Aerobic Systems

But now, sadly, after that all-out minute we are going to have to draw breath, because we need the oxygen to power the aerobic energy systems. First up we get aerobic glycolysis, this is converting glucose into pyruvate by burning it with oxygen in a really complicated 10 stage cycle. The conversion rate is limited by the amount of oxygen the lungs can absorb (VO2max) and the available fuels. It can take anywhere between 1-3 minutes to get up to ‘peak’ production and then dies away slowly over time. Once all the glucose is gone, we will bonk, which is why gels and powders are high in easily digested glucose – to refuel this process. Lastly, from about 6-7 minutes we start to rely upon lipolysis that utilises an almost limitless source of energy; fat and water. So stay hydrated!

The Future of Power-Duration models

Our Extended Power Duration Model extracts the likely contribution of these energy systems to predict the energy production (or watts per second). This is akin to reading the fuel gauge to work out how fast you’re going in a car. It’s not an exact science and so yields an approximated answer, which can be slightly overestimated because it doesn’t really consider why we fatigue.

It is likely that in the next 2-3 years current research will help to explain muscular, neural and psychological fatigue or constraints. These in turn can be used to refine our models. Research is also likely to expand our understanding of W’ and CP and how they reflect underlying physiology and associated dynamics (maybe even CP fluctuates depending upon how we ride).

The speed that a muscle fibre contracts is known as twitchiness; we have type I known as slow-twitch fibres that, true to their name, don’t contract quickly but can keep doing it for a long time and a type II also known as fast-twitch that are, unsurprisingly, very quick, but also fatigue more quickly too. Those fast-twitch fibres are further broken down into type IIa and IIx/d. Typically, slow-twitch reach their peak tension in about 100ms, with IIa at 50ms and IIx/d at 25ms. The other important distinction between the fibre types is the force that they apply; slow-twitch type I fibres apply a low force, where fast-twitch type IIa apply a medium force, whilst IIx/d are unsurprisingly the strongest. We typically have very few IIx/d which is why we can exhaust them with relatively few repetitions of an exercise when weight training.

Slow-twitch fibres contain a high number of mitochondria; often referred to as cellular power-plants. They 'generate' energy on demand in that complex 10-step process mentioned above (its actually called The Krebs Cycle). In contrast, fast-twitch muscles contain far fewer mitochondria and instead have greater stores of glycogen and the enzymes needed to to produce energy without oxygen. As a result, slow-twitch muscles are fuelled primarily from fat at endurance intensities, but will utilise glycogen at tempo and higher intensities. IIa are fuelled primarily from glycogen but can utilise fat whilst those strong and quick IIx/d only use glycogen.

In healthy adults the distribution of type I/II differs significantly by muscle group and also to a lesser extent upon your genetics but a typical basal distribution of fibre types in the quads would be in the range 55% I, 32% IIa, 11% IIx and 2% IId. With the right kind of training it is possible to 'convert' type IIa to type I which improves aerobic endurance performance but at the cost of a loss of some strength. This is typically achieved through lots of hours riding at a lower intensity, below LT1. Explosive and high intensity workouts will typically signal greater type II muscle fibre growth. So the worst thing for a sprinter to do is ride a 3 week multi-stage race like the Tour de France!

Recruitment and Central and Peripheral Muscle Fatigue

As we increase the force we want to generate our brains will recruit more and more of the motor units to meet the demand. As the demand gets higher we reach a point where all motor units available will be firing. When fresh and all motor units are available we will hit our P-max power, but if we’re fatigued some of those muscles will be exhausted and unavailable – and our maximum power will be reduced. This peripheral fatigue occurs much earlier for fast-twitch than slow-twitch muscles. So our brain will always recruit from slow-twitch before fast-twitch muscles to meet demand – so the fast twitchers are saved for when we really need them.

Aside from fatigue within the muscle fibres, our brains and neurons will also limit what we can achieve; in some cases this keeps us well within our natural capabilities, perhaps as a ‘safety mechanism’. And there is lots of research being done into this central fatigue to see how if differs by individual and how much it can be changed with training and the right ‘motivation’.

Lactate Threshold

Lactate is always produced even at lower intensities of exercise. Initially our blood flow will clear lactate away as it is produced to the liver, heart, kidneys where it is slowly converted and stored as fuel for re-use. Additionally whilst we are working at these lower intensities some of the lactate produced is converted back into glucose within the muscles themselves (which also helps to clear lactate when we rest or "lift off the gas for a moment"). The reason this only occurs at lower intensity is because it is the slow-twitch muscle fibres that contain a transporter called MCT-1 that controls lactate re-use within the fibre mitochondria. And slow-twitch muscles will all be busy when we exercise at higher intensities.

As we work a little harder lactate will be created a bit faster, but at the same time blood flow increases our heartrate goes up so we keep clearing it. But eventually we get to a point where lactate levels increase above the baseline (typically 1mmol above), this point is known as “LT1”. At this point we will feel that we are working, but no more than a tempo pace. As we continue to go harder, blood lactate accumulation will increase and so will blood flow as our heart rate rises. Performance at the LT1 point has been shown to be an excellent predictor of performance in endurance races like the marathon or a cycle race lasting two or more hours.

We will eventually get to a tipping point where clearance and accumulation will be at a maximum point we can sustain; this is the intensity that best relates to a TT pace and is called “LT2” or the lactate turnpoint (LTP) and more technically referred as the maximal lactate steady state (MLSS).

The power output at this point has been shown to be closely related to FTP and CP (although CP is typically a bit higher). From here if we go harder then lactate will build up much faster and we will start to feel a heavy burning sensation in our legs. Eventually we will crash and burn as we hit our maximum HR and can’t get enough oxygen in, let alone clear the lactate in our legs.

Performance at the LTP has been shown to be a good predictor of performance in shorter events like the 10KM or a cycling 40km TT, with most athletes able to hold power at the LTP for between 45 and 65 minutes. It is not such a good predictor of performance in events of a longer duration — hence CP and FTP are not good indicators of endurance performance but as they change it will indicate if the lactate curve is shifting to the left or the right.

Shifting the lactate curve to the right

So, if we can shift the blood lactate curve to the right we can exercise harder for longer at the same level of perceived effort. Or as Greg Lemond famously once said 'It doesn't get easier, you just go faster' — we will still hit the LTP, just at a higher power output.

In order to do this we need to train our bodies to to burn less glucose for fuel, get better at shuttling pyruvate into muscle cells before resorting to producing lactate, and once we have lactate we need to get better at clearing it away or reusing it for fuel.

So, increasing the volume and density of mitochondria within the slow-twitch fibres will give us a much greater capacity to re-use pyruvate and less lactate will be produced in the first place. Secondly, these mitochondria will also help in clearing and reusing lactate. Lastly, fat metabolism doesn’t create lactate at all, so the greater power we can develop solely from this (again using our slow-twitch muscles) the less reliance we will have on glucose energy and lactate clearance.

So, training interventions that increase the volume and density of slow-twitch fibres and mitochondria will shift that curve to the right and improve endurance performance. Typically this is the purpose of 'long slow distance' where we ride below LT1 at an 'endurance pace' for many hours.

Your VO2max is largely determined through genetics; you won't become Greg Lemond (92.5) or Flavia Oliveira (76) if you work really hard. Other factors will also affect it; it tends to peak when you are about 20 and can drop by 30% by the age of 65; at altitude it is reduced due to the thinning of oxygen in the air. But VO2max can be improved with the right sort of training interventions and weight management and it remains the best way of tracking improvements in aerobic fitness as well as comparing athletes and determining their likely potential.

For those that don't own a gas exhange analyzer, HR may be an alternative way of tracking changes. There have been numerous studies that show that HR and oxygen consumption are closely correlated; so it is potentially viable to monitor average power to average HR ratios to track trends in aerobic fitness over time. But take care as HR can fluctuate day to day depending upon hydration, caffeine, sleep and other factors.

"The Slow Component"

When we exercise at a constant intensity below LT1 (moderate domain) oxygen uptake rises over about 1-2 minutes until it reaches a steady state level that is well below VO2Max.

When we exercise at a slightly higher constant intensity, between LT1 and LT2 (heavy domain) oxygen uptake will rise over 10-20 minutes before reaching a steady state. When compared with the moderate effort the heavy effort causes oxygen uptake to rise more slowly and appears to be delayed. Additionally, the percentage of VO2max that we settle at is higher than you would predict; suggesting efficiency has been impaired in some way.

As you might guess, when we exercise at an intensity above LT2 (severe domain) oxygen uptake, like lactate accumulation, just keeps getting higher until we have to stop due to accumulated oxygen debt and excessive lactate. It does not plateau or reach a steady state.

The increase in oxygen uptake as time progresses in the heavy and severe domain is known as the ‘VO2 slow component’. It suggests that the efficiency with which the body uses oxygen to produce energy is progressively lost while exercise continues. It has even been shown that if exercise is continued at the same intensity for long enough we will eventually reach VO2max.

The cause for this is not really known for sure. It could be caused by the gradual recruitment of fast-twitch fibres as slow-twitch fibres fatigue; as we run out of slow-twitchers the brain uses more and more fast-twitch muscles to maintain the same power. But those fast-twitch muscles need more oxygen to generate the same power. So slowly, our oxygen uptake increases.

Either way, for endurance athletes, we need to shift the LT1 and LT2 as far to the right as we can to enable us to work at a higher intensity or power so that what might have been severe becomes heavy, and what was heavy may become moderate. Greg Lemond was only half right; 'It might still be hard, and you might go faster, but you can go faster for even longer when you go easy.' That may not sound so snappy but it sure is the most important thing to know about endurance training — its all about shifting the lactate curve to the right.

Regardless of this, stroke volume is most definitely improved with aerobic training; the size of the ventricles will increase with the right training, and as they become thicker and stronger they make larger and more powerful contractions. In cycling power terms that means we will see power output increase at the same heartrate as more blood is pumped with each beat.

Blood vessels – Arteries, Capillaries and Veins

There are three main types of blood vessels; the arteries that carry oxygenated blood away from the heart, veins that carry de-oxygenated blood back to the heart and the capillaries that provide the interface with tissues and muscles. Arteries and veins are flat muscle; they have a layer of muscle surrounding them that contracts and expands to help pump blood around the body. With training their overall performance will be improved along with some increase in capillary density. Astonishingly, laid end-to-end our blood vessels would stretch 100,000km.

Their role in circulation is managed by the central nervous system and will prioritise blood delivery in response to exercise (amongst many other things). For example when cycling at a moderate intensity, this will result in more blood going to working muscles in the lower body (our legs) whilst leaving other parts of the body (e.g stomach) less well supplied. In fact, at rest only 15% goes to working muscles but that can increase to as much as 60-7% during intense exercise. Which is why its hard to digest an energy bar when you're gunning for it!

Blood – Red-Blood Cells and Blood Plasma

You typically have about 5-6 litres of blood in your body at any one time which is of course 5-6 kg of vital weight; certainly not weight you can afford to lose. It is comprised of 55% blood plasma and 45% red blood cells. Blood plasma contains mostly water, sugar, protein and fats used to fuel exercise. Red blood cells carry mostly oxygen and CO2, so increasing the volume of blood and the percentage of red-blood cells through training (or doping) can have a dramatic effect on aerobic performance.

99% of oxygen is delivered to working muscles attached via haemoglobin (Hb) in red blood cells, with the other 1% dissolved within other blood fluid. In contrast, 85% of CO2 is carried away from working muscles dissolved in the water inside those red blood cells, with a further 10% carried away by the haemoglobin that was used to transport the oxygen in, the remaining 5% just dissolved in the blood plasma.

Haemoglobin and Oxygen

O2-Hb Affinity
The affinity between oxygen (O2) and the Haemoglobin (Hb) in red-blood cells is used to describe the attraction between the two. A higher affinity means more O2 will be bonded to the Hb, when low it means those bonds will break and O2 will be released. As red-blood cells pass through the lungs it is important for this affinity to be high so the blood becomes oxygenated; but as it passes through the legs it is important for this affinity to be low so oxygen is released into the working muscles.

During intense exercise there is a significant difference between the lungs and the capillaries where temperature and levels of blood pH, CO2 and phosphates are much higher. These reduce the O2-Hb affinity in the tissue the capillaries service causing the O2 to be released in the muscles where they are consumed. But there is also some reduction in the lungs too during intense exercise, this limits the amount of oxygenation that can occur. Worse, as total blood flow is also increased during intense exercise there is less time for the oxygen to enter the blood; all of which leads to a situation where blood is passing through the lungs faster than it can be fully loaded with oxygen — we have reached our limits; any increase in blood flow isn't going to deliver more oxygen to our muscles.

NIRS, SmO2 and tHB, O2Hb and HHb
Using near-infrared spectroscopy (NIRS) devices like the Moxy Muscle Oxygen monitor it is now possible to monitor oxygen delivery and consumption as we ride. NIRS shines a light through the blood in the capillaries inside muscles to identify the amount of haemoglobin present, and what percentage of that haemoglobin is carrying oxygen.

NIRS devices provide two measures; SmO2 — what percentage of haemoglobin is carrying oxygen at the muscle; and tHb — haemoglobin concentration measured at the muscle. Using these two pieces of information we can derive two further metrics; O2Hb — concentration of oxygenated haemoglobin and HHb — the concentration of deoxygenated (hydrogenated) haemoglobin. We can then plot HHb and O2Hb alongside, say, power and heartrate to analyse oxygen delivery and extraction during exercise right at the working muscle!

Use of this data to assess training and development is an exciting new development that may yield entirely new training and analysis methods in the very near future. For example; there is a direct relationship between oxygen extraction at the muscle and the Lactate Turn Point; we could use data collected from an NIRS device with a power meter during an incremental ramp test to pinpoint power at MLSS with some precision. This could provide a reliable and accurate protocol for establishing CP

Stress – Work, Intensity, TSS and TISS

Given that work in joules can be calculated by multiplying power by time it is very tempting to use this to measure the stress of a ride. But as we get stronger and more efficient those joules become easier to produce, and thus the training stress accrued in the workout should reflect that.

To account for this we need some kind of score that takes into account how hard the ride is based upon our current capability. This is precisely what BikeScore and TSS do. They reflect the stress by taking into account the relative intensity of the workout. This intensity factor is computed as a ratio of the xPower to our current CP. This intensity is then multiplied by the ride duration to get an overall stress score; the higher the stress score the bigger impact it will have had and likely the more recovery we will need the day after.

But there is still a problem, we know that work at high intensities for short durations elicits a different strain to work at low intensities for longer durations and there comes a point where more pain will give little gain. To counter this Dr Skiba introduced Ae and An TISS that are weighted differently for low and high intensity work and allow us to track these training stresses separately.

The Impulse-Response Model

Just under 40 years ago, in 1976, Eric Banister and colleagues introduced an ‘impulse-response’ model that summarized the principles of training into an elegant and coherent framework that is individualised to the athlete.

Daily training load (TRIMPS) is used as an input to the model, but could be replaced with any metric, including power-based metrics like BikeScore or TSS. These represent the training impulse of the impulse-response model. Next, to adapt the model to the individual a ‘gain’ factor is derived from past responses to training impulses; or more specifically, the past improvements in performance (e.g. 5 min power) and the training stress are observed and quantified. We can look at the gain factor to identify if the athlete is a low or high responder, or likely somewhere in between. This gain factor helps to link the training impulse to the predicted response.

These load and gain factors are then used as an input into two functions that quantify the response; fitness (or PTE, positive training effect) that accumulates over the longer term (50 days) and fatigue (or NTE, negative training effect) that accumulates over the shorter term (15 days).

The output from the IR model is then a prediction of ‘performance’ that is computed by subtracting ‘fatigue’ from ‘fitness’.

Importantly, these functions will also reflect de-training too – if your training load eases or stalls your fitness and fatigue will reduce; and changes in your performance will either increase slightly if it is a short-term taper, or drop if it is a longer term training break. Lastly, Banister recommended resetting and checking the gain and time constants every 60-90 days; by fitting recent performance outcomes with the recent training load. More recent studies have confirmed that these constants will change over time, but will also differ according to training intensity and training modality.

The Performance Manager

In 2001, working with a group of coaches and athletes interested in training with power meters Dr Andrew Coggan developed the ‘Performance Manager Chart’ (PMC). It is a variant of the Banister IR model and is described in a comprehensive article on the TrainingPeaks website .

The PMC is claimed to address a number of shortcomings of the Banister IR model that; (1) it is not tied to physiology (2) it assumes there is no upper limit to performance (3) fitting model parameters every 60-90 days requires valid data to model against (4) it is over parameterised (5) these parameters can vary by individual, intensity and sport. It is debatable whether these perceived shortcomings have any material impact on the utility of the IR model or if they are addressed by the PMC. But it is clear that the PMC has been embraced by the cycling community and has been instrumental in providing a means for the layman (and many professional coaches) to adopt an IR approach to managing their training. It is often described as the most important tool for the cyclist lucky enough to own a power meter.

The PMC does not support individualisation for low/high responders and it does not attempt to predict performance, but the core concepts from the Banister model remain; (1) there are functions that attempt to describe fatigue and fitness (2) training load is used as an input (3) training load can be represented using any metric (4) decay is essentially the same (maths differ only slightly in practice), with default values of 42 and 7 days used as opposed to 50 and 15 (but these can be adjusted anyway).

ATL, CTL and TSB

Acute and Chronic Training Load (ATL, CTL) represent the accumulated training stress (TSS) in the short-term and long-term respectively. As long-term training load (CTL) increases one would expect the accumulated stress to have elicited positive adaptations as a result CTL is widely used as a proxy for ‘fitness’, assuming a linear relationship between impulse-response. Rapid increases in short-term load (ATL) (often called the 'ramp rate') tend to indicate that training load is increasing too fast to sustain, but this is also indicated by the third metric Training Stress Balance (TSB) which is computed as the difference between the long and short term loads.

Where the Banister-IR model output predicts ‘performance’ the PMC model provides an output TSB that will indicate the athlete status; a negative value indicates a level of residual fatigue whilst a positive value indicates no residual fatigue is present or the athlete is rested. It is often claimed that a TSB of between 0 to +5 represents an ideal state to achieve your best performances and should be a target when planning training. A TSB lower than -35 to -40 indicates a risk of injury or illness through overtraining or insufficient rest (assuming you have more than 42 days of data).

Essentially, TSB indicates if the athlete is likely to perform well, whilst CTL indicates how much training they have performed (volume, frequency and intensity combined) and, therefore, their likely current ‘fitness’.

Crr, Rho and Friends

Aside from CdA there are a number of other factors that will affect how fast you go for any given power output.

Given we spend so much effort pushing air out of the way it should come as no surprise that the density of the air ( Rho ) can make a massive difference to how fast we go for any given power output. Air gets thinner as you go to altitude, its why hour records might be attempted there (lets ignore the fact there is also less air to breath). Aside from altitude, air density is also affected by humidity, temperature and air pressure; we can calculate the air density if we have all three of these.

Pushing air out of the way isn't the only thing you pedal against, the tyres on the road have a coefficient of rolling resistance or Crr ; even skinny road tyres might have a range from 0.0025 up to 0.005. Luckily there are lots of folks testing them so you don't have to. But changing tyres really can make you faster (or slower).

Remaining factors include; weight if you're riding on the flat or downhill then extra weight can be advantageous as momentum and gravity help you go faster; but as the road tilts upwards its gonna need more power to overcome. Typically, on a 2% slope an 80kg bike and rider will need 233w to maintain 25 km/h, every 1kg of weight extra costs another 2w to go the same speed. Similarly for 17km/h on 5% and 10km/h on 10% every kilo will take 2-3w of power to lift to the top.

And of course, wind is the most obvious problem. Riding with a 20 km/h headwind or sidewind is no fun; but riding with a 20km/h tailwind is great! So windspeed and just as importantly wind direction ( yaw ) can have the biggest impact on how fast we can go for any given power. Lastly we have acceleration ; every time you speed up you use power to do that, unless you're rolling downhill.

Virtual Elevation - aka The Chung Method

Ultimately we all want to get faster on the bike. Assuming you have done all you can to shed unwanted pounds there really isn't much you can do to change the wind, air density the course profile or gravity. That leaves our tyres (Crr), bike and posture (CdA) to work on. To avoid spending lots of money on time in a wind-tunnel there is a practical approach called 'Virtual Elevation' (VE) devised by Dr Robert Chung that can be done outside using a power meter and speed sensor.

In the past, in order to test position and equipment and calculate our CdA we needed to know accurate values for; weight, speed, windspeed and yaw, power Crr, Rho, incline, gravity and acceleration. So a field test would typically be performed on a still day on a flat road; removing the need for the windspeed, yaw, incline and gravity terms. Then looking at speed for each run it would be possible to check if a position was faster or slower. But riding without wind and hills was almost impossible to do outside of a velodrome. And even then velodromes have problems because (believe it or not) riding around the track you (and others there at the same time) will create your own tailwind!

How Virtual Elevation works

The single most important thing we do is to run multiple loops on the same course with a power meter; every run will have the same overall elevation change (none), same distance and experience the same environmental conditions whilst the power output and speed will vary.

Because wind can change direction or bluster it is still a good idea to perform these tests in a sheltered environment on as windless a day as possible. We need to eliminate it from our calculations.

The effects of slopes, gravity, air density will be the same for each run; we have not eliminated their effect by riding a loop but we have made them identical for each run. We can also assume that as a rider we weigh the same in each run. But we need to make sure we don't brake, lose air from our tires or change position, because none of these things are going to be taken into account.

If we do this then the power we used for each lap was used to overcome;

• rolling resistance in the tyre (Crr)
• elevation changes (slope changes)
• accelerations (speed changes)
• air resistance (CdA)

This can then be converted to a relatively simple formula to calculate power used based upon Crr, Cda, speed and accelerations, gravity and slope, acceleration, weight etc.

The clever bit is what Dr Chung does with this formula; it is solved for slope instead of watts. So we end up with a formula that combines all of those opposing forces into a virtual slope we had to ride up and down to get around our loop. Hence the name 'Virtual Elevation'.

We can then make some educated guesses about what Crr and CdA were and plot the associated virtual profile. Now since each lap is performed in a single position and the physical elevation change at the start and end of a loop is zero we need to adjust Crr and CdA until the start and end of a lap in the VE plot are the same.

Estimating using Virtual Elevation

The example shown to the right (courtesy of Dr Chung) shows a field test of 7 laps where the rider had his hands in one position for the first several laps then changed hand position part way through the test.

When the estimate for CdA and Crr are correct the VE plot for a lap will show the start and finish point at the same elevation (i.e. they will be level). We can see that the top left plot is clearly wrong as each lap finishes higher than it started; the CdA estimate is too low. The top right shows the CdA has gone up but still each lap finishes slightly higher than it started. Its only in the bottom two plots that we can see a level start and end for any given lap; those are the laps that were performed with the associated CdA and Crr.

In fact, the exact point at which the rider switched his hands from one position to the other is easily spotted – two-and-a-half laps from the end. The change in hand position was actually quite small: the first 4 laps were with the hands on the bar tops, the last two-and-a-half laps were with the hands on the brake hoods. The wind conditions were not quite calm (though the wind was neither strong nor blustery) so this example shows that small differences in aerodynamics can be spotted even under non-ideal conditions. Of course, the better the conditions, the fewer the laps and the more precisely and reliably you can pin down the differences.

This is what Aerolab in GoldenCheetah does; it plots this virtual elevation from a ride as you adjust estimates for Crr and CdA until you can see a good fit for the elevation profile. If you have sufficient laps and variations in positions you will be able to determine which lap yielded the best results – and thus identified a good position and its associated CdA.