Sub-kilometre Resolution Numerical Weather Prediction of the Convective Urban Boundary Layer – a Comparison with Observations for London, UK

With rapid increases in computing power, many operational meteorological centres are running kilometre- scale Numerical Weather Prediction (NWP) models over relatively small domains in order to improve local forecasting of weather. The primary motivation for high resolution is often to reach the so-called “convection permitting regime”, where convective overturning is explicitly simulated rather than parametrized. High resolution simulation of convection over urban areas can be motivated by the need to predict heavy rainfall more accurately to manage potential flooding events, or improving boundary layer evolution and mixing for air quality applications. Until recently, a lack of urban meteorological observations above roof height has meant that verification of urban boundary layer simulations was not possible. This presentation reports results from a comparison of the UK Met Office forecast model for a set of nested models with 500m, 100m and 50m grid lengths with observations of boundary layer turbulence in London, UK. Observations of turbulence and fluxes were carried out at the BT Tower (height 191 m) in central London for a cloud-free convective boundary layer case from September 2011.  Boundary layer depth and wind velocity profiles using Doppler lidar were also available. Comparisons were carried out of the characteristics of the convective boundary layer between the high resolution models. It was found that spin up effects at the inflow boundary of the model can be very important and need to be avoided by using a sufficiently large domain. The 100m and 50m resolution models superficially represented the convective boundary layer well with approximately the correct magnitude and depth of roll-like convective overturning. In contrast, the 500m model demonstrated much lower explicitly-resolved vertical velocities. It was clear that the size of convective structures was smaller for 50 m than 100 m resolution runs, which implied that the model was not properly converged even at such high resolutions.  However, spectral analysis of the model vertical velocity time series compared to the BT Tower and Doppler lidar data confirmed that the 50m and 100m models just about had sufficient resolution to represent the most energetic eddies. Mixing height compared very well with observations and its spatial variation across the city showed a clear deepening in response to the warmer urban surface. The results show that 100 m resolution simulations are promising but suggest optimisation of the sub-grid mixing scheme is required to properly handle the turbulence grey zone.