The water extraction of deep-rooted perennial trees and shrub belts integrated with annual cropping/grazing systems was studied at 3 sites in the 300–450 mm rainfall zone of the Murray–Darling Basin of south-eastern Australia. Within 4 years of planting alley farming systems on cropland, the soil directly below and near the belts had dried the deep profile. Between 82 and 261 mm of extra soil water storage capacity was created in the 2.5 to 5.5–6 m profile. At Palamana (the only site monitored to greater depth), living roots were found 16 m below the surface. The cumulative water content of the soil to 12 m under the belts was 600 mm less than of soil cores extracted from nearby cropland. This water storage difference created under the belts is greater than the largest episodic event likely in this region and it is therefore unlikely that leakage will occur directly under or within a few metres of the belts. The early growth of the belts was rapid and the leaf area of the belts far exceeded that of remnant mallee eucalypt vegetation. The belts used water that had accumulated deep in the profile below the annual cropping systems they replaced. However, the belts only used water from below or within a few metres from the edge with the adjacent cropland. As suggested by RJ Harper et al. (2000), a much greater amount of potential recharge could be controlled if deep-rooted perennials were planted more closely across the landscape (compared with widely spaced belts). However, although the belts may be beneficial for the catchment water balance, they would be commercially unacceptable to farmers. In practice, farmers put the belts usually no less than 50–70 m apart so that less cropland is displaced and there is less belt/crop competition. In such cases alley farming only controls a small percentage of the total leakage, similar to the amount of crop yield lost by displacement and competition. It would be better to use a full coverage of perennials on soils where annual systems are the leakiest, rather than belts across all of the landscape, some of which may not be very leaky and could be highly profitable for annual cropping. Leakage could be controlled under cropland in a few years by growing easy to establish perennial species to retrieve moisture deep in the profile. At Pallamana the belts utilised 600 mm of accumulated leakage from deep in the profile in less than 4 years. Based on the average annual recharge rates under annual cropping (11–35 mm) the land could be cropped again for between 17 and 55 years before that leakage accumulated again.

Annual crops were grown in alleys between belts of perennial shrubs or trees over 3–4 years at 3 sites across low rainfall (<450 mm) south-eastern Australia. At the two lower rainfall sites (Pallamana and Walpeup), crop grain yields within 2–5 m of shrub belts declined significantly with time, with a reduction equivalent to 45% over 9 m in the final year of cropping. At the third, wetter site (Bridgewater), the reduction in crop grain yields adjacent to tree belts was not significant until the final year of the study (12% over 11 m) when the tree growth rates had increased. The reductions in crop yield were associated with increased competition for water between the shrub or tree belts and the crops once the soil profile immediately below the perennials had dried. At all 3 sites during the establishment year, estimates of water use under the woody perennials were less than under annual crops, but after this, trends in estimates of water use of alley farming systems varied between sites. At Pallamana the perennial shrubs used a large amount of stored soil water in the second summer after establishment, and subsequently were predominantly dependent on rainfall plus what they could scavenge from beneath the adjacent crop. After the establishment year at the Walpeup site, water use under the perennial shrubs was initially 67 mm greater than under the annual crop, declining to be only 24 mm greater in the final year. Under the trees at Bridgewater, water use consistently increased to be 243 mm greater than under the adjacent annual crop by the final year. Although the shrub belts used more water than adjacent crop systems at Walpeup and Pallamana, this was mostly due to the use of stored soil water, and since the belts occupied only 7–18% of the land area, increases in total water use of these alley farming systems compared with conventional crop monocultures were quite small, and in terms of the extent of recharge control this was less than the area of crop yield loss. At the wetter, Bridgewater site, alley farming appeared to be using an increasing amount of water compared with conventional annual cropping systems. Overall, the data support previous work that indicates that in lower rainfall environments (<350 mm), alley farming is likely to be dogged by competition for water between crops and perennials.