The purpose of this paper is to study the link between the occurrence of injury road accidents, the prevailing traffic conditions, and the occurrence of rain. This is useful for assessing, before its implementation, the safety impact of a new traffic management. Traffic conditions were extracted from a one year traffic database which covers 150 kilometres of two or three lanes urban motorways near the city of Marseille, in the south of France. 208 loop detectors provide the individual speeds, headways, arrival times and lengths of vehicles. Based on this information, thirteen aggregated traffic variables were constituted every six minutes, such as the average speed, occupancy, short time headways and a few combinations of speed, relative speed and time gaps. 292 injuries or fatal accidents occurred on the network during the same year. The French accident database provides their characteristics - location, time and type of accidents, meteorological conditions and other parameters addressing the infrastructure, the driver and the vehicle. The rain occurrence is provided, every six minutes, from a meteorological station. A set of safety performance functions were estimated, each one giving the risk of injury accident by vehicle-kilometre according to the level of one traffic variable and according to the occurrence of rain. Generally based on logistic regression models, analyses were carried out separately by lane and for two types of accidents -single vehicle accidents and crashes between vehicles. Some relations linking accidents with traffic variables are significant: the occurrence of single vehicle accidents is related to the speed on the fast lane; the occurrence of multiple vehicle accidents is related to occupancy.