Odor dispersion: models and methods

Web extra: dynamics of dispersion

Web extra: puff models and plume models

Web extra: peak-to-mean elements

Modeling the dispersion of contaminants in the atmosphere has become a routine method for air quality management enabling us not only to predict future air quality effects but also to propose rational and cost-effective control strategies. For these reasons, modeling has been used increasingly in the regulatory arena for permitting and environmental assessment.

When it comes to odors, atmospheric dispersion modeling holds the same promise, and it is being used increasingly to determine not only source culpability, but to evaluate and rank control measures. The traditional atmospheric dispersion models and modeling guidance were originally developed to assess specific compound concentration effects; assessing the effects of odors, however, requires significant differences in approach. The implications of these differences must be evaluated before applying any model to an odor assessment.

Odor modeling characteristics

Differences between traditional dispersion modeling and odor modeling appear in at least three areas:

• Source (wastewater treatment plant, rendering plant, and so forth)
• Travel from the source to receptor
• At the receptor (the human nose).

The methodology used for an odor assessment is frequently based on only one of these factors, such as the short detection or recognition time characteristic of odors at the receptor, without regard for the effect of the other factors. This can lead to results that appear to overlook the physical phenomena associated with the project. It is important, therefore, when planning an odor assessment, to look at the "big picture" before deciding on the appropriate approach.

Factors	Traditional air quality modeling	Odor modeling
Standards	USEPA and states	Varied, no consistent approach
Averaging time	Hourly, annual	Seconds to minutes
Source: release time	Continuous	Short-term or continuous
Source: characterization	Pollutant-specific	Complex combinations

Odor perception

The science of odor evaluation is subjective because many facets (character, acceptability, intensity, hedonic tone, and so forth) can only be quantified by a subjective instrument (the human nose). This subjectivity leads to a good deal of complication when it comes to selecting appropriate odor criteria and relevant averaging times.

Several potential odor levels might be used as an odor criterion or standard. The detection threshold can be defined as the lowest concentration of a substance that can be detected above a blank sample by an odor panel. The recognition threshold, on the other hand, is the lowest concentration of a substance that can be recognized based upon the character of the odor. Published odor threshold values for specific compounds have generally been derived in the laboratory and represent the concentration at which a compound can be detected by the "average" person. These odor threshold values can vary widely for a given population and a given odor. H₂S, for example, has an odor threshold that varies from 1 ppb to 130 ppb (1).

Most odor assessments are performed to prevent or mitigate odor complaints. Willhite and Dydek (2) questioned whether the odor threshold is the same as the nuisance level (a level that would generate complaints) when an ambient criterion is needed for regulatory application. They noted, for example, that the nuisance level appears to be related to the "odor acceptability," which is based upon an individual's attitude and experience with the odor. The results of a field study (3) which they reported implied that people will complain, in general, when the odor reaches approximately four times the odor threshold. They also noted that the level at which people complain differs for unpleasant and pleasant odors. In this case, chemicals with unpleasant odors had a complaint level approximately three times the odor threshold, but pleasant odors were not recognized as a nuisance until the ambient odor levels exceeded five times the odor threshold.

Averaging time

Figure 1. Smoke plume observed instantaneously and averaged over 10 min and over 2 hr (Slade 1968)

One of the primary characteristics of standard dispersion models that must be modified for use in odor assessment is appropriate averaging time. Figure 1 is a schematic of hypothetical plume boundaries, much in the way that a camera with different exposure times might photograph a smoke plume. An instantaneous snapshot of the smoke plume primarily shows the meander of the plume under the influence of atmospheric turbulence larger than the plume. As the exposure time increases, the photograph captures both the meander and the internal spread of the plume; detail within the plume is smeared, and the plume boundaries increase with increasing distance and time to account for the plume meander. As the distance increases far downwind, even the envelope of the time-averaged plume can fluctuate around the centerline as it comes increasingly under the influence of large-scale atmospheric turbulence.

Figure 2.

In Figure 2, the crosswind concentration profiles (corresponding to plume depicted in Figure 1) show that the centerline concentration for the instantaneous plume is significantly higher than that for the time-averaged plume; that is, the shorter the sampling time, the larger the fluctuations from the ensemble mean. The use of a finite averaging time removes the very high and very low frequency fluctuations. This variation in "peak" concentration with averaging, or sampling time, is a key issue in odor modeling, or in the modeling of any substance where it is important to determine short-term effects, such as highly toxic or explosive contaminants.

Web extra: dynamics of dispersion

A two-fold problem arises. Most of the standard dispersion models use Gaussian dispersion equations. Gaussian models are empirically based and rely on sampling data that are necessarily time-averaged, such as the turbulent diffusion parameters, [sigma]_y and [sigma]_z. In addition, Gaussian models assume a steady-state condition. That makes Gaussian-based dispersion models, such as the Industrial Source Complex (ISC) model, applicable for averaging times of 3 min to 1 hr. If odor criterion needed for the analysis is on the order of seconds, then the time-averaged formulas could underestimate a shorter-term peak odor effect. Most odor sampling, however, is also time-averaged to collect sufficient sample for analysis. This makes validation of proposed averaging-time adjustments difficult.

Source characteristics

Time, in terms of release time, is also a consideration when characterizing a source. Traditional modeling generally assumes that the emissions of a pollutant are continuous and that the rate of emissions do not vary over the time frame of interest. An instantaneous or short-term release, such as a puff of gas from a pressure relief valve, can have a different dispersion pattern than a continuous release, especially in the near-field region where it is transported bodily by large-scale turbulent eddies in the atmosphere. To represent these releases accurately, alternative models, such as puff and statistical models, have been derived.

An additional distinction between standard regulatory modeling and odor modeling is in the characterization of the emissions themselves. Modeling performed for air permits are generally pollutant-specific. Emission rates or fluxes (in units of mass emitted/unit time and mass/unit time/unit area) are determined from source sampling, emission factors, or theoretical emission formulas. Many odorous emissions, on the other hand, are complex combinations of compounds. A single indicator compound, with a low odor threshold and high emission rate can be used if it is truly representative of the sources under consideration. However, Duffee and O'Brien (10) have pointed out that this approach can lead to significant underestimates of offsite odor effects.

Odor emission rates frequently have the dimensions of odor units/unit time or odor units/unit time/unit area. This "emission rate" is based on determining a source concentration in odor units and multiplying this concentration by a volume flow rate.

It is important to model an emission rate that truly represents the source, whether that be in dimensions of mass/unit time or as odor units/unit time. The latter approach requires site-specific source sampling to characterize the odor emissions correctly.

Atmospheric dispersion models

Atmospheric dispersion results from turbulence in the atmosphere, and the instantaneous concentration downwind of a source varies continuously with the turbulence in the wind. A continuous plume from a source of emissions is expanded around its center by the smaller eddies (turbulence) in the atmosphere (that is, smaller than the size of the plume). Field observation of dispersion in the atmosphere also indicates the presence of large-scale, short-term fluctuations in concentration that are a characteristic feature of atmospheric dispersion. These larger-scale atmospheric eddies transport the plume bodily, primarily in the lateral and vertical directions (meander), while providing little in the way of dilution. Eddies equivalent to the size of the plume both dilute and transport the plume.

If the effects of a plume's spread and meandering are viewed at a fixed location, such as a sampling location, the monitor would "see" periods of turbulent concentration fluctuations as a plume travels past the monitor and periods of zero concentration, or intermittency, when the plume meanders away from the monitor. Based upon these observations, the dispersion of the plume can be viewed as the result of two distinct processes: the instantaneous spreading out of the plume in the vertical and crosswind directions (from the small eddy turbulence), and the meandering, or fluctuation of the entire plume about its mean position as it travels downwind (from the large-scale eddy turbulence). A "true" model of atmospheric dispersion should be able to simulate both of these processes.

The ultimate goal of an atmospheric dispersion model is to predict accurately concentrations downwind of any source (or sources) under any and all atmospheric conditions. Unfortunately, atmospheric processes are so complex, and our understanding so elementary, that current models have limitations on their applicability. Models have been developed to evaluate different source types (point, area, volume), different terrain (simple or complex), different locales (urban, rural), different release rates (plume, puff), and different meteorological conditions (stable, convective). Investigators must select the model(s) that most closely approximates the parameters of the source or characteristics of the dispersion process under analysis.

Similar to the decision process used to select the appropriate model for regulatory purposes, the selection of the appropriate dispersion model for odor assessment starts with the source type and release scenario. In general, most sources can be categorized as point, area, or volume sources, with continuous or instantaneous releases. The sources responsible for odor complaints are generally continuous sources, such as from stacks, scrubbers, or basins; although routine but instantaneous or very short-term releases (for example, from digester pressure release valves) can also pose problems at nearby receptors. Depending upon the rate of release relative to odor perception's short time frame, intermittent sources can be classified as either continuous sources (release rate on the order of minutes or longer), or instantaneous sources (release rate on the order of seconds).

Web extra: puff models and plume models

The Gaussian model of diffusion is the most widely used model for plume dispersion. Its most attractive feature is that it fits what we see and experience in the real world for a range of conditions. In addition, the mathematics of the model are fairly straightforward. On the other hand, Gaussian models need significant empirical input to be used for practicable dispersion estimates, making the model results highly dependent on the conditions of the sampling used to derive the empirical values.

Problems with the Gaussian model arise because the model assumes a time-averaged distribution in the plume and assumes that the meteorological conditions (including wind direction) are constant of the time required for the plume to travel from the source to the receptor. Under these conditions, results are applicable for times of approximately 3 min to 1 hr. This time averaging may not fully account for the turbulent concentration fluctuations within the plume, nor the meander of the plume from the mean direction. This could lead to under-prediction of the short-term concentration.

Gaussian models, such as the Industrial Source Complex model, have significant advantages, however. They have been widely applied and modified to consider numerous source types with an assortment of site-specific characteristics, such as terrain and building wake. The ISC model, for example, has undergone extensive field testing and validation so that it has widespread regulatory approval. In addition, many of the Gaussian models are in the public domain, and the source codes can be obtained from regulatory agencies or through governmental electronic bulletin board systems. This significantly reduces the cost of an odor assessment and allows the modeler the opportunity to match the model to the specific project. Adapting the results from time-averaged Gaussian model output, through the use of a peak-to-mean ratio, has been used as an alternative method of determining peak concentrations.

Peak-to-mean ratio

In general, most concentration and turbulence field data are collected over relatively long sampling times (on the order of minutes) because of the difficulty of measuring high speed fluctuations. For any fixed sampling time, the mean concentration (mean), which is assumed to remain nearly constant, can be determined. As shown above, however, within that sampling time there are significant short-term fluctuations which may exceed the mean by as much as two orders of magnitude. To account for this difference, considerable effort has been spent to determine the peak concentration (peak) (Gifford [14], Singer et al. [16], Hino [8], Islitzer and Slade [17], Pasquill [18]). Analysis of numerous field data have led to estimates of a "peak-to-mean ratio" for different source-receptor configurations. Once a peak-to-mean ratio appropriate for the project has been determined and the site has been modeled using a Gaussian model like ISC, the average model output can be multiplied by the peak-to-mean factor to estimate a peak concentration. The advantage of this approach is that the analysis retains the benefits of using standard dispersion models, such as substantial validation and peer-review and regulatory approval. Before applying a peak to mean ratio, several elements must be considered:

• Downwind distance
• Stack height
• Terrain effects
• Building-induced turbulence.

Web extra: peak-to-mean elements

Summary

Odor analysis is distinguished from routine atmospheric dispersion modeling by its subjective criteria and short time scale. The short recognition time for odors, in particular, is an important issue in odor modeling because the standard diffusion models, such as ISC, are valid for averaging times on the order of minutes. This averaging time can eliminate shorter-term peak concentrations.

A number of methods have been developed to determine peak concentrations. One approach is to use models, such as the fluctuating plume-puff model and puff model, that were developed to include the effects of the fluctuating concentrations. These models divide the diffusion process into two components: a meandering component and a spread component. Most models designed to account for the concentration fluctuation consider only one of these components. Some models assume that the spread within the plume puffs is Gaussian and assess the effect of the meander; others look strictly at the spread. An additional problem with these models is the limited amount of data needed to determine the diffusion coefficients. This has led to the questionable use of Gaussian plume coefficients in some of the puff models. The effects of terrain and building wake in modifying the short-term concentrations predicted by these types of models is also unknown.

Evaluation of field experiments to estimate the dependence of concentration on sampling time (recognition time) has led to a power law relationship with time. However, the selection of the proper exponent is a function of source type and height, distance downwind, and terrain and topography.

A similar approach to the time-scale problem is the use of a "peak-to-mean" ratio. Once the mean concentration is determined using standard dispersion methods, a peak concentration can be calculated with the use of a peak-to-mean factor. The advantage of this method is that standard dispersion models, which have had substantial validation and have regulatory approval, can be used to assess odor as well as to analyze air quality. As with the power-law relationship, however, the peak-to-mean concentration ratio depends on distance, source type and height, terrain, and the influence of building wakes.

The ultimate decision as to which model to use for an odor assessment must be made after the characteristics of the source and the site are understood. Focusing solely on the effects of averaging time, while ignoring other real influences on the flow, implies a potential for under-prediction of odor effects with the use of standard (Gaussian) models because these models predict time-averaged concentrations. However, the effects of buildings, terrain features, and other flow obstructions, as well as other factors such as source height and receptor distance from the source, can lead to considerable smoothing of the flow. These project-specific influences should be factors in the selection of the appropriate modeling technique for odor assessment.

References

1. Odor Thresholds for Chemicals with Established Occupational Health Standards. American Industrial Hygiene Association, 1989.

2. Willhite, M. T. and Dydek, S. T., "Use of Odor Thresholds for Predicting Off-Property Odor Impacts," in Recent Developments and Current Practices in Odor Regulations, Controls and Technology. Transactions of the Air & Waste Management Association, 1991.

3. Procedure for the Determination of Odor Impact Models by the Binary Port Odor Panel Method. Ontario Ministry of the Environment, Toronto, 1988.

4. Slade, D. H.; Meteorology and Atomic Energy, D. H. Slade, Ed., U. S. Atomic Energy Commission, Office of Information Services, 1968.

5. Inoue, E., J. Met. Soc. Japan, 1950 28, 219.

6. Stewart, N.G., Gale, H. J., and Crooks, R.N. Int. J. Air Pollution, 1958, 1, 87-102.

7. Cramer, H. E. "A practical method for estimating the dispersal of atmospheric contaminants," in Proceedings of the Conference on Applied Meteorology, American Meteorological Society, 1957.

8. Hino, M. Atmospheric Environment, 1968, 2, 149-155.

9. Nonhebel, G. J. Inst. Fuel, 1960, 33, 479-511.

10. Duffee, R. L. and O'Brien, M. A., "Establishing Odor Control Requirements by Odor Dispersion Modeling"; 92-153.01, 85th Annual Meeting of the Air and Waste Management Association, Kansas City, KS, 1992.

11. Högström, U. Tellus, 1964, 16, 205-251.

12. Hanna, S. R., Briggs, G. A. and Hosker Jr, R.N. Handbook on Atmospheric Diffusion, DOE/TIC-11223; U. S. Department of Energy, 1982.

13. Randerson, D. Atmospheric Science and Power Production, DOE/TIC-27601, U.S. Department of Energy, 1982, 584-619.

14. Gifford, F. Int. J. Air Pollution, 1960, 3, 253-260.

15. Högström, U. Atmospheric Environment, 1972, 6, 103-121.

16. Singer, I. A., Imai, K., and Gonzalez del Campo, R. J. Air Pollution Control Assoc., 1963, 13, 40-42,

17. Islitzer, N. F. and Slade, D. H.; Meteorology and Atomic Energy, D. H. Slade, Ed., U. S. Atomic Energy Commission, Office of Information Services, 1968.

18. Pasquill, F.; Atmospheric Diffusion, 2nd. ed.; Ellis Horwood Limited, Chichester, 1974.

19. Stewart, N.G., Gale, H. J., and Crooks, R.N. The Atmospheric Diffusion of Gases Discharged from the Chimney of the Harwell Pile (BEPO), A.E.R.E.-H.P./R.1452, 1954.

20. Wipperman, F., J. of Air and Water Pollution, 1961 4.

21. Meroney, R. N.; Engineering Meteorology, E. Plate, Ed., Elsevier, Amsterdam, 1982, 481-526.

22. Csanady, G.; Turbulent Diffusion in the Environment, D. Reidel, Dordrecht, 1973, 248 pp.

23. Ramsdell, J. V., and Hinds, W. T. Atmos. Environ. 1971 5, 483-495.

24. Wilson, D. J.; Contamination of Building Air Intakes from Nearby Vents, University of Alberta, Dept. Mech. Eng. Report No. 1, Edmonton, Canada, 1976, 126 pp.
____________
Richard J Pope, PE is a vice president with Malcolm Pirnie, Inc. Phyllis G. Diosey, Ph.D., QEP is a senior air quality specialist with Malcolm Pirnie.