Validity of a novel device for indoor analysis of cyclists’ drag area

Abstract Purpose: Cyclists need to measure aerodynamic resistance accurately and reliably, as well as economically. Devices such as Notio Aerostick, an equipment device that includes one pitot tube, have appeared for this purpose. The aim of this study is, therefore, to test the reliability and degree of agreement in the evaluation of the CdA (coefficient of aerodynamic drag), assessed by means of the Notio Aerostick compared to the Virtual Elevation (VE) and Martin mathematical models. Method: Seventeen professional cyclists rode in a 250-metre-long velodrome covered with a concrete surface with their own time trial bikes. Each cyclist completed three rides of 15 laps at constant speed for the evaluation of the CdA, each of them in a different position [Baseline (B), Change 1 (C1) and Change 2 (C2)]. Results: The differences in CdA between methods were found for Martin in comparison with VE in all positions (p <.001) and with Notio Aerostick in B and C2 (p> .05). About differences of CdA for each method, considering between position changes, the results were the same for VE and Martin, but different for Notio Aerostick. Conclusions: Findings suggest that, notwithstanding Notio Aerostick is valid if we compare CdA values with respect to VE, since the direction of their between-positions CdA changes differs, the results of their aerodynamic evaluation could lead us to recommend different final setups. We need studies that evaluate different units of the Notio Aerostick device as well as the reliability and precision of each sensor that includes Notio Aerostick. Highlights The CdA calculated by the Notio Aerostick and VE, a mathematical model previously validated, can be interchangeable, however the final position recommended by each method may be different, since the changes in the following position are given by the changes of the CdA in the previous position. None of the three methods allowed elite cyclists to measure statistically significant differences between the proposed setups. Although the CdA differences between positions were not significant, they can be decisive in the final result of a time trial competition.


Introduction
In recent years, interest in the study of aerodynamics in cycling has been increasing (Malizia & Blocken, 2021). Composed of different categories of resistance such as form, frictional, induced and interference (Jobe, 1984), aerodynamic drag becomes a key factor, especially on flat stages and time trials where speeds of over 40 km/ h can be achieved (Martin et al., 2006). At this speed, 96% of the total power required is used to overcome air resistance (Martin et al., 2006).
Knowing how, and at what magnitude, resistive aerodynamic forces are generated when moving forward, enables us to take the necessary measures to optimise the aerodynamic component generated by the cyclists during pedalling (Jeukendrup & Martin, 2001), improving their performance. Among the techniques described for assessing this aerodynamic resistance in cycling, the wind tunnel is the gold standard; it is a reliable and reproducible method, although limited by accessibility and cost (Brownlie, 2019). Furthermore, the conditions simulated in a wind tunnel are not quite the same as those experienced by a cyclist in a velodrome or on the road (Malizia & Blocken, 2020).
A second approach is Computational Fluid Dynamics (CFD), which is very popular in research and other sports such as Formula 1, but this is equally difficult to access for most athletes; with a high cost when it comes to processing and interpreting data reliably (Malizia & Blocken, 2020). Alongside these is the so-called Ring of Fire, used to assess the punctual aerodynamic drag, although this is not very popular as it requires specific equipment that is not very accessible (Spoelstra et al., 2019). Finally, there are also field tests such as deceleration tests, constant speed tests and tests of section. All these use formulas that are solved according to the principle of energy conservation, using variables measured by different sensors that are adjusted with fixed values reported in the scientific literature (Malizia & Blocken, 2020).
In view of the aforementioned popularity of these sports, several methods have recently appeared with the aim to measure aerodynamic drag in a reliable and accurate, but also practical and affordable way. These include mathematical models, such as Virtual Elevation (VE) (Dyer & Disley, 2018) and the Martin's Model (Martin et al., 2006), but also aerosticks devices such as the Notio Aerostick (Argon 18; Montreal, Canada), AeroLab.tech (Strathcona Calgary, Canada), and Velocomp LLC (Stuart, Florida, EE.UU). The main feature of this devices is that it includes a pressure sensor capable of measuring wind speed, which together with other external sensors resolve an energy balance equation to obtain a composite value that includes the aerodynamic drag coefficient multiplied by the frontal area; that is, the so called coefficient of aerodynamic drag (CdA) (Kordi et al., 2021;Valenzuela et al., 2020). In the case of aerostick Notio Aerostick, it uses its own equation that has not been published. With regard to the mathematical models VE & Martins, and their energy balance equations, further details can be consulted at references (Dyer & Disley, 2018;Martin et al., 2006) respectively.
To our knowledge, only two articles have been published to date that analyse the reliability and validity of the Notio Aerostick device (Kordi et al., 2021;Valenzuela et al., 2020). Neither of these analysed its sensitivity for detecting position changes when assessing CdA in an aerodynamic assessment protocol, and in a real context with professional cyclists. The aim of this study is, therefore, to test the reliability and degree of agreement in the CdA evaluation of the aerostick Notio Aerostick (Kordi et al., 2021;Valenzuela et al., 2020) compared to the VE (Dyer & Disley, 2018) and Martin (Martin et al., 2006) mathematical models, both previously validated with the gold standard wind tunnel method. Changes in positional adjustments are considered for further comparisons. As main hypothesis, the authors expected a high grade of agreement and notable associations between the aerostick Notio Aerostick and the mathematical models.

Participants
Seventeen professional cyclists (63.5 ± 8.1 kg, height: 175 ± 9 cm) voluntarily participated in the study. The inclusion criteria were having extensive time trial cycling experience and being a professional cyclist. Before the participation, they were informed of the purpose and risks of the study and provided with a written informed consent form. None of the cyclists had any injuries or health problems. Written informed consent was obtained from each participant. The study was approved by the Ethics Committee of the University of Valencia (Valencia, Spain) and was conducted in accordance with the Declaration of Helsinki (2013) and the international research standards (Harriss, Macsween, & Atkinson, 2019).

Study procedure
The recording process was conducted in a 250 m-long concrete-surfaced indoor velodrome (Tafalla, Navarra, Spain), where the cyclists rode their own time trial bicycles. The participants were weighed together with their bikes and equipment at the start of the test. Each rider completed three rounds of 15 laps at a constant speed to assess the CdA, each time in a different position; Baseline (B), Change 1 (C 1 ), and Change 2 (C 2 ).
In all cases the 15 test laps were preceded by 12 additional laps to calibrate the Notio Aerostick device, with the riders maintaining the same position and speed at which the test was performed, according to the manufacturer's instructions. In addition, 2 launched laps were included between the 12 calibration laps and the 15 test laps to reach the target speed and position. A slightly lower speed than competition speedbetween 40 and 45 km / hwas chosen, in order to favour a stable position that would not be compromised by the intensity of the exercise. The rest time between tests was 10 minutes, which was used to make the appropriate changes to the position of the bike.
The speed of the bicycle was measured using a speedometer placed on the front wheel. Power was measured with a Power2max powermeter (Saxonar GmbH, Walhufen, Germany) installed on the cranks. The potentiometer was calibrated to zero before the measurement. Riders received speed feedback through a cycle computer Garmin 520 installed on their bike. At the same time, we checked on the track that the lap times were as indicated.
The instructions given to the cyclist were that, in each position, they should not change gear; they should maintain a stable speed and position on the bike, and they should try to ride the entire test above the black line painted on the concrete that delimits the length of the velodrome. The position changes were achieved by modifying the height, reach and width of the handlebars and elbow-rests, as the commercial contracts for professional cyclists limited any possibility of changing the material. On the other hand, when testing sensibility, it may be a limitation of the study the need to make all the position changes within the UCI regulations. Each change was made taking into account the initial position of the cyclist, looking for the best possible improvement in each of them in terms of reducing the frontal area without compromising stability and comfort. We took into account the effect of changes in the CdA caused by three positional adjustments.
Finally, temperature, humidity, partial pressure, and atmospheric pressure were continuously recorded by the Notio Aerostick device. As a result, the average weather conditions were: 20°C ± 2.6 temperature, 51.9 ± 4% humidity, and 1193 ± 11 kPA atmospheric pressure.
Aerodynamic drag coefficient assessment As indicated above, three methods were used to analyse the changes in position, with the aim of studying the degree of agreement between them.
Notio Aerostick (Argon 18; Montreal, Canada) is an aerostick device that incorporates a differential pressure sensor, a barometric pressure sensor, an accelerometer and a gyroscope. The data provided by these sensors, together with external speed and power sensors, and an own energy conservation formula, allow the Notio Aerostick to estimate the CdA. The data recorded in the test must be transferred from the device to a computer for processing with a specific programme, based on the original software of Golden Cheetah.
The VE method (Chung, 2012) uses a mathematical model incorporated into the Golden Cheetah training analysis software through the "Aerolab" app. This method was originally intended to calculate the result of two unknown drag variables, the Coefficient of Rolling Resistance (Crr) and the CdA, by recording at different speeds. However, in this case, the Crr value was incorporated into the system as a known measure -see below-.
The third system is a mathematical model proposed by Martin, specifically for track cycling. The special feature of this formula is its inclusion of the variable height of the centre of mass, to account for the centripetal force experienced by the cyclist in curves, which causes the speed of the bicycle to differ from the speed of the centre of mass (Martin et al., 2006).
Hence: Crr was set at 0.0025 for all three models in accordance with previous research involving concrete velodrome surfaces (Martin et al., 2006). We can find different Crr values in the literature, such as those of 0.003 reported by Grappe (Grappe et al., 1997) also for velodrome, although in this case, as the main objective of the study are to report changes in the CdA, it is not so important that the value are exact, as this is not modified throughout the trials.
Importantly, we ensured that temperature, speed, and weight, the three factors that can modify the Crr, were constant throughout the test. More specifically, the transmission system efficiency was set at 97.7%, in line with previous research on components of wearfree transmission and well-lubricated (Martin et al., 2006); the centre of mass height was considered as the distance between the top of the saddle and the ground (Martin et al., 2006). The wind speed for the VE and Martin mathematical models -as it was an indoor velodrome-, was considered zero, following the indications of other authors (Underwood & Jermy, 2010a, 2010b, while for the Notio Aerostick model it was calculated by the partial pressure sensor, according to the manufacturer's indications. When calculating the formula, the Notio Aerostick system includes acceleration; in the mathematical models of VE and Martin this acceleration was considered to be zero, since the instructions given to the cyclist were to maintain a stable speed at all times.
To conclude, the rider fatigue was not considered since the duration of the intervals at the target intensity is below the limit performance of our elite cyclists.

Statistical analysis
The statistical analysis was performed using a spreadsheet and SPSS software (SPSS 25.0, Inc., Chicago, IL). The alpha significance was set at 0.05. The measurements were expressed as 95% confidence intervals (CI). To ensure the normality of the sample, the Shapiro-Wilk normality test was performed. A repeated measures analysis of variance (ANOVA) was performed to determine whether there were significant differences in speed and power between the three positions. A second repeated measures ANOVA was performed to examine the CdA differences between the methods for each change of position, followed by intraclass correlation (ICC) analysis and the degree of agreement of the CdA values for the 15-lap stretches for all positions for the three methods, which were plotted on a Bland-Altman graph using RStudio software (Version 1.4.1103). Additionally, the differences between the CdA results for each method between position changes were calculated, as were the Pearson correlation coefficients (r) of these CdA differences, and Typical Error of Measurement (TEM). Finally, a linear regression was run to determine the predictive ability of the different measurement systems with respect to one another.
The calculation of the CdA showed significant differences depending on the evaluation method used (F = 12.282; p = .003, η 2 = 0.434). Considering all the rounds and without taking into account the changes in position, Notio Aerostick gave the lowest CdA value (0.238 ± 0.024 m 2 ), and Martin the highest (0.248 ± 0.021 m 2 ), with significant differences between them (p=.002). Between these two methods, VE (vs. 0.243 ± 0.020 m 2 ) only differed significantly from Martin (p<.001). These differences were reduced to a trend when comparing between positions (F = 2.951 p = .085, η 2 = 0.156; CdA baseline = 0.249 ± 0.007 km / h; CdA C1 = 0.239 ± 0.003 km / h; CdA C2 = 0.240 ± 0.004 km / h), and disappeared when considering the interaction Method*-Change of position (F = 0.150; p = .738, η 2 = 0.009). Despite this, and as shown in Figure 1, posterior Bonferroni adjustments showed that while no method detected these minimal position changes in the aerodynamics (p>.05), the differences between the methods were confirmed and were evident for Martin compared to VE for all the positions (p<.001), as well as to Notio Aerostick at CdA baseline and CdA C2 (p>.05), Figure 2.
In terms of the agreement between the systems, Notio Aerostick presented a high ICC with VE but also a high TEM (ICC = 0.820, 95% CI = 0.685-0.898; p<.001), similarly to the Martin model (ICC = 0.832, 95% CI = 0.705-0.904; p<.001), although with a low bias (± LoA) of −0.004 ± 0.017 m 2 and −0.009 ± 0.016 m 2 , respectively ( Figure 2). The two mathematical models, Martin and VE showed a very high ICC, (ICC=0.999, 95% CI = 0.998-0.999; p<.001), with a low bias (± LoA) of −0.005 ± 0.001 m 2 . When we observe the regression line calculated for the differences, in the Bland-Altman plot, between Notio Aerostick against VE and Martin, we find a proportional (non-constant) asystematic bias, with a positive trend in the differences as the magnitude of the measured variable increases. Meanwhile, when we analyse this between the two mathematical models, we can see that Martin vs. VE shows a proportional systematic bias (not constant), with a negative trend of the differences as the magnitude of the measured variable increases (Figure 3).  The analysis of Pearson's correlation for the data of CdA difference between position changes, between Notio Aerostick and the two mathematical models was significant and moderated; Notio Aerostick-VE: r = 0.411 p=.003; Notio Aerostick-Martin r = 0.409 p=.003, while between the mathematical models it was perfect: VE-Martin r = 1 p<.01. The results of the coefficient of determination (R2) for this same data set are shown in Figure 4.

Discussion
The purpose of this study was to evaluate the validity and sensitivity to changes in cyclist position with respect to the measured CdA between the Notio Aerostick device (Kordi et al., 2021;Valenzuela et al., 2020) and two other mathematical models, VE (Dyer & Disley, 2018) and Martin (Martin et al., 2006). Our main finding is that, although the level of agreement of Notio Aerostick with both formulas is satisfactory (ICC>80%), with low and non-systematic error bias, and the differences in CdA are not significant in any of the positions, our results show that Notio Aerostick gives slightly lower values than those obtained by the formulas. As expected, these confirm the differences and systematic bias, as a result of including the height of the centre of mass for indoor cycling, or not. On the other hand, none of the three systems (Notio Aerostick, VE, or Martin) was sufficiently sensitive to detect position changes, probably due to the already optimised initial position of the participants, as they were all professional cyclists. Again, Notio Aerostick exhibited a different behaviour, correlating moderately with VE and Martin, with respect to the changes in CdA measured between positions.
The CdA results calculated by Notio Aerostick were lower than those calculated by the mathematical models, 2% lower than VE and 4% lower than Martin. This contrasts with the data published by (Valenzuela et al., 2020), in which the CdA results calculated using the Notio Aerostick device were 1.6% higher than those calculated by the Martin model. The differences in these results could be explained by several factors. The first could be the possible variation between the pitot-tube registers of different Notio Aerostick devices, as the reproducibility between devices has not yet been analysed. The second reason is that, in the case of Valenzuela et al.  (2020), different temperature, humidity and pressure sources were used to compare the Notio Aerostick device with other measurement systems.
The fact that we obtained a higher CdA result with the Martin model can be explained by the variable that this model incorporates in an attempt to account for the specific physics of track cycling, correcting the speed measured by the speedometer downwards (Martin et al., 2006). This suggests that both Notio Aerostick and VE may actually give CdA lower values than the real ones in specific indoor cycling conditions.
The differences found between the baseline position and change 2 in all three systems were 0.009 m 2 , with the largest change recorded being 0.011 m 2 , as measured by Notio Aerostick, between the baseline position and change 1. Although we found no statistical significance between the position changes, the improvements in the CdA according to our calculations could mean a speed increase of 0.55 km/h, which would translate into a time improvement of around 1 minute over a 40 km time trial, or a power saving between 10 and 11 watts at constant speed. This shows that, even in the case of professional cyclists with extensive experience and an optimised baseline position, small positional adjustments could lead to significant performance improvements. The results are close to those measured by Valenzuela et al. (2020), although in that case they tested amateur cyclists and compared upright and aerodynamic positions. Our data offers a more realistic view of aerodynamic optimisation for professional cyclists, involving small adjustments from an already optimised position, in search of improved aerodynamics.
Using the same data sources for speed and power, as well as temperature, humidity, and barometric pressure for all three models, we were able to determine to what extent the differences in CdA are due to the particularities of each system. The differences between the results from the Notio Aerostick device and the two mathematical models can be explained by the data provided by the pitot-tube, the data processing, and the mathematical formula used by Notio Aerostick, which is neither published or reported by the manufacturer. The variation between the VE and Martin models can be explained by the centripetal force correction in the Martin model, observing that these measurement differences are systematically present in all the results.
On the other hand, although they shared the data sources mentioned above, and taking into account the fact that the wind conditions were the same in each measurement round, the differences between the CdA calculated by the Notio Aerostick device and the mathematical models were not systematic, a fact that we attribute to the recording and processing of the data from the partial pressure sensor.
Regarding the data of change of CdA between positions, we found no significantly different values between the Notio Aerostick and VE, although we were able to see a certain degree of significance for the baseline position and change 2 recordings. On the other hand, we did find significant differences between the two mathematical models, VE and Martin. This contribution has a limited application in the field of aerodynamic assessment, as we assume that the CdA results from the different methods will not be interchangeable, since they include different sensors and variables in their calculations. However, it would be interesting to see if they respond in a similar way to the same changes of position, since the main objective of the exercise is to find the most aerodynamic configuration (Malizia & Blocken, 2020).
When we look at the CdA change data between the different positions, we can see that Notio Aerostick correlates moderately with the mathematical models. This allow us to state that, in an aerodynamic assessment protocol, the final position recommend, could be different depending on if we are using the Notio Aerostick device, or the mathematical models of VE and Martin. Our data contrasts with that obtained by other authors, (Kordi et al., 2021;Valenzuela et al., 2020) in which the Notio Aerostick recordings followed the direction of change measured in the wind tunnel in one case, and in the other, that of the mathematical models (Kordi et al., 2021;Valenzuela et al., 2020).
To sum up, the practical application of the new systems based on pitot-tube represents an important technical challenge that needs to be studied in depth to determine clearly the methodology of use, as well as its real applications and limitations.

Conclusion
The differences and the moderate correlation found between Notio Aerostick and the two mathematical models for the changes of CdA calculated between positions, suggests that these systems would not be interchangeable in the aerodynamic field-test evaluation in elite samples. Despite the absence of significant differences between the CdA assessment by the Notio Aerostick, and the validated mathematical-model Virtual Elevation, the differences in the direction of the CdA scores might lead to different setup recommendations, because changes in a following position are given by the improvement of the CdA in the previous one. These setup differences might become very relevant in an elite time trial competition, where minimum differences play a key role in the results.
Noteworthy, nor the Notio Aerostick, nor the mathematical models Virtual Elevation and Martin, had enough sensibility to reflect changes in the cycling position of our elite sample when the proposed changes are small (inside the rules). Future studies are required to compare measurements between different units of Notio Aerostick, as well as the reliability and precision of each of the sensors that make up this device.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
The author(s) reported there is no funding associated with the work featured in this article.