-spaced alleys 37 [+ or -] 4 Pasture treatments Annual pasture replaced with 1.1 [+ or -] 0.5 deep-rooted perennial pasture Annual pasture replaced with 91 [+ or -] 1 shallow-rooted perennial pasture Combined treatment Pr+60-m-spaced alleys+deep-rooted 28 [+ or -] 2 perennial pasture in mid-slopes Treatment Seepage volume Block treatments Pr 84 [+ or -] 3 Pr+annual pasture replaced with 5 [+ or -] 1 pristine native vegetation Remnant native vegetation 134 [+ or -] 9 replaced by annual pasture B+Conservation 87 [+ or -] 4 Pr+Conservation 74 [+ or -] 3 B+Maringee/Batalling 72 [+ or -] 7 Pr+Maringee/Batalling 58 [+ or -] 6 Alley treatments B+120-m-spaced alleys 78 [+ or -] 1 B+90-m-spaced alleys 75 [+ or -] 4 B+60-m-spaced alleys 61 [+ or -] 4 B+30-m-spaced alleys 43 [+ or -] 5 Pasture treatments Annual pasture replaced with 4 [+ or -] 2 deep-rooted perennial pasture Annual pasture replaced with 57 [+ or -] 5 shallow-rooted perennial pasture Combined treatment Pr+60-m-spaced alleys+deep-rooted 33 [+ or -] 3 perennial pasture in mid-slopes Treatment Streamflow Block treatments Pr 87 [+ or -] 2 Pr+annual pasture replaced with 31 [+ or -] 2 pristine native vegetation Remnant native vegetation 112 [+ or -] 7 replaced by annual pasture B+Conservation 89 [+ or -] 3 Pr+Conservation 80 [+ or -] 3 B+Maringee/Batalling 75 [+ or -] 6 Pr+Maringee/Batalling 64 [+ or -] 5 Alley treatments B+120-m-spaced alleys 81 [+ or -] 2 B+90-m-spaced alleys 79 [+ or -] 5 B+60-m-spaced alleys 68 [+ or -] 5 B+30-m-spaced alleys 52 [+ or -] 6 Pasture treatments Annual pasture replaced with 14 [+ or -] 4 deep-rooted perennial pasture Annual pasture replaced with 48 [+ or -] 5 shallow-rooted perennial pasture Combined treatment Pr+60-m-spaced alleys+deep-rooted 44 [+ or -] 2 perennial pasture in mid-slopes
The conservation-type treatment occupies about 15% of the cleared area of the catchment but only reduces the seepage area to 86% of the base case (equivalent to a reduction of 4% of the cleared area of the catchment), the seepage volume to 87%, and the streamflow to 89%. The Maringee/Batalling-style treatment occupies about 14% of the cleared area of the catchments and reduces the seepage area to 67% of the base case (equivalent to a reduction of about 13% of the cleared area of the catchments), the seepage volume to 72%, and the streamflow to 75%. If these 2 treatments are combined with the remnant native vegetation in pristine condition there is a further reduction of 11-13% of the base case in seepage area, 13-14% in seepage volume, and 9-11% in streamflow.
The effect of the variously spaced alley plantings of trees, with the remnant native vegetation in the condition it was in January 1991, on seepage area, seepage volume, and streamflow is contained in Table 2. Simulating alleys with spacings between 120 m and 30 m reduces the seepage area to between about 73% and 37% of the base case, the seepage volume to between 78% and 43% of the base case, and the streamflow to between 81% and 52% of the base case. The results in Appendix 1 show that if the alley treatments are combined with the remnant native vegetation in pristine condition, this further reduces the seepage area by 10-14% of the base case, seepage volume by about 10%, and the streamflow by 3-6%.
The effect on seepage area, seepage volume, and streamflow of completely replacing the present annual pasture with deep- and shallow-rooted perennial pastures is shown in Table 2. The deep-rooted perennial pasture substantially reduces all 3 parameters, although the effect is smallest on streamflow, reduced to 14% of the base case. The shallow-rooted perennial pasture reduces the seepage area to 91% of the base case, the seepage volume to 57%, and the streamflow to 47%.
The effects of the treatment consisting of deep-rooted perennial pasture planted in mid-slope positions combined with 60-m-spaced alleys and the remnant native vegetation in pristine condition are shown in Table 2. The area covered by perennial pasture varied from 19% to 28% of the cleared area of a catchment. The treatments cut seepage area to 28% of the base case, the seepage volume to 33%, and the streamflow to 44%.
The program has been calibrated by modelling the streamflow and comparing it to the measured streamflow and deep groundwater discharge for 2 catchments in the region (Mauger 1996b). The rainfall in the study area from the Bureau of Meteorology isohyets is between 650 mm and 600 mm. The Lemon Catchment received 740 mm and Batalling 610 mm of rainfall for the period of calibration and they are both in a landscape similar to the study area. There is a difference between the gauged value and the modelled value of +11% and -14% for the streamflow (Table 3). For the deep groundwater discharge there was virtually no difference in one case and +4% in the other. Thus calibration against streamflow suggests that the model can be usefully applied to similar catchments such as those of the present study.
Table 3. A comparison of gauged and modelled deep groundwater discharge ([m.sup.3]/year) and streamflow (mm) using the MAGIC model for two catchments
Streamflow is given as mm per unit area of the catchment. Discharge is calculated from measured salt load of the stream at typical deep groundwater salinities for the gauged row and as an output of the model for the modelled row. From Mauger (1996b)
Lemon Catchment Batalling Catchment Discharge Flow Discharge Flow Gauged 68 156 83 233 185 44 Modelled 68 184 92 241 435 38
The study area has also been modelled in MODFLOW. The MODFLOW predictions of the location of the seepage are generally the same as those predicted by MAGIC; however, the seepage volume is 7% greater and the number of cells is 23% greater (Mauger 1996b).
The differences in area are due to 2 effects. Firstly, MAGIC reports discharge from a cell that cannot carry its inputs, whereas MODFLOW reports discharge from cells flowing into such a cell, thus displacing the discharge upslope. Secondly, the high pressure at the base of the primary discharge cell that drives water out of the ground in MODFLOW tends to spread discharge to adjacent cells, whereas MAGIC locates all the discharge at the primary discharge cell.
Throughflows in saturated areas were strongly correlated ([r.sup.2] = 0.87) between the 2 models and were close to the line of equality in a scatter diagram (Mauger 1996b). For throughflow in unsaturated areas, the MAGIC estimate is about 60% of the MODFLOW value ([r.sup.2] = 0.34). In such areas, MODFLOW may give a higher estimate of discharge because saturation is reached earlier when following a flow-line downslope, accounting for some of the 7% higher estimate of seepage volume in MODFLOW.
The differences in the absolute value between the MODFLOW and MAGIC estimates do not appear crucial given the objective of the modelling, which was to predict the relative impact of the revegetation treatments; all results in this paper are reported as percentages of the base case. Besides, MAGIC has been calibrated in 2 nearby catchments against measured streamflow and deep groundwater discharge. Its estimates are reasonably close to regional estimates (based on soil surveys) for the ultimate extent of land salinisation, 31% of the total area of the catchments compared with 36% (Ferdowsian et al. 1996). The MODFLOW estimate is 41%.
Returning the entire catchment to its pristine condition renders the seepage area very small, and reduces the seepage volume to 5% and the streamflow to about 30%. This is probably a reasonable response from catchments in this rainfall zone (between 650 mm and 600 mm per annum) and supports the validation of the model. The model of the study area shows that if the area is restored to its pristine condition, about 2% of the rainfall becomes runoff. In another catchment in a higher rainfall zone, Ruprecht and Schofield (1989) showed that before clearing of the Wights Catchment, 11% of the rainfall (~1060 mm) became streamflow. The modelling results are in agreement with these experimental results, particularly given the differences in the areas.
In the base case (80% of the catchments cleared), the model predicts that 8% of the rainfall (taken as 600 mm) will become streamflow, whereas across the divide in the Collie River records from the Batalling Catchment (49% cleared) show 7% of the rainfall (640 mm) became streamflow (Bari 1992). In the higher rainfall zone to the south (~950 mm for the period studied), Bari et al. (1996) showed that for the 4 years after clearing of the native forest, streamflow was ~20% of the rainfall, compared with ~10% for the 7 years before clearing. Compared with a control catchment, the post-clearing streamflow in the cleared catchment would have been equivalent to 23% of the rainfall. Comparison of the predicted occurrence of seepage from the bottom layer with the present piezometric surface and its rate of change (Clarke et al. in press) shows that the modelling results are consistent with the expected position of the piezometric surface when the new hydrological equilibrium is achieved.
In conclusion, although the model has not been calibrated on the study area, it has been closely calibrated within the region and its estimates compare favourably with other estimates and measurements. Since it is used to estimate the relative effects of various treatments, its use in the study area appears justified.
Modelling of revegetation strategies
If nothing is done to the catchments and the native vegetation remains as it was in 1991, then the model predicts that about 40% of the cleared area of the catchments will become saline because of groundwater discharge. As noted above, the magnitude of this prediction is similar to that estimated for the broader region by Ferdowsian et al. (1996). By contrast, this result can be compared with estimates of the extent of saline land in 1991 being 11% of the catchment for Catchment 1 and 1% for both Catchments 2 and 3 (Clarke et al. in press). Thus Catchment 1 is still far from equilibrium and in Catchments 2 and 3 salinisation has only recently begun. The predicted equilibrium state will have a dramatic economic impact on the farmers and the local communities and, if, as seems likely, the same phenomenon occurs throughout the wheatbelt, on the economy of the State. The seepage volume will be about 20 times greater than if the entire catchment were under pristine native vegetation, and the streamflow will be 3 times greater. Since the deep groundwater seepage will be saline (C. J. Clarke unpubl, data), this has significant consequences for streamflow salinisation.
Whereas the base case assumed that remnant native vegetation in the study area would retain its 1991 form and condition, remnant native vegetation in the agricultural landscape is subject to the degrading influences of tree decline, weed invasion, and grazing (Hobbs 1994). Unless this deterioration is halted or reversed, the remnants will die out. In this event, the seepage area will increase to about 134% of the base case, which will worsen the economic problems for the farmers. The seepage volume will be about 30 times what it would have been if the entire catchment were under pristine native vegetation, and the streamflow will be 4 times greater, thus also exacerbating the waterway salinisation and flood risk. However, if the condition of the native vegetation was allowed to improve by fencing, and possibly by reseeding or replanting the understorey to approach its pristine state, there would be a reduction in seepage area to 87% of the base case, and seepage volume and streamflow would decrease to about 15% of the base case.
The effects of restoring the native vegetation are large because of the relatively high proportion of remnants in this landscape, about 20-35% of the catchment area. This proportion is typical of the western wheatbelt, but is higher than the central and eastern wheatbelt, where the protection of the remnants is therefore likely to have a smaller impact on these parameters. For example, Hobbs et al. (1993) state that only about 7% of the native vegetation remains in the Kellerberrin area of the central wheatbelt.
The conservation treatment (which involved revegetation in re-entrants and between blocks) occupies about 15% of the cleared area of the catchments but will only reduce the seepage area by about 5% of the cleared land. However, when combined with the return of the remnant native vegetation to its pristine condition, the seepage area reduces to 75% of the base case (equivalent to a reduction in seepage area to 10% of the cleared area of the catchment). Since at least part of the strategy to protect the remnants requires this type of planting (Hobbs 1994), and the net effect of the loss of the remnants compared with their protection is about 20% of the cleared area, it could be argued, in order to persuade farmers to implement an otherwise apparently land-expensive treatment, that the overall net effect of the conservation planting, as part of a strategy to restore the native vegetation, is a gain of about 15% of the cleared area of the catchment. Although the demise of the remnant native vegetation will notionally increase the cleared area of the catchment, it is unlikely to have much impact on agricultural production since, in general, native vegetation was left on land considered to be of low productivity.
Block plantings in the valley bottoms and the lower valley slopes of the type used in the rehabilitation of the Wellington Dam Catchment to reduce the salt load entering a water-supply dam (Maringee/Batalling-type treatments) (Schofield et al. 1989) reduce the seepage area to 67% of the base case, and reduce the seepage volume and streamflow to 72% and 75%, respectively. At first sight, it would therefore appear that the present results justify the continued use of this sort of treatment to combat the salinisation of waterways. However, the mixed treatment of 60-m alleys combined with remnant native vegetation in pristine condition and deep-rooted perennial pasture in mid-slope bays (discussed further below) reduces the seepage area to 24-32% of the base case, and the seepage volume and streamflow to 27-36% and 40-49%, respectively. Under this treatment, only about 10% of the catchment is required for trees, compared with about 15% for the Maringee/Batalling-type treatment. The mixed treatment is thus both more effective and requires less agricultural land to be taken out of farming. Thus, the present modelling predictions suggest the possibility of alternative revegetation strategies for the water-supply dams of south-western Australia that are as effective as, or possibly more effective than, the Maringee/Batalling-style treatments, and are also more compatible with continued agricultural production.
Treatments with alleys at spacing from 120 m to 30 m reduce the seepage area to 73-37% of the base case, which still leaves 15% of the cleared catchment salt-affected, even under the densest treatment. In view of the report from Lefroy and Scott (1994) of several alley farming experiments that have successfully lowered groundwater levels, this at first sight seems surprising, and might suggest some failure in the model. However, the strip replantings referred to by Bari and Schofield (1992) failed to lower the groundwater levels, and Ruprecht and Schofield (1991) reported rates of groundwater rise under strip clearing in the Dons Catchment of the Collie River of 0.6-1 m/year. In addition, modelling of rows of trees by Bolger (1991) failed to show any significant reduction in groundwater levels. This evidence suggests that, although in some circumstances trees in alley systems may be able to eliminate land salinisation on their own, in other circumstances they will not. Given the present interest in promoting alley systems in south-western Australia (Lefroy and Scott 1994), the present modelling results suggest that a better understanding of their impact on groundwater systems in different landscapes is necessary before they can be recommended as a general revegetation strategy for salinity control.
One explanation for the paradox may be that many of the implemented alley systems are installed on a paddock scale only on the valley slopes, whereas the treatment imposed on the study area covers 3 entire catchments. Thus, although alley systems may be able to lower groundwater levels on valley slopes they may not be able to reduce them over the entire catchment, leaving the predicted residual area of salt-affected land in the valley bottoms. As discussed above, recharge is widespread throughout the landscape and can occur anywhere that is not actively discharging (Engel et al. 1989; McFarlane et al. 1989; George et al. 1991; Salama et al. 1991); thus, any revegetation system that leaves gaps in deep-rooted perennial vegetation (as do alley systems) will allow increased recharge and hence increased discharge. This view is supported by conclusions made by Richardson and Narayan (1995) and Pavelic et al. (1997) that in South Australia revegetation with lucerne needed to be broad scale, and that small-scale revegetation was not sufficient.
Replacing the current annum pasture with deep-rooted perennial pasture was the most promising revegetation treatment modelled to correct the problems of saline groundwater seepage, as is the case in North America (Halvorson and Reule 1980) and as is also predicted by MODFLOW in South Australia (Richardson and Narayan 1995). This is probably because a ubiquitous, deep-rooted, perennial system (the native vegetation) has been replaced with another (the perennial pasture). By comparison, the shallow-rooted perennial pasture option modelled, although a great improvement on annual pasture, was much less effective than the deep-rooted perennial pasture, probably because it behaves more like an extended-season annual pasture (Scott and Sudmeyer 1993).
However, deep-rooted perennial pasture such as lucerne poses some difficulties to farmers since it cannot be constantly stocked as the present annual pasture can be, and it may not be possible to establish and maintain in all parts of the landscape. In practice, blanket revegetation treatments of any sort are unlikely to be adopted by farmers for the prevention of saline seepage (Nulsen 1993). For reasons of personal preference, farmers will choose a variety of treatments and for economic reasons a mixture of revegetation treatments is preferable. Moreover, the species chosen for each revegetation treatment will have site- and soil-type requirements that limit their use in some parts of the landscape.
For this reason the combination treatment of 60-m spaced alleys (a compromise between achieving sufficient effect from the trees and giving the farmer as wide an alley as possible to farm) with deep-rooted perennial pasture planted in the mid-slopes was modelled. The mid-slope position was chosen to avoid water stress on the plants in summer on the upper slopes, and water logging (and salinity) in the lower slopes. The treatment was not as good as blanket treatments such as a deep-rooted perennial pasture plant being established instead of the annual pasture; however, it is probably a reasonable compromise between the needs of the farmer for a flexible management system, and the needs of the community to reduce land degradation, the salination of water supplies, and the consequent socio-economic losses to a minimum.
One treatment class that was not modelled in the present study was that of different crop rotations, which Nulsen and Baxter (1982) suggested could be designed to increase water use by taking advantage of the higher water use of crops such as lupins and barley compared with annum pastures. The effects of different cropping treatments and rotations as suggested by Nulsen and Baxter (1982) were not modelled since, as a steady-state model, they could only be put into MAGIC as a mean water use over an extended period. Although the maximum effect implied by Nulsen and Baxter (1982) of 50% higher water use is large, agronomic restrictions such as rotation flexibility and seasonal variation might make any likely change in the real world much smaller than this. However, if it is considered important in the future a soil water-balance model such as SWIM described by Smettem et al. (1994), which can model the effect of various pasture and crop rotations on soil water content, could be used to provide the water use of various rotations to the upper layer of the MAGIC model. Certainly, for use in the eastern wheatbelt of Western Australia where cropping is more dominant than in the region of the study area, the impact of crop rotations on saline groundwater discharge would need to be modelled.
The MAGIC model has been effectively calibrated for the study area, and since it has been used for the prediction of the relative outcomes of various treatments, it is concluded that its use is justified in the present study. The MAGIC model suggests that neither block planting nor alley planting of trees on their own will rid the catchments of salinised land or saline seepage and increased streamflow. The rehabilitation of remnants of the native vegetation to their pristine condition gives a significant improvement in all of these parameters and should be implemented in areas where there is a significant proportion of such vegetation, although as for block and alley plantings, it failed on its own to prevent completely land and stream salinisation. If all of the annual pasture is replaced by a deep-rooted perennial pasture or pristine native vegetation, seepage area and seepage volume are reduced to minimal values, whereas streamflow is less affected.
The MAGIC model suggests that the best compromise to minimise degradation and the disruption of the current agronomic system would be to incorporate a deep-rooted perennial pasture plant in mid-slope bays of an alley-farming system, where the remnants of native vegetation have been fenced and rehabilitated. Such a treatment would reduce the seepage area to about 10% of the cleared area of the catchment, the seepage volume to about 33%, and the streamflow to less than 50%.
The second of these 2 papers (Clarke et al. 1998) deals with the effect that the higher hydraulic conductivity within the major fault that underlies one of the catchments in the study area (Clarke et al. in press) has on the modelled impact of the revegetation strategies on land and water salinisation.
Except where stated otherwise, this research and interpretation was carried out by C.J.C. towards the degree of Doctor of Philosophy from Murdoch University, Western Australia, whilst receiving an Australian Postgraduate Award, and was supervised by R.W.B. and R.J.H., together with Dr Richard George of Agriculture Western Australia (AgWA) whom the authors thank for his involvement in the research effort. The research was, in part, funded by an Australian Research Council Collaborative Grant together with the Water and Rivers Commission (WRC) and AgWA.
The authors thank Alex Rogers of Jim Davies and Associates who undertook some of the computer processing of the vegetation treatments designed by C.J.C., in the model created by C.J.C. and G.W.M.; the calibration and validation modelling were carried out by G.W.M. The authors thank Greg Beeston (AgWA) for his help and contribution to the research effort and the local farmers for their cooperation. Sincere thanks also to Mic Andacich (AgWA) for creating the Fig. 1. The suggestions of 2 anonymous reviewers are also acknowledged.
Anon. (1996). Trees on farms to reduce salinity in the clearing control catchments. Volume 4: Helena Catchment. Water and Rivers Commission, Technical Series No. WRT8.
Arumugasamy, V., and Mauger, G. W. (1994). Trees on farms to reduce salinity in the clearing control catchments. Volume 2: Wellington Catchment. Water Authority of Western Australia, Report No. WS 148.
Bari, M. A. (1992). Early streamflow and salinity response to partial reforestation at Batalling Creek Catchment in the south-west of Western Australia. Water Authority of Western Australia, Report No. WS107.
Bari, M. A., and Schofield, N. J. (1992). Lowering of a shallow, saline water table by extensive eucalypt reforestation. Journal of Hydrology 133, 273-91.
Bari, M. A., Smith, N., Ruprecht, J. K., and Boyd, B. W. (1996). Changes in streamflow components following logging and regeneration in the southern forests of Western Australia. Hydrological Processes 10, 447-61.
Bell, R. W., Schofield, N. J., Loh, I. C., and Bari, M. A. (1990). Groundwater response to reforestation in the Darling Range of Western Australia. Journal of Hydrology 115, 297-317.
Bird, P. R., Bicknell, D., Bulman, P. A., Burke, S. J. A., Leys, J. F., Parker, J. N., Van Der Sommen, F. J., and Voller, P. (1992). The role of shelter in Australia for protecting soils, plants and livestock. Agroforestry Systems 20, 59-86.
Bolger, P. (1991). Modelling the effect of trees on water tables. In `Trees as a Tool for Resource Management in Irrigation Areas'. (Ed. A. Heuperman.) pp. 46-58. (Institute of Sustainable Agriculture: Tatura, Victoria).
Borg, H., Stoneman, G. L., and Ward, C. G. (1988). The effect of logging and regeneration on groundwater, streamflow and stream salinity in the southern forests of Western Australia. Journal of Hydrology 99, 253-70.
Clarke, C. J., George, R. J., Bell, R. W., and Hobbs, R. J. (in press). Major faults and the development of dryland salinity in the western wheatbelt of Western Australia. Hydrology and Earth System Sciences 2 (In press.)
Clarke, C. J., Mauger, G. W., and Bell, R. W. (1998). Computer modelling of the effect of revegetation strategies on salinity in the western wheatbelt of Western Australia. 2. The interaction between revegetation strategies and major fault zones. Australian Journal of Soil Research 36, 131-42.
Dawes, W., and Hatton, T. (1991). The impact of tree planting in the Murray-Darling Basin: the use of the TOPOGIRM hydrogeological model in targeting tree planting sites in catchments. CSIRO Division of Water Resources, Technical Memorandum No. 91/14.
Dawes, W., and Hatton, T. (1993). TOPOGIRM 1. Model description. CSIRO Division of Water Resources, Technical Memorandum No. 93/5.
Engel, R., McFarlane, D. J., and Street, G. (1989). Using geophysics to define recharge and discharge areas associated with saline seeps in south-western Australia. In `Groundwater Recharge'. (Ed. M. Sharma.) pp. 20-40. (Balkema Publishing: Rotterdam.)
Ferdowsian, R., George, R. J., Lewis, M. F., McFarlane, D. J., Short, R., and Speed, R. (1996). The extent of dryland salinity in Western Australia. In `Proceedings of the Productive Use of Saline Lands Conference'. pp. 89-97. (Promaco Conventions: Perth.)
George, R. J. (1992). Hydraulic properties of groundwater systems in the saprolite and sediments of the wheatbelt, Western Australia. Journal of Hydrology 130, 251-78.
George, R. J., McFarlane, D. J., and Lewis, M. F. (1991). A review of recharge and implications for management in saline agricultural catchments, Western Australia. In `Proceedings of the International Hydrology and Water Resources Symposium'. pp. 193-7.
Grayson, R. B., Moore, I. M., and McMahon, T. A. (1992). Physically based hydrological modeling. 1. A terrain-based model for investigative purposes. Water Resources Research 28, 2639-58.
Halvorsen, A. D., and Reule, C. A. (1980). Alfalfa to control saline seeps. Journal of the Soil Science Society of America 44, 370-4.
Hobbs, R. J. (1994). Fragmentation in the wheatbelt of Western Australia: landscape scale problems and solutions. In `Proceedings of the Third Annual Conference of the International Association for Landscape Ecology'. (Ed. J. Dover.) pp. 3-20. (AILE: UK.)
Hobbs, R. J., Saunders, D. A., Lobry de Bruyn, L. A., and Main, A. R. (1993). Changes in biota. In `Reintegrating Fragmented Landscapes Towards Sustainable Production and Nature Conservation'. (Eds R. J. Hobbs and D. A. Saunders.) pp. 65-106. (Springer-Verlag: New York.)
Johnson, C. D. (1987). Distribution of environmental chloride in relation to subsurface hydrology. In `Hydrology and Salinity in the Collie River Basin Western Australia'. (Eds A. J. Peck and D. R. Williamson.) Journal of Hydrology 94, 67-88.
Kubicki, A., Denby, C., Stevens, M., Haagensen, A., and Chatfield, J. (1993). Determining the long-term costs and benefits of alternative farm plans. In `Reintegrating Fragmented Landscapes Towards Sustainable Production and Nature Conservation'. (Eds R. J. Hobbs and D. A. Saunders.) pp. 245-78. (Springer-Verlag: New York.)
Lefroy, E. C. (1994). A tale of two alleys: parallel evolution in agroforestry. Agroforestry Today 6, 5-7.
Lefroy, T., and Scott, P. (1994). Alley farming: new vision for Western Australian farmland. Journal of Agriculture, Western Australia 35, 119-26.
Lefroy, E. C., Bicknell, D., Hobbs, R. J., Scheltema, M., and Bartle, J. (1992). Towards a revegetation strategy for the Western Australian wheatbelt. In `Proceedings of the Catchments of Green Conference'. pp. 43-57. (Greening Australia: Canberra.)
Lefroy, E. C., Salerian, J., and Hobbs, R. J. (1993). Integrating economic and ecological considerations: a theoretical framework. In `Reintegrating Fragmented Landscapes Towards Sustainable Production and Nature Conservation'. (Eds R. J. Hobbs and D. A. Saunders.) pp. 209-44. (Springer-Verlag: New York.)
MacDonald, M. G., and Harbaugh, A. W. (1988). A modular three-dimensional finite difference groundwater flow model. United Stares Geological Survey Open File, Report No. 83-875.
Mauger, G. W. (1988). Landsat evaluation of the impact of logging, burning and dieback in the jarrah forest of south Western Australia. Institution of Engineers, Australia. Transactions of Multi-disciplinary Engineering. GE12 No. 1, 24-9.
Mauger, G. W. (1994a). Trees on farms to reduce salinity in the clearing control catchments. Volume 1: Upper Denmark. Water Authority of Western Australia, Report No. WS 148.
Mauger, G. W. (1994b). Hydrologic assessment of vegetation in catchments affected by dryland salinity in southwestern Australia. In `Proceedings of Water Down Under 94'. Congress of the International Association of Hydrogeologists and International Hydrology and Water Resources Symposium, Adelaide. Vol. 2. pp. 379-500. (Institution of Engineers, Australia: Canberra.)
Mauger, G. W., (1996a). Trees on farms to reduce salinity in the clearing control catchments. Volume 3: Kent Catchment. Water and Rivers Commission, Technical Series No. WRT20.
Mauger, G. W. (1996b). Modelling dryland salinity with the M.A.G.I.C. system. Water and Rivers Commission, Technical Series No. WRT7.
McArthur, W. M. (1991). `Reference Soils of South-western Australia.' (Department of Agriculture Western Australia and the Australian Soil Science Society: Perth.)
McFarlane, D. J., Engel, R., and Ryder, A. T. (1989). The location of recharge areas responsible for valley salinity in the Lake Toolibin Catchment, Western Australia. In `Groundwater Recharge'. (Ed. M. Sharma.) pp. 255-68. (Balkema Publishing: Rotterdam.)
McFarlane, D. J., George, R. J., and Farrington, P. (1993). Changes in the hydrologic cycle. In `Reintegrating Fragmented Landscapes Towards Sustainable Production and Nature Conservation'. (Eds R. J. Hobbs and D. A. Saunders.) pp. 65-106. (Springer-Verlag: New York.)
Monteith, J. L. (1965). Evaporation and environment. Symposium of the Society of Experimental Biology 19, 205-34.
Nulsen, R. A. (1993). Opportunities and limitations for using agronomic techniques to control dryland salinity. In `Proceedings of the Land Management for Salinity Control Conference'. pp, 24-31. (Bendigo: Victoria.)
Nulsen, R. A., and Baxter, I. N. (1982). The potential of agronomic manipulation for controlling salinity in Western Australia. Journal of Australian Institute of Agricultural Science 48, 222-6.
Pavelic, P., Narayan, K. A., and Dillon, P. J. (1997). Groundwater flow modelling to assist dryland salinity management of a coastal plain of southern Australia. Australian Journal of Soil Research 35, 669-86.
Peck, A. J. (1978). Salinization of non-irrigated soils and associated streams: a review. Australian Journal of Soil Research 16, 157-68.
Peck, A. J., and Williamson, D. R. (1987). Effects of forest clearing on groundwater. In `Hydrology and Salinity in the Collie River Basin Western Australia'. (Eds A. J. Peck and D. R. Williamson.) Journal of Hydrology 94, 47-66.
Prathapar, S. A., Meyer, W. S., Jain, S., and van der Lelij, A. (1994). SWAGSIM: a soil water and groundwater simulation model. CSIRO Division of Water Resources, Divisional Report No. 94/3.
Richardson, S. B., and Narayan, K. A. (1995). The effectiveness of management options for dryland salinity control at Wanilla, South Australia. Agricultural Water Management 29, 63-83.
Ruprecht, J. K., and Schofield, N. J. (1989). Analysis of streamflow generation following deforestation in southwest Western Australia. Journal of Hydrology 105, 1-17. Ruprecht, J. K., and Schofield, N. J. (1991). Effects of partial deforestation on hydrology and salinity in high salt storage landscapes. II. Strip, soils and parkland clearing. Journal of Hydrology 129, 39-55.
Salama, R. B., Farrington, P., Bartle, G. A., and Watson, G. D. (1991). Identification of recharge and discharge areas in the wheatbelt of Western Australia using water level patterns in relation to basin geomorphology. In `Proceedings of the International Hydrology and Water Resources Symposium'. pp. 841-46. (Institution of Engineers, Australia: Canberra.)
Saunders, D. A., and Hobbs, R. J. (1992). Impact on biodiversity of changes in land use and climate. In `Biodiversity of mediterrenean ecosystems in Australia'. (Ed. R. J. Hobbs.) pp. 61-75. (Surrey Beatty: New South Wales.)
Schofield, N. J. (1990). Determining reforestation area and distribution for salinity control. Hydrological Sciences Journal 35, 1-19.
Schofield, N. J., Loh, I. C., Scott, P. R., Bartle, J. R., Ritson, P., Bell, R. W., Borg, H., Anson, B., and Moore, R. (1989). Vegetation strategies to reduce stream salinity in water resource catchments in southwest Western Australia. Water Authority of Western Australia, Report No. WS 33.
Scott, P. R., and Sudmeyer, R. A. (1993). Evapotranspiration from agricultural plant communities in the high rainfall zone of the southwest of Western Australia. Journal of Hydrology 146, 301-19.
Sharma, M. L. (1984). Evapotranspiration from a eucalyptus community. Agricultural Water Management 8, 41-56.
Smettem, K. R. J., Bristow, K. L., Ross, P. J., Cook, S. E., and Johnson, A. K. L. (1994). Trends in water balance modelling at field scale using Richard's equation. Trends in Hydrology 1, 383-402.
Strahler, A. N. (1964). Quantitative geomorphology of drainage basins and channel networks. In `Handbook of Applied Hydrology'. (Ed. V. T. Chow.) Section 4-11. (McGraw-Hilh New York.)
Wood, W. E. (1924). Increase of salt in soil and streams following the destruction of the native vegetation. Journal of the Royal Society of Western Australia 10, 35-47.
Manuscript received 20 January 1997, accepted 22 August 1997
[TABULAR DATA NOT REPRODUCIBLE IN ASCII]
C. J. Clarke(A), G. W. Mauger(B), R. W. Bell(A), and R. J. Hobbs(C)
(A) School of Environmental Science, Murdoch University, Murdoch, WA 6150, Australia.
(B) Water and Rivers Commission, PO Box 6740, Hay Street, East Perth, WA 6892, Australia.
(C) CSIRO, Division of Wildlife and Ecology, Locked Bag No. 4, Midland, WA 6056, Australia.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Computer Modelling of the Effect of Revegetation Strategies on Salinity in the Western Wheatbelt of Western Australia, part 1|
|Author:||Clarke, C.J.; Mauger, G.W.; Bell, R.W.; Hobbs, R.J.|
|Publication:||Australian Journal of Soil Research|
|Date:||Jan 1, 1998|
|Previous Article:||Runoff and micromorphological properties of a grazed haplargid, near Cobar, NSW, Australia.|
|Next Article:||The interaction between revegetation strategies and major fault zones.|