The Case for a Variable CdA in Cycling

Most riders, coaches and even researchers treat CdA as a fixed number. Our first paper shows why it isn't. Using Reynolds number theory, published wind tunnel data and our own field tests with riders including Anna Kiesenhofer, Mattie Dodd and Casper von Folsach, it shows that a rider's CdA can change by more than 10% across their speed range. That's a bigger effect than most of the error sources usually blamed in aero testing. It also traces how the "constant CdA" assumption took hold through a chain of papers from the 1970s to the 1990s, and why ignoring it can show up as implausible rolling resistance values in field testing.

A cyclist rides a road bike on a country road with overlaid text reading 'Streamlines Tyre Testing Methodology' on a green field background.

Tyre Test Methodology

How can rolling resistance be measured when the process starts from an estimate? This article sets out the energy-balance equations behind CIRRUS’s gated CdA and Crr measurements and explains where Streamlines’ approach differs from Virtual Elevation (the Chung method). It then works through how much an error in the initial Crr estimate carries into the results. The worked examples show that a 10% Crr error shifts absolute CdA by only about 1% and leaves CdA comparisons essentially unchanged. On smooth roads, tyre-to-tyre Crr differences are reliable even though absolute values are not, and on rough surfaces such as unpaved roads or cobbles, absolute Crr becomes much more accurate.

Effect of Variable CdA on Combined CdA and Crr Solutions in Field Testing

Can Virtual Elevation really solve for CdA and Crr at the same time? The method assumes CdA stays constant as speed changes, which is rarely true in cycling, and this article shows what that mismatch does to the results. Using wind tunnel drag curves from Grappe (2009) and Brown et al. (2023), it simulates 24 noise-free hill-climb and coast-down rides and runs each through a VE analyser that solves for both values together. With a realistic variable CdA, Crr came back 9–62% too high and CdA up to 8% too low, even with no wind, noise or drift. The bias changes with test speed and no speed combination removes it. It also can’t be seen in the VE plot, because the wrong answer fits the elevation better than the true values do.