Accurate predictions of wind and dispersion in the atmospheric boundary layer (ABL) can provide essential information to support design and policy decisions for sustainable urban areas. However, computational fluid dynamics (CFD) predictions of the ABL have several sources of uncertainty that can affect the results. An important uncertainty is the definition of the inflow boundary condition, which is influenced by larger scale weather phenomena. In this paper, we propose a method to quantify the effect of uncertainty in the inflow boundary conditions using input from an ensemble of mesoscale simulations. The mesoscale mean velocity and turbulent kinetic energy at the inflow of the CFD domain are used to define probability density functions for the uncertain wind direction and magnitude. A non-intrusive method is used to propagate these uncertainties to the quantities of interest. The methodology is applied to two different cases for which field experimental data are available: the Askervein hill and the Joint Urban 2003 measurements. For the latter case, the results are similar to those of a previous study that characterized the uncertain input parameters based on measurements. Hence, the results show that the proposed mesoscale simulation-based approach provides a valuable alternative in absence of sufficient measurement data.

Computational fluid dynamics predictions of urban flow are subject to several sources of uncertainty, such as the definition of the inflow boundary conditions or the turbulence model. Compared to Reynolds-averaged Navier-Stokes (RANS) simulations, large eddy simulations (LES) can reduce turbulence model uncertainty by resolving the turbulence down to scales in the inertial subrange, but the presence of other uncertainties will not be reduced. The objective of this study is to present an initial investigation of the relative importance of these different types of uncertainties by comparing urban flow predictions obtained using RANS and LES to field measurements. The simulations are designed to reproduce measurements performed during the Joint Urban 2003 field experiments. The time-averaged velocity measured at an upstream wind sensor is used to define the inflow boundary condition, and the results are compared to time-averaged measurements at 34 locations in the downtown area. For the turbulence kinetic energy, the LES is found to be more accurate than the RANS in 80% of the available high-frequency measurement locations. For the mean velocity field, this number reduces to 50% of all stations. Comparison of the LES results with a previous inflow uncertainty quantification study for RANS shows that locations where the LES is less accurate than the RANS correspond to locations where the RANS solution is highly sensitive to the inflow boundary conditions. This suggests that inflow uncertainties can be a dominant factor, and that their effect on LES results should be quantified to guarantee predictive capabilities.

Large-eddy simulations (LES) of the atmospheric boundary layer (ABL) require the specification of a turbulent inflow condition with appropriate turbulence intensities and length scales. When using a synthetic turbulence generator, the statistics obtained downstream of the inlet might deviate considerably from the intended values. In the present work we propose a fully automated approach to modify the input parameters for the turbulence generator such that the desired turbulence statistics are obtained at the downstream location of interest. The method employs a gradient-based optimization in combination with the divergence-free version of the digital filter method developed by Xie and Castro [1, 2]. A sensitivity analysis showed that the spanwise and vertical Reynolds stresses and length scales are the most influential input parameters. Hence, the optimization adjusts these parameters until the desired turbulence statistics are obtained downstream in the domain. The results demonstrate the promising capabilities of the method: the mean velocity profile is correctly maintained using an appropriate wall function, while the optimization results in Reynolds stresses, integral length-scales and turbulence spectra that compare well to ABL wind tunnel measurements.

Detailed aerodynamic information of local wind flow patterns in urban canopies is essential for the design of sustainable and resilient urban areas. Computational Fluid Dynamics (CFD) can be used to analyze these complex flows, but uncertainties in the models can negatively impact the accuracy of the results. Data assimilation, using measurements from wind sensors located within the urban canopy, provides exciting opportunities to improve the quality of the predictions. The present study explores the deployment of several wind sensors on Stanford’s campus to support future validation of CFD predictions with uncertainty quantification and data assimilation. We focus on uncertainty in the incoming wind direction and magnitude, and identify optimal sensor placement to enable accurate inference of these parameters. First, a set of Reynolds-averaged Navier-Stokes simulations is performed to build a surrogate model for the local velocity as a function of the inflow conditions. Subsequently, artificial wind observations are generated from realizations of the surrogate model, and an inverse ensemble Kalman filter is used to infer the inflow conditions from these observations. We investigate the influence of (1) the sensor location, (2) the number of sensors, and (3) the presence of noise or a bias in the measurement data. The analysis shows that multiple roof level sensors should enable robust assimilation of the inflow boundary conditions. In the future field experiment, sensors will be placed in these locations to validate the methodology using actual field measurement data.

Numerical simulations of flow and pollutant dispersion in urban environments have the potential to support design and policy decisions that could reduce the population’s exposure to air pollution. Reynolds-averaged Navier-Stokes simulations are a common modeling technique for urban flow and dispersion, but several sources of uncertainty in the simulations can affect the accuracy of the results. The present study proposes a method to quantify the uncertainty related to variability in the inflow boundary conditions. The method is applied to predict flow and pollutant dispersion in downtown Oklahoma City and the results are compared to field measurements available from the Joint Urban 2003 measurement campaign. Three uncertain parameters that define the inflow profiles for velocity, turbulence kinetic energy and turbulence dissipation are defined: the velocity magnitude and direction, and the terrain roughness length. The uncertain parameter space is defined based on the available measurement data, and a non-intrusive propagation approach that employs 729 simulations is used to quantify the uncertainty in the simulation output. A variance based sensitivity analysis is performed to identify the most influential uncertain parameters, and it is shown that the predicted tracer concentrations are influenced by all three uncertain variables. Subsequently, we specify different probability distributions for the uncertain inflow variables based on the available measurement data and calculate the corresponding means and 95% confidence intervals for comparison with the field measurements at 35 locations in downtown Oklahoma City.

When dealing with Atmospheric Boundary Layer (ABL) simulations, commercial computational fluid dynamics (CFD) acquires a strategic resonance. Thanks to its good compromise between accuracy of results and calculation time, RANS still represents a valid alternative to more resource-demanding methods. However, focusing on the models’ performances in urban studies, LES generally outmatches RANS results, even if the former is at least one order of magnitude more expensive. Consequently, the present work aims to propose a variety of approaches meant to solve some of the major problems linked to RANS simulations and to further improve its accuracy in typical urban contexts. All of these models are capable of switching from an undisturbed flux formulation to a disturbed one through a local deviation or a marker function. For undisturbed flows, a comprehensive approach is adopted, solving the issue of the erroneous stream-wise gradients affecting the turbulent profiles. Around obstacles, Non-Linear Eddy-Viscosity closures are adopted, due to their prominent capability in capturing the anisotropy of turbulence. The purpose of this work is then to propose a new Building Influence Area concept and to offer more affordable alternatives to LES simulations without sacrificing a good grade of accuracy.

Large-eddy simulations (LESs) are frequently used to model the planetary boundary layer, and the choice of the grid cell size, numerical schemes and sub grid model can significantly influence the simulation results. In the present paper the impact of grid spacing on LES of an idealized atmospheric convective boundary layer (CBL), for which the statistics and flow structures are well understood, is assessed for two mesoscale models: the Regional Atmospheric Modeling System (RAMS) and the Weather Research and Forecasting model (WRF). Nine simulations are performed on a fixed computational domain (6 × 6 × 2 km), combining three different horizontal (120, 60, 30 m) and vertical (20, 10, 5 m) spacings. The impact of the cell size on the CBL is investigated by comparing turbulence statistics and velocity spectra. The results demonstrate that both WRF and RAMS can perform LES of the CBL under consideration without requiring extremely high computational loads, but they also indicate the importance of adopting a computational grid that is adequate for the numerical schemes and subgrid models used. In both RAMS and WRF a horizontal cell size of 30 m is required to obtain a suitable turbulence reproduction throughout the CBL height. Considering the vertical grid spacing, WRF produced similar results for all the three tested values, while in RAMS it should be ensured that the aspect ratio of the cells does not exceed a value of 3. The two models were found to behave differently in function of the grid resolution, and they have different shortcomings in their prediction of CBL turbulence. WRF exhibits enhanced damping at the smallest scales, while RAMS is prone to the appearance of spurious fluctuations in the flow when the grid aspect ratio is too high.

Reynolds-averaged Navier–Stokes (RANS) simulations are often used in the wind engineering practice for the analysis of turbulent bluff body flows. An approach that allows identifying the uncertainty related to the use of reduced-order turbulence models in RANS simulations would significantly increase the confidence in the use of simulation results as a basis for design decisions. In the present study we apply a strategy that enables quantifying these uncertainties by introducing perturbations in the Reynolds stress tensor to simulations of the flow in downtown Oklahoma City. The method is combined with a framework to quantify uncertainties in the inflow wind direction and intensity, and the final result of the UQ approach is compared to field measurement data for the velocity at 13 locations in the downtown area.

Computational Fluid Dynamics (CFD) methods are widely used to investigate wind flow and dispersion in urban environments. Validation with field experiments that represent the full complexity of the problem should be performed to assess the predictive capabilities of the computations. In this context it will be necessary to quantify the effect of uncertainties in simulations of the full-scale problem. The present study aims at quantifying the uncertainty related to the variability in the inflow boundary conditions for Reynolds-averaged Navier–Stokes (RANS) simulations of the flow in downtown Oklahoma City to address validation with the Joint Urban 2003 field measurements. Three uncertain inflow parameters were defined: the wind speed and wind direction at a reference height, and the aerodynamic roughness in the logarithmic velocity inlet profile. An ensemble of 729 RANS simulations were performed to determine the polynomial chaos expansion coefficients that define the response surfaces for the velocity magnitude and direction at 13 field measurement stations, and the results are compared to the experimental data. For the velocity magnitude the mean experimental velocity magnitude is encompassed within the 95% confidence interval for the magnitudes predicted by the Uncertainty Quantification study in all stations. For the velocity direction this holds in 11 out of 13 locations. The study demonstrates the significant potential of applying advanced uncertainty quantification methods to address validation with field measurements and to develop a more realistic approach to the definition of inflow boundary conditions in atmospheric CFD simulations.