Winter road maintenance activities are intuitively beneficial due to their critical roles in maintaining the safety and mobility of highway networks in winter seasons. There is however no robust methodology currently available for quantifying these benefits, which are needed for comprehensive cost-benefit analyses of all winter road maintenance decisions. This paper introduces a set of collision and mobility prediction models that have the potential to address this need. The models were developed using a unique data set containing detailed hourly records of road weather and surface conditions, traffic counts, and collisions for a number of Ontario highway routes. Several case studies are used to illustrate the applications of these models for evaluating alternative winter maintenance policies and operations, such as changing of bare pavement recovery time, changing maintenance operation deployment time, and changing level of service standards.