Abstract/Summary

Two urban schemes within the Joint UK Land Environment Simulator (JULES) are evaluated offline against multi-year flux observations in the densely built-up city centre of London and in suburban Swindon (UK): (i) the 1-tile slab model, used in climate simulations, (ii) the 2-tile canopy model MORUSES (Met Office–Reading Urban Surface Exchange Scheme), used for numerical weather pre- diction over the UK. Offline, both models perform better at the suburban site, where differences between the urban schemes are less pronounced due to larger vegetation fractions. At both sites, the outgoing short- and longwave radiation is more accurately represented than the turbulent heat fluxes. The seasonal varia- tions of model skill are large in London, where the sensible heat flux in autumn and winter is strongly under-predicted if the large city-centre magnitudes of anthro- pogenic heat emissions are not represented. The delayed timing of the sensible heat flux in the 1-tile model in London results in large negative bias in the morning. The partitioning of the urban surface into canyon and roof in MORUSES improves this as the roof-tile is modelled with a very low thermal inertia, but phase and amplitude of the gridbox-averaged flux critically depend on accurate knowledge of the plan-area fractions of streets and buildings. Not representing non-urban land- cover (e.g. vegetation, inland water) in London results in severely under-predicted latent heat fluxes. Control runs demonstrate that the skill of both models can be greatly improved by providing accurate land-cover and morphology information and using representative anthropogenic heat emissions, which is essential if the model output is intended to inform integrated urban services.