gas dispersion - a definitive guide to accidental releases of heavy gases

Refinery Process Stream Purification Refinery Process Catalysts Troubleshooting Refinery Process Catalyst Start-Up / Shutdown Activation Reduction In-situ Ex-situ Sulfiding Specializing in Refinery Process Catalyst Performance Evaluation Heat & Mass Balance Analysis Catalyst Remaining Life Determination Catalyst Deactivation Assessment Catalyst Performance Characterization Refining & Gas Processing & Petrochemical Industries Catalysts / Process Technology - Hydrogen Catalysts / Process Technology – Ammonia Catalyst Process Technology - Methanol Catalysts / process Technology – Petrochemicals Specializing in the Development & Commercialization of New Technology in the Refining & Petrochemical Industries

Web Site: www.GBHEnterprises.com

GBH Enterprises, Ltd.

Process Safety Guide: GBHE-PGP-020

GAS DISPERSION A Definitive Guide to Accidental Releases of Heavy Gases

Process Information Disclaimer

Information contained in this publication or as otherwise supplied to Users is believed to be accurate and correct at time of going to press, and is given in good faith, but it is for the User to satisfy itself of the suitability of the information for its own particular purpose. GBHE gives no warranty as to the fitness of this information for any particular purpose and any implied warranty or condition (statutory or otherwise) is excluded except to the extent that exclusion is prevented by law. GBHE accepts no liability resulting from reliance on this information. Freedom under Patent, Copyright and Designs cannot be assumed.



Process Safety Guide: Gas Dispersion CONTENTS 0 INTRODUCTION 0.1 AIMS 0.2 SCOPE 0.5 FURTHER INFORMATION 1 METEOROLOGICAL PARAMETERS WHICH AFFECT DISPERSION 1.1 ROUGHNESS LENGTHS (Zo) 1.2 WIND SPEEDS 1.3 ATMOSPHERIC STABILITY 1.3.1 General 1.3.2 Pasquill - Gifford Methods of Characterizing Atmospheric Stability 1.3.3 Monin-Obukhov Length Methods of Representing Atmospheric Stability TABLES 1.1 TYPICAL ROUGHNESS LENGTHS 1.2 KEY TO PASQUILL - GIFFORD STABILITY CATEGORIES 1.3 METHOD OF ESTIMATING LEVEL OF INCIDENT RADIATION 1.4 EXAMPLE PASQUILL-GIFFORD STABILITY ANALYSIS FIGURES 1.1 THE EFFECT OF ATMOSPHERIC STABILITY ON PLUME DISPERSION 1.2 RELATIONSHIP BETWEEN PASQUILL-GIFFORD STABILITY CATEGORY AND MONIN-OBUKHOV LENGTH 2 AIR QUALITY STANDARDS 2.1 WHAT ARE AIR QUALITY STANDARDS? 2.2 WHAT AIR QUALITY STANDARDS EXIST 2.2.1 General Background 2.2.2 United States 2.2.3 European Union 2.2.4 The Netherlands 2.2.5 Japan 2.2.6 Taiwan 2.2.7 United Kingdom Air Quality Strategy



2.2.8 Non-Governmental Organizations 2.2.9 Occupational Exposure Limits 2.2.10 General Comparison 2.2.11 Air Quality Standards for Odor Impacts 2.3 WHAT IS THE LAW - AND WHAT ISN’T 2.4 FUTURE DEVELOPMENTS 3 MODEL COMPARISON AND SELECTION

3.1 CLASSIFICATION OF DISPERSION MODELING PROBLEMS 3.2 WHAT MODELS ARE AVAILABLE? 3.3 DESCRIPTION OF AVAILABLE MODELS 3.3.1 General 3.3.2 ADMS (Atmospheric Dispersion Modeling System) 3.3.3 ALOHA (Areal Locations of Hazardous Atmospheres) 3.3.4 DISP2 3.3.5 ISC (Industrial Source Complex) 3.3.6 PHAST (Process Hazard Assessment Tools) 3.3.7 Other Models 3.3.8 Summary of Model Applications 3.4 COMPARISON OF MODEL RESULTS 3.4.1 General 3.4.2 Buoyant gas releases 3.4.3 Dense Gas Dispersion 3.5 RECOMMENDATIONS TABLES 3.1 COMMONLY USED DISPERSION MODELS 3.2 SUMMARY OF MODEL APPLICATIONS



FIGURES 3.1 CATEGORIZATION OF DISPERSION MODELING PROBLEMS 3.2 BUOYANT GAS RELEASE: ADMS RESULTS 3.3 BUOYANT GAS RELEASE : DISP2 RESULTS 3.4 BUOYANT GAS RELEASE: ISC RESULTS 3.5 BUOYANT GAS RELEASE: PHAST RESULTS 3.6 SINGLE PHASE DENSE GAS RELEASE UNDER STABLE ATMOSPHERIC CONDITIONS 3.7 CATASTROPHIC DENSE GAS RELEASE UNDER UNSTABLE ATMOSPHERIC CONDITIONS 3.8 SINGLE PHASE DENSE GAS RELEASE: ALOHA RESULTS 3.9 TWO PHASE DENSE GAS RELEASE: PHAST RESULTS

4 STACK DESIGN

4.1 INTRODUCTION 4.2 STACK DESIGN 4.2.1 Stage A: Preceding Design Work 4.2.2 Stage B: Estimate Mass Emission Rates 4.2.3 Stage C: Identify Acceptable Process Contributions 4.2.4 Stage D: Identify Significant Pollutants 4.2.5 Stage E: Initial Stack Design 4.2.6 Stage F: Model On-site Concentrations 4.2.7 Stage G: Model Off-site Concentrations 4.2.8 Stage H: Assess Results 4.3 FURTHER CASE STUDY FIGURE 4.1 FLOW CHART FOR STACK DESIGN



5 DENSE GAS DISPERSION

5.1 INTRODUCTION 5.2 MODELING METHODOLOGIES

5.2.1 Instantaneous Catastrophic Releases 5.2.2 The Dispersion of a Continuous Dense Gas Plume

5.3 POINTS TO NOTE 5.4 VALIDATION WORK 5.5 DISPERSION MODELS AVAILABLE

5.5.1 DISP2 5.5.2 HGSYSTEM5 5.5.3 ALOHA 5.5.4 PHAST 5.5.5 EFFECTS 5.5.6 GASTAR 5.5.7 LORIMAR Model

FIGURES 5.1 CLOUD SHAPE AS A FUNCTION OF TIME 5.2 BEHAVIOR OF A DENSE GAS PLUME WITH VERTICAL MOMENTUM 6 SOURCE TERMS 6.0 INTRODUCTION 6.1 SOURCE CHARACTERISTICS AND HOLE SIZES 6.1.1 Ammonia Storage Tank Example 6.1.2 Estimation of Hole Sizes 6.1.3 Inventories and Time Dependent Behavior 6.2 THE DISCHARGE OF GASES THROUGH HOLES 6.2.1 Compressible Choked Flow 6.2.2 Compressible Unchoked Flow 6.2.3 Incompressible Flow 6.2.4 Discharge Coefficients



6.3 TWO-PHASE RELEASES 6.3.1 Catastrophic Releases of a Liquefied Gas 6.3.2 Two-Phase Releases Arising from Guillotine Failures of Pipe work 6.4 LIQUID POOL SPREADING AND EVAPORATION 6.5 SOURCE TERMS FOR ENVIRONMENTAL RELEASES 6.5.1 General 6.5.2 A Real Example 6.5.3 A Source Data Checklist for Environmental Applications 6.6 REFERENCES FIGURES 6.1 POSSIBLE RELEASE SCENARIOS FROM A LIQUEFIED AMMONIA STORAGE TANK 6.2 COMPARISON OF PLUME CHARACTERISTICS vs. TARGET DISTANCE 6.3 DIAGRAMMATIC REPRESENTATION OF PSEUDO SOURCE DIAMETER 6.4 EVAPORATION RATE OF CHLORINE FROM AN INSTANTANEOUS 10 TONNE SPILL 7 BUILDING WAKE EFFECTS

7.1 WHY ARE BUILDING WAKE EFFECTS IMPORTANT? 7.2 HOW DO BUILDINGS INFLUENCE ATMOSPHERIC DISPERSION? 7.3 SCIENTIFIC UNDERSTANDING OF BUILDING WAKE EFFECTS 7.4 THE BUILDINGS MODULE IN ADMS: PRINCIPLES 7.5 THE BUILDINGS MODULE IN ADMS: APPLICATION 7.5.1 When Should the Buildings Module be Used? 7.5.2 Points to Note About Using the Buildings Module 7.5.3 Interpreting the Results of the Buildings Module



FIGURES 7.1 THE INFLUENCE OF A BUILDING WAKE ON PLUME DISPERSION 7.2 SCHEMATIC DIAGRAM OF TURBULENT ZONES USED IN ADMS BUILDING MODULE 7.3 EFFECT OF BUILDING WIDTH ON WAKE DISPERSION 7.4 AREAS OF CONCERN DUE TO BUILDING EFFECTS 7.5 THE BUILDINGS MODULE OF ADMS (STABLE CONDITIONS)

8 MODELING THE DISPERSION OF OXIDES OF NITROGEN 8.1 GENERAL 8.2 ASSESSING NOx LEVELS 8.2.1 Approach 1 8.2.2 Approach 2 8.2.3 Approach 3 8.2.4 Approach 4 8.2.5 Suggested Method 8.3 EXAMPLE: DISPERSION OF NOx FROM A BOILER HOUSE FIGURE 8.1 SAMPLE NO2 NOx RATIO CALCULATION 9 THE COMPLEX TERRAIN MODULE IN ADMS 9.1 WHAT IS THE COMPLEX TERRAIN MODULE? 9.2 HOW DOES THE COMPLEX TERRAIN MODULE OF ADMS WORK? 9.2.1 Wind Flow 9.2.2 Dispersion Calculations 9.3 WHEN AND HOW SHOULD THE COMPLEX TERRAIN MODULE BE USED?



9.4 WHAT IS THE EFFECT OF USING THE COMPLEX TERRAIN MODULE? 9.4.1 Conclusions - Terrain Elevations 9.4.2 Conclusions - Variations in Surface Roughness 9.4.3 Conclusions - Buoyant Releases TABLES 9.1 COMPARISON OF REPRESENTATIVE CONCENTRATIONS FOR RELEASES UPWIND OF HILL 9.2 COMPARISON OF REPRESENTATIVE CONCENTRATIONS FOR RELEASES DOWNWIND OF HILL FIGURES 9.1 WIND FLOW AROUND A HILL (SIDE VIEW) 9.2 WIND FLOW AROUND A HILL UNDER STABLE ATMOSPHERIC CONDITIONS (PLAN VIEW) 9.3 TOPOGRAPHY OF THE RUNCORN AREA 9.4 MODELED CONCENTRATIONS DUE TO EMISSIONS FROM 60 m STACKS UPWIND OF HILL B STABILITY / 2 m/s 9.5 MODELED CONCENTRATIONS DUE TO EMISSIONS FROM 60 m STACKS DOWNWIND OF HILL B STABILITY / 2 m/s 9.6 MODELED CONCENTRATIONS DUE TO EMISSIONS FROM 60 m STACKS UPWIND OF HILL D STABILITY / 5 m/s 9.7 MODELED CONCENTRATIONS DUE TO EMISSIONS FROM 60 m STACKS DOWNWIND OF HILL D STABILITY / 5 m/s 9.8 MODELED CONCENTRATIONS DUE TO EMISSIONS FROM 60 m STACKS UPWIND OF HILL F STABILITY / 2 m/s 9.9 MODELED CONCENTRATIONS DUE TO EMISSIONS FROM 60 m STACKS DOWNWIND OF HILL F STABILITY / 2 m/s



10 THE DEPOSITION MODULE OF ADMS - A BRIEF GUIDE 10.1 INTRODUCTION 10.2 DEPOSITION MODELING METHODOLOGY USED IN ADMS 10.3 DEFAULT AND RECOMMENDED INPUTS USED IN ADMS 10.3.1 Wet Deposition 10.3.2 Dry Deposition 10.4 RECOMMENDATIONS FOR USING DEPOSITION MODULE 10.5 EXAMPLE APPLICATION OF THE DEPOSITION MODULE TABLES 10.1 WET DEPOSITION COEFFICIENTS 10.2 DRY DEPOSITION VELOCITIES FOR GASEOUS COMPOUNDS 10.3 DISTANCES AT WHICH DEPOSITION PROCESSES HAVE A SIGNIFICANT EFFECT ON AIR CONCENTRATIONS FIGURES 10.1 PARTICULATE DRY DEPOSITION VELOCITIES AS A FUNCTION OF PARTICLE DIAMETER 11 EXAMPLE GAS DISPERSION CALCULATIONS FOR ENVIRONMENTAL APPLICATIONS USING ADMS

11.1 INTRODUCTION 11.2 SOURCE DATA 11.3 EXAMPLE CALCULATIONS

11.3.1 EXAMPLE ONE - CONTINUOUS EMISSIONS 11.3.2 EXAMPLE TWO - MULTIPLE STACK CALCULATION 11.3.3 EXAMPLE THREE - ODOR DISPERSION CALCULATION 11.3.4 EXAMPLE FOUR - DISPERSION AROUND A BUILDING 11.3.5 EXAMPLE FIVE - ANNUAL AVERAGE STATISTICAL CALCULATION FOR AN AREA SOURCE 11.3.6 EXAMPLE SIX - DISPERSION OF PARTICULATES FROM A PRILLING TOWER

11.4 ACCURACY OF ADMS-2



11.5 CHOICE OF WIND AND WEATHER CONDITIONS FOR DESIGN 11.6 RUN TIMES

11.6.1 GENERAL 11.6.2 RUNNING BATCH FILES

11.7 WIND AND WEATHER DATA 11.8 SUMMARY OF ROUGHNESS LENGTHS (Z O) 11.9 CALCULATION TRENDS FIGURES 11.1 OUTPUT FROM X-Y PLOTTING OPTION 11.2 THE DISPERSION OF SULFUR DIOXIDE FROM A 40 M STACK 11.3 SAMPLE ADMS LINE PLOT : PLUME HEIGHT (M) 11.4 SAMPLE ADMS LINE PLOT : MAXIMUM CONCENTRATION IN PLUME 11.5 THE DISPERSION OF THE OXIDES OF NITROGEN FROM A PLASTICS WORKS 11.6 THE DISPERSION OF THE OXIDES OF NITROGEN FROM A PLASTICS WORKS 11.7 THE DISPERSION OF THE OXIDES OF NITROGEN FROM A PLASTICS WORKS 11.8 THE DISPERSION OF THE OXIDES OF NITROGEN FROM A PLASTICS WORKS 11.9 ANNUAL AVERAGE CONCENTRATION OF THE OXIDES OF NITROGEN - BOTH STACKS AT 40 m 11.10 MEAN GROUND-LEVEL CONCENTRATION - EXAMPLE THREE 11.11 THE DISPERSION OF ETHYL ACRYLATE FROM A 15 m HIGH STACK - 98th PERCENTILE OF CONCENTRATION FLUCTUATIONS - 5 m/s NEUTRAL ATMOSPHERIC STABILITY



11.12 THE DISPERSION OF ETHYL ACRYLATE FROM A 15 m HIGH STACK - 2 m/s UNSTABLE ATMOSPHERIC CONDITIONS - 98th PERCENTILE OF SHORT TERM CONCENTRATIONS 11.13 THE DISPERSION OF ETHYL ACRYLATE FROM A 50 m HIGH STACK - 2 m/s UNSTABLE ATMOSPHERIC CONDITIONS - 98th PERCENTILE OF CONCENTRATION FLUCTUATIONS 11.14 EFFECT OF WIND DIRECTION ON CONCENTRATION - EXAMPLE FOUR 11.15 DISPERSION OF SO2 FROM A SULFURIC ACID RECOVERY PLANT - EXAMPLE FOUR 11.16 DISPERSION OF SO2 FROM A SULFURIC ACID RECOVERY PLANT - 30 m STACK - EXAMPLE FOUR 11.17 ANNUAL AVERAGE BENZENE CONCENTRATIONS FROM A SMALL LAGOON 11.18 THE DISPERSION OF PARTICULATES FROM A PRILLING TOWER - EXAMPLE SIX 11.19 TOTAL ANNUAL DEPOSITION RATE FROM THE PRILLING TOWER (µg/m2s) 11.20 MAXIMUM 24 HOUR MEAN PARTICULATE CONCENTRATION FROM A PRILLING TOWER

12 DISPERSION MODELING OF ODOROUS RELEASES

12.1 ODOR EMISSIONS - CHARACTERIZATION AND MEASUREMENT 12.2 AVERAGING TIMES

12.2.1 Concentration Fluctuations 12.2.2 Change in Mean Wind Direction 12.2.3 Accounting for Dependence on Averaging Time

12.3 ODOR THRESHOLDS 12.4 ODOR DISPERSION MODELING 12.5 EXAMPLE ODOR DISPERSION MODELING STUDY



FIGURES 12.1 INSTANTANEOUS AND AVERAGED PLUME DISPERSION 12.2 ACTUAL CONCENTRATIONS OF RELEASED MATERIAL 12.3 ACTUAL AND MEASURED CONCENTRATIONS OF RELEASED MATERIAL 12.4 STATISTICAL DESCRIPTIONS OF MEASURED CONCENTRATIONS 12.5 WIND DIRECTION ENVELOPES FOR SHORT AND LONG-TERM MEANS 12.6 EXAMPLE STUDY : SITE DIAGRAM TABLES 12.1 APPROPRIATE AVERAGING TIMES 12.2 EXAMPLE STUDY: PLANT ODOROUS RELEASES 12.3 EXAMPLE STUDY: MODELED CONCENTRATIONS

13 BIBLIOGRAPHY 14 GLOSSARY

APPENDICES APPENDIX A WIND GENERATION OF PARTICULATES APPENDIX B TABLE OF PROPERTY VALUES FOR SPECIFIC CHEMICALS DOCUMENTS REFERRED TO IN THIS PROCESS SAFETY GUIDE



0 INTRODUCTION 0.1 AIMS This Process Safety Guide has been written with the aim of assisting process engineers, hazard analysts and environmental advisers in carrying out gas dispersion calculations. The Guide aims to provide assistance by:

• Improving awareness of the range of dispersion models available within GBHE, and providing guidance in choosing the most appropriate model for a particular application.

• Providing guidance to ensure that source terms and other model inputs

are correctly specified, and the models are used within their range of applicability.

• Providing guidance to deal with particular topics in gas dispersion such as

dense gas dispersion, complex terrain, and modeling the chemistry of oxides of nitrogen.

• Providing general background on air quality and dispersion modeling

issues such as meteorology and air quality standards.

• Identifying personnel within GBHE's Alliance Network with expertise and experience of dispersion modeling.

• Providing example calculations for real practical problems.

0.2 SCOPE The gas dispersion guide contains the following Parts: 1 Fundamentals of meteorology. 2 Overview of air quality standards. 3 Comparison between different air quality models. 4 Designing a stack. 5 Dense gas dispersion.



6 Calculation of source terms. 7 Building wake effects. 8 Overview of the chemistry of the oxides of nitrogen. 9 Overview of the ADMS complex terrain module. 10 Overview of the ADMS deposition module. 11 ADMS examples. 12 Modeling odorous releases. 13 Bibliography of useful gas dispersion books and reports. 14 Glossary of gas dispersion modeling terms. Appendix A : Modeling Wind Generation of Particulates. APPENDIX B TABLE OF PROPERTY VALUES FOR SPECIFIC CHEMICALS The two models referred to by name are the currently preferred models for dense gas dispersion (PHAST) and neutral/buoyant gas dispersion (ADMS).



1 METEOROLOGICAL PARAMETERS WHICH AFFECT DISPERSION 1.1 ROUGHNESS LENGTHS (Zo) The roughness length is a parameter which quantifies the effect ground roughness has on the turbulent flow properties of the wind - the higher the roughness length, the more turbulent the wind flow. For an elevated stack, the higher the roughness length, the more rapidly the plume centerline concentration decreases with distance. However, the higher the roughness length, the more rapidly the plume spreads in the vertical direction, counteracting the effect of roughness on plume centerline concentrations. Hence it is not possible to generalize the effect surface roughness has on ground level concentrations. For a ground level release of a heavier-than-air gas cloud, the higher the surface roughness, the more rapid is the dispersal rate of the cloud Estimating roughness lengths can be difficult - rarely is the terrain uniform around a source - in general, consider the roughness of the ground upwind of the source. Typical values are as given in Table 1.1 below:- TABLE 1.1 TYPICAL ROUGHNESS LENGTHS



GBHE suggests that if in doubt one should choose a roughness length of h/30 where h is the average height of the obstacles e.g. if the typical size of the roughness elements is 9-10 m, use Zo = 0.3 m. This is only a simple rule of thumb. Commercial programs, can only accept roughness length inputs of 0.01, 0.1 and 1 m - use a roughness length of 0.1 m for an industrial site. 1.2 WIND SPEEDS Wind speed varies as a function of height and ground roughness. In general, whenever a wind speed is quoted, it refers to the speed at a height of 10 m, although sometimes data from the US or from a small local weather station may be measured at a height of 2 m. The velocity profile as a function of height is dependent on atmospheric stability. Models such as the US-EPA models, commercial programs assume a power law velocity profile:-

where n is a function of roughness and atmospheric stability; z is the height above the ground (m), and uz is the velocity at height z m. More widely used is a log-law relationship based originally upon Prandtl mixing length theory for the turbulent boundary layer over a flat surface:-

where L mo is the Monin-Obukhov length. Y is a function that takes into account the effect of atmospheric stability - usually found empirically. u* is a term known as the friction velocity defined as √(τ/ρa), where τ is the surface shear stress and ρa is the air density. k is the von Karman constant, which has a value close to 0.4. The effect of the variation in wind speed as a function of height does have a significant effect on gas dispersion modeling. For example, the advection velocity of a dense gas box-type model is usually taken to be the wind speed at half the height of the gas cloud. As more air is entrained into the cloud, its height increases and hence the bulk velocity of the cloud increases.



1.3 ATMOSPHERIC STABILITY 1.3.1 General The term atmospheric stability describes the degree of stratification of the atmosphere, which plays a vital part in the dispersion of atmospheric pollutants. On hot sunny days with cloudless skies, the ground absorbs radiation from the sun at a faster rate than the air above it. The ground then re-radiates and convects heat back into the atmospheric boundary layer setting up large scale convective motions. These cause rapid plume spreading in the vertical direction and large scale plume meandering. This rapid spreading brings elevated plumes down to ground level. For elevated stacks, the highest ground level concentrations occur in low wind speed, unstable atmospheric conditions. During cold winter evenings and nights with little or no cloud cover, the ground is at a lower temperature than the air above it and heat is transferred from the air to the ground. This sets up a stratified layer of colder air close to the ground which dampens out atmospheric turbulence. Gaseous effluent from elevated stacks form narrow pencil-shaped plumes which rarely strike the ground. Hence, stable conditions, in general, give low ground level concentrations from elevated stacks. However, stable conditions would give the worst case conditions if the plume directly impacted an adjacent plant structure or hill nearby. Low wind speed, stable atmospheric conditions always give the worst case scenario for catastrophic releases of a heavier than air gas cloud and for any ground level release. In practice, for at least 60% of the time in the USA, there is neutral atmospheric stability where the effect of heat transfer from the ground into the plume is negligible. In this case, mechanical turbulence generated by the wind flow in addition to turbulence generated by the initial momentum of the plume, control the dispersion rate. Neutral conditions usually prevail when the wind speed exceeds 5 m/s. The effect of atmospheric stability on plume dispersion is illustrated in Figure 1.1.



There are two commonly applied ways of characterizing atmospheric stability:- Pasquill -Gifford stability scheme and Monin-Obukhov length scaling. The former methodology is used by many models, including DISP2 and PHAST but there is increasing tendency for the latest dispersion models such as UK-ADMS, to adopt the latter approach. 1.3.2 Pasquill - Gifford Methods of Characterizing Atmospheric Stability Pasquill - Gifford stability analyses assign a letter in the range A to G in order to characterize atmospheric stability. The most unstable atmospheric conditions, characteristic in the USA of a few really hot summer afternoons, are represented by the letter A; neutral conditions by the letter D and stable conditions by F. A few modelers in Northern Latitudes use G conditions to represent really stable conditions (e.g. winter evenings in Norway). The actual choice of stability category is governed by wind speed and cloud cover and is defined in Tables 1.2 and 1.3.



A typical wind speed/direction/Pasquill-Gifford atmospheric stability analysis is shown in Table 1.4.



TABLE 1.4 EXAMPLE PASQUILL-GIFFORD STABILITY ANALYSIS



Typically, the atmospheric stability categories in the USA occur with the following probabilities:

A-stability < 1% B-stability 1-2% C-stability around 10% D-stability 50-70% E-stability 10-20 % F-stability 5-15% G-stability <2%

In general the further the meteorological station is from the sea, the higher is the frequency of stable and unstable conditions. Also, note that in the categorization of atmospheric stability category in Table 1.2, there is no link between temperature and stability category. In the USA we automatically associate F-stability conditions with cold weather - in fact, the definition of atmospheric stability is linked with cloud cover and incident radiation levels. In the Far East, cloudless skies at night often occur far more frequently than in the USA. This can lead to F-stability frequencies of 30%, even though temperatures do not fall below freezing. 1.3.3 Monin-Obukhov Length Methods of Representing Atmospheric Stability Many gas dispersion models developed since 1990 have adopted Monin-Obukhov length scaling methods. The Monin-Obukhov length (Lmo) is defined by:-

where ρa is the air density (kg/m3); Ta is the air temperature (K); u* is the friction velocity as defined above; k is the von Karman constant (0.4); H is the surface heat flux (W/m2) - the heat flow from the ground into the atmosphere rather than the incident radiative heat flow.



The Monin-Obukhov length is a parameter for the ratio of the mechanical turbulent energy to that produced by buoyancy. It is an extremely awkward parameter to use since in neutral atmospheric conditions (Pasquill - Gifford stability category D), the surface heat flux is zero and hence L mo is infinite. Consequently many models use as an input the reciprocal of the Monin-Obukhov length. Note that the L mo is negative in unstable conditions and positive in stable conditions. For the gas dispersion practitioner, the Monin-Obukhov length is very difficult to measure. To estimate u*, it is necessary to take measurements in order to quantify the velocity profile of the wind flow with height above the ground. Additionally the surface heat flux would have to be measured. In practice, standard values for the Monin-Obukhov length are used. Also, because the friction velocity is dependent on the ground roughness, the Monin-Obukhov length is both a function of roughness length and atmospheric stability category. The following Figure 1.2, derived from Golder (1972) enables a direct comparison to be made between Pasquill Gifford stability category and Monin-Obukhov lengths. Typical values of the reciprocal of the Monin-Obukhov length for a roughness length of 0.1 m are:-

1 m/s A-stability - 0.5 m-1 2 m/s B-stability - 0.075 m-1 (or possibly as high as -0.1 m-1) 5 m/s C-stability - 0.01 m-1 5 m/s D-stability 0.00 m-1 3 m/s E-stability 0.01 m-1 2 m/s F-stability 0.05 m-1



2 AIR QUALITY STANDARDS 2.1 WHAT ARE AIR QUALITY STANDARDS? Air quality standards are limits on concentrations of pollutants in the air. The limits are usually set on the basis of the health effects of the particular pollutants. In some cases, there are also limits designed to protect vegetation (for example, World Health Organization guidelines for ozone). In other cases, where pollutants interact, standards have been set for two pollutants in combination: for example, European Union limits on smoke and sulfur dioxide. The limits are designed to ensure that there would be no significant adverse effects to the most vulnerable in society arising from exposure to the pollutant at levels below the air quality standard. Air quality standards are set by international or national governments. Recommendations are also made by interested bodies, notably the World Health Organization. The standards are used by licensing agencies such as the United States Environmental Protection Agency, or the Environment Agency/Scottish Environmental Protection Agency in the UK. These bodies would use the standards to determine whether pollution levels in their areas are acceptable. This will feed into their readiness or otherwise to license new or existing processes, and may also be used to limit the contribution that each individual process can make to off-site levels of air pollution. Air quality standards apply to environmental levels of pollutants from all sources in combination, rather than to emissions from a single source, or works. Air quality standards need the following components:

• Identification of the pollutant (for example, sulfur dioxide, or "particulate matter which passes through a size selective inlet with a 50% collection efficiency cut-off at 10 microns ( PM10)").

• A numerical concentration (for example, 100 parts per billion by volume (ppb), or 50 micrograms per cubic meter (μgm-3)).

• An averaging time for the numerical concentration (for example, 15-minute mean, or running 24-hour mean).

• An acceptable level of compliance (for example, 99th percentile, or complete compliance) - see Box 1.



Additionally, air quality standards may have other relevant information, such as an indication of their status (for example, legislative limit, or government objective), details of their applicability (for example, appropriate for use in sensitive areas, or particular designated planning zones), and specification of the conditions to which the standards refer to enable conversion between units (for example, 20°C, 760 mmHg pressure) Once all this information is known, it is possible to investigate measured pollution levels to determine whether compliance with a quality standard has been achieved. An example is given in Box 2. When compliance or non-compliance has been established, it is also necessary to consider the status of the standard to determine how significant this result is. For example, could non-compliance result in prosecution for the company, or significant expenditure in the period leading up to the implementation of an objective?



As well as looking at measured pollution levels, it is also possible to consider the results of dispersion models in the light of air quality standards. This would enable a similar assessment to be carried out at locations where measurements have not yet been carried out, for future years at existing plants, or for new plants and developments. This kind of assessment is very useful in obtaining licenses to operate new plant, and in planning the extent of investment that will be necessary to meet forthcoming air quality standards. In the next sections, we will consider the various types of standards that exist; what the standards actually are, and how they should be applied in various situations.



2.2 WHAT AIR QUALITY STANDARDS EXIST? 2.2.1 General Background There are two complementary approaches to regulating air pollution emissions. You can either place limits on emissions (that is, what goes up), or place limits on ambient concentrations (that is, what comes down) - or both. Placing limits on emissions is an attractive approach, because it enables the regulator to ensure that emissions from each source are appropriately limited, and measurement is relatively straightforward. In

principle, this approach avoids the need to work backwards from high ambient levels of air pollution to establish which sources should be controlled.

The disadvantage is that careful specification and enforcement of emissions controls is required to restrict levels of pollutants in air to acceptable levels. The lack of overall controls of air pollution impacts in the UK culminated in the smogs of the 1950s and 1960s, when as many as 4,000 additional deaths were caused by air pollution within a few days. Nowadays, emissions from individual sources of pollutants (including road vehicles) are regulated. However, the lack of overall controls on emissions of oxides of nitrogen and volatile organic compounds (VOCs) particularly from road traffic results in high levels of ozone and photochemical smog in many parts of the world.

2.2.2 United States The United States has specified air quality standards since the introduction of the Clean Air Act in 1970. Recently, revisions have been made to the air quality standards for ozone and fine particulate matter (July 1997). The current standards are given in Box 3. Many other countries adopt the USEPA standards for use where there are no local standards.



2.2.3 European Union The European Union has specified ambient air quality standards for pollutants in a series of directives in the 1980s. These are now implemented into environmental legislation throughout Europe. The European Union standards are given in Box 4.



For some pollutants, the existing standards comprise mandatory "limit values" and discretionary "guide values". The limit values are mandatory standards for application throughout member states, whereas the guide values are intended to contribute to the long-term protection of the environment, particularly in setting up specific environmental improvement projects. These would not generally be directly relevant to all businesses in Europe, although they may influence the policy of regulatory bodies. In some European countries, additional standards have been specified. These include The Netherlands, where standards have additionally been specified for carbon monoxide and benzene. The Dutch standard for benzene is an annual mean concentration of 10 µgm-3. The European Union has recently implemented a directive known as the "Air Quality Framework Directive". This directive lays down a mechanism of establishing a sliding scale of air quality standards. Two levels can be specified for a pollutant, the first for immediate application and the second for application at a specified future date. In the intervening period, the standard is progressively tightened towards the second more stringent level. At the time of writing, proposed standards for sulfur dioxide, nitrogen dioxide, PM10 and lead have been published (see Box 5). The link between levels of sulfur dioxide and particulates (see Box 4) has not been carried through into this new generation of air quality standards.



2.2.4 The Netherlands Ambient air quality standards have been specified in the Netherlands which exceed the current requirements deriving from the EU directives. The relevant standards are given in Box 6.

2.2.5 Japan Air quality standards are specified in the Basic Law for Environmental Control. The standards were set between 1969 and 1978. The standards of relevance to select chemical companies are summarized in Box 7.



2.2.6 Taiwan Air quality standards are based on the Release of Air Quality Standard in Taiwan. The standards of relevance to select chemical companies are summarized in Box 8.

2.2.7 United Kingdom Air Quality Strategy

Recent developments in air quality policy in the UK are highly significant in the development of air quality standards. A government advisory panel known as the Expert Panel on Air Quality Standards (EPAQS) has made recommendations for standards for 8 pollutants, with several more due to be produced by the end of 1998. These recommendations do not have any legal basis, but they have formed the basis of the UK air quality strategy objectives.



The Environment Act 1995 provided for the preparation of a national air quality strategy and guidance on its implementation. In 1996, the Government consulted on the air quality strategy for the UK. In March 1997, this air quality strategy was published in its final version. The EPAQS recommendations were used in this document as objectives to be achieved by 2005 - see Box 9. The UK air quality strategy and the air quality objectives contained in it will be very influential in guiding Environment Agency and Local Authority thinking on air quality.

Regulations implementing the air quality objectives have been made under the Environment Act, and commenced in December 1997, specifying that compliance is to be achieved by 2005. The Environment Act also introduced a program of "Local Air Quality Management" in which local authorities are required to assess their air quality. If it appears that the statutory air quality objectives will not be met by 2005, then a local air quality management plan should be devised and implemented to ensure that the objectives will be met. This may include additional controls on industrial emissions and traffic pollution, although the plan should ensure that the burdens on various sectors are "proportionate".

2.2.8 Non-Governmental Organizations The World Health Organization published an influential set of air quality guidelines in 1987 ("Air Quality Guidelines for Europe", WHO European Office, Copenhagen). These were, in general, relatively stringent guideline values for levels of air pollutants, and included guidelines for pollutants not covered in legislation. Guidelines were specified to protect not only human health, but also components of the natural environment. These guidelines have been used by select chemical companies as objectives for ambient air quality for pollutants which do not have air quality standards (for example, vinyl chloride and toluene). The guidelines are also used by some countries in place of specific local air quality standards (for example, Pakistan). The guidelines are due to be updated during 1998, but the process has currently stalled due to financial difficulties within the WHO. The draft guidelines are given in Box 10.



The World Bank has also specified air quality standards for use in assessing projects which it funds - see Box 11. These have been adopted for use in some countries including Pakistan



2.2.9 Occupational Exposure Limits Occupational exposure limits exist for a very wide range of pollutants. These are specified to protect the health of employees in the workplace. Ambient air quality guideline values are frequently derived from these occupational exposure limits for pollutants which do not have any specific air quality standards or WHO guidelines. These occupational limits themselves should not be used directly for ambient air, as they are appropriate to fit and healthy adults (making no allowance for sensitive members of the population such as children and those suffering from respiratory disease), and they are specified on the basis that exposure takes place during working hours only. With suitable adjustments to allow for these constraints, however, ambient air quality guideline values can be derived from the occupational exposure limits. This is achieved by dividing the occupational exposure limit by a specified factor to give the ambient air quality standard. A range of factors have been used for this purpose in the past, ranging from one twenty-fifth to one hundredth. In the UK, the Environment Agency has issued guidance on how this conversion should be addressed in a recent publication (Technical Guidance Note (Environmental) E1, "Best Practicable Environmental Option Assessments for Integrated Pollution Control", 1996). It indicates that "environmental assessment levels" for pollutants can be determined as follows:

• Hourly mean concentration: 2% of the 15-minute maximum exposure limit (MEL: these are occupational exposure limits for carcinogens) or 10% of the 15- minute occupational exposure standard for materials where no MEL has been specified (i.e., non-carcinogens).

• Annual mean concentration: 0.2% of the 8-hour maximum exposure limit or 1% of the 8-hour occupational exposure standard. The guidance note indicates that individual processes should be a "priority for control" if they contribute more than 2% of the environmental assessment level for a given pollutant.



This is a highly restrictive constraint, and in practice, a process that contributes less than 10% of the environmental assessment level will generally be considered acceptable. Further guidance on acceptable off-site concentrations is given in Box 3 of Part 4.

2.2.10 General Comparison In general, newer air quality standards tend to be more stringent than older standards, and guidelines tend to be more stringent than regulatory limits. The least stringent standards are standards specified during the 1970s and 1980s such as the US National Ambient Air Quality Standards, and the existing set of EU standards. Standards based on occupational health guidelines also tend to be relatively lax. For example, the hourly average environmental assessment level for use in the UK based upon 10% of the occupational health standard for nitrogen dioxide would be 500 ppb. In contrast, the UK Air Quality Strategy objective for hourly mean nitrogen dioxide concentrations is 150 ppb. Newer standards such as the UK Air Quality Strategy objectives, the EU daughter directive proposals and the US standards for ozone and PM2.5 are more stringent than the existing legislative standards, and cover a wider range of pollutants. The UK Air Quality Strategy objectives are similar to the WHO guidelines of 1987 in most respects, although for particulate matter, new information has led to a significantly tighter objective. The WHO guidelines also cover a wider range of pollutants. It may be expected that the revised WHO guidelines to be issued during 1998 will be more stringent than the 1987 document.

2.2.11 Air Quality Standards for Odor Impacts Odor impacts are likely to become an increasingly important driver of limits on air pollution emissions. In many countries, process operators are required to ensure that there is no off-site odor. Odor impacts can be forecast, or estimated from process emissions data, but the procedure is very uncertain, and because a large number of safeguards must be built in, the assessments are of necessity very stringent in terms of acceptable release conditions.



A reasonable standard for off-site Odor would be that hourly average concentrations of an odorous chemical should not exceed 2.5% - 5% of the Odor threshold. Odor thresholds are discussed in Box 12. This standard would provide a sufficient safety margin to protect against the uncertainty in Odor threshold measurements, short-term fluctuations in concentration that can give rise to transient Odor, and the variability in human response to different Odors. The Odor standard is likely to be much more stringent than the corresponding health-based guidelines, reflecting the fact that Odor is generally significant at lower concentrations than health effects, and also reflecting the additional safety margin in the Odor standard. It should be noted that for a few chemicals such as ethylene dichloride, the health impacts occur at concentrations below the Odor threshold.

2.3 WHAT IS THE LAW - AND WHAT ISN’T A clear distinction should be made between air quality standards which comprise legal limits in particular countries, and other recommendations and guidelines which are not limits. In practice, air quality standards are frequently exceeded in many parts of the world - particular problems surround standards for ozone and fine particulate matter. This does not translate into legal action against emitters of pollution. Process operators would be affected by air quality standards under the following circumstances: • A new process is highly unlikely to be permitted if emissions will lead to a contravention of an air quality standard.



• In some countries (for example, China), different air quality standards apply in different planning zones. Thus, the locations where chemical industries may be located could be restricted by the more stringent standards in some areas. • Continued operation of a process may be at risk in an area where air quality standards are frequently exceeded. Under these circumstances, the process operator may be required to reduce emissions to enable the air quality standards to be met. Legislative air quality standards currently applicable in various parts of the world are as follows: USA: National Ambient Air Quality Standards (see Box 3) European Union: Directives 80/779, 82/884, 85/203 as enacted in individual Member States (see Box 4) Netherlands: Legal and non-legal air quality standards (see Box 6) Japan: Basic Law for Environmental Control (see Box 7) Taiwan: Release of Air Quality Standard in Taiwan (see Box 8) Air quality standards are progressively tightening. The EU is due to propose a range of new and progressively tightening standards for air quality over the coming year. New limits for nitrogen dioxide, smoke, particulate matter and lead have been specified (see Box 5). These will be made under the "Framework Directive", and will eventually have legal force. In the period between the standards being adopted by the EU and their implementation in individual member states, they should be treated as if they were legal limits. For design of new plant in the EU at any time, the new limits should also be treated as if they had legal force to ensure that plant design is adequate. In the UK, the new standards are unlikely to lead to a significant additional burden on industry, over and above the burden imposed by the new UK air quality objectives. There are now objectives for air quality in the UK, specified as part of the UK air quality strategy. These objectives are shown in Box 9. The objectives will be reviewed during 1998, and may be tightened, and/or brought into line with any new European air quality standards.



The objectives will have legal force, but the onus will be on local authorities to implement air quality management plans to achieve compliance by 2005 rather than on individual process operators. Thus, local authorities and/or the Environment Agency are likely to require action to be taken in any areas where it is likely that the objectives will not be achieved. These objectives may impact select chemical companies as the regulatory bodies assess their requirements for reductions in the impact of industrial air pollution to meet the air quality objectives by 2005. Select chemical companies operating in the UK may need to be prepared to undertake independent assessments of the impact of their air pollution emissions in order to ensure that any additional regulatory burden is appropriate and proportionate (see Box 13 for an example). Apart from these legislative and proposed air quality standards and objectives, a number of other guidelines for ambient air quality may be used. These do not have legal force. They would be used where businesses are releasing compounds for which there are no other air quality standards. This covers a wide range of Select chemical companies process emissions, whereas combustion emissions would generally be covered by the air quality standards and objectives. The World Health Organization standards and the application of occupational health standards in ambient air quality assessments is described in Section 2.2 above.

National Air Quality Standards do not apply in plant areas to which the public cannot gain unrestricted access. In these areas, Occupational Exposure Standards (OESs) and Maximum Exposure Limits (MELs) are appropriate.



Materials with a Maximum Exposure Limit present serious concerns about possible health effects in workers. In practice, MELs have most often been allocated to chemicals for which there is no clearly defined safe concentration level and for which there is no doubt about the seriousness of the hazard posed by the substance. Usually MELs are defined for chemicals which are carcinogenic or can cause occupational asthma. OESs are set at levels below which it is believed (based on current scientific knowledge) that the substance would not damage the health of workers exposed to it day after day. For listings of OESs and MELs, see either the UK Health and Safety Executive’s EH 40 document - “Occupational Exposure Limits”, which is published annually. The HSE provides the following guidance on how to apply OESs and MELs: “Applying OESs:- if exposure to a substance that has an OES is reduced at least to that level, then adequate control has been achieved. If this level is exceeded, the reason must be identified and measures to reduce exposure to the OES put into action as soon as reasonably practicable. Applying MELs:- Exposure should be reduced as far below the MEL as reasonably practicable and should never exceed the MEL when averaged over the appropriate reference period.”

2.4 FUTURE DEVELOPMENTS In general terms, the most significant future development is the progressive tightening of air quality standards around the world. One example is the recent introduction of a tighter ambient air quality standard for ozone, and a new standard for PM2.5 in the USA. New air quality standards have been drafted by the European Commission (see Box 5). A further standard for ozone is expected to be published by the end of 1998, with proposals for polycyclic aromatic hydrocarbons and some heavy metals to follow. These represent a considerable tightening of standards in comparison to current air quality standards in Europe.



The planned revisions to the World Health Organization air quality guidelines are unlikely to have as profound an impact as the original 1987 guidelines. Many of the considerations adopted by the WHO have been taken on board by bodies such as the US EPA and European Union in setting air quality standards. A significant future development in the UK will be the implementation of the air quality strategy up to 2005. This may well lead to additional constraints on industrial emissions in some areas where the UK air quality objectives would not otherwise be met. These constraints may be implemented through limits on emissions agreed between individual process operators and the regulator (Environment Agency and/or Local Authority). The current set of air quality objectives (see Box 9) are under revision, with revised targets and/or dates to be published during 1998. Again in the UK, the implementation of Technical Guidance Note E1 may lead to tighter restrictions on emissions of pollutants not covered by the air quality strategy. This is because of restrictions on the contribution of individual processes to ambient levels of air pollutants. The guidance indicates that those pollutants contributing more than 2% of the Environmental Assessment Level off-site will become "a priority for control". It will not be possible to apply this process in practice because of the large number of industrial processes which will become "priorities for control". A value of 10% of the EAL is generally considered to be acceptable. However, the Guidance Note does indicate a significant shift in Environment Agency policy. Odor issues are likely to become an increasing driver for restrictions on emissions. This reflects some success in dealing with emissions of the health effects of pollutants, and also sustained public awareness and concern regarding air pollution. There is very little formal guidance on the assessment of odor emissions, but it is likely that plants which have known odor problems are likely to come under increasing pressure to control the emissions. If this cannot be achieved via process improvements, investment in end-of-pipe control equipment may be required. Finally, aesthetic effects may well become more important. Already, local authorities are often unwilling to allow new tall stacks to be constructed because of their visual impact. In the next few years it is likely that industry will be under pressure to reduce the visual impact of large plumes of water vapor from vents.



3 MODEL COMPARISON AND SELECTION 3.1 CLASSIFICATION OF DISPERSION MODELING PROBLEMS

Dispersion modeling problems are commonly categorized as “safety” or “environmental.” - see Figure 3.1. “Safety” issues involve the assessment of the consequences of unplanned releases which may present a significant direct hazard to the health of individuals located either on or off-site. Because the majority of chemicals used by chemical businesses are heavier than air, these are usually dense gas releases. Storage at low temperature also tends to result in releases of gases which are denser than air. These are seen as safety issues because the effects are potentially serious, and the release will only take place over a short period. In contrast, “environmental” issues generally arise from continuous or intermittent releases of material of similar density to air (“neutral”), or lighter than air (“buoyant”). Occasionally, continuous releases may be more dense than the air. These are generally planned releases of material arising from normal process operations. Any effects of these releases tend to be most significant off-site. As well as short-term toxicity effects, the assessment of environmental releases also takes into account the effects of long-term exposure to released materials. In some cases, consideration is given to effects on the natural environment, as well as on the human population. For the purposes of dispersion modeling, there is some overlap between the two categories of problem.



FIGURE 3.1 CATEGORIZATION OF DISPERSION MODELING PROBLEMS

The range of potential release scenarios means that a large number of dispersion modeling tools have been designed to assess their consequences. The aim of this Part of the guide is to provide guidance on selecting the appropriate tool for a particular problem. The appropriate model(s) to use for a particular application is dependent on the initial density, duration and location of the release.



3.2 WHAT MODELS ARE AVAILABLE?

To deal with the situations for which dispersion modeling is required, a range of modeling tools have been developed. For the purposes of this guide, a “model” is defined as a computational code which provides an airborne concentration of material given a set of release conditions, a set of meteorological conditions, and a location relative to the source. These have been developed to varying specifications over and above the minimum model definition. For example, some models contain algorithms for calculating loss rates of material, given some assumptions regarding the quantity of material, the size and location of a leak, etc. Some models permit highly flexible specification of the locations at which concentrations are to be calculated, or permit the use of long-term meteorological data to calculate long-term mean concentrations of material. A number of commonly-used models are listed in Table 3.1, together with an indication of the type of situations in which they can be applied, and their functionality.

TABLE 3.1 COMMONLY USED DISPERSION MODELS



3.3 DESCRIPTION OF AVAILABLE MODELS 3.3.1 General

As indicated in Table 3.1, the dispersion models listed in the table all have advantages and disadvantages associated with their use. The aim of this section is to set out the pros and cons of each model, and to provide some practical guidance in using each model. Finally, Table 3.2 summarizes the type of problem for which each model should be used. The use of dispersion models is regulated to varying extents in different countries of the world. In some countries, a specific model needs to be used in a specific way; in other countries, the applicant is free to use any appropriate model. Some examples are as follows:

• Germany: Dispersion modeling to be carried out as laid down in TA Luft regulations. These specify the dispersion equations to be used, and appropriate values for many of the inputs.

• UK: ADMS is preferred by the Environment Agency for regulatory

applications, but no formal guidance exists.

• Netherlands: EFFECTS is the preferred model for dense gas releases, and PLUIM for buoyant/neutral releases.

• USA: A variety of different models are approved by the US EPA for

various situations, as laid down in Appendix W to the 40th Congressional Federal Register part 51. For modeling point source emissions in non-complex terrain, ISC is recommended (section 4.1 of Appendix W; see the USEPA web site for further details: www.epa.gov). For dense gas dispersion modeling, any appropriate model is permitted.

Attempts are currently under way to harmonize the approach to dispersion modeling across national boundaries, but many countries (e.g. Hungary) insist on the use of a national dispersion model.



3.3.2 ADMS (Atmospheric Dispersion Modeling System) This model is produced and developed by Cambridge Environmental Research Consultants on behalf of the Environment Agency, Health and Safety Executive, and a consortium of industry and government bodies including select chemical companies. The model flexibility is enhanced with a number of additional modules for dealing with specific cases, as listed in Table 3.1. The model is designed for neutral density and buoyant releases. ADMS can also be used for releases of dense gases from elevated sources providing the plume does not slump to ground level. This can be a matter of judgment, but some indication can be gained from consideration of the plume centerline height, and/or by considering near-source results from a dense gas dispersion model such as PHAST. The model is straightforward to use, with a series of screens providing rapid data entry. The model is supplied with a range of example source and meteorological data files, which can be used as a basis for compiling inputs for other applications. The file "r91a-g.met" is particularly useful, as it provides a set of 7 meteorological conditions representative of the range of conditions encountered in temperate regions. ADMS has a straightforward x-y plotting program, and can provide contour plots via a link to the SURFER package. The program can be linked to a GIS system if required, to facilitate data input and results presentation. The program uses state-of-the-art understanding of meteorology to represent the atmospheric boundary layer. Output is provided in a set of separate ASCII text files, which can be imported into other applications if required. Percentile concentrations can be obtained provided the appropriate meteorological data is used: this is useful for obtaining predictions in terms of air quality standards and objectives. ADMS is the preferred model for regulatory applications in the UK. In view of its technical merits and the wide range of problems it can deal with, it is also recommended for use outside the UK in situations where no other model is specified by the regulatory authority. ADMS only permits modeling to be carried out for a limited number of receptors (maximum grid size: 31 x 31 x 2 receptors). This may be a restriction for some applications. Model run times can be very long when long-term meteorological data is being used, particularly where building effects or complex terrain are incorporated into the model.



3.3.3 ALOHA (Areal Locations of Hazardous Atmospheres)

This model is a user-friendly version of the US Coastguard/University of Arkansas model DEGADIS. It is extremely user-friendly, and enables even a novice user to set up appropriate meteorological and source inputs rapidly. The tank source options are particularly user-friendly with graphical images to assist the source specification. It is probably best used as an emergency response tool, with more complex planning cases being handled by a more flexible model such as PHAST. It may also be appropriate for use in Risk Management Planning in the US. The model can provide indoor concentrations of pollutants, based on certain assumptions relating to air exchanges in the building. The model can handle a variety of source types including mixed aerosol/vapor releases arising due to a tank rupture, and liquid puddles. Because ALOHA is set up to model releases from a relatively simple set of cases in an emergency situation, more complex cases cannot easily be modeled. The major disadvantage of ALOHA for planning purposes is that receptors must be specified individually, and the model re-run for every receptor. The model only allows for a one-hour run time, and so concentrations are not predicted at locations where the maximum concentration from a release would not have been reached one hour after the release. The model has also been found to reset parameters without warning (for example, changing units from mgm-3 to ppm), and frequently clears the values of modeling parameters which have already been entered. This can occur for example when attempting to edit the source details if the wrong type of source is selected in error.

3.3.4 DISP2

DISP2 was developed by a European chemical company, and has two components. The BURST model gives concentrations arising from a short emission period, and the PLUME model gives concentrations arising from a continuous emission. It has been shown to be robust in handling a wide range of cases for both environmental and safety applications. However, it cannot handle two phase releases. The model is not used outside of the European chemical company that developed it, and there may be problems in justifying its use to regulatory authorities. It should, in general, not be used for new applications as there are externally validated models available which can cover most of the situations for which DISP2 was designed.



DISP2 is straightforward to operate. Study inputs are entered via a series of screens in a logical sequence. The "mass fraction" option for burst releases (i.e., instantaneous catastrophic releases) of mixtures should not be used, as there is an error in the software associated with this option. The model is limited to three values for surface roughness (0.01 m, 0.1 m and 1 m). This can be a restriction in the common situation of a release in a typical urban/industrial area where a value of around 0.3 - 0.5 m would be appropriate. Using a surface roughness of 0.1 m could underestimate the influence of the terrain on dispersing emissions from a ground level release. Data for each run is saved in a file named "c:\windows\temp\analysis.lis": the data in this file is overwritten each time the model is run, and must be extracted between runs if required. More information is given in the model problem file (*.prb) A maximum of 21 receptors downwind of the source is permitted. A single source, wind direction and downwind line of receptors is considered in each model run, although the model does provide off-axis concentration isopleths if required. The meteorological data entry can be unclear: the available conditions are specified using codes such as "B2." The default setting is for the number 2 in this context to refer to Force 2 on the Beaufort scale, rather than 2 ms-1. Care needs to be taken when considering dense gas releases to ensure that two-phase effects are not significant. STACK2 is a multi-source neutral/buoyant release dispersion model. It should not be used for new applications as both ADMS and ISC can be used to carry out all the calculations that are possible with STACK2.

3.3.5 ISC (Industrial Source Complex)

ISC is available as ISCST (Short Term) and ISCLT (Long term) and was developed by the US-EPA. This model is very widely used throughout the world for environmental modeling applications. It is prescribed for regulatory use in the USA. The core of the model is now some 20 years old, and it contains some major shortcomings - for example, the inability to specify the terrain surface roughness (see below). The model is due to be replaced by a new version currently known as “Aermod” within the next 12-18 months. The use of ISC is prescribed for regulatory calculations in the US and some other countries. Where the use of ISC is not prescribed, it would generally be preferable to use ADMS in view of the more advanced modeling methods and greater flexibility of ADMS.



ISC is not straightforward to use. In principal, a logically ordered input (.dat) file is written, and the model produces an output .lst file. However, the user interface requires a number of apparently unrelated inputs to compile an appropriate .dat file, and it can be difficult to provide a combination of inputs that is acceptable for the model. For example, a parameter "RUNORNOT" must be set to "RUN" for the model to proceed. Some processing of the output .lst file is necessary to produce contours which can be incorporated into a graphical plotting or numerical analysis package. This can be time consuming and is a potential source of error. The model is very flexible in terms of the number and type of sources that can be included. Also, concentrations can be modeled at a very large number of receptors. The model runs relatively quickly, which is useful for producing long-term statistics based on measured meteorological data. A serious disadvantage is that the effects of surface roughness can only be incorporated by running the model in "urban" or "rural" modes, and it is not clear what values of Zo these modes correspond to. As a rough guide, "rural" is likely to correspond to Zo ~ 0.1 m, and "urban" is likely to correspond to Zo ~ 1.0 m. Some example meteorological data is provided with the model; however, it is not straightforward to provide data in the correct format for the model to use. User friendly versions of the US-EPA models are produced by various Consultants in the USA. File driven versions of models, available free of charge from the US-EPA Bulletin Board, are very difficult to use.

3.3.6 PHAST (Process Hazard Assessment Tools)

This model has been developed by DNV Technica. It is probably the most sophisticated general purpose hazard assessment software package currently available - for example, it covers high momentum jet releases at a range of angles, catastrophic dense gas releases, pool evaporation, two phase releases, fires and explosions. The model has an extensive physical properties database. The two main drawbacks are firstly its cost - with an annual maintenance fee of around; secondly, there are a number of bugs in the program. Some of these are inconvenient, but others could give rise to serious errors in executing the model.



One bug relates to the model treatment of the dispersion of non-dense gas releases: hence, PHAST should not be used to model releases that could be modeled with a package such as ADMS or ISC. If it is necessary to model non-dense gas releases with PHAST, the “Density Tolerance for Cloud Buoyancy” should be changed from its default value of 0.005 to a value of 10-6. PHAST is not straightforward to use, essentially because of the wide variety of problems that it sets out to solve. Setting up a mixture of chemicals from the materials database is difficult, and the materials database itself is not compatible with some computer operating systems. It should be noted that surface roughness is entered as a default parameter rather than a model option, which means that a separate default parameters file should be set up for each project. The program uses an idiosyncratic value called the "Surface Roughness Parameter" (SRP) rather than the surface roughness length (Zo) to describe the terrain. This is defined as SRP = 0.4 / ln(10/Zo) The program operates with a single source, and a series of downwind receptors. Three "locations of interest" and three "concentrations of interest" can be specified, but otherwise receptor locations cannot be specified. The program produces results at a series of irregularly spaced receptors. The data for these receptors is tabulated in a file named c:\...\<filename>.dir\<filename>.msu. This data is hard to process, as it is interrupted with informational messages; the receptor downwind spacing is not regular; and two sets of data at different receptor spacing's are provided for releases which lead to subsequent pool vaporization. No information is given on concentrations away from the plume centerline, although these can be plotted. Ground-level concentrations are given in the penultimate column of data: it should be noted that for mixtures, this data is for the bulk mixture, rather than for the "component to monitor" for which plume centerline concentrations are given. PHAST has a graphical plotting facility, but this frequently fails to operate satisfactorily. It can be useful for some diagnostic work and investigation of the model inputs. Printing from this package is a particular problem.



3.3.7 Other Models

Some of the models are very rarely used for practical dispersion modeling purposes, as follows:

• EFFECTS: This model is developed by TNO, the Netherlands environmental research institute. It is not widely used outside the Netherlands and hence has not been incorporated into this assessment.

• HGSYSTEM: This model was developed by Shell on behalf of a

consortium of hydrogen fluoride producers. It is complex to use, and its chief advantage is the ability to model the thermodynamics and chemistry of HF oligomerization. It is not recommended for general application.

• GASTAR: This model is being developed by Cambridge

Environmental Research Consultants on behalf of the UK Health and Safety Executive. The current release of the model is at b-testing stage. It may prove to be useful for applications in the future, but as it is now at an evaluation stage, and has a small materials database, it is not currently recommended for use.

3.3.8 Summary of Model Applications TABLE 3.2 SUMMARY OF MODEL APPLICATIONS



3.4 COMPARISON OF MODEL RESULTS 3.4.1 General

A series of comparisons have been carried out to evaluate the performance of the various dispersion models relative to one another. These comparisons were made by specifying a set of example runs which were representative of the type of dispersion modeling study commonly encountered. Each model was run using identical inputs, so far as this was possible. The results were compared to evaluate the performance of each model relative to the other models under consideration, and also to evaluate the internal consistency of the models. This evaluation was not intended to comprise a validation study in which dispersion model results are compared with measured data. The following example runs were set up:

• Scenario A: comprised a continuous thermal release of sulfur dioxide in air from an elevated stack. This comprised a buoyant release, and was analyzed using models suitable for this purpose: ADMS, DISP2, ISC and PHAST.

• Scenario B: comprised a catastrophic burst release of chlorine

from a storage tank. This is typical of the cases studied as part of a hazard study, and although it is an unlikely event, the potential effects are very serious. This scenario was analyzed using models appropriate for studying the dispersion of a dense gas release: ALOHA, DISP2 and PHAST.

• Scenario C: comprised a single phase leak of gaseous chlorine

from the vapor space in a storage tank. The release took place close to ground level. This scenario was included to avoid the complications associated with a two-phase release so that the different dispersion characteristics of the model could be compared. This scenario was analyzed using models appropriate for studying the dispersion of a dense gas release as in Scenario B.

• Scenario D: comprised a two-phase leak of liquid and gaseous

chlorine from a storage tank close to ground level. This is a complex scenario, but one which is commonly encountered in hazard assessments.



The differences in model performance may arise from differences in dispersion, and/or in specifying the release conditions. This scenario was analyzed using models appropriate for studying the dispersion of a dense gas release as in Scenario B.



3.4.2 Buoyant gas releases

The results for Scenario A are illustrated in Figures 3.2 to 3.5. It can be seen that all four models gave similar values for the maximum ground-level concentration arising from the particular release under consideration. The maximum concentrations varied from 190 μgm-3 to 540 μgm-3. These maxima all occurred under unstable atmospheric conditions. The distance downwind at which the maximum concentration occurs may be significant when undertaking modeling aimed at estimating concentrations at nearby receptors. For ADMS and PHAST, the maximum concentration occurs 150 m downwind, whereas for ISC and DISP2, the maximum occurs some 300 m downwind. Concentrations at a particular location due to emissions from an elevated source should be checked by considering a range of downwind distances between half and twice as far from the source. The highest concentration obtained over this range of distances should be used to give a more conservative estimate at the location of interest. This procedure allows for the uncertainty in the distance downwind of maximum ground-level concentrations arising from elevated sources. The representation of the atmospheric boundary layer is more advanced in ADMS than in the other three models under consideration. This model predicts lower concentrations at distances above 1000 m downwind of the source than the other three models. This is a matter that deserves further investigation by the model developers, but the difference in results in this region should be noted. If results further than 1000 m from a source are required, it is recommended that more than one model should be used to provide an indication of the likely uncertainty range associated with the modeled concentrations. PHAST gave consistently high results for unstable and stable atmospheric conditions, which is surprising as it uses similar physical descriptions of the dispersion process to DISP2 and ISC. Furthermore, the dependence of results on atmospheric stability was not as expected. Because of this and other known problems with PHAST, it is recommended that PHAST should not be used to model releases of density less than 110% that of ambient air.



3.4.3 Dense Gas Dispersion

The model results for dense gas dispersion demonstrated a reasonable consistency for the majority of cases considered. Under neutral and stable atmospheric conditions, modeled concentrations from the three models ALOHA, DISP2 and PHAST were generally within a factor of 10. For example, see Figure 3.6: this result is for a single phase release under stable conditions, and all three models predict concentrations within a factor of two across almost the entire distance range studied. However, other release combinations gave more variable results - see Figure 3.7 for example. In general, the agreement between models is good, reflecting the large amount of work carried out in recent years to validate and reconcile dense gas dispersion model forecasts.

ALOHA predicts high concentrations in the near field for the catastrophic release case. Indeed, in the immediate vicinity of the source, concentrations can exceed 5 million ppm. The model developers state that ALOHA is a far-field model because it is set up to predict events at distances of more than a few yards from the source, and hence over predicts in the near field (ALOHA User’s Manual, June 1995, p187). This over prediction may be extended further from the source for the catastrophic release considered here.



The sensitivity of the models to atmospheric stability is variable. In some cases, there is almost no difference between the model results for convective and neutral conditions - for example see Figure 3.8. This is due to a coincidental cancelling of the effects of reduced turbulence and increased wind speed. In other cases, the models demonstrate very significant dependence on atmospheric stability - for example, see Figure 3.9.



It is recommended that any of ALOHA, DISP2 or PHAST could be used for modeling dense gas releases. ALOHA and DISP2 are straightforward to use, and could be applied rapidly in, for example, an emergency situation. DISP2 could be used for single phase releases, and ALOHA for two-phase releases. One European chemical company has been developing an emergency response system. This system is intended to be used by non-experts in an emergency situation, and is set up to select the most appropriate of a set of prepared dispersion model cases. The results of these cases are superimposed on a local area plan to assist in emergency response. The system currently uses results given by DISP2, and is set up for use at select European chemical company sites. PHAST would be recommended for planning purposes where dense gas releases may be involved, particularly in more complex situations such as fire/explosion hazard assessment, and combined gas/liquid pool releases. Because of the complexity and unreliability of PHAST, it should only be used by experienced hazard assessors. At all times, model users should bear in mind the variability of the different model predictions, as illustrated in Figure 3.7. For a given set of source and atmospheric conditions, a dispersion model could be expected to give results that are reliable to within a factor of approximately 5.

3.5 RECOMMENDATIONS

(a) Releases with a density less than 110% of air or low inventory releases of dense gases should be modeled using ADMS. ISC or DISP2 could be used if specifically indicated - for example, for continuity or where specified by a regulatory authority. Dispersion over distances greater than 1 km should be checked using an

alternative dispersion model. (b) Releases with a density greater than 110% of air should be modeled as follows: ALOHA Emergency response; US RMP DISP2 Emergency response; project continuity PHAST Planning; complex situations such as fire/explosion hazard assessment, and combined gas/liquid pool releases.



(c) The effects of buildings and hills on dispersion together with information on fluctuations, percentiles, deposition and plume visibility can be modeled using ADMS (d) GASTAR should be considered as an alternative to PHAST and DISP2 for modeling dense gas releases once it has been released in fully supported form with an expanded materials database. (e) PHAST should not be used for modeling non-dense releases for which alternative models are available (f) Prior to using PHAST on any project, users should be aware of the bugs in the model. A current list of identified problems can be obtained. (g) STACK2 should no longer be used except to check earlier calculations. (h) Guidance on estimating source terms is given in Part 6 of this guide. PHAST and ALOHA could be used for source term estimation, but some understanding of the processes involved is required to use these models. (j) Concentrations at a particular location due to emissions from an elevated source should be checked by considering a range of downwind distances between half and twice as far from the source. The highest concentration obtained over this range of distances should be used to give a more conservative estimate at the location of interest.

4 STACK DESIGN 4.1 INTRODUCTION

Dispersion modeling is a vital tool in designing stacks for new plant. It is also very useful in redesigning stacks - for example, in response to changes in regulator policy which are likely to arise in the near future from the new Government air quality targets. This Part of the Guide sets out a simple guide for designing a new stack or redesigning an existing stack.



The steps to be taken are as follows:

• Ensure that emissions of all pollutants from the source are minimized, having regard to the Best Practicable Environmental Option (BPEO), and the Best Available Techniques Not Entailing Excessive Cost (BATNEEC).

• Identify what pollutants are being emitted from the source, and their

emission rates.

• Identify which pollutant(s) is/are the most significant.

• Using a preliminary design, model concentrations of the most significant pollutant(s) on and off-site.

• Amend the design if necessary, and repeat the modeling until

modeled concentrations on and off-site are satisfactory. The steps for designing a stack are summarized in the following flow diagram (see Figure 4.1). Each step is discussed in detail in 4.2, together with a case study to illustrate the procedure. Finally, a further case study is given in 4.3.



FIGURE 4.1 FLOW CHART FOR STACK DESIGN



4.2 STACK DESIGN 4.2.1 Stage A: Preceding Design Work

Some degree of process design will have been carried out prior to arriving at the need to design a stack. This may be very much in outline, or the design may be virtually complete. If the stack is being designed for an existing process, then the details of the process design should be well known. The design should be reviewed to assess whether emissions have been “prevented, or minimized and rendered harmless.” This means that releases should be prevented where possible to avoid the need to disperse emissions. If it is not possible to prevent a release, then the quantity released from the process should be minimized. It is almost always preferable to achieve reductions in emissions by process alterations rather than end-of pipe treatments, which tend to be unreliable and expensive. Process alterations can impose costs in terms of down time, or small reductions in yield, but these are generally not as great as the ongoing costs associated with emissions control equipment. However, other controls on emissions (including end-of-pipe treatment techniques) should also be considered, and if found to be cost-effective, they should be applied to the process. Detailed advice on emissions minimization and the BATNEEC/BPEO regimes can be found in the following sources: • Technical Guidance Note E1: “Best Practicable Environmental Option

Assessments for Integrated Pollution Control”, Environment Agency, The Stationery Office, 1996.

• “Best Practicable Environmental Option Assessments for IPC: A Summary”, Environment Agency, 1996.

• GBHE Process Guide (A Current Technical Practice Guide on the Selection of Process Technology for the reduction of discharges to atmosphere of volatile organic compounds).

• GBHE Process Guide (Vent Collection and Thermal Oxidizer Systems).

• Environmental Technology Best Practice Program (tel. 0800 585794; web site www.etsu.com/etbpp).



Once the process has passed through a review of this nature, the remaining pollutants likely to be emitted from the source should be known. These pollutants should be listed.

4.2.2 Stage B: Estimate Mass Emission Rates

The mass emission rates of each pollutant should be estimated. The mass emission rate is the quantity of each pollutant emitted in a given time, in terms of grams per second, tonnes per hour or an equivalent unit. There may be some complications - for example, many pollutants are not emitted continuously. In this case, it may be appropriate to record the maximum emission rate likely to occur, and the average emission rate over a long time period. Sources of information for these mass emission rates include the following: • Measurements of emissions from an existing process. These could be

from the process for which the stack is being designed, if it is operational. Otherwise, measured emissions from a similar process could be used. These should be checked to ensure that they are representative of the normal operating regime.

• The Environment Agency produces guidance notes for its own

inspectors and for local authority inspectors. These give guidance on the limits likely to be appropriate for a wide range of industrial processes including all those operated by select chemical companies (an example is given in Box 1).



The limits in these documents could be used as estimates of maximum emission rates, on the basis that they are achievable in practice, and compliance with the limits is likely to form the basis of the license to operate the process. Advice on the most appropriate guidance note to use for a proposed process can be obtained from the local business environment adviser.

• For some processes, it may be possible to estimate emissions based

on the design. For example, numerical process design models such as Aspen can be used to estimate emissions from combustion sources. Alternatively, it may be possible to relate emissions of a particular chemical to its vapor pressure in an effluent stream.



Emissions data are frequently given in terms of the concentration in the effluent stream. This may be in terms of volume fraction (“parts per million/billion by volume” or “% by volume”), mass fraction (“parts per million/billion by mass” or “% by mass”), or a concentration (“milligrams/micrograms per cubic meter”). These concentrations may be corrected to a set of reference conditions (for example, 5% oxygen content and 298K). To convert these to estimates of mass emission rates, any values that have been corrected to reference conditions should be returned to the actual measured stack conditions. The formulae in Box 2 can then be applied to obtain a value for the emission rate. The emission rates are used firstly to determine which of the emitted pollutants are the most significant (Stage D). The key parameter is mass emission rate (G) divided by the toxicity (occupational exposure standard or air quality standard). This information is used to narrow down the scope of the modeling that needs to be carried out. It could also be used to return to the process design/emissions minimization stage and focus attention on reducing the impact of the most significant pollutants. Secondly, the emission rates of the most significant pollutants are also an important input to the modeling part of the study (Stages F and G).



4.2.3 Stage C: Identify Acceptable Process Contributions

The next stage of the assessment is to determine the maximum contribution to pollutant levels on-site and off-site due to emissions from the process which would be acceptable. The “process contribution” is defined as the concentration of a given pollutant at a point due solely to emissions from the process under consideration. Legislation relating to air quality is divided into standards which are appropriate for application to on-site impacts, and standards which should be applied off-site. In general, there are more stringent standards for off-site impacts, because a wider range of the population may be exposed, and the exposure will be continuous rather than restricted to the working day. The legislation for on-site impacts does not actually require the occupational exposure standards to be met throughout the site, but does require that site personnel should not be exposed to levels above the standards. Thus, in areas where the standards may be exceeded, it would be acceptable to restrict access to personnel wearing suitable breathing apparatus. There are various sets of standards for both on-site and off-site atmospheric concentrations. Care should be taken to ensure that the most appropriate standard is considered for any pollutant. Acceptable Process Contribution On-site Standards for use on-site are generally referred to as “occupational exposure limits”. The sources of these standards are as follows:



• GBHE, "Hazard Data Sheets" (these would be used if the chemical

is not covered in EH40/98). • Health and Safety Executive, EH40/98 Occupational Exposure

Limits 1998, The Stationery Office, 1998. • The local company Occupational Health Department would be able

to advise on appropriate standards for materials not covered in the Hazard Data Sheets or EH40/98.

Occupational exposure standards are generally given in the form of long-term (8-hour) exposure standards and/or short-term (15-minute) exposure limit. The most appropriate standard for the likely exposure should be used for materials with both a short-term and a long-term exposure limit. In cases of doubt, the long-term exposure standard should be used. Occupational exposure standards should be determined for all the pollutants listed at Stage A. Once standards have been determined, other on-site or nearby sources of the pollutants should be reviewed to determine whether there is likely to be a significant “background” concentration of the pollutant on-site. If this is the case, an estimate of the highest background concentration likely to arise should be subtracted from the occupational exposure standard. The acceptable on-site process contribution is then given by 50% of the occupational exposure standard minus the background concentration. If there is no significant background contribution, then the acceptable on-site process contribution concentration is 50% of occupational exposure standard itself. The acceptable on-site process contribution is referred to as “C1”.



Acceptable Process Contribution Off-site Standards for use off-site are referred to as “air quality standards.” These are discussed in detail in Part 2 of this Guide. An air quality standard should be determined for each pollutant listed at Stage A. There are a number of sources of air quality standards, as follows (in descending order of priority):

• UK Air Quality Strategy Targets. The Strategy document was published in March 1997, and contains targets for the following eight pollutants: benzene, 1,3-butadiene, carbon monoxide, lead, nitrogen dioxide, ozone, fine particles (PM10) and sulfur dioxide. These targets will be legal requirements to be met by 2005.

• European Union Air Quality Standards. Standards already exist for

nitrogen dioxide, smoke/sulfur dioxide in combination, and lead. However, the Air Quality Strategy targets are more stringent than these standards, and should be used in preference to the existing EU standards. The EU is due to specify air quality standards for a range of pollutants during 1997 and 1998. It is likely that the UK air quality strategy targets will be harmonized and broadened to coincide with the new EU standards. However, if there is a new EU standard and no Air Quality target for the chemical in question, then the EU standard should be used.

• World Health Organization Guidelines. The WHO has published a

set of air quality guidelines ("Air Quality Guidelines for Europe", WHO Copenhagen, 1987). A revised set of guidelines is due to be published during 1998. These guidelines do not have legislative backing or official Environment Agency/governmental support, but these guidelines should be applied to pollutants which do not have air quality targets or EU standards.

• In the absence of air quality standards or guidelines, off-site air

quality standards should be derived from the on-site occupational exposure limits already determined. The ambient air quality standard for a pollutant with a Maximum Exposure Limit (that is, which appears in Table 1 of EH40/97) is one fiftieth (2%) of the MEL. Otherwise, the off-site air quality standard is one tenth (10%) of the 8-hour or 15 minute exposure limit (this guidance is based on the Environment Agency Technical Guidance Note E1).



The air quality standard is a limit on concentrations arising from all sources, not just the individual process under consideration. Once an ambient air quality standard has been determined for each pollutant, an “acceptable off-site process contribution” must be obtained from the standard. There is no official guidance on what constitutes an acceptable contribution, except that EA Technical Guidance Note E1 states that a contribution less than 2% is not a priority for control.

Elsewhere, the EA states that a contribution less than 2% would not require alternative emissions control techniques to be evaluated (“Best Practicable Environmental Option Assessments for IPC: A Summary”, Environment Agency, 1996). This document also indicates that a predicted environmental concentration of 80% of the air quality standard due to emissions from all sources would not be acceptable. In practice, an acceptable process contribution which lies between “insignificant” and “unacceptable” levels is specified. In Box 3, tentative guidance for acceptable process contributions is given, based upon practice adopted by chemical company environmental advisers, regulators and external consultants.

For some chemicals, the odor impact is more significant than the health impacts on which air quality standards are based. An acceptable off-site would be one fortieth (2.5%) of the odor threshold. This would be appropriate for use in a residential area: a higher fraction (5-10%) may be acceptable in an industrial area. Hourly means are compared against a small fraction of the threshold to take into account that one’s nose picks up peaks in concentration over a few seconds. These peaks may be 10-40 times higher than the hourly means (see Part 13). Odor threshold data can be obtained from the following sources:



• M Woodfield and D Hall, Odor Measurement and Control - an Update, AEA technology, 1994.

• K Verschueren, Handbook of Environmental Data on Organic

Chemicals, Van Nostrand, 1996.

• WS Cain et al., Reference Guide to Odor Thresholds for Hazardous Air Pollutants Listed in the Clean Air Act Amendments of 1990, USEPA Report No. EPA-68-D9-0173, 1992.

• JE Amoore and E Hautala, Odor as an Aid to Chemical Safety, J

App Toxicol 3(6), 1983, p272. The lower of the acceptable off-site process contributions due to health effects (based on air quality standards) and odor effects (based on odor threshold measurements) should be used in the remainder of the study. The acceptable off-site process contribution is referred to as “C2”.



Summary of Acceptable Process Contributions For each pollutant identified at Stage A, an acceptable on-site process contribution (the concentration of the pollutant at any point on-site due solely to emissions from the process) should be determined as follows: (a) Identify an appropriate occupational exposure standard for the pollutant. (b) Subtract any background contributions due to other sources from the occupational exposure standard. (c) The acceptable on-site process contribution is 50% of this value. An acceptable off-site process contribution should be determined as follows: (1) Identifying an air quality standard for the pollutant. A fraction of the occupational exposure standard can be used if there is no specific air quality standard for the pollutant. (2) Multiply this standard by the appropriate percentage from Box 3 to give the acceptable off-site process contribution. (3) If the compound has a strong odor, then an acceptable off-site process contribution due to odor impacts is given by 2.5 - 10% of the odor threshold.

(4) The lower of these two acceptable off-site process contributions should be adopted.



4.2.4 Stage D: Identify Significant Pollutants

At this stage, the pollutants with the most significant on-site and off-site impacts are identified. The pollutant with the most significant on-site impact is the pollutant with the highest value of mass emission rate divided by acceptable on-site process contribution (C1). Similarly, the pollutant with the most significant off-site impact is the pollutant with the highest value of mass emission rate divided by acceptable off-site process contribution (C2). Thus, to establish the pollutant with the highest on-site impact, divide each mass emission rate by the acceptable on-site process contribution (C1). This will give an on-site index value for each pollutant in units of (mass)/(concentration x time) - for example, gs-1(ppb)-1, or tpa (mgm-3)-1. The exact form of the units is not significant, provided the index for each pollutant is in the same units. The pollutant with the highest index value will have the most significant on-site effects, and will determine whether a given stack design provides adequate protection against on-site health impacts. † A similar set of indices should be obtained for off-site impacts. Again, the pollutant with the highest index value should be selected. This may or may not be the same pollutant identified for on-site impacts. If two pollutants have similar indices and the air quality standards for the two pollutants have different averaging times, the material with the shorter averaging time should be selected. At this stage, it may be apparent that one pollutant is dominating the contribution from all other pollutants. Under these circumstances, there may well be some benefit in reviewing the process design and emissions control provisions to focus more directly on reducing emissions of this pollutant. If reductions can be achieved, there is likely to be a significant reduction in the stack design requirements, with associated reductions in costs. To proceed with a stack design, note the mass emission rate and acceptable process contributions of the most significant pollutant(s) with regard to on-site and off-site impacts.



4.2.5 Stage E: Initial Stack Design An initial stack design is now required. This design will be refined at subsequent stages to ensure that dispersion provided by the stack is adequate to protect health and safety both on-site and off-site, while at the same time not incurring excessive costs due to overdesign. Some initial guidelines for stack design are as follows:

• A location and height should be selected to minimize the chance of the plume striking adjacent buildings and plant structures.

• The stack should be at least 5 meters above any nearby buildings,

walkways, windows, ventilation inlets, or any other location where personnel may gain access. This is to ensure that personnel are not exposed to undiluted stack emissions, and also restricts the influence of turbulence around buildings in bringing elevated plumes down to ground level.

• If possible, the stack should be fixed to existing plant structure to

reduce installation costs.



• A reasonable first estimate for a stack design is given by the formulae in HMIP Technical Guidance Note D1. This is due to be revised and reissued by the Environment Agency, using ADMS as its basis. An alternative approach is to use a stack height already in use on a similar process. Existing stacks have generally been designed under a less stringent regime than that currently used by the Environment Agency, and (other factors being equal), the height may need revising upwards.

Other parameters such as the composition and temperature of the bulk gas mixture are usually known for an existing process, or can be estimated from the design of a planned process. Dispersion modeling can now be carried out to refine the initial stack design to an appropriate final design.

4.2.6 Stage F: Model On-site Concentrations

All the necessary information should now have been obtained to enable dispersion modeling to be undertaken. An appropriate model should be selected for use in determining on-site concentrations of the pollutant with the greatest on-site effect. Considerations in selecting a dispersion model include the following: • Is the release significantly more dense than air? • Is the release likely to be significantly affected by turbulent building wakes? • Is the release flammable? • Is the release single phase or two-phase (liquid and gas)? • How close to the source are personnel likely to be located?



ADMS would be appropriate for modeling a release which is at most 10% more dense than air. It can be used for high molecular weight materials, if they are released at a concentration low enough such that the molecular weight of the bulk release is less than 10% higher than that of air. It can be used to take account of building wake effects, but is not appropriate for use within 10 m of the source. It cannot be used for two-phase releases. Part 11 of this Guide contains examples of studies carried out using ADMS. PHAST would be appropriate for modeling a jet release or a dense gas release, but cannot take account of building wake effects. With care, it can be used for two-phase releases. It can be used to investigate the likelihood and consequences of flammability hazards, although these lie outside the scope of this guide. There are currently a number of unresolved problems with PHAST, and expert advice should be sought before it is used. DISP2 could be used as an alternative to PHAST for dense gas releases. Dispersion modeling calculations should be made to determine the highest concentrations to which personnel are likely to be exposed. The possibility of exposure at elevated locations (for example, on adjacent plant structures, or via ventilation inlets) should be considered. If necessary, exposure at these locations should be investigated by modeling concentrations at elevated receptors, or by considering plume centerline concentrations.

A reasonably wide range of meteorological conditions should be considered to ensure that the maximum concentration which is reasonably likely to be experienced on-site is modeled. As a general guide, wind speeds between 2 ms-1 and 10 ms-1 should be modeled, together with atmospheric conditions ranging from unstable (Pasquill-Gifford Stability Class B), through neutral (P-G Class D) to stable (P-G Class F). An appropriate set of conditions would be the following:



The result of on-site modeling should be to provide an estimate of the highest concentrations of the most significant pollutant likely to be experienced on-site. This value should be compared with the acceptable on-site process contribution (C1) determined at Stage C. If the value is less than 70% of C1, then the design is more than adequate with regard to on-site impacts. It may be possible to make some cost savings by relaxing the stack design. If the value is between 70% and 100% of C1, then the design is acceptable with regard to on-site impacts. If the value is above 100% of C1, then the design is not acceptable with regard to on-site impacts, and the design should be adjusted to reduce the on-site impact (see Stage H, but note also the following comments regarding intermittent releases). If the release is intermittent, then some consideration should be given to the frequency of the release. Where a chemical has a maximum exposure limit (see Table 1 in EH40/97), C1 should never be exceeded. Thus, it is not generally appropriate to make an allowance for the frequency of releases of such chemicals, even if the release is only likely to happen a few times a year. For chemicals without a maximum exposure limit, a small number of accidences of C1 may be acceptable, if it would be excessively costly to design the stack to eliminate these. If this route is adopted, the possibility of restricting access to areas likely to be affected during the release should be considered. Other management measures should certainly be implemented if overall levels of the chemical are likely to exceed the occupational exposure standard by more than a factor of 2. The overall aim should always be to avoid contraventions of the occupational exposure standards.



4.2.7 Stage G: Model Off-site Concentrations

The aims and considerations of modeling off-site concentrations are similar to those for modeling on-site concentrations. Firstly, an appropriate model should be selected for use in determining off-site concentrations of the pollutant with the greatest off-site effect. In general, the same considerations apply to the selection of a model for determining on-site impacts. ADMS would be appropriate for modeling a release which is at most 10% more dense than air. PHAST or DISP2 should be used for other dense gas and two phase releases. Dispersion modeling calculations should be made to determine the highest concentrations beyond the site boundary. Particular attention should be given to concentrations over schools, hospitals, old people’s homes and sites of special scientific interest. A range of meteorological conditions should be considered, as for on-site impacts: a reasonable range of conditions is given in Box 4. These conditions would be appropriate for modeling short-term peak concentrations. The standards for some pollutants are specified in terms of long-term statistics (for example, limits on annual average concentrations or the 99th percentile of hourly average concentrations.



An example of air quality standards expressed in terms of percentiles is given in Box 5). For these pollutants, it may be necessary to model levels of air pollutants using long-term datasets measured at an appropriate meteorological station. Advice on meteorological data can be obtained from GBHE dispersion modeling specialists. If a pollutant is released intermittently, this can be incorporated into an assessment of long-term concentrations. For example, if benzene is released from a source for two hours out of every day, then the annual mean concentration may be estimated assuming a release rate of one twelfth of the 2-hour rate. However, some consideration should also be given to the levels that would arise during the 2-hour release period. Similarly, if nitrogen dioxide is released from a source for two hours out of every day, the 99.9th

percentile concentration would be equivalent to the 98.8th percentile concentration modeled assuming the 2-hour rate. Where the assessment standard is a limit on short term concentrations, then the comments made at Stage F apply also to off-site impacts.



The result of off-site modeling should be to provide an estimate of the highest concentrations of the most significant pollutant likely to be experienced at the site boundary and at nearby properties. The higher of these values should be chosen, except where there is little or no likelihood of public exposure at the site boundary. In this case, the value at the nearest point where public exposure is likely should be used. The highest modeled concentration should be compared with the acceptable off-site process contribution (C2) determined at Stage C. If the value is less than 70% of C2, then the design is more than adequate with regard to off-site impacts. It may be possible to make some cost savings by relaxing the stack design. If the value is between 70% and 100% of C2, then the design is acceptable with regard to off-site impacts. If the value is above 100% of C2, then the design is not acceptable with regard to on-site impacts, and the design should be adjusted to reduce the on-site impact.



4.2.8 Stage H: Assess Results The results from the modeling of on-site impacts (Stage F) and off-site impacts (Stage G) comprise an assessment of whether stack design needs to be altered, or whether it is acceptable. The greater impact (that is, the higher percentage of acceptable on/off-site process contributions) should be considered in this stage of the design process. For example, if the design is more than adequate with regard to on-site impacts (<70% C1), but in the acceptable range with regard to off-site impacts (70 - 100% C2) then the off-site impacts should be used to determine the next step. If the higher percentage is below 70% of the acceptable process contribution, then the stack is more than adequate for dispersion. The design can be relaxed in one or more of the ways given in Box 6, or in some other way. This should not be taken as an indication that emissions from the source can be increased, in view of the requirement on process operators to “prevent, or minimize and render harmless” releases of all pollutants. Once a new design has been specified, return to Stage F to model the impacts of this design.



If the higher percentage is above 100% of the acceptable process contribution, then the stack design needs to be revised to improve the dispersion of pollutants. The design can be revised in one or more of the ways given in Box 7, or in some other way. Once a new design has been specified, return to Stage F to model the impacts of this design.

If the higher percentage is between 70% and 100% of the acceptable process contribution, then the stack design is acceptable - neither over-designed not under designed. A vital stage of the process is to apply “common sense” to the design to check whether it seems to be reasonable. For example, a medium-sized boiler house might have a stack between 30 - 50 m tall: it is unlikely to require either a 6 meter or a 130 meter stack. Unusually high or low values may indicate an error in the calculations, or an unjustified assumption at some stage in the stack design process.



The final stack design can be used with some confidence for budget forecasts, feasibility studies, outline plant design and similar applications. It is recommended that the advice of dispersion modeling specialists should be sought to verify the calculations carried out prior to purchasing decisions being made.

4.3 FURTHER CASE STUDY

Stage A: Process Design: The boiler house will have an output of 55 MW. The boiler will produce 4000 m3/hr of waste gases at a temperature of 300°C. The principal trace components will be carbon dioxide, carbon monoxide and oxides of nitrogen: of these, only emissions of nitrogen dioxide are regulated. When burning oil, particulates and sulfur dioxide will also be generated. Although no detailed design has been undertaken, it is presumed that the plant will meet the requirements of BATNEEC. This will be evaluated by assuming that the plant will emit all pollutants at 50% of the limit given in the appropriate IPR Note (1/1)



Stage B: Mass Emission Rates: Mass emission rates were taken to be 50% of the emission rates in PG1/1. These are as follows: Nitrogen dioxide: Gas burning - 175 mgm-3, equivalent to 83 mgm-3 at release conditions, corresponds to 0.093 gs-1. Oil burning - 248 mgm-3, equivalent to 118 mgm-3 at release conditions, corresponds to 0.131 gs-1. Particulates: Oil burning - 28 mgm-3, equivalent to 13 mgm-3 at release conditions, corresponds to 0.015 gs-1. Sulfur dioxide: Oil burning - 935 mgm-3, equivalent to 445 mgm-3 at release conditions, corresponds to 0.495 gs-1. Stage C: Acceptable Process Contributions: Acceptable on-site process contributions (C1) and off-site process contributions (C2) are determined as follows: Nitrogen dioxide: Occupational exposure standard is 5 mgm-3. Annual mean on-site concentration determined by diffusion tube monitoring is 0.06 mgm-3 (not significant). C1 = 0.5 X 5 mgm-3 = 2.5 mgm-3. Air quality standard is 150 ppb (0.297 mgm-3) for hourly mean concentrations, or 21 ppb (0.042 mgm-3) for annual mean concentrations. Industrial area/common pollutant, therefore C2 = 0.05 ´ 0.297 = 0.015 mgm-3 (hourly mean) or 0.0021 mgm-3 (annual mean) Sulfur dioxide: Occupational exposure standard is 5 mgm-3. Annual mean concentration from historical study is 0.40 mgm-3 (significant, but pessimistic, since majority of SO2 would have been emitted from old boiler house). C1 = 0.5 X 4.6 mgm-3 = 2.3 mgm-3. Air quality standard is 100 ppb (0.276 mgm-3) (15 minute mean). Industrial area/industrial pollutant, therefore C2 = 0.05 X 0.276 = 0.014 mgm-3. Particulates: Occupational exposure standard for respirable dust is 5 mgm-3. Annual mean concentration at nearest monitoring site is 0.025 mgm-3 (likely to be higher on-site in view of number of on-site sources; assume 4 X off-site value). C1 = 0.5 X 4.9 mgm-3 = 2.4 mgm-3. Air quality standard is 0.05 mgm-3 (24 hour mean). Industrial area/common pollutant, therefore C2 = 0.05 X 0.05 = 0.0025 mgm-3.



Summary: Nitrogen dioxide G = 0.093 gs-1

(gas) C1 = 2.5 mgm-3 C2 = 0.0021 mgm-3

(annual) G = 0.131 gs-1

(oil) C2 = 0.015 mgm-3 (hourly)

Sulfur dioxide: G = 0.495 gs-1 C1 = 2.3 mgm-3

C2 = 0.014 mgm-3 (15 min)

Particulates G = 0.015 gs-1 C1 = 2.4 mgm-3

C2 = 0.0025 mgm-3 (24-hour)

Stage D: Identify most significant pollutant: Short-term air quality standards need to be compared with emissions during oil burning as these will be greater than emissions during gas burning. Long-term air quality standards need to be considered with emissions during gas burning. The pollutant on-site and off-site indices are calculated as follows: Nitrogen dioxide index: On-site: 0.131/2.5 = 0.052 Off-site: 0.093/0.0021 = 44 (gas/annual) Off-site: 0.131/0.015 = 8.7 (oil/hourly) Sulfur dioxide index: On-site: 0.495/2.3 = 0.215 Off-site: 0.495/0.014 = 35 Particulate index: On-site: 0.015/2.4 = 0.006 Off-site: 0.015/0.0025 = 6 The highest index is for sulfur dioxide (on-site), and nitrogen dioxide (off-site). However, the off-site index for sulfur dioxide is similar to that for nitrogen dioxide, but the averaging time is much shorter (15 minutes, compared with an annual mean). Thus, sulfur dioxide was also used for off-site impacts. Stage E: Specify initial stack design: An initial stack design was determined from knowledge of similar installations. A stack height of 40 m with a release velocity of 15 m/s was used, giving a stack internal diameter of 0.3 m. A height of 40 m would be sufficient to raise the stack well above the level of any nearby buildings. The stack would be located at the western edge of the site, with the site extending 800 m to the east. Stage F: Model concentration on-site: ADMS was used to model ground-level concentrations of sulfur dioxide on-site. The maximum modeled concentrations were: Stability Class B; wind speed 2 ms-1

Max ground level concentration 12.6 µgm-3

Stability Class C; wind speed 5 ms-1 Max ground level concentration 5.1 µgm-3

Stability Class D; wind speed 5 ms-1 Max ground level concentration 3.5 µgm-3

Stability Class E; wind speed 3 ms-1 Max ground level concentration 1.0 µgm-3

Stability Class F; wind speed 2 ms-1 Max ground level concentration 0.0 µgm-3

Stability Class D; wind speed 10 ms-1Max ground level concentration 2.2 µgm-3



C1 is 2.3 mgm-3 (2300 µgm-3). Thus, the stack is well within acceptable design

parameters with regard to on-site impacts, and a relaxation of the design could be considered. Stage G: Model concentration off-site: ADMS was used to model ground-level concentrations of sulfur dioxide at the nearest properties, which are located 190 m west of the site. The maximum modeled concentrations were: Stability Class B; wind speed 2 ms-1 Max ground level concentration 8.8 mgm-3 Stability Class C; wind speed 5 ms-1 Max ground level concentration 5.2 mgm-3 Stability Class D; wind speed 5 ms-1 Max ground level concentration 3.7 mgm-3 Stability Class E; wind speed 3 ms-1 Max ground level concentration 1.1 mgm-3 Stability Class F; wind speed 2 ms-1 Max ground level concentration 0.1 mgm-3 Stability Class D; wind speed 10 ms-1 Max ground level concentration 2.3 mgm-3.

C2 is 0.014 mgm-3 (14 mgm-3). Thus, the stack is within acceptable design parameters with regard to off-site impacts, and a limited relaxation of the design could be considered. Stage H: Adjust stack design: The maximum on-site concentration is 0.5% of C1. The maximum off-site concentration is 63% of C2. Thus, the design is reasonable, but there is some limited scope for relaxing the design. A revised stack height of 30 m was considered, which gave a maximum off-site ground-level concentration of 12.3 mgm-3. This is 88% of C2, and this design was therefore used for budgeting purposes.

5 DENSE GAS DISPERSION

5.1 INTRODUCTION

The behavior of a dense gas released into the atmosphere is markedly different from that of a continuous neutral or buoyant gas release. A heavier than air release (often defined to be a gas cloud with a density 10% greater than that of air) will slump rapidly to ground level remaining in contact with the ground. This is what makes dense gas releases so dangerous because the highest concentrations occur where people are likely to reside. The highest concentrations generally occur in stable atmospheric conditions. It is of great importance that a neutral density or a buoyant gas dispersion model is not used to represent dense gas dispersion - concentration predictions would be far too low.



(It should be noted that both the PLUME model used in the modeling package DISP2 and the UDM model used in PHAST can cope with buoyant, neutral and dense gas clouds)

5.2 MODELING METHODOLOGIES Most dense gas dispersion models were developed between 1975 and 1990 based upon remarkably similar physics. Calculation methods are split into two categories to reflect two distinct types of problem:-

5.2.1 Instantaneous Catastrophic Releases For worst case analyses of catastrophic incidents, such as the complete loss of containment of the contents of a storage tank, many models have the capability of representing the likely consequences of a large dense gas cloud released instantaneously to atmosphere. In the initial dense gas phase, it is often assumed that the gas cloud takes the form of a vertical cylinder, with its height equal to its radius. The concentration is assumed to be uniform within the cloud. The cloud then falls under its own weight, with a decrease in height and a corresponding increase in its radius (Figure 5.1). Empirical relationships are used to estimate the entrainment rate of air across the vertical face of the plume and across the top of the plume. The key variables controlling the dilution rate are the rate of increase in plume diameter (dependent on the height and density of the plume) and a term known as the Richardson number, which characterizes the effect of density on the air entrainment process.



The higher the Richardson number, the more dense is the gas cloud. Different models use different definitions for the velocity scale (u) and the depth/length scale (h). For example, when estimating the effect of mechanical turbulence on the entrainment rate, the friction velocity (u*) is used. Sometimes the length scale is selected to be the depth of the cloud, other times a turbulence length scale is adopted.



Simultaneous with the dilution processes, it is assumed that the plume as a whole moves bodily downwind (a process known as advection). Usually it is assumed that the centroid of the plume moves at the same speed as the wind at the height of the centroid above the ground. As the cloud dilutes, the density becomes closer to that of air and it behaves increasingly like a neutral density gas cloud. When the density of the cloud is close to ambient or when the plume slumping ceases, most models switch from a dense gas model to a passive dispersion model.

5.2.2 The Dispersion of a Continuous Dense Gas Plume

Credible (rather than catastrophic) accidental gaseous releases will often produce a continuous dense gas plume. Source conditions might include evaporation from pools, releases from holes in vessels or guillotine failures of pipelines. Many dense gas dispersion models have source models to represent pool spreading and evaporation, two phase release thermodynamics and methods for predicting flow rates through holes. Most dispersion models require as input a constant discharge rate whereas in practice, discharge rates will vary with time. Thus it is necessary for models to average discharge rates with time. For pool evaporation, some models use the maximum evaporation rate. If a dense gas plume is discharged above ground level, or released with momentum at an angle above the horizontal, in the first stage of the dispersion the plume slumps down to ground level (Figure 5.2). With the possible exception of the dispersion of a refrigerated gas with a molecular weight less than that of air (e.g. methane or hydrogen fluoride), the plume will remain in contact with the ground. The density of the plume will lead to the plume height decreasing and compared with a neutral density gas cloud, the plume will spread rapidly perpendicular to the wind direction.



At a certain point downwind, the density of the plume becomes close to that of air. Beyond this point, the plume will disperse in a similar way to a ground level passive release.

5.3 POINTS TO NOTE (a) For most dense gas releases, worst case conditions occur during low wind speed stable atmospheric conditions (usually 2 m/s F- stability). Under those conditions, a loss of containment of say 1000 tonnes of Hydrogen Fluoride could result in fatalities tens of miles from the source. (b) Dense gases dilute far more slowly than neutral density gases. (c) There are specific source models for titanium tetrachloride and hydrogen fluoride (d) A dense gas cloud released into the atmosphere with high momentum, but at quite a low flow rate, may not necessarily slump to ground level. It could dilute to ambient density before it has slumped to the ground.



(e) Dense gas dispersion only takes place if the bulk properties of the gas cloud are heavier than air. Many continuous environmental releases will comprise a low concentration of a toxic high molecular weight gas in a large bulk flow of air or nitrogen. This will disperse as a passive gas cloud rather than as a heavier-than air cloud. (f) Low emission rates of heavier than air gases, say, less than 1 g/s, will dilute so rapidly at the source that they will always disperse as a passive gas cloud. (g) Downwind dispersion distances are always sensitive to the discharge rates, which for hazard analyses can be very difficult to define. For example, choosing a one inch source diameter instead of half an inch will result in a fourfold increase in the emission rate. For pool spreading and evaporation, the key parameter in defining the source conditions is often the minimum pool depth - obviously the greater the minimum depth, the smaller the pool area and the lower the total evaporation rate from the pool. (h) In estimating discharge rates from a pressurized source, it is important to remember that at exit velocities greater than around 0.3 times the speed of sound, it is not appropriate to use simple incompressible flow equations (e.g. Bernoulli’s equation). More complex compressible flow equations are required. When the discharge velocity reaches the speed of sound (300 - 400 m/s for most gases) the flow becomes choked. This occurs for source pressures around 0.8 bar gauge. (j) Many worst case dispersion scenarios which specify hazard zones of 50 km or more are probably over conservative. The time taken for a plume to reach a distance of 50 km at a wind speed of 2 m/s is around 7 hours. In practice, the wind direction and atmospheric stability would not remain constant over such a long period. Should such a catastrophic release occur, variations in wind direction would probably result in a broader, shorter plume with lower concentrations. The broader plume, with its large surface area would cause a more rapid rate of dispersion.



(k) It is important to select an appropriate averaging time for the hazard considered. A flammable cloud can ignite immediately after the release occurs - thus an averaging time of a few seconds is appropriate. For comparing concentrations against occupational exposure limits, averaging times of 15 minutes or 8 hours are appropriate.

5.4 VALIDATION WORK

Surprisingly, there are more data sets available for validating dense gas dispersion models than there are for environmental neutral density continuous plume models. The two key sets of experimental data are the Maplin Trials carried out by Shell to represent a spill from a tanker over the sea, and the Thorney Island trials which represented a catastrophic dense gas release over flat ground. Most dense gas dispersion models have been extensively validated against these data sets. However there is very little data available in highly stable atmospheric conditions, which usually represent worst case conditions. Program of full scale release trials still take place at regular intervals. For example, the US organization PERF have recently sponsored trials in the Nevada Desert and in 1999 Hydrogen Fluoride release trials (Project URAHFREP) will take place at a test site in France. In neutral atmospheric conditions, most models may predict downwind concentrations to an accuracy of a factor of around 3, but for catastrophic dense gas release models, their accuracy is unknown. For the same source conditions, predictions made by different models can vary by a factor of up to 10.



5.5 DISPERSION MODELS AVAILABLE

5.5.1 DISP2 DISP2 is proprietary dispersion model, developed by a major European chemical company. It includes a continuous plume model (PLUME) and an instantaneous release (BURST) model. It is recognized by UK regulators, but it should not be used for regulatory applications outside the UK. DISP2 is easy to use, however, its plume module has not been extensively validated, it cannot cope with two phase releases, and has no specific source models, to represent, for example pool evaporation. There are also no fire or explosion models.

5.5.2 HGSYSTEM HGSYSTEM is the Shell suite of dense gas dispersion models which can cope with steady or transient plume problems (HEGADAS) and catastrophic releases (HEGABOX). It also has a range a source models including pool evaporation. HGSYSTEM has additional modules to cope with the thermodynamics and oligomerization of Hydrogen Fluoride. It can be difficult to use and run time errors are frequent. However, it is available as Shareware from US-EPA bulletin board. HGSYSTEM should be used for all dispersion calculations involving Hydrogen Fluoride.

5.5.3 ALOHA

ALOHA is a user-friendly version of the US Coastguard/University of Arkansas model DEGADIS. It is available from the US Bureau of Safety. Possibly the world’s easiest to use dispersion model, there are some concerns over its accuracy, especially close to the source where concentrations can be predicted to be greater than 100%. It can cope with two phase releases and does include source term modeling. ALOHA is possibly the first choice for applications in the USA.



5.5.4 PHAST

Probably the most sophisticated general purpose hazard assessment software package available, PHAST has been developed into a tool that can model most loss of containment incidents. The dispersion model has recently been improved, and the new Unified Dispersion Model can cope with high momentum jet releases, catastrophic dense gas releases as well as passive releases. PHAST is widely used by European chemical companies Hazard Assessors. Its main drawback is its cost. The model has a range of source models, for example, the depressurization of a pressurized storage tank, pool evaporation and two-phase release thermodynamics. Downstream, there is a selection of fire and explosion models, including TNO's multi-energy method. PHAST has an extensive physical properties database RDIPPR.

5.5.5 EFFECTS

Effects is a DOS based Hazard Assessment tool developed by TNO. It is an excellent general purpose hazard analysis tool, using the US Dispersion model SLAB for dense gas dispersion. This would probably be the first choice software package for applications in Holland. It can cope with a wide range of loss of containment incidents - it is a surprise that it is not more widely used outside of Holland.

5.5.6 GASTAR

GASTAR is a dense gas dispersion model developed by Cambridge Environmental Research Consultants. It is possibly the most technically sophisticated dense gas dispersion model, being unique in having a non-uniform concentration distribution in the vertical direction. The UK Health and Safety Executive have funded further developments to the model to include jet dispersion and a user friendly front end. However, it does not have downstream models and its physical properties database is currently rather limited.

5.5.7 LORIMAR Model

A development of DEGADIS for titanium tetrachloride producers.



6 SOURCE TERMS 6.0 INTRODUCTION

Possibly the most complex part of gas dispersion modeling is the specification of the source terms, i.e. the estimation of the emission rate of the discharge into the atmosphere. From experience, the major difficulties lie in hazard analysis applications - particularly important are:

• The estimation of the flow rate through a hole - what are the typical hole sizes produced when pipeline joint gaskets fail, or when reaction vessels are punctured?

• The effect of the source pressure on the flow characteristics though

a hole. For a high pressure gas source, it is necessary to use compressible flow equations.

• The two phase flow characteristics of accidental releases of gases

stored as a liquid under pressure, or from reaction vessels, distillation columns etc.

6.1 SOURCE CHARACTERISTICS AND HOLE SIZES

The first vital step in estimating emission rates to the atmosphere, is the assessment of the hole or source size. There are no generalized methods for estimating hole sizes, each potential source should be considered as a separate case.

6.1.1 Ammonia Storage Tank Example Consider a spherical vessel storing liquefied ammonia under pressure.



The source conditions used as input to a dispersion model need to be specific to the process plant under analysis - the full range of process conditions, possible release points and equipment failures should be considered. Obviously, the higher the source pressures, the larger the hole sizes, the more toxic the materials, the more serious will be the consequences of a loss of containment incident. For the example of the ammonia storage vessel above, the position of the hole relative to the liquid level is vital. If the hole is below the liquid level, then liquid would be emitted; above the level, gas would be produced. Thus for a given hole size and source pressure, the mass emission rate of ammonia produced as liquid would be over 100 times greater than that produced as a gas, simply because at the liquefaction pressure of ammonia the density of the liquid is over 100 times greater than the gas. Also the type of failure of pipe work leading to and from the vessel will lead to very different source conditions. A failure of a weld joining the pipe to the vessel would cause the direct discharge of ammonia as a liquid, which would then rapidly evaporate. However, if there was a guillotine fracture in pipe work, a meter or more from the vessel, the decreasing pressure gradient along the pipe work from the vessel would cause some of the ammonia to evaporate within the pipe work. Thus the discharge into the atmosphere would be two-phase in nature.



Finally, it should be noted that the failure of a pipe often leads to the loss of containment from two sources. For example, the weld failure discussed above would lead to the loss of containment of material from the process plant at the far end of the pipe work as well as the ammonia vessel itself.

6.1.2 Estimation of Hole Sizes

The choice of appropriate hole sizes is case specific - for example, if a credible release scenario arose from a fork lift truck hitting a reaction vessel, then an appropriate hole size might be the cross sectional area of one the prongs on the lifting platform of the truck. Over the years however, a few simple rules of thumb have developed: (a) For large scale (often catastrophic) releases, it is appropriate to assume that the entire inventory is released over a certain period. For example, for US Risk Management Planning, the worst case scenario assumes that the entire contents stored in a tank or vessel is released over a period of 10 minutes. Note, that assuming that the entire inventory is released instantaneously can give over conservative dispersion predictions. (b) For leaks from pipe joints with compressed asbestos fiber, it is often assumed that the worst case hole would arise from a segment of the gasket blowing out between two bolts.

The hole size is estimated to be 2πrt/N where t is the gasket thickness, N is the number of bolt-holes and r is the inside pipe radius. If it is felt that gasket blow-out could not occur (e.g. from spirally wound gaskets), then smaller hole sizes are appropriate.



Based upon GBHE guidance, the cause of a leak from a spirally wound gasket would be flange distortion and a typical leakage gap would be 0.05 mm.

Multiplying this gap by the distance between bolt holes gives an appropriate leakage area. One of the best reviews of hole sizes can be found in Cox, Lees and Ang (1990). They review standard hole sizes for particular process plant items: (1) Leaks from valves typically come from holes of 0.25 mm2 but values as high as 2.5 mm2 are appropriate for catastrophic incidents. (2) Centrifugal pumps. For a centrifugal pump with a 25 mm shaft diameter, the hole sizes recommended are: Mechanical seal, no throttle bush 25 mm2 Mechanical seal with throttle bush 5 mm2 For other pump sizes, the hole size is taken as proportional to the square of the shaft diameter. (3) Reciprocating compressors: Typical hole sizes tend to lie in the range 1 - 5 mm2 (4) Centrifugal compressors: For a 150 mm diameter shaft compressor, the hole sizes recommended are:- Purged labyrinth seal 250 mm2

Floating ring seal 50 mm2

6.1.3 Inventories and Time Dependent Behavior

The duration and hence severity of a release will be highly dependent on the total inventory which could be discharged into the atmosphere. A release from a pressurized vessel will have a decaying discharge rate as the source pressure inside the vessel reduces.



Consequently, concentrations found at any point downwind of the source will also vary with time. It is often useful to compare the duration of the release with the time of flight of the gas cloud to the target point of interest. Consider the discharge of a toxic gas arising from a hole in a process vessel. If the vessel contains 5 tonnes of material and the predicted mean discharge rate is 10 kg/s, then the release would have a duration of 500 seconds. If the wind speed is 5 m/s, then as a very rough estimate, the plume could travel 2.5 km as a continuous plume. A target located 300 m from the source would see the plume as a continuous plume for the duration of the release, whereas a target 10 km away would see the plume as a puff (finite duration) release. Use of a continuous plume model for the point 10 km away would lead to concentration predictions which are too high - the inventory of material would not be sufficient to produce such concentrations over the larger area (see Figure 6.2). Models such as PHAST and ADMS have both continuous plume and finite duration release modules. The balance between the duration of the release and the time taken to reach the target of interest will govern which module should be used for a particular analysis.



6.2 THE DISCHARGE OF GASES THROUGH HOLES

The discharge of a gas from a pressurized source can be characterized by three distinct flow regimes: • Compressible choked flow. • Compressible but unchoked flow. • Incompressible flow. The key factor which governs the flow regime appropriate for a particular source is the source pressure. Choked flow occurs when the source pressure is sufficiently high that the velocity at the hole reaches the speed of sound. Increasing the source pressure further increases the density of the gas flow but does not increase the gas velocity. If the maximum speed of the gas jet is greater than around 0.3 times the speed of the sound, then the flow is compressible but unchoked. Below a velocity of 0.3 times the speed of sound, the flow is incompressible. In the following sections, methods for estimating release rates for the three distinct flow regimes will be presented.

6.2.1 Compressible Choked Flow

Compressible choked flow occurs if the source pressure exceeds the critical pressure (pc) which is defined as:-

For a typical gas, values of γ range from 1 to 1.5. For chlorine, γ is 1.355 and hence the critical pressure is 1.87 times atmospheric pressure.



Typically, for many gases, source pressures around 2 bar absolute and higher will cause choked flow through a hole. If choking occurs, then for an ideal gas exiting through an orifice under isentropic conditions, the gas emission rate is given by the following equation:-

Note that the discharge rate is independent of the ambient pressure downstream of the hole. Pseudo Source Diameters for choked flow For choked flow through a hole or from a nozzle, the pressure in the jet as it is released into the atmosphere is greater than ambient. The jet expands via a series of shock waves down to ambient pressure, with a corresponding change (usually an increase) in jet diameter. See Figure 6.3 Birch, Hughes and Swaffield (1987) analyzed the near field behavior of turbulent jets into the atmosphere. They deduced the following formula by conserving mass and momentum and assuming that the downstream temperature after expansion of the unexpanded jet is roughly the same as the source temperature.



This equation only applies if the source pressure is significantly greater than the critical pressure and that no air entrainment occurs as the jet expands to ambient pressure. (There appears to be an error in the formula quoted in the original paper. The formula above, is the corrected version - a factor of g was missed out in the original paper. Also Birch’s definition of Cd is in error).

Few, if any dispersion models, explicitly take into account the effect of the compressibility of the gas flow on jet dispersion. A very simplistic but conservative alternative to the above involves: (a) The calculation of the mass flow rate of the gas through the hole dependent on the appropriate compressible or incompressible flow regime. (b) The specification of the source diameter to be that required to maintain conservation of mass if the exit velocity is 0.3 times the speed of sound. (c) The use of a standard gas dispersion model for assessing concentration as a function of downwind distance. (Note that momentum is not conserved using this method).



This would not lead to serious errors for toxic releases which could have an impact in the far field but could lead to overestimates of the likely area to be affected by a relatively small release of a flammable gas.

6.2.2 Compressible Unchoked Flow

If the source pressure is less than the critical pressure but the exit velocity is greater than around 0.3 times the speed of sound in the gas flowing through the hole, then the following equation should be used to predict the discharge rate:-

where T1 is the temperature upstream of the hole (K), M is the molecular weight of the gas (kg/kmole) and R is the universal gas constant (8314 J/kgK).

6.2.3 Incompressible Flow

If the exit velocity is below 0.3 times the speed of sound (a), then it can be assumed that the flow is incompressible, and Bernouilli’s equation applies:-



6.2.4 Discharge Coefficients

There is surprisingly little information available on the estimation of discharge coefficients. For a thin sharp edged orifice, C d has a value around 0.6, whereas for a nozzle the discharge coefficient is typically 0.97. For a leak from a thick walled pipe or vessel, the key parameter is the ratio of the wall thickness to the hole size. Smith (1981) suggests the following discharge coefficients:



6.3 TWO-PHASE RELEASES 6.3.1 Catastrophic Releases of a Liquefied Gas

Many gases are stored or processed in liquefied form, with liquefaction being achieved either by high pressure or low temperature. For a loss of containment of a pressure liquefied gas through a hole in a storage vessel below the liquid level, the liquefied gas would pass through the hole as a liquid and a proportion would then rapidly evaporate (known as flashing). A simple way of estimating the amount that flashes is to assume that the vaporization process occurs so quickly that the process can be considered to be adiabatic (i.e. the latent heat required to evaporate the liquid is provided by the heat released by the rest of the liquid cooling to its boiling point).

The flashing process is very rapid and highly energetic and it is found that as the gas flashes, liquid droplets are entrained and released into the atmosphere. These droplets subsequently evaporate. A rule of thumb often used by safety practitioners is that when a pressure-liquefied gas is released into the atmosphere, twice the adiabatic flash fraction is discharged as a gas. Other sources suggest that for low boiling point liquids, the key parameter for estimating the quantity of gas released into the atmosphere is the degree of superheat of the liquid. If the liquid boiling point is greater than 30 deg C below ambient temperature, a loss of containment incident would result in all of the liquid evaporating.



Most dense gas dispersion software packages include modules which explicitly take into account the adiabatic flashing of liquefied gas combined with models which predict the spreading and evaporation rate of the liquid pool formed. From a safety viewpoint, it is usually better for a gas to be stored in liquefied form at its boiling point rather than under high pressure at ambient temperature. If a refrigerated liquefied gas is released into the atmosphere, the consequences will be less serious, since the amount of material that would flash at the point of discharge would be much reduced most of the liquefied gas would form a liquid pool from which the gas would relatively slowly evaporate.

6.3.2 Two-Phase Releases Arising from Guillotine Failures of Pipework

The most widely used technique for estimating two-phase flow for a one component liquid vapor mixture is the homogenous equilibrium flow model which assumes:

• The liquid-vapor mixture is homogenous. • The liquid and vapor phases are in thermal equilibrium. • There is no slip between the liquid droplets and the gas, i.e. no

relative motion between the liquid droplets and gas around it. • The expansion process is isentropic.

Experiments by Fauske (1985) showed that homogeneous equilibrium models are appropriate if the distance between the pipe break and source vessel are greater than10 cm. Fauske and Epstein (1988) derived relatively simple analytical expressions for two-phase choked flow. If the source pressure is significantly greater than that of the liquid vapor pressure at the source temperature, a Bernouilli-type equation is proposed:-



For the above equation it is assumed that the velocity of the gas at the source is zero. If it is not zero then stagnation conditions should be used. For saturated liquid conditions, where ρ o = ρ v(To), Fauske and Epstein (1988) suggest:-

The friction loss factor, F accounts for frictional dissipation based on the length to diameter ratio (L/D) of the exit tube. Fauske and Epstein (1988) suggest the following table of values:

For the transition between region from sub-cooled to saturated stagnation conditions, Fauske and Epstein (1988) suggest a combination of the above two equations:-



For an excellent book on compressible flow, see Aksel and Eralp (1994). If in doubt, ask for specialist advice.

6.4 LIQUID POOL SPREADING AND EVAPORATION A large scale loss of containment of a liquefied gas or volatile liquid, would lead to the creation of a liquid pool on the ground. For safety analyses, we usually need to know the size of the pool and the evaporation rate of the flammable or toxic gas from it. The physics of pool spreading and evaporation is complex and involves the following physical processes occurring simultaneously: (a) Pool Spreading: Driving the spreading of the pool is the static head of the pool (ρgH) where r is the liquid density, g is the acceleration due to gravity, and H is the depth of the pool. Counteracting this, is the roughness of the ground and the surface tension of the liquid. The diameter of the pool is critical - in general the evaporation rate is directly proportional to the area of the pool. Most models do not model the surface tension of the liquid explicitly, instead it is necessary for the user to insert a minimum pool depth. This can be critical, since this minimum pool depth controls the maximum diameter of the pool. As a rough rule of thumb the following table of minimum depths provides general guidance.

(b) Heat transfer from the ground: For the liquid in a pool to evaporate, heat has to be supplied from the ground to provide the latent heat of evaporation. This causes the ground to cool leading to a reduction in the rate of heat transfer into the pool. In turn this reduces the evaporation rate of liquid from the pool. Thus the evaporation rate varies significantly with time, especially for the first few minutes of the release.



(c) The dispersion of the gas evaporating from the pool: The characteristics of the gas cloud around the pool is dependent on both the wind and weather conditions as well as the emission rate. A highly volatile liquid producing a gas with a molecular weight greater than that of air, such as a pool of liquefied chlorine or natural gas, can produce a continuous dense gas plume which remains in contact with the ground. Such a cloud can present a toxic and flammable hazard since the plume can enter and fill enclosed spaces, not just process plant but also drains, underground pipe channels and effluent pits. For such applications, a specific dense gas dispersion model should be used. For worst case analyses of flammable materials, the maximum evaporation rate should be used. Most models will still represent the dispersion from a pool by dispersion from a point source. To provide a margin of safety, the origin for the dispersion calculations should be defined to be the downwind edge of the pool rather than its centre. Less volatile materials, even those with a molecular weight significantly greater than air will behave like a passive gas cloud. As a rule of thumb for outdoor applications, emission rates of 1 g/s or less are unlikely to show dense gas effects. The wind flow will dilute rapidly the low flow rate of gas evaporating from the pool. Consequently, the bulk molecular weight of the gas cloud above and around the pool will be similar to that of air. For these applications, an area source passive gas dispersion model should be used, e.g. the area source module of ADMS. For pool evaporation source modeling, a specific computer based pool spreading and evaporation model should be used rather than simple formulae. The hazard assessment packages, PHAST and HGSYSTEM both have appropriate models. Also available is the AEA Technology model GASP which has a very sophisticated model for both the spreading and thermodynamics of a liquid pool. Note the significant difference in evaporation rates predicted by PHAST and GASP.



Historically, GBHE have used the Clancey Formulae for pool evaporation. Whilst not recommended for new hazard analyses, for the record the formulae are:-



There are a number of assumptions in the derivation of these formulae - values for the latent heat and kinematic viscosity are for a typical (unknown) hydrocarbon! No explicit account is made for heat transfer from the ground - it is assumed that the evaporation rate is controlled by the rate of diffusion into the atmosphere.

6.5 SOURCE TERMS FOR ENVIRONMENTAL RELEASES 6.5.1 General

Usually the dispersion modeler is reliant on measured plant data as source information for environmental impact assessments. This data is notoriously unreliable - take care to understand the following: (a) Concentration measurements are rarely uniform with time. Ensure that there is sufficient data to ensure that the mean and maximum concentrations can be estimated with confidence. (b) Worst case conditions often occur when plants are being started up or shut down. Check that separate measurements have been recorded for start-up and steady state conditions. (c) A myriad of units are used for measured data - often emissions are referenced against standard conditions, such as NTP, STP etc. If the actual source exit temperatures are say 300 deg C, but the volume used for the calculations is the volume normalized to 0 deg C then the dispersion calculations will be seriously in error. NTP often causes confusion - unlike STP, (1 atmosphere, 0 deg C), the definition of NTP varies, including 25 deg C, 1 atm, and 15 deg C, 1 bar. GBHE often uses a term Rbar, which is 20 deg C, 1 atmosphere. Watch for the units of source concentrations. Usually concentrations are quoted as a concentration by volume, or as a mass per unit volume say mg/m3. Note that the conversions from concentrations by volumes to concentrations expressed as mg/m3 are dependent on the source temperature. Usually concentrations expressed as ppm, ppb etc. are concentrations by volume, but they are occasionally used to represent the mass of the pollutant divided by the mass of air.



(d) Sometimes emission rates are estimated from process design data. Often plants are uprated without all the plant documentation being altered accordingly. For a particular emission, plant staff are sometimes uncertain from which vent the emission is being discharged. (e) The presence of buildings close to the source can result in significantly higher ground level concentrations. It is important to be aware of applications where building wakes could significantly affect the dispersion. Typically, building effects are unimportant if the height of the building is less than one-third of the height of the nearby stack. (f) Beware of data supplied in reports to regulators. Emission rates in such reports are often very conservative (i.e. much higher than actually occurs). (g) There can be confusion over the height of stacks. Some plant drawings define heights above a datum rather than the height above ground level. This datum is not always a height above sea level.

6.5.2 A Real Example Recently, the following source data was received from a plant outside of the UK. The plant was a continuous process, making a single product. Regulators had asked for short term and long term concentration predictions.



How should such data be interpreted? Would all consultants treat the problem in the same way? This data can be explained as follows: (a) These are not monthly averages based upon continuous measurement. In practice, these were single measurements recorded once a month over a period of one hour. (b) The plant was being refurbished in May. (c) The measuring equipment was being repaired in February - but this was recorded as 0! The fact that the air flow rate was normal, made me query whether this figure was correct. (d) Only one technician knew how to operate the equipment and he was on holiday in August. (e) For a few minutes every day during a particular loading operation, source concentrations are higher than average. This was the cause for the higher concentration recorded in March. (Subsequent calculations showed that the peak concentration would be around 90 ppm). Thus, for worst case dispersion modeling, a figure of around 90 ppm should be used and for annual average calculations a source concentration figure around 20 ppm would be appropriate.

6.5.3 A Source Data Checklist for Environmental Applications

Below is a list of source data items usually required for a gas dispersion modeling study to assess the environmental impact of a continuous discharge: (a) Stack height and diameter. (The inside diameter of the stack tip). (b) Stack location. (c) Gas temperature at stack exit. (d) Volume flow rate at exit temperature from stack.



(e) Emission rates of all the components of the discharge not just the toxic components. If concentrations are provided then ensure that the units are clearly defined. (f) How does the emission rate vary with time? What is the annual average emission? What would be the peak concentration over a few seconds/minutes/one hour? Remember to consider process start-up conditions. (g) Location and heights of any adjacent buildings within 20 stack heights of the source. Note that process plant structures should only be considered as a building if a significant area of the plant has cladding or walls. (h) Are there any slopes nearby with a gradient steeper than 1 in 10? If so, a complex terrain model should be used. (j) An indication of the general environment around the plant is necessary to estimate an appropriate roughness length (Zo) - see Part 1 of this Guide. (k) Are there any possible targets above ground level, e.g. multi-storey office blocks, adjacent process plant? (l) How far away is the nearest residential/public amenity area beyond the site boundary? (m) A site plan is always useful. (n) Locations of Sites of Special Scientific Interest.



6.6 REFERENCES The best overview of source conditions can be found in chapter 4 of ‘Guidelines for the Use of Vapor Cloud Dispersion Models published by the CCPS (1996), ISBN 0-8169-0702-1. Aksel, MH and Eralp OC (1994), ‘Gas Dynamics’, Prentice Hall, ISBN 0- 13-497728-9. Birch, AD, Hughes, DJ and Swaffield F (1987), Velocity Decay of High Pressure Jets, Combustion Science and Tech vol. 52. pp161-171. Clancey VJ (1974), The evaporation and dispersion of flammable liquid spillages, Chemical Process Hazards I Chem E Sym 39a. Cox, Lees, and Ang (1990), ‘Classification of hazardous locations’, IChemE publications. Fauske HK (1985), ‘Flashing flows: some practical guidelines for emergency releases’, Plant/Operations Progress, vol. 4, pp132-134. Fauske HK and Epstein M (1988), ‘Source term considerations in connection with chemical accidents and vapor cloud modeling’, J Loss Prevention in the Process Industries, vol. 1 pp75-83.



7 BUILDING WAKE EFFECTS 7.1 WHY ARE BUILDING WAKE EFFECTS IMPORTANT?

Buildings have a very significant effect on the dispersion of pollutants from sources released at heights comparable with that of the buildings. A building increases air turbulence which can bring pollution from an elevated source down to ground level. This leads to much higher ground-level concentrations of pollutants than would be predicted in the absence of the building. Typically, the presence of a building can increase the ground-level concentration of a pollutant by up to a factor of 8 under neutral conditions. Under stable conditions, the effects can be much greater. Effects of this scale are clearly a major concern, and one of the most significant factors in pollutant dispersal to air. If building effects are ignored or incorrectly modeled, then ground-level concentrations could seriously be in error. It is vital that building wake effects are correctly modeled to provide an assessment that is conservative, but not unduly pessimistic.

7.2 HOW DO BUILDINGS INFLUENCE ATMOSPHERIC DISPERSION? A building blocks the wind flow, setting up additional turbulence around the building. In the immediate vicinity of the building, a recirculation zone is established, with low wind speeds and consequently long residence times. The turbulent zone typically extends several building lengths downwind of the building before reattaching to the ground. The extent of this zone depends on the atmospheric stability, the size and shape of the obstruction and the wind speed. Wind tunnel and field measurements indicate that the turbulence can extend upwind as well as downwind of the building.



This local air turbulence field can be very significant in atmospheric dispersion modeling. In general terms, an increase in air turbulence leads to an increase in the dispersion of a plume of pollutants. This leads to lower concentrations of material, and a wider plume spread. However, the practical consequence is often that the increased spread of an elevated plume in the vertical direction in a building wake leads to a substantial increase in ground-level concentrations - and ground-level concentrations are generally the major concern for gas dispersion applications. This effect outweighs the increased dilution in the cross-wind direction. Additionally, the lower wind speeds in the recirculation zone lead to longer residence times and hence higher concentrations in this region. In the example in Figure 7.1, part of the plume is affected by building wake turbulence. The part of the plume which crosses the recirculation zone is dispersed very effectively throughout this zone. The part of the plume which passes through the turbulent zone will also experience more vigorous dispersion than in the undisturbed atmosphere. Consequently, the vertical plume spread will be much greater at point A with the building present than if the building were not present. The line B represents the lower boundary of the plume (e.g., one standard deviation (σ z) below the plume centerline) that would arise in the absence of the building. With the building present, the plume is mixed to ground level, and no longer has a straightforward Gaussian form because of the presence of the recirculation zone.



This effect is particularly significant under stable atmospheric conditions. These conditions would typically occur during cold winter nights, when the lowest layer of the atmosphere is cooled by the transfer of heat to the ground. This cold layer of air does not mix well with the warmer air above, and consequently, emissions from elevated stacks spread very slowly in the vertical direction (this behavior is known as "fanning"). This means that ground-level concentrations due to emissions of material from elevated stacks with no nearby buildings are very low. However, if the atmospheric stratification is interrupted by the presence of a building close to the release, a significant proportion of the released material can be dispersed much more vigorously in the turbulent and recirculation zones near the building. The turbulent zone can extend a long way downwind as the stable stratification of the atmospheric boundary layer hinders the dissipation of the turbulent energy in the airflow downwind of the building. Further downwind, dispersion of material at or close to ground level will be restricted due to the stably-stratified atmospheric structure. These factors combine to give high ground-level concentrations of material downwind of the building. Individual buildings can have an influence on the dispersion of material in this way over distances up to approximately 20 times the building height. Further afield, the influence of buildings and other surface features are incorporated into gas dispersion modeling through a parameter used to represent the surface roughness. The surface roughness describes the overall characteristics of the ground over which a plume of pollutants will travel. A rough surface (such as a city centre or chemical works) will set up more turbulence in the atmosphere than a smooth surface (such as a river estuary or open farmland). The increased turbulence for a rough surface will facilitate the dispersion of a plume of pollutants. This is discussed further in the Part 1 (meteorology) and Part 9 (complex terrain) of this Guide.

7.3 SCIENTIFIC UNDERSTANDING OF BUILDING WAKE EFFECTS

The majority of field work studying the effects of buildings on atmospheric plume dispersion was carried out by the CEGB and other power generators during the 1960s - 1980s.1 This work focused on dispersion of power station combustion products from elevated stacks between 100 m and 200 m above ground level. A small amount of work has been carried out more recently on smaller buildings and lower sources.



A considerable amount of work has been carried out in wind tunnels and using computational fluid dynamics (see, for example, the review carried out on behalf of the HSE3). These studies have focused on neutral atmospheric conditions, leaving considerable uncertainty with regard to building wake effects in stable and unstable conditions. It has been found that the turbulent wake around buildings can have a very significant influence on ground-level concentrations arising from an elevated plume. Concentrations of material can be increased by up to an order of magnitude over those that would be experienced in the absence of the building. This effect can extend a short distance upwind of a building, but the main influence is likely to be downwind of the building. A greater effect would be expected under stable atmospheric conditions. The latest gas dispersion models contain algorithms to reproduce the influence of buildings on the dispersion of pollutants from stacks and vents. These algorithms are relatively poorly validated, particularly for low-level sources, and for dispersion in stable conditions. This is because there is little or no experimental data under these conditions. There are no models currently available for investigating the effects of buildings on dense gas dispersion. An extension of PHAST is at the initial stages of development, but there will be a need for experimental data to verify any models that are produced.

7.4 THE BUILDINGS MODULE IN ADMS: PRINCIPLES

The buildings module in ADMS operates by dividing the air surrounding an idealized rectangular building into various zones. These zones are illustrated in Figure 7.2.



The region of recirculating flow, R, may be formed from the leading edge of the building (as shown in Figure 7.2), or from the trailing edge (as shown in Figure 7.1). The region is formed from the trailing edge if the downwind building length is greater than the height or half the cross-wind width, whichever is the smaller. Small changes in wind direction can lead to a change-over between these two cases. This can cause unexpected sensitivity to the precise specification of the building dimensions and wind direction. Any material found to be entrained in this region is assumed to be uniformly mixed throughout the entire region R for most building shapes. If the cross-wind length of the building is more than three times the height, then a sub-region of width three times the height is specified as the well-mixed zone. Experimental and wind tunnel measurements are used to estimate the size and shape of the regions influenced by the building, and to provide expressions for the wind speeds and turbulence parameters in the various regions. The parameters are constrained to converge with the undisturbed values at the boundaries of each region. The dispersion of a plume through part or all of the region influenced by the building (B = R+U+A+W+E) is calculated using dispersion parameters appropriate to the sub-region.



Thus, the plume centerline through region W, for example, follows the wind flow determined by the wind speeds appropriate for this region. The plume spread is modeled using dispersion parameters derived from the normal atmospheric values enhanced by a measure of the excess turbulence in this region. The source material is divided into two parts: the fraction which is fully entrained into the building wake, and the remainder. The dispersion of fully entrained material in region R is calculated using a “virtual” ground-level source located so as to match the conditions found in the recirculation region. The total concentration of material at a given point is the sum of contributions from the recirculation region, and the remainder of the plume.

7.5 THE BUILDINGS MODULE IN ADMS: APPLICATION

This section will cover the following points:

• When should the buildings module be used? • Points to note about using the buildings module. • Interpreting the results.

7.5.1 When Should the Buildings Module be Used?

Building wake effects are generally likely to be significant if a release takes place directly upwind or downwind of a building; the source is within 15 building heights of the building; and the release point is no more than 50% higher than the building. For more distant or lower buildings, there is unlikely to be a significant effect solely associated with this building. The presence of any buildings near to the release should be incorporated into the model via an appropriate value for the surface roughness parameter (using the Complex Terrain Module if necessary).

7.5.2 Points to Note About Using the Buildings Module

The buildings module is generally straightforward to use. Up to 10 buildings can be entered in the appropriate ADMS input screen. However, it is very important to note that these buildings are NOT treated individually by the model. An “effective” building is used by the model which has the height of the “main” building (as specified by the user).



The dimensions of the effective building perpendicular and parallel to the wind direction are calculated from the dimensions of all the buildings entered by the user. Note that only buildings close to the main building and at least half as high as the main building are included in specifying the effective building. For applications in which only one wind direction is to be considered, it is recommended that the user should specify a single representative building which is considered to best represent the complex of buildings that may affect dispersion. This may or may not be identical to the effective building that the model would derive from the user-specification of several buildings. For applications in which a range of wind directions is to be considered (e.g., when measured statistical or sequential weather data is used), it may be preferable to specify the buildings individually. In this case, care should be taken in specifying the “main” building as this is used to determine the height of the effective building. To ensure a conservative assessment, the highest building would be used. However, this may be unduly pessimistic, and an alternative would be to carry out two or more runs with different buildings specified as the main building. The most appropriate calculated concentration at a given receptor would be used, bearing in mind its location relative to the source and the nearby buildings. Exceptionally, for critical applications in complex situations, physical modeling using a wind tunnel may be necessary. If the purpose of the study is to establish the worst-case short-term peaks in concentrations, then the most unfavorable building complex in terms of air pollution impacts should be specified. It may be necessary to try more than one combination of wind direction, main building, receptor location etc. to ensure that the worst case has been covered, particularly where there are multiple sources as well as multiple buildings. Many plant structures are not solid buildings, but “porous” structures with gaps between the plant equipment. This permits some passage of the wind through the structure. A judgment needs to be made as to whether the building is sufficiently dense to give rise to the pressure differences that are the cause of turbulence and recirculation. For example, structures which are sheeted should be modeled using the buildings module in the normal way. The buildings module is probably not appropriate for use with a structure that is judged to be “porous”, e.g. an open chemical plant structure.



A long, narrow building transverse to the wind direction is likely to have a more significant influence on atmospheric dispersion than the same building parallel to the wind direction. Wind tunnel results shown in Figure 7.3 below, indicate that increasing the width to length ratio from 3:1 to 6:1 results in an increase in ground-level concentrations by a factor of 7. Further increases in aspect ratio do not significantly affect ground level concentrations (data taken from DJ Hall et al., Plume dispersion from chemical warehouse fires, BRE Client Report 56/95). If the release point is at or below the height of nearby buildings, the plume may impinge directly on the building face. This is a serious concern, particularly if the building structure is open, if there are windows that may be open, or if there are ventilation inlets on the building. Personnel may be exposed to very much higher concentrations of material than would be experienced at ground level. For this reason, vents should always be located above the roof apex height of nearby buildings.



It is frequently necessary to use the buildings module to assess the influence of storage tanks on gas dispersion - for example, from the vents attached to the tanks themselves. At present, the only way that these can be modeled is to treat the storage tanks as rectangular buildings with the same height and cross-sectional area. Although a rectangular building would have a greater influence on the wind flow than the cylindrical storage tank, this provides a conservative assessment. As no field measurements have been carried out, it is not currently possible to evaluate how conservative this would be. A wind tunnel experiment funded by a European chemical company to investigate this issue is due to start shortly based at the EnFlo centre at the University of Surrey. The advice to treat cylindrical storage tanks as rectangular structures will be updated as new information becomes available. For some applications, ground-level sources of material may need to be assessed. The influence of buildings on ground-level or low-level releases is to increase the rate of dispersion, and hence to reduce ground-level concentrations. Model run-times using ADMS when using statistical weather data are greatly increased when the buildings module is applied. This is because, in the absence of building wakes or other complex effects, plume dispersion is independent of the wind direction. Thus, the model carries out dispersion calculations for a single wind direction, and derives concentrations at the specified receptors from the calculated parameters. If the buildings module is applied, calculations should be carried out for each wind direction, as well as carrying out the calculations of building wake effects. Consequently, run times are increased by a factor of approximately n, where n is the number of wind directions in the meteorological data file, (usually 12).

7.5.3 Interpreting the Results of the Buildings Module Because of the difficulties in specifying a generally applicable module for complex effects, the buildings module can give unreasonable results under some circumstances. Considerable care is therefore needed in interpreting the results of ADMS when the buildings module is used.



In general, the presence of a single building close to elevated sources of air pollutants leads to local increases in ground-level concentrations of material for wind directions from the source towards the building, and vice versa. Figure 7.4 gives an indication of the areas that may be affected. Note that effects can be significant when the source is located downwind of the building. Concentrations in these areas are likely to be a source of concern, and these increased concentrations should be used for design purposes unless there are good reasons to the contrary (for example, the area in question is limited to personnel wearing breathing apparatus). It should be noted that these elevated concentrations will only arise when the wind direction falls in a narrow range. For the majority of the time when the wind is not blowing directly to or from the building, it will not influence pollution levels.

A problem with the dispersion module arises from the use of the module under stable atmospheric conditions. As discussed above, elevated ground-level concentrations are expected under these conditions. However, the model predicts concentrations which can be 70 times higher than those obtained without the buildings module, and also much higher than results obtained under convective conditions. This is illustrated in Figure 7.5 for the case of a release at a building edge, below the top of the building. This scale of effects seems to be greater than what would reasonably be expected, even bearing in mind the relatively small amount of data for these conditions.



This problem may arise from the use of a “virtual” source to represent dispersion from the recirculation zone. While recirculation in building wakes is undoubtedly a real effect, the model algorithms can result in the prediction of excessively high ground-level concentrations of material downwind of the building. It seems likely that some of the assumptions used to apply the building wake module in stable atmospheric conditions are not valid. For example, it may be that under stable conditions, there is not complete vertical mixing in the recirculation zone. For many applications, concentrations determined downwind of buildings in stable atmospheric conditions will be much greater than under any other conditions, and will therefore have a major influence on the study conclusions. It is strongly recommended that any such applications should be referred to dispersion modeling specialists to ensure that design and purchasing decisions are not based on excessively pessimistic data.



8 MODELING THE DISPERSION OF OXIDES OF NITROGEN 8.1 GENERAL

Oxides of nitrogen, or "NOx" are commonly emitted from chemical and combustion processes. NOx consists almost exclusively of two oxides of nitrogen: nitric oxide (NO) and nitrogen dioxide (NO2). Nitrogen dioxide is the more toxic form - the occupational exposure limit for nitrogen dioxide is one sixth of the limit for nitric oxide (see Box 1). Furthermore, there are European and new UK limits on the concentrations of nitrogen dioxide in ambient air (see Box 2). Exposure to elevated levels of nitrogen dioxide can cause respiratory and bronchial problems, particularly for those already suffering from respiratory diseases such as asthma. The issue is complicated by the reactions of oxides of nitrogen in the atmosphere following their release (see Box 3). These reactions eventually result in an equilibrium between NO and NO2. The equilibrium concentrations are determined by the strength of sunlight and levels of ozone and volatile organic compounds in the air, but in general, most NOx would be present as NO2. In contrast, industrial emissions are frequently rich in NO. These conditions are very far from equilibrium, and it can take some time for the oxidation of NO to NO2 to take place.



Site operators are often required to assess the health and environmental impacts of NOx emissions. This might be as part of a SHE assessment or an IPC authorization for example. Dispersion models are routinely used to estimate concentrations of NO2 at locations where individuals may be exposed. To do this, some assumptions must be made about the balance between NO and NO2 at the exposure point.

8.2 ASSESSING NOx LEVELS 8.2.1 Approach 1

It is common practice to assume that all the NOx is present as the more toxic form, NO2. This is a conservative "worst-case" assumption. In practice, the majority of the release is frequently in the form of the less toxic nitric oxide - more than 90% for a typical combustion process. This approach may therefore lead to a substantial overestimate of the impact of the release. This in turn may lead to unnecessary constraints on plant operation, and over-design of stacks or abatement equipment.

8.2.2 Approach 2 An alternative would be to measure or estimate the breakdown of NO and NO2 at the source. This could then be used to assess levels of both compounds against their respective occupational exposure and air quality standards. The disadvantage of this method is that it does not take any account of the conversion of NO to NO2 which takes place in the atmosphere. This method would be appropriate for assessing near-field impacts, such as on-site occupational health impacts. Further from the release point, the oxidation of NO to NO2 will be increasingly significant, and should be taken into account.



8.2.3 Approach 3 A satisfactory alternative would be an interpolation between approaches 1 and 2. Close to the source, the proportion of NOx present as NO2 would be similar to that in the discharge. Further from the source, the proportion present as NO2 would increase as a result of the reactions summarized in Box 2. At a sufficiently long distance from the plant, the NOx would tend towards the equilibrium concentration based on the balance between the formation and loss processes.

8.2.4 Approach 4 Specific computational packages have been compiled in which dispersion models are linked to atmospheric chemistry models of varying complexity. Three models of this type will be implemented in ADMS and these will be evaluated by the GBHE dispersion modeling team. Otherwise, there are no models considered to be sufficiently reliable in both the dispersion and chemistry aspects to be applied within GBHE.

8.2.5 Suggested Method A suggested method for assessing NOx emissions based on Approach 3 is as follows: Step 1: Measure or estimate emissions of NOx from the source(s) in units of mass/second, assuming all NOx is present as NO2. Step 2: Assuming that all NOx is present as NO2, use a suitable modeling technique to estimate concentrations of NO2 at locations likely to be affected (in units of mgm-3 or equivalent). These calculations should be carried out for a range of wind and weather conditions: see Part 4 of this Guide - stack design, for guidance on appropriate sets of meteorological data. If it can be shown that the proposed emissions are effectively harmless as defined by the appropriate regulators, then no further assessment is required. For example, the Environment Agency Technical Guidance Note E1 indicates that the release would not be a priority for control if maximum off-site concentrations are less than 2% of the air quality standard. A higher percentage could be appropriate in some areas and/or for existing plant: a figure of 10% would typically be appropriate for an existing process. This would need to be agreed with the appropriate regulator.



Step 3: Measure or estimate the proportion of NOx present as NO2 in the source (f o). Multiply the modeled NO2 concentrations from step 2 by this fraction. If these concentrations of NO2 are not effectively harmless (i.e., less than say 10% of the air quality standard), then the proposed process shall be reviewed to reduce the impact of NOx emissions. Measures to be considered may include: (a) changing the process to reduce NOx generation; (b) the use of emissions control technologies; (c) coning the vent to increase the velocity of the release; (d) raising the stack height. Return to step 1 with the new design. Step 4: This step should be reached if the emissions are unacceptable on the assumption that all NOx is present as NO2, but acceptable on the assumption that the source proportion of NOx present as NO2 does not change in the atmosphere. (a) Estimate the maximum levels of ozone, nitric oxide and nitrogen dioxide commonly found in the vicinity of the site and receptor location. These may be obtained from national monitoring networks (see http://www.aeat.co.uk/netcen/aqarchive/home.html) - for example, the 95th percentile of hourly average measurements for the previous full year at a representative monitoring site could be used. If no local data is available, use [O3]max = 40 ppb. Levels of oxides of nitrogen will be much more dependent on the source location (urban/suburban/rural; proximity to major roads etc.). (b) Use the spreadsheet " NO2 NOx Ratio" to estimate the proportion of NOx which is present as NO2 at each location. This spreadsheet is available in Excel format, and can be obtained from the authors of this guide. Multiply the modeled NO2 concentrations from step 2 by this fraction. If these concentrations of NO2 are not effectively harmless, then the proposed process shall be reviewed and redesigned to reduce the impact of NOx emissions. Return to step 1 with the new design. If the revised concentrations of NO2 are found to be harmless then the process can be operated satisfactorily.



8.3 EXAMPLE: DISPERSION OF NOx FROM A BOILER HOUSE Are emissions of NOx from this boiler house acceptable, or is additional mitigation required? Step 1: Source details: 8 gs-1 NOx emitted from a 55 meter stack with a diameter of 1.2 meters at a temperature of 360K and a velocity of 12 ms-1. The nearest properties are located 280 m from the source. The area is European suburban fringe, with surface roughness ~ 0.3 m. The source would not be directly affected by any building wake effects. Step 2: ADMS was used to model NOx emissions from the plant for the range of meteorological conditions suggested in Part 4 of this Guide. The highest concentrations at the nearby properties were identified for B stability/2 ms-1; slightly further from the stack, C stability/3 ms-1 gave the highest ground-level concentrations. The maximum modeled NOx concentration at the nearest properties was 60 mg m-3. This is more than 10% of the air quality target value for hourly-average NO2 concentrations of 300 mg m-3. Further assessment is therefore required. Step 3: It is estimated by the process engineers that "almost all " the NOx is emitted as NO: this is taken to mean that f 0 = 0.10. Multiplying the modeled NOx concentration by this value gives 6.0 mgm-3. This is below 10% of the air quality target value. Redesign is not definitely indicated at this stage, and the assessment should proceed to Step 4. Step 4: (a) A review of European Air Quality Archive data for 1995 and 1996 at two nearby sites indicates that ozone levels are below 39 ppb in the area for 95% of the time. This value is taken as [O3]max. The 95th percentile of nitric oxide concentrations at the more representative site is 44 ppb, and nitrogen dioxide 36 ppb. (b) The location-specific proportion of NOx present as NO2 is estimated using the spreadsheet for the two sets of meteorological conditions which gave rise to the highest concentrations. B 2 m/s: [NO2]/[NOx] = 0.74 C 3 m/s: [NO2]/[NOx] = 0.75 (average of results for B and D stability) Thus, the location-specific modeled NO2 concentration at the nearest property is:



B 2 m/s 60 mgm-3 x 0.74 = 44 mgm-3. C 3 m/s 60 mgm-3 x 0.75 = 45 mgm-3. These concentrations of NO2 are greater than 30 mgm-3 and hence not effectively harmless. The release should be redesigned to reduce NOx concentrations at the nearest properties by at least 63%.



9 THE COMPLEX TERRAIN MODULE IN ADMS 9.1 WHAT IS THE COMPLEX TERRAIN MODULE?

The Complex Terrain Module (CTM) is an optional component in the state-of-the-art gas dispersion model, ADMS. It is used to take account of the influence of terrain variability on plume dispersion. Two aspects of complex terrain are considered by the model: variability in terrain height, and variability in surface roughness. Both these characteristics of the terrain can influence dispersion. Changes in surface roughness can significantly affect the local wind flow. The shear stress at the ground-air interface is reduced or increased, leading to changes in wind speed and turbulence across the area under consideration. This in turn affects the dispersion of pollution in the atmosphere.



Where the terrain gradients are not particularly steep, the wind flows will tend to follow the ground contours without significantly affecting ground-level concentrations of trace contaminants. However, steeper hills will influence the dispersion of pollutants. For most conditions, a plume blown towards a hill will not strike the hill directly. Instead, the wind flow will be deflected upwards and outwards so that it flows over and around the hill. However, the hill will distort the boundary layer, often resulting in a speed-up of the wind flow close to the ground, and also affecting the turbulence properties of the atmosphere. This tends to assist the dispersion of pollutants on the upwind slope of the hill. At the same time, the plume centerline is closer to ground level (that is, it rises at a shallower gradient than the hillside), which tends to reduce atmospheric dispersion. Downwind of the hill, there may be a turbulent zone with some air recirculation. All these effects are illustrated in Figure 9.1.

Quantifying the effect of a hill on dispersion under stable atmospheric conditions is more difficult. In these conditions, the plume may bifurcate and pass round the sides of the hill rather than passing over it. Pollutants following the stagnation streamline could directly strike the hill, giving rise to high ground-level concentrations in the area of slow recirculation at the front face of the hill (see Figure 9.2).



9.2 HOW DOES THE COMPLEX TERRAIN MODULE OF ADMS WORK? 9.2.1 Wind Flow

The module uses data representing the surface height and surface roughness length over a given area. This is combined with data produced by the ADMS meteorological preprocessor, based on the meteorological data specified for the model run. The CTM then calculates wind velocity components at a grid of points covering the area, and up to an altitude of several kilometers. As well as the wind velocity, the CTM calculates parameters describing the turbulence of the atmosphere at each grid point. The wind velocity components are calculated using a Fourier transform technique. It is possible to derive formulae for the Fourier transforms of the changes in the velocity distribution over an area of non-uniform terrain. These can be inverted numerically to evaluate the actual flow at each point. This technique is substantially faster than the alternative of computing the full equations of motion at each grid point (as would be done for a computational fluid dynamics study, for example).



The atmosphere over a hill is divided into three layers. The specification of these layers is dependent on the significant influences on atmospheric structure. In the lowest layer, shear stress (that is, friction between the atmosphere and the ground) is important. In the middle layer, the shear stresses themselves are less important than the effects of the shear on atmospheric turbulence. The outer layer contains the remainder of the turbulent boundary layer and may also include part of the free atmosphere. In this region, shear stresses are unimportant.

9.2.2 Dispersion Calculations Having established the wind field (direction, strength, degree of turbulence etc.) across the area of interest, calculations of pollutant dispersion are carried out. In calculating concentrations at a particular point due to emissions from a particular source, allowance is made for the modifications to the wind field and the surface roughness at each point over which the plume passes. The amended parameters are used in the Gaussian dispersion model to reproduce the effects of complex terrain on dispersion. For example, the plume centerline direction may be altered by the presence of a hill: this will have a variable effect on ground-level concentrations depending on whether the centerline is moved closer to or further from a particular point. Also, a hill will cause the plume centerline to approach closer to the ground.



A plume dispersing across an area of high surface roughness (e.g., an urban area) will undergo rapid spreading. If the plume then crosses an area of low roughness (e.g., a river estuary), then further spreading will occur more slowly.

9.3 WHEN AND HOW SHOULD THE COMPLEX TERRAIN MODULE BE USED?

As stated in the ADMS User Guide, the Complex Terrain Module should be used in areas with hill slopes of greater than about 1 in 10, and less than 1 in 2. There is essentially no restriction on the range of values of surface roughness length that could be used in the CTM. The CTM cannot currently be used in conjunction with long-term meteorological datasets to obtain annual mean or percentile concentrations; in any case, this would require unacceptably long calculation times. This option will be available with the next release of ADMS and run times should also be significantly shorter to make this a viable option. Guidance for deciding when to use the CTM to take account of terrain heights is given in Box 1. Guidance for deciding when to use the CTM to take account of variations in surface roughness is given in Box 2. Using the CTM can be time consuming, both in terms of preparing the model input data (files of terrain elevation and surface roughness lengths), and in terms of the long model run times. For example, one of the runs referred to in Section 9.4 (5 sources; 31 x 31 receptors; 3 meteorological conditions; terrain height and surface roughness included). The most significant point to note when using the CTM for stack design or assessment is that the worst case conditions may well change from those that would give the highest pollution levels without the complex terrain option. Secondly, using the CTM is unlikely to lead to dramatic changes in maximum modeled pollution levels. A factor of two is the greatest increase that is likely. Higher changes are likely in some cases, but these are unlikely to be the conditions giving the highest ground-level concentrations.



Terrain elevation data suitable for using in ADMS may be difficult to obtain. This data needs to be processed using a utility provided with ADMS to give a file covering the area of interest for the particular application. In the example shown in Figure 9.3, the grid reference of the datum (0,0) point is SJ480760, and the range extends 10 km north and east from this point. A grid size of 64 x 64 was used: note that a grid size of 16 x 16 is not recommended for design use. Surface roughness data is not available routinely in the same way that terrain elevation data can be obtained. The most straightforward method of obtaining a surface roughness (.ruf) file is to create a terrain elevation (.ter) file, and import it into a spreadsheet package. The elevation data (the fourth column of data) can be replaced with values for surface roughness length, and the file saved in text file format. Typical values for surface roughness lengths are as follows:

One useful technique for creating appropriate values of surface roughness length in a coastal area is to assign a value of 0.001 m to all points where the terrain height is less than (say) 5 meters, and a representative value for the type of area at all other points. In other cases, it will be necessary to use a map to identify regions of different surface roughness - for example, urban areas, industrial zones and open farmland.



9.4 WHAT IS THE EFFECT OF USING THE COMPLEX TERRAIN MODULE?

An analysis of the effects of the Complex Terrain Module on modeled ground-level concentrations has been carried out. The conclusions reached as a result of this assessment are summarized in Boxes 1 and 2. This Section gives the background information supporting the guidance given in these boxes. A range of source combinations has been considered, using the topography around a select European chemical site as an example of an area where the terrain may significantly affect the dispersion of pollutants. Hill slopes in the near vicinity of a chemical works are unlikely to exceed those at this select European chemical site. The European chemical site is located at the foot of a Hill, as shown in Figure 9.3.



The Hill rises steeply to the east of the European chemical site to a height of 70 m above sea level. The gradient of the hill exceeds 1 in 10 on this west-facing slope, and thus could significantly affect dispersion. The majority of the hill comprises built up urban or suburban districts. The estuary of the a river is immediately to the west of the European chemical site. The estuary area will have a much lower surface roughness than the industrial/urban areas of the European chemical site, and this may also affect the dispersion of pollutants. To investigate and (if possible) quantify the scale of effects likely to arise, dispersion modeling of emissions from 5 sets of sources on the European chemical site has been carried out. Each set consists of 5 sources located at intervals on the site from north to south. These sources are not intended to represent any actual release points on the site itself. The first set of sources consists of 5 release points, each at a height of 20 m above local ground level. The second set consists of 5 release points at 40 m above ground level. The third set are 60 m above ground level; the fourth set are 80 m above ground level; and the fifth set are 100 m above ground level. All sources were assigned a diameter of 0.5 m. The release velocity from each source was 9 m/s of gas of neutral density at 15°C containing 1 g/s of pollutant. Releases from each of the five sets of sources have been modeled for each of the following scenarios: Scenario 1: No complex terrain effects; westerly wind Scenario 2: Inclusion of terrain heights; westerly wind Scenario 3: Inclusion of terrain heights and variability in surface roughness between land and sea areas; westerly wind Scenario 4: No complex terrain effects; easterly wind Scenario 5: Inclusion of terrain heights; easterly wind Scenario 6: Inclusion of terrain heights and variability in surface roughness between land and sea areas; easterly wind In scenarios 2 and 3, the sources are located upwind of the hill. Thus, the dispersion of emissions may be affected by the reduction in plume centerline height, and by the acceleration of the wind flow.



Overall, it would be expected that ground-level concentrations of released material would tend to increase when the topography is taken into account. In scenarios 5 and 6, the sources are located in the turbulent wake region downwind of the hill. In these scenarios, it would be expected that ground-level concentrations would be reduced by taking the topography into account. The difference in model forecasts are illustrated in Figures 9.4 to 9.9. Each figure contains results for 3 scenarios obtained using stack heights of 60 m. The scenarios shown in each figure are as follows: Figure 9.4: Stability Category B (convective); scenarios 1, 2 and 3 (stacks upwind of hill) Figure 9.5: Stability Category B (convective); scenarios 4, 5 and 6 (stacks downwind of hill) Figure 9.6: Stability Category D (stable); scenarios 1, 2 and 3 (stacks upwind of hill) Figure 9.7: Stability Category D (stable); scenarios 4, 5 and 6 (stacks downwind of hill) Figure 9.8: Stability Category F (unstable); scenarios 1, 2 and 3 (stacks upwind of hill) Figure 9.9: Stability Category F (unstable); scenarios 4, 5 and 6 (stacks downwind of hill) The results obtained for a range of meteorological conditions and stack heights are compared in Table 9.1 (stacks upwind of hill; Scenarios 1, 2 and 3) and Table 9.2 (stacks downwind of hill; Scenarios 4, 5 and 6).



Note 1: The “representative concentration” is the 90th percentile of concentrations modeled at all 961 grid points (i.e., the 48th highest ground-level concentration for each scenario/set of meteorological conditions). This concentration occurs in the downwind region of the plume away from the immediate vicinity of the source, which is typically a region of concern in dispersion modeling studies. Concentrations are given in arbitrary units.



9.4.1 Conclusions - Terrain Elevations

Convective conditions Incorporating the effects of terrain height has little effect under unstable (convective) atmospheric conditions. For releases from elevated sources (more than 20 m above ground level) which are not directly influenced by building wake effects, unstable conditions frequently give the highest ground-level concentrations. In cases where short term concentrations of pollutants released from elevated sources are of concern, there is therefore little to be gained from incorporating the CTM into a dispersion modeling study. If challenged for not taking the effects of complex terrain into account in a dispersion modeling study, it would generally be reasonable to state that “the effects of complex terrain are of little significance when studying dispersion in convective conditions, as the thermal atmospheric turbulence dominates the dispersion process.”



Neutral conditions Under neutral atmospheric conditions (which occur for the majority of the time in Europe), complex terrain can have a significant influence on ground-level concentrations. The results shown in Table 9.2 indicate that a reasonable sized hill close to the release point may lead to an increase in ground level concentrations of 50%. Over the region downwind of the sources, ground level concentrations were found to increase by 25 - 75% with incorporation of terrain effects. If this scale of increase could be significant, then complex terrain effects should be incorporated into a dispersion modeling study. This would apply if modeled off-site concentrations without complex terrain effects were within a factor of 2 of the acceptable level for a particular application. On the other hand, if modeled concentrations are a factor of 10 below the acceptable level, there would be little to be gained from incorporating complex terrain effects. Stable conditions Stable atmospheric conditions frequently lead to the highest ground-level concentrations at long distances from an elevated source, or close to a low-level source. In some cases, the model was found to give higher concentrations without incorporating terrain effects; in other cases, incorporating terrain effects led to an increase in modeled ground-level concentrations. For cases where a plume released under stable conditions reaches ground level, the most significant effect of incorporating terrain elevations is likely to be the increased atmospheric turbulence. This leads to a reduction in ground-level concentrations of up to 50%. In cases where a plume released under stable conditions does not reach ground level, ground-level concentrations are likely to be much smaller than those predicted under neutral or unstable conditions, although they may be increased by applying the CTM. It is therefore concluded that incorporating terrain elevations into a dispersion modeling study is unlikely to significantly increase the highest modeled ground-level concentrations.

9.4.2 Conclusions - Variations in Surface Roughness Incorporating the effects of surface roughness leads to variations in dispersion in the crosswind and vertical directions from the values that would be obtained if these variations were not incorporated. In general, variability in surface roughness is only likely to have a significant effect on the dispersion of pollutants from industrial sources if there are large areas with very different surface roughness scales. The obvious example is dispersion taking place over mixed land



and water areas. Dispersion in an urban fringe area, or from an industrial site in a rural location could also take place over areas with significant variability in surface roughness. Furthermore, variability in surface roughness only has a significant effect on modeled ground-level concentrations in the region immediately surrounding the source. Thus, as a general guide, surface roughness variability only needs to be incorporated into a model run if the surface roughness length within 500 m of the source varies by more than a factor of 100. Under these conditions, a region with large roughness elements (such as a chemical plant or city centre) would give increased dispersion. This would encourage the dispersion of a ground-level plume, but could bring an elevated plume down to ground level more rapidly. A relatively flat region such as a river estuary or farmland would inhibit the dispersion of pollutants at ground level, but at the same time would not facilitate the mixing of an elevated plume down to ground level. These effects are likely to change modeled ground-level concentrations by less than 30%. Variations in surface roughness are therefore unlikely to be a critical influence on modeled concentrations, compared with the other common uncertainties in dispersion modeling studies. It is recommended that a single representative value of Z o should be used for the region over which dispersion is taking place, unless there are very strong discontinuities in the type of terrain. If strong discontinuities exist, then the Complex Terrain Module could be used. Alternatively, model runs could be carried out using two values representative of the different types of terrain to give an envelope of results. These are unlikely to lead to

estimates of the highest concentrations which differ by more than 30% at any location, and would avoid the long run times resulting from the use of the CTM.



9.4.3 Conclusions - Buoyant Releases The conclusions presented above are based on results obtained for a plume of neutral density - i.e., the same density as the air. A buoyant plume will rise more rapidly than a neutral plume, and hence will tend to have less interaction with the ground. The guidance given above can therefore be applied to buoyant releases because, if anything, it will tend to be more conservative than for neutral releases.



10 THE DEPOSITION MODULE OF ADMS - A BRIEF GUIDE 10.1 INTRODUCTION

"Deposition" refers to the mass transfer of dissolved gases or suspended particles from the atmosphere to the earth's surface by precipitation ("wet deposition") or in the absence of precipitation ("dry deposition"). The process of wet deposition is similar for particles and gases; however, different dry deposition processes operate for large/heavy particles compared to small/light particles and gases. The purpose of the deposition module in ADMS is to model the deposition processes which are very important in determining the exposure of humans, plants and animals to environmental pollutants. The deposition process can also be significant in affecting atmospheric concentrations of material, by depleting atmospheric levels of the pollutants uniformly or non-uniformly across the plume. Situations where GBHE users may need to investigate deposition models include the following: (a) Modeling the deposition close to the source of coarse particulate material emitted from vents and stacks - for example, ammonium nitrate dust from prilling towers. (b) Dispersion over distances such that deposition processes may significantly affect airborne concentrations of material. In these cases, the main interest is in the concentrations of material in the air, rather than the quantity of material deposited on the ground. (c) Modeling the deposition of materials such as mercury which may accumulate in the environment. (d) Modeling the deposition of materials such as dioxins for which deposition can be a significant human or ecological exposure pathway (in the case of dioxins, by deposition on grass, consumption by cows and human consumption of milk). (e) Modeling the deposition of non-volatile liquid sprays. The remainder of Section 10.1 provides background information on the deposition process. Section 10.2 describes how deposition processes are modeled in ADMS. Section 10.3 provides recommendations for values to be used in deposition modeling.



Some points relating to the use of the deposition module of ADMS are summarized in Section 10.4, and an example is given in Section 10.5. The deposition process takes place via two principal mechanisms: (1) Wet deposition: the process by which pollutants are removed from the plume of material by precipitation. This results in a uniform loss of material from the whole of the plume, and consequently does not affect the concentration distribution in the vertical direction. (2) Dry deposition: the process by which pollutants are removed from the plume of material by contact with the ground. This takes place at the ground surface, and consequently dry deposition selectively removes material from the bottom of an airborne plume. This can significantly affect the concentration distribution in the vertical direction, with reduced concentrations at or near ground level. Dry deposition is more complex than wet deposition. The dry deposition of gases is very dependent on the chemical characteristics of the material, and also on the surface type. Deposition takes place when gases are brought into contact with the ground surface by the turbulent flows in the air. Vertical turbulence movements are by definition more vigorous under convective conditions, and so deposition rates may be expected to be greater under these conditions. Dry deposition will also be encouraged onto ground types with high surface area (such as trees in leaf, or flowering plants). To reproduce these effects, it is necessary to specify dry deposition parameters for individual combinations of pollutants/surface types/meteorological conditions. A summary of measured dry deposition velocities is given in Section 10.3, but it is recognized that this list is incomplete. The variability in dry gas deposition rates is a significant source of uncertainty in deposition modeling. The dry deposition of small particles will tend to follow a similar pattern. "Small" particles are defined as those for which the velocities of turbulent motion are greater than the terminal velocity. Dry deposition velocities for small particles will consequently be dependent on the surface type, although the mechanism for attachment to surfaces will be different to that of gases. The dry deposition of large particles will be dominated by gravitational settling of the material. Thus, for large particles, the deposition velocity approaches the terminal velocity given by the balance between gravitational forces and air friction (Stokes' Law).



The parameters that control deposition rates are, in general, not well known. For example, wet deposition is governed by the amount of precipitation at any point. This is a parameter which is inherently variable, and cannot be forecast with any degree of certainty. Complications are introduced if the rainfall rate is variable across a plume. Deposition rates are also dependent on meteorological factors (wind speed, stability, friction velocity, relative humidity etc.); factors related to the pollutant (particle size and mass, solubility, chemical reactivity, etc.); and the nature of the terrain (pH, prior loading, areal density, type of vegetation cover etc.). In general, wet deposition rates vary in time, but do not vary much over distances of a few tens of kilometers. Dry deposition rates vary both in time, and over very short distances. The following sections aim to provide guidance on appropriate parameters to use, and also to give an indication of the reliability of the deposition module. Dry deposition is characterized by a "deposition velocity" vd, defined by:-



As discussed above, the dry deposition velocity can be broken down into a gravitational settling component and a gaseous deposition component. For a gaseous or very fine particulate material, the gaseous deposition pathway will predominate, and dry deposition will be strongly dependent on the surface type. For coarse particulate material, the gravitational forces will predominate. For particles of intermediate size, there is a minimum in the dry deposition velocity when both gravitational settling and turbulent deposition are relatively weak (see Section 10.3). Wet deposition is characterized by a "washout coefficient" Λ, defined by Λ = D/C where D is the loss rate of material (gs-1m-3)

C is the concentration of material (gm-3)

10.2 DEPOSITION MODELING METHODOLOGY USED IN ADMS

ADMS uses a simple methodology to represent deposition, based on the following assumptions:

• Dry and wet deposition can be represented by proportionality between the removal rates and the atmospheric concentration of the material.

• The proportionality constants for wet and dry removal are constant

over the study area - for example, saturation of the rainwater does not occur; there are no chemical or physical transformations within the plume (such as aerosol agglomeration) which could affect the deposition rates; and the deposition rates for all ground types are represented by a single value.

• The proportionality constants for wet and dry removal are constant

over all meteorological conditions.

• Removal processes act independently.

• Removal processes are irreversible.

• All plume material lies in or below the rain clouds.

• The rainfall rate is uniform over the study area.



The principal uncertainties in using the ADMS deposition module are likely to be as follows:

o Selecting appropriate values for the dry and wet deposition coefficients. It may not be possible to select values that are entirely satisfactory - for example, there is considerable variability in deposition rates between urban, rural and water surface areas, and deposition rates on vegetation depend strongly on the type of vegetation, the time of day and the time of year. Deposition rates are higher during the day and in the summer when there is a relative abundance of foliage and other surfaces for dry deposition to occur.

o Identifying rainfall rates. For short-term applications, a "worst

case" approach will probably need to be used, based on (say) the 90th percentile of 24-hour rainfall rates. For long-term calculations, the rainfall rate is specified in very coarse "bins" - for example, 0 mm/hr, <0.6 mm/hr and >0.6 mm/hr, with a single value used to represent each class. This gives a very coarse representation of the variability in rainfall rates.

o Particulate density, diameter and shape. The size distribution of the particles is not likely to be well known, and some assumptions will probably be required. Additionally, the dry deposition rate for coarse particles relies on the assumption that the particles are spherical, which may well not be the case.

o Variability in the atmospheric dispersion process and other uncertainties associated with dispersion modeling in general.

The deposition module requires the following inputs for gaseous pollutants: (a) Washout coefficient, Λ If Λ is not known, values for a and b to be used in the equation Λ = a.J b where J is the rainfall rate in mm hr-1. (b) Deposition velocity, v. If v is not known, identification of the pollutant as reactive, non- reactive or inert. The deposition module requires the following inputs for particulate pollutants:



(1) A size distribution of the particles (i.e., a range of mean particle diameters and mass fractions having that mean diameter). (2) Washout coefficient, L If L is not known, values for a and b, as above. (3) Deposition velocity, v, and terminal velocity for each size fraction If v is not known, specification of the particle diameters and mass fractions If the terminal velocity is not known, specification of the particle diameters, mass fractions and densities.

10.3 DEFAULT AND RECOMMENDED INPUTS USED IN ADMS 10.3.1 Wet Deposition

In the absence of a specific washout coefficient supplied by the user, ADMS uses default values of a = 10-4 s-1 and b = 0.64. These values are appropriate for a range of heavy metal radionuclides, and are taken from K Jylha, Empirical Scavenging Coefficients of Radioactive Substances Released from Chernobyl, Atmospheric Environment 25A, 263- 270, 1991. The US EPA recommends a = 1.6 x 10-6 and b = 1 for Hg2+ (Mercury Study Report to Congress Volume III: Fate and Transport of mercury in the Environment, 1997). For particulates, the US EPA Mercury study recommends b = 1, and values for a dependent on the mean particle diameter (the following equation is a fit to data given in the report): a = 6.3 x 10-5 + 0.0002 log10d + 0.00035 (log10d)2 s-1 0.1 < d < 10 mm a = 6.6 x 10-4 s-1 d ³ 10 mm where d is the mean particle diameter (mm). Wet deposition coefficients for other materials are given in Table 10.1. These are taken from R Singles, MA Sutton and KJ Weston, A multi-layer model to describe the atmospheric transport and deposition of ammonia in Great Britain, Atmospheric Environment 32, 393-399, 1998, assuming a value of b = 1.



10.3.2 Dry Deposition

The user has the option of specifying dry deposition velocities directly. It is recommended that the user should do this, using the data given in Figure 10.1 (for particles) and Table 10.2 (for gases). In the absence of specific dry deposition coefficients supplied by the user, ADMS estimates dry deposition velocities. For particulates, the estimated value is calculated from wind tunnel measurements, together with the terminal velocity calculated using Stokes Law. For gases, the user is asked to categorize the pollutant as "reactive", "unreactive" or "inert." Inert gases have zero deposition velocity; reactive and unreactive gases have values of approximately 1 cms-1 and 0.1 cms-1 respectively. The use of these default values is not recommended - while they are not unreasonable, they could easily be in error by a factor of 100. Dry deposition rates are strongly dependent on the type of area and the prevailing meteorology, as well as the characteristics of the material, as discussed in Section 10.1. For particulate material, the dry deposition rates shown in Figure 10.1 are recommended (taken from GA Sehmel, Particle and Dry Gas Deposition: A Review, Atmospheric Environment 14, 983-1012, 1980.)



The following dry deposition velocities given in Table 10.2 are recommended for gaseous compounds. These were obtained from a short literature survey. For situations where the compound/surface of concern is not listed, alternative sources of data could be used, or reasonable values to use could be estimated from the data given. For example, a reasonable value for the deposition velocity of chlorine over an urban area might be similar to the deposition rate over the root crop alfalfa, in view of the similar velocities for sulfur dioxide over urban and crop surfaces.



10.4 RECOMMENDATIONS FOR USING DEPOSITION MODULE

There are very significant uncertainties associated with using the deposition module. The main sources of uncertainty are as listed below: • Selecting appropriate values for the dry and wet deposition coefficients. • Identifying suitable rainfall rates for short-term applications. • Uncertainties associated with dispersion modeling in general. This combination of uncertainties probably means that the uncertainty in an annual average calculation is likely to be of the order of a factor of 10. Where possible, model predictions should be compared with measured data to provide an indication of the reliability of the model results. Wet deposition rates are dependent on the integrated quantity of the pollutant above a given area of ground. Close to a source of pollution, crosswind plume spreading is at its smallest, and consequently, wet deposition rates are greatest close to a source of any height. Wet deposition rates decrease further away from the plume, but can still be significant. Conversely, dry deposition is dependent on the ground level concentration of material, and follows the pattern of ground level concentrations. Thus, dry deposition may be expected to be a maximum at a distance of 5-20 stack heights from an elevated source. Deposition may be a significant factor in determining airborne concentrations of material, due to the loss of material to the ground. The deposition parameters that would give a 10% reduction in ground-level concentrations at a range of downwind distances from a 15 meter source are as follows. Two values are given for wet deposition. The first is based on a typical annual mean rainfall rate of 0.067 mm/hr (i.e., 0.6 m/year). The second is based on a typical precipitation rate during a period of rainfall of 0.7 mm/hr.



Thus, for example, the ground-level concentration of ammonia would be appreciably reduced (10% reduction) by dry deposition at a distance of around 1800 m from a source (assuming a deposition velocity of 2.5 cms-1). Wet deposition is unlikely to be a significant plume depletion process for heavy metals and particulates.

10.5 EXAMPLE APPLICATION OF THE DEPOSITION MODULE

The annual emissions budget from the site is 2500 tonnes per year of SO2. The majority of this is from elevated release points at a temperature of around 200°C. Because the target area is some distance from the site, the results will not be sensitive to individual release points, and the releases will be modeled as an area source covering the entire site. A release height of 30 m will be used. The site covers an area of 1 km (E-W) by 0.7 km (N-S). The north edge of the moorland area is located 4 km south of the south edge of the site, and covers an area 2 km (E-W) by 1.5 km(N-S). The majority of the land between the site and the moorland area is farmland, and a surface roughness of 0.1 m was used. Annual hourly meteorological data for the most recent complete year from a nearby measurement station was used to provide a reasonable breakdown of rainfall rates with other meteorological conditions, in spite of the additional processing time associated with running each hour of data separately.



The following deposition coefficients were used:

Wet deposition coefficient a: 1.1 x 10 -4 s-1 (mmhr-1)-1 (taken from Table 10.1) Wet deposition coefficient b: 1 Dry deposition velocity: 0.8 cms-1 (taken from Table 10.2) Wet and dry deposition rates, and ambient concentrations of NOx were calculated at 500 m intervals over the moorland area, using ADMS. It was found that the mean wet and dry deposition rates over the moorland area were 1.31 x 10-10 gm-2s-1 (wet deposition) and 6.18 x 10-12 gm-2s-1 (dry deposition). This gave an annual mass deposition rate of 13 kg/year on the moorland area.

11 EXAMPLE GAS DISPERSION CALCULATIONS FOR ENVIRONMENTAL APPLICATIONS USING ADMS 11.1 INTRODUCTION

The purpose of this Part of the Guide is to provide guidance to inexperienced environmental gas dispersion modelers on how to apply the ADMS (Atmospheric Dispersion Modeling Software) package to practical gas dispersion modeling problems. The following examples are based upon real applications found in the chemical industry, although the emission rates are not actual figures from any particular plant. From past experience, the greatest misunderstandings in gas dispersion modeling occur over the criteria for determining whether or not an emission is acceptable. The examples shown here represent typical assessment methods and criteria used by gas dispersion practitioners in environmental consultancies and industry. However, these examples are not meant to provide a definitive statement on how to decide whether an emission is acceptable. The following examples have been carried out using ADMS.



11.2 SOURCE DATA It is very rare for the source data used in a dispersion study to have been measured or estimated by the person actually having to carry out the study. Almost always the modeler is reliant on source data which has been measured in a rush on the plant. Often it is far from comprehensive and few data sets are error free. A morning spent scrutinizing the source data or an hour spent talking to the technician or scientist who has measured the data can often save a week of analysis. The most common source data errors are caused by: (a) Source concentration data being based upon too small a number of measurements. The following is a real example:- To estimate an annual emission rate of a Volatile Organic Compound (VOC) from a process manufacturing an additive used in the plastics industry, one vent was sampled continuously for around 15 minutes once a day for a week. The concentration figures were 8.0 ppm, 7.3 ppm, 7.0 ppm, 112.5 ppm and 6.8 ppm. The inexperienced scientist would either have worked out an average figure of 28.3 ppm or assumed that the figure of 112.5 ppm was erroneous. When this data was discussed with the plant’s process engineer, it was discovered that when the process is cleaned between production runs - a 20 minute operation, once a fortnight - VOC emission rates over 100 ppm could occur. Thus for estimating an annual average, a concentration figure of 7.3 ppm would probably be an appropriate figure to use rather than 28.3 ppm. However, when assessing the maximum concentration likely over a period of 15 minutes (e.g. for an occupational health assessment), the appropriate concentration to use is the 112 ppm figure. (In practice for this study, plant technicians subsequently took emission rate measurements every minute during the cleaning phase to quantify the rate more accurately). (b) Plant staff providing absolute worst case discharge conditions and assuming that these source conditions occur all year - important for annual average calculations.



(c) Upgrades of process plant, which can lead to emission rates significantly higher or lower than the original plant design. For example, reducing condenser temperatures usually reduces emission rates from processes discharging VOCs. Dispersion modelers should be very wary of theoretical source data based upon design information more than a few years old. (d) Missing out significant emissions - especially from low stacks. Relatively small emissions from low stacks can make significant contributions to ground-level concentrations. (e) Failing to adequately take into account the effect of building wakes. Detailed information on buildings is vital for any calculations carried out using ADMS. For a given stack height, the presence of a nearby building can lead to much higher ground-level concentrations. (f) Using emissions data based on computer simulations of processes. Sometimes the predictions of emission rates from the simulations are based upon incorrect assumptions, which can lead to serious errors. 11.3 EXAMPLE CALCULATIONS 11.3.1 Example One - Continuous Emissions A small boiler house has a 40 m high, 1 m diameter stack and is located in the middle of a chemical complex more than 5 km away from any major town. It discharges 20 m3/sec of gaseous combustion products at a temperature of 60°C; a molecular weight of 29.6 and a specific heat capacity of 1000 J/kgK. The initial concentration of sulfur dioxide discharged in the combustion products is 50 mg/m3 (based upon a reference temperature of 60°C). There is a 70 m high plant structure, which does not have any cladding, 200 m away from the source. Determine whether the emissions from this stack are acceptable.



Inputting the source data:- (a) Set up screen: (1) For this location, an appropriate roughness length to use would be around 0.3 m, although if the plants were densely packed with high structures, a roughness length as high as 1 m could be appropriate. (2) The latitude is required in order to estimate the incident radiation on the ground, which is then used to calculate the heat transfer rate from the ground back into the atmosphere. For a given wind speed, the higher the heat transfer rate, the more unstable is the atmosphere. (3) The UK air quality standard for sulfur dioxide is based upon an averaging time of 15 minutes.

(4) The mass units of emission should be in the same form as required in the output. Thus if an output of micrograms per cubic metre are required, then data input here should indicate micrograms (in this case put as ug).



(b) Source data:

The data input to the table is not identical to that of most spreadsheets. Clicking once on a cell in the table highlights the existing data in blue and it is then possible for the value to be overwritten. Clicking twice on the cell enables the value to be edited. Ignore the create groups option for single stacks. There are six additional columns not shown in the above image - they can be accessed by moving the bottom slider to the left. The data in the first three columns (gas molecular weight, specific heat and density) only need to be changed if the bulk properties of the gas cloud discharged are different from that of air. In this case, by calculation, the molecular weight was calculated to be 29.6 but the specific heat capacity was identical to that of air. The exit velocity (Exit V) or volume discharge rate (Volume) option toggles by clicking on the appropriate cell. Note also the non-standard definition of NTP (Column 5) of 1 bar, 25 deg C.



Having created the table for the bulk properties of the gas flow from the stack, to specify the mass flow rate of the toxic component it is necessary to click on ’Pollutants’. It is very easy to forget to do this next and thus not enter the toxic gas discharge rate. However, if this section is left blank, an error message is flagged up. An emission rate has to be specified before the problem can run.

To input the pollutant discharge rate, it is necessary to click on ’New’ and then specify a pollutant name. Note that you can specify more than one pollutant from a single stack. The discharge rate of the toxic component must be inputted in the same format as the set up screen i.e. for this example, in micrograms/second rather than g/second. (1 g/s = 1,000,000 mg/s). If the user requires an output in terms of ppb rather than mg/m3, then an appropriate units conversion factor can be selected in the options screen (see the Part 4 of this Guide for guidance on converting between units). The boxes to the right of the discharge rate only need to be changed if modeling gaseous or particulate deposition.



(c) Meteorological data: The meteorological data used by ADMS can either be read from a file or directly inputted into a table that can be accessed using the ’by hand’ option. The file R91A-G.met gives a range of typical wind and weather conditions and an amended version is detailed below:-

There are 4 columns of data. The first is the wind speed (m/s) at a height of 10 m, the second is the reciprocal of the Monin-Obukhov length which characterizes the atmospheric stability. A negative Monin-Obukhov length implies unstable atmospheric conditions; a positive value represents stable atmospheric conditions. The reciprocal is used because the Monin-Obukhov length has an infinite value in neutral atmospheric conditions.



The third column specifies the direction from which the wind is blowing - a wind direction of 270 degrees aligns the plume centerline parallel with the x-axis. The fourth column is the height of mixing layer in meters - effectively above this height the wind speed remains constant. This factor is important in defining the wind speed versus height profile. For each line of metrological data, a separate results file is produced. In this case, concentration data can be found in files prob1.c01 to prob1.c08.



Knowing the minimum and maximum co-ordinates of the area over which concentration data is required, then the grid spacing in the x-direction (Dx) is (xmax - xmin) / (nx-1) where nx is the number of points in the x-direction. Correspondingly the grid spacing in the y-direction (Dy) is (ymax - ymin)/ (ny-1). As a rule of thumb, for a single stack the grid spacing in the downwind direction should be roughly equal to the stack height, with the grid spacing perpendicular to the wind direction being equal to half the stack height. However, the grid can be chosen to cover whatever area the modeler wishes. There is no point in outputting data upwind of the stack - concentrations will be zero everywhere. To resolve as much of the detailed concentration of the plume it is generally advisable to use a 31 x 31 grid. For a symmetrical grid either side of the plume centre-line, it is necessary to choose an odd number of grid points perpendicular to the wind direction to ensure that concentration predictions are made along the plume centerline. Let us now consider what grid is required for obtaining a ground-level concentration footprint plot for emissions from this 40 m power plant stack. For a single stack assessment, it is usual to assign the stack a position x=0, y=0. For a given atmospheric stability and wind speed, the actual value of the peak concentration is independent of the wind direction, so for convenience we have chosen the wind direction to be from the west. (This is specified in the wind and weather file (*.MET) or via the direct met data input screen). The plume centerline will then run along the x axis. The likely peak in ground-level concentrations will be around 3-10 stack heights from the source for unstable and neutral atmospheric conditions. Using the rule of thumb of choosing a grid spacing equal to the stack height then with 31 grid points, the concentration predictions would be made up to a downwind distance of 1200 m. The footprint plots for a single stack are always oval in shape with the long axis being in the downwind direction. Using the rule of thumb of choosing half the stack height in a direction perpendicular to the wind direction, then the maximum cross wind extent would be 300 m either side of the plume centre-line, if 31 grid points are chosen. Thus the initial points for the grid specification should be 0 m in the x-direction, -300 m in the y-direction and 0 m in the z-direction. The grid spacing's should be 40 m in the downwind direction and 20 m in the crosswind direction. 31 grid points should be chosen in the x and y-directions.



It should be noted that any grid arrangement could have been chosen within a 32 x 32 grid points limit. If there was a particularly sensitive ’target’ for the pollution, such as a hospital, say 2 km from the source, then an 80 m grid spacing could be selected in the x-direction, generating concentration predictions up to 2.4 km from the source. If a finer resolution of the peak in concentration was required, then the grid spacing in the x-direction could be reduced to 20 m. Note that concentration data can only be outputted at two heights - it is not possible to produce a vertical concentration profile. It is not unusual to get the grid spacing wrong, especially with multiple stacks or with wind directions off-axis. With short calculation times, a trial and error method for choosing the grid is sometimes the best way of quickly choosing the best grid. However, for calculations which can take many hours, such as annual average calculations around buildings or for percentiles, then it is important to get the grid right first time.

(e) Options: For a single stack dispersion calculation using hourly meteorological data, it is necessary to select ’Source’. For an annual average single stack problem or for multiple source long- term and short-term calculations, it is necessary to select ’Groups’.



(f) Running the problem: Having selected ’Run’, ADMS automatically prompts the user to specify a file name for the problem. The problem file has an extension .apl - in this case prob1.apl is the name of the problem file. All the output files have the name prob1 and can be found in the same directory as the prob1.apl file. If the problem is completed successfully, the message ’program terminated with exit code 0’ is generated at the end of the run - if unsuccessful then the run will be ’terminated with exit code 1’ as well as producing other error messages. To keep all the input and output files in order, it is advisable to save each problem in a separate directory.



(2) Line plotting It is possible to directly plot the maximum ground-level concentrations using the x-y plotting facilities under ’Results’. Usually, the gas dispersion modeler wants to superimpose plots to determine the effect of wind and weather conditions on ground-level concentration. For this example, we want to determine the maximum ground-level concentrations for 2 m/s unstable, 5 m/s slightly unstable and 5 m/s neutral atmospheric conditions. The numerical output data files for these atmospheric conditions are prob1.c02, prob1.c03 and prob1.c04, respectively. ADMS uses the equivalent data files prob1.!02, prob1.!03 and prob1.!04 for x-y plotting. To highlight more than one file, it is necessary to hold the "control" button down whilst clicking on the left mouse button. Using the x-y plotting facility it is also possible to plot the plume centerline concentration - this would be the highest concentration found anywhere in the plume. This is of use in chemical plants and urban areas where there could be plant structures or buildings higher than the vent. The plume centerline concentration would be the highest concentration found anywhere on these structures above ground level.



The highest ground-level concentration occurs for the 2 m/s unstable case with a peak ground level concentration of 14 micrograms/m3 occurring at 160 m from the source. The exact value can be found directly from the xy graph, by positioning the mouse pointer over the data point of interest and then pressing the left mouse button. The values appear in the bottom right hand side of the output. Note that it is possible to label the axes of the graph (click on ’Graph Setup’ for options), and the legend (click on ’Graph Setup’ - ’Data’ - ’Legend Text’). Normally for a single stack emitting a plume that is not highly buoyant or has an exit velocity less than say 50 m/s, the highest ground-level concentrations occur in unstable atmospheric conditions. Usually the most unstable case of 1 m/s, reciprocal of the Monin-Obukhov length = -0.5 /m is ignored for design, since in the UK, these conditions (equivalent to Pasquill-Gifford stability class A) usually occur for less than 1% of the time.



The UK Air Quality Standard for sulfur dioxide is 100 ppb by volume which is equivalent to 266 mg/m3 at 20°C. These concentrations are based upon a 15 minute averaging time. The peak concentration under unstable conditions (14 mg/m3) would be around 5% of the air quality standard, which, for an existing installation, is usually acceptable to regulators. However, there are no clear cut recommendations for an acceptable ground-level concentration from a single stack, except that according to the Environment Agency's Guidance for Operators of IPC Processes, 'Best Practicable Environmental Option Assessments for IPC', the sum of the background concentration and predicted environmental concentration arising from the emission must not exceed 80% of the air quality standard. If concentrations arising from the emission are 2% or less than the air quality standard, then the emission can be assumed to not have a significant impact on air quality. In practice, for the design of a new power plant stack, most dispersion practitioners would regard concentrations as high as 10% of air quality standard as acceptable, and concentrations higher than 25% of the standard as unacceptable (see Part 2).

(3) Contour plots To show the concentration distribution at ground level, isopleths (concentration contours) can be plotted using the package SURFER (Figure 11.2). SURFER is a very user-friendly package - to alter any plot, place the mouse pointer over the part of the plot to be altered and then double click. This brings up a window that, for example, enables the contour values, labeling or titles to be changed. Often, difficulties occur when the labels on the contours overlap. Sometimes this can be cured by reducing the Font Size of the text. It is possible to import a digitized site plan or Ordnance Survey Map as a base map. There is a choice of being able to fill concentration contours with colored areas. There is also an option which makes the colored areas opaque, so that the detail of the map beneath is not obscured.



The highest concentration on the nearby plant structure would occur when the plume could continuously impact directly on the building. The worst case atmospheric conditions would be stable conditions where there is very little mixing in the vertical direction and the plume centerline concentrations are high. The first fact to determine is whether the plume could directly and continuously impact the building. This can be determined by plotting out the plume centerline height which is one of the options which can be selected from the line plotting definitions screen. There are two options for centerline height - the ’plume centerline’ is calculated by carrying out a momentum balance to predict the plume rise above the source. The ’mean height’ is calculated by working out the centre of mass from the concentration calculations. For, a near field calculation, the former is the most appropriate to use, since this represents the actual trajectory of the plume. Figure 11.3 shows the predicted centre-line height of the plumes. As can be seen, the plume centerline height for all weather conditions (except unstable - the !02 plot (filled triangles)) is less than 70 m above the ground - thus the plume could make a direct impact against the top of the plant structure.



Worst case conditions for direct plume impact occur in stable atmospheric conditions (Figure 11.4). The plume centre-line concentration is predicted to be around 265 mg/m3, 200 m from the source. At this point, it should be noted that the UK Air Quality Guidelines do not apply within industrial sites or any premises where access to the general public is not permitted. Within site boundaries, the air quality criterion to use are the occupational exposure limits which for sulfur dioxide are 5000 mg/m3 over 8 hours, 13,000 mg/m3 over 15 minutes. (The HSE’s EH40 document provides information on occupational exposure limits). Thus emissions from the stack are acceptable.

11.3.2 Example Two - Multiple Stack Calculation A plant producing plastic chips thermally oxidizes volatile organic compounds in a tail gas burner and discharges 0.5 g/s of oxides of nitrogen from a 40 m high stack. It is planned to build a second larger plant with a stack 300 m south east of first stack. The discharge rate would be 0.75 g/s. At what height should this new stack be constructed? The discharge temperature of both stacks is 50 deg C, the bulk properties of the emissions are assumed to be similar to that of air and the exit velocity of the emissions from the stacks are designed to be 15 m/s. The diameter of the existing plant stack is 1 m, for the new stack 1.25 m. The highest concentrations at any point from the stacks would occur when the wind flow is along a line joining the two stacks (i.e. when the wind is blowing from the North West or South East). When this occurs, the two concentration footprints overlap to give the highest ground-level concentrations. If there were three or more stacks then it would be necessary to carry out preliminary dispersion calculations for a range of wind directions, say 12 wind directions, equally spaced 30 degrees apart. This, then, would enable the likely worst case wind direction to be determined. Always remember that the wind can blow in any direction, not just the prevailing wind direction. In this particular example, to the north west of the site there is a very wide estuary and there are no houses or schools etc. within 3 kilometers of the source. South east of the site is a residential area extending to within 100 m of the site boundary. Thus, it is only necessary to consider the area to the south east of the source.



When designing stack heights using ADMS, it is not possible to select an option that will directly calculate a suitable stack height - it is necessary for the dispersion modeler to use a mixture of trial and error (i.e. putting in a range of stack heights) combined with educated guesswork. Inputting the source data:- (a) Set up screen: The averaging time selected should be one hour. The air quality standard for nitrogen dioxide is 150 ppb (282 mg/m3 ) as a one hourly mean and 21 ppb (40 mg/m3 ) as an annual average. Again, the air quality standard is expressed as mg/m3, thus the mass unit of emission has been selected to be ug.



The discharge rate of NOx from the new plant is 50% higher than the existing plant, hence, as a first guess it is likely that a higher stack would be required. Since the first stack was specified at 40 m, as a first guess, a stack height of 55 m was selected Note that the table does not behave in the same way as a spreadsheet. Having inputted the data for one source, it is necessary to click on new to generate a second source. A default name ’Source xxx’ is produced automatically and can be edited as required. The ’Create groups’ option has not been selected in this analysis. This option is useful when analyzing the emissions from many sources on a site. It is possible to group together the emissions from individual plants, and to assess both the impact of the site as a whole as well as the emissions from a particular group of sources (e.g. those associated with a particular process).

(c) Pollutants: Again the discharge rate of the NOx from the stack needs to be specified in terms of micrograms/sec. If the same pollutant is discharged from more than one stack, then it is vital to use the same name for the pollutant. If the pollutant is called NOx from one stack and ’Oxides of nitrogen’ from the other, ADMS would not automatically detect that the pollutant is the same species.

(d) Meteorological data



For this study, the meteorological data has been input ’by hand’. Note that unlike most gas dispersion models, ADMS (and other recently developed gas dispersion models) does not require the input of Pasquill-Gifford stability categories, but calculates more accurately the properties of the atmospheric boundary layer from data such as, the boundary layer height, cloud cover, time and day of the year. It is also possible to input directly the surface heat flux - typical values can be found in ’Atmospheric Diffusion’ by F Pasquill and F B Smith, Ellis Horwood Press (2nd Edition - 1983). It can also be measured directly. The first line of data represents unstable conditions found on a hot sunny afternoon on the 1st August and the second line of data represents typical windy conditions on the 1st of October. It should be noted that the boundary layer height does not remain constant over a whole day. During a sunny summer’s day, the boundary layer may increase from 100 m to 1500 m over a period of eight hours from daybreak, remain at a roughly constant level during the afternoon, collapsing rapidly back down to 100 m over an hour or two as the sun sets. The lateral spread sq can be found from measurements using equipment such as sonic anemometers - this option is rarely used. It should be noted that the value of the mean speed varies with height - most standard meteorological measurements of wind speed are taken at 10 m. If real measured wind data is used, then the actual anemometer height must be specified.

(e) Grid definition: It is always good practice to have as many grid points as possible within the cloud footprint. This then enables the contour plotting package SURFER to resolve the concentration profile as accurately as possible. From the experience of many ADMS users, for multiple sources, it is quite often necessary to repeat calculations, because the first attempt at guessing the grid was incorrect. This can be very frustrating if the run times are several hours or more. Using a 21 x 21 grid rather than 31 x 31 can significantly speed up run times at the expense of the detailed resolution of the concentration footprint. The following grid gave a good resolution of the concentration footprint:-



(f) Options:

For this application it is necessary to select the ’Groups’ option and then double click on the box to the right of ’All sources’ - a red ’tick’ should be produced. Note that it is possible to output the concentration from each of individual sources.

(g) Running the problem: (h) Results:

There is no clear well-established methodology for designing stacks specifically for the oxides of nitrogen. From industrial combustion processes, most of the NOx is nitric oxide (NO) which has a 15 minute occupational exposure limit five times higher (i.e. five times less toxic) than that for nitrogen dioxide. Therefore, a conservative approach would be to assume that all of the emissions are nitrogen dioxide. Although widely used, this is probably unnecessarily conservative - the oxidation process of nitric oxide to nitrogen dioxide is slow, and by the time the plume reaches the ground, only a small proportion would have oxidized. Typical nitric oxide to total NOx measurements can be found at the UK Air Quality Archive internet website. The most useful summary tables can be found at:-www.aeat.co.uk/products/centres/netcen/aqarchive/auto.html One method for predicting the levels of nitrogen dioxide would be to scale the output concentration predicted by the model by the ratio of the nitrogen dioxide concentrations to total oxides of nitrogen measured at an appropriate monitoring site. An alternative method and more detailed discussion is given in Part 8 of this Guide. Figures 11.5 and 11.6 show the predicted concentration footprint for the new stack with a height of 55 m



The square region in the centre of the plots represents the site boundary. Calculations for 2 m/s unstable atmospheric conditions and 5 m/s neutral atmospheric conditions are shown. The hourly average NOx concentration beyond the site boundary is predicted to be around 4 mg/m3 - below 2% of the air quality standard and hence not a cause for concern. However, what is a cause for the concern would be the difficulty in persuading the local authorities to give planning permission for a new 55 m stack so close to a residential area. (A 55 m stack would be the highest point on the site). The calculations were repeated with both stacks at a height of 40 m. As can be seen in Figure 11.7 in unstable atmospheric conditions, reducing the stack height increases the peak value of ground-level concentration, but has very little effect on the concentrations beyond the site boundary. In neutral atmospheric conditions, the peak in concentration is close to the site boundary but concentrations are still below 2% of the air quality standard.



Thus a 40 m high stack is acceptable for the new plant. In fact, there could be scope to reduce the stack further - however, in this particular example, there was a redundant plant structure up to a height of 35 m to which the vent could be attached. Often, in the construction of stacks above 25 m, the major cost of a stack is its support structure.

Annual average calculations To calculate annual average concentrations, a new APL file should be created by editing the existing file. It is necessary also to select a file of statistical weather data most appropriate for the site (for this calculation Prestwick Airport); to change the grid so that contour plots are outputted 360 degrees around the source and on the ’Options’ screen, it is necessary to select the Long Term Average option. The annual average problem file must be saved with a different name to that of the short term calculation, otherwise existing output files will be overwritten.

When editing an APL file to produce annual average calculations, usually the grid has to be changed. This is necessary to take into account the variation in wind direction throughout the year. For this example, a square 3 km x 3 km grid centered on the existing plant stack has been specified using an initial point of -1500 m, a grid spacing of 100 m and 31 grid points in both the x- and y- directions.



In the met screen, the user must choose the sector size to cope with the particular format of the statistical weather data. The wind direction, always defined to be the direction from which the wind is blowing, may be categorized into 36 groupings (10 degree sectors) or 12 groupings (30 degree sectors). Usually, statistical met data from the UK Meteorological Office is based upon a 30 degree sector size. The contour plot is produced using the ADMS contour plotting routine selecting the .C01 file which for annual average calculations always contains the concentration data. From Figure 11.9, the maximum annual average concentration of the oxides of nitrogen is around 0.40 mg/m3 (i.e. 1% of the annual air quality standard for nitrogen dioxide which is 40 mg/m3). This concentration level is acceptable. Thus the new stack should have a height of 40 m.



11.3.3 Example Three - Odor Dispersion Calculation

As part of the development of a new paints process, a pilot plant design requires the venting of 12 mg/s of ethyl acrylate from a 15 m high stack. The ethyl acrylate is part of a bulk flow from a 10 cm diameter stack with ambient temperature and properties identical to that of air. The plant is located 1 km from a large city centre with a mixture of light industrial and domestic housing around it. The site boundary is 100 m from the source and there is a large hospital 500 m away. Determine whether the emissions could cause an odor nuisance. Ethyl Acrylate is a highly odorous material with a very low odor threshold of 0.082 µg/m3. ( Reference :- ’Odor measurement and control - An update’ by M Woodfield and D Hall AEA Report CS/REMA-038). Very low concentrations could produce an odor nuisance.

The human nose can pick up peaks in concentration over a few seconds - thus to assess the likely scale of an odor nuisance, it is necessary to predict the magnitude of these peaks.



In practice, it is possible for the predicted mean concentrations to be below the odor threshold, but because the peak concentration can be 10-50 times the mean concentration then the emission could still cause an odor nuisance. By using the fluctuations options in ADMS, the magnitude of the peaks can be estimated. There are two ways of presenting fluctuations data for given wind and weather conditions either: (1) Choosing a percentile level e.g. 98th. The output provided is a concentration which for a 98th percentile would only be exceeded for 2% of the time. (2) Probability density function. A series of concentrations are inputted and the output takes the form of corresponding percentile figures. From practical experience, if the mean concentration is less than 1/40th of the odor threshold, or if the 98th percentile of concentration is not significantly greater than the odor threshold, then it is unlikely that the emissions would cause an odor nuisance. If the mean concentration is greater than the odor threshold, then an odor nuisance is very likely.

Inputting the source data:- (a) Set up screen: A location close to a city centre implies a high roughness length - a figure of 1 m has been selected. The fluctuations option has to be selected under the ’other model options’ heading. Although, we are interested in the likely peak concentration over a few seconds, the emission could be continuous over a period of an hour or more and so for the mean concentration predictions, the appropriate averaging time to use is one hour. The fluctuations averaging time is selected on the fluctuations screen.



(b) Source data:

The source data is easy to input - note that the user can choose either a stack exit velocity (m/s) or a discharge rate (m3/s) by toggling the two options available in the cell directly below the ’Vel or Vol’ heading in the Source table.

Concentrations predictions will be made for the standard range of wind and weather conditions found in the r91a-g.met file.



The fluctuations options are selected via the fluctuations screen. A fluctuations averaging time of 5 seconds is representative of the time taken for someone to breathe ’in’ and ’out’. The 98th percentile concentration is specified by entering ’98’ in the appropriate small box and then pressing the spacebar.

(c) Grid definition: The wind direction for the standard wind conditions in the R91a- g.MET file is from the west - hence the plume centerline will be parallel to the x direction. The grid is selected to have a maximum downwind extent of 600 m - so that the hospital is included within the plume footprint. Thus with a maximum number points of 31, a grid spacing of 20 in the x direction is required.

The width of the plume will be less than half its length, so that the grid has been selected as stretching from y = -150 m to +150 m - 31 points with a spacing of 10 m. The concentrations predicted will be independent of wind direction.



(d) Results: Surprisingly, when the fluctuation option is selected, ADMS does not produce mean concentration output files which can be plotted using the X-Y plotting option. Thus it is necessary first to run the problem without the fluctuations options selected in the set up screen and then to re-run it using the fluctuations option.

Figure 11.10 shows the mean ground-level concentration. At distances greater than 100 m from the source, concentrations are predicted to be as high as 1 mg/m3 - 12 times greater than the odor threshold. Note that the .!02 file represents 2 m/s highly unstable conditions, the !03 file represents 5 m/s slightly unstable conditions, the !04 file represents 5 m/s neutral conditions and the !08 file represents 10 m/s neutral conditions. Thus for this example, it is very likely that there will be an odor nuisance under most atmospheric conditions. Figures 11.11 and 11.12 show the predicted 98th percentile of short term fluctuations for the 5 m/s neutral stability and 2 m/s B-stability cases.



As can be seen, for these wind conditions, the 98th percentile of concentration is predicted to be over 5 mg/m3 beyond the site boundary (i.e. around 60 times the odor threshold). To reduce the 98th percentile of short term concentrations to around the odor threshold, the discharge rate of ethyl acrylate would have to be reduced by a factor of 60 - a demanding requirement.

Figure 11.13 shows the likely 98th percentile of short term concentrations for the same discharge rate from a 50m high stack. As can be seen, raising the stack has virtually no impact on reducing the concentration beyond the site boundary. For a 15 m stack the peak in ground-level concentrations occurs within the site boundary, for a 50 m stack, it is beyond the site boundary.



Thus for this application, the discharge of ethyl acrylate at a rate of 12 mg/s is unacceptable. Process changes are required to reduce the emission rate by roughly a factor of 60. Raising the stack height will have little impact on reducing the likely odor impact.



11.3.4 Example Four - Dispersion Around a Building

As part of a sulfuric acid recycling process, 0.1 g/s of sulfur dioxide in a larger flow of air is discharged from a vent, located centrally 1.5 m above the roof of a rectangular building 60 m long x 20 m wide x 25 m high. The vent diameter is 0.2 m, the air flow rate is 0.16 m3/s and there is a Chinaman’s hat on top of the vent to keep the rain out. The discharge temperature is ambient. Determine the maximum concentrations that could occur - the long side of the building runs north-south and parallel to the nearest site boundary 80 m away. For these calculations it is necessary to use the buildings option of ADMS. For a rectangular building, the highest concentration usually occurs with the wind perpendicular to the longest side of the building. However, this will be checked below. A file ROSE5D.met has been created to represent 5 m/s neutral conditions for a range of wind directions from the south clockwise to the north. Thus the Met data for this problem is ’from file’.

Inputting the source data:- (a) Set up screen: It is necessary to select ’buildings’ under the complex effects option. Remember to choose a 15 minute averaging time, because of the 15 minute SO2 air quality standard of 266 µg/m3.



(b) Source data:

The effect of the Chinaman’s hat is to remove any exit momentum in the plume. Hence, even though there is a considerable air flow out of the vent, the exit velocity has been set to zero.

The stack height is defined to be the height of the stack tip above the ground rather than its height above the roof of the building.



(c) Buildings option: The buildings option of ADMS only models dispersion around a single building. It is, however, possible to input the dimensions of a number of buildings. ADMS would then calculate the dimensions of the single building that best represents the dimensions of the buildings taken in combination. The point (x,y) defines the position of the centre of the building relative to the same origin in the source menu. The ’angle’ is the angle the longest side makes with north. Thus, if the longest side is parallel to the north-south direction, angle = 0 degrees, if it is parallel with the east-west direction, the angle is 90 degrees. ADMS can only model dispersion around a rectangular building - there are not separate models for say storage tanks, or buildings with an apex roof.

Figure 11.14 shows the effect of wind direction on the predicted maximum ground level concentration. If building wakes are significant then the model assumes that within the wake concentrations are constant.



The highest concentrations are produced by the !04 file which corresponds to the wind direction being from the west. Thus, for this study to assess the effect of the wind and weather conditions, only calculations with the worst case wind direction (270 degrees) will be carried out.

Note that the concentration is constant in the near building wake region, followed by a peak in concentration due to dispersion from a strong ground-level source. The calculation was then repeated, again using the file R91a-g.met which represents a range of wind and weather conditions. For a stack where building effects are unimportant, it is only necessary to consider unstable and neutral conditions when assessing the likely highest ground-level concentrations. The buildings effect module of ADMS represents the building wake as creating two sources, one at ground level, one at the stack height. The percentage of the emission from the ground-level source is dependent on the proportion of the emission which enters the wake. If the ground-level source is significant, high ground level concentrations can be predicted in stable atmospheric conditions.



Thus it is necessary to consider the full range of wind and weather conditions (except for the highly unstable case *.!01 file and the highly stable case *.!07 file which occur very infrequently in the UK). See Figure 11.15.

Compared with the 15 minute air quality standard for SO2 of 266 mg/m3, the concentrations predicted are too high - remedial action is necessary. To reduce the impact of the building wake, it is necessary to raise the vent above the roof - for a large building, it is good practice to have a minimum clearance of 5 m above roof level. For this example, a height of 5 m above roof level is the highest possible without construction costs becoming prohibitive. Additionally, redesign of the vent to eliminate the Chinaman’s hat would give the plume an increased exit momentum. This which would reduce the proportion of the plume entering the building wake leading to lower ground-level concentrations.



Figure 11.16 shows the predicted ground-level concentration for a 30 m high stack with the Chinaman’s hat removed. As is immediately evident, the effect of raising the stack is to significantly reduce the predicted ground-level concentration beyond the site boundary to less than 3% of the air quality standard, which is acceptably low. Within the site boundary concentrations are less than 1% of the 8 hour occupational exposure standard of 5000 mg/m3.

11.3.5 Example Five - Annual Average Statistical Calculation for an Area Source As part of the planned up rate of a process manufacturing solvents, one simple option would be to discharge the waste liquid from the process into an existing small lagoon. This has dimensions of 70 m x 50 m with the longer side running in an east-west direction. The average emission rate of benzene from the lagoon is estimated to be 0.5 g/s - determine whether this emission is acceptable.



The site is located away from any major towns but the nearest housing is 300 m to the north of the lagoon. Other sources of benzene from the site are negligible. The air quality standard for benzene is 5 ppb (16 mg/m3) expressed as a running annual mean. This problem requires an annual average area source calculation with the source at ground level. Inputting the source data:- (a) Set up screen: The appropriate figure for the roughness length is 0.3 m. The annual average calculations give averages of modeled hourly mean concentrations - thus the appropriate averaging time to use is 1 hour. (b) Source data:

The source data to be input into this table is not straightforward. Firstly, the choice of the molecular weight of the gas flow should be selected to be the molecular weight of air instead of the molecular weight of benzene. The small quantity of benzene evaporating will dilute so rapidly with the wind flow above the lagoon, that the gas will behave like a passive neutral density gas rather than as a heavier than air gas plume. (If however there was a huge spillage of benzene into the lagoon and the emission rate was tens or hundreds of grams per second, then the concentration of benzene vapor above the lagoon could be sufficiently high that the plume would demonstrate dense gas effects - ADMS cannot cope with dense gas behavior). The ’area source’ option is selected from the table of source data. By clicking on the cell directly below ’Source Option’ a menu is selected with options ’P’; ’A’; ’V’; or ’L’. ’P’ is the default (Point source), ’V’ selects a volume source and ’L’ selects a line source. ’A’ selects the area source option.



The area source is defined from the co-ordinates of the bottom left and top right corners of the rectangular area.

The emission rate per square meter of source has to be inputted rather than the total emission from the area source. Thus if 0.5 g/s of benzene is being discharged from the 70 m x 50 m area, this is equal to 142.9 mg/m2s.



(c) Meteorological data: Statistical weather data based upon 10 years of information recorded at a European airport has been used for this analysis.

(d) Options: It is necessary to select the ’Long Term Average’ option as well as choosing Groups with a red arrow alongside the ’All sources’ cell.

(e) Grid definition: The nearest residential area is 300 m from the source - this governs the size of the grid. An area stretching up to 450 m from the source has been selected.

(f) Running the problem: (g) Results: Plotting the annual average concentrations (Figure 11.17), it is evident that the predicted annual average benzene concentration over the nearest housing is close to 80% of the benzene air quality standard. Bearing in mind the uncertainty of any gas dispersion model predictions, the uncertainty in the evaporation rate calculations, and the World Health Organization’s long term aim to reduce benzene air quality standards from 5 ppb to 1 ppb imply that the proposal to send the waste stream to the lagoon is unacceptable.



11.3.6 Example Six - Dispersion of Particulates from a Prilling Tower The final stage of the production of solid fertilizer involves spraying hot liquid downwards from the top of a prilling tower into a vertical flow of air. This air flow helps to cool and solidify the droplets. This process has been found to be the best way of producing uniform spherical granules that enable farmers to spread fertilizer uniformly over the ground.



An air flow of 350,000 m3/hr at 20 deg C is discharged from the tower having passed through a filtration system. The particulate concentration and size distribution has been estimated to be:-

6 mg/m3 of particulates with a particle size greater than 10 microns with a mass weighted mean diameter of 25 microns 2 mg/m3 of particulates with a particle size in the range 5-10 microns with a mean diameter of 7 microns 1 mg/m3 of particulates with a particle size in the range 1-5 microns with a mean diameter of 3 microns.

The vent diameter is 5 m and the density of the particulates is 1050 kg/m3. Determine whether these emissions are acceptable. (The site boundary is 50 m from the base of the prilling tower which is located in a chemical works in the outer suburbs of a town). Also estimate the annual average rate of particulate deposition. The air quality standard for particulates is 50 mg/m3 for fine particulates, 10 microns or below, based upon a 24 hour averaging time. Particulates larger than 10 microns in diameter are not respirable and thus do not present a cause for concern. (However, it should be noted that emissions of particulates should not have a significant visible impact). Thus, for carrying out these dispersion calculations, the dispersion of particulates greater than 10 microns in diameter can be ignored.



The tower has a circular cross-section and a height to diameter ratio of 7. For this analysis, it is assumed that the wake effects of the tower are negligible - visual observation of the plume shows that assumptions of a uniform concentration in the wake of the tower would be inappropriate. The details of the particulate emission are specified in the ’Pollutants’ screen.



Note that it is not possible to directly input a particulate concentration. The total mass of the particulate has to be calculated and the appropriate mass fraction selected. For this analysis only dry deposition has been modeled. If the terminal or deposition velocities of the particulates are known then these figures can be inputted directly. In practice, at best, the particulate size distribution and density are known and ADMS will calculate the appropriate deposition velocity assuming the particulates are spheres. To obtain deposition rates the appropriate wet/dry/total deposition options need to be activated in the options screen (for further information see Part 10). (However, if the particulates are highly non-spherical, e.g. they are platelets or agglomerates, then it may be necessary to measure deposition velocities directly).



The concentrations of particulates plotted above are the sum of the concentrations for each of the two particulate size ranges. The highest concentrations (as expected) occur in unstable atmospheric conditions (!02 file). The predicted maximum ground-level concentration is just over 1.0 micrograms per cubic meter (i.e. concentrations are around 2% of the air quality standard and are thus acceptable). The annual average deposition flux enables the total quantity of material discharged over a year to be calculated. For the following analysis, the wet deposition rate of material is included with the dry deposition.



Figure 11.19 shows the predicted total annual average deposition rate. The highest deposition rate beyond the site boundary is predicted to be around 0.0025 mg/m2/s. Thus, using this figure, over a whole year, 0.08 g of particulate would be deposited per square meter. Once every six months, the filtration system needs to be changed, which leads to a twenty fold increase in particulate emissions below 10 microns for around 24 hours. What would be the likely maximum 24 hour mean concentration? For the analysis shown in Figure 11.18, the worst case conditions occur in highly unstable atmospheric conditions. However, such conditions only occur for a few hours in the middle of day.



Therefore 24 hourly mean concentration based upon highly unstable atmospheric conditions would overestimate concentrations. The standard ADMS file oneday.met is a sequential list of 24 sets of weather conditions representing a day when convective conditions occur. It is reproduced below:-

This is a true representation of a hot sunny day in the UK with a constant cloud cover of 2 oktas (eighths), no rainfall and a constant wind direction from the west. To model this situation, it is necessary to change the averaging time to one hour, multiply the discharge rate by twenty, activate the sequential calculation box in the met screen and run a long term average calculation, remembering to highlight the ’Groups - all sources’ option.



As can be seen, the maximum hourly average concentrations with the filtration system out of use is around 20 micrograms/m3. Although 40% of the air quality standard, these emissions would be acceptable for a day or two, but unacceptable for continuous operation.

11.4 ACCURACY OF ADMS-2 ADMS-2 has been extensively validated against the limited number of datasets available. Most of the data available are studies of emissions from power stations - thus the data available is inevitably for stacks of a height of 100 m or more. For such stacks, heat transfer to and from the ground into the atmosphere has a greater effect on the dispersion processes than for stacks, say 25 m or less, where mechanical turbulence generated by the roughness of the ground predominates. For elevated stacks, versions of ADMS before v2.11 were shown to overestimate concentrations in unstable atmospheric conditions for such high stacks. V2.11 of the model has been altered to give significantly lower concentrations leading to lower stack heights. The unsteady nature of the atmosphere can result in significant differences between predicted and measured data, whatever the dispersion model. However, it is felt that the basic single stack ADMS model will predict concentrations to within an accuracy of a factor 2.



The accuracy of the more sophisticated modules of ADMS is less well known. Areas of the model which really need detailed validation over the coming years are:- (a) Dispersion around buildings, especially in stable atmospheric conditions. Currently very high ground-level concentrations are predicted. (b) Ground-level/area sources in stable atmospheric conditions - always a problem, whatever the dispersion model.

(c) The ’discrete release’, percentiles, annual average, wet and dry deposition and hills modules all require detailed validation. The complex terrain wind flow model has been extensively validated against full scale data - it is reliable upwind of the apex of the hill, but there are concerns about the accuracy in the wake region of the hill.

11.5 CHOICE OF WIND AND WEATHER CONDITIONS FOR DESIGN Every dispersion modeler has his/her own way of specifying appropriate wind and weather conditions for designing a vent. For an elevated stack, worst case ground-level concentrations occur in unstable atmospheric conditions. Stacks must be designed to ensure that concentrations are acceptable in unstable conditions, but do not need to be designed so stringently to cope with the most unstable hour of weather conditions during over a 10 year period. Instead, as a rule of thumb, most UK designers will use as their basis, Pasquill-Gifford B stability conditions rather than the more unstable A conditions. B-stability conditions, will occur for 1-2% of the time in the UK. If however, a stack was being designed for central Spain or Texas, then it would probably be necessary to design the stack to cope with A-stability conditions. For use with ADMS, a reciprocal of the Monin-Obukhov length of -0.075 /m is representative of typical unstable conditions found in the UK. As discussed in the above text, the stack height is specified usually by trial and error to ensure that peak concentration over publicly accessible areas is a small fraction of the appropriate air quality standard.



Certain practitioners, however, estimate stack heights by using a sequential list of real met data - sometimes 10 years’ worth of data - and then find a worst case hourly average. Then the background concentration is added to the worst case hourly data and then compared against 80% of the air quality limit. This method can (but not always) lead to slightly shorter stack heights. However, there is no account taken for errors in the model - likely to be significant for the worst case hourly data and run times can be prohibitively long. 11.6 RUN TIMES 11.6.1 General As would be expected, run times are dependent on the number of sources and the actual module of ADMS used. For a single stack, and say 10 different wind conditions, run times are less than a minute. Annual average calculations for a single isolated stack may take up to 10 minutes. However, annual average calculations with building effects can take up to 2 hours. (For building calculations, concentrations are a function of wind direction, for a given wind speed and stability, concentrations from an isolated stack over flat ground are independent of wind direction).

Multiple stack, multiple source component problems obviously take longer than for an isolated stack. Also 31 x 31 grid calculations can take much longer than a 21 x21 grid. A finer grid requires more RAM. If the capacity of the RAM is exceeded, then it is necessary to write to and from the virtual memory on the hard disc, which significantly slows down the problem. If emissions from a whole site are to be assessed simultaneously, I would strongly suggest upgrading a PC to 16 Gbytes of RAM. 98th percentile calculations for multiple sources and multiple components can take days. It is particularly frustrating to wait 3 days for a 98th percentile result to be produced only to find the grid has been incorrectly specified. Always run a problem without percentiles to check that the grid is correct, before running a percentile problem.



Calculations using hills can take hours, especially for multiple sources. Real care needs to be taken with the grids, to enable the predictions to be married together well with the terrain data file. Annual average calculations using sequential met data take many hours.

11.6.2 Running Batch Files For runs taking several hours, it can be useful to carry out a series of runs overnight or over a weekend. The ADMS manual suggests a method of running ADMS in batch mode. However, this is avoided if ADMS is run from the ’Run’ option in Windows. Then, it is necessary to run either admsnh.exe or admsh.exe, depending on whether a flat ground or complex terrain problem is being run. Then directly after the command, type in the APL files that need to be run. For both the ADMS executable file and the APL files, the entire pathname should be specified. Thus if there are three APL files (FILNAM1.APL, FILNAM2.APL and FILNAM3.APL) in directory C:\ADMS-2\WORK, then it is necessary to insert into the box associated with the run command the following:- C:\ADMS-2\ADMS211\ADMSNH.EXE C:\ADMS-2\WORK\FILNAM1.APL C:\ADMS-2\WORK\FILNAM2.APL C:\ADMS-2\WORK\FILNAM3.APL Only a space needs to be inserted between the APL file pathnames.

11.7 WIND AND WEATHER DATA The UK Meteorological Office is the sole supplier of wind and weather data in ADMS format. Most commonly, it is necessary to acquire statistical weather data based upon a 10-year statistical analysis. This is necessary for assessing chronic hazards associated with low level continuous releases of hydrocarbons or for making comparisons against annual average air quality standards for, say, nitrogen dioxide. The cost of such data (December 1997) is around £400. The appropriate contacts at the Met Office are Linda Rigby and Lynette Rogers, Tel 01344 856971 Fax 01344 854906.



Rarely, is there a meteorological station very close to the source - the closest met station can be up to 50 miles away from the site being modeled. There can be cases where the nearest station is not the most appropriate one to use, especially where terrain effects can play an important role in the wind statistics at the nearest met station. If in doubt, however, use the nearest weather station. For most stations the prevailing wind direction is from the south west. However, it should be noted that the frequency of stable and unstable weather conditions increases as the distance from the coast increases. There is a world wide net of meteorological data and the Met Office can configure weather data from around the world in ADMS format.

11.8 SUMMARY OF ROUGHNESS LENGTHS (Z O) The roughness length is a parameter which quantifies the effect ground roughness has on the turbulent flow properties of the wind - the higher the roughness length, the more turbulent is the wind flow. For an elevated stack, the higher the roughness length, the more rapidly the plume centerline concentration decreases with distance. However, the higher the roughness length, the more rapidly the plume spreads in the vertical direction, counteracting the effect of roughness on plume centerline concentrations. Hence, it is not possible to generalize the effect surface roughness has on ground-level concentrations. For a ground-level release of a heavier-than-air gas cloud, the higher the surface roughness, the more rapid is the dispersal rate of the cloud Estimating roughness lengths can be difficult - rarely is the terrain uniform around a source - in general, consider the roughness of the ground upwind of the source. Typical values are as below:-



If in doubt choose a roughness length of h/30 where h is the average height of the obstacles (e.g., if Zo = 0.3 m), the typical size of the roughness elements would be 9 m. This is only a simple rule of thumb.



11.9 CALCULATION TRENDS Finally, a few comments in general about environmental gas dispersion. (a) In gas dispersion modeling there are no simple ’back-of-the- envelope’ calculations which enable rough estimates of concentration to be determined. (b) For design, it is always preferable to minimize the emission at source rather than to raise stack heights. (c) As the height of a stack increases, the value of the peak ground- level concentration decreases and it occurs at point increasingly distant from the source. (d) For ground-level releases, highest concentrations occur under stable atmospheric conditions. (e) For elevated stacks, worst case conditions usually occur in low wind speed unstable atmospheric conditions. For high momentum sources however, higher concentrations can occur in high wind speed neutral atmospheric conditions. (f) If there is an adjacent building higher than the stack, worst case conditions will occur when the plume makes a direct impact against the building under stable atmospheric conditions. In this case, the plume centerline concentrations should be used rather than ground-level concentrations. If it is necessary to attach a vent to a building say 10 meters high, the stack height should be at least 3 meters above the apex of the building.

The building module is only really suitable for use, if there is a single building which will significantly affect the dispersion from a stack. For a whole site with many buildings, but with no single building controlling the dispersion, the best approach is to increase the roughness length (e.g. from 0.3 m to 0.6 m), but not use the buildings option. Increasing the ground roughness will increase the atmospheric turbulence, spreading the plume more in the vertical direction, resulting in higher ground-level concentrations for elevated stacks.



(g) For an emission from an elevated source, raising stack heights a few meters will not significantly reduce ground level concentrations, especially if the peak in ground-level concentration occurs upwind of the nearest target point of interest.

(h) For an elevated stack, ground-level concentrations can be reduced by increasing the buoyancy of the plume and by increasing the exit velocity of the gas flow from the stack.

(j) Always take care with units, especially when converting from mg/m3 to ppm. In general, concentrations expressed as mg/m3 are equivalent to ppm, mg/m3 are equivalent to ppb. Note also that concentrations, mg/m3 and mg/m3 are dependent on temperature, ppm and ppb are not.

12 DISPERSION MODELING OF ODOROUS RELEASES 12.1 ODOR EMISSIONS - CHARACTERIZATION AND MEASUREMENT When considering the environmental effects of emissions of volatile organic compounds (VOCs), it is frequently found that odor effects are more critical than toxicity effects. Thus, many issues such as stack design and process authorizations are determined by the odor impact of a material rather than its toxicity. It is therefore very important to ensure that odor impacts are correctly assessed. Odor is unusual in air pollution issues in that the critical timescale for an odor nuisance is the time taken to breathe in and out - of the order of 5 seconds. Consequently, people are sensitive to short-term peaks in concentration which arise only intermittently, rather than the continuous average level. In other words, an odor does not need to be present all the time for an odor nuisance to arise. If an odor is present only intermittently, this will be perceived by members of the public as an odor nuisance. In assessing whether a predicted level of odor is likely to cause a nuisance, it is vital to use an appropriate measurement of pollutant levels that takes this issue into account. Most occupational or public health predictions relate to concentrations averaged over a period of 15 minutes or one hour, and dispersion models are set up to produce concentrations averaged over these lengths of time.



However, during these periods, the instantaneous concentrations of material will fluctuate producing intermittent values up to 40 times higher than the longer-term average. These short-term concentrations should be considered when designing plant to avoid odor nuisance. As well as the difficulties associated with averaging time, odor is a subjective matter, and the analysis of air samples for odor content is time-consuming and expensive. The variability of measurements of the odor strength of different materials is also a serious issue. Measured odor strengths for a given chemical can differ by several orders of magnitude. A robust value should be used that will provide a reasonable degree of public protection. However, the use of a worst-case value could easily result in serious overdesign and additional expenditure. A definition of odor strength is given in Box 1.

The question of averaging times is discussed in Section 12.2. Methods of characterizing the odor strength of an emission are discussed in Section 12.3. Dispersion modeling for odorous emissions is outlined in Section 12.4, and an example is given in Section 12.5. 12.2 AVERAGING TIMES

In dispersion modeling, averaging times are frequently ill-defined. However, the effect of averaging time on measured concentrations of material is highly significant. There are three principal effects that can lead to a variation in concentration with averaging time.

• Concentration fluctuations.

• Changes in mean wind direction.



• If the averaging time is significant relative to the release duration. This effect is not considered in more detail here, in view of the very short averaging times under consideration with regard to odor impacts.

12.2.1 Concentration Fluctuations As a plume travels downwind, it does not move in a straight line and spread uniformly. The wind direction fluctuates about its mean value, leading to a meandering plume. At the same time, packets of clean air are entrained into the plume of released material, leading to zones of high and low concentration of material within the plume. Consequently, at any instant, concentrations of material will be high at some points and low at other points. Averaging these highs and lows results in the familiar Gaussian distribution of concentration across the mean wind direction. This is illustrated in Figure 12.1.

FIGURE 12.1 INSTANTANEOUS AND AVERAGED PLUME DISPERSION

Instead of considering the plume as a whole, we now consider concentrations of material recorded at a single point - for example, point P in Figure 13.1. For the reasons given above, there will be short periods when the air contains very high concentrations of material, and periods when concentrations of material are very low. Low concentrations will arise when the wind direction has shifted away from line between the source and the measurement point, or when a packet of clean air entrained in the plume passes over the point.



At point P, for example, there would currently be very low concentrations of material, as the point lies outside the path of the plume. However, there would have been high concentrations prior to the plume reaching the position shown in Figure 12.1, and high concentrations are likely to arise at P again as the portion of the plume currently at point Q moves downwind. The concentrations at point P would typically have the form shown in Figure 12.2. In practice, the instrumentation may not be sufficiently accurate to measure low concentrations of material: these are likely to be recorded as zeros. Furthermore, the instrument will take a finite time to take the measurement, and this may be significant in averaging out short-term fluctuations in the signal. This is illustrated in Figure 12.3. The “actual” concentration profile between times 600 and 700 (as shown in Figure 12.2) would give a “measured” concentration profile which averages out the short-term peaks in the signal.

FIGURE 12.2 ACTUAL CONCENTRATIONS OF RELEASED MATERIAL



FIGURE 12.3 ACTUAL AND MEASURED CONCENTRATIONS OF RELEASED MATERIAL

From this measured data, it is possible to define a number of descriptions of the data. The “total mean” value is the mean of all measurements including zero values. The “conditional mean” value is the mean of all measurements after zero values have been removed from the data set. The peak concentration is the concentration that would arise once during an event. The nth percentile is the value below which n% of the measurements lie. These are illustrated in Figure 12.4.



FIGURE 12.4 STATISTICAL DESCRIPTIONS OF MEASURED CONCENTRATIONS

As can be seen from the two sets of data in Figure 12.3, the peak concentration is very sensitive to the averaging time. As the averaging time increases, the peak concentration decreases. For example, increasing the averaging time from one time unit (“actual” data shown in Figure 12.2) to 9 units (“Measured” data in Figure 12.2) to 1000 time units (“total mean” value shown in Figure 13.4) results in a change in peak concentration from 92 units to 71 units to 8 units. Dispersion models are generally set up to predict long-term mean concentrations, whereas odor impacts are determined by the peak concentrations likely to occur on a time scale of a few seconds. Hence, some adjustment to the predicted concentrations is necessary to provide useful data for the assessment of odor impacts. Concentration fluctuations do not occur uniformly under all conditions. Convective meteorological conditions give rise to much greater fluctuations in concentration than stable meteorological conditions because of the greater turbulence in the atmosphere (by definition). So, for example, under convective conditions, the 99th percentile of 5-second mean concentrations of odorous material from a source 10 meters above ground level might be typically 40 times the mean level. In contrast, under



stable conditions, the 99th percentile of 5-second mean concentrations might be typically 15 times the mean level.

This could be significant for example in an application where stable conditions give the worst case hourly mean concentrations: convective conditions could give the worst case conditions for odor impacts.

12.2.2 Change in Mean Wind Direction

In general, the mean wind direction does not vary significantly over periods of less than one hour. Experience tells us that the wind direction can change very rapidly over periods of a few seconds, leading to the kind of plume meandering shown in Figure 12.1. However, it is usually possible to determine an overall mean wind direction. When calculating pollutant concentrations, it is generally appropriate to assume that the wind could blow consistently from a source to an individual receptor for periods of up to one hour. The exception to this is periods of very low wind speed, when dispersion models are in any case not generally applicable. Over longer periods of several hours or more, the wind direction may vary more significantly - commonly referred to as the wind direction veering. This leads to a reduction in concentrations averaged over this length of time because the wind will typically not blow consistently from a source to a receptor over this length of time. This is illustrated in Figure 12.5.

FIGURE 12.5 WIND DIRECTION ENVELOPES FOR SHORT AND LONG-TERM MEANS



It is important to note that wind direction veering is the only effect that is taken into account by the dispersion model ADMS when the averaging time in the set up screen is changed. This is because the model is aiming to predict mean values appropriate for the averaging time, rather than peak values. Using an averaging time of (say) 10 seconds will give a result which is the expected value of the 10-second mean concentration, given the meteorological data. This is almost identical to the expected value of hourly mean concentrations, and is not the same as the highest 10-second mean concentration that would arise during 1 hour of measurements. To obtain this value, it would be necessary to use the fluctuations module of ADMS. Using an averaging time of (say) 8 hours will give a result which is the expected value of the 8-hour mean concentration. This will be less than the hourly mean concentration, because some veering in wind direction would be expected over this 8 hour period.

12.2.3 Accounting for Dependence on Averaging Time The correct approach to be adopted in modeling studies depends on the nature of the information used in the study, and the information required from the study. Measurements of meteorological parameters are invariably provided in the form of hourly mean values. Dispersion modeling studies are generally either: (a) calculations using individual meteorological cases; or (b) calculations of annual statistics using long-term (1-year or 10-year) sets of meteorological data. Table 12.1 provides a guide to calculating statistics for various applications.



TABLE 12.1 APPROPRIATE AVERAGING TIMES

As indicated in Table 12.1, an appropriate method to assess odor impacts is to select the combination of meteorological conditions likely to give rise to the worst case impacts. Dispersion modeling should be carried out to give the hourly mean concentrations likely to arise from this combination of conditions. These hourly mean concentrations then need to be adjusted to give a reasonable estimate of the highest 5-second peak in concentration likely to arise during the conditions. This requires a view to be taken on what “likely” means in this context. The choice of an appropriate adjustment to the modeled hourly mean concentrations is discussed in Section 12.4.

12.3 ODOR THRESHOLDS

As discussed in Section 12.1, it is important to know the odor strength of the materials being dispersed. In general, this is achieved in one of two ways.



• Use of measured/estimated release rates of odorous chemicals combined with laboratory measurements of their odor thresholds. This is cheap and relatively straightforward, but there remains the possibility that odor could be caused by chemicals not included in the assessment. The quality of odor threshold measurements is often poor. • Taking samples of the released material and carrying out an olfactometric study to establish the odor strength of the material directly. This is expensive and unless a large number of samples are taken, can give highly variable results due to variations in emissions, variability in individual responses to odor, decay of sample before analysis etc. In general, sampling followed by olfactometry is unlikely to be a cost-effective method of estimating the source strength of an odorous emission. It might be an appropriate tool for assessing a known source of odor which is giving rise to consistent complaints of odor nuisance. Otherwise, the use of measured odor thresholds is recommended. A number of criteria have been used to define odor threshold, such as the level at which 50% of a panel can detect an odor, or the level at which 100% of a panel recognized the odor as being representative of the material being studied. The 50% detection level (referred to as “absolute odor threshold” or “population perception threshold,” PPT50) is the most appropriate value to use in odor assessments. Compilations of odor threshold data can be obtained from the sources given in Box 2.



It should be noted that odor threshold data is extremely variable. For example, the measured odor thresholds for ammonia quoted in Verschueren range from 7 x 10-4 ppm to 110 ppm. It is recommended that a median or mean value from a range of referenced sources should be used, and the interpretation of the results of odor assessments should have regard to the variability in measured odor thresholds. It is particularly important to be aware of the potential uncertainty in odor threshold for materials with just one measurement of odor threshold, or where a single compilation value with no indication of its reliability is used.

12.4 ODOR DISPERSION MODELING

The stages in carrying out a dispersion modeling study of odorous emissions are as follows: (a) Estimate emissions of odorous material, considering routine operations and conditions likely to give the highest emission rates.



(b) Use an appropriate dispersion model to evaluate odor concentrations at relevant locations (e.g., site boundary; closest residential properties; properties from which complaints have been received; public amenities such as parks). (c) Use the following criteria to evaluate whether the modeled odor levels could be detectable:

(1) 40 times maximum hourly mean concentration above the odor threshold (1 OU/m3) - guideline traditionally used within GBHE; (2) 98th percentile of 5-second average concentrations above the odor threshold (1 OU/m3), taking the frequency of the release into account. This can be determined using the fluctuations module of ADMS.

(d) Use the following criteria to evaluate whether the modeled odor levels could comprise a nuisance:

(1) 40 times maximum hourly mean concentration above 5 times the odor threshold (5 OU/m3); (2) 98th percentile of 5-second average concentrations above 5 times the odor threshold (5 OU/m3), taking the frequency of the release into account.

As always, attention should be given to the options for preventing or minimizing releases of material. New plant should be designed to ensure that there will be no odor nuisance arising from any emissions. Two equivalent approaches can be used in carrying out the dispersion modeling part of a study. The more straightforward method is to model the dispersion of the chemical components of the release. This will give ambient concentrations of these components at the locations of interest. Dividing the concentration by the odor threshold gives the odor strength in odor units per cubic meter.



Equivalently, the chemical components of the release can be converted to an odor strength in the release by dividing the concentration in the effluent by the odor threshold. Multiplying this by the volumetric flow rate (m3s-1) gives a value for the source strength in OU/s. The dispersion of odor can then be modeled using “OU” rather than a mass measurement of emissions to give ambient levels of odor in OU/m3.

12.5 EXAMPLE ODOR DISPERSION MODELING STUDY

The known odorous components of releases from the plant are given in Table 12.2. The odor thresholds of the materials listed in the table are derived from the references given in Box 2.

TABLE 12.2 EXAMPLE STUDY: PLANT ODOROUS RELEASES

A schematic plan of the site and the nearest location from which complaints have been received is shown in Figure 12.6.



FIGURE 12.6 EXAMPLE STUDY : SITE DIAGRAM

Releases from the site were modeled using ADMS under the 6 sets of meteorological conditions listed in Box 4 of Part 4 of this Guide. The following results were obtained. TABLE 12.3 EXAMPLE STUDY: MODELED CONCENTRATIONS

“Mean” refers to the modeled hourly average concentration at the property “98th %ile” refers to the modeled 98th percentile of 5-second average concentrations at the property. The highest modeled concentrations are shown in bold.



Comparing the modeled concentrations to the odor thresholds indicates that the continuous releases are very unlikely to be the cause of the odor complaints. The peak concentration of methyl acetate is 449 mgm-3, less than 0.5% of the odor threshold. The peak concentration of formaldehyde from the continuous vent is less than 10% of the odor threshold. The mean concentrations of odorous material arising from the process changeover which takes place every 48 hours are below the odor threshold. However, as discussed above, this does not necessarily imply that there will be no odor due to these releases at the property. Short term peak concentrations from these releases have been estimated by multiplying the hourly mean value by 40, as the fluctuations module is not available for area sources.

The peak concentration of acetic acid from this source is close to the odor threshold of the material. Thus, while an intermittent acetic acid odor may occur at the property, it is unlikely to comprise a nuisance. The peak concentration of formaldehyde from this source is five times the odor threshold. Thus, an odor nuisance could arise from this source during the process changeover. The uncertainty in odor threshold of formaldehyde should also be borne in mind: the threshold could be a factor of 10 higher or lower than the value of 400 mgm-3 used in the study. The reports of “acrid chemical fumes” are not entirely consistent with the pungent hay/straw odor of formaldehyde. The study conclusions are that formaldehyde released during the process changeover is the only likely source of odor at the nearby property. Odor would only occur during the first 10 minutes of the process changeover period, and then only if the wind was blowing from the site towards the property. The plant is unlikely to be the cause of a persistent odor problem. It would be worth repeating the exercise for other properties from which complaints have been received, and also investigating whether complaints are correlated with the plant changeover period and/or wind direction.



13 BIBLIOGRAPHY

Britter RE and Griffiths RF (1982), ’ Dense Gas Dispersion’, Journal of Hazardous Materials, vol 6 issues 1 and 2, Elsevier Press. Carter D (1995), ’Hazardous Substances on Spillage’, IChem E. CCPS (1990), ’Instructors Guide for Safety, Health and Loss Prevention in Chemical Processes’. CCPS (1995),’Understanding Atmospheric Dispersion of Accidental Releases’. CCPS (1996), ’Guidelines for Use of Vapor Cloud Dispersion Models (Second Edition)’. Colls J (1997), ’Air Pollution - An Introduction’, E+FN Spon Publishers. De Nevers N (1995), ’Air Pollution Control Engineering’, McGraw-Hill. DOE (1997), ’The United Kingdom National Air Quality Strategy’, HMSO. Department of Health (1997), ’Handbook on Air Pollution and Health’, The Stationery Office. Environment Agency (1996), ’Guidance for Operators and Inspectors of IPC Processes: Best Practicable Environmental Option Assessments for Integrated Pollution Control Volume 1: Principles and Methodology Volume 2: Technical Data (for Consultation) (Technical Guidance Note E1)’, HMSO. Environmental Analysis Co-operative (1996), ’Released Substances and their Dispersion in the Environment’, HMSO. HMIP (1993), ’ Guidelines on Discharge Stack Heights for Polluting Emissions, Technical Guidance Note D1’, HMSO. HSE (1998), ’EH40/98,Occupational Exposure Limits 1998’. NSCA (1997), ’ 1997 Pollution Handbook’, NSCA.



Seinfeld J H (1986), ’Atmospheric Chemistry and Physics of Air Pollution’, Wiley Press. Turner D B, (1994), ’Workbook of Atmospheric Dispersion Estimates (2nd edition)’, Lewis Press. US-EPA (1992), ’Reference Guide to Odor Thresholds for Hazardous Air Pollutants listed in the Clean Air Act Amendments of 1990, Report EPA/600/R-92/047. Valentin FHH and North AA (1980), ’Odor control - a concise guide’, Warren Springs Laboratory report ISBN 0 85624 2144. Wilson DJ (1995), ’Concentration Fluctuations and Averaging Time in Vapor Clouds’, CCPS Concept Book. Woodfield M and Hall D (1994), ’ Odor measurement and control - an update’, AEA Report AEA/CS/REMA-038.



14 GLOSSARY

Adiabatic lapse rate The temperature of the atmosphere generally reduces with height over the first 10 - 15 km of the atmosphere. This arises because there is no significant solar heating to counterbalance the drop in atmospheric pressure over this region. The adiabatic lapse rate is the rate at which the temperature of a parcel of air would change on rising through the atmosphere without exchanging heat with the surrounding atmosphere, approximately equal to 1°C per 100 m. If the actual lapse rate in the boundary layer is greater than this, then a rising parcel of air will cool less rapidly than the surrounding air and tend to continue rising. This corresponds to unstable conditions. Conversely, a smaller lapse rate, or an increase in temperature with height leads to a rising parcel of air being more dense than the surrounding atmosphere. The air parcel would thus tend to sink down. This corresponds to stable conditions. ADMS ADMS (Atmospheric Dispersion Modeling System, also known as UKADMS). This is a modified Gaussian dispersion model, which is the model of choice for most neutral-density or buoyant gas releases. The model uses an advanced understanding of the structure of the boundary layer to model the vertical dispersion of pollutants. This is particularly important in unstable atmospheric conditions. ADMS also incorporates modules to take account of the influence of buildings and complex terrain on dispersion, and to assess concentration fluctuations. Advection The process by which a gas plume is moved bodily downwind without any dilution. This would occur when the turbulent air velocity scales are small in comparison with the wind speed. Air quality A description of air in terms of the quantity of air pollutants it contains. “Good” air quality means low levels of all pollutants; “poor” air quality means that levels of one or more pollutants are high. Various definitions of low and high levels of air pollutants exist. For example, a range of air quality standards and guidelines have been specified to assist in defining low and high levels of air pollution, generally in terms of the human health effects of the pollutant.



Air quality framework directive A European Union directive which lays down a mechanism of establishing a sliding scale of air quality standards. Two levels can be specified for a pollutant, the first for immediate application and the second for application at a specified future date. In the intervening period, the standard is progressively tightened towards the second more stringent level. A series of “daughter directives” will set immediate and target levels for a range of pollutants. Air quality strategy/ Air quality objectives In March 1997, the UK Government published an air quality strategy for the UK. This strategy contains objectives for air quality to be achieved by 2005. Local Authorities are required to carry out assessments of local air quality, and take action to improve air quality if it appears likely that the objectives will not be met in any area. Air quality standards Limits on concentrations of pollutants in the air, usually specified to protect public health. Air quality standards are set by international or national governments; recommendations are also made by interested bodies, notably the World Health Organization. Air quality standards apply to environmental levels of pollutants from all sources in combination, rather than to emissions from a single source. Air quality targets See Air quality objectives. Ambient Conditions that would be experienced by members of the public and/or the wider environment. Ambient air quality standards, guidelines and objectives are applicable in areas where the public may be expected to be present for a significant proportion of the time - for example, homes, offices and shops. Atmospheric stability The stability of the atmospheric boundary layer has a profound effect on the dispersion of air pollution. Atmospheric conditions can vary from highly turbulent with considerable vertical mixing of air, to stably stratified conditions with very restricted vertical mixing.



Convectively turbulent conditions occur when the atmosphere is strongly warmed by the transfer of heat from the ground under low wind speeds: these are known as unstable conditions. These conditions would typically occur on a warm summer afternoon, and favor rapid mixing and dispersion of pollutants. Under these conditions, high concentrations can occur close to elevated sources. Stable conditions occur at low wind speeds when the temperature is lowest close to the ground. Under these conditions, which occur during winter nights, the dispersion of pollution is restricted. Consequently, stable conditions give the highest concentrations for ground-level sources. Neutral conditions occur when there is no appreciable heat transfer to/from the ground, or when wind speeds are above 5 m/s. Averaging time Any atmospheric concentration measurement has an associated averaging time, ranging from less than a second to a year or more. The averaging time affects the measurement in two ways. Firstly, dispersion tends to take place by splitting the plume of concentrations into smaller and more widely separated packets. This can dominate the entrainment process by which air is mixed into the released material. Thus, a short-term measurement may be very low (if no released material is present during the measurement), or very high (if a packet of material is present). Consequently, short-term measurements can be highly variable. Secondly, the wind direction can be variable over periods greater than about an hour. This means that a particular location will not be downwind of the source for all of the measurement period, resulting in a reduction in concentrations averaged over a longer period. Background concentrations Emissions of a pollutant from a particular source will be additional to the concentrations of that pollutant which are already present in the environment from other sources. These may include natural sources, domestic emissions, road traffic or other industrial sources. To assess whether emissions from a particular source would result in the contravention of a standard, it is necessary to add the forecast contribution from the source to the background levels of the pollutant(s) arising from other emissions. Background levels of air pollutants can be assessed by analyzing local air quality monitoring data, by considering data obtained in other comparable locations, or by modeling the emissions from these other sources.



Boundary layer The lowest part of the atmosphere, in which air flows are significantly affected by friction with the surface of the earth. The boundary layer is the region within which the wind speed increases with height. The boundary layer typically extends to a height of one hundred meters under stable atmospheric conditions, to one kilometer in neutral conditions, and to two kilometers under unstable conditions. This is also referred to as the “boundary layer depth”. Pollution released within the boundary layer generally remains within the boundary layer. However, if the plume reaches the top of the boundary layer, some material can penetrate to the free troposphere above. Building wake The zone of turbulent air caused by an obstruction to the wind flow such as a building or plant structure. The zone generally extends up to 5 or 6 building lengths downwind of the building, and can extend some distance upwind of the building. Building wakes can be highly significant in affecting the dispersion of pollution from an elevated source by transporting material to ground level from the elevated plume. Buoyant Plume A gas release with an initial density lighter than that of ambient air. This density difference may be due to the cloud having a molecular weight less than that of air and/or having a significantly higher temperature than the ambient conditions. Centerline concentration This is the highest concentration that can be found anywhere in the plume at or above ground level. Sometimes the term centerline concentration is used to represent the highest ground-level concentration which will be directly below the real plume centerline. Choked Flow The exit velocity of a gas release through a hole from a high pressure source is limited by the speed of sound in the gas. When the exit velocity is at the speed of sound, then the flow is choked. The source pressure at which choking occurs is termed the critical pressure. If the pressure is increased above the critical pressure, the density of the gas flow will increase, but the exit velocity through the hole remains constant.



CIMAH study A CIMAH (Control of Industrial Major Accident Hazards) study is a safety case demonstrating to regulators that the major hazards on a chemical plant are adequately controlled. COMAH study COMAH regulations have been developed by the EU into safety regulations covering a wider range of chemical and other major hazards. COMAH stands for Control of Major Accident Hazards. Complex Terrain A general term used to define any terrain which is not flat or does not have a uniform roughness (ZO) in all directions from the source. Typically, ground is regarded as flat if there are no slopes steeper than 1 in 10. Concentration fluctuations Over short averaging periods, concentration measurements can be highly variable (see Averaging time). Although a large number of these short-term measurements would have the same mean value as a longer period measurement, there would be a small number of extremely high measured values. These high values are of significance with regard to issues such as odour. Their occurrence can be described in terms of percentiles - for example, values of the 95th or 98th percentile short-period concentration could be used to describe the upper range of fluctuations likely to arise. ADMS contains a module to assess concentration fluctuations. Typically, the maximum ground-level concentration over a few seconds might be 40 times the mean recorded over a whole hour. These peaks are caused by the highly turbulent nature of the atmosphere bringing parcels of high concentration air down to ground level. Convective conditions Conditions of high turbulence in the atmospheric boundary layer. See Atmospheric stability Dense Gas A gas release with an initial density (10% or more) heavier than that of ambient air. This density difference may be due to the cloud having a molecular weight higher than ambient air and/or a significantly lower temperature than the ambient conditions. For example, liquefied methane (molecular weight 16, compared to a bulk value of 29 for air) will show dense gas dispersion properties in the near field when the



cloud is cold. In the far field it will exhibit properties of a lighter than air gas. To predict plume concentrations for a dense gas release it is vital that a model specifically for heavier-than-air gases is used, e.g. PHAST’s UDM model, DISP2’s BURST model or ALOHA. Dispersion modeling The use of mathematical models to describe how a plume of material from a specified source is likely to disperse in the environment. The most widely used gas dispersion models are Gaussian models. This assumes that the concentration of material downwind of the source has a bell-shaped distribution about its maximum in both the vertical and cross-wind directions. Empirical values are used to estimate the width of these distributions (usually expressed as the standard deviation of the distribution in the y and z directions, σy and σz). For a given downwind distance under a range of atmospheric conditions. A simplified expression for downwind concentrations is given by:

Dispersion coefficients The parameters that describe the dispersion of material in the crosswind and vertical directions are referred to as dispersion coefficients, and written σy and σz. (see Dispersion modeling). Various sets of dispersion coefficients have been derived from experimental data, which depend on distance downwind from the source, and atmospheric stability class (see, for example, Turner, 1994). These are then applied in a dispersion equation to give an estimate for the concentration of material at a given point. This can be done by hand, or by using a dispersion model. Dose The total quantity of material absorbed or experienced by an individual. The dose is given by:-



Entrainment The flow of air into a toxic or flammable gas cloud, leading to dilution of the components in the release. The entrainment rate is the flow rate of air into the gas cloud. Environmental assessment levels Levels of ambient air pollution considered by the Environment Agency to be a suitable benchmark for assessing industrial process emissions (See EA Technical Guidance Note E1). Where air quality objectives or standards exist for a particular pollutant, these are used as the EALs. Otherwise, a fraction of the occupational exposure standard is used. EPAQS The Expert Panel on Air Quality Standards. A body advising the UK government on standards for ambient air quality. EPAQS has produced a series of recommendations, which have informed the air quality objectives given in the UK air quality strategy. Fluctuations See Concentration Fluctuations. Friction velocity A reference velocity used to describe atmospheric conditions, defined as u* = Ö(t/r), where t is the surface shear stress, and r is the air density. The friction velocity increases with wind speed and surface roughness. It is generally of the order of 10% of the mean wind speed. Fumigation An elevated plume released under stable atmospheric conditions disperses poorly. If the stable layer passes over an unstable layer (for example, following sunrise in the summer), mixing down to ground level will occur. This is known as fumigation, and can give rise to high ground-level pollution concentrations.

Gaussian model See Dispersion modeling. Ground Level Concentrations This term refers to the concentrations that are found at ground level downwind of the plume. The highest ground level concentrations will occur along a line vertically below the line of centerline concentrations. Determining these concentrations is frequently the aim of dispersion modeling studies.



In other cases, concentrations at non-ground level locations such as storage tanks or air ventilation inlets may be of interest. Where nearby structures are of a similar height to the release point, attention should be paid to centerline concentrations as well as ground level-concentrations. Integrated Pollution Control IPC is the basis on which the Environment Agency regulates emissions to air, land and water from around 2000 major industrial processes. Operators of these major processes have a duty to demonstrate that they are applying the Best Available Techniques Not Entailing Excessive Cost (BATNEEC) to prevent, or minimize and render harmless emissions to all media. Inversion Layer In stable atmospheric conditions, the ambient temperature rises with height above the ground At the top of the stable boundary layer, there is usually a step change in density and temperature above which the temperature begins to fall with increasing height. This step in density serves to trap pollutants within a layer 200 m or less from ground level. Inversion conditions are synonymous with high concentrations and sustained inversion conditions led to the smogs of the 1950s in London. However, if gases are discharged from power station scale stacks (100m+), then pollutants can be discharged at a height above the inversion layer, leading to negligible near field concentrations. Isopleth A line of constant concentration, similar to a contour line on a map. The results of dispersion modeling studies are frequently shown as isopleths of concentration superimposed on a site plan or map. ISC ISC (Industrial Source Complex) is a widely used Gaussian dispersion model, formulated by the United States Environmental Protection Agency. The model is formulated in two ways: ISCST (ISC - Short Term) is for use with individual meteorological conditions or Sequential Weather Data. ISCLT (ISC - Long Term) is for use with Statistical Weather Data. ISC is due to be superseded by a newmodel, AERMOD during 1998.



LC50 The concentration that would result in death for 50% of the exposed population. LC1 The concentration that would result in death for 1% of the exposed population. Local Authority Air Pollution Control LAAPC is the basis on which local authorities in the UK regulate emissions to air from around 12000 industrial processes and 15000 small waste burning operations. Local authority inspectors have regard to guidance from the Secretary of State in determining authorizations under LAAPC. Process operators have a duty to demonstrate that they are applying the Best Available Techniques Not Entailing Excessive Cost (BATNEEC) to prevent, or minimize and render harmless emissions to air. Lower flammable limit The concentration below which a cloud of gas will not ignite. Because concentration fluctuations can lead to local increases in concentrations, it is usual practice to use a fraction (one half or one quarter) of the lower flammable limit for design purposes Maximum Exposure Limit (MELs) If there are serious concerns about the toxicity of a chemical and no known safe threshold has been quantified, a Maximum Exposure Limit rather than an Occupational Exposure Standard is defined. For the specified averaging time for the MEL, workers must not receive concentrations above this limit and concentrations of the chemical must be reduced to as low a figure below the MEL as reasonably practicable. Micrograms per cubic meter (µgm-3) A unit of concentration. A pollutant concentration of one microgram per cubic meter in air means that each cubic meter of air contains a mass of one microgram (10-6 grams) of the pollutant. A concentration of 1 µgm-3 is equivalent to 0.001 mgm-3. At typical atmospheric pressure, 1 µgm-3 = 0.082 (T/M) ppb (M: molecular weight (g/mol); T: temperature (K)). Milligrams per cubic meter (mgm-3)



A unit of concentration. A pollutant concentration of one milligram per cubic meter in air means that each cubic meter of air contains a mass of one milligram (10-3 grams) of the pollutant. A concentration of 1 mgm-3 is equivalent to 1000 µgm-3. At typical atmospheric pressure, 1 mgm-3 = 0.082 (T/M) ppm (M: molecular weight (g/mol); T: temperature (K)). Monin-Obukhov Length A reference length used to describe atmospheric conditions. The Monin-Obukhov length is positive and small (<20 meters) in stable or very stable conditions. It is negative and of similar magnitude in unstable conditions. Its value tends to infinity under neutral conditions (see atmospheric stability). The Monin-Obukhov length is given by:-

Neutral atmospheric conditions Atmospheric conditions under which heat transfer is less significant than mechanical turbulence which prevail for the majority of the time in the Europe. See Atmospheric stability. Neutral density gas A gas release with a density equal or close to (+/- 10%) that of air. Normal Temperature and Pressure (NTP) Normally NTP is defined to be 1 atmosphere and a temperature of 0 degree C. Many gas dispersion models have non-standard definitions of NTP, such as ADMS’s 1 bar, 25 deg C. Occupational exposure standards/limits Limits on concentrations of pollutants in workplace air. In the UK, occupational exposure standards are set by the Health and Safety Executive. They take the form of limits on 15-minute (short term) and/or 8-hour (long term) average concentrations of air pollutants. They are further divided into Maximum Exposure Limits (generally for substances which may cause cancer or workplace asthma) and Occupational Exposure Limits for other materials. This data is published annually in a HSE document referred to as EH40. They are not directly applicable to ambient air quality.



Odor thresholds The odor strength of a particular chemical is normally determined using a laboratory technique known as olfactometry. The concentration of the chemical at which 50% of a panel of 6-8 people can identify that the odor is present is known as the odor threshold. Odor threshold measurements can be highly variable: a measurement range of several orders of magnitude is not uncommon. Oligomerization When released to the atmosphere, Hydrogen Fluoride molecules tend to stick together in groups of 2-5 molecules. This effectively increases the molecular weight and density of the gas cloud, making it more resistant to dispersal. For Hydrogen Fluoride, this oligomerization converts a potentially buoyant gas cloud into one which is heavier than air. Glacial acetic acid also exhibits oligomerization. Oxides of nitrogen A generic term usually taken to mean the sum of concentrations or masses of nitrogen dioxide (NO2) and nitric oxide (NO, as NO2). These two compounds are frequently treated together, as the processes converting one to the other take place rapidly in the atmosphere. The reactions resulting in loss of NO or NO2 are generally less significant. Consequently, to a reasonable approximation, the quantity of NOx in a plume is conserved. Part per billion (ppb) A unit of concentration. A pollutant concentration of one part per billion by volume in air means that one billionth (10-9) of the air volume is occupied by the pollutant. This also means that one billionth of the molecules in the air are pollutant molecules. A concentration of one ppb is equivalent to 0.001 ppm. At typical atmospheric pressure, 1 ppb = 12.2 (M/T) mgm-3 (M: molecular weight (g/mol); T: temperature (K)).



Part per million (ppm) A unit of concentration. A pollutant concentration of one part per million by volume in air means that one millionth (10-6) of the air volume is occupied by the pollutant. This also means that one millionth of the molecules in the air are pollutant molecules. A concentration of one ppm is equivalent to 1000 ppb. At typical atmospheric pressure, 1 ppm = 12.2 (M/T) mgm-3 (M: molecular weight (g/mol); T: temperature (K)). Pasquill-Gifford Stability Class A widely used scheme for classifying air conditions according to the atmospheric stability. See Stability Class. Percentile Some air quality standards need not be met all the time. Instead, a percentage is assigned to the standard. This is the percentage of measurements which must be below the specified level for the standard to be met. For example, the current European Union air quality standard for nitrogen dioxide is a limit of 200 mgm-3 on the 98th percentile of hourly average nitrogen dioxide concentrations. Compliance is calculated by recording hourly average nitrogen dioxide levels for a calendar year. The n values are listed in ascending order. The value at position (0.98 x n) is the 98th percentile. If this value is less than 200 mgm-3, then compliance with the standard has been achieved. PM10 Particulate matter with a diameter of less than 10 microns (more strictly, particulate matter which passes through a size selective inlet with a 50% collection efficiency cut-off at 10 microns). Plume center-line The plume centre-line is the locus of the centre of mass through successive vertical slices of the plume. Usually, it is the location at which the concentration within the plume is a maximum. For a neutral density gas plume from a stack, the height of the plume centre-line will rise close to the source and then reach a plateau as the downwind distance increases. Sometimes the term ‘plume centre-line’ incorrectly refers to the straight line at ground level along which the concentration is a maximum. This line can be found directly beneath the true centre-line.



Receptor A particular location at which pollutant concentrations are calculated. Reference conditions Emissions from a stack are frequently expressed in terms of some standard reference conditions rather than the actual stack conditions. Before emissions modeling can be carried out, these must be converted back to the actual stack conditions using the following formula:

d denotes discharge conditions; s denotes conditions at STP; H2O is the % moisture content; O2 is the % oxygen content. Richardson Number A non-dimensional number, in general the ratio of potential energy to kinetic energy in a gas. The Richardson number is used to characterize the degree of stratification of a dense gas release, and is also applied in determining whether a particular set of ambient conditions are convective or stably stratified. The Richardson number is sometimes called densimetric Froude number. RMP Regulations In the USA new regulations are in place, requiring operators of hazardous plants to carry out worst case hazard assessments for emergency planning. The techniques for quantifying the hazards are highly prescribed. Running mean An n-hour running mean concentration is calculated by taking the average of n consecutive hourly concentration readings (1/nSi=1 i=nci). The next value is obtained by averaging the group of n concentrations starting from the second value in the previous set ( 1/nSi=2 i=n+1ci ). This procedure is continued so that a smoothed set of running mean concentrations is built up - for example, 24-hour running mean concentrations will eliminate the variability in concentrations during a single day, and give an indication of the variation in concentrations from day to day.



Sequential Weather Data A list of wind and weather conditions recorded every hour or three hours. A typical data set covers a full calendar year of readings (for example, 8760 hourly readings of wind speed, wind direction, Pasquill Stability Class, and boundary layer height. US Weather data is often in this format. Slumping Close to the source of a release, a dense gas cloud will fall under its own weight down to ground level. Once it is in contact with the ground, the hydrostatic head will cause the plume to spread laterally and reduce in depth. This process is known as ’slumping’. Stability Class Atmospheric stability classes are a classification of air conditions according to the atmospheric stability. The most widely used scheme was derived by F Pasquill, and developed by F A Gifford. The scale ranges from Class A (highly convective) to Class G (highly stable). Classes A and G occur only rarely in the UK. Gaussian dispersion models have been specified using sets of parameters appropriate for each of the Pasquill-Gifford stability classes. While useful, this parameterization of the atmosphere has been found to be an oversimplification for some applications, and the most recent models use other measures of atmospheric stability such as the Monin- Obukhov length and friction velocity. Stable atmospheric conditions Atmospheric conditions under which the air in the boundary layer flows in stable layers. See Atmospheric stability.

Statistical Weather Data A list of wind and weather conditions representing the distribution of conditions recorded at a particular weather station over a period of, typically, ten years. The various measured parameters are divided into groups - for example, wind direction may be divided into 16 arcs; wind speed into 5 bands; rainfall into 3 bands; surface sensible heat flux into 4 bands. Each combination of weather conditions is then assigned a frequency of occurrence by analyzing the meteorological records for the weather station. This procedure is known as “binning”. For example, the above analysis would result in 16 x 5 x 3 x 4 = 960 “bins”, each with a measured frequency of occurrence. This analysis is particularly useful in predicting long-term average levels of air pollutants.



Calculating the data for every hour of a 10-year dataset would require almost 88000 separate calculations to be made, rather than the 960 calculations for the statistically averaged data set. Surface roughness length (Z O) A term describing the influence of buildings, vegetation etc. on wind flows across the land surface. Surface roughness is frequently quantified using a “surface roughness length,” referred to as Z o. The surface roughness length is typically 3 - 10% of the average obstruction height. For industrial zones, values of 0.1 to 1.0 m may be appropriate, depending on the density and height of the infrastructure. A value of 0.3 m may be appropriate for suburban zones, and 0.1 m or less for rural areas. Surface sensible heat flux The balance of heat transfer at the earth’s surface. A net flux of heat from the earth to the atmosphere favors unstable atmospheric conditions; a net flux to the earth results in cooling of the lower atmosphere, and the formation of stable conditions (see atmospheric stability). Technical Guidance Note Two relevant Technical Guidance Notes (TGN) have been released by the Environment Agency. TGN D1 (Dispersion) provides guidance on minimum stack heights, and supersedes the Third Edition of the 1956 lean Air Act Memorandum on Chimney Heights (DOE, 1981). The memorandum provides a reasonably straightforward methodology for designing stack heights, and will give conservative results under most conditions. It is not recommended for use in complex terrain, where deposition may be a problem, or where individual buildings or other structures may affect dispersion. The guidance also states that the heights produced are a guide, rather than a mathematically precise definition of discharge stack height. A specific dispersion modeling study should be carried out for more complex situations, or where a more detailed design is required. TGN E1 (Environmental) provides guidance on Best Practicable Environmental Option assessments for IPC processes. In particular, it sets out the approach used by the Environment Agency for assessing the contributions of individual processes to ambient air pollution. Criteria are given for determining whether a release is (in the Agency’s view) a priority for control, or insignificant, or whether it lies between these two extremes.



Two-phase release When material is released from a cooled or pressurized source such as a storage vessel or pipe work, the behavior of the material can be very complex. The material may be released as a liquid alone, or as a gas alone. A liquid spill may then subsequently evaporate and disperse in the atmosphere. Alternatively, a two-phase release may take place, in which the material is released simultaneously as a liquid and a gas. The process of expansion through a small orifice leads to cooling of the release, and possible formation of aerosols in a gas phase release. Further complications are introduced by the cooling that takes place during the evaporation of liquids. A range of methods of varying complexity exist for modeling the phase changes associated with releases of this nature, but it should be noted that subtleties associated with the physics of compressed or cooled releases can have a major influence on material release rates. Unstable conditions A synonym of Convective conditions. See Atmospheric stability. Validation The procedure by which the results of a dispersion model are compared with the results of field measurements. The field measurements should not have been used in constructing the model. A number of statistical measures can be used to assess the performance of the model in reproducing the observations. It is common practice to compare a range of appropriate dispersion models with the results of field experiments of dense gas dispersion. Bodies such as the EU fund work in this area with the aim of providing guidance to model users on appropriate models for a particular situation. Verification The procedure of ensuring that a computer code correctly reproduces the mathematical formulations. This would include checking that formulae have been entered correctly, checking that approximations used in the model do not introduce significant errors; and checking that iterative procedures converge correctly.



Volatile organic compounds Volatile organic compounds (VOCs) is a collective term for organic compounds emitted from industrial, commercial, road traffic, domestic and other sources. Adverse health effects directly from these releases are unusual; however, VOCs contribute to ground-level ozone and photochemical smog formation. Wake region This is a low pressure region downwind of an obstruction such as a building or a hill. Usually, the flows within a wake region recirculate - pollution entrained into such a zone becomes trapped, leading to high concentrations.



APPENDICES



APPENDIX A WIND GENERATION OF PARTICULATES

This appendix contains two excerpts from the United States Environmental Protection Agency AP42 document. Section 12.2.4 contains a methodology for estimating the rates of generation of wind-borne dust from aggregate handling and storage piles, and Section 12.2.5 is applicable to wind erosion from aggregate piles and other open areas.

12.2.4 Aggregate Handling And Storage Piles

12.2.4.1 General

Inherent in operations that use minerals in aggregate form is the maintenance of outdoor storage piles. Storage piles are usually left uncovered, partially because of the need for frequent material transfer into or out of storage. Dust emissions occur at several points in the storage cycle, such as material loading onto the pile, disturbances by strong wind currents, and load out from the pile. The movement of trucks and loading equipment in the storage pile area is also a substantial source of dust.

12.2.4.2 Emissions And Correction Parameters The quantity of dust emissions from aggregate storage operations varies with the volume of aggregate passing through the storage cycle. Emissions also depend on 3 parameters of the condition of a particular storage pile: age of the pile, moisture content, and proportion of aggregate fines. When freshly processed aggregate is loaded onto a storage pile, the potential for dust emissions is at a maximum. Fines are easily disaggregated and released to the atmosphere upon exposure to air currents, either from aggregate transfer itself or from high winds. As the aggregate pile weathers, however, potential for dust emissions is greatly reduced. Moisture causes aggregation and cementation of fines to the surfaces of larger particles. Any significant rainfall soaks the interior of the pile, and then the drying process is very slow.



Silt (particles equal to or less than 75 micrometers [mm] in diameter) content is determined by measuring the portion of dry aggregate material that passes through a 200-mesh screen, using ASTM-C-136 method.1 Table 13.2.4-1 summarizes measured silt and moisture values for industrial aggregate materials.

Table 12.2.4-1. TYPICAL SILT AND MOISTURE CONTENTS OF MATERIALS AT VARIOUS INDUSTRIES



12.2.4.3 Predictive Emission Factor Equations Total dust emissions from aggregate storage piles result from several distinct source activities within the storage cycle: 1. Loading of aggregate onto storage piles (batch or continuous drop operations). 2. Equipment traffic in storage area. 3. Wind erosion of pile surfaces and ground areas around piles. 4. Load out of aggregate for shipment or for return to the process stream (batch or continuous drop operations). Either adding aggregate material to a storage pile or removing it usually involves dropping the material onto a receiving surface. Truck dumping on the pile or loading out from the pile to a truck with a front-end loader are examples of batch drop operations. Adding material to the pile by a conveyor stacker is an example of a continuous drop operation. The quantity of particulate emissions generated by either type of drop operation, per kilogram (kg) (ton) of material transferred, may be estimated, with a rating of A, using the following empirical expression:

The particle size multiplier in the equation, k, varies with aerodynamic particle size range, as follows:



The equation retains the assigned quality rating if applied within the ranges of source conditions that were tested in developing the equation, as follows. Note that silt content is included, even though silt content does not appear as a correction parameter in the equation. While it is reasonable to expect that silt content and emission factors are interrelated, no significant correlation between the 2 was found during the derivation of the equation, probably because most tests with high silt contents were conducted under lower winds, and vice versa. It is recommended that estimates from the equation be reduced 1 quality rating level if the silt content used in a particular application falls outside the range given:

To retain the quality rating of the equation when it is applied to a specific facility, reliable correction parameters must be determined for specific sources of interest. The field and laboratory procedures for aggregate sampling are given in Reference 3. In the event that site-specific values for correction parameters cannot be obtained, the appropriate mean from Table 13.2.4-1 may be used, but the quality rating of the equation is reduced by 1 letter. For emissions from equipment traffic (trucks, front-end loaders, dozers, etc.) traveling between or on piles, it is recommended that the equations for vehicle traffic on unpaved surfaces be used (see Section 13.2.2). For vehicle travel between storage piles, the silt value(s) for the areas among the piles (which may differ from the silt values for the stored materials) should be used. Worst-case emissions from storage pile areas occur under dry, windy conditions. Worst case emissions from materials-handling operations may be calculated by substituting into the equation appropriate values for aggregate material moisture content and for anticipated wind speeds during the worst case averaging period, usually 24 hours. The treatment of dry conditions for Section 13.2.2, vehicle traffic, "Unpaved Roads", follows the methodology described in that section centering on parameter p. A separate set of non-climatic correction parameters and source extent values corresponding to higher than normal storage pile activity also may be justified for the worst-case averaging period.



12.2.4.4 Controls

Watering and the use of chemical wetting agents are the principal means for control of aggregate storage pile emissions. Enclosure or covering of inactive piles to reduce wind erosion can also reduce emissions. Watering is useful mainly to reduce emissions from vehicle traffic in the storage pile area. Watering of the storage piles themselves typically has only a very temporary slight effect on total emissions. A much more effective technique is to apply chemical agents (such as surfactants) that permit more extensive wetting. Continuous chemical treating of material loaded onto piles, coupled with watering or treatment of roadways, can reduce total particulate emissions from aggregate storage operations by up to 90 percent. References For Section 12.2.4 1. C. Cowherd, Jr., et al., Development Of Emission Factors For Fugitive Dust Sources, EPA-450/3-74-037, U. S. Environmental Protection Agency, Research Triangle Park, NC, June 1974. 2. R. Bohn, et al., Fugitive Emissions From Integrated Iron And Steel Plants, EPA-600/2-78-050, U. S. Environmental Protection Agency, Cincinnati, OH, March 1978. 3. C. Cowherd, Jr., et al., Iron And Steel Plant Open Dust Source Fugitive Emission Evaluation, EPA-600/2-79-103, U. S. Environmental Protection Agency, Cincinnati, OH, May 1979. 4. Evaluation Of Open Dust Sources In The Vicinity Of Buffalo, New York, EPA Contract No. 68-02-2545, Midwest Research Institute, Kansas City, MO, March 1979. 5. C. Cowherd, Jr., and T. Cuscino, Jr., Fugitive Emissions Evaluation, MRI-4343-L, Midwest Research Institute, Kansas City, MO, February 1977. 6. T. Cuscino, Jr., et al., Taconite Mining Fugitive Emissions Study, Minnesota Pollution Control Agency, Roseville, MN, June 1979. 7. Improved Emission Factors For Fugitive Dust From Western Surface Coal Mining Sources, 2 Volumes, EPA Contract No. 68-03- 2924, PEDCo Environmental, Kansas City, MO, and Midwest Research Institute, Kansas City, MO, July 1981. 8. Determination Of Fugitive Coal Dust Emissions From Rotary Railcar Dumping, TRC, Hartford, CT, May 1984. 9. PM-10 Emission Inventory Of Landfills In the Lake Calumet Area, EPA Contract No. 68-02-3891, Midwest Research Institute, Kansas City, MO, September 1987.



10. Chicago Area Particulate Matter Emission Inventory — Sampling And Analysis, EPA Contract No. 68-02-4395, Midwest Research Institute, Kansas City, MO, May 1988. 11. Update Of Fugitive Dust Emission Factors In AP-42 Section 11.2, EPA Contract No. 68-02-3891, Midwest Research Institute, Kansas City, MO, July 1987. 12. G. A. Jutze, et al., Investigation Of Fugitive Dust Sources Emissions And Control, EPA-450/3-74-036a, U. S. Environmental Protection Agency, Research Triangle Park, NC, June 1974. 13. C. Cowherd, Jr., et al., Control Of Open Fugitive Dust Sources, EPA-450/3-88-008, U. S. Environmental Protection Agency, Research Triangle Park, NC, September 1988.

12.2.5 Industrial Wind Erosion 12.2.5.1 General1-3

Dust emissions may be generated by wind erosion of open aggregate storage piles and exposed areas within an industrial facility. These sources typically are characterized by non-homogeneous surfaces impregnated with non-erodible elements (particles larger than approximately 1 centimeter [cm] in diameter). Field testing of coal piles and other exposed materials using a portable wind tunnel has shown that (a) threshold wind speeds exceed 5 meters per second (m/s) (11 miles per hour [mph]) at 15 cm above the surface or 10 m/s (22 mph) at 7 m above the surface, and (b) particulate emission rates tend to decay rapidly (half-life of a few minutes) during an erosion event. In other words, these aggregate material surfaces are characterized by finite availability of erodible material (mass/area) referred to as the erosion potential. Any natural crusting of the surface binds the erodible material, thereby reducing the erosion potential.

12.2.5.2 Emissions And Correction Parameters If typical values for threshold wind speed at 15 cm are corrected to typical wind sensor height (7 - 10 m), the resulting values exceed the upper extremes of hourly mean wind speeds observed in most areas of the country. In other words, mean atmospheric wind speeds are not sufficient to sustain wind erosion from flat surfaces of the type tested. However, wind gusts may quickly deplete a substantial portion of the erosion potential. Because erosion potential has been found to increase rapidly with increasing wind speed, estimated emissions should be related to the gusts of highest magnitude.



The routinely measured meteorological variable that best reflects the magnitude of wind gusts is the fastest mile. This quantity represents the wind speed corresponding to the whole mile of wind movement that has passed by the 1 mile contact anemometer in the least amount of time. Daily measurements of the fastest mile are presented in the monthly Local Climatological Data (LCD) summaries. The duration of the fastest mile, typically about 2 minutes (for a fastest mile of 30 mph), matches well with the half-life of the erosion process, which ranges between 1 and 4 minutes. It should be noted, however, that peak winds can significantly exceed the daily fastest mile. The wind speed profile in the surface boundary layer is found to follow a logarithmic distribution:

The friction velocity (u*) is a measure of wind shear stress on the erodible surface, as determined from the slope of the logarithmic velocity profile. The roughness height (z o) is a measure of the roughness of the exposed surface as determined from the y intercept of the velocity profile, i. e., the height at which the wind speed is zero. These parameters are illustrated in Figure 13.2.5-1 for a roughness height of 0.1 cm.

Figure 12.2.5-1. Illustration of logarithmic velocity profile



Emissions generated by wind erosion are also dependent on the frequency of disturbance of the erodible surface because each time that a surface is disturbed, its erosion potential is restored. A disturbance is defined as an action that results in the exposure of fresh surface material. On a storage pile, this would occur whenever aggregate material is either added to or removed from the old surface. A disturbance of an exposed area may also result from the turning of surface material to a depth exceeding the size of the largest pieces of material present.

12.2.5.3 Predictive Emission Factor Equation The emission factor for wind-generated particulate emissions from mixtures of erodible and non-erodible surface material subject to disturbance may be expressed in units of grams per square meter (g/m2) per year as follows:

The particle size multiplier (k) for Equation 2 varies with aerodynamic particle size, as follows:

This distribution of particle size within the under 30 micrometer (µm) fraction is comparable to the distributions reported for other fugitive dust sources where wind speed is a factor. This is illustrated, for example, in the distributions for batch and continuous drop operations encompassing a number of test aggregate materials (see Section 13.2.4). In calculating emission factors, each area of an erodible surface that is subject to a different frequency of disturbance should be treated separately. For a surface disturbed daily, N = 365 per year, and for a surface disturbance once every 6 months, N = 2 per year.



The erosion potential function for a dry, exposed surface is:

Because of the nonlinear form of the erosion potential function, each erosion event must be treated separately. Equations 2 and 3 apply only to dry, exposed materials with limited erosion potential. The resulting calculation is valid only for a time period as long or longer than the period between disturbances. Calculated emissions represent intermittent events and should not be input directly into dispersion models that assume steady-state emission rates. For un-crusted surfaces, the threshold friction velocity is best estimated from the dry aggregate structure of the soil. A simple hand sieving test of surface soil can be used to determine the mode of the surface aggregate size distribution by inspection of relative sieve catch amounts, following the procedure described below. FIELD PROCEDURE FOR DETERMINATION OF THRESHOLD FRICTION VELOCITY (from a 1952 laboratory procedure published by W. S. Chepil): 1. Prepare a nest of sieves with the following openings: 4 mm, 2 mm, 1 mm, 0.5 mm, and 0.25 mm. Place a collector pan below the bottom (0.25 mm) sieve. 2. Collect a sample representing the surface layer of loose particles (approximately 1 cm in depth, for an encrusted surface), removing any rocks larger than about 1 cm in average physical diameter. The area to be sampled should be not less than 30 cm by 30 cm.

3. Pour the sample into the top sieve (4-mm opening), and place a lid on the top.



4. Move the covered sieve/pan unit by hand, using a broad circular arm motion in the horizontal plane. Complete 20 circular movements at a speed just necessary to achieve some relative horizontal motion between the sieve and the particles. 5. Inspect the relative quantities of catch within each sieve, and determine where the mode in the aggregate size distribution lies, i. e., between the opening size of the sieve with the largest catch and the opening size of the next largest sieve. 6. Determine the threshold friction velocity from Table 13.2.5-1. The results of the sieving can be interpreted using Table 13.2.5-1. Alternatively, the threshold friction velocity for erosion can be determined from the mode of the aggregate size distribution using the graphical relationship described by Gillette.5-6 If the surface material contains non-erodible elements that are too large to include in the sieving (i. e., greater than about 1 cm in diameter), the effect of the elements must be taken into account by increasing the threshold friction velocity.

Threshold friction velocities for several surface types have been determined by field measurements with a portable wind tunnel. These values are presented in Table 12.2.5-2.



The fastest mile of wind for the periods between disturbances may be obtained from the monthly LCD summaries for the nearest reporting weather station that is representative of the site in question.7 These summaries report actual fastest mile values for each day of a given month. Because the erosion potential is a highly nonlinear function of the fastest mile, mean values of the fastest mile are inappropriate. The anemometer heights of reporting weather stations are found in Reference 8, and should be corrected to a 10-m reference height using Equation 1.

To convert the fastest mile of wind (u+) from a reference anemometer height of 10 m to the equivalent friction velocity (u*), the logarithmic wind speed profile may be used to yield the following equation:

This assumes a typical roughness height of 0.5 cm for open terrain. Equation 4 is restricted to large relatively flat piles or exposed areas with little penetration into the surface wind layer. If the pile significantly penetrates the surface wind layer (i.e., with a height-to-base ratio exceeding 0.2), it is necessary to divide the pile area into subareas representing different degrees of exposure to wind. The results of physical modeling show that the frontal face of an elevated pile is exposed to wind speeds of the same order as the approach wind speed at the top of the pile.



For 2 representative pile shapes (conical and oval with flattop, 37-degree side slope), the ratios of surface wind speed (us) to approach wind speed (ur) have been derived from wind tunnel studies.9 The results are shown in Figure 13.2.5-2 corresponding to an actual pile height of 11 m, a reference (upwind) anemometer height of 10 m, and a pile surface roughness height (zo) of 0.5 cm. The measured surface winds correspond to a height of 25 cm above the surface. The area fraction within each contour pair is specified in Table 13.2.5-3.



The profiles of U s/U r in Figure 13.2.5-2 can be used to estimate the surface friction velocity distribution around similarly shaped piles, using the following procedure: 1. Correct the fastest mile value (u+) for the period of interest from the anemometer height (z) to a reference height of 10 m 10 u+ 4 using a variation of Equation 1:

2. Use the appropriate part of Figure 13.2.5-2 based on the pile shape and orientation to the fastest mile of wind, to obtain the corresponding surface wind speed distribution

3. For any subarea of the pile surface having a narrow range of surface wind speed, use a variation of Equation 1 to calculate the equivalent friction velocity (u*):

From this point on, the procedure is identical to that used for a flat pile, as described above.



Implementation of the above procedure is carried out in the following steps: 1. Determine threshold friction velocity for erodible material of interest (see Table 13.2.5-2 or determine from mode of aggregate size distribution). 2. Divide the exposed surface area into subareas of constant frequency of disturbance (N). 3. Tabulate fastest mile values (u+) for each frequency of disturbance and correct them to 10 m (u10) using Equation 5.

4. Convert fastest mile values (u10) to equivalent friction velocities (u*), taking into account (a) the uniform wind exposure of non- elevated surfaces, using Equation 4, or (b) the non-uniform wind exposure of elevated surfaces (piles), using Equations 6 and 7. 5. For elevated surfaces (piles), subdivide areas of constant N into subareas of constant u* (i. e., within the isopleth values of us/ur in Figure 13.2.5-2 and Table 13.2.5-3) and determine the size of each subarea. 6. Treating each subarea (of constant N and u*) as a separate source, calculate the erosion potential (Pi) for each period between disturbances using Equation 3 and the emission factor using Equation 2. 7. Multiply the resulting emission factor for each subarea by the size of the subarea, and add the emission contributions of all subareas. Note that the highest 24-hour (hr) emissions would be expected to occur on the windiest day of the year. Maximum emissions are calculated assuming a single event with the highest fastest mile value for the annual period. The recommended emission factor equation presented above assumes that all of the erosion potential corresponding to the fastest mile of wind is lost during the period between disturbances. Because the fastest mile event typically lasts only about 2 minutes, which corresponds roughly to the half-life for the decay of actual erosion potential, it could be argued that the emission factor overestimates particulate emissions.



However, there are other aspects of the wind erosion process that offset this apparent conservatism: 1. The fastest mile event contains peak winds that substantially exceed the mean value for the event. 2. Whenever the fastest mile event occurs, there are usually a number of periods of slightly lower mean wind speed that contain peak gusts of the same order as the fastest mile wind speed. Of greater concern is the likelihood of over prediction of wind erosion emissions in the case of surfaces disturbed infrequently in comparison to the rate of crust formation.

12.2.5.4 Example 1: Calculation for wind erosion emissions from conically shaped coal pile

A coal burning facility maintains a conically shaped surge pile 11 m in height and 29.2 m in base diameter, containing about 2000 megagrams (Mg) of coal, with a bulk density of 800 kilograms per cubic meter (kg/m3) (50 pounds per cubic feet [lb/ft3]). The total exposed surface area of the pile is calculated as follows:

Coal is added to the pile by means of a fixed stacker and reclaimed by front-end loaders operating at the base of the pile on the downwind side. In addition, every 3 days 250 Mg (12.5 percent of the stored capacity of coal) is added back to the pile by a topping off operation, thereby restoring the full capacity of the pile. It is assumed that (a) the reclaiming operation disturbs only a limited portion of the surface area where the daily activity is occurring, such that the remainder of the pile surface remains intact, and (b) the topping off operation creates a fresh surface on the entire pile while restoring its original shape in the area depleted by daily reclaiming activity. Because of the high frequency of disturbance of the pile, a large number of calculations must be made to determine each contribution to the total annual wind erosion emissions.



This illustration will use a single month as an example. Step 1: In the absence of field data for estimating the threshold friction velocity, a value of 1.12 m/s is obtained from Table 13.2.5-2. Step 2: Except for a small area near the base of the pile (see Figure 13.2.5-3), the entire pile surface is disturbed every 3 days, corresponding to a value of N = 120 per year. It will be shown that the contribution of the area where daily activity occurs is negligible so that it does not need to be treated separately in the calculations.



Figure 12.2.5-3. Example 1: Pile surface areas within each wind speed regime. Step 3: The calculation procedure involves determination of the fastest mile for each period of disturbance. Figure 12.2.5-4 shows a representative set of values (for a 1-month period) that are assumed to be applicable to the geographic area of the pile location. The values have been separated into 3-day periods, and the highest value in each period is indicated. In this example, the anemometer height is 7 m, so that a height correction to 10 m is needed for the fastest mile values. From Equation 5,

Figure 12.2.5-4. Example daily fastest miles wind for periods of interest.



Step 4: The next step is to convert the fastest mile value for each 3-day period into the equivalent friction velocities for each surface wind regime (i. e., us/ur ratio) of the pile, using Equations 6 and 7. Figure 12.2.5-3 shows the surface wind speed pattern (expressed as a fraction of the approach wind speed at a height of 10 m). The surface areas lying within each wind speed regime are tabulated below the figure. The calculated friction velocities are presented in Table 12.2.5-4. As indicated, only 3 of the periods contain a friction velocity which exceeds the threshold value of 1.12 m/s for an un-crusted coal pile. These 3 values all occur within the Us/Ur = 0.9 regime of the pile surface.

Step 5: This step is not necessary because there is only 1 frequency of disturbance used in the calculations. It is clear that the small area of daily disturbance (which lies entirely within the Us/Ur = 0.2 regime) is never subject to wind speeds exceeding the threshold value. Steps 6 and 7: The final set of calculations (shown in Table 13.2.5-5) involves the tabulation and summation of emissions for each disturbance period and for the affected subarea. The erosion potential (P) is calculated from Equation 3.



For example, the calculation for the second 3-day period is:

The emissions of particulate matter greater than 10 mm (PM-10) generated by each event are found as the product of the PM-10 multiplier (k = 0.5), the erosion potential (P), and the affected area of the pile (A). As shown in Table 12.2.5-5, the results of these calculations indicate a monthly PM-10 emission total of 780 g.

12.2.5.5 Example 2: Calculation for wind erosion from flat area covered with coal dust

A flat circular area 29.2 m in diameter is covered with coal dust left over from the total reclaiming of a conical coal pile described in the example above. The total exposed surface area is calculated as follows:

This area will remain exposed for a period of 1 month when a new pile will be formed. Step 1: In the absence of field data for estimating the threshold friction velocity, a value of 0.54 m/s is obtained from Table 13.2.5-2. Step 2: The entire surface area is exposed for a period of 1 month after removal of a pile and N = 1/yr.



Step 3: From Figure 12.2.5-4, the highest value of fastest mile for the 30-day period (31 mph) occurs on the 11th day of the period. In this example, the reference anemometer height is 7 m, so that a height correction is needed for the fastest mile value. From Step 3 of the previous example, u+ 10 = 1.05 u+ 7, so that u+ 10 = 33 mph. Step 4: Equation 4 is used to convert the fastest mile value of 14.6 m/s (33 mph) to an equivalent friction velocity of 0.77 m/s. This value exceeds the threshold friction velocity from Step 1 so that erosion does occur. Step 5: This step is not necessary, because there is only 1 frequency of disturbance for the entire source area. Steps 6 and 7: The PM-10 emissions generated by the erosion event are calculated as the product of the PM-10 multiplier (k = 0.5), the erosion potential (P) and the source area (A). The erosion potential is calculated from Equation 3 as follows:

References For Section 13.2.5 1. C. Cowherd, Jr., "A New Approach To Estimating Wind Generated Emissions From Coal Storage Piles", Presented at the APCA Specialty Conference on Fugitive Dust Issues in the Coal Use Cycle, Pittsburgh, PA, April 1983. 2. K. Axtell and C. Cowherd, Jr., Improved Emission Factors For Fugitive Dust From Surface Coal Mining Sources, EPA-600/7-84-048, U. S. Environmental Protection Agency, Cincinnati, OH, March 1984.



3. G. E Muleski, "Coal Yard Wind Erosion Measurement", Midwest Research Institute, Kansas City, MO, March 1985. 4. Update Of Fugitive Dust Emissions Factors In AP-42 Section 11.2 — Wind Erosion, MRI No. 8985-K, Midwest Research Institute, Kansas City, MO, 1988. 5. W. S. Chepil, "Improved Rotary Sieve For Measuring State And Stability Of Dry Soil Structure", Soil Science Society Of America Proceedings, 16:113-117, 1952. 6. D. A. Gillette, et al., "Threshold Velocities For Input Of Soil Particles Into The Air By Desert Soils", Journal Of Geophysical Research, 5(C10):5621-5630. 7. Local Climatological Data, National Climatic Center, Asheville, NC. 8. M. J. Changery, National Wind Data Index Final Report, HCO/T1041-01 UC-60, National Climatic Center, Asheville, NC, December 1978. 9. B. J. B. Stunder and S. P. S. Arya, "Windbreak Effectiveness For Storage Pile Fugitive Dust Control: A Wind Tunnel Study", Journal Of The Air Pollution Control Association, 38:135-143, 1988. 10. C. Cowherd, Jr., et al., Control Of Open Fugitive Dust Sources, EPA 450/3-88- 008, U. S. Environmental Protection Agency, Research Triangle Park, NC, September 1988.



APPENDIX B TABLE OF PROPERTY VALUES FOR SPECIFIC CHEMICALS



DOCUMENTS REFERRED TO IN THIS GUIDE Legislation Environment Act 1995 (referred to in Part 2 - 2.2.7). Code of Federal Regulations CFR 40 : Part 68 Protection of Environment (referred to in Appendix A). US EPA Reports No. EPA-68-D9-0173, 1992 (referred to in Part 4 - 4.2.3; Part 13 - Box 2). No. EPA-600/7-80-036 (referred to in Part 8 - 8.3). No. EPA/600R-92/047 (referred to in Part 15). European Union Directives 80/779 Air Quality Limit Values and Guide Values for Sulfur Dioxide and Suspended Particulates (referred to in Part 2 - 2.3). 82/884 Title Unknown (referred to in Part 2 - 2.3). 85/203 Title Unknown (referred to in Part 2 - 2.3). HSE Documentation Technical Guidance Note E1 : Guidance for Operators and Inspections of IPC Processes (referred to in Part 2 - 2.4; Part 4 - 4.2; Part 15). EH40/98 : Occupational Exposure Limits (referred to in Part 2 - 2.3; Part 4 - 4.2.3 and Appendix C). HMIP Documentation Technical Guidance Note D1 : Dispersion (referred to in Part 4 - 4.2.5; Part 15).

gas dispersion - a definitive guide to accidental releases of heavy gases

Technology

process safety guide

atmospheric stability

pasquill gifford methods

gas dispersion contents

situ exsitu sulfiding

air quality standards

plume dispersion

definitive guide