Assessing the ability of a high-resolution forecast model to represent the weather in London

The ability to accurately forecast local conditions in cities is important for a wide variety of applications. For example, alerting local authorities of low temperatures so that roads can be gritted, or issuing health-related warnings to vulnerable residents. The representation of the urban surface in numerical weather prediction models requires parameterization of the increased thermal inertia and aerodynamic roughness which is typical in cities, resulting in a modified surface flux scheme within the model. Until recently, a lack of urban meteorological observations above roof height has meant that verification of numerical weather prediction models in urban areas can be difficult.

Observations of the boundary layer of Central London have been carried out at the BT Tower (190 m) and at rooftop height since summer 2010. A Doppler lidar has been used over the same period to obtain profiles of wind and turbulence, and to determine boundary layer height. These observations have been used to evaluate the behavior of the Met Office 1.5 km forecast model (UKV) during a strong urban heat island event. The results of this case study suggested that the UKV underestimated both daytime temperature and sensible heat flux in the city. The model also displayed a lag in the diurnal cycle of temperature and sensible heat flux of 1-2 hours. Observed differences between modeled and lidar-derived wind profiles were thought to be related to a difference of up to 500 m between the boundary layer heights diagnosed by the model, and the surface-driven mixing height derived from lidar data.

These observations have now been used to assess the model performance over a period of one year. The results of this assessment suggest that the daytime boundary layer depth is frequently underestimated by the model. A lag in the diurnal cycle of temperature and sensible heat flux is not generally evident, and may be particular to strong urban heat island events, which are rare in the data-set.