The incorporation of perennials in general, and lucerne in particular, into farming systems of southern Australia has been proposed as a possible means to slow or stop the spread of dryland salinity. In order to be effective, lucerne roots must remove substantially more water from the soil than roots produced by annual crops and pastures. The term ‘buffer’ is used here to denote the extra water storage created by lucerne in addition to that normally created by an annual crop or pasture. In trials across southern Australia, lucerne has proved variable in its ability to create a buffer. In this research, we established 3 new trials, and collated results from current and published trials across Australia, to determine the effect of various edaphic (soil pH, texture, depth, and density for A and B horizons), climatic (average and actual seasonal rainfall), and agronomic (lucerne age, plant density, dry matter production, and rooting depth) factors on buffer size created by lucerne. Data from 26 trials were analysed, representing 84 site × year comparisons. The mean lucerne buffer for all comparisons was 91 mm, and increased with lucerne age. Buffers were generally greater for heavier-textured soils, but standard deviations of the means were large. Within a broad soil type, regression equations were developed to predict buffer size from climatic, edaphic, and agronomic factors, with r² values ranging between 0.96 and 0.84, and standard errors ranging between 40 and 44 mm. For all soil types, average summer rainfall (but not actual summer rainfall) was a significant component of the regression, suggesting that management of the lucerne stand, in terms of maintaining leaf area during summer, may have an important role in buffer development.