Full Terms & Conditions of access and use can be found at
European Journal of Sport Science
ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/tejs20
Reliability and Sensitivity of the Notio Konect to
quantify Coefficient of Drag Area in Elite Track
Mehdi Kordi , Gert Galis , Teun van Erp & Wouter Terra
To cite this article: Mehdi Kordi , Gert Galis , Teun van Erp & Wouter Terra (2021): Reliability and
Sensitivity of the Notio Konect to quantify Coefficient of Drag Area in Elite Track Cyclists, European
Journal of Sport Science, DOI: 10.1080/17461391.2021.1891296
To link to this article: https://doi.org/10.1080/17461391.2021.1891296
Accepted author version posted online: 16
Submit your article to this journal
Article views: 47
View related articles
View Crossmark data
Publisher: Taylor & Francis & European College of Sport Science
Journal: European Journal of Sport Science
Reliability and Sensitivity of the Notio Konect to quantify Coefficient of Drag Area in Elite
Mehdi Kordi1,2, Gert Galis3, Teun van Erp4 and Wouter Terra3
1Royal Dutch Cycling Federation (KNWU), Papendal, Arnhem, Netherlands
2 Department of Sport, Exercise and Rehabilitation, Northumbria University, Newcastle,
3Aerospace Engineering Department, TU Delft, Delft, Netherlands
4Division of Orthopaedic Surgery, Faculty of Medicine and Health Sciences, Stellenbosch
University, Tygerberg, South Africa.
Royal Dutch Cycling Federation (KNWU)
Elite level cycling events are performed at speeds in excess of 50 km/h. At these
speeds, over 90% of the resistance forces come from aerodynamic resistance (CDA). Recently
bicycle-mounted pitot tubes, such as the Notio Konect (NK) have become more
commercially available making CDA easier to measure. Its reliability and sensitivity would be
useful for riders and coaches to be able to understand what constitutes as a change in CDA.
Accordingly, the aim of this study was to establish the intra- and inter-effort reliability and
sensitivity of the CDA measures of the NK. Seven elite level track riders were used in this
study which was broken into two parts: 1) Reliability and 2) Sensitivity. For both parts of the
experiment, riders performed identical efforts, riding at ~50 km/h for six laps of a 250m
indoor velodrome. For reliability, the riders performed six efforts without any changes in
position or resistance. For sensitivity, they performed the efforts with a rod with discs of a
known diameters attached at each end to vary the CDA by a known amount. For the reliability
assessment, low coefficient of variation of intra- (0.2%) and inter-effort (0.9%) reliability
were measured. With regards to sensitivity, the smallest changes in resistance (from 5 – 6cm
i.e., 1.2% or 0.002m2) was identified by the NK. The data in this experiment suggests that the
NK is a highly reliable in measuring CDA can detect changes up to at least 1.2% in an indoor
velodrome using elite level track riders.
Keywords: Aerodynamics; Performance; Testing; Velodrome.
The speeds that are reached in elite cycling races, generally range between 50 km/h in
road race time trials (TT) and can reach almost 80 km/h in track sprinters. The mechanical
power to overcome aerodynamic resistance is a third-order polynomial of the velocity and
thus, at these high velocities, it is as very important to reduce or minimise aerodynamic drag
as well as maximising mechanical power output in order to improve performance
(Underwood & Jermy, 2010). At these forementioned velocities, the aerodynamic drag
contributes to more than 90% of the overall resistance riders need to overcome (e.g. Martin et
al., 2006). Hence, reducing the aerodynamic drag is a widely used approach in order to
increase a rider‟s speed and, in turn, improve its performance.
The „gold standard‟ for measuring the aerodynamic drag is by installing a rider and
the bike on a force balance in a wind tunnel, resulting in highly repeatable conditions, testing
with professional cyclists in a wind tunnel the coefficient of variation is generally around 1%
(Juan García-López et al., 2008) and high precision measurement of the aerodynamic forces.
However, the wind tunnel environment generally differs from the conditions a rider
experiences on a track or outdoors (e.g. model supports, turbulence characteristics of the
flow) and, hence, the obtained results may not always be representative for race conditions.
The aerodynamic drag is also evaluated in the field, for example by means of towing
methods, coast-down methods and mathematical models (Debraux et al., 2011). The latter,
calculates the aerodynamic drag or drag area (CDA) of a rider using the measured mechanical
power output and speed (Martin et al., 2007; Underwood & Jermy, 2010). However, these
methods usually make use of a number of assumptions, for example, default rider trajectory
and stagnant air (Lukes et al., 2012), which generally do not match the real conditions. Whilst
the CDA measures of the aforementioned methods have not been compared extensively in the
literature, it is likely that each method will produce different CDA values. As a consequence,
comparing the absolute CDA measures between methods is futile. Understanding the
reliability and sensitivity of each methodology will be more practical and beneficial to riders,
coaches and practitioners.
One of the commercially available products for aerodynamic drag evaluation in the
field using mathematical models is the Notio KonectTM (NK). A particular advantage of the
NK is its capability to measure the speed of the rider relative to the air using a pressure sensor
mounted in front of the handlebars. This allows to use the actual air speed in the
mathematical model, instead of using the measured ground speed and assuming stagnant air
conditions, and so avoid one source of error. A previous study has tried to investigate the
validity, reliability and sensitivity of the NK (Valenzuela et al., 2020). The authors concluded
that the aerodynamic drag measured with the NK is comparable to that obtained by other
mathematical models (e.g. Track Aero System) when riding in an indoor velodrome. In
addition, whilst the NK could detect large (or more obvious) differences in CDA (in this case,
between an upright position and a TT position), it remained unclear if the NK can detect
smaller differences. For example, they could not detect any significant differences between
wearing a training helmet or a more aerodynamic racing helmet. The present authors identify
three main limitations to this study. Firstly, it was unclear what differences to expect when
changing helmets between riders because of a missing benchmark. Secondly, the study did
not systematically vary rider configuration in order to determine the accuracy of the
measured drag area. Note, that this is also not possible without the benchmark. Finally, the
CDA was only measured for one minute rather than a given number of full laps, which could
have influenced the final CDA measures (Lukes et al., 2012) and, related to it, no inter-effort
reliability has been established (i.e. comparing the reliability between laps in the same effort).
Being able to further know the (intra- [within effort] and inter-[between effort]) reliability
and sensitivity of the NK would help riders, coaches and support staff to make informed and
evidenced-based decisions of which positions and/or attire optimise CDA and therefore,
Accordingly, there were two aims of this study. Firstly, to ascertain the intra- and
inter-effort reliability of the NK for measurement of the CDA of elite track cyclists in an
indoor velodrome setting. Secondly, to evaluate its sensitivity against a known CDA
In total 7 (4 women and 3 men) elite level track endurance riders volunteered for this
experiment. All were either currently competing at senior World or European championship
level (including two current World Champions). All riders participated in experiment 1. Of
those, 4 riders participated in experiment 2. They performed the experiment(s) as part of their
training and on their preferred track bike set-up. Three participants performed the efforts on
their TT bike holding a TT position, while four participants held a dropped position on their
bunch bikes with handlebars similar to those of regular road bikes. Before their involvement,
the riders were informed of the purpose and potential risks of the study and provided written
informed consent. All riders were healthy and had no underlying injuries or health conditions.
This study sets-out to ascertain the reliability and sensitivity of the NK and the
cyclists as a system rather than the reliability and sensitivity of the NK. The study was split
into two experiments to assess the reliability and sensitivity of the NK. In experiment 1, the
inter- and intra-effort reliability was assessed by comparing the measured CDA of 6 identical
efforts. The goal of experiment 2 was to determine the sensitivity of the NK. Discs of
increasing size (5, 6, 8 and 10cm) were mounted to the bike by means of a one-meter-long
rod (Figure 1). Discs have been selected as a means to add aerodynamic drag to the system,
in contrast to spheres for example, because their drag area remains constant along all the
speeds considered in this study. Considering the relatively large distance between the discs
and the rider (~40 cm), it can be assumed that the interaction of the flow around the discs and
the rider is negligible and, hence, the aerodynamic drag of the discs can be obtained from
wind tunnel measurements in absence of a rider. Both experiments were performed in an
indoor 250m velodrome (Omnisport, Apeldoorn, The Netherlands) and the average
conditions were: 53.1% humidity, 1015kPa and 27oC. In order to minimise any
environmental influences (e.g., wind and turbulence), the environmental conditions of the
velodrome were kept constant (i.e., all doors closed, and the velodrome track was used
exclusively without any other riders). Each experimental session was carried out on separate
days but within 1 – 7 days of each other. Each rider had a statically calibrated portable power
crank (SRM, Jülich, Welldorf, Germany), magnet-based speedometer (SRM, Jülich,
Welldorf, Germany) attached to their bikes which were wireless connected to a NK unit
(Notio Technologies, Montréal Canada) which was mounted on the handlebars of the bike.
Each rider rode the same front 5-spoke and a rear disc wheel which were each inflated to 1.1
MPa. The tyres (Diamant, DuGast, The Netherlands) used had an assumed Crr of 0.0025
which is what has been used before (J. García-López et al., 2014; Martin et al., 2006). Each
rider wore their own race attire (e.g. helmets, skinsuit and socks) throughout.
Before each session, using previous training and race data, a pre-defined speed was
agreed with each rider to perform the calibration and efforts (the group average speed for the
men: 52.2 ± 2.8 km/h [14.5 ± 0.8 m/s]; and the group average for the women: 46.0 ± 1.5
km/h [12.8 ± 0.4 m/s]). The rider performed a 2 x 3000m calibration run (with two laps of
light pedalling in between each 3000m) as per manufactures guidelines. Subsequent to the
calibration efforts, the respective experimental session was performed. Each effort was aimed
to be identical: the system (rider and bicycle) mass was measured then the rider was given 2-
lap build-up to their target speed. Once they had reached their target speed, they were asked
to hold a constant position and average lap speed for 6 laps. Real-time lap splits were
communicated by a coach that was trackside each lap. The riders also had access to their real-
time speed, power and cadence via their cycling computer that was attached to their bike.
In experiment 1, the riders performed six identical efforts (i.e. 2 build-up laps and 6
laps at constant speed) without any changes or modifications to their bicycle. The riders were
given approximately 10 min passive rest between each effort. During experiment 2, an
aerofoil shaped rod was attached to the handlebars of the bicycles with a disc of 5, 6, 8 or 10
cm in diameter attached to both ends of the rod (Figure 1). Prior to the data collection for
experiment 2, the CDA of the rod and each set of discs was measured in a wind tunnel at 14
m/s (which equates to 50.4 km/h). The drag increments between these configurations are used
to benchmark the expected change in riders‟ CDA in comparison to the NK data. When the
changes in CDA measurements of the discs in the wind tunnel were compared to theoretical
calculations, an average difference of 0.0003 ± 0.0001 m2 was observed. The addition of
CDA from the discs was also calculated theoretically showed was 0.0003 m2 from the Similar
to experiment 1, riders performed two separate efforts (i.e. 2 build-up laps and 6 laps at
constant speed) with each disc set attached to their bike. The estimated CDA of the rider was
ascertained (again) from another preliminary calibration run and the additional CDA for each
disc set was added. The relationship between changes (both absolute [m2] and relative [%]) in
CDA measured from the NK and the calculated CDA were compared.
For both experiments, the NK edition of Golden Cheetah (http://goldencheetah.org/)
was used to process and calculate an average CDA. The principles of the CDA calculations are
based from previously published literature (Martin et al., 1998). Each lap was identified
using the “Velodrome” function. This function differentiates laps by using the inbuilt
gyroscope to identify every other peak roll value (i.e. maximum lean angle) and well as
setting all recorded elevation values (the most variable measure) to zero then performs the
Absolute reliability was measured using absolute standard error and calculated
relative reliability using coefficient of variation (CV), standardised typical error and
intraclass correlation coefficient (ICC). CV was used to measure inter-effort reliability by
comparing of the individual lap CDA measures from experiment 1. Intra-effort reliability was
measured by taking the average CDA of each 6-lap effort and comparing it between efforts
during experiment 1. For standardised typical error, the results were doubled prior to
interpretation using modified effect size thresholds (trivial, ≤ 0.2; small, > 0.2–0.6; moderate,
> 0.6–1.2; large, > 1.2) as previously advocated (Smith & Hopkins, 2011). ICC was
interpreted according to the following thresholds: high, > 0.90; moderate, 0.80–0.90; low, <
0.80 (Vincent, 2012). Raw and relative typical error, as well as ICC, were determined by
using the MS Excel Reliability spreadsheet developed by Hopkins (Hopkins, 2015). Results
are presented as mean ± SD or mean (± 90% CI)
As a group average, the reliability trials were performed at 49.1 ± 3.5 km/h and 334 ±
64 W. The individual and average ±SD intra- and inter-effort CV and CDA are summarised in
Table 1. Intra-effort CV average was 0.20% and ranged from 0.16 – 0.69%. Absolute typical
error was 0.0017 (0.0013 – 0.0023) m2 and standardised typical error was 0.13 (0.11 – 0.18).
Both of which are classed as a trivial effect size. ICC was 0.99 (0.98 – 1.00), which
represented high repeatability. Inter-effort CV average was 0.90% and ranging from 0.46 –
1.52%. Absolute typical error was 0.00158 (0.00122 – 0.00232) m2 and standardised typical
error was 0.12 (0.09 – 0.18) which is classed as a trivial effect size. ICC was 0.99 (0.98 –
1.00), which showed high repeatability.
The mean relationship between changes and measured changes in CDA exhibited a very large
and positive relationship (r2 = 0.996) which is shown in Figure 2. Individually, the changes in
absolute and relative CDA each of the riders all displayed near perfect linear relationships (r2
= 0.95 – 0.99) which are shown in Figure 3.
The aim of this experiment was two-fold. First, to establish the intra- and inter-effort
reliability of CDA when using the NK. Second, to establish the sensitivity of the NK. With
regards to the first aim, a very high level of reliability was observed with this cohort of riders.
The average intra-level reliability of a 6-lap effort 0.20% with the range from as low as 0.16
– 0.69%. The group inter-effort reliability was 0.90% ranging from 0.46 – 1.03%. With
regards to the second aim, the NK was able to distinguish between all levels of changed
resistance with the smallest step change being a change of 0.002m2 (or average of 1.2%) in
The results of experiment 1 suggest that the high levels of (intra- and inter-) effort
reliability. These findings suggest that the NK can be used to assess changes in CDA of a
rider during a training session as well as between training sessions. Making it possible for
riders, and support staff to understand how CDA could change throughout a particular effort
or how a constant riding position is held at different speeds or different levels of fatigue.
Intra-effort reliability of the NK was also assessed by Valenzuela et al. 2020. Intra-effort ICC
values were identical between studies (0.99). However, the typical error presented in this
study was noticeably lower than was is shown by Valenzuela et al. 2020 by a factor of ~10
(0.0015 vs. 0.015 m2) (Valenzuela et al., 2020). This may be due to two reasons: First, the
cohort of riders in this data collection are elite riders who have competed at a number of
World and European championships (which include several World Champion winners and
medallists). As such, they are very experienced in riding the track and even though four of the
seven riders were on their upright/bunch race bike, they are accustomed to holding low and
aggressive positions. In addition, typical error represents absolute values of variation. The
riders in this data collection had much lower average CDA (~26% lower) values than those of
Valenzuela et al., 2020. When put in this context, the typical errors should be closer than it
initially seems. As the relative reliability of Valenzuela et al., 2020 was not reported, this is
no way of being completely certain whether this is true and is largely speculation.
The “velodrome” lap function in the analysis provided by the NK allowed inter-effort
reliability to be also measured. This was very low throughout this experiment (Table 1). To
the best of the author's knowledge, this function has not been used in any other experiments
in the literature as at the time of writing, this is a new feature provided by NK. In any case, it
has shown to be a useful tool which allows coaches and riders to analyse each lap of the
larger effort. This is particularly useful when trying to understand how a rider‟s CDA changes
under fatigue, throughout an effort or at different speeds.
The results from the addition of known resistances suggest that the NK is sensitive to
the full range of drag increments that was presented to it. All of the four riders who
participated in this part of the experiment showed perfect linear relationships with the
changes in CDA. All of which also were able to detect the smallest presented changes, going
from 5 to 6 cm discs which equates to an increase of 0.002 m2 (or an average of 1.2%).
With the NK being able to detect changes (as small as 0.002m2) makes it very
beneficial for its use in the field. The NK can help riders, coach and practitioners to find
small differences in riders‟ position or attire which are extremely valuable at all levels of
performance especially at the elite level track cycling. To put this in perspective, in a simple
model suggests that if a rider can lower their CDA from 0.198 (average of this experiment) to
0.196 m2 and performs a ride at 50 km/h, the difference is ~4 W of aerodynamic drag. Over
an individual pursuit this can equate to an improvement of ~0.8s in an individual pursuit and
over 5s in a 10-mile TT.
The average data of the sensitivity data was closely matched with the line of identity
(Figure 2). However, when as resistance increased (by adding larger discs), the further the
line of regression deviates from the line of identity. This could be as attributed to the increase
in aerodynamic wake. As the size becomes significant (~10% of total drag) there is likely to
be some significant interaction between the wake of the discs and the wake of the rider. This
could be somewhat mitigated by mounting the discs at a wider spacing.
When displayed individually (Figure 3), the strong linear relationships are still seen of
each rider with the addition of resistance and being able to distinguish between small
differences in CDA. However, each rider had their own “off-set” which was not systematic.
There does not seem to be an obvious reason to this. The authors speculate that this “off-set”
may partly be explained by the flow around the rod that interacts with the measurement of the
air speed of the NK, considering that the two were installed relatively close together, in
particular for the TT setup. In addition, it is possible that the orientation of the NK was
altered (by accident and without noticing) when installing the rod and, so, that the air speeds
measured after installing the rod are different from those before installation. The conclusion
of the sensitivity of the Valenzula et al. 2020 study suggested the NK can differentiate
between larger differences in CDA (i.e. between upright and aero positions) but questioned its
sensitivity when trying to differentiate between smaller differences (between training and
racing helmet) (Valenzuela et al., 2020). Unfortunately, the difference between the helmets
were not quantified and as such they were not able to conclude the exact sensitivity of the
NK. This study builds on the work of Valenzuela and colleagues and suggests that the NK
can differentiate in differences as small as 0.002 m2 (or 1.2%).
There are limitations and improvements that could have been made to this study. The
limits of sensitivity of the NK were not established i.e., to what point can the NK no longer
differentiate between changes in CDA, so the exact sensitivity has yet to be determined. so the
exact sensitivity of the NK has been found. Furthermore, the results of this experiment are
only applicable to an indoor velodrome setting. The influence of outdoor environmental
conditions (e.g., wind) and elevation change have not been factored in. Future experiment
evaluating the impact of outdoor environmental conditions and/or outdoor velodrome
settings. Lastly, therefore, the sampling rate is limited by the smallest sampling measure. In
this case it was power and cadence was transmitting data at 1Hz, while the NK samples at
4Hz. Higher sampling frequency could mean higher levels of sensitivity and could limit the
number of laps necessary at these high speeds. Future experiments could ascertain the limits
of sensitivity of the NK with smaller changes in aerodynamic drag.
In conclusion, this study showed a high reliability and sensitivity of the NK, this
means it can be useful for riders to perform aerodynamic testing with the NK at a relatively
constant speed in a velodrome. Furthermore, the new function “velodrome‟ function
introduced by NK is highly valuable for it use in the field.
The authors would like to thank the riders who participated in this study as well as Fulco van
Guilk and Adriaan Helmantel for facilitating the study.
Debraux, P., Grappe, F., Manolova, A. V., & Bertucci, W. (2011). Aerodynamic drag in
cycling: Methods of assessment. Sports Biomechanics, 10(3), 197–218.
García-López, J., Ogueta-Alday, A., Larrazabal, J., & Rodríguez-Marroyo, J. A. (2014). The
use of velodrome tests to evaluate aerodynamic drag in professional cyclists.
International Journal of Sports Medicine, 35(5), 451–455. https://doi.org/10.1055/s-
García-López, Juan, Rodríguez-Marroyo, J. A., Juneau, C.-E., Peleteiro, J., Martínez, A. C.,
& Villa, J. G. (2008). Reference values and improvement of aerodynamic drag in
professional cyclists. Journal of Sports Sciences, 26(3), 277–286.
Hopkins, W. G. (2015). Spreadsheets for analysis of validity and reliability. Sportscience, 19,
Lukes, R., Hart, J., & Haake, S. (2012). An analytical model for track cycling. Proceedings of
the Institution of Mechanical Engineers, Part P: Journal of Sports Engineering and
Technology, 226(2), 143–151. https://doi.org/10.1177/1754337111433242
Martin, J. C., Davidson, C. J., & Pardyjak, E. R. (2007). Understanding sprint-cycling
performance: The integration of muscle power, resistance, and modeling.
International Journal of Sports Physiology and Performance, 2(1), 5–21.
Martin, J. C., Gardner, A. S., Barras, M., & Martin, D. T. (2006). Modeling sprint cycling
using field-derived parameters and forward integration. Medicine and Science in
Sports and Exercise, 38(3), 592–597.
Martin, J. C., Milliken, D. L., Cobb, J. E., McFadden, K. L., & Coggan, A. R. (1998).
Validation of a Mathematical Model for Road Cycling Power. Journal of Applied
Biomechanics, 14(3), 276–291. https://doi.org/10.1123/jab.14.3.276
Smith, T. B., & Hopkins, W. G. (2011). Variability and Predictability of Finals Times of Elite
Rowers: Medicine & Science in Sports & Exercise, 43(11), 2155–2160.
Underwood, L., & Jermy, M. (2010). Mathematical model of track cycling: The individual
pursuit. Procedia Engineering, 2(2), 3217–3222.
Valenzuela, P. L., Alcalde, Y., Gil-Cabrera, J., Talavera, E., Lucia, A., & Barranco-Gil, D.
(2020). Validity of a novel device for real-time analysis of cyclists‟ drag area. Journal
of Science and Medicine in Sport, 23(4), 421–425.
Table 1: Individual and group intra- and inter-effort coefficient of variation (CV) as well as
average (±SD) and individual coefficient of drag area (CdA) of all the identical repeat
Intra-effort CV (%)
Inter-effort CV (%)
0.195 ± 0.001
0.179 ± 0.001
0.197 ± 0.001
0.213 ± 0.002
0.202 ± 0.003
0.185 ± 0.002
0.212 ± 0.002
0.198 ± 0.013