The following text is the section 'Summary of discussions in Work Group II' from the proceedings of the workshop "Objectives for Next Generation of Practical Short-Range Atmospheric Dispersion Models".
The workshop was organized by DCAR (Danish Centre for Atmospheric Research).
The proceedings are now out of print, but certain sections of the proceedings are available via the Word Wide Web.
Model input - especially meteorological input
Model structure and model management
Summary of discussions in Work Group II:
Rapporteur: H. R. Olesen, Denmark.
Chairman: J. Kretzschmar, Belgium.
The main theme discussed in Work Group II was the requirements posed by regulators and model users on model capabilities. This entailed discussions concerning model input, model output, model structure and model management.
The discussion covered a rather broad range of subjects; clearly, some subjects attracted more attention than others. For one thing, the discussion centred more on "regulatory models" than on "real-time models".
The discussion in the work group reflected the fact that the whole workshop was of an introductory nature in the sense that it was meant to initiate a dialogue between the communities of model users and model developers. As a consequence, only a few subjects were discussed in depth.
The present summary concentrates on the most discussed matters and only to a minor extent covers those topics that were only briefly touched.
For those who wish to obtain some more insight into the issues discussed than provided by this summary, it can be recommended to read the pertinent papers in the proceedings. These include Paper 5 by S. Hanna (on meteorological input), Paper 6 by N. Thompson (meteorological input), Papers 3 and 8 by J. Irwin (meteorological input and harmonisation of models), Paper 9 by H. Noordijk (standards for model development), Paper 24 by C. Williams (British requirements for model capabilities), Paper 27 by J.A. Jones (harmonisation of models), Paper 28 by A. Robins (modelling standards) and Paper 29 (the Dutch standard on model development).
The present summary has been divided into sections on
- model input
- model output
- model structure and model management.
One last introductory remark concerns the terms "regulatory models" and "real-time models". It was discussed in the group how appropriate these terms are for describing the models discussed, and a number of alternatives for them were suggested. The above terms are used throughout the summary, but some views on their usage deserve mentioning.
Regulatory models: The term is normally used for models which can be used to determine whether a polluter complies with certain rules and regulations with regard to air quality management. Thus, an appropriate alternative term is compliance models. Also, the expression model handling routine releases was suggested.
Real-time models: The term was considered somewhat too restrictive for some of the models that were considered at the workshop. These models could be named - possibly more appropriately - episodic models or emergency response models. They need not necessarily be run in real time but can be used for hazard assessments etc. at any time.
Model input - especially meteorological input
Input for atmospheric dispersion models includes source description, topographic information and meteorological input. Only the meteorological input was discussed in any detail.
The basic premise underlying the whole workshop is that we are presently developing a "new generation" of models which has a better representation of the physics in the atmospheric boundary layer than the "old generation" of Pasquill-Gifford type models.
New-generation models require input of not only routinely measured meteorological data, but also of fundamental parameters that describe the atmospheric dispersion, such as the Monin-Obukhov length, friction velocity and boundary-layer height. The fundamental dispersion parameters can be derived from the measured meteorological parameters, but at the expense of a considerable amount of preprocessing.
Now, with the "post-Pasquill" dispersion models becoming increasingly available, it is timely to consider harmonisation within the preprocessing of meteorological data. This subject is treated in Paper 6 by N. Thompson, which formed the basis of some of the discussions in WG II.
According to Thompson, harmonisation within the preprocessing of meteorological data does not mean that a single method is sought for supplying the fundamental dispersion parameters. However, the various methods used should be well documented, and tested against experimental data and against each other.
Further, Thompson suggests the use of a class division of a set of relevant meteorological parameters. It should be used when generating meteorological input for models. If agreement on such a classification is eventually reached, it can be used by varying meteorological offices and similar institutions. With this procedure, a limited set of intervals would be defined for each chosen parameter, and corresponding to each interval there would be a parameter "bin" containing the frequency of that interval. For the entire set of meteorological parameters, there would be a multi-dimensional matrix, where each bin would contain the frequency of the thus defined meteorological scenario. Such a matrix would have many bins, as there are several relevant parameters (Thompson mentions seven in his paper, and there could be more), and there would be several classes for each parameter.
This "matrix" approach is attractive, because once the matrix has been defined, it can be used internationally, and it is relatively straightforward to compile and compare statistics for various locations. However, a number of disadvantages of the matrix approach were pointed out as compared with the alternative "time-series approach" which is more flexible. A severe disadvantage is that the matrix approach cannot be applied to all purposes. For instance, for some pollutants (i.a. ozone) it is not so important whether the concentration is high during a single hour, but it is critical when high concentrations persist during several consecutive hours. A matrix approach is not suitable to handle this sort of problem, as necessary information is lost in the class division. Another difficulty which concerns preprocessing of data is the following: if meteorological data are available for one location, but are needed for another, then the data have to be transformed using an appropriate method. This is much more difficult to do if the meteorological data are only available as a set of statistics, because useful information has been lost.
Conclusively it can be said that the "matrix approach" may be useful for some purposes, but it cannot be used exclusively; a time series of meteorological data having relevant parameters must also be available.
The discussions on harmonisation of meteorological input are likely to continue; Dr. Thompson agreed to act as a focal point. He is also a representative in a COST committee to which a proposal for work on this issue has been submitted.
It is a recommendation of the work group that
- There should be an action for harmonisation of meteorological input for "next-generation models".
If work on meteorological input is not coordinated, each model is likely to be equipped with its own preprocessor. Different preprocessors would probably produce different results, entailing confusion about the interpretation of preprocessor output. Further, the amount of work spent on preprocessing would be unnecessarily large in this situation - compared to the alternative, where there is a general acceptance of a limited number of preprocessing methods.
A study was suggested to
- conduct a review of meteorological data availability and access for modelling in EC countries. What is available and how does a user get it? Recommendations should be given to harmonize access to original meteorological data.
The output desired by a model user depends on the nature of the problems being studied. Therefore, there seemed to be a consensus in the work group that a basic requirement to models is flexibility in their output capabilities. It would not be adequate that models can only produce one or two kinds of statistics, e.g. 99 percentiles and averages. As an example, the various sorts of output delivered by the British UK-ADMS model were listed. This model is expected to deliver the following types of output:
1. For emissions of long duration:
Average concentrations and deposition rates; statistics of concentrations for a range of averaging times.
2. For short emissions, time-integrated results:
Average doses; average deposits; statistics of dosage.
3. For short emissions, instantaneous quantities:
Average concentrations and average deposition rates; statistics of instantaneous concentrations.
The desired flexibility can be obtained if models use time series as input and are also able to produce output in the form of time series. The output can then be treated by a postprocessor.
Further, it was stated by several participants that it is desirable to have as output expressions of the uncertainties involved in the estimates given.
The request for output in the format of time series implies that meteorological input must also be present in a time series.
The point of view was expressed, though not allocated much time in the discussion, that there should software for graphical presentation of output, and preferably also of input.
It was noted, but not commented much, that next-generation models should have a better treatment of chemistry than previous models.
A number of activities were proposed which could be carried out subsequent to the workshop by appropriate institutions. These studies could help in identifying more precisely requests for model output.
Suggested activities concerning model output:
- Compilation of a summary of national and international air quality standards.
- Review of regulatory use of models. Which models are actually used in the EC member states? What guidance or requirements do the different countries have?
- Comparison of output from different models for a single case (in the style of a product review in a journal).
- A study of the uncertainties inherent in model calculations and the setting up of guidelines on how to present them to the model user.
Model structure and model management
When model structure was discussed, a widely expressed desire amounted to the statement that a model should be flexible. Further, a modular structure was seen as desirable for a model. Such a structure has the advantage that each module can be replaced by another which is either better or more appropriate for the task in question.
The desire for flexibility could be interpreted as a desire of having some, very versatile - hence complex - models. However, a counterview was strongly expressed that it is not feasible to construct an "all-singing-and dancing model" - it is too overwhelming a task.
The situation can be described in the following way: There is on the one hand a set of regulatory needs and on the other hand a set of models which have been developed for various tasks.
What is important to the regulators is which models are fit for the various tasks. Thus, the key question to be answered by the modelling community is how well existing and forthcoming models fit the various needs - and where improvements can be made.
The discussions in the work group revealed that the above question, although essential, is difficult to answer for the time being. Models are developed and documented in a rather disorganized manner, and there seems to be a lack of some standards and tools which would make it feasible to make statements on model merits in a satisfactory way.
It is the general attitude in the modelling community that traditional models for regulatory purposes are in many respects inadequate. There are large uncertainties connected to their results, and in some instances the results are directly unreliable. This implies that either environmental decisions based on these models are arguable, or that excessively costly decisions are taken in order to be "on the safe side". However, systematic and extensive studies of the performance and shortcomings of models are few.
The question of quality management of models is addressed in some of the workshop papers: H. Noordijk (Paper 9) advocates a better description of models as a means to stimulate better use of models, and A. Robins (Paper 27) discusses standards for model development. Also, in a report which was cited at the workshop R. Britter speaks of the need to encourage quality management of technical models.
Much of this can be condensed to the desires that
- Work on model evaluation should have a high priority.
- There should be an action for setting up guidelines for model development and documentation. The aim is to promote more correct use of models.
In the work group, H. Noordijk of RIVM, the Netherlands, volunteered to be the contact person concerning the issue of guidelines for model documentation.
Having guidelines for model development and documentation imposes a burden on model developers. It was the subject of discussion (particularly at a plenary session), whether it is realistic to impose this overhead on modellers. There seemed to be agreement, however, that in principle it is advantageous to have such guidelines, but there were different opinions on how specific they should be.
The topic of whether harmonisation of models is desirable was touched upon in some papers (e.g. Papers 3, 10, 27 and 28) and in some discussions. There does not seem to be much support for the view that a single model should be enforced in the regulatory process; however, some forms of harmonisation are seen as beneficial.
As a practical step to improve communication within the modelling community, the concept of electronic Bulletin Board Services was brought into the discussion (BBS services are described in Paper 7 by H.R. Olesen). The idea was encouraged that
- One or more electronic Bulletin Board Services (BBS) or similar services should be established to help in the coordination of model development and management.
Several participants announced their interest in the establishment of such services. Thus, following the workshop, there may be established BBS's on the initiatives of A. Skouloudis (of the steering committee for the initiative on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes), P. Mestayer (of the ERCOFTAC Special Interest Group on Turbulence and Dispersion in the Urban Atmosphere), as well as EURASAP.
Homepage of the
Initiative on Harmonisation within Atmospheric Dispersion Modelling for Regulatory Purposes