Data Assimilation of Urban Weather Observations in WRF to Model the Urban Climate of Amsterdam.

Ongoing world-wide climate change and urbanization illustrate the need to understand urban hydrometeorology and its implications for human thermal comfort and water management. Weather and climate models can help increasingly to understand these issues, as they progress towards finer scales. With high model resolutions (100m), the representation of cities becomes crucial. The complex structures of cities, made up of buildings, streets and vegetation of various sizes and shapes, require a different modelling approach than the homogeneous rural surroundings. The current urban canopy-layer schemes account for these city specific characteristics, but differ substantially amongst each other due to uncertainty in land use parameters and incomplete physical understanding. Therefore, meteorological forecasting and hindcasting of the urban environment requires improvement.

In this study, we improve the WRF (Weather Research and Forecasting) mesoscale model performance at 100m grid spacing for Amsterdam by incorporating observations of a variety of sources using data assimilation and nudging techniques. Data assimilation aims to accurately describe the most probable atmospheric state by steering the model fields in the direction of the observations. Using the WRF-3DVAR, hindcasts were made for July 2014 for Amsterdam at a resolution of 100m. July 2014 is characterized by both a warm dry period and two days with extreme precipitation (more than 100mm in two days). Assimilated data consists of WMO synoptic weather observations, volume radar data. and urban weather observations recorded by hobby meteorologists. The final goal of the project is to create a 15 year climatological urban re-analysis data archive of (hydro)meteorological variables. This will enable us to trace trends in thermal comfort and extreme precipitation.

Specific to urban boundary layers, a novel approach has been developed to nudge modelled urban canyon temperatures with quality controlled urban weather observations. Crucial is to adjust the urban fabric accordingly, because of the large heat storage within urban canopies. The road and wall layers of the urban canopy model are adjusted considering the bulk heat transfer of the urban canopy which is dependent on stability and wind. In addition, the anthropogenic heat release within the urban canopy model of WRF is improved by incorporating this flux predominantly into the canyon instead of the first model layer above the canyon.

Data assimilation has been conducted every 2 hours and has been verified on the intermediate time steps. The largest improvement is made by assimilating the air temperature, dew point temperature and pressure from WMO synoptic stations. Applying radar data assimilation in addition, slightly improves the location of the precipitation indicated by the fraction skill score. The urban data assimilation, lastly, reduces the cold biases within the canopy which appeared in WRF.