Forecasting municipal pavement condition index through sigmoid deterioration model

Abstract

This study investigates the application of a sigmoidal deterioration model with separate regression parameter sets to estimate the pavement condition index ( PCI ) for 2022 across both residential and non-residential streets — including main, collector and industrial roads — in Skellefteå Municipality, Sweden. Separate best-fit sigmoid curves were also developed for various pavement maintenance treatments applied to non-residential streets. The analysis is based on data gathered through windshield surveys conducted in 2014, 2018 and 2022. The findings reveal that the sigmoidal model exhibited poor performance for residential streets, with low R ² values and weak correlations between predicted and observed PCI ratings — probably due to inconsistent pavement age and disruptions such as utility cuts within segments. Conversely, the model demonstrated strong predictive accuracy for non-residential streets, with high correlations and low prediction errors. Moreover, the sigmoid curve for pavements treated with surface levelling (SL) yielded the highest prediction accuracy, followed by sigmoid curves for surface treatment (ST) and for milling and resurfacing (MR) treated pavements, which showed moderate performance. Given its cost-efficiency, SL treatment is recommended for addressing early-stage deterioration, while MR may be more appropriate for extending pavement life, offering a balance between cost and predictive strength. The study offers insights to enhance timely, proactive, data-driven pavement maintenance in municipalities.