Gravel Loss Prediction Model Using Beta Regression

Abstract

Unpaved roads consist of considerable portions of the roadway network in many countries. These roads play crucial roles in the development of infrastructure systems, advancements of socio-economic activities and improvements of the agricultural and production sectors. Thereby, unpaved roads benefit the underdeveloped, rural, and remote neighborhoods, and act as lifelines for these geographically disadvantaged communities. Frequent and regular maintenance activities keep the roadway system operational at a desired level of service. Resurfacing is one of the major maintenance treatments for unpaved roads. A gravel loss prediction model (GLPM) can evaluate the impacts of varying magnitude of resurfacing treatments on the roadway performance. Thus, a GLPM can provide valuable insights for roadway maintenance budget scheduling and decision-making tasks. In this paper, the backgrounds, input requirements, and output results of three popular GLPMs were reviewed. These models were (i) Highway Development and Management Model 4 (HDM-4), (ii) South African Technical Recommendation for Highways Model 20 (TRH-20), and (iii) Australian Road Research Board (ARRB) Model. In addition, the practicality of roadway resurfacing frequency charts which were developed based on these models was also evaluated. This study determined that the existing GLPMs and the corresponding roadway resurfacing frequency charts were often unreliable and impractical. In this study, a beta regression (BR) analysis methodology was utilized to develop and calibrate a GLPM for Iowa. Because of its simplified yet effective nature, the BR model outperformed the popular GLPMs and offered a practical approach to quantify annual roadway gravel loss.