VO2 Measured to Modeled

Recently I was asked to compare how closely WKO modeled VO2 tracks against actual measured VO2. My last few articles looked at how measured VO2 tracked with HR for both continuous and intermittent workouts. So let’s look again at a few VO2max intervals and see how well the modeled numbers track to measured VO2.

VO2 is measured with the VO2 Master Pro wearable VO2 analyser.

Modeled VO2 in WKO is estimated based on real-time power and the athlete’s 90-day historical power-duration curve. The data below are visualized in WKO4 with some heavily customized charts. I’ll discuss the charts in much more detail over the coming weeks.

WKO5 was actually just released this week, and I’m already seeing some massive improvements in functionality. But I’m also seeing some bugs on my machine at least. I may have to wait to completely transfer over. I might be posting a mix of WKO charts for posts in the near future.


Hard-start VO2max Intervals

The first workout we’ll look at is the same 3x6min hard-start decreasing-power interval from back in December, 2018 that we looked at previously. Briefly, these are intervals designed to rapidly bring the athlete up to near VO2max with a high power start, then decrease power as the athlete fatigues to extend the duration sustained above 90% VO2max.

WKO4_2019-07-16_09-12-38INKED

3x6min Hard-start VO2max Intervals
Measured VO2 from VO2 Master Pro and Modeled VO2 from WKO4

  • Power in yellow
  • Heart Rate in red, highlighted above 90% HRmax
  • Modeled VO2 in Dark Blue (VO2 in mL/min) in the foreground, highlighted above 90% Modeled VO2max
  • Measured VO2 in Light Blue in the background, highlighted above 90% Measured VO2max
  • Time >90% HRmax and >90% VO2max reported at top right, along with R2 which shows how closely the two VO2 lines are correlated

First thing worth pointing out is the difference between measured & modeled VO2max. WKO modeled VO2max is estimated from the athlete’s 90-day historical power-duration curve. This workout was performed early in base training, so I expect my PD curve was not completely filled in. This would have resulted in a lower modeled VO2max estimate than what I was actually capable of in reality. This explains why the 90% VO2max targets occurred at different magnitudes for measured VO2 (5000 mL/min) and modeled VO2 (4270 mL/min).

Looking at absolute VO2 through the workout, we see good agreement between the VO2 lines during the warm-up and recovery intervals. This suggests that the model was accurately able to predict VO2 at submaximal intensities. The lines are also very close during the initial rise in VO2 at the start of each work interval.

Importantly, the times accumulated >90% VO2max (reported at top right) also show good agreement and track well to time >90% HRmax. Despite the different magnitudes for VO2max, both measured & modeled VO2 reported me above 90% VO2max for around the same duration.

The difference comes during the decreasing-power work intervals. Modeled VO2 is derived from power (which I’ll talk about more in an upcoming post) so as power decreased through the interval, modeled VO2 therefore also declined. However in reality measured VO2 remained elevated during these intervals, as you would expect for a continuous near-maximal effort.

Without going into detail again, this difference may have been exaggerated by my Left leg blood flow limitation. This condition significantly degrades the power I can sustain for high intensity efforts. So let’s take a look at a similar workout from another athlete.


Steady-power VO2max Intervals

This workout was done in May, 2019. It was performed outside on Cypress mountain on the North Shore of Vancouver. This is kind of amazing that we can collect VO2 data in real-time on the road, literally in a real-world environment.

These weren’t targeted to be specific VO2max intervals. They were steady-power 5min intervals at a specific target power above the athlete’s FTP. Based on the athlete’s subjective report that he “was worried about puking in my mask by the end” I think we can assume he approached VO2max during the workout! 😅

WKO4_2019-07-16_09-02-23INKED

3x5min Steady VO2max Intervals

Note, since we didn’t establish true VO2max with a ramp test, I will be using the athlete’s VO2peak, the highest 30-second VO2 value he achieved during the workout, instead of VO2max.

This athlete showed a much closer estimate for measured VO2peak (4850 mL/min) and modeled VO2max (4500 mL/min). This likely indicates that his PD Curve was more filled in and better reflected his true fitness.

We can again see that the major difference between measured & modeled VO2 was during the work intervals. While the decreasing-power intervals above show a steady-VO2 trend, the steady-power intervals here show an increasing-VO2 trend, as expected. This is caused by the VO2 slow component, which is basically a manifestation of fatigue and loss of efficiency during high intensity efforts above Critical Power or FTP.

Modeled VO2 derived from power isn’t designed to account for fatigue or the VO2 slow component during a constant work rate interval. As a consequence the model ends up over-estimating VO2 during the first interval, and under-estimating VO2 during the third interval.

There is also a difference in VO2 onset kinetics during the first two intervals. VO2 onset kinetics refers to the speed at which VO2 ramp-ups in response to a change in workload (demand). Especially during the first interval, measured VO2 was slower to rise than modeled VO2 predicted. Very interestingly, measured VO2 kinetics sped up through the workout, to be nearly identical to the predicted model by the third interval.


Looking Closer at VO2 Onset Kinetics

More recently I experimented with VO2 onset kinetics by going straight into a 5min VO2max interval without any prior warm-up.

One of the major reasons we perform a warm-up is to stimulate our aerobic system to work faster (ie. improved VO2 kinetics) and more efficiently. This ensures our body is primed to respond quickly to any change in workload (demand) with a rapid delivery of oxygen (supply) to the working muscles.

VO2 onset kinetics are markedly faster after a prior hard effort (ie. a priming effect, also see Bailey et al, 2008). So I decided to observe for myself how VO2 would react to an initial ‘cold’ effort, compared to subsequent ‘warm’ efforts.

WKO4_2019-07-16_09-17-16INKED2

3x5min No Warm-up Intervals

The first ‘cold’ interval shows the expected slower kinetics in measured VO2 and a lower peak VO2 compared to subsequent intervals. Modeled VO2 remains the same across all three intervals. Once again modeled VO2 is derived directly from power. The predicted rise in VO2 will be tied to the rise in power, regardless of all the internal and external factors (fatigue, fueling, temperature, etc.) that influence VO2 onset kinetics in reality.

Even considering the difference in magnitude of measured & modeled VO2max as previously discussed, I was surprised to see the differences in accumulated times >90% VO2max. The model simply thought I wasn’t going hard enough to achieve 90% VO2max at all. While measured VO2 was reporting that I was well above 90% VO2max during the two ‘warm’ intervals.

Heart rate shows an interesting comparison. Directionally it correlates better to measured VO2, with a gradual increase in peak HR during each subsequent interval. However HR barely reached 90% HRmax, and therefore time >90% HRmax correlated closer to modeled VO2 at essentially zero…

I think the most important take-away here is the value of a warm-up for getting the highest quality out of your interval training. Also the realization that VO2 response is not constant for repeated intervals even through a single workout session. This is a rationale for persisting through those later intervals during a workout instead of calling it quits. There still may be value to completing those final fatigued reps at a lower power target, as long as your body is still able to reach near VO2max. This is something I will collect more data on and discuss in the future.


Summary

  • Modeled VO2max is dependent on an accurate power-duration curve. This is no different than most of the other models in WKO derived from the PD curve. With enough relevant historical power data it seems reasonably accurate. At least from the few athletes I’ve measured.
  • Modeled VO2 derived from power during a workout appears to be fairly accurate at submaximal intensities. However as it approaches VO2max the model is unable to account for variations in internal metabolic state or external conditions. Such as fatigue onset through a workout, changes in gross efficiency, VO2 onset kinetics, temperature, fueling, hydration, etc.

Conclusions

I was hoping to conclude that modeled VO2 was ‘good enough’ or at least ‘better than nothing’ for use in training analysis and prescription. Unfortunately my opinion is that modeling VO2 from power alone is too inconsistent and insufficient to provide additional value. I think trying to manipulate power to estimate VO2 is a case of over-extrapolating the data and ending up with less valuable and less meaningful information as a result.

Power already provides us with so much meaningful and prescriptive information, but it is limited as the mechanical output of a chaotic internal metabolic cascade. Power alone tells us very little about internal energy production or how our bodies are working to produce that power.

Let me be clear, I love the debate around the value of power and heart rate. I currently come down on the side that HR, for all it’s limitations, has descriptive and prescriptive value as the only accessible measurement of internal metabolic state. Power of course has value as probably the best measurement of external mechanical work. I think it’s a clear win-win to use both together to build a more accurate picture of training effect and physiological response.

I’ll finish with an open question: could Power & HR be combined somehow to more accurately model VO2? I’ve talked before about how I don’t quite fully understand the Power-HR-VO2 relationship and how all three components interact. Could combining power & HR compensate for eachother’s limitations to produce a more reliable estimate of VO2?


Next post I’ll look at some intermittent intervals. Then I want to go into more detail on the WKO4 (and soon WKO5) charts I’m working with for VO2 & SmO2, and bring estimates of anaerobic power into the discussion. I’m already writing those articles so it shouldn’t be another 3 months before the next post!

2019-06-09 12.24.27

Thanks for sticking with me so far!