Force-velocity (FV) relationships are used to evaluate athletes and individualise training in several tasks. In rowing, only one previous study has determined FV relationships using ergometers. This study aimed to develop a method to obtain FV relationships during on-water rowing, to assess their quality and inter-day reliability and to compare them with ergometer profiles. Fourteen participants performed 8 all-out starts in instrumented double scull boats with 3 scull lengths under two conditions: two active rowers or one active and one passive rower. The FV relationships, reconstructed with all the conditions, were linear with a high goodness of fit (median R2: 0.96), covering 24.6 ± 5.0% of the velocity range. Ninety-five per cent confidence intervals for maximal force (F0: 3.9 ± 0.5%) and maximal velocity (V0: 6.2 ± 1.5%) were relatively narrow. The coefficient of variation for inter-day reliability was below 10% and intraclass correlation ranging from 0.69 for V0 to 0.96 for F0. Correlations between on-water and ergometer rowing were significant for F0 (r = 0.70) and Pmax (r = 0.82) but not for V0 (r = 0.11). F0 and V0 were significantly higher (23%) and lower (31%) on the ergometer, respectively. These results highlight a reliable linear FV relationship during on-water rowing to characterise rowing force production capacities, largely affected by on-water technical skills.