This study utilised on-water graded exercise tests (GXT) to determine the power output (PO) corresponding to the first and second lactate thresholds (LT1PO and LT2PO), subsequently examining their relationship to the mean maximal power (MMP) and race time achieved across three on-water sprint kayak time-trials. Twelve well-trained sprint kayak athletes completed an on-water GXT and a 200-, 500- and 1000-m time-trial utilising novel instrumented paddle technology. Stepwise multiple regression was used to determine whether equations incorporating 200-, 500- and 1000-m MMP data could be used as an alternative method for estimating LT1PO and LT2PO. On-water GXT derived LT1PO and LT2PO were 151 ± 34 and 194 ± 39 W, respectively. For the 200-, 500- and 1000-m time-trials, MMP were 528 ± 143, 358 ± 92 and 287 ± 67 W, respectively. Athletes’ LT1PO and LT2PO had very-large inverse relationships to 200-, 500- and 1000-m time-to-completion (r = −.71 to −.85, P ≤.010) and very-large, to near-perfect positive relationships to 200-, 500- and 1000-m MMP (r =.81 to.94, P ≤.001). The equation incorporating 1000-m MMP alone provided the best prediction of LT1PO and LT2PO, explaining 78% and 88% of the variance, and yielding a standard error of estimate (SEE) of 11.3% and 7.1% for these measures, respectively. The results of this study provide further evidence to support the ecological validity of recently developed on-water GXTs graded by PO, since LT1PO and LT2PO were significantly correlated to 200-, 500- and 1000-m performance. Practitioners could also predict LT2PO with reasonable accuracy based solely from a 1000-m time-trial; potentially providing an alternative, non-invasive, competition-specific protocol for threshold determination. Highlights The fact that LT1PO and LT2PO had very-large, to near-perfect positive relationships to 200-, 500- and 1000-m MMP suggests that coaches should consider these relative submaximal aerobic-fitness variables when evaluating the performance of sprint kayak athletes, regardless of their race specialty. While the SEE and 95% limits of agreement (95%LoA) values for the prediction of LT1PO may be too large to be practically meaningful, measures of LT2PO could be predicted with a reasonable level of accuracy based upon 1000-m MMP. The ability to inform athletes' LT2PO from a single 1000-m time-trial is advantageous since it would provide a more feasible, and time-efficient testing protocol within the athletes' training schedule compared to GXTs, potentially allowing coaches and practitioners to monitor changes in LT2PO, and subsequently review individual training zones, more regularly. Given that LT1PO and LT2PO derived from on-water GXTs had very-large, to nearly perfect relationships to 200-, 500- and 1000-m performance, practitioners may prefer to use on-water, rather than laboratory-based GXTs given their greater practical significance and ecological validity.