Can we represent urban areas in operational numerical weather prediction models?

Numerical weather forecasting for urban areas is an important but challenging problem. Whilst it is well known that the urban environment can be very different to the surrounding rural areas, often in operational numerical weather prediction models both the cities and the surrounding countryside lay within (at best) a few gridboxes. So it is difficult to forecast the urban effects with confidence, yet many communities require these differences to be accurately predicted. Higher resolution mesoscale or pollution models also require accurate forcing data from these relatively course resolution numerical weather prediction models, so at the very least the numerical weather prediction models must be able to capture general urban phenomenon, such as the urban heat island.

There are two ways in which the surface exchange of sub-gridscale heterogeneity in landuse (such as rural and urban areas) can be modelled. The first is to model the surface exchange using aggregate parameters which are determined from averaging the parameters for each landuse type. The alternative is to explicitly model the surface exchange over all of the landuse types and then aggregate the resulting surface fluxes. Results from both approaches will be presented and the benefits from calculating the surface exchange over all landuse types will be discussed.

In addition, two approaches for modelling the explicit surface exchange from urban areas will be examined. It will be shown that whilst one method is limited in the urban-rural differences that can be modelled, the other method can represent many of the known impacts of urban areas.