The dispersion models discussed so far are principally concerned with assessing the effects of specific sources on specific locations. Their key features are:
- the sources are well identified (although there may be quite a lot of them)
- they can be point, line or area sources
- the impact locations are specified
- they are usually focused on worst-case circumstances (or at least identifiable cases leading to some specific undesirable effect).
This section provides an introduction to a number of specialised dispersion models which to date have not been frequently used for regulatory purposes. However, with the ongoing development in the number and type of effects that are required for an assessment of environmental effects (AEE), these models are likely to be used more often in the future.
3.1 Airshed modelling
In many circumstances, particularly in urban areas, we want to know about the effects of all the sources of emissions on a whole area. Standard dispersion models (whether Gaussian-plume or puff) cannot do this very well, so a different class of model is applied - the airshed model. In theory, provided all the sources can be identified, a standard dispersion model could perform the task, but in most cases the number of sources is too large for the model to handle and the computing resources required would exceed the capacity of most desk-top computers.
3.1.1 What is airshed modelling?
Rather than tracking the plumes of contaminants from point (or line, or area) sources, an airshed model divides the whole region of interest into a series of cells, and models what happens as contaminants are moved by the wind from one cell to the next. Airshed models also include formulations of the chemistry of the atmosphere, so they can account for chemical transformations that occur in the atmosphere.
The models have variable grid sizes and time steps, depending on the requirement. These can range from very large-scale models, with grid cells of hundreds of kilometres examining air quality parameters over entire continents for many years, to micro-scale models, with grid cells of a few tens of metres and time steps of a few minutes.
Airshed models also need to take account of what is happening in the vertical structure of the atmosphere and thus have a number of layers in the grid boxes, often with variable depth. These range from a few metres deep (particularly in the lower layers close to the ground) to hundreds of metres deep. The depth almost always includes the boundary layer (up to a few km), but in some larger-scale models needs to cover the entire troposphere (up to 30 km).
As for dispersion models, airshed models require two key types of input information: emissions and meteorology.
The emissions of all contaminants of interest, including anything they might react with, needs to be known for each grid cell and time step. This is no simple task, even for small, uncomplicated areas. Therefore a reasonable emissions inventory is required, identifying all sources from point, area and line emissions including domestic, mobile and natural (particularly vegetation) sources. In-depth information about how to calculate and compile the emissions from a particular airshed is provided in the Good Practice Guide for Preparing Emission Inventories (Environet, 2001).
Before compiling the required emissions data, the following choices need to be made. Some of these can require considerable resources to quantify.
- Spatial resolution - 1 to 3 km is typical. Higher resolution is desirable, but the number of grid cells, effort required to identify emissions, and computer time needed to run any model increases dramatically.
- Time resolution - the ideal is hourly. However, it is often impractical to explicitly determine many emissions on an hourly time frame. In these cases sub-models can be used. For instance, mobile emissions can be estimated using a traffic model, or vegetation emissions from a broad understanding of the daily and seasonal cycle of plant growth.
- Contaminants - these depend on the application, but most models would require CO, SO2, NOx and particulates, plus a good description of volatile hydrocarbons. Most models include NOx and ozone, and these are intimately involved in most chemical reactions of contaminants. These reactions are strongly influenced by volatile (or reactive) hydrocarbons. Some models require these to be known to great accuracy, with each compound specified, and over 100 can be involved. Other models can perform adequately by using a single estimate of reactive hydrocarbon emissions. Commonly used photochemical transformation mechanisms include:
- SAPRC 99 (http://pah.cert.ucr.edu/~carter/SAPRC99.htm)
- Carbon Bond IV (i.e. http://airsite.unc.edu/soft/cb4/cb4main.html
- RACM (http://pah.cert.ucr.edu/~carter/epacham/stockwel.pdf)
- Generic Reaction Set (http://www.dar.csiro.au/pollution/IER_Calc/IERdescription.htm #Azzi_1992).
- Period to model - this depends on the application and computer power, with two periods being typical. The first is examining a case study over one to three days, with the objective of understanding the response of the airshed to a particular event (e.g. a high ozone day). The other is over periods of a year or longer, examining trends and scenarios.
In addition to the emissions, meteorology is also a strong determinant of air quality, so a good description of the meteorology is vital because the effects are cumulative. Unfortunately, however, meteorology is often the most uncertain input into airshed modelling assessment.
Consider two grid cells in close proximity, both with high emission rates. If wind speed and/or direction are not known accurately, then the emissions from one grid cell can either add to or miss the other completely. This can result in errors of up to ±100% in the total concentration of emissions. If this situation is extended to a number of grid cells, or to a complex meteorological case with re-circulation, the errors can become enormous.
A similar situation arises if we do not know enough about the vertical structure of the atmosphere. Consider two extreme cases: firstly, a light-wind day with a very stable atmosphere, and perhaps an inversion at 100 m; and secondly, a light-wind day with a good mixed layer 1000 m deep. The model will mix contaminants through this layer, and the concentration difference in these cases - which are entirely realistic - is a factor of 10.
Meteorological fields for use in airshed models are usually derived from a specialised mesoscale meteorological model, which can be part of the airshed modelling system or a completely separate component.
3.1.2 Where is airshed modelling used?
Airshed models are used to examine the air quality characteristics of entire regions, incorporating all the relevant features of the region. The region, or airshed, is defined by the application.
For some purposes the airshed might be quite small, as when looking at a single contaminant in a valley, without major chemical reactions such as carbon monoxide or particles from wood burners. In this case the sources might be well known, the chemical reactions negligible, and the meteorological cases of interest straightforward. In the extreme, this case could lead to the application of a highly simplified form of airshed modelling, the 'box' model, which has just a single grid cell into which all contaminant emissions are mixed.
A more common application of airshed models is for an urban airshed, which is from a few kilometres to tens of kilometres. In this case, it is desirable to know all the emissions from a city, where they are going, and what affect they have on the air quality of the city and surrounds.
Larger-scale applications such as modelling the airshed of entire states or countries are not used in New Zealand. These are often used for photochemical pollution studies, or to assess acid rain, where transport of hundreds of kilometres is relevant. Such models are used extensively in Europe, Asia and the US. The most extreme airshed is the whole hemisphere.
3.1.3 What information do airshed models provide?
Model outputs contain a huge amount of information. Imagine a typical, realistic, upper-end domain of a 100 km x 100 km grid with 1 km spatial resolution, ten vertical layers, run for a year with one-hour time resolution, specifying eight typical contaminants. This is 7 gigabytes of data, and if printed out would be a stack of pages slightly higher than Auckland's Sky Tower.
Key statistics, multi-dimensional charts, time series at selected points and short-term case study periods are all necessary to display and assimilate airshed model outputs. The choice depends very much on the application. Some example applications are provided in Table 3.1.
Table 3.1: Some example applications for airshed modelling
|Frequently asked questions||Method|
"What's the peak value occurring, and where and when?"
A global search on the output files will determine this easily.
"Where should we locate our monitors to make sure we get a representative picture of air quality in the area?"
A visual examination of a selected series of maps of ground level concentrations will provide a good indication.
"What will happen if we allow vehicle emissions to double?"
Select either areas, or times, with peaks, and re-run the model with the extra emissions.
"What effect will an unusually warm summer have on our air quality?"
Alter the meteorological file appropriately and re-run.
"We can't monitor the air quality at all of our schools, but can we get an idea of what sort of air quality they experience through a typical year?"
For each of the school locations, print out a time series of the contaminant of interest.
"We need to know people's exposure as part of a health study - can we get a map of average and peak concentrations?"
For a suitable-length model run, produce GIS maps of the relevant statistic.
"I've got an idea about what causes poor visibility on some days, but do we have any information on chemicals in the air above the ground?"
Produce maps or time series of the contaminant in the relevant layer.
"What is the effect on urban air quality of new industrial sources?"
Run the airshed model with and without the new emissions.
An example model output for epidemiological exposure research in Auckland is shown in Figure 3.1. This shows the relative exposure to NO2 in July 1999 by analysing airshed model output and counting the number of occurrences of NO2 above a pre-specified level. This information can help to determine how many people are affected by different levels of pollutants and hence what the likely health effects are.
Although running an urban airshed model is a fairly specialised task, the final example in Table 3.1 is an important one in the context of the effects of new point sources. The effects of new NOx emissions cannot be considered in isolation from those already present, because the production of NO2 from NO is limited by the amount of ozone present, which must be 'shared' between all sources. Consequently, the NO2 produced by the new source - if considered to be geographically isolated - will be overestimated compared to the extra NO2 produced by that source in the presence of other emitters. If the new source is close to a complex system of air pollution emitters (e.g. a traffic network), then it is more realistic to apply an airshed model to the whole system as part of the resource consenting process for the new industrial source. The need for an assessment of the 'regional' effects of new industry has already arisen in New Zealand.
3.1.4 What models are available for airshed use?
A large number of airshed models are continually being developed. Some examples are mentioned here.
UAM-V (Urban Airshed Model - variable grid) is one of the most common airshed models and was developed by Systems Applications International (SAI). It is available free from SAI or through the EPA website (www.epa.gov/scram001) and can be downloaded in executable form with all supporting guidance documentation (http://www.uamv.com/). It performs reasonably well, but is not particularly user-friendly and requires a great deal of effort to run. UAM has undergone continual development over the last decade and in many ways is now a more advanced model than other more recent models. It has numerous modules and a complex structure.
CALGRID is an increasingly popular model, developed by EarthTech Inc. It is freely available from either EarthTech or the California Air Resources Board website (http://www.arb.ca.gov/eos/soft.html#calgrid), and is more user-friendly. It links to its companion model, CALMET, which produces a realistic meteorological file (the dispersion model CALPUFF is also part of the series). CALGRID is easier to install and run, but still needs a focused effort to obtain the appropriate emissions inventory files and links to a good meteorological file. CALGRID is used in New Zealand, and has been applied to Auckland successfully.
The US EPA has recently released the Models-III photochemical-aerosol modelling system (http://www.epa.gov/asmdnerl/models3/cmaq.html). This is a complex, multi-scale modelling system which comprises an emissions inventory, prognostic meteorological modelling and chemical transport-transformation modelling sub-systems. The CMAQ (Community Multi-scale Air Quality) is the main component of the system, modelling the processes of pollutant transport, chemical transformation and wet and dry deposition for a variety of primary and secondary gaseous and aerosol species.
TAPM (http://www.dar.csiro.au/tapm/index.html) has recently been coupled with the CSIRO chemical transport model that includes a highly condensed chemical transformation mechanism. This enables more complex chemical transformation mechanisms to be considered and allows TAPM to be usefully applied to modelling inter-seasonal and inter-annual variations in photochemical smog concentration. TAPM has been applied in Auckland, Christchurch, Timaru and Nelson.
A number of companies offer 'commercial' airshed models or, more commonly, complete packages which provide a specific deliverable. These are usually achieved by the use of an in-house model specifically developed by the company.
There are also a large number of 'research' grade models. These are not discussed here, because many are highly specific to the issue being studied, or are too complex for third-party users.
Airshed (rather than point-source) models should be used when assessing the air quality characteristics of entire regions, incorporating all the relevant features of that region.
Airshed models should be used to:
a) design and assess management programmes
b) examine long-term air quality trends
c) provide air quality information where there is no monitoring data
d) provide data for use in exposure assessments
e) assess the effects of new sources in urban areas, when their emitted pollutants interact chemically with those from sources already present.
Identify and include all emissions from point, domestic and area, mobile and natural sources (particularly vegetation), into the airshed model.
Calculate and compile the emissions using methods consistent with the guidance provided in The Good Practice Guide for Preparing Emission Inventories (Environet, 2001).
The New Zealand Traffic Emissions Rates Model (Ministry of Transport, 2000) should be used to provide data for vehicle emissions.
Airshed models require three-dimensional, time-dependent meteorological fields, which may be obtained from a specialised mesoscale meteorological model or extrapolated from local monitoring site data (see section 5.3 on meteorological inputs for advanced models).
3.2 Roadway emissions modelling
Until recently the adverse effects of proposed roadways on air quality have not been considered in New Zealand. The Ministry of Transport and Ministry for the Environment commissioned a study to identify the key factors affecting roadside pollution concentrations and to predict pollutant concentrations near roadways in New Zealand in 1997. The report, Transport Emissions Study - Modelling and Monitoring (Ministry for the Environment, 1997), provides a good introduction to the effects of roadway emissions in New Zealand. This report can be downloaded from http://www.mfe.govt.nz/publications/air/transport-emissions-study-sep97.html.
Assessing the effects of emissions from vehicles and their transport and transformation at the urban scale is complex. Atmospheric dispersion modelling is one tool used internationally to assess the impact of proposed roadway developments.
3.2.1 What is roadway modelling?
Figure 3.2 is a schematic representation of the emission and mixing processes associated with vehicle emissions. Thermal and mechanical turbulence occurring behind a vehicle contributes to mixing the emissions, so that the air behind a vehicle is relatively well-mixed. If the situation described in Figure 3.2 is expanded to show a flow of vehicles travelling both ways along a road, a 'line source' of contaminants is developed. In many situations the modelling of roadway emissions is carried out using a Gaussian-plume model configured to emulate the dispersion of contaminants from this type of line source.
Obtaining an accurate emission rate is critical to the success of any modelling project. Estimating the type and quantity of contaminants emitted from roadways is inherently complex because emissions vary according to:
- the driving cycle (e.g. accelerating, steady speed, decelerating, idling)
- roadway conditions (e.g. free flow or congested)
- vehicle fleet composition
- traffic volume.
Because of these complexities, roadway modelling often requires an emissions model to meet the input data requirements of the model. In New Zealand the Ministry of Transport has funded the development of a vehicle emissions database, New Zealand-Traffic Emission Rates (NZ-TER). Details and the availability of NZ-TER are described at: http://www.transport.govt.nz/business/multimodal/environment/vfecs/traff....
Users of NZ-TER have raised a number of issues to be aware of when using the model, including:
- emission factors being based on a small number of tailpipe tests
- emission factors only including tailpipe emission - no fugitive or non-tailpipe emissions are accounted for
- NZ-TER includes a number of undocumented assumptions (e.g. future vehicle fleet compositions and the rate of adoption of emission control technology)
- the unquantified but most likely significant effect of badly maintained or otherwise high-pollutant-emitting vehicles - although these vehicles comprise the minority of the fleet they may account for the majority of traffic-related emissions.
To date NZ-TER has not been peer reviewed or validated against real world monitoring data. The Ministry of Transport is currently working towards having NZ-TER peer reviewed. In addition to this, NIWA plans to use data collected on the tailpipe emissions from 40,000 vehicles in Auckland to validate the emission rates calculated by NZ-TER.
The Ministry of Transport has also produced a New Zealand Rail Emissions Inventory, which is available upon request.
3.2.2 What roadway models are available?
Dispersion modelling is widely used for assessing the effects of roadway emissions in Europe, England and the United States. Information about the commonly used roadway models can be found at the following internet sites:
- US EPA - http://www.epa.gov/scram001/index.htm (choose dispersion models)
- UK Department of Environment, Food, and Rural Affairs - http://www.defra.gov.uk/environment/airquality/laqm/guidance/pdf/laqm-tg...
- European Environmental Agency - http://22.214.171.124/mds/bin/allmodels.
EPA Victoria (Australia) has recently released the dispersion model AusRoads, a simple line-source Gaussian-plume dispersion model for predicting the near-road impact of vehicle emissions. The methodology is based on the US CALINE4 model. Although the functionality of the original CALINE4 near-road model has been retained, AusRoads has been written so that data entry is easier and a number of artificial limitations have been removed. For example:
- AusRoads has increased the number of links and receptor locations that can be modelled
- a full year of local meteorological information can be read into the program from an external file
- road geometry, traffic density and emission factors and receptor location information can now be entered either directly from the graphical user interface or read from external files.
- AusRoads is available from EPA Victoria (www.epa.vic.gov.au).
3.2.3 An example of a roadway model: CALINE4
This section illustrates some of the important generic issues associated with roadway modelling using the California Line Source Model (CALINE4) as an example. This model is commonly used in New Zealand, the USA and England, is quite user-friendly and is freely available. CALINE4 is only one of many roadway models available. Other well-used models include CAR-FMI (Karppinen et al., 2000) and AEOLIUS (UKMO, 1995).
CALINE4 was developed by the California Department of Transportation and the US Federal Highways Agency for assessing roadway traffic emissions. It is based on the Gaussian diffusion equation and employs a mixing zone concept to characterise dispersion over the roadway. CALINE4 is a Gaussian-plume model and as such is subject to the same limitations of other steady-state Gaussian-plume models (see section 2.1.6).
CALINE4 can model roadways, intersections, street canyons, parking areas, bridges and underpasses. Each CALINE run allows the prediction of up to eight one-hour mean concentrations. Therefore it is useful for investigating one-hour concentrations of NO2 and CO and eight-hour concentrations of CO.
The US EPA lists CALINE4 as the preferred/recommended roadway model (US EPA, 1999). The UK Department of the Environment, Transport and the Regions lists CALINE4 as an advanced model (UK DETR, 2000) but does not indicate any form of approval or endorsement.
a Where can I get CALINE4?
The source code, GUI and manual are available free at: http://www.dot.ca.gov/hq/env/air/calinesw.htm
b What input data does the model need?
CALINE4 requires the user to define:
- hourly meteorological conditions
- line-source emission rates
- number of vehicles
- roadway configuration
- receptor locations.
c What care needs to be taken when using this model?
CALINE4 requires meteorological data of a similar type and format to other Gaussian-plume models such as ISCST3 or AUSPLUME. However, unlike other dispersion models, the CALINE4 GUI will only allow eight hours of meteorological data to be processed during each run. The modeller must therefore choose a 'worst case' set of meteorological conditions. Identifying this will take some experimenting with the meteorological input data. In addition to this, the maximum ground-level concentration for each receptor is likely to occur under different meteorological conditions. Run times longer than eight hours can be achieved if the model is run in batch mode via the DOS prompt.
The estimation of roadway emission data is a very important but complex task. The New Zealand-Traffic Emission Rates (NZ-TER) database is considered more likely to estimate emissions accurately from the New Zealand vehicle fleet than using overseas emission factors, because the New Zealand vehicle fleet contains a unique mix of vehicles, emission control standards and fuel-type uses.
The Gaussian formulation used in CALINE4 is based on two somewhat restrictive assumptions:
1) horizontally homogeneous wind flow, and
2) steady-state meteorological conditions.
Complex topography can bring the validity of each of these assumptions into question. For these reasons, use of CALINE4 in complex terrain should be approached with care.
CALINE4 is a steady-state model and is not designed to emulate the changing rate of emissions from decelerating, idling and accelerating vehicles (i.e. the emission rate for each roadway element in the model is an hourly average). To accurately assess the effects of vehicles passing through an intersection, careful consideration of fluctuating emission rates associated with decelerating, idling and accelerating becomes important. While CALINE4 contains an intersection module it may be more appropriate to use an intersection-specific model such as CAL3QHC (CALINE3 with queuing and hot spot calculations). The source code, GUI and manual are available free of charge, and can be found under Screening Tools at http://www.epa.gov/scram001/index.htm.
Unlike many other dispersion models CALINE4 does not allow gridded receptors to be used. The user can define a maximum of 20 receptors. The location of these should be set to assess exposure at the most sensitive sites and in the approximate locations where maximum impacts are likely to occur (e.g. at the down-wind end of long straights).
3.2.4 Future developments in roadway modelling
Modelling roadway emissions is developing in parallel with other fields of atmospheric dispersion modelling. The trend is towards non-steady-state models that will more accurately emulate dispersion in complex situations such as within complex terrain or street canyons. Three examples of advanced roadway models are:
- Operational Street Pollution Model OSPM (Fu et al., 2000)
- SPRAY (Nanni et al., 1996)
- HYROAD (Ireson and Carr, 2000)
- Lagrangian wall model (LWM) (M. Cope, CSIRO Atmospheric Research, Australia).
The Lagrangian wall model (LWM) solves a similar set of chemistry equations to those in a complex chemical transport model (CTM). The two-dimensional wall is moved at the speed of the vertically averaged wind, allowing considerable speed-up of the solution of the model equations. This allows the model to be operated at very high resolution (10 m), making it suitable for modelling near-road air quality impacts. Initial concentrations in the 'wall' (upwind of the road or sources of interest), and boundary conditions at the edges of the wall, are obtained from the CTM or a model such as TAPM, albeit at a larger scale. Alternatively, the boundary conditions can be merely specified as a typical background concentration.
At the time of publication the Ministry is considering the development of further detailed national guidance on assessing discharges to air from transport. Such guidance would be prepared in conjunction with the Ministry of Transport and the Auckland Regional Council. In the meantime, the following recommendations are offered as guidance.
Specialised roadway models should be used when assessing the effects of contaminants discharged from transport corridors.
The currently available roadway models should only be used in non-complex terrain.
When calculating an emission rate of a contaminant from a particular roadway consider the following factors:
a) traffic volume
b) vehicle speed
c) vehicle fleet composition
d) roadway conditions (e.g. free flow or congested).
Use the vehicle emissions database, New Zealand-Traffic Emission Rates (NZ-TER). If an alternative is employed, specifically justify this.
Take care to identify the worst-case meteorological conditions that will generate the maximum ground-level concentration for each receptor.
If assessing the effects of discharges from vehicles passing through an intersection, consider using a model that accounts for the variation of emissions with driving cycle (e.g. accelerating, steady speed, decelerating, and idle).
The meteorological and emission input data for roadway models require a different approach to that used more commonly in point-source modelling.
3.3 Modelling coastal fumigation
Consider a tall stack, located on a shoreline, which emits a narrow plume towards land (Figure 3.3). The plume is embedded in the stable boundary layer and is intercepted by a growing thermal internal boundary layer (TIBL) over land. The height of the TIBL increases with solar heating of the land surface. Convective mixing over land can rapidly bring the elevated pollutants to the ground, causing local high ground-level concentrations. Unlike the fumigation events associated with the erosion of nocturnal ground-based inversions (see section 4.5.7c), coastal fumigation may persist for several hours, and in the same location.
Coastal fumigation is an important issue in Western Australia, for example. The effects have not been observed to the same degree in New Zealand because the substantial flat land mass and relatively deep, cool water needed to generate this effect are not as common here as in continental land masses. However, the potential effect of coastal fumigation should be considered carefully if you are dealing with a large source located on a coastline. DISPMOD, a steady-state Gaussian-plume model developed by the Western Australia Department of Environmental Protection, and the US EPA model OCD (Offshore and Coastal Dispersion), contain algorithms to model TIBL effects and may be appropriate tools if coastal fumigation is identified as a significant issue at a particular site.
Other steady-state models cannot simulate the high ground-level concentrations in the TIBL. However, advanced models, such as CALPUFF and TAPM, which give more realistic representations of the meteorology in the coastal area, are arguably the most suitable for simulating the fumigation process.
When modelling the dispersion of pollutants from a source located near the coast, the effects of coastal fumigation may be simulated in either:
a) a Gaussian-plume model, which has the ability to handle this specific effect (e.g. DISPMOD, OCD, ADMS3 or US EPA's SCREEN 3); or
b) an advanced model, which gives a realistic representation of the meteorology in the coastal area.
3.4 Visibility modelling
Visibility is a good measure of how humans perceive the atmosphere. It is measured by how far people can see and what colour the sky is, and is therefore an amenity value rather than a health hazard. Predicting how visibility will be degraded is complex because it involves estimating not only the dispersion of contaminants but also the way they are transformed by reactions in the atmosphere and how people perceive visibility. For more information on managing and monitoring visibility in New Zealand, refer to the Good Practice Guide for Monitoring and Management of Visibility in New Zealand (Ministry for the Environment, 2001c).
Atmospheric physicists and chemists have developed and produced a number of tools that enable modellers to study visibility degradation at several spatial scales. CALPUFF is the US EPA regulatory model for regional visibility modelling. PLUVIEW is the US EPA regulatory model for calculating the visual impact of a single plume. The FOG model for calculating the visual impact of a single plume is built into the CALPUFF model.
In New Zealand, there is no specific visibility monitoring programme and no attempt to quantify any future risks using modelling. It is, however, recognised that visibility is an environmental issue in this county (Ministry for the Environment, 1999). The sources of visibility degradation are many and complex, ranging from wind-blown sea spray to dust, fine particles, gases, haze, rain, fog and clouds. Visibility modelling in New Zealand is still in its infancy and is only performed for research.
To model the effect of emissions on visibility:
a) use the approach and models recommended by the US EPA as a starting point
b) consider the influence of site-specific (i.e. New Zealand) conditions.
3.5 Dispersion modelling on larger scales
Atmospheric processes and air pollution dispersion phenomena are commonly classified with regard to their spatial scale. The most common use of dispersion models in New Zealand is to predict near-field effects that occur within 10 km of the source. The effects of urban-scale air pollution can often be observed at greater distances; for instance, the Auckland urban plume reaching the Coromandel Peninsula, or the Christchurch urban plume reaching the southern Canterbury and Otago regions. In such complex cases, advanced models must be used - the straight-line steady-state Gaussian-plume model in a uniform wind field is simply too unrealistic. The same types of models are used for all scales from urban upwards; it is largely only the processes - meteorological and pollution dispersion - that may differ.
3.5.1 The regional scale
The regional scale may be defined as having characteristic distances of between 10 and 1000 kilometres. At this scale, airflow is influenced by thermal and dynamic effects such as flow channelling and the variation of the region's energy balance with land use, vegetation type and water bodies. This scale is commonly referred as the 'mesoscale' when discussing meteorological features, and their importance for pollution dispersion must be considered when choosing a model for regulatory impact assessments.
Usually, the simulation of mesoscale, or regional pollutant, dispersion must be carried out by advanced meteorological and dispersion models. Only at the small-scale end of the range might a Gaussian-plume model be appropriate.
All prognostic meteorological models - referred to simply as 'mesoscale models' - and their associated dispersion models are applicable for regional-scale simulations (e.g. for cases in which dispersion over the whole of New Zealand is important). CALMET (Scire et al., 2000b) and CALPUFF are also suited to this application if there is good meteorological data coverage over the domain of interest. CALMET can also be driven by meteorological output from mesoscale models if data are sparse. On a regional scale the production of secondary pollutants (e.g. ozone and fine particles) becomes more important, and so atmospheric chemistry cannot be ignored.
To model the dispersion of pollutants on a regional scale:
a) use prognostic meteorological models and their associated dispersion models
b) consider the transformation of pollutants due to atmospheric chemistry.
3.5.2 Long-range transport
At much larger scales, with distances of, say, more than 1000 km, meteorology is governed by large-scale pressure-gradient forces and the rotation of the earth. This scale includes weather systems such as anticyclones, depressions and fronts, although the effects of these feed into the mesoscale and smaller scales. Long-range transport may be thought of as dispersion on these 'continental' scales, as controlled by weather systems. It is important when considering, for example, trans-boundary transport across mainland Europe, the transport of pollutants from the USA across the Atlantic Ocean to Europe and, occasionally, the transport of air pollution from Australia across the Tasman Sea to New Zealand (e.g. smoke from fires around Sydney).
All regional modelling systems can be used for long-range transport, provided they can account for the Earth's curvature in their co-ordinate systems, and the chemical transformations that occur on the associated (extended) timescales.
Long-range transport models are used to provide information on the transport and fate of emissions from locations such as the UK to Europe or from state to state in the USA. The chemistry in these models is very important, as pollutants are both removed and chemically converted with distance from the source. Some examples of long-range transport models are STEM (Carmichael et al., 1991), ATMOS (Arndt and Carmichael, 1995), CRIEPI (Ichikawa and Fujita, 1995) and NAME (Ryall and Maryon, 1998). Another well-known model in Europe is RAINS (Regional Air Pollution Information and Simulation), which was developed as a tool for the integrated assessment of alternative strategies to reduce acid deposition in Europe and Asia (Alcama et al., 1990). This model describes the pathways of emissions of SO2, NO3, HNO3 and NH4, and explores their impact on acidification and eutrophication. The use of long-range transport models has not been explored in New Zealand.
3.6 Accidental releases
Accidental releases can result in very high ground-level concentrations of pollutants, albeit for a short time period. A number of accidental releases are of particular concern. A catastrophic rupture of a pipeline or tank, or a spill from a tank, can produce a release that lasts from a few seconds to a few minutes. This results in a burst of material or a puff-type release. Gases or liquids may leak from around seals, pipe joints and valves, or from cracks or holes in vessels. This type of release may start slowly and increase in size. High-pressure releases of both gases and liquids from pressure relief valves or pressure seal ruptures could occur, and may be accompanied by flashing of the fluid to a vapour-liquid mixture. If a release lasts from 10-30 minutes, it could be described as a small continuous release.
Modelling accidental releases may be required for a variety of reasons.
For long-term industrial site planning - modelling could be used to see the effects of different accidental release scenarios, and thereby help to determine the terrain, meteorology and surrounding residential areas most favourable for the prevention of significant harm to the residents.
To identify the types of accidental releases that could result in significant downwind adverse effects - this would enable prevention planning through the design of mitigation equipment and emergency planning through evacuation strategies. This type of modelling is also used as a consequence tool for risk assessment.
These models are also used by emergency response services if an accidental release does actually occur. The model outputs are used to identify potentially affected people and estimate requirements for evacuation.
Modelling accidental releases requires both a source emission model (because of the nature of accidental releases, you can not measure them), as well as a transport and dispersion model. It is important to know whether the discharge involves gases that are heavier than air (known as dense gases). Many accidental releases do involve dense gases that cannot be modelled by the commonly used plume or puff models. Dense gases require specialised treatment from dense gas models such as DEGADIS+ (US EPA, 1998c), SLAB (Ermak, 1990) or AUSTOX (Ross 1994; Ross and Koutsenko, 1993).
ALOHA (Areal Locations of Hazardous Atmospheres) is one of the most commonly used accidental release models and is used worldwide for response, planning, training and academic purposes. ALOHA uses information provided by its operator and physical property data from its extensive chemical library to predict how a hazardous gas cloud might disperse in the atmosphere after an accidental chemical release. ALOHA can predict rates of chemical release from broken gas pipes, leaking tanks and evaporating puddles, and can model the dispersion of both neutrally buoyant and heavier-than-air gases. ALOHA is intended for use during hazardous chemical emergencies and was designed to be easy to use so that inexperienced responders can use it during high-pressure situations. Further detail on ALOHA can be found on the US EPA's CAMEO (Computer Aided Management of Emergency Operations) website (http://www.epa.gov/ceppo/cameo/techdes.htm).
When there is an instantaneous release or a burst of buoyant material, a puff model like CALPUFF (version 6) can be used.
When modelling accidental releases, choose a tool that is capable of handling the specialist requirements associated with this task.
When planning emergency response procedures, extra care must be taken to accurately calculate the emission and dispersion of these contaminants because of the potentially high (perhaps even lethal) impact of accidental releases.
When modelling an actual accidental release a fast response emergency response will obviously be required. In this instance, an approximate but conservative estimate of the area potentially affected by the release is appropriate.
Readers can refer to Schnelle and Dey (1999) for more information on modelling accidental releases.
3.7 Salt and steam effects: cooling towers
Steam effects are visible steam plumes from cooling towers, such as those found at power stations and petroleum product refineries. The central North Island of New Zealand is a large geothermal area where geothermal power stations are often located next to major roadways. Fogging from plumes released by geothermal cooling towers frequently occurs (Godfrey and Fisher, 1994), and modelling is often used to determine if the plume could cause visibility problems for motorists or aircraft.
Salt effects may be caused if a steam plume contains salts (from the use of salt water in the cooling tower). They can be detrimental (in the long term) to vegetation and can also enhance the corrosion of building materials.
When modelling for the visual extent of a plume, the model calculates the plume path, length and radius under a particular set of emission and environmental conditions and includes assessment of parameters such as liquid water content and the plume/environment temperature difference. An example of a visual plume calculation model is ATCOOL, a standard fogging model developed by Dr Steven Hanna of the US EPA. ATCOOL is used to calculate the variation of cooling-tower plume parameters with height and distance downwind. Plume-length calculations from ATCOOL have been well validated by observations at many locations.
Spray-water drift deposition assessment can be carried out using the SACTI model, developed by the Argonne National Laboratory for the Electric Power Research Institute (EPRI) in California, USA. SACTI contains a number of models, which include drift deposition, visibility and shadowing. Which is to say, SACTI is not just a drift deposition model. This model is capable of calculating the amount of water deposition that can occur from droplets of spray out of cooling towers. This requires an assumption of a drift rate (a rate that water is emitted) and a typical droplet size distribution. The model has strict input data requirements for accurate humidity and vertical profile information because model results are extremely sensitive to humidity. If there are contaminants in the water droplets, such as salt, deposition rates for those contaminants can also be calculated by assuming a concentration of the contaminant in the droplets.
FOG was a stand-alone model but its algorithms have now been incorporated into the CALPUFF model. CALPUFF is designed to simulate the transport and diffusion of water vapour emissions from multiple point sources and is set to become the default US EPA model. See http://www.epa.gov/scram001/ for more details.
To model salt and steam effects:
a) use the approach and models recommended by the US EPA as a starting point.
b) consider the influence of site-specific (i.e. New Zealand) conditions.