Parametrization and evaluation of a coupled Urban Canopy-Weather Research Forecast Model for Singapore

The impact of anthropogenic heat sources in an urban area on the background atmosphere, so called the urban heat island effect, is an interesting phenomena which is extensively studied. The advanced and more sophisticated urban models which can capture the heterogeneity in urban surfaces and coupled with a numerical weather prediction model are useful for this purpose. From these models one can also expect more accurate forecasted meteorological fields, useful for several applications such as air quality modeling and prediction. Motivated by this, we parametrize and evaluate a single layer Urban Canopy Model (UCM) coupled with Weather Research Forecast (WRF) model for the tropical region of Singapore.

The urban morphology of Singapore is complex, posing significant challenge in estimating the canopy parameters in UCM. Moreover, a high resolution and detailed land use, which is essential to configure a UCM, is not available for the region. In the present work this data is constructed using the Urban Redevelopment Authority land use data and additional features are extracted from onemap for Singapore. The built-up areas are classified as low and high intensity residential, commercial and industrial. The other areas in the URA data are matched with the MODIS land use categories. This new data is ingested in our model, configured with four two-way nested domains of 27-9-3-1 km spatial resolution. A total number of 50 vertical sigma layers are used, with the first 15 layers below 1km to better resolve the boundary layer. The initial and boundary conditions are derived from the GFS model.

The model is evaluated against the surface observations: temperature, humidity, wind speed and direction, obtained from a network of stations operated by the National Environmental Agency of Singapore. The classical statistical measures of accuracy: root mean square error, mean bias, hit rate are used to assess the model performance. We also study the sensitivity of urban parameters on the overall model performance.