Quoting the press release:
As well as learning the race tracks to gain split-second advantages, players will start to see, hear and feel nuances in taking one racing line or another. For example, players might see a contour made by another car through some gravel, which if they follow will help them gain speed, whereas if they take a fresh route through the gravel it could potentially slow them down. It's these gameplay elements that give extra depth, is what players expect from a next-gen console and is something no game has done before.

HOW DOES IT WORK?
Most games use a 1 metre polygon grid to drive upon, but that was never going to work for the level of details aimed for in SEGA Rally. Because of this, the entire drivable surface of every track within SEGA Rally is modelled at the massively high detail of 6 centimetres for every single polygon - that's 17 times more detail and means that each tyre of each Rally car interacts with up to 12 polygons at any time. Importantly, the physics engine powering SEGA Rally also runs the detail at a high frame-rate, reacting to every bump in the highly detailed scenery. For every surface, there are a number of characteristics - in addition to the normal static and dynamic slip components SEGA Racing Studio has modelled wear rate and how ruts form for every polygon, how the friction changes as players dig down into the surface, and the profile of the debris which tyres leave as they churn each surface up. SEGA Racing Studio are even able to model the higher grip levels expected on tarmac as traffic lays down rubber - and that means real live racing lines forming that can and must be reacted to as in a real race.