Continuing research to improve the subroutines would assure that the model representing the best science would gradually improve over time.

Transcription

1 CREATING HARMONY FROM DISSONANCE IN MODEL DEVELOPMENT A BUSINESS APPROACH Richard H. Schulze, Chairman Trinity Consultants North Central Expressway, Suite 1200 Dallas, Texas Telephone: (972) FAX: (972) rschulze@trinityconsultants.com Abstract The development of dispersion models has highly unusual attributes compared to most scientific processes. Dispersion model development is characterized by a relatively low cost to develop a model, followed by an enormous cost to validate the models under a variety of conditions. Since harmonization in Europe is more of a bottom up phenomenon contrasted to the top down approach in the United States building a framework for model use and development is essential. There will be two essential parts of this framework. First, there should be a hierarchy of models that can be used to analyze situations. Simplistic models are adequate where there is clearly no threat to air quality standards. More refined models are necessary as air quality concentrations approach standards. Second, there needs to be framework to develop a sophisticated model in which various research groups in Europe undertake the development and validation of a specific part of a model. Models for the 21 st century will have many subroutines to address specific issues including: point sources, stack tip downwash considerations, building wake effects, convective turbulence, line sources, area sources, volume sources, moving sources, puff sources, stack plume rise both momentum and buoyancy, plume rise from flares, plume rise from buoyant line sources, buoyancy induced dispersion, particulate matter deposition, computations of concentrations for averaging times specified in each nation s regulations, processors for meteorological inputs that will include computed data at grid points at several elevations, computation of mixing heights, development of an appropriate scheme for sea breeze situations, and the ability to relate to varying surface roughness and anthropogenic heat sources. The creation of a universally acceptable model will involve the development and validation of dozens of subroutines, each one addressing a special topic. Only when all the parts are well designed and tested, will the finished model withstand the scrutiny and gain acceptance of the many regulatory agencies in Europe. Continuing research to improve the subroutines would assure that the model representing the best science would gradually improve over time. Keywords Dispersion modeling Meteorological data Air quality Validation Atmospheric transport 1

2 Introduction Well over one hundred governmental bodies regulate air pollution in Europe. In some countries, the regulations are quite centralized, while in others bodies at the provincial, regional or canton levels perform regulation. Individual countries, abetted by industry and labor, are generally reluctant to cede on air quality matters much authority to a central authority such as the European Union. This highly fragmented situation impairs the development of both a uniform approach to air quality regulation as well as mutually acceptable models for assessing air quality. The United States, in contrast, has a moderately strong central government and thus most regulations are developed at the national level by the Environmental Protection Agency (EPA) and then delegated to the states. It is the U.S. EPA models that are universally accepted by industry and agencies throughout the country. Still, it took some 15 years after EPA was established for models developed in states like Texas to be supplanted by the EPA models. The metaphor for the contrasting situation in dispersion models can be found in the development of the symphony orchestra. Until about 200 years ago, ensembles played without conductors as we know them today. 1 Mozart conducted either from the keyboard of a harpsichord or from the concertmaster s position. With the advent of larger scale works, it became necessary for someone to be in charge and direct the group. Some orchestras like the major London orchestras and the Vienna Philharmonic still have player associations that hire conductors. More common, however, is an orchestra board that hires the music director. Harmony came to music when individual players all received and accepted some direction. Harmony will come to dispersion modeling when there is common acceptance of a leader or a group providing direction. The balance of this paper will deal with a framework for model use, suggestions for development of the next generation refined models, the development of future meteorological processors, and conclusions. Framework for Model Use The development of air quality dispersion models is characterized by a relatively low model development cost followed by an enormous cost to validate the model under a variety of conditions. As was summarized in the Harmonisation workshop in Ostend, there are relatively few data sets with which to validate models. Thus, it is easier to validate models that set an upper bound on concentrations than those that purport to accurately portray concentrations. Given this situation, it is generally wise to categorize models according to use. Simple screening models can be used to analyze one or two sources and determine whether the incremental impact of the source warrants further attention. That level can be set from one to five percent of a standard or guideline air quality concentration. The principal role of such a screening model is ensuring that it generally overstates concentrations that would be computed if a more sophisticated model were used. The U.S. EPA model SCREEN is such a model. The estimates are on the high side of what actually occurs, and it is used as a tool to determine if further detailed analysis is required. Such a screening approach could be used with most of the 100,000 to 150,000 sources located in Europe that are or could be potentially subject to regulation. 2 The next level of modeling could be called advanced screening, and it should be able to analyze many sources of various types using meteorological data routinely collected at airports. If the results of this level of modeling show that both the proposed source(s), as well as neighboring sources, are below about 60 to 75 percent of a standard or guideline, then the authorities could confidently approve the project. This type of model - similar to the U.S. EPA ISC model should also generally over-predict actual concentrations. The 2

3 levels at which individual new source contributions become important should be stated in a guideline for model use. Similarly, the guideline should state the level of concentrations from all sources near a proposed source at which a more sophisticated model should be used. The third level of model could be termed refined models. It is likely that fewer than 1000 sources have emissions large enough to justify the use of this type of model. Refined models would almost always use onsite meteorological data and would only be required where concentrations found using models of lesser accuracy show possible problems. This type of model should try to estimate more accurately the magnitude of the upper bound of concentrations. Refined models are the principal thrust of my paper. In addition, there would be an even higher level of models called research models, which would require extensive local validation to be approved. Over the years, nearly all the presentations on these workshops have focused on refined and research models, even though the bulk of all modeling required for the approval of an application can be accomplished with a screening or advanced screening model. Development of a Refined Model Models developed for the 21 st century will consist of a rather long list of algorithms, each of which can be individually tested and validated. These include: Point sources Line sources Area sources Volume sources Moving sources Algorithm Puff sources Plume rise from point sources both buoyant and momentum Plume rise from flares and fires associated with accidents Plume rise from buoyant line sources Buoyancy induced dispersion Deposition Convective turbulence modeling Building wake effects Algorithm Stacks, vents Application Highways, unpaved roads Waste treatment lagoons, mining operations Multi-story parking structures, chemical plants, oil refineries Ships at sea Explosions and accidents Stacks emitting gases of various densities Flares, opening burning accidents Aluminum smelters, glass factories and similar sources Increase in plume spread caused by entrainment of ambient air into a rising plume Any source with particular matter emissions, the emission of gases that react to form particles, and the deposition caused by rainout Daytime conditions Turbulence induced by structures and other obstructions; sources with adjacent buildings which affect transport and diffusion of pollutants Application 3

4 Stack tip downwash Representation of concentration Mixing height and partial plume penetration of the mixing height Sea breezes Surface roughness Antropogenic heat sources Treatment of terrain effects Wind and turbulence changes with height Directional wind shear Loss of plume elevation when winds at stack top exceed certain values Related to standards or guidelines of a regulatory body in terms of frequency of events and computation of averages; relate concentration to health effects Stacks with hot effluents Sources near large bodies of water Account for increased turbulence caused by the nature of the underlying surface Modeling in urban and/or industrial areas Sources and receptors near complex terrain Improved modeling of all sources Accounts for horizontal wind direction shift with altitude From a standpoint of development, addressing in a timely fashion all of the issues listed above, plus a number of other topics, is a gigantic task. What can be done is for individual research organizations to select one topic, or just a few, and then investigate these, develop algorithms and procedures, and validate them to ensure that they replicate real world situations. If 15 of these issues were addressed by about 10 research organizations, then the algorithms could be assembled onto a framework or skeleton to create a model with broad scale application. This model would then jointly represent the best research Europe has to offer. Continuing research to improve subroutines would assure that the model representing the best science would gradually improve over time. The approach outlined will enable an individual research organization to really probe deeply in a rather narrow area rather than single-handedly trying to develop a comprehensive model to address, at least crudely, all the issues. If 10 or 15 research organizations each chose specific areas to study, I believe that more progress would be made in developing a comprehensive up-to-date model for use in Europe as well as elsewhere. Meteorological Data The amount and timeliness of meteorological data has increased in the past decade and further improvements are likely in the decade ahead. 3 A great deal of information is available from aircraft as they land and takeoff, from geo-stationery satellites and polar orbiting satellites, from various weather radars, as well as from manned and automated surface observation stations. Atmospheric scientists have developed models that provide data on up to 80 parameters on a 50-kilometer grid at over 40 elevations. As the European Meteorological Agency acquires more supercomputers, the grid size will shrink and the number of elevations will increase. Moreover, this data will be available in almost real time. In fact, the amount of data is so overwhelming that development of some systems to select that data of greatest interest to dispersion modelers is required. The progress has been so rapid in this field, that the use of on-site data may have a limited future. Instead, a site will be able to download more comprehensive data on modeled atmospheric parameters 4

5 based on the assimilatio n of information from many sources than could be economically supplied by an onsite location. Terrain and sea breeze wind flow models to accurately replicate on-site conditions nearly anywhere in Europe could process such data. All of this means that resources will have to be devoted to abstracting relevant data from the large quantity available, archiving this abstracted information, and developing a meteorological processor to produce data for use with the next generation of dispersion models. Conclusion There are three principal thoughts in this paper. First, there needs to be a hierarchy of models so that scarce resources, both computer and human, are effectively used. Screening models can be used for the vast bulk of sites in Europe. Advanced models, using readily available meteorological data, can be used if the site has numerous sources or if predicted concentration using a screening model show excessive concentrations. Refined models need to be developed to improve the accuracy of predictions, but I doubt that we will ever achieve accuracy of ±10 percent. Finally site specific models can and should be used if they have been adequately validated for a site. The organization for the development of a European refined model will require the development of a model skeleton. Then 10 or 15 research institutes throughout Europe can work on one or two of the subroutines that would be hung on the skeleton. Each research institute would be responsible for developing and validating the particular section of the model. This form of organization, where each institute goes deep on one or two subjects, will produce a better result than if all the research institutes try to cover all aspects of a model in a less-than-thorough manner. Finally, significant changes on the type of meteorological data available to modelers are currently taking place. It is vital that dispersion modelers work closely with meteorological agencies to help define the type of data most useful to modelers. The data currently available are of Brodingnagian proportions. There is an immediate need to develop more useful meteorological parameters for direct input into models. As an American, I have always admired the scope and skill of research activities in Europe. By organizing an Europe-wide effort to develop better models using the improved meteorological data now available, you have the opportunity to create a 21 st century model free of many of the historical legacies that hobble model development and acceptance in the Unites States. References 1. Stanley Sadie, ed., The New Grove Dictionary of Music and Musicians, Vol. 4, 1980, Macmillan London, pp Schulze, Richard H., (1994), Balancing Simplicity with Accuracy in the Use of Dispersion Modeling in the United States, Third Workshop on Harmonisation Within Atmospheric Dispersion Modelling for Regulatory Purposes, Mol, Belgium, November 21-24, Turner, D. Bruce, Schulze, Richard H., Potential Use of NOAA-Archived Meteorological Observations to Improve Air Dispersion Model Performance, EM, March 1998, pp , Air and Waste Management Association, Pittsburgh, PA, USA 5