Dataset of microscale atmospheric flow and pollutant concentration large-eddy simulations for varying mesoscale meteorological forcing in an idealized urban environment

• DOI: 10.1016/j.dib.2025.111285
• Publication Date: 2025-01-10T07:51:29Z
• Authors: Eliott Lumet; Thomas Jaravel; Mélanie C. Rochoux
• Keywords: Microscale chemistry; Forcing (mathematics); Large-Eddy Simulation

By 2050, two-thirds of the world's population will live in urban areas under climate change, exacerbating environmental and public health risks associated with poor air quality heat island effects. Assessing these requires development microscale meteorological models that quickly accurately predict wind velocity pollutant concentration high resolution, as heterogeneity environments leads to complex patterns strong gradients. Computational Fluid Dynamics (CFD) has emerged a powerful tool address this challenge by providing obstacle-resolved flow dispersion predictions. However, CFD are very expensive require intensive computing resources, which can hinder their systematic use practical engineering applications. They also subject significant uncertainties, particularly those arising from mesoscale forcing internal variability atmospheric boundary layer, some aleatory thereby irreducible. Given issues, construction datasets account for uncertainty would be an interesting avenue research science. In context, we present PPMLES (Perturbed-Parameter ensemble MUST Large-Eddy Simulations) dataset, consists 200 large-eddy simulations (LES) characterizing interactions between turbulent airflow, tracer dispersion, idealized environment. These reproduce canonical field campaign while perturbing model's parameters. includes time series at human height within built environment track release over time. complete 3-D fields first- second-order temporal statistics concentration, sub-metric resolution. The induced layer is provided. computation required consuming 6 million CPU core hours, equivalent emission approximately 10 tCO2eq greenhouse gases. This computational effort carbon footprint motivates sharing data generated. added value dataset twofold. First, perturbed-parameter LES enables quantify understand effects been identified major predicting environments. Secondly, reference used benchmark different levels complexity, extract key information about physical processes involved inform more operational modeling approaches, example through learning surrogate models.