• Open Access

Managing water in agricultural landscapes with short-rotation biomass plantations

Authors

  • Richard J. Harper,

    Corresponding author
    1. School of Veterinary and Life Sciences, Murdoch University, Murdoch, WA, Australia
    2. Forest Products Commission, Perth Business Centre, Perth, WA, Australia
    Search for more papers by this author
  • Stanley J. Sochacki,

    1. School of Veterinary and Life Sciences, Murdoch University, Murdoch, WA, Australia
    2. Forest Products Commission, Perth Business Centre, Perth, WA, Australia
    Search for more papers by this author
  • Keith R. J. Smettem,

    1. School of Veterinary and Life Sciences, Murdoch University, Murdoch, WA, Australia
    2. School of Environmental Systems Engineering, The University of Western Australia, Nedlands, WA, Australia
    Search for more papers by this author
  • Nicole Robinson

    1. Forest Products Commission, Perth Business Centre, Perth, WA, Australia
    2. School of Agriculture and Food Sciences, The University of Queensland, Brisbane, Qld, Australia
    Search for more papers by this author

Abstract

Bioenergy production using woody biomass is a major climate change mitigation strategy but is often considered in terms of competitive effects on water. This paper describes the use of a short-rotation biomass system (Phase Farming with Trees PFT or ‘Kamikaze Forestry’) to manage water in dryland farming systems where this has accumulated below the root zone and has on and off-site environmental impacts. This excess water can be utilized for growth by deep-rooted, high-density biomass plantations inserted as short rotations into agricultural land. The objective is to promote rapid growth and mining of deep stored water through strategies such as high planting densities, the use of fast-growing species or fertilization each of which increases leaf area. Once the water is used, the trees are harvested and excess water is allowed to build up again in the subsequent cropping phase. Biomass production and water depletion were measured in a five-year rotation of trees inserted into a dryland (367 mm yr−1 mean annual rainfall) cereal farming system in south-western Australia. Both were markedly affected by tree age, planting density, and landscape position on a very minor slope. The greatest biomass production was achieved with high-density (4000 stems ha−1) plantings of Eucalyptus occidentalis and Eucalyptus globulus in lower landscape positions. High-density plots of these species in mid and upper landscape positions succumbed to drought after 3–4 years, but depleted available soil water to depths of >8 m, equivalent to 771 mm of stored available water. These results suggest that biomass yield can be readily manipulated through planting density and site selection. Moreover, biomass production can produce positive water management co-benefits.

Introduction

Bioenergy production using woody biomass materials is a major component of climate change mitigation strategies (Chum et al., 2011). In some cases, this can rely on forest residues (Canadell & Raupach, 2008; Mitchell et al., 2012) but it can also be derived from purpose grown woody crops (Canadell & Raupach, 2008). Where these crops are established on farmland, a major issue is the competition with water (Koh & Ghazoul, 2008; Gerbens-Leenes et al., 2009; Dornburg et al., 2010) with these concerns mirroring previously well canvassed problems following plantation establishment including the loss of water yield in ground and surface water catchments (Calder et al., 1993; Jackson et al., 2005). An additional issue is the displacement of food production particularly given the scale of competing demands for carbon mitigation, increased food production, and climate change impacts on crop yields (Erb et al., 2012; Smith et al., 2013).

Various strategies have been advocated to avoid this competition with food and water production, such as producing biomass from abandoned or lower productivity land (Wicke et al., 2011; Sochacki et al., 2012) or integrating belts of trees with agriculture (Stirzaker et al., 2002; Bartle & Abadi, 2010). In water-limited environments, tree belts will however result in displacement of food production and also competition between the trees and adjacent crops (Lefroy & Stirzaker, 1999; Robinson et al., 2006; Sudmeyer et al., 2012).

In some situations, such as the dryland cropping areas of southern Australia and various irrigated areas globally the farming systems are often associated with a hydrological imbalance (Peck & Hatton, 2003; Khan et al., 2006), with the accumulation of excess water and consequent impacts on productivity and off-site environmental values. In the former case, this has been induced by the widespread replacement of deep-rooted native vegetation with shallow-rooted annual plants for agriculture. Annual crops only transpire water for part of the year, and as a consequence, recharge under agricultural systems is one to two orders of magnitude greater than under native vegetation (Allison et al., 1990; Smettem, 1998), resulting in wetting of deep soil horizons, rising water tables, and the mobilization of salt stored within the regolith (Peck & Hatton, 2003). In irrigated areas, recharge rates can be much higher than in dryland areas due to leakage from both rainfall and irrigation. Whereas trees are advocated for landscape scale hydrological management (Calder, 2005), it has been suggested that extensive planting over as much as 80% of the landscape may be required to have a measureable response (George et al., 1999), however this will obviously depend on the nature of the local hydrological systems.

An alternative approach, in a system termed phase farming with trees (PFT) (Harper et al., 2000) or ‘Kamikaze Forestry’, is to incorporate short (3–5 year) rotations of trees with agriculture (Harper et al., 2000, 2010; Hatton et al., 2002; Sochacki et al., 2007, 2012). This system relies on the use of fast-growing species and the manipulation of silviculture (planting density, fertilization) to produce biomass as rapidly as possible (Harper et al., 2010), thus minimizing the length of the tree phase as this is often less profitable than the agricultural (cropping) phase (Harper et al., 2000). Whereas the aim of traditional dryland forestry is to develop forest stands that are parsimonious with water, through species selection, planting density or fertilization, here the aim is to accelerate the system to both rapidly deplete excess soil water and produce biomass, and thus allow a faster return to crop production. Because biomass as a bioenergy feedstock is the end product, rather than timber or other wood products, tree death doesn't matter as the trees can be harvested dead or alive. Although a harvesting system for PFT has yet to be developed (Harper et al., 2010), no particular problems are envisaged from this approach. Removal of the need for trees to remain alive opens up a range of possibilities for management, including very high planting densities, fertilization, and also use of species that wouldn't otherwise be considered in particular environments (Harper et al., 2010).

Although this will also result in the displacement of food producing land for biomass production, the concept is that the area of land required to produce biomass will be less as the trees, and as will use moisture from both incoming rainfall during the tree growth phase and also water stored in the agricultural phase. In this sense, it is akin to fallow systems used for dryland cereal production and is also similar to green fallow (Sanchez, 2002) and silage sycamore (Steinbeck et al., 1972; Devine et al., 2006) biomass production systems. Whereas, the occurrence of this deep stored water is often considered a problem, it also represents an additional source of water for biomass production. Similarly, in some cases nutrients such as nitrogen may also have leached below the root zone of annual crops or pastures (Anderson et al., 1998) and hence the tree phase provides an opportunity to scavenge these nutrients (Rockwood et al., 2004). The PFT system may thus provide a range of environmental co-benefits in terms of improving the sustainability of farming systems as well as a source of bioenergy (Harper et al., 2010).

The rates of biomass accumulation and depletion of soil water by a full rotation of PFT are untested, although growth (Sochacki et al., 2007; Harper et al., 2010) and nutrient removals (Sochacki et al., 2013) to 3 years of age have been reported from an experiment in a dryland farming area of south-western Australia. In the same region, Robinson et al. (2006) found evidence of water depletion to depths of up to 10 m beneath 7 year old mallee eucalypts on similar soils. The consequence of tree root water extraction has been simulated using the WAVES model (Zhang & Dawes, 1998) for several sites in southern Australia with annual rainfall ranging between 320 and 580 mm yr−1 (Harper et al., 2000, 2010; Hatton et al., 2002) using a range of scenarios based on soil texture, depth of soil profile, and the presence or absence of an unconfined aquifer. These simulations suggested that the technique was applicable to sites with soil profiles several meters deep without root inhibiting layers or water tables.

This paper reports final results for a full five-year PFT rotation from a field experiment (Sochacki et al., 2007, 2013) and sets out to resolve: (i) whether the fundamental premise of the system is valid, i.e., rapid dewatering of soil profiles occurs to several meters depth; and (ii) whether the amount of soil water depletion and thus biomass production and tree rotation length, can be manipulated by species selection, stand density management or placement in the landscape.

Materials and methods

Study site

The study site was located near Corrigin, Western Australia, approximately 240 km east of Perth (Fig. 1, 117°41′47.13″E; 32°23′24.67″S), the state capital and has been previously described (Sochacki et al., 2007, 2013; Harper et al., 2010). This site was selected as having soils and landforms representative of the general region (McArthur, 1991), which has a semiarid Mediterranean climate, with an annual seasonal drought from November to April. Rainfall and potential evaporation were estimated using the SILO climate data base (Jeffrey et al., 2001). The site has a mean annual rainfall (1889–2006) of 367 mm yr−1 and mean annual evaporation of 1803 mm yr−1. The climatic conditions during the year prior to planting, and the subsequent 5 years of the experiment are given in Table 1.

Table 1. Summary of the climate of the experimental site. Data from SILO (Jeffrey et al., 2001)
YearTemperatureRainfall (mm)Pan evaporation (mm)
Max. (°C)Min. (°C)
  1. Temperatures are mean annual values

200023.69.84041734
2001(planted)23.49.34031729
200224.29.82881671
200323.810.32221826
200423.910.03661826
200523.49.63031836
200624.09.73731755
Figure 1.

Location of the Corrigin experimental site, Western Australia, with the 300 mm and 600 mm annual rainfall isohyets.

Conventional farming on this site involves annual rotations of cereal (Triticum aestivum, Hordeum vulgare) or legume (Lupinus angustifolius) crops with improved annual legume (Trifolium subterraneum) and grass (Lolium rigidum) pastures, grown during the winter rainfall season. It is thus similar to the farming systems of broad areas of southern Australia (Squires & Tow, 1991).

Experimental details

Trial design

An experiment was established in August 2001 (winter) to determine the rates of removal of soil water to depth and whether water use and biomass production could be manipulated by species selection, planting density or fertilizer application.

The experiment comprised a randomized complete block design with three replicate blocks, each with 15 treatments. The treatments comprised three species (Eucalyptus globulus, Eucalyptus occidentalis, Pinus radiata) planted at four densities (500, 1000, 2000, and 4000 stems ha−1), as well as 500 stems ha−1 plus nitrogen fertilizer applied at 100 kg N ha−1. As there was no response to the fertilizer treatment, these plots are not considered further in this analysis.

The three blocks were situated in the same field, but arrayed in different landscape positions. Block 2 (upper slope) was on a gravelly ridge, Block 1 (mid-slope) in a concavity with a sandy duplex profile, and Block 3 (lower slope) with a sandy duplex profile with a confined saline water table with a potentiometric surface at 2–3 m below the ground surface. The relative elevation between these plots was approximately 15 m. A single core was extracted from each Block prior to trial establishment to depths of between 7 m and 10 m depth (Table 2). Each block was underlain by deeply weathered profiles most likely developed from granite, typical of the region (McArthur, 1991). Soil analysis included 1 : 5 soil/water suspensions of electrical conductivity (EC) and pH measured in CaCl2 (Rayment & Higginson, 1992), clay content of the <2 mm fraction by the pipette method (Gee & Bauder, 1986), and an assessment of the >2 mm (gravel) fraction by sieving.

Table 2. Main soil properties in a representative bore within each Block
Slope positionBlockDepth (m)Claya (%)Gravel (%)Hard layerspHEC (mS m−1)
  1. a

    Clay content <2 μm of the <2 mm fraction; gravel >2 mm; pH in CaCl2 and EC (electrical conductivity, mS m−1) of a 1 : 5 soil/water extract. Hard layers were identified in the drill spoil.

Upper20.51247 5.42.1
1.02222 5.42.8
2.030 4 5.53.3
2.82348Silcrete6.04.1
4.011 1Silcrete6.511.9
5.322 2 6.325.8
6.010 7 6.248.0
7.012 3 6.462.2
8.326 4 6.676.2
9.028 1 6.661.8
Mid10.52433 4.93.5
1.04353 5.53.6
2.02417 5.94.0
3.053 2 7.35.7
4.3  Ferricrete  
5.422 3 7.74.6
6.520 6 7.411.0
7.019 6 7.313.9
8.016 3 6.742.1
9.016 1 6.568.5
Lower30.533 0 8.114.9
1.048 1 7.311.5
2.050 0 8.021.8
3.02612 7.619.8
4.021 1 6.241.5
5.042 1 5.767.7
6.048 1 5.474.8
7.01815 5.730.9

Trees were planted by hand in early August 2001 in 50 × 50 m treatment plots. This followed ripping to a depth of 0.5 m and treatment of the site with standard cropping herbicides. The site was not mounded.

Biomass

Biomass estimation

Measurements were made of all treatments, on an annual basis, to obtain estimates of biomass. Predictor variables likely to be used in the allometric relationships were measured on all trees within 20×20 m permanent measurement plots (Sochacki et al., 2007). Plots of this size were considered unlikely to be affected by edge effects between contrasting treatments. Attributes measured included diameter at 10 cm, diameter at 130 cm and height. Survival was estimated as a proportion of the plants alive at the time of measurement compared to the number planted.

The development of allometric relationships for trees at 3 years of age at this site is described elsewhere (Sochacki et al., 2007). Further trees were sampled at 57 months of age in May 2006 to extend the allometric relationships to larger trees. Tree roots were excavated to a depth of 0.35 m with an excavator and collected by hand to a nominal root diameter limit of 5 mm. Derived relationships were applied to the plot tree measurement data to develop estimates of the oven-dry biomass yield of different tree components and total biomass. Resultant equations are summarized in Table 3.

Table 3. Prediction equations for whole tree biomass (kg tree−1) at age 5 derived from measurements of stem diameter and tree height for Eucalyptus globulus, Eucalyptus occidentalis, and Pinus radiata
SpeciesPrediction equation n r 2 FI SE (kg)CV (%)
  1. Bt, total biomass; Bl, leaf biomass; Bs, stem biomass; Br, root biomass; n, number of sample trees; r2, proportion of variation explained; FI, Fit index; SE, standard error of estimate; CV, coefficient of variation.

E. globulus Bt = 1.019e−2.576d101.934ht0.6331160.980.952.222
E. occidentalis Bt = 1.01e−2.358d101.916ht0.6481150.940.883.336
P. radiata Bt = 1.059e−2.985d101.783ht0.8881260.950.961.834

Predictor variables measured were tree height, crown volume, and diameter over bark at 10, 50 and 130 cm above ground level. E. occidentalis often had more than one stem at 130 cm and a diameter equivalent was calculated when more than one stem was measured (Avery & Burkhart, 1983). No estimates were made of the energy content of the materials produced in this study.

Water

Soil water

A neutron moisture meter was used to estimate changes in soil moisture content. Neutron moisture meter tubes were installed in selected treatment plots in January 2002 in all three landscape positions. This encompassed all planting densities of E. globulus and the 1000 and 4000 stem ha−1 treatments for the other species.

A total of 41 holes were drilled to 9 m and 50 mm polyvinyl chloride (PVC) tubes were capped and inserted into a bentonite soil slurry using the method of Greacen (1981). PVC was chosen over other tube materials due to concerns of corrosion as a result of saline soil profiles. The installation of a 50 mm PVC tube resulted in a clearance of less than 4 mm inside the access tube thus reducing error due to eccentric positioning of the probe.

The first measurements using a neutron moisture meter were made in March 2002, with measurements subsequently made at bi-monthly intervals. Soil water contents to 8 m were also measured in a blank plot (i.e., with volunteer pasture) in each replicate and a tube in the adjacent field that was still being cropped.

Estimation of soil water content

Neutron moisture meter counts/second were converted to gravimetric soil moisture values (g g−1) using a calibration equation provided by Dr Phil Ward (CSIRO Plant Industry, Floreat, WA, Australia). Further conversion to volumetric values (mm3 mm−3) was made using an assumed bulk density of 1.7 g cm−3, with this value obtained from the regional study of Robinson et al. (2006).

Soil water deficits were calculated from these data by assuming the measurements in June 2002 (i.e., at 1 year after establishment) represented a baseline, and subtracting the volumetric soil moisture contents for each subsequent measurement. Values were summed over the first 8 m of the neutron moisture tubes.

Estimation of water use efficiency

A broad index of water use efficiency (WUE, kg ha−1 mm−1) was developed based on the biomass production (kg ha−1), soil water depletion at a particular age (mm), and the total rainfall to that age. Thus, at age 2, the total water supply included the rainfall for the 2 years (691 mm), plus the respective soil water deficit (mm). This index was not calculated for different slope positions as the effective water contributions in each slope position were uncertain due to unknown run-off/on dynamics and contributions from groundwater.

Results

Biomass

There were clear effects on biomass production as a result of age, stocking, landscape position, and species. Individual plot data are shown in Table 4 with main effects presented in Fig. 2. Biomass production increased with age, with mean values across all plots of 8.1 ± 0.8, 11.5 ± 1.1, and 16.0 ± 1.3 t ha−1 at 3, 4, and 5 years after establishment, respectively. For each species at age 5, there was an increase in yield with planting density, and slope position (Table 4), as previously reported for these trees at age 3 (Sochacki et al., 2007).

Table 4. Biomass production (t ha−1) for trees between 2 and 5 years of age for each species, planting density, and slope position
SpeciesSlope pos.Tree ageMeanSE
2345
Stocking (stems ha−1)
500100020004000500100020004000500100020004000500100020004000
E. glob.Up0.10.30.82.20.51.74.17.71.02.86.610.11.75.811.414.54.51.11
Mid1.83.15.66.05.911.510.29.57.516.613.19.811.824.319.510.910.41.48
Low1.52.34.48.38.69.310.615.911.412.115.522.515.015.417.827.012.31.72
E. occi.Up1.02.44.16.73.95.55.48.96.47.98.19.110.612.911.510.57.20.84
Mid3.1 7.53.95.5 8.18.96.0 8.79.112.0 10.39.27.70.76
Low2.54.85.314.210.613.711.621.616.518.316.029.521.022.121.233.516.42.13
P. rad.Up0.20.20.84.61.41.65.214.32.94.011.321.57.07.722.331.28.52.32
Mid0.30.52.74.82.02.911.913.83.86.119.316.87.512.629.321.39.72.14
Low0.31.63.31.32.210.113.64.94.616.720.710.87.725.826.412.610.22.14
Mean1.21.93.85.84.57.09.011.76.710.613.315.510.515.818.819.011.9 
SE0.40.60.71.31.11.71.11.71.62.21.72.51.82.72.33.22.0 
Figure 2.

Main effects of biomass production (dry matter, t ha−1) with (a) planting density (stems ha−1), (b) species and (c) slope position for ages 2–5.

Mean total biomass yields varied with planting density and ranged from 1.7 to 27.0 t ha−1 5 yr−1 for E. globulus, 9.2 to 33.5 t ha−1 5 yr−1 for E. occidentalis and 7.0 to 31.2 t ha−1 5 yr−1 for P. radiata. The highest biomass yields for E. globulus and E. occidentalis were from 4000 stems ha−1 treatments located on the lower slope site (Table 4). For P. radiata, the highest yields were produced with 4000 stems ha−1 in the upper-slope site. Mean yields of the three species, in the high planting density plots, were not significantly different and ranged from 17.5 to 21.7 t ha−1 5 yr−1.

At age 5, the yields were 10.5 ± 1.8, 15.8 ± 2.7, 18.8 ± 2.3, and 19.0 ± 3.2 t ha−1, for the 500, 1000, 2000, and 4000 stems ha−1 treatments, respectively (Fig. 2a). These yields were affected by high-density plots of E. occidentalis and E. globulus collapsing (Table 5) after exhausting soil moisture (Table 6) in upper and mid-slope positions at 3 or 4 years after planting, and the change in survival in the mid-slope P. radiata plots from 92.8 to 68.7% at age 4. As the trees got older, the yield advantage from higher planting densities decreased, such that at age 5 the mean yields of the 2000 and 4000 stems ha−1 treatments were similar (Fig. 2a). Overall, there were few differences in biomass production between species (Fig. 2b).

Table 5. Survival (% of initial planting) for trees between 2 and 5 years of age for each species, planting density, and slope position
SpeciesSlope pos.Tree ageMeanSE
2345
Stocking (stems ha−1)
500100020004000500100020004000500100020004000500100020004000
E. glob.Up20283698152536351525301915253019294.9
Mid958891859588766290886058588605727.2
Low95989694909375829088647485886473842.8
E. occi.Up75787079757870775786477575647616.8
Mid95 989995 15190 13180 1115013.1
Low100939699959095979590949795909497950.8
P. rad.Up1005398981005398979553989095539890854.9
Mid1001009996100100999610010099311001009830905.9
Low85839485858394858583898585838985860.9
Mean8577869383767362827568457975674569 
SE8.68.86.92.58.98.89.813.08.78.710.213.58.58.710.313.42.7 
Table 6. Soil water deficit (mm) developed in soil profiles to a depth of 8 m for each species, planting density, and slope position for 2–5 years after establishment. The soil water deficit is the maximum value achieved in each year compared to the initial value 1 year after planting
SpeciesSlope positionTree ageMeanSE
2345
Stocking (stems ha−1)
10004000100040001000400010004000
E. glob.Up15928421645523758923327430650.7
Mid20920629128139348440423231336.4
Low23721229423627027819625124711.8
E. occi.Up16246537977148476429315143484.9
Mid 345 476 365 11332576.3
Low18536726042728153228749435443.1
P. rad.Up17120519441321353720747030151.9
Mid13613721329626433426611422029.3
Low13914918922628022125420820817.1
Mean175263255398303456267256297 
SE12.336.722.956.531.957.223.046.717.1 

The differences in plot yields with planting density and slope position were marked, for E. globulus, at age 3, for example, the yield at 3 years old for the 500 stems ha−1 upper slope plot was 0.5 t ha−1, whereas the lower slope 4000 stems ha−1 treatment had a yield of 15.9 t ha−1, or around 30 times greater. This difference was maintained at age 5 with respective yield differences of 1.7 and 27.0 t ha−1. Similar patterns were apparent with E. occidentalis and P. radiata, albeit with smaller ratios between the low density-upper slope and lower slope high-density treatments (Table 4).

There were also clear differences in biomass yield with slope position (Fig. 2c). Across all species and densities, the yields of the lower, mid- and upper slope positions were 11.0 ± 1.4, 8.2 ± 1.1, and 5.0 ± 1.1 t ha−1 at 3 years. These differences were maintained at age 5 with respective yields of 20.4 ± 2.1, 15.3 ± 2.1, and 12.3 ± 2.3 t ha−1. As described, survival of the plots varied markedly with slope position (Fig. 3c), with 86% survival at age 5 for the lower slope, 60% for the mid-slope, and 54% for the upper slope. Most drought deaths had occurred in upper slope plots by age 3 and in the mid-slope plots at age 4. Drought deaths first occurred in the 4000 stems ha−1 E. occidentalis treatment in upper and mid landscape positions at age 3, with 7% and 1% survival, respectively (Table 5). The same pattern was repeated at age 4 in E. globulus, with 19 and 5% survival. In general, there were fewer deaths in the lower slope plots.

Figure 3.

Main effects of tree survival (%) with (a) planting density (stems ha−1), (b) species and (c) slope position for ages 2–5.

The response to slope position also varied with species. For the eucalypts, the highest yields were achieved in the lower slope positions. At age 5, this was 18.8 ± 2.8 vs. 8.4 ± 2.9 t ha−1 for E. globulus and 24.4 ± 3.0 vs. 11.4 ± 0.6 for E. occidentalis lower and upper slopes, respectively. For P. radiata, however, the overall effects of slope position were negligible, with similar yields for each position. At age 5, the respective yields for the lower, mid, and upper slope positions were 18.1 ± 4.7, 17.7 ± 4.8, and 17.1 ± 5.9 t ha−1, respectively (Table 4).

Soil moisture deficit

Changes in maximum soil water deficit (mm) from 1 to 5 years are plotted for each slope position, species and the 1000 stems ha−1 and 4000 stems ha−1 planting densities (Fig. 4), with each of these factors contributing to differences in response. Over all plots and years, the greatest soil water depletion was achieved by the 4000 stems ha−1 treatment of E. occidentalis in an upper slope position at 3 years, with 771 mm of depletion, with all the trees in this plot dying in the preceding months. Across all plots, there was clearly a smaller amount of water depletion with 1000 stems ha−1 compared to 4000 stems ha−1 (Fig. 4a), with these having mean respective values across all treatments of 180 mm vs. 269 mm at age 3, and 303 mm vs. 456 mm at age 4. By age 5, there was no difference, this being due to the tree deaths in the high-density plots in the earlier years (Table 5).

Figure 4.

Main effects of maximum annual water deficit to 8 m depth (mm) with (a) planting density (stems ha−1), (b) species and (c) slope position for ages 2–5.

The induction of a soil water deficit varied markedly between E. occidentalis and the other species, with this being greater each year, with values of 466 mm at age 4, compared to 322 mm for E. globulus and 259 mm for P. radiata (Fig 4b). Whereas the rate of water depletion for E. occidentalis slowed between 4 and 5 years (19 mm), it was greater for P. radiata (50 mm), this likely reflecting different in patterns of growth, as the biomass increment between years 4 and 5 was also the greatest for P. radiata (Fig. 2b). There were also widespread deaths in the higher density E. occidentalis plots, this reducing water use.

There were clear differences in soil water depletion with slope position (Fig. 4c), with this being greatest in the upper slope positions. Respective values for the upper, mid- and lower slope positions, respectively, were 471, 368, and 310 mm, when averaged across all species (Table 6). This is despite the opposite trend in growth with slope position as discussed earlier; with the trees in the lower slope positions consistently having greater biomass yields than those in upper slope positions.

Variation in soil moisture with depth

The soil moisture deficit provides an indication of the overall performance of the trees in extracting available soil moisture, however patterns of water depletion can also be considered in terms of depth and season, in order to provide an insight into the pattern of root water extraction. For example, changes in water content between early spring (September) and late autumn (March–April) provide an indication of the depth of rooting because there is generally little summer rainfall in this region.

As the highest rates of water depletion occurred under the 4000 stems ha−1 treatment of E. occidentalis, this species was examined in more detail and compared to P. radiata (Fig. 5). Within 2 years of planting, there had been a reduction in soil water content at depths of up to 4.0–5.5 m in the different landscape positions. This change persisted and was more pronounced in the upper slope position. By 4 years, this was >8 m in the upper slope E. occidentalis plot (Fig. 5a), but was less pronounced in mid- and lower slope positions (Fig. 5), indicating refilling of the soil moisture store in these landscape positions, either from surface runoff or through flow within the soil profiles. Soil moisture depletion under P. radiata was generally lower in all slope positions (Fig. 5d–f).

Figure 5.

Change in soil moisture content (mm m−1) measured to a depth of 8 m, over a period of 3 years from June 2002 (1 year after commencement) to February 2005 for the 4000 stems ha−1 treatments in upper, mid- and lower slope positions for E. occidentalis (a–c) Pinus radiata (d–e). Corrections made after online publication August 29, 2013: Fig.5 y-axis label have been updated.

Water use efficiency

The water use efficiency (WUE) was estimated from both the rainfall and soil moisture deficit (Fig. 6) for trees from 2–5 years after establishment. WUE tended to plateau after 3 years, with that of the higher planting density (4000 stems ha−1) being consistently greater than that of the lower density (1000 stems ha−1, Fig. 6a). Despite WUE being greater in 2-year-old E. occidentalis trees, in the latter years (3–5) there were no apparent differences with the other species (Fig. 6b).

Figure 6.

Change in water use efficiency (WUE; kg ha−1 mm−1) over time for different (a) planting densities (stems ha−1) and (b) species. WUE was calculated from the total dry matter and both the soil water deficit induced by the different treatments and the rainfall received up to the time of measurement.

Discussion

Water use and biomass production

Both planting density and the location of trees in the landscape had a strong influence on biomass yield as previously reported for this site at 3 years of age (Sochacki et al., 2007). Maximum biomass production at age 5 for the eucalypts was achieved in lower landscape positions with high planting densities (4000 stems ha−1), with yields of 27.0 and 33.5 t ha−1 for E. globulus and E. occidentalis, respectively. The highest P. radiata yield (31.2 t ha−1) was achieved in a upper slope position with a planting density of 4000 stems ha−1; overall this species was less responsive to landscape position both in terms of biomass production and drought death.

Results from this study also support the basic premise of the phase farming with trees system. That is, short rotations of trees can utilize stored soil water over the depth of rooting and dried soil can be used as a buffer to prevent groundwater recharge from subsequent agricultural crops. Moreover, this deficit can be substantial, with E. occidentalis planted at 4000 stems ha−1, having a soil water deficit ranging from 443 to 771 mm at 4 years of age dependant on slope position (Table 6), with this being 1.5–2.5 times mean annual rainfall. Assuming a recharge rate of 40 mm yr−1, this could result in a tree/cropping rotation of 3–4 years of trees, then 11–20 years of agriculture.

Tree roots exploited soil profiles to depths of >8 m at 4 years of age (Fig. 5a). The tree roots are likely exploiting old root channels and other macropores particularly as the subsoils in this region have bulk densities of >1.7 g cm−3 (Robinson et al., 2006). The results are also consistent with the depths of soil water depletion reported in other studies in the region (Brooksbank et al., 2011; Mendham et al., 2011).

The marked variation in soil water depletion between different species and with planting density indicates that the amount of soil water depletion can be managed by manipulating these factors. Thus, it may be possible to further increase the soil water deficit with higher planting densities or by selection of other species. With the wide genetic diversity and water use within the Eucalyptus genera alone (e.g. White et al., 2002), there is potential to select species with higher water depletion potential. Although rotations longer than 3 years resulted in more water depletion for P. radiata, the water deficit for these species still did not approach that of E. occidentalis at 4 years.

There were also marked differences in tree response to landscape position across the site. Soil water was depleted more effectively in the upper landscape positions, despite smaller biomass production compared to sites in lower landscape positions (Fig. 2c). This is most likely due to the redistribution of water across the landscape, particularly overland and through flow following peak rainfall events, and possibly trees accessing groundwater systems (George, 1990; Brooksbank et al., 2011). The relative contribution of these sources was not measured, however the trees in the lower landscape positions thus have a smaller soil water deficit due to a greater supply of water. The subsoils of each of the blocks were semi-saline with EC values of 60–70 mS m−1 at depth (Table 2); and it is highly likely that there will be differences between the species in their ability to extract water from both subsoils and groundwaters (Niknam & McComb, 2000; Nosetto et al., 2008). Carefully matching species to site conditions will thus be an important component of deploying the PFT system.

The relief in this landscape is very subtle, with only 15 m difference in elevation between the upper and lower slope positions and generally very shallow (<2–3%) slopes. These differences however clearly have a profound effect on landscape hydrology and need to be taken into account in management. The extra accession of water into these lower landscape positions suggests that trees or other woody perennials could be maintained in some of these areas on a permanent basis as a coppice system.

Quantifying co-benefits and environmental services

Reforestation in this environment can produce two major outcomes; biomass as a feedstock for bioenergy, and landscape scale water management. Whereas bioenergy markets are emerging with improvements in technology and demand (Chum et al., 2011), obtaining payment for water management related environmental services is more problematic (Weersink & Wossink, 2005), particularly in dryland environments. Where reforestation results in the restoration of watershed water quality and an increase in water to sell (Townsend et al., 2012), there is a case for payments per hectare reforested. In this environment, salinization results from a landscape hydrological imbalance with a range of off-site impacts such as on biodiversity and built infrastructure (Peck & Hatton, 2003) and there is no fresh water runoff that could be captured and sold.

In the absence of a market for water management benefits, the PFT system does provide a means of boosting biomass productivity per unit land area, through use of both ambient rainfall and stored soil water. It will also provide some flexibility in land use; whereas forestry systems mostly involve a permanent land-use change, this approach will allow decisions on land use (e.g., length of forestry rotation) to respond to the prevailing markets for food and biomass. This is one of the underlying precepts of the silage sycamore system (Steinbeck et al., 1972; Devine et al., 2006).

Biomass plantings could be considered in terms of optimizing the dual values of water use and biomass production. If there was a payment for hydrological services, 3–4 year rotations of eucalypts would achieve the desired outcome. If however, the aim was to maximize biomass production, then the strategy would be the establishment of higher planting densities of eucalypts in lower slope positions, where additional water is received. Depending on the nature of the local aquifer systems and the sustainability of excess water supply, an alternate strategy would be to establish permanent plantings of trees in these landscape positions.

An economic analysis based on these full results has not been undertaken. Earlier economic analysis (Harper et al., 2000) made assumptions about likely yields and costs of production, in the absence of biomass yield and water depletion data, and no formal renewable energy market. That analysis suggested that profitability could be increased by decreasing planting and harvesting costs. Although there are still uncertainties about the best methods of harvest (Harper et al., 2010), future economic analysis could consider likely economic returns from the system from biomass, valuing the improvements in sustainability in the farming systems and the profitability of the agricultural component of the system. Such an analysis should allow optimization of the system in terms of planting density and rotation length on a site-specific basis.

Potential application of PFT to irrigation schemes and mine dewatering systems

The removal of stored water in soil profiles raises the possibility that the PFT system can be applied to water management in other situations. For example, many irrigation systems accumulate water in the deeper soil profile as a result of excess water percolating past the root zone (Khan et al., 2006) and the application of a rotation of trees may allow dewatering providing subsoil salinity levels are tolerable. Interestingly, such an approach was advocated for the long-term management of irrigated cotton systems in Sudan (Greene & Snow, 1939). This will also produce a biomass feedstock, while improving the long-term sustainability of agricultural production, rather than displacing it.

A similar approach could be used to dispose of the waste produced from mine dewatering operations. Because mines generally have a limited production life, the water source is temporary and long-term forestry is not feasible. However, this situation may lend itself to short forestry rotations, again, provided the water quality is suitable. Anderson & Davies (2006) note a successful trial using coal mine wastewater to grow fodder crops in New South Wales, Australia and there are many examples of biomass production from wastewater using fast-growing species such as willow (Börjesson & Goran, 2002).

The results of this study thus suggest that biomass plantings, and the PFT system in particular, could have a role in providing a range of water management benefits and increase the sustainability of different agricultural systems. This is contrast to the common position that considers that biomass production will have negative consequences.

Acknowledgements

Funding for this work was from the Australian Government's Joint Venture Agroforestry Program (Projects CAL-3A, CAL-6A) and Farm Forestry Program (Project 983197) and the Western Australian Government's Department of Conservation and Land Management. We thank Dr John McGrath, Dr Stuart Crombie, Mr John Bartle, Professor Lyn Abbott, Dr Tom Hatton, Dr David McKenzie and Prof. Chris Mitchell for stimulating discussions, and Dr Phil Ward CSIRO Plant Industry for help with the conversion of NMM measurements. Andrew Stilwell, Robert Archibald, Yvette Oliver, Wesley Hibbitt, Yumiko Bonnardeaux, Sylvain Micola, Nick Spencer, and Dr Augustine Okom assisted in the field and laboratory. We also thank Lawrie and Jenny Pitman for access to their land and enthusiastic support of the work.

Ancillary