Review of CFD Guidelines for Dispersion Modeling

This is the review of CFD (Computational Fluid Dynamics) guidelines for dispersion modeling in the USA, Japan and Germany. Most parts of this review are based on the short report of the special meeting on CFD Guidelines held at the International Symposium on Computational Wind Engineering (CWE2014), University of Hamburg, June 2014. The objective of this meeting was to introduce and discuss the action program to make worldwide guidelines of CFD gas-dispersion modeling. The following six gas-dispersion guidelines including Verification and Validation (V&V) schemes are introduced by each author; (1) US CFD guidelines; (2) COST/ES1006; (3) German VDI (Verein Deutscher Ingenieure) guidelines; (4) Atomic Energy Society of Japan; (5) Japan Society of Atmospheric Environment; (6) Architectural Institute of Japan. All guidelines were summarized in the same format table shown in the main chapters in order to compare them with each other. In addition to the summary of guidelines, the overview of V&V schemes and many guidelines of CFD modeling in the USA are explained.


Introduction
Computational Fluid Dynamics (CFD) is widely acknowledged today as a useful tool to assess problems in engineering, process sciences, power production, pharmaceutical and biomedical science, and the geophysical sciences.The breadth of application is revealed by the wide range of applications and specialty fields and the annual conferences found on the internet links of CFD Online [1].It is also now extensively accepted for regulatory purposes applied to dispersion aspects of nuclear safety, environmental impact assessment, architectural design and ventilation.However, the protocols and regulations concerning CFD application in dispersion modeling are not uniform.The present paper is a review of the existing CFD guidelines applied to its use focusing on atmospheric dispersion modeling.
Air pollution and noxious gas releases are a common problem of commercial, industrial, health facilities, and even residences throughout the world.As a result, engineers and scientists have sought means to predict and mediate the dispersion and transport of such gases.In the past, field scale measurements and fluid modeling with wind-and water-tunnels have been used to study and understand the mixing processes.However, recently, CFD has also become an important tool Fluids 2016, 1, 14 3 of 16 simulation validation.This can replace the earlier approach found in the historic scientific methodology that does not have a place for computing and simulation.A relationship diagram between the steps of such a methodology is shown in Figure 1 [27].The outer circle together with data validity are the technical processes that must be addressed to show that a model is credible.Assessment activities are spawned from each of these technical processes [28].
Fluids 2016, 1, 14 3 of 15 between the steps of such a methodology is shown in Figure 1 [27].The outer circle together with data validity are the technical processes that must be addressed to show that a model is credible.Assessment activities are spawned from each of these technical processes [28].
It is not really possible to say conclusively that our CFD solutions have a known accuracy and reliability.We can only point to our verification and validation bench marks.Even when one has carefully considered equation verification and application validation, there are still issues concerning Error Propagation.Precision with computers does not equate to accuracy.Small errors in initial and boundary conditions can propagate into solutions [29].
One must also realize that validation is a learning process, and the process is never complete.Inherent uncertainties can exist in model physics, initial conditions, boundary conditions, and even turbulent randomness which can exist as a barrier to repeatability of both physical experiments and numerical comparisons.

CFD Guidelines Used in the United States
To a large extent, V&V practice in the United States builds upon the experience and recommendations of previous CFD modelers from different disciplines and different fluid mechanics communities.One can also classify advice into several subcategories:  It is not really possible to say conclusively that our CFD solutions have a known accuracy and reliability.We can only point to our verification and validation bench marks.Even when one has carefully considered equation verification and application validation, there are still issues concerning Error Propagation.Precision with computers does not equate to accuracy.Small errors in initial and boundary conditions can propagate into solutions [29].
One must also realize that validation is a learning process, and the process is never complete.Inherent uncertainties can exist in model physics, initial conditions, boundary conditions, and even turbulent randomness which can exist as a barrier to repeatability of both physical experiments and numerical comparisons.

CFD Guidelines Used in the United States
To a large extent, V&V practice in the United States builds upon the experience and recommendations of previous CFD modelers from different disciplines and different fluid mechanics communities.One can also classify advice into several subcategories: AESJ guidelines [30] were made to utilize CFD models for dispersion calculations for nuclear safety evaluation of radiation dose.In Japan, wind tunnel experiments have been used for nuclear safety evaluation for more than 40 years by Japanese electric power companies.Today, it is customary to utilize CFD models instead of wind tunnel experiments for a preliminary evaluation to confirm the effect of additional structures on gas dispersion.
An outline of the guidelines is summarized in Table 1 in the same format as the other guidelines introduced in this paper.It took two years to complete the guidelines by the JAES (Japan Atomic Energy Society) working group, and one year was consumed to establish a V&V scheme based on comparisons with experimental wind tunnel data.Criteria were integrated from the COST (European Cooperation on Science and Technology) 732 guidelines [31] and the AESJ wind tunnel guidelines [32].A safety factor is defined to be modified with the final calculated results based on the uncertainty factor of calculated results suggested in the ASME (American Society of Mechanical Engineering) V&V guidelines [33].
The AESJ guidelines will be revised periodically every five years based on the established general rules of the AESJ guidelines.Verification criteria (1) Wind velocity profile: 1/7 power law (2) Boundary layer thickness: Over 400 m (3) Turbulent intensity of downwind direction at Z = 30 m: 8%-16% (4) Vertical dispersion width: Atmospheric stability C-D

Validation criteria
Step-1: Concentration distribution around rectangular box (1) Ground level concentration along the plume axis: FAC2 > 0.89 (2) Total spatial concentration : FAC2 > 0.54 Step-2: Effective source height of actual nuclear station (1) Slope of the regression line: 0.9-1.1 (2) Correlation coefficient: Over 0.9 (3) Variation coefficient: Under 20 Uncertain analysis the calculated effective source height is adjusted by the variation coefficient Remarks Every model must be validated with the disclosed wind tunnel data at the actual nuclear power station, which will be applied.

Public comments
No public comments Although there is a rule reviewed by the JAES's working group

Best Practice Guidelines Defined by COST Action ES1006 in EU
The COST Action ES1006 was established to pursue a substantial improvement in the implementation of local-scale emergency response tools.The main objective is to evaluate and improve the reliability of local-scale emergency response tools on the basis of a comprehensive, concerted and harmonized cross-national approach.The main focus is the evaluation of the atmospheric dispersion models when used in urban or industrial environments with complex building structures, and their integration in emergency response systems.
The COST ES1006 Action is structured in three Working Groups, whose main activity is hereafter briefly summarized.
Working Group 1-Threats, Models and Data Requirements: characterizes and categorizes existing models as well as typical release scenarios.
Working Group 2-Test, Evaluation and Further Development: defines open and blind test scenarios, tests and assesses different modelling approaches, and outlines scientific strategies for improving the implementation of corresponding tools.
Working Group 3-Applicability, Implementation and Practical Guidance: deals with the practical constraints in the use of local-scale emergency response models.
The Best Practice Guidelines (BPG) [35] are formulated as a concise and focused document that can be straightforwardly consulted by potential users.The most important aspects that guide the final recommendations are summarized above, while, for in-depth analysis, one is referred to the specific and detailed publications documents prepared in the frame of the COST ES1006 Action work.The outline of BPG [35] is summarized in Table 2. JSAE guidelines [36] were made to utilize CFD models for environmental impact assessment for relatively short distance dispersion in urban complex.The CFD model used for this purpose is named "Diffusion Model with Computational Fluid Dynamics" abbreviated as DiMCFD to distinguish it from other popular CFD models which are mainly used for the analysis of fluid dynamics.In Japan, Gaussian plume and puff models have been used for the environmental impact assessment in the atmosphere for long time; however, it is apparent that Gaussian models produce large errors for dispersion in urban complex in which there are many buildings and complicated emission sources.Wind tunnel experiments were previously employed for impact assessment in urban areas; however, now relatively few wind-tunnel studies are performed for this purpose because of the higher costs.It is planned to utilize DiMCFD models instead of wind tunnel experiments and to confirm whether the DiMCFD actually give better performance than Gaussian models in urban areas.
Outline of the guidelines is summarized in Table 3 in the same format as the other guidelines introduced in this paper.It took four years to complete the guidelines by the working group of the JSAE.Criteria were also integrated from the COST 732 guidelines [31] and AIJ's (Architectural Institute of Japan) Guidebook for Practical Applications of CFD to Pedestrian Wind Environment around Buildings [37].

Revision period Not specified
Referenced document (1) EPA guideline on the development, evaluation, and application of environmental models [39] (2) COST 732 report [31] (3) AIJ, Guidebook for practical applications of CFD to pedestrian wind environment around buildings [37] Working group members Chairman: Dr. Tateki Mizuno Members: Experts, users (consultants for environmental impact assessment), government Fluids 2016, 1, 14 7 of 16

Guidelines for Microscale Modeling Defined by VDI in Germany
In Germany, the Commission on Clean Air within VDI (Verein Deutcher Inginieure) and DIN (Deutsches Institut für Normung) ensures fitness-for-purpose of tools and methods in environmental meteorology through the preparation of technical guidelines.The guideline Prognostic micro-scale wind field models-Evaluation for flow around buildings and obstacles was published in 2005, and the draft review has been published in 2015 for public consultation [40].The guidelines consider micro scale meteorology models that typically calculate wind, turbulence, temperature, humidity in the vicinity of buildings considering orography and Coriolis force effects.A typical use of this type of model is environmental impact assessment.For this, the results of models evaluated according to this guideline can be used to calculate tracer dispersion.However, a specific guideline for the evaluation of tracer dispersion does not currently exist.
These guidelines ensure fitness-for-purpose of obstacle resolving prognostic models through an evaluation procedure involving three distinct steps: general evaluation, scientific evaluation and validation.The general evaluation checks that the model is comprehensible in the sense that it is well documented and the evaluation is traceable.The scientific evaluation checks the set of equations solved and the physical and numerical properties of the model by requiring e.g., that all three wind components are calculated from prognostic equations and that orographic and obstacle structure is explicitly resolved.It assesses whether Coriolis force, temperature and possibly phase changes are considered depending on the application range.Finally, the model results are validated based on comparisons with results of the same model, with analytic solutions and with wind tunnel data.Here, the guidelines do not distinguish between verification and validation.The comparison metric in all cases is the hit rate q which is calculated as q = (Σn i )/N, where otherwise.P i and O i are co-located normalized model results and comparison data, respectively, and N is the number of measurement points.W (absolute allowed deviation) and D (relative allowed deviation) are defined for each test case and the model is required to reach a minimal hit rate for each comparison in order to fulfill the guidelines (see Table 4 below for values).This metric reflects the aim of the guideline to identify fitness-for-purpose rather than ranking models according to their performance.To test fundamental properties such as homogeneity of results over homogeneous terrain, symmetry of the results for symmetric obstacles and stationarity for stationary boundary conditions model results are compared with results of the same model.To ensure that the results are physically realistic, wind profiles for an empty domain are compared with an analytic solution.Finally, several test cases involve the comparison with reference data from wind tunnel measurements.The obstacle configurations include individual obstacles of different shapes (beam, cube, cuboid) as well as one case with a highly complex building configuration corresponding to that of a generic European urban center.The comparison data used in the guideline is based on the CEDVAL (Compilation of Experimental Data for Validation of Microscale Dispersion Models) database (EWTL (Environmental Wind Tunnel Laboratory), 2016) [37].
Since their first publication in 2005, the guidelines have been applied to several microscale models [41][42][43][44].The guidelines are revised when deemed necessary (e.g., when new relevant datasets become available), normally every five years.An update of the guidelines is currently in public consultation.Comparison metric: Hit rate q = (Σn i )/N is the fraction of measurement points for which the model result differs by less than the allowed absolute (W) or relative (D) difference from the comparison data.The hit rate is required to be above a threshold for each case in order for a model to fulfill the test criterion.(2) COST 732 (2010): Final report [31].
( Computer facilities and computational fluid dynamics (CFD) software have significantly improved in recent years; therefore, using CFD to predict and assess the pedestrian wind environment around buildings has become practical at the building design stages.Consequently, guidelines that summarize the important aspects relevant to using the CFD technique for appropriately predicting the pedestrian wind environment are a necessity.In 2008, the Architectural Institute of Japan (AIJ) proposed practical guidelines for applying CFD to the pedestrian wind environment around buildings [45].These guidelines were based on the results of benchmark tests, in which cross comparisons were conducted between the results of the CFD predictions, wind tunnel tests, and the field measurements of seven test cases.The aim of these seven test cases was to investigate the influence of numerous types of computational conditions for various flow fields [46][47][48].Researchers worldwide have been referring to these guidelines, as is evident from the 335 citations in ScienceDirect up to 2015.
The abovementioned guidelines and the relevant investigations mainly targeted the problems pertaining to the strong wind around high-rise buildings.However, major environmental problems (urban heat island, thermal environment, air pollution, and particle dispersion, etc.) are often caused in weak wind regions, such as at the back of buildings and within street canyons.Generally, CFD predictions relevant to a weak wind flow region are more challenging than those for an area of strong wind flow.Therefore, although the previous guidelines are quite effective in the pollutant Fluids 2016, 1, 14 9 of 16 dispersion, the guidelines needed to be re-examined and extended in terms of their applicability to comprehensive environmental problems.Furthermore, while the abovementioned guidelines mainly focus on steady RANS (Reynolds-averaged Navier-Stokes equations) models, it is highly expected to establish appropriate practical guidelines for application of Large Eddy Simulation (LES).For this purpose, the benchmark tests are being re-conducted, using the new reliable experimental database provided by the AIJ group, as well as the previous guidelines.In the benchmark tests, a systematic approach is being taken, i.e., from a simple to a complicated case, and the influence of various computational conditions on the prediction results is being investigated methodically.Parts of the results of the benchmark tests have already been published in scientific literature [49][50][51][52][53][54].In addition, the endeavor to extend the guidelines to include urban pollutant/thermal dispersion problems is continuing as noted in Table 5.

Applied purpose
Environmental assessment in built-up area (pedestrian wind environment, atmospheric air pollution, and thermal environment, etc.)

Validation criteria
Intended codes and methods have to be validated by the results of the benchmark tests provided by our group: (1) A single building (2) Building arrays (3) Actual buildings Under discussion, pertaining to the criteria for validation metrics.

Uncertainty analysis
Uncertainty in the measurements of the benchmark tests are considered in evaluating the accuracy of CFD.

Remarks
Many wind tunnel experiments and computations, using different CFD codes, are being conducted to investigate the influence of several types of computational parameters for various flow fields.The guidelines will derive from the findings, based on these comparisons.The contents of the original AIJ guidelines will be included as part of the extended guidelines.

Public comments
The guidelines will be reviewed by specialists who are not members of the committee.

Working group members
Researchers and consultants from universities and construction companies who have expertise in CFD: A. Mochida (Committee chair), Y. Tominaga (Secretary), R. Yoshie (WG chair)

Conclusions
The intent of this paper was the review of CFD guidelines used by different national professional bodies as currently proposed to guide the realistic prediction of dispersion and transport in the atmosphere, within and around buildings, terrain features and urban complexes.This paper is not a comprehensive review of CFD modeling techniques, nor is it intended to provide a final specification of best practices to be used by the modeling community.There are many similarities in the protocols currently employed as well as differences between individual guidelines.At some point in the future, it may be possible to arrive at a consensus for an international guideline for dispersion modeling, but, in the meantime, this paper summarizes existing methodologies.The paper provides a basis for discussion and improved guidance.

( 1 )
Qualitative analysis: Draw the following three plots and confirm that the scatters are less than those of examples in the guideline.(a) Scatter plot with correlation and regression coefficients (b) Quantile-Quantile plots (c) Residual analysis through residual plots (2) Quantitative analysis (COST 732 [31]) FAC2 > 0.5, |FB| < 0.2, 0.7 < MG < 1.3, NMSE < 4, VG < 1.6 Uncertain analysis Included in validation process Remarks Treat only neutral stability and stationary state Public comments No public comments.The contents were opened in general assembly of JSAE and reviewed by a third party.

VDI 3783 Part 9 .
Environmental Meteorology.Prognostic Micro-Scale Wind Field Models.Evaluation for Flow around Buildings and Obstacles (2015) [40] Organization Commission on Clean Air within VDI (The Association of German Engineers) and DIN Recommended model None specified (any obstacle resolving prognostic model that fulfills the criteria of the guideline Output 3d wind and turbulence fields Applied purpose Model evaluation in the context of air quality assessment near buildings Validation and Verification criteria (guideline classifies all as validation)

( 1 )
Model internal comparison for (a) stationary solution; (b) scalability; (c) symmetry; (d) homogeneity; (e) grid independence; (f) grid orientation.(a-e) W = 0.01-0.05,D = 0.05, q ě 0.95; (f) W = 0.06, D = 0.25, q ě 0.66 Comparison with analytic solution (a) wind profile; (b) ageostrophic angle.(a,b) W = 0.01, D = 0.05, q ě 0.95 Comparison with wind tunnel data for (a) quasi-2d beam; (b) individual cube and cuboid obstacles; (c) multiple obstacles in a realistic urban configuration.D = 0.06-0.08,W = 0.25, q ě 0.66 Uncertainty analysis not quantified Remarks Guidelines aim to identify fitness for purpose in a binary check.No ranking of models.Public comments Public consultation period with every review of the guideline.Revision period Considered for review by Commission on Clean Air every five years or if necessary.Referenced document Environmental Meteorology.Prognostic Micro-Scale Wind Field Models.Evaluation for Flow around Buildings and Obstacles; VDI 3783 Part 9 [40].
for Aerodynamic and Fluid Mechanic Codes,

Table 1 .
Outline of the guidelines defined by Japan Atomic Energy Society.

Table 2 .
Outline of the best practice guidelines defined by COST Action ES1006.

Table 3 .
Outline of the guidelines defined by Japan Society for Atmospheric Environment.

for Atmospheric Environmental Impact Assessment Using Diffusion Model with Computational Fluid Dynamics [36]
Verification criteriaBecause commercial CFD model is usually used, it is assumed that verification as a CFD model has been already finished.

Table 4 .
Outline of the guidelines defined by the Commission on Clean Air (KRdL (Kommission Reinhaltung der Luft)) and German Engineering Society (VDI).
Environmental Wind Tunnel Laboratory (EWTL) Compilation of Experimental Data for Validation of Microscale Dispersion Models (CEDVAL, CEDVAL-LES).

Table 5 .
Outline of the extended AIJ (Architectural Institute of Japan) guidelines (ongoing).