The backhaul forwards aggregated traffic from massive users, machines, and sensors to the core network, and is the key infrastructure to facilitate a smart city. The beamforming and spatial multiplexing technologies enable the wireless multipoint-to-point backhaul to connect dense small cell eNodeBs (eNBs) and macrocell eNBs. Because the spatial characteristics of the backhaul channels, such as the angular power spectra and spreads, determine the interference and capacity performance of the backhaul links, accurate 3D channel modeling is vital for the evaluation and comparison of candidate proposals. In this article, the concepts and methodology of spatial channel modeling are first introduced, and the state-of-the-art models developed by the standardization bodies in recent years are surveyed. Then we present a field measurement campaign on the 3D backhaul channels in an urban street, including the channel sounder implementation, field measurement, multipath parameter estimation, and propagation modeling. The Rx of the sounder (emulating a donor eNB) was installed on the rooftop of a five-story building, and the Tx was located along a street and at different altitudes to emulate a relay eNB. The channel angular power spectra in the elevation and azimuth domains were measured, and the impact by the relay°s distance and altitude on the angular spreads was evaluated. In addition, a Laplace model is proposed for the power spectra in both domains. This article not only explains the methodology and implementation of the spatial channel measurement, but also reveals the urban wireless backhaul channel characteristics which are useful for the design and deployment of wireless backhaul for a smart city.