Flow deep within the urban canopy is spatially complex and unsteady. Its coupling to roof-level flow, or surface layer flow at a reference height well above the canopy is dependent on turbulent processes. Given the complex influences of local morphology, traffic and stability on turbulence, the question can be asked: can simple models of flow at street level be formulated, and to what height should they be referenced? This paper presents flow and turbulence data measured during the second DAPPLE campaign (see www.dapple.org.uk) which took place from 20th April to 12th June 2004. 11 sonic anemometers were deployed in and around a heavily trafficked intersection in central London, UK. Reference measurements were made on a nearby roof-top and at a height of 190m on the BT Tower, one of the tallest structures within the neighbourhood with excellent exposure to the flow, and located within a mile of the site. The relationship between mean and turbulent flow processes at street level with flow at both reference heights has been investigated. The results show that an idealised model of street canyon flow explained more of the mean flow variability for certain measurement positions when a reference velocity at the upper reference height was used. Street level turbulence also showed a strong correlation with turbulence at the upper reference height throughout the campaign. This was consistent with few observations of overnight stable conditions which could have decoupled the flow. The results are discussed in the context of modelling dispersion in urban areas.