4.1 Where does it come from?
Models attempt to simulate a real situation using mathematics. The types of physical situations that result in the release of pollutants to the atmosphere are classified as types of 'sources', with user-specified characteristics. This section details the criteria used by the models to describe the characteristics of the sources being modelled.
4.1.1 Physical characteristics
There are four main types of sources used in dispersion modelling (Figure 4.1):
- point source - discharges from a small opening such as a stack or vent
- area source - a source with a large surface area, such as a landfill surface, contaminated site, a pile of solid material, or a liquid surface (pond, tank, lagoon)
- line source - a long, narrow source such as a roadway, conveyor belt, or roofline vent along a long, narrow building (usually a line source must be redefined as a chain of volume sources for modelling)
- volume source - bulky, diffuse source such as emissions from within a building.
Some models (e.g. TAPM) allow various types of gridded emission sources to represent complex emission inventories, and can include emissions from many different anthropogenic and biogenic/natural sources.
The ISCST3 model also includes a special algorithm for modelling concentration and deposition impacts of particle emissions from open pit sources, such as surface coal mines or rock quarries.
a Point sources
Point sources require the specification of emission temperature and exit velocity in addition to emission rates. The importance of exit velocity is discussed in section 4.1.4. AUSPLUME requires temperature data in degrees Celsius, and ISCST3, CALPUFF, AERMOD and TAPM require temperature data in Kelvins.
If you want the emission temperature to be the same as the ambient temperature in the meteorological data file, as when your stack is discharging building ventilation air at the same temperature as the atmosphere, then in ISCST3 set the emission temperature to 0 K, as this defaults the model to use the ambient temperature. In AUSPLUME, if the emission temperature is below the ambient temperature, the model will round the emission temperature up to the ambient temperature. So if you want the emission to discharge at ambient temperature, set the emission temperature to be lower than the lowest ambient temperature in the meteorological file. However, this will cause an error if the ambient temperature is lower than 0°C as the model cannot handle an emission temperature lower than 0°C.
Special care must be taken when modelling sources that are not point sources. Figure 4.1 illustrates a number of different source types. The model will specify the source as a release of plumes or puffs, which have Gaussian concentration distributions. The model will automatically assign 'effective' locations of the initials plumes/puffs and their horizontal and vertical spreads - these are not necessarily the actual location or size of the physical source (but are, of course, closely related). This means that, in general, modelled concentrations close to the source are not realistic, as the true source is represented as a series of Gaussian-shaped plumes or puffs. To circumvent this, and obtain realistic near-source concentrations, the physical source should be divided into smaller elements. For instance, a roadway is divided into segments, each represented by a point source in the Gaussian-plume model CALINE-4 and similar models.
b Area sources
Some models assume that individual area sources have square dimensions. In practice, sources are usually irregular in shape. You can approximate these irregular shapes by using an appropriate selection of a number of square area sources. However, AUSPLUME 5, ISCST3 and CALPUFF also allow area sources to be rectangular, circular or polygons with irregular angles. The rectangular area sources can be rotated to any angle rather than oriented parallel to a north-south grid.
For square area sources, the models handle area sources as a line source that rotates to maintain its crosswind orientation. In this sense, the area source more closely resembles a circular source than a square. The initial vertical dispersion is calculated as if it were a distance L metres upwind (where L is the area source side length) in order to simulate the diffusion that occurs as the pollutant moves across the source. Receptors should not be located closer to an area source than the area source's width (VicEPA, 2000). Estimating concentrations close to the source may require several smaller sources to be used to represent the actual source.
The models use a numerical integration approach for modelling impacts from non-square area sources. While this option is slower, it allows more exact estimates to be made close to the source (VicEPA, 2000).
CALPUFF allows for thermally buoyant area sources. This may be significant if the area source is much hotter than the ambient temperature, such as a tank containing hot trade wastes from an industrial process or a surface fire. Although AUSPLUME and ISCST3 do not calculate thermal buoyancy effects from area sources, it is possible to estimate the effective release height using standard buoyancy flux equations. The model may then be run with the source at the effective height and at actual height to give an envelope of concentrations. (Note this treatment is only applicable for small area sources. Large, hot area sources such as open burning, detonation of explosives and/or bushfires should be addressed using source-specific models. For more information, refer to the US EPA SCRAM website www.epa.gov/scram001/). In most cases however, the area source will be at a temperature close to ambient, and thermal buoyancy effects will be negligible.
Terrain effects and building downwash effects are not computed for area sources in either AUSPLUME, ISCST3, CALPUFF or TAPM.
c Line sources
Line sources are not modelled by AUSPLUME and ISCST3, and must instead be treated as volume sources (see section 4.2.1d). In AERMOD, line sources are represented as a string of volume sources or elongated area sources. In CALINE-4, line sources are divided into strings of point sources. Buoyant line sources are modelled by CALPUFF and TAPM.
d Volume sources
The size of volume sources is specified using initial horizontal and vertical spreads. These depend on the type of volume source and what the source is representing (such as a ridge-line vent), and can be obtained from tables in the user's guide or online help system for each model.
A common application of a volume source is to model the fugitive emissions escaping from a building. Here, the initial vertical and horizontal plume spreads may be taken as a quarter of the building height and width respectively (VicEPA, 2000). This assumes that the bulk of the material (about 86%) is captured in the building wake. To be conservative, the building width should be taken as the minimum of the building dimensions. In this example, the model creates a virtual source so the emissions are released at half building height.
To model a line source, a row (or arc) of volume sources must be arranged along the centreline of the real line source with separations less than about one-quarter of the distance to the nearest receptor. Results will then be insensitive to the value chosen for the horizontal spread of the volume sources, although a recommended distance is half of the separation between volume sources. The vertical spread should be set as one-quarter of the line source's height. Since line sources adjacent to a building are affected by the building wake, the volume sources should be located along the building centreline with initial vertical spreads equal to one-quarter of the building height.
Building downwash effects are not computed for volume sources in either AUSPLUME, ISCST3, CALPUFF or TAPM. Terrain effects are computed for volume sources.
Lagrangian particle models represent pollutant releases as a stream of particles. Since the model particles have no physical dimensions, source types may be specified to have any shape and size, and the emitted particles may be distributed over an arbitrary line, area or volume.
Select the most appropriate source type offered by the model. Justify the source type selection if it is not obvious.
Use measured (or planned) dimensions and characteristics to describe the source.
If the dimensions and characteristics of the source are not accurately known or have not been calculated, run both a worst-case scenario (i.e. the configuration of source that results in highest ground-level concentration) and the most likely realistic dimensions. Present and compare the results of both scenarios.
Consider using an alternative model, or even a different form of assessment, if a particular source does not fall easily into one of the source types that a particular model is designed to handle.
4.1.2 Emission rates: the number one critical parameter
Emission rates can be a major source of error and inconsistency in any modelling analysis. For inert pollutants, the modelled concentration is directly proportional to the emission rate, so any errors in the emission rate data translate directly into errors in the model results. It is therefore important to use emission rates that are as accurate as possible.
Sources of emission data include:
- measurements from a particular (or similar) source
- manufacturer's specifications or process information
- published data (e.g. US EPA's AP-42 database)
- regulatory authority files and data
- calculated emissions from emission models (e.g. NZ-TER).
Emission rate data should ideally be sourced from measurements undertaken at either the site in question (for an existing site) or a similar site (if available). A significant number of emission tests have been undertaken at different sites in New Zealand and so there is a substantial amount of local emission data available. Regional councils are probably the best source of this data.
Alternatively emission rates may be calculated from manufacturer's specifications or directly using industrial process knowledge.
When no appropriate measured emission rates are available, published emission factors can be useful. Published emission factors give the mass of pollutants discharged per mass of fuel consumed, or product processed, and are useful as a first estimate of emission rates for pollutants where collection of actual emission rate measurements is impractical or impossible. Examples of these include the US EPA's AP-42 Emission Factors (www.epa.gov/ttn/chief/) or the UK Emission Factors Database (www.naei.org.uk/emissions/). Australia employed a set of 82 calculation handbooks for the National Pollutant Inventory, largely based upon US EPA emission factors. These handbooks are available at www.npi.gov.au.
New Zealand has its own database for calculating transport emissions. This is the NZ-Traffic Emission Rates (NZ-TER), and is available at http://www.mot.govt.nz/publications/14_26doc.shtml. This is discussed further in section 3.2.1
Emission factors are typically assigned a rating to reflect their accuracy. The accuracy rating and the applicability of any emission factor to New Zealand conditions must be considered and should be discussed when used in any study.
Consider using a different form of assessment if the emission rate from a particular source cannot be confidently measured or accurately calculated.
For simple dispersion model scenarios with only one or two sources, the maximum measured emission rate from the source(s), or the calculated emission rate corresponding to maximum production or fuel burning rate, is typically used for dispersion calculations.
Special care is required where there are several sources of a pollutant on one industrial site. In most cases all discharges are unlikely to emit their highest emission rates simultaneously, resulting in high ground-level concentrations. In such cases, it is more realistic to model all the sources together using some kind of average or variable emission rate for each source (section 4.1.3). Each source should also be modelled individually using its worst-case maximum emission rate to determine its maximum potential effect. If a few sources are likely to discharge worst-case emissions at the same time, such as in the case of a treatment system upset at a sewage plant or a cyclone failure at a processing plant, then all of those sources should be modelled together at their maximum emission rates. If this is the case, be realistic about the number of failures that are actually likely to coincide and provide an approximation of how frequent the failure events are likely to be.
When dealing with measured emission rates from a source with multiple measurements over a period of time (e.g. a vent monitored on a number of occasions):
a) use the maximum measurement if only a few data points are available (e.g. fewer than 4-5)
b) if a number of data points are available (e.g. greater than 5-10), then draw a curve and use a certain percentile (such as 50th or 70th percentile) emission rate (the percentile should be selected on a case-by-case basis after examination of data). Vic EPA (1985) provides guidance on what percentile values may be useful.
When dealing with measured emission rates from a large area source, measurements should be taken at a number of locations over the surface in a short space of time, and a mean calculated which represents the average emission over the surface at that point in time.
When dealing with an area source that contains separate zones of different emission rates, each zone should be measured as a separate source and then modelled as such.
4.1.3 Variable emission rates
If your sources exhibit variable emission rates, either over short- or longer-term periods, it may be important to consider programming the model to simulate the variation in the emissions. This would reduce the possibility of the model over-predicting long-term (longer than one-hour) averages, which could occur if the maximum emission rate was assumed to apply for 24 hours per day, 365 days per year.
There are a number of ways in which emission rates may vary. Some of the processes that can drive the need to use variable emission rates are when:
- a process does not operate 24 hours per day
- the rate of process (e.g. rate of combustion or production) varies throughout the day
- when the emission rate (e.g. odour or dust) varies with temperature or season
- the emission is from a large area source where the emission rate varies over the surface
- the emission is from a large liquid area source such as an oxidation pond, where the emission varies with wind speed over the surface.
AUSPLUME, ISCST3 CALPUFF and TAPM allow you to vary the emission rates by:
- hour of day
- hour and season
- wind speed and stability category
- ambient temperature
- hour by hour variation read from an external file.
If you are modelling the discharge from a combustion source such as a coal boiler, where the combustion rate varies throughout the day, remember that emission rates may not decline in proportion with a decline in combustion rate as the combustion efficiency changes, and at lower combustion rates the efflux velocity and temperature will also change. These factors all have an effect on the rate of dispersion from the stack and should be accounted for (see section 4.1.4).
For some processes, emission rates may vary substantially in a random fashion. An example would be a waste combustion process in which emission spikes of sulphur dioxide or hydrogen chloride will occur unpredictably because of the variable sulphur and chlorine content of the waste fuel. In such situations, using maximum emission rates to estimate maximum short-term concentrations (e.g. hourly averages) is likely to result in gross over-estimation, because there is only a small chance of maximum emissions coinciding with worst-case meteorology. However, it is possible to use probabilistic methods that take into account the frequency distributions of predicted concentrations and emission rates (measured or estimated) to assess the likelihood of maximum emission rates coinciding with unfavourable meteorological conditions.
Detailed descriptions of the application of probabilistic methods to account for the effect of varying emissions rates can be found in Amr Abdel-Aziz and Frey (2003a, 2003b).
Use variable emission rate data when:
a) there is evidence that shows how much and how often the emission rate will vary from the maximum potential emission rate as operational conditions change
b) assessing average ground-level concentrations for periods longer than the time that maximum emissions actually occur for
c) assessing the actual rather than the potential frequency of pollution events.
When using variable emission rates, account for other factors that also vary with operational conditions, such as lowered efflux velocities and temperatures.
4.1.4 Exhaust flow - influence of stack hoods
Stacks are modelled as point sources. Parameters that control plume rise, such as the initial vertical momentum and thermal buoyancy of the plume (section 4.3.2), are calculated from the characteristics of the stack emissions, in particular the temperature and exit velocity. If the exit of the stack is not pointed in a vertical direction, and unimpeded by any obstruction, the exit velocity used by the model will be different to the gas velocity within the stack itself. The exit velocity used by a model in this case is the vertical component of the exhaust flow (see Figure 4.3).
An example of the effects of stack exit velocity on dispersion is shown in Figure 4.2. The improvement in dispersion diminishes with increasing exit velocity, and the amount of improvement will vary depending on the stack height. When designing a stack, it is necessary to optimise both the stack height and the exit velocity.
If you have a stack which discharges vertically into the air with no impediment to the vertical movement of the exhaust gases, then the exit velocity for the model is simply the volumetric flow of gas (at discharge temperature) divided by the stack-tip diameter. Diameter-reducing cones fitted to the stack tip can be used to increase the exit velocity, although in practice this may not achieve the proportional increase in velocity that you expect. This is because of the increased flow resistance and pressure drop generated by the smaller exit diameter. If the process generating the discharge has fans upstream, the change in pressure drop may change the performance and flow rate of these fans, and bigger fans may be needed. If the process does not have any fans, the change in pressure drop will decrease the flow rate out through the stack, and fans may need to be installed.
It is important when modelling discharges from a stack being fitted with a diameter-reducing cone to account for the increase in resistance and pressure drop (resulting in decreased flow). For example, it would be incorrect to assume that the flows shown in the first two stacks in Figure 4.3 are equal (unless fans were introduced to the second stack). It requires the expertise of process or ventilation engineers to accurately calculate the effect of the introduction of a diameter-reducing cone on exit velocities.
The relationship between fan selection, flow rate and pressure drop is not simple. If you are faced with this situation it will be helpful to seek advice from someone with qualifications and experience in this area.
If you have a stack that discharges exhaust gases at an angle to the vertical, you must calculate the vertical component of the flow vector and use that component as the exit velocity in the model (Figure 4.3).
If the stack has a fitting such as a 'chinese hat' to prevent rain ingress, or discharges horizontally or at an angle to the ground, the effective vertical exit velocity is zero. This can have a large effect by reducing the rate of dilution and dispersion. For this reason many regional councils require that this type of rain hat be removed from stacks.
It should be noted that the models do not account for significant sideways or downward momentum from the stack, such as could occur if the discharge is sideways or deflected down by the stack geometry. This will influence ground-level concentrations, particularly for receptors close to the source. This additional uncertainty in the model results should be acknowledged if the results are to be presented in a report. Because of the relatively large uncertainties associated with modelling the dispersion of contaminants from this type of source, it is not advisable to rely on the accuracy of the results. More realistic results may be obtained if the modelling is undertaken using a volume source.
The vertical efflux velocity of a gas stream leaving a stack must be adjusted to account for the influence of any object that may increase resistance to, or change the direction, of the flow.
4.1.5 Multiple sources
Most dispersion models allow you to enter a large number of sources at a time and model the ground-level concentration from each source, either together, separately or in subgroups (called 'source groups'). The models assume that if source A causes a concentration XA at a particular receptor, and source B causes a concentration XB at the same receptor, then the combined ground-level concentration at that receptor is XA + XB. This assumption is valid for pollutants where mass is conserved, such as particles or hydrogen sulphide, but is not valid for odours. More discussion on the applicability of this assumption for odours can be found in the Ministry for the Environment's Odour Management Guide (Ministry for the Environment, 2002b).
To improve the quality of modelling results, multiple sources can be used to break up large area sources into smaller sources, such as for the purpose of defining parts of the area source with different emission rates, or for modelling a long chicken shed with a roof vent along its length as a chain of volume sources.
The time required by a computer to carry out the dispersion calculations increases in proportion to the number of sources. In some cases, if you wish to reduce the calculation time, small multiple sources can be grouped and modelled as one larger source. This can be particularly useful if you are modelling a large number of sources and the ones you wish to group together do not dominate the major individual sources (e.g. multiple clarifiers at a sewage plant). In doing so, you must check that your receptors of interest are far enough away from the source so that the model results are not affected by the grouping.
If running a screening model with multiple sources, you must:
a) use a number of screening meteorological data input files that contain different wind directions to determine which will generate the highest concentrations
b) ensure you account for the potential effect of building downwash.
A example of the effect that different source and wind alignments can have on ground-level concentrations is shown in Figure 4.4.
A stack containing multiple flues, or multiple stacks close together, will have enhanced buoyancy and a higher plume rise, and will therefore generate lower ground-level concentrations than would be the case if the flues or stacks were modelled as separate sources. The effect of enhanced plume rise due to multiple stacks was described and then modelled by Briggs (1975) using algorithms. Manins et al. (1992) demonstrated that plume rise could be enhanced by 10 to 45% by the presence of multiple stacks. Therefore, when modelling a source with a number of stacks in close proximity, unless the effects of enhanced buoyancy are accounted for the ground-level concentrations may be over-predicted. The effect of enhanced plume rise appears to be greatest with large sources such as power plant cooling towers and fossil-fuelled electricity plants. The effect may not be so significant with small or medium-sized sources. Approaches to account for enhanced plume rise have been described by Bornoff et al. (2001), Kong et al. (2002) and Overcamp and Ku (1988).
When modelling multiple flues, or multiple stacks that are close together, take account of the effect of enhanced plume buoyancy.
In some cases, the flues may have similar diameters and flow rates. In this case, if the flues are adjacent to one another (within three diameters), enter a stack having the same effective cross-sectional area as the sum of the individual flue cross-sectional areas. Calculate an effective exit velocity though the single flue that gives the equivalent volumetric flow rate to the combined flues.
4.2 Where does it go to?
All dispersion models require the specification of the co-ordinates downwind from a source where the ground-level concentrations are to be recorded. The grid of receptors can be an evenly or unevenly spaced Cartesian or polar grid (Figure 4.5).
Guidelines for defining the extent, spacing, and elevation of a grid are covered in sections 4.2.1 to 4.2.3. Dealing with complex terrain and surface roughness is discussed in section 4.2.4.
4.2.1 Grid extent
The extent of the grid should be chosen to include any regions of sensitive or important receptors such as residential areas, and should also be sufficiently large to capture peak downwind pollutant predictions. For sources emitting pollutants close to ground level, the maximum ground-level concentration will be close to the source. However, for stack sources the maximum ground-level concentration can be some distance away, and the model may have to be run more than once with increasing grid ranges to make sure the peak is captured.
If the model is being run in screening mode with a single wind direction, the grid extent can be reduced by removing any receptors upwind of the source(s). In addition, if the sources are in line with the wind direction blowing either north-south or east-west, the grid extent can be further simplified by removing all receptors in the cross-wind direction except the one directly downwind of the plume centreline (Figure 4.6). Graphical results plotted from a simplified grid such as this will have a two-dimensional format.
If you are using a Gaussian-plume model, you should note the guidelines for range applicability of Gaussian-plume models (section 4.3.1) when defining your grid extent.
The extent of the grid must be chosen to include any regions of sensitive or important receptors (e.g. residential areas, schools and hospitals). The spacing and extent should also be configured to capture peak downwind ground-level concentrations.
4.2.2 Grid spacing
Selecting the spacing between individual receptor points is a compromise between processing time and the required results resolution. If you double the number of receptors in your grid, the processing time will also double. However, if the spacing is too large, the peak ground-level concentration may fall between two receptor points and not be captured in the results file.
Check that the grid spacing is small enough by running the model with increasingly smaller grid spacings near the location of the peak ground-level concentration, until halving the grid spacing effects a change in peak ground-level concentration of less than 10%.
Irregularly spaced grids can be used to concentrate the number of receptors close to the location of the peak ground-level concentration. If the run-time or memory requirements become too great when a grid is subdivided, one solution is to reduce the domain to one or more smaller areas over several runs of the model, then merge the results later. Alternatively, run the model on a coarse grid first, then again on a fine grid with a smaller extent but centred on the area of maximum ground-level concentrations.
If you are using a polar grid, to ensure capture of peak ground-level concentrations round off all wind directions in the meteorological data so that they align with the rays of the polar receptor grid. This only works if all the sources are located at or near the origin of the grid so that the plume's centreline always travels directly over a ray of receptors (or almost so). This adjustment procedure will only provide reasonable ground-level concentration predictions if the emission source is not embedded in elevated terrain, and only if one-hour averages are being compared against guideline limits. In other instances, adjusting the orientation of the wind direction to coincide with polar receptor grid will over-estimate the frequency at which the peak concentration will occur in any one location. In these (more complex) situations a better approach would be to increase the number of rays in the receptor grid rather than adjust the meteorological data file.
If you are planning to use SURFER to plot the results of your model (section 6.1.3), set the number of grid points in the 'Grid Data' option within SURFER to the same as your number of receptors. This will ensure that SURFER graphs the modelled results exactly at each receptor, rather than interpolating different numbers if the concentration is changing rapidly at this location.
Take special care when positioning receptors if you have small or elongated area or line sources. A line of receptors may need to be specially placed parallel to and running through the longest dimension of the source, so that the plume centreline from this source is actually picked up by the model (Figure 4.7).
4.2.3 Elevated (flagpole) receptors
Unless otherwise specified by the user, models produce concentrations at ground level. It is standard practice that assessments of pollutant concentrations are conducted at ground level. Some modellers use concentrations that occur at the (approximate) height of a person (e.g. 1.8 m). In reality there is unlikely to be a significant gradient between the concentrations occurring at 1.8 m and at ground level. For the sake of consistency it is recommended that unless specific circumstances require otherwise, assessments are conducted at ground level.
Flagpole receptors are available in AUSPLUME, ISCST3 and CALPUFF.
Use flagpole receptors when there is a requirement to calculate pollutant concentrations at some height above the ground.
An example of where you might use flagpole receptors is when you are modelling pollutant emissions in an area where there are tall buildings that have air conditioning intake ducts on their roofs, or where upper-storey windows may be opened for ventilation. The flagpole receptor option may also be used to look at changes in plume rise and vertical spread downwind from the source. Two examples of the use of flagpole receptors are shown in Figure 4.8.
Flagpole receptors should be distinguished from elevated receptors that are at ground level but at a higher elevation than the base of the source due to changes in terrain. Ground-level concentrations for the latter type of elevated source are calculated by a different method to those for flagpole receptors, so you should be careful to specify the nature of the elevation of your receptors in the correct way in the model.
4.2.4 Complex terrain
Changes in terrain around an air discharge source can significantly affect the pattern of dispersion of the discharge plume. Steady-state Gaussian models like AUSPLUME and ISCST3 contain simple algorithms to attempt to account for the effects of terrain in a limited fashion. CALPUFF contains much more sophisticated procedures for modelling the effects of terrain, with correspondingly greater effort required by the modeller to specify the terrain data.
TAPM handles complex terrain by using a terrain-following co-ordinate system and solving fundamental equations for wind, temperature, moisture, rain, turbulence and pollution dispersion. Using TAPM in complex terrain is relatively easy, as TAPM comes with an easy-to-use GUI that allows users to configure a simulation for any region using provided data sets of terrain, land use and synoptic-scale meteorology. User-defined data sets can also be provided to TAPM to allow more detailed local data to be used if available.
Because of the assumption that the wind speed and direction remain constant over the full length of the plume, Gaussian-plume models like AUSPLUME can only partially simulate terrain effects (VicEPA, 2000). Complex terrain may produce wind channelling around or between hills and ranges, especially under stable atmospheric conditions.
In complex terrain, Gaussian-plume models should only be used to provide a screening assessment and then only for impacts on terrain features that are adjacent to the source.
The limitations of Gaussian-plume models to accurately simulate dispersion in complex terrain must be acknowledged and accounted for in any assessment where terrain may influence dispersion, or where receptors are located on topographical features.
If terrain is potentially significant to a simulation then consider using CTDM, CTSCREEN or an advanced model. Justify the choice of model if a steady-state one is used.
However, if local terrain is not significant and a decision has been made to use AUSPLUME or ISCST3, sensitivity studies (NIWA, 1998) suggest that:
- terrain more than 10% of stack height and in the range 5-50 m should not be ignored
- for terrain heights smaller than 10% of stack heights a 10% underestimate in peak concentration is likely to occur.
a How does terrain affect plume dispersion?
Local topography can have several influences on plume transport and diffusion (Katestone Scientific, 1998):
- Upwind terrain can alter the wind flow and turbulence characteristics from those measured at the nearest meteorological station. Hills or rough terrain can change wind speeds, directions and turbulence characteristics, and nearby water bodies can considerably dampen turbulence levels.
- Significant valleys can restrict horizontal movement and dispersion and encourage the development and persistence of drainage flows. Night-time values of horizontal turbulence can be considerably reduced.
- Sloping terrain may help to provide katabatic or anabatic flows (i.e. drainage of air down or up hillsides in response to changing vertical temperature profiles).
b How does AUSPLUME handle terrain?
AUSPLUME 5 offers three options for terrain adjustment calculations. These are outlined below, using the descriptions from the AUSPLUME online help system. The differences between these three options are shown in Figure 4.9.
- ISC method (horizontal plume): The simplest terrain correction assumes that the terrain has no influence on the plume height above sea level (i.e. the plume is assumed not to be uplifted at all by the terrain below it). Although called the 'ISC method' in the AUSPLUME help, the terrain approach used by the current version of ISCST3 is more complex than this, as described below in (c).
- Egan half-height approach: In neutral or unstable conditions, a plume will tend to be uplifted by broad terrain features. Under stable conditions, this lifting will generally be less and the plume will pass closer to the face of the hill and may even impact on the surface. On the other hand, plumes passing into a valley will tend to move further from the ground. In both situations the variation in plume centreline height above the local terrain becomes more apparent as atmospheric stability increases. These situations are simulated by allowing the plume axis to remain at the plume stabilisation height above mean sea level under stability categories E and F (stable). For unstable and neutral conditions (categories A to D) the half-height correction factor is used for changes in plume axis height above varying terrain (see Figure 4.9). The plume axis is constrained to be at least 10 m above ground level. This is the preferred terrain correction option.
- Modified Egan approach: The third option is simply a generalisation of the Egan method that allows the user to specify the constant of proportionality of approach for each of the six Pasquill stability classes. This option is only recommended where observational data exist to justify its use.
When using AUSPLUME to assess the impact of discharges on elevated receptors, use the Egan half-height approach as the primary assessment approach.
If using steady-state plume models to assess the impacts of discharges on elevated receptors, present the results from both ISCST3 and modified Egan approaches to provide some indication of the potential uncertainty contained in the assessment.
c How does ISCST3 handle terrain?
Terrain below release height is referred to as simple terrain. Receptors located in simple terrain are modelled with the ISC option described above for AUSPLUME (section 4.3.4b).
ISCST3 incorporates the US EPA's COMPLEX1 screening model algorithms for use with complex terrain (some or all receptors above final plume rise height) and intermediate terrain (terrain located between the release height and the plume height). For receptors located on intermediate terrain, the model will select the higher impact from the simple and complex terrain algorithms on an hour-by-hour, source-by-source and receptor-by-receptor basis. The COMPLEX1 algorithm is based on the Egan half-height method for AUSPLUME (section 4.3.4b), where the plume height relative to the stack base is deflected upwards by an amount equal to half of the terrain height as it passes over complex terrain during unstable and neutral conditions (this is the same in AUSPLUME). The plume height is not deflected by the terrain under stable conditions (in AUSPLUME the plume height is deflected by 35% by the terrain under stable conditions in the default Egan half-height method (Earth Tech, 2001)). However, sector average is used to account for plume deflection around the hill. This approach is not used in AUSPLUME.
The complex terrain screening algorithms apply only to point source and volume source emissions - area source and open pit emission sources are excluded - and do not include building downwash effects.
d How does CALPUFF handle terrain?
A considerably more sophisticated approach that permits the plume to flow both around and over terrain obstacles is used by CALPUFF. In the steep, complex terrain often encountered at sites in New Zealand, the use of CALPUFF is sometimes necessary. However the input and processing resources required to run this three-dimensional non-steady-state model are significantly greater than for either AUSPLUME or ISCST3. CALMET uses terrain and meteorological data to define a wind field for the modelling domain. Figure 4.10 shows an example of a wind field and illustrates how the wind vectors change to reflect the terrain.
CALPUFF then uses this wind field to model the dispersion and movement of the puffs. CALPUFF also uses the same Egan half-height method as AUSPLUME to model the movement of puffs over terrain. A more complicated option, the 'strain-based dispersion adjustment', is also offered. In addition, CALPUFF simulates the interaction of puffs within sub-grid-scale terrain. The sub-grid terrains feature is used when concentration estimates are required at locations on terrain elements not resolved by the grid in the flow-field model. It requires that the terrain feature being considered is identified and described within CALPUFF. Ideally, though, all important terrain features should be resolved by CALMET.
Users wanting more information on these options should refer to the CALPUFF user's guide.
Use an advanced model in an assessment which involves complex terrain only when either:
a) a good quality meteorological file is available; or
b) a good-quality meteorological file can be produced for the site in question.
e How accurate are results when terrain effects are significant?
The more complex the situation a model is required to simulate, the poorer its performance is likely to be. However, some models will handle complex terrain much more realistically than others.
Plume model results must be treated with due caution when terrain effects are significant. There are three inherent limitations with plume models in complex terrain:
- rudimentary treatment of terrain effects on plume lift
- no consideration of causal effects (i.e. the time it takes pollutants to travel to the terrain features), which means that only effects on terrain adjacent and close to the source should be considered
- the straight-line trajectory of the airflow.
Puff models are fundamentally different from plume models. Their non-steady-state nature allows them to account for causal effects and non-straight-line trajectories, so, in principal, puff models will produce more accurate results than plume models when terrain effects are significant. However, in a complex terrain situation, one should never assume an advanced model will overcome all potential problems. It may transpire that the specific issues of a situation determine that none of the available models will provide useful information.
A comparison between ISC3 and CALPUFF showed that in a steady-state environment results were similar (US EPA 1998b). However, in a variable state environment (i.e. complex terrain) there is a trend toward higher concentrations being simulated by CALPUFF. This trend is reversed when the terrain is at greater distances from the source.
4.2.5 Where can you get detailed digital topographical data?
If terrain data are required they can be read from a 1:50,000 scale topographical map for simple cases, although the resulting precision and digital reproduction will not be particularly accurate. A terrain reading will be required for every receptor on the grid, which can be a tedious task. If the terrain detail precision is important, and/or reading from a topographical map will be too time-consuming, then digitised terrain data for New Zealand can be purchased from several agencies such as Terralink International Ltd (http://www.terralink.co.nz/) or GeographX (NZ) Ltd (www.geographx.co.nz). Although the original terrain data are public data and free to all, these agencies charge a licence fee for providing the data in a useable format and for updating the data set.
Terralink International Ltd also provides a new database called the New Zealand Landcover Database, which contains both terrain and land-use data at 100 m resolution. The land-use data are necessary for modelling with CALPUFF.
These data sets are upgraded regularly, and are becoming increasingly easier to access.
Obtain and use terrain data with a resolution that will identify terrain features that are important for the problem under consideration.
4.3 Which buttons do I push?
There are many different ways to configure a model, and variations to the input parameters can have significant effects on a model's results. These include:
- dispersion parameter schemes (section 4.3.1)
- plume rise parameters (section 4.3.2)
- partial plume penetration (section 4.3.2)
- building wake effects (section 4.3.3)
- averaging times (section 4.3.4).
It is important to use the most accurate and appropriate parameters, both for your model and for the particular situation you are trying to simulate. It can be tempting to choose an option that gives you favourable results, without considering how sensitive your model is to that option.
Configure the model to reflect the reality of the situation as closely as possible.
Clearly describe and explain the options chosen in the method.
If unsure which option to use, or if several options could be applied, run the model using each option and present both the most conservative and the realistic set of results.
4.3.1 Dispersion coefficients
Dispersion coefficients are the horizontal and vertical dispersion parameters used to define the rate of dispersion of contaminants in the plume in the horizontal and vertical directions (i.e. plume width and height). In Gaussian-plume/puff models such as AUSPLUME, ISCST3 and CALPUFF, these coefficients are a function of atmospheric stability and distance from the source. In models that use more fundamental equations of atmospheric dispersion, such as TAPM, the dispersion is calculated internally by the model using the predicted meteorology and turbulence.
There are three types of horizontal dispersion coefficients available in AUSPLUME, ISCST3 and CALPUFF: Pasquill-Gifford (P-G), Briggs Rural (B-R), and the standard deviation of wind direction known as sigma-theta (σΘ).
The P-G dispersion curves were developed from 10-minute average experimental data from near ground level, released in flat terrain. The B-R formula was derived from experiments on dispersion from tall stacks (of which there are very few in New Zealand).
For any model, hourly-averaged conditions can lead to over-prediction of concentrations, as the wind direction is assumed constant for each hour and pollutants arrive at the same location for the full period. If this causes pollutant levels of concern, the hour-by-hour meteorological conditions should be checked. If conditions are roughly constant for a number of hours, it is safe to assume that they are constant within each hour. Variations on sub-hour time scales should be parameterised through the horizontal dispersion coefficients (σΘ). Higher variability implies a larger dispersion coefficient and a lower hour-averaged concentration, which is more realistic. Wherever possible, horizontal dispersion coefficients should be derived from observed wind-direction fluctuations, and this is generally included in plume models. Otherwise, the coefficients have to depend on the stability class.
Use of the σΘ coefficient requires data on the standard deviation of wind direction to be included in the meteorological data file, although this will often not be available. Use σΘ coefficients with caution. When highly stable conditions coincide with very low σΘ values (i.e. very limited dispersion in both the horizontal and vertical directions), unrealistically high concentrations can be predicted. When using σΘ coefficients it can be helpful to cross-check your results by rerunning the model using P-G dispersion curves and comparing results.
The P-G is the most commonly used dispersion coefficient scheme used in ISCST3 and is selected when rural conditions are modelled. However the urban option automatically uses the McElroy Pooler coefficients calculated from the Briggs formula. CALPUFF does the same if the default dispersion coefficients are selected. Note that when you choose the urban option in ISC3ST3 you get both the urban (McElroy-Pooler) dispersion coefficients and the urban wind profile.
Compared with the commonly used dispersion schemes, CALPUFF also has more sophisticated methods to calculate dispersion coefficients from meteorological variables. Scire et al. (2000a) provide a detailed description of the 'similarity theory' and 'real turbulence' dispersion schemes. These schemes are widely recognised as being more scientifically robust and are therefore a potentially better option for the modeller to use.
When using steady-state Gaussian-plume models, use:
a) Pasquill-Gifford dispersion curves for all sources except those with very tall stacks
b) Briggs-Rural dispersion curves for stacks higher than 100 m
c) the same type of vertical dispersion coefficient as selected for the horizontal equations
d) measured turbulence parameters (i.e. σΘ) only when good-quality and appropriate meteorological data are available and when it can be demonstrated that using σΘ produces superior results to using P-G or B-R schemes.
When using advanced models, consider the available turbulence/dispersion schemes, and use the option that:
a) makes best use of the available meteorological data (e.g. measured turbulence parameters)
b) is the most appropriate and useful scheme for the case under consideration - use a sensitivity analysis of the model's performance to the different turbulence/dispersion schemes to identify the most appropriate and useful option.
For more detailed information on the use of dispersion coefficients, refer to the user's guide for the particular model being used.
If you are using a steady-state Gaussian model, you should note the following guidelines for range applicability of Gaussian-plume models (CASANZ, 1998).
- The P-G curves used to define dispersion coefficients are applicable over distances up to about 1 km from the source.
- The P-G curves can also be extrapolated out to distances of about 10 km from the source but with a loss of accuracy due to changes in wind speed, direction, terrain and surface roughness, which would occur as the plume travels over that distance.
- The P-G curves are much less reliable for receptors less than about 100 m from the source.
The approximate range applicability of plume models is:
a) receptors < 50 m from source - acknowledge large uncertainties and do not rely on model results (applies to most models, but AERMOD, ADMS, then ISCPRIME and AUSPLUME-PRIME may perform better in this circumstance)
b) receptors 50 m - 100 m from source - use model results with some caution
c) receptors 100 m - 10 km from source - this is the usually accepted range of model applicability, although results for distances greater than about 5 km will lose accuracy due to wind shifts over that distance
d) receptors >10 km from source - do not rely on plume model results; instead use a mesoscale or regional model which uses wind fields over the extent of the grid.
Wind directional shear, surface roughness and the selection of rural or urban wind profile exponents also affect the rate at which the plume spreads and must be considered when modelling. These are discussed below.
a Wind direction shear
Any variation in wind direction over the plume depth (wind shear) will give rise to an effective lateral dispersion which becomes a significant additive effect on the horizontal spread of the plume at long distances downwind from the source (>10 km). Some models provide an option to estimate the enhancement in the lateral dispersion due to wind direction shear.
Adjustment for wind direction shear should only be used where there are receptors more than 10 km from the sources.
Before configuring the model to adjust for wind shear, check the model's manual to ensure that any other prerequisite criteria are met.
It is unlikely that the option in AUSPLUME or ISCST3 for 'adjustment for wind directional shear' will ever need to be used. However, in CALPUFF shearing of puffs is accounted for, and the use of wind direction shear is appropriate.
b Surface roughness
Topographic features, buildings or vegetation increase the ground's surface roughness. For all but the unstable categories (where convective turbulence dominates), surface roughness increases the vertical mixing of a plume and changes the wind-speed profile at elevated heights because of the enhanced mechanical turbulence generated as the air moves over the ground (VicEPA, 2000).
Both the height and spacing of roughness elements at the surface will influence the frictional effect on the wind. Typical values of surface roughness length are shown in Table 4.1.
Table 4.1: Surface roughness lengths for typical surfaces
|Type of surface||Surface roughness length (m)|
Cultivated land (summer)
Cultivated land (winter)
Source: Schnelle and Dey, 1999
AUSPLUME uses a simplified range of land-use categories and their approximate surface roughness heights for dispersion calculations (see Table 4.2), although other surface roughness heights can also be used. The type of land use both upwind and downwind of the source should be considered when choosing a land-use category, as both will affect the wind-speed profile. In situations where a range of land-use types must be considered, the modeller may choose to:
- use the category that has the lower surface roughness value, which will produce a more conservative ground-level concentration
- estimate an average surface roughness value for the area and choose the land-use category that is closest to this value.
The AUSPLUME user's guide (VicEPA, 2000) notes that reported values of the surface roughness height vary greatly even for the same nominal land use, and that the surface roughness heights entered into the model when you select a land-use category should be a general guide only.
Table 4.2: Land-use categories and surface roughness length in AUSPLUME
|Land-use category||Surface roughness length (m)|
Figure 4.11 is an example (using AUSPLUME) of a sensitivity analysis for surface roughness. This suggests that the relative effect of changing surface roughness increases with distance from the source. The effect of surface roughness will vary depending on the type of source being modelled. Surface or near-surface releases tend to show a greater sensitivity to changes in surface roughness than do releases from tall stacks.
ISCST3 and ISC-PRIME do not incorporate surface roughness into their dispersion calculations but offer a choice of either urban or rural land use to account for the effect of land use on dispersion characteristics. CALPUFF incorporates the effect of surface roughness via land-use data used at the CALMET meteorological pre-processing stage.
When assigning a surface roughness to the domain being modelled, attempt to capture characteristics representative of the area in which the maximum concentrations are likely to occur (i.e. within 200 m of a surface or near-surface release and up to 10 km for tall stacks).
When choosing a surface roughness length, consider the sensitivity of the model's output to the value of the factor chosen.
Select the smallest relevant surface roughness factor for the area being studied.
The use of the most relevant surface roughness height will not always give the highest concentration. For example, from an elevated emission, particularly close to the source, the increased vertical dispersion resulting will bring the plume down to ground sooner for larger surface roughness heights.
c Wind profile exponents (rural versus urban)
To calculate plume rise, the wind speed at the stack height is estimated by extrapolating from the wind speed measured by the anemometer (often located at a standard height of 10 metres) (VicEPA, 2000). The extrapolation calculation uses a power law equation with a wind profile exponent. Wind profile exponents are most strongly influenced by surface roughness and are a function of the Pasquill stability category.
Wind profile exponents can be specified as rural or urban. The dispersive characteristics of the atmosphere are different over large urban areas, which might have significant surface fractions of concrete, roads and buildings (Figure 4.12). In some circumstances, the choice of urban or rural dispersion curves can make a substantial difference to the maximum ground-level concentration predicted. Results using urban values are typically one to five times higher due to additional convection caused by urban surfaces, although the reverse can be true for ground-level area sources.
The models' users guides contain guidance for determining whether the surrounding land should be classified as rural or urban.
Using height and density of building development as criteria, classify the land use for the area contained within a 3 km radius (approximately) from the source. If more than 50% of this area is urban (i.e. reasonably high density of buildings), then urban dispersion coefficients should be used, otherwise rural coefficients should be used.
Where the surrounding area may have an equal mix of both urban and rural land use, run the model using both options and present the range of results. The actual effects are likely to be somewhere between the two.
Bodies of water are not included in this calculation (Trinity Consultants, 1996). It is generally considered that the majority of New Zealand cases will be classified as rural, because urban development tends not to be as heavily built up as in many overseas examples (NIWA, 1998).
4.3.2 Plume rise
The height of the plume centreline at some downwind distance is the sum of the initial discharge height, the plume rise due to buoyancy and the plume rise due to the initial momentum of the discharge, minus the stack-tip down-wash (VicEPA, 2000). In addition to the effects of building wakes on plume rise described in section 4.3.3, there are three types of plume rise options that can be selected:
a) gradual rise of a buoyant plume
b) partial penetration of elevated inversions
c) plume downwash from behind a stack (stack-tip downwash).
In AUSPLUME you must select either option (a) or (b), because the model will not calculate both at the same time. It may be necessary to run the model with first one and then the other option selected to see which option creates the greatest ground-level concentration. Option (b) is not available in ISCST3. In CALPUFF, both options (a) and (b) can be simulated together.
In TAPM, a point source can be represented by either the Eulerian Grid Module (EGM), or by a hybrid Lagrangian Particle Module (LPM) for near-source dispersion, converting to EGM mode far from the source. TAPM calculates plume rise using predicted 3-D meteorology and turbulence. Gradual plume rise and stack-tip downwash are automatically used in LPM mode, while only stack-tip downwash is used in EGM mode. TAPM developers recommend that when using TAPM, significant point sources should be run in LPM mode wherever possible/practical when near-source maximum concentrations need to be modelled accurately.
a Gradual plume rise
When gradual plume rise is selected, the plume rises gradually to its final height as it travels downwind (Figure 4.13). The gradual plume rise mechanism is physically realistic but mathematically simplistic. If this option is not selected, the model assumes the plume is at the final plume height everywhere when calculating ground-level concentrations, rather than calculating plume height as a function of downwind distance. If the partial penetration of plumes through an elevated inversion layer (option b) is required, then in AUSPLUME the gradual plume rise option cannot be selected. This is likely only in cases of tall stacks whose plumes quickly rise towards the inversion height; otherwise it is physically unrealistic to not select the gradual plume rise option.
In AUSPLUME, select the gradual plume rise option unless there are tall buoyant sources and low mixing heights.
In ISCST3 the 'regulatory default' combination of dispersion and plume rise parameters include no gradual plume rise, so if the gradual plume rise option is required the regulatory default settings will need to be overridden (US EPA, 1995).
b Partial penetration of inversions
An inversion layer can be described as thermal stratification between layers of the atmosphere that provides a resistance to vertical movement of the plume. If the plume has enough energy from its initial thermal buoyancy or vertical momentum to penetrate the inversion layer, some or all of the plume may become trapped above the layer and be unable to return to ground. Some Gaussian-plume models assume that plumes in the mixed layer below the inversion cannot penetrate upwards through the inversion base. However, observations show that in some cases the plume may penetrate the inversion layer. Partial penetration is shown in Figure 4.14.
AUSPLUME, CALPUFF, CTDMPLUS and AERMOD provide an option for simulating the partial penetration of buoyant plumes through an elevated inversion. Partial penetration of elevated inversions is automatically handled by TAPM using the predicted 3-D meteorology and turbulence. For any simulated hour, concentrations will either increase if the plume penetrates downwards through the inversion base, or decrease whenever material is lost upwards from the mixed layer below. The AUSPLUME user's guide (VicEPA, 2000) notes that this option assumes that the plume reaches its maximum height instantaneously as it leaves the stack. Sacrificing the ability to model the gradual rise of the plume as a function of the downwind distance is only warranted for tall, buoyant sources such as power stations and smelters, because partial penetration of inversions is important in these cases and some models cannot simulate both partial penetration and gradual plume rise. Neglecting the gradual plume rise will underestimate concentrations wherever building downwash or impacts on nearby hillsides are important.
If the partial plume penetration option is selected, plume rise and emission rates are modified only if the:
- plume is buoyant
- plume height is within a certain range
- atmospheric conditions are unstable or neutral
- the simulated hour is between 6 am and 8 pm (i.e. during daytime).
For more information on how AUSPLUME calculates partial penetration, refer to the model's user guide.
Use the partial penetration option with caution, and only when there are tall buoyant sources and low mixing heights.
Describe the influence of partial penetration on modelling results by rerunning the model and comparing the results with this option switched off.
c Stack-tip downwash
This option allows for the downwash effects of chimney wakes; i.e. the downwind turbulence created by the stack itself (Figure 4.15). The amount of stack-tip downwash depends on the stack height and diameter, exit velocity and wind speed (VicEPA, 2000). A maximum reduction of plume rise of three stack diameters occurs when the exit velocity is zero. Since the reduction in plume height is never more than three chimney diameters, stack-tip downwash usually makes little difference to the results except for the receptors close to the stack.
Stack-tip downwash is also included in the regulatory default options in ISCST3. An exception may be large, stumpy stacks like tunnel vents, in which case you may want to consider turning stack-tip downwash off.
Select the stack-tip downwash option unless there is a specific reason not to do so.
4.3.3 Building downwash effects
Airflow around buildings is often very complicated and may create zones of strong turbulence and downward mixing on the lee side of a building (Figure 4.16). This effect is known as building downwash. In such cases, the entrainment of exhaust gases released by short stacks or rooftop vents in the wake of a building can result in much higher ground-level concentrations close to the source than the model would otherwise predict. A well-designed stack can minimise building downwash effects. It is generally accepted that if a stack complies with the criteria in the Good Engineering Practice (i.e. 2.5 times higher than any nearby building), then building downwash is unlikely to occur (US EPA, 1985).
Much research has concentrated on ways to simulate building downwash. As a result, there are four main types of building downwash algorithms now in use:
- Huber-Snyder (H-S)
- Schulman-Scire (S-S)
- hybrid scheme (a combination of H-S and S-S)
- Plume Rise Model Enhancements (PRIME).
These algorithms are described in VicEPA (2000). In AUSPLUME, ISCST3, AERMOD, CALPUFF and TAPM building downwash effects are computed only for point sources.
The recently available PRIME algorithm has been proven to be superior to the H-S and S-S algorithms in verification studies (Paine and Lew, 1997). The PRIME algorithm is the default algorithm in AUSPLUME v5.4, although other algorithms can be selected by the user. The US EPA has modified ISCST3 to allow use of the PRIME algorithm, resulting in a new model called ISC-PRIME, which is now becoming the preferred model instead of ISCST3 in the United States. CALPUFF v6 and AERMOD also contain the PRIME algorithm. TAPM uses an approach based on the PRIME algorithm for point sources in LPM mode. The PRIME algorithm requires more detailed building dimension information than the H-S and S-S algorithms.
A full description of PRIME's treatment of plume rise and building downwash is given by Schulman et al. (1997 and 2000). It is recommended that modellers who intend using the PRIME algorithm make themselves familiar with the technical issues associated with this model's development and use.
In AUSPLUME (pre version 5) and CALPUFF (pre version 6) the user may choose one of H-S, S-S or the hybrid approach. One problem with the H-S algorithm is that it tends to overestimate ground-level concentrations in stable, low wind speed conditions (Thistle et al., 1995). If PRIME is not available, the S-S algorithm is recommended over either the H-S algorithm and the hybrid approach. ISCST3 uses the hybrid option by default, and the user cannot change this. A more productive option to take is to upgrade your model to a PRIME compatible version, rather than attempting to use any of the S-S , H-S or hybrid algorithms.
Figure 4.17 shows an example of the effect of selecting S-S or PRIME algorithms on model results where downwash is important.
Building dimensions for ISCST3, AUSPLUME, AERMOD and CALPUFF can be generated either manually or via the Building Profile Input Programme (BPIP), which is available free from http://www.epa.gov/ttn/scram/. With BPIP, the location and height of all buildings and stacks are entered and the subroutine calculates the direction-specific building dimensions. If the PRIME algorithm is being used, a modification to BPIP called BPIPPRM must be used, as special building dimension data are required. BPIP and BPIPPRM are easy to use, and the process is made even simpler by AUSPLUME v5, which has integrated BPIP and BPIPPRM into its graphical user interface (GUI). Some proprietary GUIs for ISCST3, such as those of Trinity Consultants or Lakes Environmental (1996), also include the BPIP subroutine in their GUI.
The output from BPIP and BPIPPRM includes building height and width information for wind directions at 10 degree intervals. Note that although there is a difference in wind direction convention for the input meteorological data between AUSPLUME and ISCST3, both models correctly use the BPIP/BPIPPRM output.
TAPM doesn't use the BPIP-based approach (described above) for treating multiple buildings. The effects of overlapping wakes from multiple building blocks, whether from the same multi-level or multi-tiered physical building, or from multiple physical buildings, are treated by combining the meteorology and turbulence from overlapping wakes. Plume rise is affected by the modified meteorology and turbulence for point sources in both EGM and LPM modes, while dispersion is influenced only for point source plumes in LPM mode. LPM calculations in the model are done for both the cavity and wake regions, rather than specifying a uniform concentration in the cavity as is done in PRIME.
In initially assessing whether building wake effects are likely, a building should be considered sufficiently close to a stack to cause wake effects if the distance between the stack and the nearest part of the building is less than or equal to five times the lesser of the height or the projected width of the building. For downwash analyses with direction-specific building dimensions, wake effects are assumed to occur if the stack is within a rectangle comprising two lines perpendicular to the wind direction, one at 5 L downwind of the building and the other at 2 L upwind of the building, and by two lines parallel to the wind direction, which are each 0.5 L away from the side of the building, as shown in Figure 4.18.
If building wake effects are likely, the algorithms discussed above permit a simple treatment of the effect on ground-level concentrations. However, conventional diffusion approaches employed in such models can be misleading in complex flow situations. Consequently, where significant interactions between several flow obstacles are likely, and model predictions within the downwash area are critical, recourse to either wind-tunnel simulations or computational fluid dynamics (CFD) models may be necessary (e.g. Georakis et al., 1995).
When using AUSPLUME-PRIME, careful note should be taken of the conditions under which the validation studies (Paine and Lew, 1997) of ISC3-PRIME were undertaken. AUSPLUME-PRIME may not default to the same dispersion curve scheme as used in the validation of ISC3-PRIME. This difference may affect the validity of AUSPLUME-PRIME results.
Building downwash effects are more likely to occur and will be most strongly observed during periods of relatively high wind speeds, so it is important that the meteorological conditions during which the maximum concentrations are predicted are considered when assessing the potential effects of building downwash. If maximum concentrations occur at very low wind speeds and at some distance from the building, it is unlikely that they are due to building downwash effects. Conversely, if the maximum concentrations occur close to the building when wind speeds are high, it is likely that building downwash effects are occurring. This may be checked by re-running the model with the building downwash algorithm switched off.
Note that there is a trade-off when using the building wake algorithms in AUSPLUME and ISCST3. When the building downwash algorithm is switched on, the gradual plume rise, stack tip downwash and ISCST3 COMPLEX1 algorithms are negated.
Assess whether building wake effects influence the results by running the dispersion model with and without building wake effects.
Use the PRIME building downwash algorithm if available. Consider upgrading your model if it does not contain the PRIME algorithm.
Use the most conservative of the S-S, H-S or hybrid algorithms if PRIME is not available.
When reporting building wake results, acknowledge the complex nature of plume/building interactions and the increased uncertainty contained in calculations of results within zones affected by building wakes.
Carefully consider the additional uncertainty in ground-level concentrations predicted for receptors within the near wake (i.e. within three building heights or widths).
If model predictions among buildings and in the downwash area are critical, then a computational fluid dynamics model should be considered.
4.3.4 Averaging times
a Averaging periods available
Most dispersion models permit ground-level concentrations to be calculated for a range of averaging times. The minimum averaging time for ISCST3 and CALPUFF is one hour, because the meteorological data are treated as hourly averages. Future versions of CALPUFF may allow meteorological data to be simulated at shorter intervals, and the minimum averaging time will also be reduced.
TAPM uses a numerical time-step of five minutes, and outputs hourly averaged meteorology and pollution concentration. Post-processing options are currently available to generate longer averaging times for pollution concentration, while post-processed pollution concentration for averaging times shorter than one hour are planned for a future version of TAPM.
AUSPLUME also uses one-hour meteorological data but allows the calculation of concentrations for averaging periods of less than an hour. The minimum averaging time that AUSPLUME will allow is three minutes.
To calculate concentrations for periods of less than one hour AUSPLUME undertakes a three-step process. Firstly, AUSPLUME modifies the original three-minute average sigma-y curves for a 60-minute averaging period. The aim of this process is "to represent the various factors involved in the dispersion process as realistically as possible" (Victoria EPA, 1985, p. 40), and it is designed to allow for increased dispersion due to wind meander during averaging times longer than three minutes. The difference between the three-minute and 60-minute sets of sigma-y P-G curves is determined by the approach recommended by Hanna et al. (1977). The following equation represents this relationship:
This approach gives a ratio of 1.8 between the three-minute and 60-minute sigma-y dispersion curves.
Secondly the 60-minute sigma-y curves are scaled for the desired shorter time period (T) using the power law equation below:
Thirdly, the concentration for the time period T is calculated by AUSPLUME using the modified sigma-y (σy Tmin) produced in step 2. Some modellers manually convert one-hour average concentrations to shorter time periods by using a power law equation. The formula most frequently used is:
where C60 = concentration for one-hour average
t =averaging time, in minutes
Ct = concentration for time of t minutes
X = is a coefficient ranging from 0.17 to 0.6.
This manual conversion will produce similar results to the modifying sigma-y approach undertaken in AUSPLUME. The exponent X in this power law equation is often quoted as being 0.2. Using this equation, the correction factors for various averaging times and where x = 0.2, are as follows:
Table 4.3: Correction factors for various averaging times
|Averaging time, t (minutes)||Factor by which to multiply one-hour concentration (C60) to get concentration for time of t minutes|
The best value of X to use depends largely on the degree to which the site experiences convection, and can range from less than 0.2 for tall stacks in a highly convective situation, to as much as 0.6 for low-level releases in stable atmospheres.
However, recent research has shown this formula to be too simplistic and that its use cannot be justified (Venkatram, 2002). For most situations it can lead to a severe under-estimation of maximum short-term concentrations (Katestone Scientific, 1998). This is most likely to be a problem for odour simulations, and is discussed in more detail in the Ministry's revised Odour Management Guide and supporting documents (Ministry for the Environment, 2001).
This formula should not be used to estimate concentrations for longer averaging times than one-hour averages (e.g. using screening meteorology to get the maximum one-hour average, and then estimate 24-hour or annual averages from the exponential relationship). The longer-term averages are determined by meteorological patterns for the particular location, which will not be represented by the exponential relationship. For example, maximum one-hour average concentrations may occur in different wind directions from maximum 24-hour or annual average concentrations because prevailing wind directions do not necessarily give maximum short-term concentrations.
Averaging times for pollutants in the Ministry's Ambient Air Quality Guidelines, as proposed in the 2001 review, require a minimum time of one hour, plus other times of eight hours, 24 hours, three months, and annual (Ministry for the Environment, 2001). These averaging times can all be computed by AUSPLUME, ISCST3 and CALPUFF, but the accuracy of the results is determined by the quality of the meteorological data used.
Configure the model to calculate concentrations for an averaging period that is consistent with the assessment criteria being used.
Calculate results only for averaging times that are equal to or longer than the time-step of the meteorological data set being used.
The use of the relationship that calculates average concentrations for periods shorter than one hour should be avoided. If this relationship is employed, its suitability for that task must be justified, especially the choice of the exponent X (refer equations in section 4.3.4).
Do not calculate average concentrations for periods longer than 1 hour if using screening meteorological data.
b AUSPLUME versus ISCST3 one-hour averages
AUSPLUME and ISCST3 calculate one-hour average concentrations using the dispersion equations that are principally the same. But if you compare one-hour average results for AUSPLUME and ISCST3 (which are configured for an identical source), the ISCST3 results will be higher by a factor of approximately 1.8. This difference is primarily a result of the dissimilar manner in which the two models treat the Pasquill-Gifford horizontal (sigma-y) dispersion curves. This different manner in which AUSPLUME and ISCST3 treat sigma-y dispersion curves is described in section 4.3.4(a).
However if the results of an AUSPLUME and ISCST3 (configured for an identical source) are compared, it is unlikely the theoretical ratio of 1.8 that should exist between the two sets of results will actually be observed. Other differences between the two models that contribute to any deviation from the theoretical ratio of 1.8 include the following.
- ISCST3 defaults to McElroy-Pooler dispersion curves when in urban mode (this option is not available in AUSPLUME).
- P-G curves are adjusted for surface roughness in AUSPLUME (this adjustment does not occur in ISCST3).
- The vertical wind profile exponents do not match. AUSPLUME defaults to either Irwin urban or rural exponents. ISCST3 defaults to its own scheme.
- Wind speeds of less than 1 m/s are treated differently by the two models (section 4.5.5(a)).
To get AUSPLUME and ISCST3 results to agree as closely as possible you would configure:
- AUSPLUME to use ISCST wind profile exponents, P-G dispersion curves, surface roughness length of 0.03 metres for both the modelling domain and the meteorological site and a three-minute averaging time
- ISCST to use the rural dispersion scheme
- the meteorological data set so all wind speeds of less than 1 m/s are equal to 1 m/s.
ISCST3 should be acknowledged as likely to produce more conservative one-hour average concentrations than AUSPLUME (assuming both models are configured appropriately).
c Fixed versus moving averaging times
There are two approaches to calculating time-averaged concentrations: fixed averages, and moving averages. To illustrate the difference between these, consider an eight-hour averaging period. A fixed eight-hour average is reported three times per day, being the average of hours 01 to 08, 09 to 16, and 17 to 24. In comparison, a moving average would be reported 24 times per day, being the average of hours 01 to 08, 02 to 09, 03 to 10, and so on.
The Good Practice Guide for Air Quality Monitoring and Data Management (Ministry for the Environment, 2000b) recommends that both fixed and moving averages should be used for examining human health and environmental effects.
Most averaging times computed and reported by models are fixed, so if an assessment requires a moving average the model output will require post-modelling data processing. This can be achieved by importing the hourly model output data for a particular receptor into a spreadsheet where moving averages can be calculated and plotted from fixed one-hourly data.
Post-processing results (from models that do not include a moving-averages option) can become unwieldy if moving averages are calculated for all receptors. A practical approach to this problem is to identify the areas of highest impact from one-hour average concentration contour plots. For a small number of high impact receptors obtain all hourly values in the time period being modelled from an additional modelling run. Then post-process these results to analyse the moving average concentration.
Report both fixed and moving averages when assessing potential human health and environmental effects.
The type of average being used (i.e. fixed or moving) and the method used to calculate these should be clearly stated in the modelling methodology and reported with the model's results.
The averaging times computed in AUSPLUME 5 are:
- minutes - any number of minutes between 3 and 60
- hours - 1, 2, 3, 4, 6, 8, 12, 24 hours (fixed average)
- days - 90 days computed as a daily moving average
- months - three months, computed as a moving three-monthly average
- long term - averaged over the full meteorological data period.
Additional running averages for a number of percentile levels can be obtained by generating an AUSPLUME binary output file. The additional statistical information for averaging times between one hour and 24 hours can be obtained by running the statistics utility from the main menu (VicEPA, 2000).
The averaging times computed in ISCST3 are as follows: 1, 2, 3, 4, 6, 8, 12, or 24 hours, monthly averages (for calendar months), the average for the entire data period, and the annual average. The type of average (i.e. fixed or moving) is not specified in the user guides (US EPA, 1995 and US EPA, 1995b), but should be taken to be fixed averages.
Averaging times from CALPUFF runs are computed by the post-processor CALPOST. Averaging times of 1, 3, 24 hours, or length-of-run, are preset, or the user can specify any other averaging time of 1 hour or greater. The type of average (fixed or moving) is not specified in the user guide (Earth Tech, 2001), but should be taken to be fixed averages.
TAPM can produce either fixed or running averages for averaging times greater than one hour.
d Modelled versus measured medium-term averages
Fixed 24-hour averages in AUSPLUME, ISCST3 and CALPUFF are calculated from midnight to midnight (i.e. hour 01 to hour 24). If you are adding measured 24-hour average background data to these modelled results, be aware that the background data may be calculated on a different basis to the modelled data (e.g. 9 am to 9 am as used by Environment Canterbury). In many cases, this may not make a significant difference unless the meteorological conditions in the two data sets are significantly different from a dispersion perspective. These differences may also apply to other averaging periods.
4.3.5 Wet and dry deposition
When pollutants are emitted from a source, they disperse vertically in the turbulent air. Particles also settle towards the ground under the influence of gravity. Depending on the characteristics of the surface, pollutants may be removed from the air and deposited at the surface. Most models allow for the influence of gravitational particle settling on ground-level concentrations, with the user entering particle size distribution and density data into the model.
In addition, AUSPLUME 5, ISCST3, CALPUFF and TAPM can simulate the deposition of particles to the surface due to both:
- dry deposition (when the deposition is due to gravitational settling, turbulence and the nature of the ground surface)
- wet deposition (when the deposition is due to removal from the air by rain).
Dry deposition leads to a steady removal of pollutants from the air, in all weather conditions. However, precipitation can remove contaminants far more rapidly, cleaning the atmosphere in a single rain event.
Dry or wet deposition can be important whenever the source discharges significant amounts of large particles or certain other contaminants (e.g. metals and dioxins), which can occur from many source types. AUSPLUME 4 and earlier versions can calculate dry deposition only using a cruder methodology (VicEPA, 2000).
AUSPLUME 5, ISCST3, CALPUFF and TAPM also calculate plume depletion, due to either dry or wet processes, and material is removed from the plume as it is deposited on the surface.
Particle information (size, distribution and density) must be specified by the user if either dry particle deposition or depletion is requested. Particle size distributions (usually given in terms of aerodynamic diameters) have major effects on deposition rates. It can be a resource intensive exercise to obtain realistic and representative particle size distributions. Scavenging coefficients must be provided if either wet deposition or depletion needs to be modelled.
Wet deposition or depletion options will also require precipitation data to be available in the meteorological data file. In addition to rainfall in the meteorological data set, ISC requires friction velocity, Monin Obukhov length, surface roughness length and precipitation code. Therefore there is substantial work in developing a conventional ISC meteorological dataset to allow wet and dry deposition calculations to be made. AUSPLUME 5 will estimate friction velocity and Monin Obukhov length (although no information is given as to how it actually does this), so dry deposition can be computed without extra met file manipulation.
Gas deposition relates to surface adsorption of gases by vegetation. The ISCST3 and CALPUFF models also include algorithms to handle the scavenging and removal by dry or wet deposition of gases such as sulphur dioxide, sulphate, nitrate, nitrous oxides, and nitric acid. Gas deposition modelling is not commonly undertaken and will be most applicable only for large-scale discharges. Scavenging coefficients and special meteorological data must be specified by the user. For more information, refer to the models' user guides.
TAPM chemistry-mode includes dry and wet deposition processes for a number of species.
If modelling the effects of wet or dry deposition, demonstrate that good-quality input data are put into the model (such as deposition rates and specialised meteorological information).
In the near field (i.e. less than 10 km) and for (relatively) simple issues AUSPLUME and ISCST3 can be used to provide an indicative assessment of the effects of deposition.
In large modelling domains and for less simple issues, an advanced model which is designed to assess the effects of deposition should be used.
4.3.6 Atmospheric chemistry
Some pollutants are chemically reactive in the atmosphere. These include two types of pollutants.
- Pollutants that are unstable when discharged, and continue to dissociate into other chemicals as they disperse and are carried downwind - the effects of such reactions on ground-level concentrations may be significant over short ranges downwind if the reactions are fast.
- Gases such as nitric oxide (NO), nitrogen dioxide (NO2), and sulphur dioxide (SO2) that react with each other in the atmosphere. The effect of such reactions on ground-level concentrations tends to be more important for airshed, ozone, visibility and long-range transportation studies, perhaps over hundreds of kilometres. These are the more common types of applications of atmospheric chemistry. In such cases, the applicability of AUSPLUME and ISCST3 is very limited due to the complexity of the chemistry, and the range over which the receptor grid must the calculated.
In AUSPLUME, the exponential decay of pollutants can be modelled by simply multiplying the computed concentrations by the factor exp(kx/U), where k is the decay coefficient, x is the downwind distance and U is the wind speed at plume height. A single default value for k can be specified for all hours, or a separate value can be specified in the hourly meteorological data file. No other types of atmospheric chemistry can be modelled by AUSPLUME.
A similar decay term is used in ISCST3. The user can specify either a half-life for exponential decay in seconds, or a decay coefficient in units of s-1. A decay half-life of four hours is automatically assigned for SO2 when modelled in the urban mode.
CALPUFF allows a more detailed simulation of atmospheric chemistry, including transformations between ozone, NO, NO2 and nitrate (NO3), and SO2 to sulphate (SO4). One of the most common atmospheric chemistry issues regulatory modellers are required to address is estimating NO2 from modelled NOx concentrations. This problem is not handled realistically in either ISCST3 or AUSPLUME. Depending on the source, the amount of NO2 in the exhaust stream as it is released is around 5-10 % of the NOx. To compensate for the transformation of NO to NO2 that occurs after the exhaust gases are discharged, modellers have adopted the practice of increasing the amount of NO2 discharged from the source. In some cases the overly conservative assumption is made that 100 % of NOx is NO2.
Expressions by Janssen et al. (1988) have been used quite commonly to estimate conversion of NOx to NO2 downwind from emissions, evidently on the mistaken impression that these expressions are based on chemical reaction rate constants. However, the observations and expressions given in the Janssen paper result from the rate of diffusion of ozone into the emission plume rather than the rates of reaction, and in this circumstance Janssen's estimation is probably applicable only to the particular 500 MW power station studied and is of questionable application to other sources.
The method described by Janssen could potentially be useful if local NOx data are available to estimate how the NO2/NOx ratio changes with distance. It may be possible to apply the same observed ratio to the new NOx emissions to estimate the new additional NO2.
A conservative method of estimating the extent of conversion of NOx to NO2 based on typical background concentrations of ozone and the fast reaction between ozone and nitric oxide has been proposed. The methodology is explained in detail in Appendix C 'Estimation of Nitrogen Dioxide Concentrations from Modelled NOx Concentrations'. The methodology explained in Appendix C is a pragmatic and potentially very useful way of estimating the rate at which NO is oxidised to NO2, and is similar to the methodology incorporated into the US EPA ozone limiting model (OLM).
The OLM is conservative for the reasons given in the Appendix, but also because it assumes all the ozone is available to the new source. The OLM will be too conservative when the new source is to be located in close proximity to existing sources. In this case the new source will be competing with the existing sources for the available ozone, and the rate of conversion of NO to NO2 will not be as great as if the new source was in an isolated location.
It is important to note that either of the methods to assess the production of NO2 described above may be the more appropriate, depending on the situation being modelled. Further consideration of these methodologies by the wider air quality professional community is required before either may be recommended for general use. This topic will be covered in the Good Practice Guide for Assessing Discharges to Air currently under development by the Ministry.
TAPM provides a choice between simple exponential decay in Tracer mode, or more complex chemistry and integrated deposition processes using the Generic Reaction Set (GRS) photochemistry and aqueous chemistry scheme in Chemistry mode. The Chemistry mode uses a similar approach to other urban airshed models, but with fewer species and chemical reactions to allow faster execution of the model. Output species include NOX, NO2, O3, SO2, PM10 and PM2.5. The TAPM chemistry scheme is suitable for most applications, and is more comprehensive than some plume/puff models can offer. But for studies that examine the effect of small perturbations to VOC urban emission inventories, a more complex chemistry scheme provided by an urban airshed model is recommended.
Urban airshed models, such as CALGRID and UAM-V, contain a comprehensive treatment of atmospheric chemistry, including reactions between NO, NO2, O3 and VOCs leading to the photochemical production of O3. No assumptions of decay rates or post-processing of chemical output is required. These models are useful on the regional scale, when chemically active pollutants arise from many sources to interact over the urban area (e.g. from vehicles, industrial and domestic sources, and vegetation).
When atmospheric chemistry may significantly change the composition of the pollutants contained within the plume being modelled, one of the following methods should be used to assess the changes.
a) Use a Gaussian-plume model if the chemistry can be described by a highly simplified first-order scheme.
b) The model results for inert pollutants may be post-processed to derive concentrations of chemical products. A particular example of this is the calculation of NO2 from emitted NOx. (See the US EPA's ozone-limiting method, as described in Appendix C.)
c) Chemistry available in model (e.g. NOx and SOx treatment in CALPUFF).
d) A full airshed model should be used for more complex interactions, when many sources are present (e.g. urban traffic, industry, home heating).
The method used to consider the influence of atmospheric chemistry should match the scale and potential effects of the source being assessed. For further advice, refer to the Ministry for the Environment's upcoming Good Practice Guide for Assessing Discharges to Air.