Wildlife roadkill is one of the most obvious impacts of roads and traffic. The usual way to quantify roadkill risk in different sections of infrastructures is to collect information about the location of casualties and the environmental and infrastructure variables associated with them. There are, however, limitations to this approximation: it is not applicable in roads with low traffic intensity as they have a small number of victims (e.g. unpaved roads), where there is a high removal rate of casualties by scavengers (e.g. in natural areas), when the estimates need to be obtained over a large network of roads or when it has to be estimated before the infrastructure is built. We developed an indirect approach to evaluate the risk of collisions with wildlife within the natural area of Doñana (SW Spain), considering the abundance and phenology of species, the characteristics of the environment, and traffic intensity. First we characterized the road network, corresponding to 2190 km of roads (4.04 km/km2) of which only 2% are paved; and extracted environmental variables for the complete network in sections of 200m. Then, we characterized the traffic by building a model of traffic intensity for the network using data from a stratified sampling design in 62 sites using magnetometers and automatic fixed counting in main roads, and applied the model to the whole network of roads to estimate the traffic intensity. We characterized the abundance of multiple species using track censuses in 183 sites using 200 m transects; obtaining information on abundance, crossing intensity and the distance moved along the road (estimator of the time of exposure to vehicles or exposure). With this information we created a model of the number of crossing events per species in sections of 200 m using environmental predictors and applied the models to the whole network of roads. We estimated the roadkill risk using the index risk = log(no. crossings x traffic intensity x exposure), standardized between 0 and 1. We calculated the index for the whole network of roads. As an example, we show the predictions corresponding to the roadkill risk for European rabbits (Oryctolagus cuniculus) and Iberian lynx (Lynx pardinus), clearly identifying areas of high risk which are localized along roads with high traffic intensity and within them, specific sections with maximum risk. This predictions match well with the observations of road-killed lynx recorded in the area. The proposed approach can be very useful in the identification of areas in which there is a high risk of collision with wildlife in already developed road networks (e.g. identifying roads to be eliminated or sites where crossing structures could be built) and before road construction to decide where to apply mitigation measures.