Improving the water-limited yield of dryland crops and farming systems has been an underpinning objective of research within the Australian grains industry since the concept was defined in the 1970s. Recent slowing in productivity growth has stimulated a search for new sources of improvement, but few previous research investments have been targeted on a national scale. In 2008, the Australian grains industry established the 5-year, AU$17.6 million, Water Use Efficiency (WUE) Initiative, which challenged growers and researchers to lift WUE of grain-based production systems by 10%. Sixteen regional grower research teams distributed across southern Australia (300–700 mm annual rainfall) proposed a range of agronomic management strategies to improve water-limited productivity. A coordinating project involving a team of agronomists, plant physiologists, soil scientists and system modellers was funded to provide consistent understanding and benchmarking of water-limited yield, experimental advice and assistance, integrating system science and modelling, and to play an integration and communication role. The 16 diverse regional project activities were organised into four themes related to the type of innovation pursued (integrating break-crops, managing summer fallows, managing in-season water-use, managing variable and constraining soils), and the important interactions between these at the farm-scale were explored and emphasised. At annual meetings, the teams compared the impacts of various management strategies across different regions, and the interactions from management combinations. Simulation studies provided predictions of both a priori outcomes that were tested experimentally and extrapolation of results across sites, seasons and up to the whole-farm scale. We demonstrated experimentally that potential exists to improve water productivity at paddock scale by levels well above the 10% target by better summer weed control (37–140%), inclusion of break crops (16–83%), earlier sowing of appropriate varieties (21–33%) and matching N supply to soil type (91% on deep sands). Capturing synergies from combinations of pre- and in-crop management could increase wheat yield at farm scale by 11–47%, and significant on-farm validation and adoption of some innovations has occurred during the Initiative. An ex post economic analysis of the Initiative estimated a benefit : cost ratio of 3.7 : 1, and an internal return on investment of 18.5%. We briefly review the structure and operation of the initiative and summarise some of the key strategies that emerged to improve WUE at paddock and farm-scale.

Rainfall is the main driver of potential yield in the dryland cropping environment of Australia. Rainfall-based empirically derived models such as that proposed by French and Schultz (1984) (F&S) can be used to determine an upper limit of water-limited potential yield, but F&S often overestimates actual yield as it does not account for rainfall distribution, runoff, drainage, or stored soil water. More complex models are available to predict potential yield more accurately; however, farmers and advisors favour easy-to-use-methods to estimate potential yield. To derive a simple yet accurate method for estimating potential yield, several adjustments to F&S were evaluated: (1) accounting for stored soil water at sowing, (2) varying the value of the intercept between yield and growing-season rainfall (GSR), (3) varying the water-use efficiency of the crops (WUE) according to soil type, and (4) adjustments to GSR depending on soil plant-available water capacity (PAWC). The water-limited potential yields predicted from these methods were compared with simulations from the daily time-step simulation model APSIM and observed wheat yields from 146 dryland wheat crops, managed to water-limited potential yield, covering the 1996–2006 seasons in the Mediterranean-type growing environments of Australia. The original F&S method overestimated actual yields, particularly at high rainfall (GSR > 220 mm) when PAWC was low, and underestimated yields at low rainfall (GSR < 220 mm). Significant improvements to the F&S were achieved with a few simple adjustments. With the addition of a variable intercept (dependent upon GSR), accounting for stored soil water at the start of the season and placing a cap on GSR that is a function of the soil PAWC, the predictive performance (RMSE 624 kg/ha) was similar to that gained with the daily time-step model APSIM (RMSE 419 kg/ha). The improved method gave more realistic estimates of water-limited potential yield, particularly at low and high rainfall and on soils of low PAWC.