The measurement and estimation of wheel loads is an interesting and complex topic relevant for vehicle dynamics control. Accurate wheel load information allows for more straight-forward, more robust and more efficient control. In this paper a novel model based wheel load reconstruction approach is presented. An Unscented Kalman Filter is used to reconstruct the unknown wheel loads by analysis of the deformation of the bearing outer-ring. The performance of the approach is demonstrated by field tests using an instrumented passenger car. Results show that the proposed approach is well able to reconstruct both tilting and self-aligning moments as well as lateral and vertical wheel forces during various steering maneuvers.