Municipal street pavement maintenance and management practices in Sweden

QC 20240412

Research funders:

Skellefteå municipality

Mistra InfraMaint Project 1.8 (DIA 2016/28)

Abstract

A well-functioning street network is pivotal in the socio-economic development of a region. Street networks not only facilitate the movement of people and goods but also allocate space for utility services. Maintaining the street network in good condition and meeting the sustainability targets necessitate implementing optimal street maintenance strategies, leading to an efficient utilization of taxpayers' money. 

The objectives of this Licentiate thesis are to analyse pavement management practices and challenges faced by Swedish municipalities, specifically focusing on asphalt concrete (AC) pavements within street networks. Additionally, it seeks to integrate a sustainability tool into pavement maintenance to select maintenance measures that contribute to sustainability goals at the municipal street network management level. Furthermore, the study aims to enhance municipal-level pavement maintenance approaches through the implementation of machine learning (ML) models within a pavement management system (PMS). 

Within this context, three individual studies were conducted—two case studies and a survey involving Swedish municipalities. One case study explores sustainability framework application, whereas the other investigates the utilization of ML models in municipal AC pavement maintenance. The survey investigates the practices and challenges faced by municipal street network administrations in AC pavement maintenance.

The sustainability framework SUNRA (Sustainability National Road Administrations) was adopted by the Swedish Transport Administration (STA) with a primary emphasis on promoting sustainability in pavement management on state-level roads. In this study, the framework has been tested, applied and further streamlined to be applicable for setting sustainability targets and monitoring sustainability performances at the project level within both the STA and municipal contexts. The aim was to simplify the framework so it is appropriate for investment, re-investments, maintenance and operation projects and also to enhance its applicability for various users. The study additionally explored how the framework could contribute to sustainability, identified the drivers and barriers for its application, and examined its applicability and adaptability to projects of varying complexities. The results indicate that the framework can be readily utilized and adapted for investment, reinvestment, maintenance, and operational pavement projects during the planning stage. Additionally, it is also suitable for small municipal establishments, construction or reconstruction of residential areas, and regular maintenance.

A web-based questionnaire survey was disseminated to municipalities across the country to gather first-hand insights into the current practices and challenges associated with street maintenance at the municipal level in Sweden. Survey responses were received from 147 of the 290 (51%) municipalities nationwide. The study reveals that predominant pavement distress encompasses potholes, surface unevenness, and alligator cracking, with the most prevalent causes being pavement ageing, heavy traffic, and patches. Likewise, cold climate and population density serve as influential factors contributing to pavement deterioration. The automated survey methods for collecting pavement condition data, such as road surface scanning vehicles and application of commercial PMS, are very limited. On the contrary, the windshield method, a subjective approach for pavement condition assessment, is widely adopted among municipalities utilizing PMS. The allocation of the budget for maintenance, rehabilitation and reconstruction is higher in the northern regions of the country, as well as in densely populated municipalities.

Manually collected pavement condition data for the years 2014 and 2018 were acquired from Skellefteå municipality to assess the performance of ML models in comparison to the observed pavement condition index (PCI) of the street network. In this context, the supervised ML models Linear Regression (LR), Random Forest (RF), and Neural Network (NN) were employed in conjunction with several variable combinations. The RF model, utilizing paired variables of pavement age (A) and pavement distresses (D) data, consistently demonstrated higher accuracy compared to the other models for residential streets. However, RF models constructed with paired variables of A and traffic (T) consistently outperformed other models in the context of non-residential streets. The significance of input variables fluctuates based on the model's complexity and the pavement performance objective. Nonetheless,  variable A consistently emerges as the predominant factor for predicting PCI in both residential and non-residential street models. 

Further evaluation of the models and simplification of the SUNRA framework to enhance pavement performance and sustainability are recommended. 

urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-345320