Urban Flow Predictions: What Is Stopping Us?
Biography
Clara García-Sánchez is an associate professor and currently leads the Urban Data Science section within the faculty of Architecture and the Built Environment at Delft University of Technology (TUDelft) in the Netherlands.
Clara completed her degree in aerospace engineering in 2011 at the Polytechnic University of Valencia, Spain. During her degree she performed international stays at Politecnico di Milano, Italy, and École Polytechnique de Montréal, Canada. After finishing her degree, she completed the Research Master in fluid dynamics at the von Karman Institute, Belgium. One year after, she was awarded a grant for Strategic Basic Research (IWT) to pursue her PhD degree in physics, in collaboration between the University of Antwerp and the von Karman Institute, where she graduated in 2017. During her PhD, she performed additional collaborations at Columbia University and Stanford University, USA. Before joining TUDelft, she was a postdoctoral research scientist in the Global Ecology department at Carnegie Institution for Science.
Clara joined TUDelft in 2019 through a Delft Technology Fellowship that supported her research focused on wind engineering problems, specifically addressing dispersion and airflow predictions in the built environment. She is interested in renewable energies, efficient urban processes, as well as mitigation techniques that can allow to prevent and reduce urban pollution. Her research is two folded, on one side it focuses on applying uncertainty quantification methodologies for the prediction of winds in the urban canopy. This approach allows to study the natural variability of the wind or the effects of geometrical uncertainties, which further improves the level of confidence in the numerical results, by adding risk assessment margins based on realistic probabilities. These type of studies grant additional knowledge related to pedestrian wind comfort and pollution maps, information that can be useful to optimize sustainable urbanizations. On the other side she focuses in developing open-source CFD practices that maximize the efficiency of the urban flow simulation pipeline. Some recent developments include City4CFD, an open source tool that allows for the automatic reconstruction of cities at diverse level of detail, depending on the available source data. For further info visit: https://3d.bk.tudelft.nl/gsclara/
Abstract
Urban flow simulations are entering a new era. With rapidly expanding CPU clusters and the accelerating capabilities of multi-GPU systems, running simulations at larger scales is becoming increasingly feasible. Yet, despite these advances, the true bottlenecks remain, critical hurdles that lie in what surrounds the solver.
The pre-processing step which encompasses domain definition, geometry generation, and mesh design remains a persistent obstacle. These steps are often constrained by limited or low-quality data, as well as an incomplete understanding of how size and geometric detail influences urban flow predictions. Meshing, in particular, continues to be one of the most intricate and failure-prone stages in CFD workflows. While open-source automation tools exist, they frequently struggle to cope with the complexity of real urban geometries.
At the other end of the pipeline, post-processing presents a different but equally pressing challenge. As simulations scale up to cover larger urban areas, they produce large datasets that strain storage capacity and complicate data handling, analysis, and visualization.
This talk will explore how we have moved from a compute-centric mindset to address these emerging bottlenecks. By rethinking data acquisition, geometry handling, meshing strategies, and data management, we can unlock the full potential of large scale urban simulations and ensure that increased computational power translates into meaningful scientific and societal advances.
Address
1151 Richmond Street, London, ON, N6A 3K7
cwe2026@uwo.ca
Organized by
