Observation and Modeling of Urban Climate Variability at Local Scale

The Urban Heat Island (UHI) is often studied at the scale of the entire city for by differentiating its effects according to homogeneous neighbourhoods. Nevertheless, the urban climate variability at very local scale may be of the same order of magnitude than at city scale. The neighbourhood is besides a quite interesting and challenging microclimate study field. The comprehension of microclimate variability and involved physical processes objective of improving thermal comfort while meeting as much as possible inhabitant expectations.

Within the framework of the EUREQUA research project, a field experiment was carried out in a neighbourhood of the city of Toulouse (France). The area covered about 1 km x 0.5 km and was composed of different urban fabrics. From January to June 2014, a permanent network was setup, composed of ten weather stations recording near-surface temperature, humidity, wind speed and direction, and completed by a roof-level reference station in order to document larger scale atmospheric variables including the incoming short- and long-wave radiation. Three intensive observational periods were conducted in January, April and June 2014. For three successive days, every three hours, mobile measurements of temperature, humidity and wind were continuously recorded along a predefined itinerary through the neighbourhood, with a GPS recording associated. In addition, a high-resolution geospatial database has been produced to document land covers and morphological characteristics.

We have first studied the microclimate variability at this study scale, on the basis of fixed and mobile measurements. We have analyzed the relationships with urban landscape heterogeneities, while considering the possible influences of seasons and of weather types.
Afterwards, a modeling exercise has been performed with the Town Energy Balance (TEB) model in order to evaluate its capability (and understand some of its limitations) in simulating fine-scale variations in meteorological parameters. For this, different configurations and parameterizations of the model, as well as various input data resolutions (from 25 to 100-m horizontal resolution) have been tested and compared.