The only direct method to evaluate safety is to monitor the number of accidents and near misses. However, vehicle accidents are statistically infrequent and near misses are heavily underreported, making this approach unfeasible. An alternative strategy is to examine and evaluate metrics which have been shown to be precursors to accidents. The widespread use of metric evaluation for measuring safety in vehicle operations is limited due to a lack of ubiquitous data collection and communication systems. In addition, the effective evaluation of a safety metric is strongly dependent on the high-level context of the situation. This paper presents a system that records and exchanges data and context information to facilitate the calculation of informative safety metrics, and shows results from a number of implementations of this system in a mining context.