Although deterministic propagation models are physically meaningful and potentially accurate, they still have some limitations to reliably reflect some propagation effects known from measurements such as diffuse multipath and keyhole effect for MIMO or smart antennas applications. Our measurement results in urban environments also observed diffused multipath components of the walls. In this paper, a hybrid physical spatio-temporal model is proposed for radio wave propagation in macrocellular urban environments. The hybrid model combines a deterministic model with a geometry-based stochastic model. The former model is a 3-D site-specific model employing a ray-tracing algorithm to describe the direct wave, specular-reflected waves, and diffracted waves due to plane surfaces or rectilinear edges of the buildings in urban. The latter model describes the diffuse multipath components (non-specular contributions) of the dominant walls and the scattered fields due to randomly positioned local scatterers around mobile stations (MSs). According to our broadband measurement results, the total number of point scatterer depends on signal bandwidth. Our model has been validated by comparing the computed TOA (Time-of-Arrival), AOA (Angle-of-Arrival), r.m.s. angle spread and r.m.s. delay spread of radio propagation in urban environments with the measured ones. It is also found that the hybrid model including the scattered fields has better prediction accuracy than the deterministic model especially in predicting the temporal channel parameters, which is because that the scattered fields of the clustered scatterers yield broader temporal responses.