Sensitivity Analysis of the MORUSES Canopy Model over Singapore

The city of Singapore represents a challenge for mesoscale modelling due to its high complexity in urban morphology and function. From the mesoscale point of view, the problem is approached by simplifying the morphology which, in case of Singapore, can lead to unrealistic results. The present study investigates the ability of the urban canopy parameterization scheme MORUSES (Met Office Reading Urban Surface Exchange Scheme) (Porson et al. 2009a,b), which is implemented in the Met Office Unified Model (MetUM), to represent the morphology of Singapore. As a first step, the sensitivity of the temperature and momentum fluxes calculated by the model to changes in roughness parameters is studied.

A method based on Voronoi cells and following Bottema and Mestayer (1998) for estimating the plan and frontal area densities and mean height used to calculate the roughness parameters, is applied with different spatial resolutions and wind directions. Comparison of the roughness length for 1 m and 5 m resolution, respectively, shows relative changes of less than 10% in 99.4% of the grid cells (see Fig. 1). The larger differences are found mainly in areas where the roughness length is less than 0.5 m. Furthermore, when comparing the roughness length for two perpendicular wind directions the relative differences are larger than 10% in 47% of the domain (Fig. 2). It is easy to infer from these two figures that wind direction has a major impact on the roughness length. This is an important result as, by default, this parameter is averaged over 360 degrees in MORUSES.

The planar and frontal area densities calculated using different methods will be implemented in an offline version of the urban scheme and their impact on the meteorological fluxes will be assessed. This step will allow a proper estimation of the degree of detail needed for the simplification of the urban morphology and the importance of wind direction in mesoscale urban models.

Figure 1:Relative difference between roughness length calculated with 1 m and 5 m resolution, respectively.

Figure 2:Relative difference between roughness length calculated for wind direction