Bringing Power-Based Training to the Water
Why Power? The Case for an External Load Metric on the Water
Mechanical power output has been recorded in road cycling since the late 1980s, when the SRM system (Schoberer Rad Messtechnik) emerged as the first commercially available power meter. In these devices, strain gauges integrated into the crank spindle, the crank arm or the pedal allow the external mechanical load to be quantified continuously during training and competition, providing a direct measure of mechanical output that was previously unavailable. Over the following decades, power measurement became established across road, track, mountain bike and other cycling disciplines and changed how training is prescribed and monitored. The underlying rationale is that power quantifies the mechanical work done, whereas heart rate reflects the physiological cost of that work (Allen & Coggan, 2010; Leo et al., 2022). Sprint canoeing has so far lacked a comparable field method, and this article outlines how a stern-mounted, kinematics-based approach can close that gap and how the resulting data can be interpreted within an established physiological framework.
From a physical standpoint, mechanical power is the rate at which mechanical work is done. In a boat, only a fraction of the metabolic energy transformed through substrate metabolism is converted into mechanical work. The ratio of mechanical work to the metabolic energy turned over defines the gross mechanical efficiency of the athlete. The mechanical work that reaches the system is used to overcome hydrodynamic resistance and to accelerate the combined mass of athlete and boat.
To date, coaches and athletes in canoe sprint have relied largely on boat speed measured by GPS together with heart rate and blood lactate. Each of these measures has limitations that constrain the precise prescription of training intensity. Boat speed is an outcome variable that is strongly influenced by external conditions such as wind and water current, meaning a given speed does not correspond to a fixed mechanical effort. Heart rate exhibits a physiological latency relative to changes in mechanical output and is subject to cardiovascular drift during prolonged efforts and under thermal strain. It is also influenced by factors that are largely independent of mechanical output, including ambient temperature, hydration status and the intake of stimulants such as caffeine (Achten & Jeukendrup, 2003). Blood lactate sampling provides valuable insights into the metabolic state but is invasive and yields discrete, delayed snapshots rather than a continuous signal. Furthermore, the measured concentrations depend heavily on the chosen step duration and the specific test protocol (Heck et al., 2022).
These limitations are compounded by a more fundamental problem with the way intensity is commonly anchored. Recent critiques of intensity-prescription methods have evaluated the construct validity of different approaches based on their ability to elicit consistent physiological responses across different intensity domains (Jamnick et al., 2020).These expected responses include oxygen-uptake kinetics and blood lactate levels that match specific intensity zones. Research indicates that prescribing intensity as a fixed percentage of a maximal anchor such as maximum oxygen uptake, maximum heart rate or maximum work rate has limited merit for producing distinct or homogeneous physiological responses between individuals. For example, exercising at 80% of maximum heart rate placed 17 of 31 participants above their ventilatory threshold while the remainder stayed below it, meaning the same relative intensity corresponded to different intensity domains in different athletes. The same fixed percentages also span wide ranges relative to physiological landmarks. Critical power, for instance, has been reported to correspond to anywhere between roughly 60% and 95% of maximum oxygen uptake across individuals (Jamnick et al., 2020). Of the methods examined, critical power and critical speed had the strongest supporting evidence for demarcating the boundary between the heavy and severe domains, which represents the critical transition from steady-state, sustainable exercise to a non-steady state where fatigue accumulates rapidly and task failure becomes imminent (see Figure 2). For an external-load device this is the central point. Because power allows intensity to be anchored to a physiological threshold rather than to a fixed percentage of a maximal value, the prescribed training stimulus can be made significantly more comparable between athletes. This ensures that a group of athletes executing a session at a defined threshold percentage are all experiencing the intended physiological strain.
Measuring Propulsive Power from Hull Acceleration
Strain-gauge power measurements, as used in cycling, are difficult to transfer to a paddle on a daily basis. Instrumenting the paddle alters its mass distribution and swing dynamics, limits transferability between athletes, and introduces additional calibration requirements such as applying known static loads before a session. The PaddlePulse sensor instead uses a kinematics-based approach mounted to the stern. In water sports driven by discrete propulsive impulses, the hull alternates between acceleration during the stroke and deceleration during the glide. A high-frequency accelerometer records this acceleration continuously, and, because the combined mass of boat and athlete is known and approximately constant, the underlying forces can be estimated.
During the glide phase, the hull is decelerated primarily by hydrodynamic resistance. Treating drag as the dominant resistive force during this phase, the resistive force can be estimated from the product of system mass and the measured deceleration. The instantaneous propulsive force during the stroke is then recovered by adding this resistive force to the product of system mass and the acceleration measured during the active phase, and propulsive power follows as the product of propulsive force and hull velocity.
A practical challenge with accelerometry on water is the signal noise introduced by small waves and surface turbulence. The PaddlePulse sensor addresses this with physics-informed neural networks. In contrast to purely data-driven networks, these models incorporate the governing equations of motion as constraints, which helps separate the propulsive signal from environmental noise while keeping the output consistent with the underlying mechanics.
External Versus Internal Load: Complementary, Not Competing
Introducing power measurement adds an external-load metric and changes how internal and external load can be combined. The aim is not to replace physiological monitoring but to provide a reference for it.
A useful way to organise the limitations of current canoe sprint metrics is to consider each signal in turn, noting that none of them alone quantifies the mechanical work done. Heart rate remains a highly valuable tool because it provides a non-invasive, continuous measure of systemic cardiovascular strain. However, it is a response to a stimulus rather than a measure of the stimulus itself. It lags changes in mechanical output, drifts upward during prolonged or hot conditions at a constant workload, and is sensitive to factors unrelated to propulsion (Achten & Jeukendrup, 2003). Prescribing training by a fixed percentage of maximum heart rate is therefore problematic. (Jamnick et al., 2020). Blood lactate testing offers critical insights into the glycolytic contribution and the metabolic state. However, it only provides information at discrete time points and with a temporal delay. Furthermore, lactate is produced in the working muscle but is measured as a concentration in the blood after a delayed distribution phase. Without precise knowledge of an athlete’s body composition, such as total body water and muscle mass, interpreting these absolute concentrations is challenging. Consequently, prescribing training based on fixed lactate thresholds with rigid concentration values lacks physiological validity unless anchored to individual baselines (Heck et al., 2022). Boat speed measured by GPS reflects the outcome of the effort but is confounded by wind, current and water state, meaning it cannot be used to standardise the mechanical effort. Stroke rate, finally, is a kinematic descriptor rather than a load measure. Used in isolation, and without the corresponding mechanical power, each of these signals gives an incomplete picture of how hard the athlete is actually working and whether mechanical output is changing over time.
The consequence is that, when training is steered predominantly by physiological strain, athletes and coaches lack an objective reference for mechanical progression. If only strain is measured, it is not possible to tell whether more mechanical power is being produced to move the boat. If only boat speed is measured, a favourable wind or current can give the impression of improved fitness. Power provides the missing external reference. Rather than prescribing a heart-rate zone and assuming that the mechanical output follows, a coach can prescribe a defined mechanical output and then monitor the physiological response to it. When internal and external metrics are used together, they create a much more robust monitoring system. The external load dictates the exact mechanical demand, while the internal metrics reveal the precise physiological cost of meeting that demand. The prescription can specify a target power and an associated amount of mechanical work in kilojoules. The heart rate and, where available, lactate response can then be observed at that fixed external load to check for cardiovascular drift and to confirm that the intended domain is being targeted. Anchoring high-intensity efforts to a mechanical threshold rather than to a percentage of a maximal value has been associated with reduced between-athlete variability in the physiological response (Meyler et al., 2023; Muniz Pumares et al., 2025). Specifically, recent research has shown that for high-intensity exercise, prescribing intensity relative to critical power produces more consistent exercise tolerance and lower inter-individual variability in physiological responses compared to prescribing relative to a fixed percentage of a maximal anchor (Meyler et al., 2023). If, under standardised conditions and over several weeks, an athlete produces more propulsion for the same physiological strain, this is consistent with a genuine performance improvement (see Figure 4).
The PaddlePulse sensor also measures stroke rate continuously, which removes the need to time a small number of strokes by hand. Combining continuous power and stroke-rate data allows the mechanical strategy behind a given output to be examined, for example whether an athlete relies more on force per stroke or on stroke rate. The same data support continuous tracking of fitness changes, more accurate interval pacing, and retrospective analysis of races.