Pollution Prevention and Abatement Handbook WORLD BANK GROUP Effective July 1998

Airshed Models

Modeling may be necessary to estimate the changes in ambient air quality—both local and at a distance—caused by a particular set of emissions. Modeling can be appropriate for new plants and for modifications to existing plants. This chapter provides guidance on some mod- els that may be useful in the context of typical World Bank Group projects.

Air quality is an issue of increasing concern in utilize different types of models, which are dis- many countries. Projects that introduce new cussed elsewhere in this Handbook. sources of emissions or are designed to reduce Although thermal power plants are often emissions require careful analysis to quantify the singled out as major polluting sources, nearly all effects as far as possible. For many sources, this industrial facilities, especially those with short will typically require mathematical modeling of stacks, have the potential to cause localized the changes in ambient concentrations that result areas of unacceptable air quality. In addition, from the new emissions. The few widely used urban areas can act as diffuse sources of air pol- models are reviewed in this chapter. lution, particularly where poor-quality fuels Air quality modeling can be a complex task, are burned in household stoves. Cases of mul- and the objectives need to be clear. The costs of a tiple point sources or area sources (or both) can study can range from US$10,000 to US$500,000, often be modeled by using simplifying assump- depending on the complexity of the situation and tions or by integrating the impacts of individual the level of detail required; in many cases, costs sources. are at the lower end of this scale. The simplest approach uses a point source dispersion model Use of Near-Field Dispersion Models to estimate the ground-level concentrations of the pollutants of interest at some distance (typically Typically, dispersion models have been used in from hundreds of meters to tens of kilometers) developing countries only in isolated cases where from a point source. More complicated models air pollution had been recognized as a serious allow the examination of multiple sources, in- problem (e.g., Mae Moh, Thailand, and Krakow, cluding area (nonpoint) sources. For an area con- Poland). However, with increasing pollution taining a number of point and nonpoint sources, problems and more emphasis on air quality stan- an air quality model can be constructed that in- dards in developing countries, dispersion mod- cludes all of the sources in the area. In practice, els are expected to be used more extensively in such models are rare because of the costs of de- the future for sector- and project-level environ- velopment and the data required to make the mental assessments, as well as for assistance in model a realistic tool. establishing specific emissions requirements. This chapter examines the application of the As a general guide, it is suggested that a basic most commonly used air quality dispersion mod- analysis of possible impacts on ambient concen- els for assessing the impact on air quality of key trations be carried out on installations that have pollutants—sulfur dioxide (SO2), nitrogen oxides the potential to emit annually more than 500

(NOx), and particulates—emitted from point metric tons of sulfur dioxide or nitrogen oxides, sources.1 Far-field dispersion and acid rain depo- or 50 metric tons of particulate matter or any sition are governed by different principles and hazardous air pollutant. In many cases, simple

82 Airshed Models 83 calculations based on loads and air volumes may Key factors that affect these calculations, and be sufficient to provide an order-of-magnitude therefore the selection of dispersion models, estimate. However, the use of formal models are: should be considered for any project involving • Topography. The area surrounding the plant is large new plant or significant modifications. For characterized either as flat to gently rolling ter- major sources, the modeling should include the rain or complex terrain (having downwind lo- planned source or sources, as well as existing cations with elevations greater than stack sources in the same general area—within a ra- height). dius of 10 to 15 kilometers (km)—so that the • Land use. Whether the surrounding area is ur- cumulative effect of all the facilities on local ban or rural is important because urban areas air quality can be assessed. In some cases, typically have large structures and heat building-wake effects are important (for ex- sources that affect the dispersion of pollutants. ample, where release points such as stacks and In addition, the density of the population af- vents are less than 2.5 times the height of fects the numbers potentially impacted. nearby buildings), and more detailed modeling • Pollutant properties. Physical and chemical may be appropriate. properties of the pollutants influence their The models described in this document per- transport. For modeling sulfur dioxide within tain to “near-field” (less than 50 km from the 5 to 10 km of a source, no chemical transfor- point source) dispersion of sulfur dioxide, nitro- mations are assumed to occur. Beyond this gen oxides, and particulates. Such models esti- distance, an exponential decay function may mate the ground-level concentration of pollutants be useful. Most nitrogen oxide is emitted as in the air, which is then compared with ambient nitric oxide (NO), but in a matter of minutes, air quality standards or guidelines.2 Other mod- depending on the availability of ozone, it be- els that address photochemical smog are not de- comes nitrogen dioxide. The deposition of par- scribed in detail here. ticulates is a function of particle size and travel time. Factors Affecting Dispersion of Pollutants • Source configuration. The height and tempera- ture of the discharge and proximity to struc- The dispersion and ground-level concentration tures affect dispersion. Effective plume height of pollutants are determined by a complex inter- is the physical height of the stack adjusted for action of the physical characteristics of the plant factors that raise the plume (as a result of buoy- stack or other emission points, the physical and ancy or momentum) or lower it (as a result of chemical characteristics of the pollutants, the downwash or deflection). meteorological conditions at or near the site, and • Multiple sources. All dispersion models assume the topographical conditions of the surrounding that the concentrations at any one target site areas. are the arithmetic sum of concentrations from In general, three different calculations are each of the sources being examined. Note that needed to estimate the time-averaged concentra- it is the effects that are summed, not the emis- tion of pollutants at a location downwind from a sions rates or stack parameters. plant: • Time scale of exposure. The recommended mod- els make calculations for the basic time period • The plume rise above the stack must be estab- of one hour. Concentrations for longer time pe- lished (effective stack height). riods, such as 8 hours or 24 hours, are the arith- • The dispersion of the pollutants between the metic averages of the hourly concentrations of source and the downwind locations of inter- those time periods. Annual averages are com- est must be mathematically modeled on the puted by averaging hourly concentrations for basis of atmospheric conditions. a full year or by using models that use a fre- • The time-averaged concentration at ground quency distribution of meteorological events level must be determined. to compute an annual average. The recom- 84 IMPLEMENTING POLICIES: AIR QUALITY MANAGEMENT

mended models have the necessary “book- well as at other specified distances, are deter- keeping” incorporated into the processing or mined. No consideration of wind direction is re- available as postprocessor routines. quired because the output represents the concentrations directly downwind. (This model Selecting an Appropriate Model is designed for average North American condi- tions; care should be taken in using it under dif- Model selection requires matching the key char- ferent climatic conditions.) acteristics of the site and the requirements of the Options in the model allow for the effects of evaluation with the capabilities of the model. a single dominant building and for terrain dif- Normally, expert advice is required in making a ferences between the source and the receptors. selection. As a general principle, modeling should To refine the estimates in complex terrain, a always begin with the simplest form possible, more sophisticated screening model is avail- moving to more complex approaches only where able in CTSCREEN, derived from CTDMPLUS their necessity and value can be demonstrated. (see below). At the most basic level, a crude mass balance can Although only a single source (stack) is con- indicate whether a new source is likely to pose a sidered, multiple nearby sources can be screened problem. Alternatively, a simple screening model, by using the sum of the emissions rates from the as described below, can provide a realistic esti- sources as the emissions rate for this single stack. mate of the order of magnitude of the impacts of This will yield an overestimate, since the effects a source. Situations involving multiple sources of geographic separation of the sources or the or varying terrain may require a more sophisti- points of maximum concentration will not have cated effort involving site-specific data collection been included. Scaling factors to estimate con- and more complex models. centrations for longer time averages (3 hours, 8 In some cases, more than one model may be hours, 24 hours, and even one year) are included required. For example, modeling of gaseous in the user’s guide. emissions and particulates in the Mae Moh Val- If the concentrations determined by using a ley, Thailand, required the use of one model for screening model are within the relevant guide- the valley floor, where the terrain is flat, and an- lines, no additional modeling should be neces- other for the mountains that surround the valley sary. If concentrations exceed the guidelines, on three sides (KBN Engineering and Applied more refined modeling should be done. Since the Sciences, Inc. 1989). simplifying assumptions made in the screening model tend to overestimate impacts, refined Commonly Used Models modeling nearly always yields somewhat lower estimates of concentration. Screening Models Screening is straightforward and does not re- quire difficult decisions as to the relevance and The preliminary scoping of the magnitude of the representativeness of meteorological data. It may air pollution problem can be accomplished by the be carried out by competent local specialists, per- use of screening models designed to determine haps with some expert assistance. quickly and easily the impacts from a single source. If it is obvious that several sources are Refined Models contributing to concentrations, screening is not appropriate. A useful screening model is More refined modeling of near-field dispersion SCREEN3 (EPA 450/4-9-006, Modeling Guide- can be carried out with one of several simple line, and EPA 454/R-92-019, Screening Proce- Gaussian plume models. These models predict dures for Stationary Sources). This approach the dispersion patterns of nonreactive pollutants requires no site-specific meteorological input, as such as sulfur dioxide, nitrogen oxides, and par- calculations are made for a spectrum of possible ticulates within 50 km of the emissions source combinations of wind speed and atmospheric and are generally expected to produce results stability (using Pasquill classes). Concentrations within a factor of 2 of the measured values. Most at the downwind point of maximum impact, as such dispersion models are similar in design and Airshed Models 85 performance and do not attempt to account for • PARADE, developed by Electricite de France complex situations such as long-range transport • PLUME 5, developed by Pacific Gas & Elec- and highly reactive chemical emissions. tric Co., San Ramon, Calif., and applicable to What distinguishes the various models is their both urban and rural areas with complex ter- capability to handle different settings. Some of rain the models (such as ISC3 and CTDMPLUS, de- • The German TA Luft procedures. scribed below) are characterized as “preferred Table 1 provides the key characteristics of the models” by organizations such as the USEPA be- commonly used air quality dispersion models. cause they meet certain minimum technical cri- More details on these, as well as on other disper- teria, have undergone field testing and have had sion models (e.g., ERTAQ, COMPTER, MPSDM, extensive peer review. This does not make the MTDDIS, MULTIMAX, LONGZ, SHORTZ, nonpreferred models less suitable for an appli- SCSTER, 3141 and 4141), are provided in USEPA cation, but it does mean there is a documented (1993). experience base for the preferred models, which These models have been developed and used may add more credibility to the analysis or elimi- mainly in industrial countries, but they are suit- nate the need for model validation. Two of the able for developing countries, as demonstrated most commonly used models for assessment of by their use in, for example, the projects in Mae pollutant dispersion are from the USEPA. Moh, Krakow, and Sri Lanka mentioned below. • The ISC3 (Industrial Source Complex) model However, they may require some adaptation to is used for point (stack), area, and volume or calibration for topography and weather pat- sources in flat or complex terrain. The com- terns that are not common in industrial countries. plex terrain analysis does not employ a sophis- For example, dispersion models have not been ticated algorithm. There are two versions: subject to a comparison of model calculations of ISCST3, for averaging periods of 24 hours or existing sources with monitored air quality data less, and ISCLT3, for averaging periods of 30 in tropical weather conditions. days or longer. • The CTDMPLUS (Complex Terrain Disper- ISC3 and CTDMPLUS Models sion) model is for use in complex terrain. A screening version of this model, CTSCREEN, The Industrial Source Complex (ISC) model is a provides estimates if only one or two stacks steady-state Gaussian plume model used to as- affect high terrain. sess pollutant concentrations from a wide vari- ety of industrial sources. It accounts for settling Other commonly used models are: and dry deposition of particulates; wake effects • UK-ADMS, the United Kingdom Meteorologi- stemming from building obstruction; plume cal Office Atmospheric Dispersion Modeling rise as a function of downwind distance; and System multiple but separate point, area, and volume

Table 1. Key Characteristics of Commonly Used Dispersion Models Terrain Source configuration Time scale Flat/ Com- Pollutant Ele- Mul- Short- Long-

gently plex/ Land use Sox / Parti- vated tiple term term

rolling rough Urban Rural Nox culate Point point point exposure exposure

SCREEN3 Yes Yes Yes Yes Yes Yes Yes Yes No Yes No ISCLT Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes ISCST Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Noa PLUME5 Yes Yes Yes Yes Yes No Yes Yes Yes Yes Yes CTDMPLUS No Yes No Yes Yes Yes Yes Yes Yes Yes No

Note: All of the above models except PLUME 5 are available through NTIS, Springfield, Va. 22161. PLUME 5 is available through Pacific Gas and Electric Co, San Ramon, Calif. a. “Yes,” if all hours for the time period are averaged (e.g. 8,760 hours in a 365-day year). 86 IMPLEMENTING POLICIES: AIR QUALITY MANAGEMENT sources. It has the ability to analyze concentra- screening models, but they can be applied in a tions in any type of terrain, and it can estimate basic manner when the necessary data are diffi- hourly to annual pollutant concentrations. This cult to obtain or in order to determine the value model is recommended for both urban and rural of collecting further data. For example, in a com- areas. plex airshed with several sources and differing An earlier version of the ISC model has been meteorological conditions, it may be appropri- used in a number of World Bank projects such as ate to calculate the impacts of each major source Mae Moh, Thailand (KBN Engineering and Ap- individually, using simplified meteorology for plied Sciences, Inc. 1990), Krakow, Poland that source. The impacts from all the sources can (Adamson et al. 1996), and Sri Lanka (Meier and then be summed, producing estimates that are Munasinghe 1994). not as precise as a multisource model but that Several private firms offer enhanced versions may give a reasonable indication of the overall of the EPA models (see, for example, Scholze impact. 1990). The enhancements include user-friendly In complex situations, significant effort and data input, the capability of easily plotting the professional judgment are often necessary for output, custom output summaries, and techni- estimation of the emissions inventory (defining cal support. Some of these firms offer training in all the sources to be included), for collection of the use of models, both in the United States and local meteorological data, and for selection of the overseas. specific combinations of conditions to be mod- Addresses for obtaining further information eled. For example, in an industrial city, it may be on the USEPA and German models are given at appropriate to group the sources into different the end of this chapter. types, such as major sources, smaller industrial or municipal sources, and indeterminate residen- Other Models tial emissions. Identification of sources and esti- mation of the emissions inventory for use in the In addition to the above-mentioned models, model would be significant tasks. For the major there are models that are not used widely but sources, site-specific details should be obtained. that may be the most suitable for specific loca- Smaller industrial sources might be handled by tions because they have been developed by lo- aggregating them into groups or by defining typi- cal institutions and have taken into account cal characteristics. Residential sources would local requirements. For example, a World Bank have to be partitioned in accordance with some study in Krakow, Poland (Adamson et al. 1996) estimates of population density or housing type. utilized a model developed by the Warsaw Efforts would have to be made to understand the University of Technology. local meteorology in some detail, addressing vari- For a facility within 200 km of an urban center ability across the area, patterns of inversions, and that has a smog problem, one may wish to exam- perhaps day/night variations to reflect patterns ine the effects of photochemical reactions be- of residential emissions. tween volatile organic compounds (VOCs) and Credible modeling in such complex situations nitrogen oxides, if the new source contributes requires significant effort and resources. Collec- more than 1–2% to the total emissions of these tion and interpretation of the data required compounds in the airshed. This analysis requires as input can take a large part of the overall an enormous amount of data, takes highly skilled resources—perhaps 50% or more. modeling personnel, and is generally quite ex- pensive. Commonly used models for such an as- Data Requirements sessment are the USEPA Urban Air Quality Model (UAM) and the Regional Oxidant Model (ROM). The data requirements of dispersion models fall into three categories: Use of Refined Models • Source data, including location of stacks and other Models such as ISCST3 are more sophisticated sources (coordinates), physical stack height and in their structure and capabilities than the simple inside diameter, stack exit gas velocity and tem- Airshed Models 87

perature, and pollutant release rate. The latter cally by an expert) to compare them with local is usually given as the time-weighted average ambient air quality standards and identify “hot (per 1 hour, 24 hours, or year). spots”—areas where pollutant concentration is above desirable levels. Some dispersion models may require addi- It should be emphasized that mathematical tional inputs such as point source elevation, modeling of complex atmospheric processes in- building dimensions (e.g., average building volves a significant level of uncertainty, which width and space between buildings), particle size can be made worse when data are lacking or un- distribution with corresponding settling veloci- reliable. Model results must therefore be treated ties, and surface reflection coefficients. with care when using them in formal decision- • Meteorological data are required for predicting making. The presentation of results should nor- the transport, dispersion, and depletion of the mally include a discussion of the probable pollutants. Most models accept hourly surface variability and the confidence limits. weather data that include the hourly Pasquill For decisionmakers, the results need to be stability class, wind direction and speed, air summarized in a clear, understandable way. temperature, and mixing height. Ideally, a year Table A.1, which sets out the key findings from a of meteorological data would be available. In modeling study of a proposed power plant, is an cases where some long-term data are available example of such a presentation. in the region (typically, readings taken at an airport), shorter-term local observations may Resource Requirements allow the long-term records to be transferred for Dispersion Studies to the site under examination. Where appro- priate, a local meteorological station can be Information on screening models is generally established (estimated cost, about US$30,000– readily available. The costs of acquiring the $40,000 for setup and one year of operation of model, some training, and the actual study one automated site). should be less than US$10,000. Local consultants • Receptor data, meaning identification of all key can rapidly acquire skill with the screening mod- receptors (e.g., areas of high population or ex- els. Where refined modeling is required, the nec- pected maximum ground-level concentration). essary skill level increases sharply. Usually, receptors are specified by their coor- Air quality monitoring and model validation dinates and elevation. can have significant costs. In the United States, air quality analysis costs for power plants have Values of input parameters can be determined ranged from US$100,000 to US$2 million. The by direct measurement, sampling, or estimates lower end of this range corresponds to the case based on sound engineering principles. The lit- of readily available meteorological data and flat erature may provide data or empirical correla- terrain in a rural area. The high end of the range tions that can be used for estimating dispersion includes ambient air quality monitoring costs model inputs. and, in some cases, the cost of demonstrating the inappropriateness of a model approved by the Interpretation of Results regulatory agency or of validating a model not approved by the regulatory agency. Although The results of dispersion modeling are typically these costs are based on experience from indus- maps showing the concentration of the consid- trial countries, costs in other countries are ex- ered pollutants (usually sulfur dioxide, nitrogen pected to be similar. Some cost reductions could oxides, and particulates) throughout the imme- be achieved by maximizing the utilization of lo- diate area surrounding the facility. The map con- cal consultants, particularly if the local consult- sists of the computed concentrations at each site ant has the opportunity to carry out four or five and a plot of the isopleths (lines of constant con- such projects annually. Unless there is frequent centrations). Since plotting results in “smooth- use of dispersion modeling, it may not be worth- ing,” the actual computed data should be while to acquire the skill because of the rapid evaluated. The maps need to be evaluated (typi- changes in the models themselves and in the com- Annex. Example of Summary Output from an Airshed Model Table A.1. Air Pollution Characteristics of a Proposed Thermal Power Plant

A. AIR QUALITY PROJECTIONS (all units are µg/m3) Reference values World Bank guidelines National standards Annual average Daily maximum Annual maximum

Background levels Monitoring point 1 Monitoring point 2 Monitoring point 3 Monitoring point 4 Annual average Daily maximum Annual maximum

Design coal: Monitoring point 1 Monitoring point 2 Monitoring point 3 Monitoring point 4 background plus two 600 MW units Annual average

88 Daily maximum Annual maximum

Design coal: Monitoring point 1 Monitoring point 2 Monitoring point 3 Monitoring point 4 background plus four 600 MW units Annual average Daily maximum Annual maximum

Check coal: Monitoring point 1 Monitoring point 2 Monitoring point 3 Monitoring point 4 background plus two 600 MW units Annual average Daily maximum Annual maximum

Check coal: Monitoring point 1 Monitoring point 2 Monitoring point 3 Monitoring point 4 background plus four 600 MW units Annual average Daily maximum Annual maximum B. ASSUMPTIONS

1. Sulfur content Design value: 0.31% Check value: 0.92%

2. Ash content Design value: 15.5% Check value: 28.4%

3. Stack height

4. ESP efficiency

C. PROJECTED EMISSIONS

World Bank guidelines National standards Two 600 MW units Four 600 MW units

SO2 (t/d) Design Check 89 TSP (mg/m3) Design Check

NOx (ng/J) Design Check

3 Note: SO2, sulfur dioxide; TSP, total suspended particulates; NOx, nitrogen oxides; MW, megawatt; ESP, electrostatic precipitator; mg/m , milligrams per cubic meter; ng/J, nanograms per joule; t/d, metric tons per day. 90 IMPLEMENTING POLICIES: AIR QUALITY MANAGEMENT

puter technology needed to effectively use the Notes models. 1. Dispersion refers to the movement of parcels of When Should the Modeling Be Done? gases, whether vertically or horizontally, and their si- multaneous dilution in the air. Dispersion modeling should be part of the ini- 2. Standards pertain to the environmental require- ments of the country or the local authority; guidelines tial environmental assessment for a power are practices suggested by organizations such as WHO project, for example. It is recommended that the and the World Bank. dispersion modeling be carried out early in project preparation (e.g., as part of the feasibility References and Sources study) before the plant location and the detailed design have been finalized. Adamson, Sebron, Robin Bates, Robert Laslett, and Alberto Pototschnig. 1996. Energy Use, Air Pollution, Additional Resources: For Further and Environmental Policy in Krakow: Can Economic Information on the Models Incentives Really Help? World Bank Technical Paper 308. Washington, D.C. • USEPA (U.S. Environmental Protection Electric Power Research Institute. 1986. “Estimating the Agency). “ISC Dispersion Model User’s Guide: Cost of Uncertainty in Air Quality Modeling.” EA- Volumes 1 and 2. ” EPA 454/B-95-003, a, b, and 4707. Palo Alto, Calif. c. Office of Air Quality Planning and Stan- KBN Engineering and Applied Sciences, Inc. 1989. dards, Research Triangle Park, N.C. “Mae Moh Air Quality Update: Units 1–11.” Pre- • USEPA. “CTDMPLUS Model User’s Guide.” pared for the Survey and Ecology Department of EPA 600/8-89/041. “Terrain Processor.” EPA the Electricity Generating Authority of Thailand. 600/8-88/003. Office of Air Quality Planning Bangkok. and Standards, Research Triangle Park, N.C. KBN Engineering and Applied Sciences, Inc. 1990. All USEPA models are available from the EPA “Sulfur Dioxide Air Quality Impact Analysis for Mae Moh: Units 1–13.” Prepared for the Survey and SCRAM bulletin board free of charge. Ecology Department of the Electricity Generating Tel: 919/541-5742 Authority of Thailand. Bangkok. e-mail: TTNBBS.RTPNC.EPA.GOV In addition all EPA models are available for a Meier, Peter, and Mohan Munasinghe. 1994. “Incorpo- rating Environmental Concerns into Power Sector fee from the National Technical Information Ser- Decisionmaking: A Case Study of Sri Lanka. World vice (NTIS), Springfield, Va. 22161. Bank Environment Paper 6. Washington, D.C. • Germany, Federal Ministry for the Environ- Scholze, R. H. 1990. “Dispersion Modeling Using Per- ment, Nature Conservation, and Nuclear sonal Computers.” Atmospheric Environment 24A(8): Safety. 1992. “Manual of Ambient Air Quality 2051–57. Control in Germany.” Berlin. USEPA (U.S. Environmental Protection Agency). • Germany, Federal Ministry for the Environ- 1993. “Selection Criteria for Mathematical Mod- ment, Nature Conservation, and Nuclear els Used in Exposure Assessments: Atmospheric Safety. 1992. “Air Pollution Control: Manual Dispersion Models.” EPA/600/8-91/038. Wash- of Continuous Emission Monitoring.” Berlin. ington, D.C.