In New Zealand, dairy farming faces increasing scrutiny for its environmental impacts, including those on soil carbon (C) stocks; hence, alternative management practices are required. One such practice is usage of deep-rooting forage, such as lucerne (Medicago sativa L.). We measured the C and water exchange of two neighbouring lucerne fields on stony, well-drained soil for 3 years, following conversion from grassland. One field received irrigation and effluent; the other received neither. Net CO2 exchange and evaporation were measured by eddy covariance, drainage and leaching with lysimeters, and water inputs with rain gauges. Biomass removal from harvesting and grazing was recorded by direct sampling. In the conversion year, irrigated lucerne was C-neutral despite two harvests and losses from the conversion process. In the 2nd and 3rd years combined, the biomass-C removal exceeded net CO2 uptake, causing net losses of 450 g C m-2 and 210 g C m-2 for irrigated and non-irrigated lucerne, respectively. Leaching losses accounted for 1 to 9 % of annual net C uptake from the atmosphere. The ratio of ecosystem respiration to gross photosynthetic productivity (GPP) increased from <0.7 in spring to ≈ 1 in autumn. Consequently, the net C balance for both lucerne crops showed gains in the first two growth periods of each year and losses in the subsequent two to four growth periods. Irrigation made no difference to the photosynthetic water-use efficiency at field scale (GPP/evaporation), but enhanced production water-use efficiency (biomass/water input). Irrigation increased both the absolute amount of drainage and the fraction of water inputs lost by drainage. In one year, significant summer drainage occurred for the irrigated lucerne. To prevent that, soil-water content should be kept well below field capacity but above the crop's water-stress level. Such practice would likely also help retain soil carbon.