Road surface conditions in winter are very transient. A small change in the friction coefficient can have a large effect on vehicular movements. A friction coefficient measured at the scene directly after an accident may be different from the one at the time when the accident occurred. A traffic accident reconstruction method that can estimate friction coefficients based indirectly on in situ data is presented. First, it is pointed out that the friction coefficient is largely dependent on the tire slip ratio, in particular that of studless tires whose tread rubber is very soft and adhesive. It is emphasized that it makes sense to use a tire model that considers the slip ratio to estimate the friction coefficients more precisely. Rear-end collision accident data were collected. The types of vehicles involved in accidents were identified, and their friction coefficients at the time of skidding were estimated by the new reconstruction model. The friction coefficients estimated by this model were compared with those estimated by a conventional rigid-body model. In addition, by inversely calculating the friction coefficient required to avoid accidents, how an increment in the friction coefficient could greatly improve traffic safety is considered. Finally, it is shown that a threshold defined as the intersection of the cumulative curves of both friction coefficients provides useful information for winter maintenance.