A proper modeling of flow and turbulence within and over urban canopies is key to properly predict air pollution and dispersion in cities. Trees are an integral part of the urban landscape. In many suburban neighborhoods, tree cover is 10 to 30% and trees are often taller than buildings. The effect of trees on drag, mean wind and turbulence in cities is not accounted for in current weather, air pollution and dispersion models. Our goal is to use high-resolution Large Eddy Simulations (LES) over a realistic urban canopy to inform about the effects of trees drag, mean wind and turbulence in the urban roughness sublayer (RSL). The simulated area is part of the Sunset-Neighborhood in Vancouver, Canada. In this area, long-term wind and turbulence measurements are available from instruments on a 28m-tall tower. Further, a high precision airborne Light Detection and Ranging (LiDAR) point cloud provides data to represent both buildings and trees at high spatial resolution in a realistic configuration. Trees are described by location-specific leaf area density (LAD) profiles. LES simulations are performed over a 512 x 512m characteristic subset of the city that contains the tower location and source area. In the LES, buildings are accounted for through an immersed boundary method, adopting a zero level-set distance function to localize the surface location, whereas drag forces from trees are parametrized as a function of the height-dependent LAD. Spectra of streamwise and vertical velocity components compare well between tower data and the model data, confirming the good performances of LES in simulations of flow over fully rough surfaces. We show how the presence of trees affects mean velocity and computed momentum flux profiles, significantly decreasing dispersive terms in the bulk of the flow, which are usually found to play a role in pressure driven boundary layer flows. The impact of trees on integral length scales in the flow is discussed.