Numerous types of mitigation measures aimed at reducing animal-vehicle collisions (AVC) are available. These measures are oriented both to the driver and the animal and show very different costs and effectiveness. This diversity can be used to implement the most cost-effective measure for each AVC hotspot according to its characteristics. Given the extensive road network, for many road sections we should select relatively inexpensive solutions that do not increase barrier effect. Measures must also take into consideration animal behavior; some species often run quickly and perpendicularly to the roadway, entering it suddenly and without any response time for the driver. In this sense, measures such as vertical road signs, aimed at reducing speed, could meet these requirements. The problem is that the measure loses effectiveness over time because drivers become habituated to the sign. We can try to reduce this habituation by focusing the warning only at certain critical moments, and, for this purpose, we can take advantage of the fact that AVC respond to quite marked temporal patterns. In line with this idea, we are developing a new road sign that informs drivers in real time about the probability of an animal invading the roadway and, only if a certain risk threshold is exceeded, emits a warning signal. The risk levels are obtained from temporal models that include the time of year, time of day, moon phases, weather conditions, winter maintenance and hunting activity. These models communicate with the road sign via a mobile data system integrated into a VPN. We are testing this prototype on several sections of the Junta de Castilla y León road network (Spain) with frequent AVC with roe deer, red deer and wild boar. The prototype has three positions corresponding to three risk levels: low (off), high (orange) and very high (red). If proven effective, it could become an inexpensive and widely used mitigation measure. The prototype is being patented.