Modeling urban flows using data-driven methods – MUFDD

Our ambition is to demonstrate that scattered and potentially noisy observations in space and time can offer much more by i) enriching physically-based models, ii) combining at best different sources of information (data assimilation), and even iii) discovering nonlinear models. This project will address challenges posed by this data revolution with the aim to obtain a simplified dynamical model of the turbulent flow within the urban canopy, at the scale of the street. In a first step, we will derive such models using data from numerical simulations limited to the portion of interest of the urban canopy, and/or to relatively low values of Reynolds number. In a second step, data from laboratory experiments measured within the canopy and/or the more accessible atmospheric boundary layer flow above will be used. In a final step, data from an already available full-scale field experiment will also be considered. To focus
our efforts on the development of novel methods and their thorough validation, the investigated flow configurations will be mostly limited to the street canyon model which offers all the basic features of urban flows. In addition, the study will be extended to buoyancy-driven flows by considering a non-isothermal street canyon in all of the above mentioned steps and dispersion processes by performing scalar measurements in laboratory facilities.

Since the start of the project (January 2023), we have been focusing on the study of the impact of the presence of an upstream tall building on the flow within the street canyon immersed in a realistic urban environment (1:200 scale model of the city center of Nantes).

The spatial complexity and heterogeneity of the urban canopy render accurate simulations of the canopy flow out of reach with the current computational capabilities. This type of simulation also requires the use of models to account for the presence and influence of the urban fabric. The models derived in this project with the different described data-driven strategies could be used as surrogate models of the canopy region. In the future, they could serve as “low-computational cost” but accurate tools to predict the unsteady transport processes based on online measurements of sparse data in the canopy or above. They will therefore constitute valuable tools for investigating the urban micro-climatology, for assessing the unsteady street ventilation and the short-term air-quality and for addressing issues due to accidental exposure. In the long run, the exploitation of the scientific results obtained during the proposed project could take the form of packages or numerical modules to be included in commercial CFD softwares to account more precisely for the unsteady turbulence existing in urban areas for the prediction and estimation of air quality or scalar dispersion in such regions. Models will also be developed with the final aim of being implemented in meso-scale computational codes to serve as novel wall or canopy models that account for the presence of the urban canopy and its interaction with the atmospheric flow.

Du, H., Savory, E., and Perret, L. “Effect of morphology and an upstream tall building on the mean turbulence statistics of a street canyon flow”. In: Building and Environment 241 (2023). doi : 10.1016/j.buildenv.2023.110428 .

The objective of MUFDD is to model in a data-driven framework the unsteady ventilation and transport processes taking place in the urban canopy. The project’s ambition is to exploit recent advances in data-driven modelling (among which Machine Learning algorithms) and data assimilation to reproduce accurately and rapidly the key features of the dynamics of interest at the scales of the street or neighbourhood. The project will cover from physically-based modelling to pure data-driven strategies capable of discovering non-linear models. These models will then be enriched with sequential data assimilation techniques using data from LES simulations, laboratory experiments and full-scale terrain. Besides gaining a deeper understanding of the flow physics, the expected outcomes are the development of operational models for investigating the urban micro-climatology, assessing the unsteady street ventilation and the short-term air-quality, and addressing accidental exposure issues.