This post addresses the error in estimating the friction factor due to errors in pipe roughness.
The Freeman data for the wrought iron pipes resulted in the following absolute roughness values, all in ft.: max = 3.4e(-4), ave = 1.8e(-4), min = 9.4e(-5). These bounds may also be appropriate for carbon steel, but I have not verified that assumption. The plot below shows the error bounds assuming the average roughness is used for the prediction but the actual roughness is either the max or min value.
For fully turbulent flows, the uncertainty in roughness can produce up to 18% error in the friction factor estimate for pipes between 2-60 inches in diameter. Naturally, the errors increase as the pipe diameter decreases. The errors decrease as the flow decreases due to the convergence of the friction factor values to the smooth pipe value. When using the Colebrook-White (CW) formula in the fully turbulent flow region, the prediction error for f will be essentially due to the uncertainty in the roughness. In the turbulent transition region, the under prediction bias of the CW formula (<8%) and the roughness error are about the same magnitude depending upon pipe diameter and Reynolds number.
Scobey provided three values for the absolute roughness of concrete pipes, .001, .003, and .01, all in ft. If we assume the middle value for the average roughness, the other two values produce the error bounds shown below.
The variability in concrete pipe roughness can result in nearly 50% error for the smallest pipes.
Now for the good news
The possible error in prediction of the friction factor isn't very encouraging. However, the situation may not be as bad as it first appears. For short pipe applications, the friction loss in the pipe is usually not as significant as the loss due to valves, elbows, entrance and exits. For long pipe applications, the pipe itself performs averaging of the roughness values due to the many pipe segments involved. For these reasons, use of the average roughness may be acceptable: that's good news since it is our only option during the design stage.
References: see 6/1/2014 post (including pdf file) and 6/12/2014 post.