Skip to main content Skip to secondary navigation

Pollutant dispersion in Oklahoma City

Main content start

by Clara Garcia-Sanchez

Project Timeline: 2012-2017

Motivation and Objective

The definition of boundary conditions for CFD simulations of urban dispersion is a major challenge, since they are determined by atmospheric conditions that are highly variable and uncertain. Improving the predictive capabilities of environmental flow simulations therefore requires quantifying the effect of these uncertainties on the simulation outcome. The goal of this project is to establish a novel computational tool that accounts for uncertainties in the inflow boundary conditions and predicts pollutant concentrations with a quantified confidence interval.

Methods and results

We defined a parameterization for the inflow boundary conditions for CFD simulations in terms of random variables that can represent the large-scale variability in the atmospheric boundary layer. Two different methods to inform the definition of the parameterization were explored: using measurement data available from a sensor near the inflow boundary of the simulations, and using a Numerical Weather Prediction code (WRF). The uncertainties are propagated using a polynomial chaos approach. The methodology is validated by performing simulations of Oklahoma City and comparing the results to data from the Joint Urban 2003 field measurements. The plot below demonstrates the predictive capabilities of the approach, in particular when using the measurement data to define the uncertain input parameters. When using the weather prediction code, the main challenge is the significant model uncertainty in the weather prediction, which, when transported through the CFD model, results in a large uncertainty in the predictions (note that the range of the y-axis in the plot on the right is an order of magnitude larger).

Concentration Predictions in OKC

The results indicate the promising predictive capability of simulation tools that represent the natural variability in the wind through uncertainty quantification (UQ). Additional research explored sensitivity analysis to identify the dominant uncertain parameters, as well as large-eddy-simulation and turbulence model UQ to explore the importance of turbulence model uncertainties. The results of this research can inform more reliable decision-making tools for air quality control, emergency response planning, and urban planning. Furthermore, the UQ method is relevant to all wind engineering problems that require realistic predictions of the velocity field in the urban canopy.


This research was funded by a Phd fellowship from the Flemish Institute for Innovation through Science and Technology. The simulations were performed using the Extreme Science and Engineering Discovery Environment (XSEDE), which is supported by National Science Foundation grant number CI-1548562.