Abstract
during harsh winter conditions. Yet it poses substantial economic and environmental challenges, particularly in countries like Norway, where winter operations constitute nearly 20% of the national maintenance budget and involve millions of kilometres of plowing and several thousands of tons of salt each season. At the same time, WRM decisions occur within a multilevel strategic-tactical-operational structure governed jointly by public authorities and private contractors. This makes it difficult to anticipate and evaluate the consequences of individual planning decisions. Despite extensive research on winter severity indices, decision-support tools, and weather-based models, there remains no transparent, spatially explicit tool capable of linking both weather and non-weather drivers to the expected number of operations at decision-relevant scales. Addressing this gap, this PhD, embedded under the WinterSim project, aims to develop the computational core for a GIS-based simulation framework for WRM in Norway. The work is guided by three research questions: (1) What are the primary triggers and key factors influencing the number of WRM operations? (2) How can these factors be used to effectively simulate WRM operations? (3) How can weather variability and climatic trends be incorporated into future-oriented WRM simulations?
The first part of the research consists of an extensive literature review answering the first research question. This review examined global WRM approaches, Winter Severity Indices (WSIs), decision-support systems, and existing simulation models. It identified fundamental weather triggers, such as snow events, cold days, and freezing rain. The review also revealed key limitations in current tools; Most index-based systems lack interpretability, cannot clearly link individual factors to expected operations, and rarely include non-weather variables or spatial variability. This led to proposing the conceptual design of the Effort Model, intended to serve as the computational core of a GIS-based simulation tool. It is a location-specific, weather- and non-weather -driven estimator of the number of plowing, salting, and combined operations.
The second main component of the thesis focuses on the development of the initial Effort Model to address the second research question. A generalized linear regression (GLR) approach was used to formulate three sub-models, one each for plowing, salting, and combined operations. The sub-models estimate number of associated operations at any given locations across the Norwegian state roads based on road geometry, annual traffic, fleet allocation, elevation, and gridded weather variables. This first operational version demonstrated reasonable explanatory power, with R² values of 0.71 (plowing), 0.66 (salting), and 0.66 (combined). It provided a quantitative framework for directly linking individual variables to operation frequencies, thereby enabling transparent scenario analysis and serving as a foundational step for simulation-based planning.
Building on this, the third component enhances the model using Multiscale Geographically Weighted Regression (MGWR), further addressing the second research question. MGWR captures spatial heterogeneity by allowing each variable to operate at its own geographical scale, reflecting Norway’s diverse topographic and climatic conditions. This upgrade improved accuracy by 0.16, 0.25, and 0.23 for plowing, salting, and combined operations, respectively, and produced interpretable coefficient surfaces aligned with operational understanding.
Recognizing the need to support future contract planning, the fourth major component of the thesis addresses the third research question by incorporating climatic variability and trend behaviour into the simulation. Using long-term time series of the model’s weather variables, the thesis applies trend and harmonic components to describe dominant climatic trends and cycles. These functions produce near-future projections of weather variables that serve as inputs to the MGWR-based Effort Model. The results show that multi-year average (over Norway’s contract horizons) forecasts achieve accuracies between 76% and 89% for separate operation types, with approximately 83% accuracy for contract-period forecasting for total operations. While long-term projections remain uncertain, the method provides a feasible and practical approach for incorporating climatic trends in WRM simulations and anticipating near-future needs under evolving climatic conditions.
The discussion chapter synthesizes these findings and evaluates the broader implications of adopting simulation as a decision frame for WRM. The thesis argues that simulation unlocks new planning capabilities by providing a direct operational measure, number of required actions, rather than severity scores or cost proxies. It also critically discusses limitations of the current model with regard to calibration duration, supporting evaluation of the conducted efforts’ socio-economic consequences, and supporting operational or real-time decisions.
Overall, this dissertation develops a methodological foundation for advancing the use of simulation in winter maintenance planning. Collectively, the three components, GLR baseline, MGWR spatial upgrade, and climate-trend integration, form a coherent simulation system capable of supporting strategic and tactical WRM decisions and planning. This work contributes to more transparent, data-driven, and climate-aware decision-making in the winter maintenance sector, supporting Norway’s broader goals of sustainable mobility, cost efficiency, and environmental responsibility.