This work focuses on a method of estimating the slope of road networks that are ground-modeled by OSM originally. The aim is to get 3-D road vectors including their 2-D location and inclination, that is an important parameter to ensure more reliable route planning. This is done from GPS data that are collected by a vehicle traveling on an existing OSM road network whose a DEM, like SRTM or ASTER data, provides a modeling of the terrain surface. GPS, OSM and DEM data are modeled as measurement equations in order to account for their errors through an UKF that fuses them in a centralized scheme. Here, the key step is to match GPS/OSM/DEM measurements successively by computing statistical Mahalanobis distances. The experimental framework show some results of road inclinations estimation and the significant contribution of a DEM as baseline.