Computational Fluid Dynamics

Reynolds-Averaged Navier–Stokes (RANS)

Some numerical studies employing CFD have been carried out since the 1990s by using the Reynolds-Averaged Navier–Stokes (RANS) approach to examine the relevance of such models about pollutant dispersion (Murakaml et al., 1990; Meroney et al., 1999).

Also, CFD has been used in examining its relevance to pollutant dispersion amongst actual buildings based on RANS (Moon et al., 1997; Pontiggia et al., 2010, 2011). However, it was found in this context that a RANS computation frequently leads to high levels of concentration that are not found in the standard measurements. It is pointed out in this context that low concentrations can occur in particular situations through pseudo-diffusivity that generally occurs in RANS computations that are characteristic of coarse grid spacing. After having recently analyzed the performance patterns of varied revised kε models in regard to the dispersion fields adjacent to buildings, the authors found that all RANS computations were characterized with under-prediction in the context of horizontal concentration diffusion, even though a few revised kε models provided more specific outcomes in comparison to the standard kε model (Tominaga and Stathopoulos, 2010).

Similar tendencies were found to occur in several studies that made use of steady-state RANS calculations about pollutant dispersion involving stable atmospheric boundary layers that were neutrally stable. It became apparent from the study that more accurate outcomes were required to be attained through transient simulations.

Steady RANS is the most common approach used in modeling turbulent flows under the assumption that the non-convective transport characterized by the turbulent flow is influenced by stochastic three-dimensional turbulence with broadband spectrum devoid of specific frequency levels, which means that it actually models the complete series of eddy length scales. However, such an approach is characterized by apparent shortcomings and creates severe ambiguities inflows with highly organized structures, such as those that exist in canyons and around buildings. Moreover, RANS approaches are governed by gradient transport that does not always prove to be relevant in the context of pollutant exchange prevailing on the top of a street canyon.

Timed-average solutions are provided by RANS simulations but the entire range of turbulence scales have to be modeled. Nevertheless, RANS is an important means to compute the concentration field and time-averaged flows and to determine the precision of the simulations.

A RANS model characterized by k-turbulence closure was used by Paterson and Apelt (1986, 1989) in arriving at the average pressure and velocity of flows in the vicinity of a flat surface and an individual prismatic building. Zhang et al. (1993) made use of the usual standard k-model and investigated how turbulence and wind shear present in upstream flows impacted flow fields near a single cuboid building. Similarly, Yamada (2002) used the higher-order turbulence closure RANS model on a complex terrain in simulating flows in the vicinity of an individual square-based building. Most of the initial work done in this field has relied on applying the RANS model to investigate three-dimensional patterns amongst flow regimes that characterize an urban canyon (Dawson et al., 1991; Hunter et al., 1992; Baik and Kim, 1999).

Large Eddy Simulation (LES)

Large-eddy simulation (LES) has been used as a numerical technique to integrate spatially filtered equations in the context of motion describing greater levels of Reynolds turbulence and is largely used in conducting turbulence research. However, according to Moeng and Sullivan (2002), this option proves to be very expensive.

The LES methodology is known to resolve the complexities associated with the large unstable turbulence scales and effectively models the impact of smaller motions on the resolved options. Hanna et al (2002) investigated the significance of LES in urban areas, while Cheng et al. (2003), and Xie and Castro (2006) researched the patterns that emerge in the context of series of regular cubes. Similarly, Camelli et al. (2005), Tseng et al. (2006), Michioka and Sato (2009) and Xie and Castro (2009) researched in the area of field scale flows. It has been found that LES capabilities are better than the RANS approach, which is known to have shortcomings on account of the long computing time taken by it.

Extensive research has been done on the use of LES in studying flow patterns in bluff areas (Murakami et al., 1987; Noda and Nakayama., 2003). Even though complexity in configurations has so far prevented the use of pollutant dispersion simulations, Walton et al. (2002) used LES to study pollution dispersion in street canyon conditions in urban areas and found the outcomes to be the same as the ones arrived at through the use of field data and the k–ε model. Similarly, Murakami et al. (1990) conducted wind-tunnel tests in comparing the k–ε model and LES. Zhong et al. (2015) used the LES to study transport and dispersion patterns in typical urban street canyons in regard to the NO, O3, and NO2 species. Liet et al. (2008a) assessed the LES approach by using a water-channel strategy and found that it was in keeping with the structure, turbulent intensities, and velocities of flows.

Comparison between RANS and LES

It is now agreed amongst researchers that in comparison with RANS, LES can provide solutions regarding high-intensity unsteady motions with low levels of modeling. It thus emerges that qualities such as the disparity of wind pressure on buildings occurring on account of large-scale motions are effectively replicated in LES. Also, many researchers have held that LES is a good option because it concurs with experimental data relative to turbulent energy and average velocity around individual buildings (Murakami, 1993; Rodi, 1997; Tominaga et al., 2008a).

Past research studies that have compared LES and RANS in measuring dispersion around buildings have found that LES provides better outcomes in regard to the measurement of concentration distribution, even though the difference between the two in regard to mean velocity is not massive (Xie and Castro, 2009; Dejoan et al., 2010; Santiago et al., 2010; Tominaga and Stathopoulos, 2010, 2011, 2012; Salim et al., 2011; Gousseau et al., 2011a,b). Such outcomes occur because LES effectively reproduces the vertical and horizontal dispersal of concentration. Moonen et al. (2012) held that unsteady flow creates strong a major impact on air exchange in the usual urban conditions.

Comparison of LES and RANS approaches was made by Cheng et al. (2003) by examining a series of cubes and they concluded that both models provided reasonable predictions in terms of mean flows. Xie and Castro (2006) concluded that even though field details were better incorporated through LES, both approaches auger well in terms of comparable outcomes in the context of canopy layers of flows and different series of cubes.

Detached-Eddy Simulation (DES)

Introduced in 1997 and employed for the first time in 1999, the Detached-eddy simulation (DES) approach combines both LES and RANS approaches in assuming that they are individually insignificant in resolving the problem in hand (Spalart et al. 1997). Therefore, it proves to be more useful than both LES and RANS. In effect, after its introduction, several DES committees were formed with its critics as well as propagators, which led to the emergence of some DES branches.

It can be said that DES has a positive future even though it is still in its infancy stage and is being constantly improved Greschner et al. 2008). It is believed to be a hybrid of RANS and LES and can be effectively used in different industries in the context of complete boundary layers and RANS functioning in the future. Nevertheless, the approach is believed to be quite complex and efforts are being made to arrive at greater predictability under different grid conditions. It is also believed that the numerical standards of codes will improve and attached flows will become more acceptable. According to Peng and Haase (2008), better abilities can be acquired by facilitating and organizing grid generation and establishing guidance practices that lead to systematic improvements in hybrid and DES approaches.

In making use of the massive information that is available from DNS (Direct Numerical Simulation) and LES, several studies have been conducted in investigating the ways through which pollutants disperse in the vicinity of buildings (Gousseau et al., 2011b, 2012). Rossi et al. (2010) conducted research and found that the counter-gradient systems controlling the mass transfer of turbulence are mostly present in the stream-wise directions, which implies that negative turbulence takes place despite the declining concentration in the center of the trail. However, the dispersion patterns of turbulence strengthen the correctness of the hypothesis relative to gradient diffusion as apparent in the lateral and vertical directions.

Branford et al. (2011) researched by using DNS results and substantiated on the processes that impact the varied plume structures such as plume-skewing, secondary source dispersion, detrainment, lateral dispersion, and channeling. It is very important to make numerical predictions about turbulence. However, constant concerns relative to the conceptual aspects of nature have the potential of classifying DSS as an approach that can be intuitively construed as right and most successful, but it does not always satisfy perfectionists. It is thus important to deal with these issues and with the practical aspects associated with them in the remaining sub-sections.

Of late, a large number of studies have been conducted on DNS and LES in the context of dispersion fields relating to obstacle arrangements that are apparent in such studies. In this regard, Boppana et al. (2010) found there was a strong impact emanating from the morphology of roughness on dispersion patterns and the strength of LES in arriving at physically significant information about flux relative to scalar turbulence. Branfort et al. (2011) researched DNS and found that significantly helped in understanding the processes that impact the plume structures; such as detrainment, lateral dispersion, and channeling. In the context of turbulent flows in tubes, it has been found that it is possible to use LES and DNS in studying dilute dispersed systems. LES appears to be less adaptable in the context of the higher Reynolds numbers because the spatial resolution is absent in situations when it is combined with simpler sub-grid approaches.

Evaluation of CFD models

According to Blocken and Gualtieri (2012), the Validation process has been acknowledged to be a crucial area of model evaluation in providing quantitative and qualitative opinions about the model. It helps a great deal in ascertaining the confidence and accuracy of the model outcomes, but it is extremely difficult to know how validation can be done based on accuracy as the concentration of pollutants adjacent to buildings varies at different times.

For example, in making a comparison of data on wind tunnels, the German VDI Guideline mandates a specific hit rate in some test cases (VDI, 2005). Chang and Hanna (2004) provided additional typical values, while Britter and Schatzmann (2007) came up with new models of evaluating urban meteorology to the micro-scale, which helps in predicting pollutant dispersion and flows in the context of urban canopy layers.

Schatzmann et al. (1997) and Schatzmann and Leitl (2002, 2011) overviewed an operational evaluation process and a model validation process in highlighting the responsibilities and needs of model users. However, predictions about plume coverage and location need to be accurate (Brown, 2004). From this perspective, past research on assessing plume location has limited applicability (Chang and Hanna, 2004; COST Action 732, 2009). However, according to Santiago et al. (2010), this kind of analysis has the potential of being misleading given the limited availability of data.

It is very important to measure fluctuations in concentration levels around buildings, specifically in situations characterized by odorous, flammable, or toxic elements. Prediction results by way of the extent to which there is a percentile increase in fluctuations are more important than the absolute maximum outcomes. This allows the outcomes to be stronger and more realistic. Overall, LES proves to be more advantageous about the RANS model (Schatzmann and Leitl, 2011; Harms et al., 2011).

Cite this paper

Select style


Premium Papers. (2021, June 18). Computational Fluid Dynamics. Retrieved from


Premium Papers. (2021, June 18). Computational Fluid Dynamics.

Work Cited

"Computational Fluid Dynamics." Premium Papers, 18 June 2021,


Premium Papers. (2021) 'Computational Fluid Dynamics'. 18 June.


Premium Papers. 2021. "Computational Fluid Dynamics." June 18, 2021.

1. Premium Papers. "Computational Fluid Dynamics." June 18, 2021.


Premium Papers. "Computational Fluid Dynamics." June 18, 2021.