Quantifying the effects of resilient modulus testing and modelling variability on flexible pavement design performance

The effects of laboratory testing variability and predictive model accuracy on pavement performance prediction in unbound materials remain underexplored. This study quantifies the repeatability of the AASHTO T 307 resilient modulus (MR ) test, evaluates the fitting accuracy of predictive models, and assesses the sensitivity of the AASHTOWare Pavement ME models to unbound‑layer MR inputs via single and multi‑variable analyses for high‑ (HD) and low‑traffic demand (LD) scenarios. The results show that single-operator tests exhibit, on average, less than 10% variability, but inter-machine/operator differences can reach ±30%–60%. The NCHRP 1‑28A model provides the best fit, outperforming simpler formulations. In sensitivity analyses, bottom‑up fatigue cracking is the controlling distress in both high- and low-traffic demand scenarios. However, the LD case showed greater sensitivity towards the changes in the unbound layer's stiffness. Findings highlight that laboratory variability has a greater influence on predicted performance than model selection, underscoring the need for stringent material-testing quality controls for ME design inputs.