Ongoing urbanization highlights the need for a better understanding and high resolution modelling of the urban climate. In this study, we combine rural observations by WMO surface stations, weather radar data and urban crowd-sourced observations with very fine-scale modelling efforts for Amsterdam, The Netherlands. As a model, we use the Weather Research and Forecasting (WRF) mesoscale model with 3D variational data assimilation at a 100-m resolution in the innermost model domain. In order to enable the assimilation of observations within the urban canopy, we develop a scheme to reduce urban temperature biases by adjusting urban fabric temperatures. The scheme is tested against independent urban observations for the summer month of July 2014 and specifically for a hot period and an extreme precipitation event. We find data assimilation reduces biases in temperature and wind speed. Within the city, the most significant improvement is the reduction of negative temperature biases during clear nights, which implies a better prediction of the Urban Heat Island (UHI). Concerning precipitation, the fractional skill score improves incrementally when additional observations are assimilated, and the largest impact is seen from the assimilation of weather radar observations.

In this study, we improve the Weather Research and Forecasting mesoscale model (WRF) performance by incorporating observations of a variety of sources using data assimilation and nudging techniques on a resolution up to 100 meter for urban areas. Our final goal is to create a 15 year climatological urban re-analysis data archive of (hydro)meteorological variables for Amsterdam which is named ERA-urban. This will enable us to trace trends in thermal comfort and extreme precipitation.