Combining the Environmental Models with other Large-Scale Models

The environmental phenomena are closely related to other phenomena. Therefore, the environmental models have to be considered together with other models. Several examples are given below to illustrate this statement.

Relationship Between Climatic Changes and Pollution Levels. Some results related to the impact of the climate changes on the ‘‘bad days’’ have been presented in one of the previous sections. It must also be mentioned here that the climate changes have significant influence also on other quantities related to damaging pollutants. Therefore, the climate changes must be considered as an important factor when decisions about measures, which will lead to reductions of harmful pollution levels, are to be taken.

Furthermore, the increased pollution levels will cause some feedback effects on the climate. This finding implies that it is desirable to combine high-resolution climatic models with environmental models to study in detail the interrelations between these two processes.

Preparation of Pollution Forecasts on Different Scales. This task involves running several weakly connected large-scale models in order to handle a set of complex problems:

- weather forecasts on different regional scales (starting with results obtained on a global scale),
- weather forecasts on an urban scale (perhaps in parallel for several urban areas),
- pollution forecasts on different regional scales,

- pollution forecasts on an urban scale (perhaps in parallel for several urban areas),
- treatment of the output results in order to prepare them for the people who will use them (data mining algorithms and high-speed visualization tools must be applied at this stage),
- sending the relevant data to appropriate media (television stations, radio stations, Internet sites, GMSs, etc.).

It is clear that computer grids will be very powerful tools in the solution of the very challenging task related to the possibility to treat efficiently the set of problems described above. Currently, such sets of problems are solved only by imposing many simplifying (and very often not physical) assumptions. At the same time, it is also clear that a lot of difficulties must be overcome in the efforts to run such complex tasks efficiently on a computer grid. The greatest difficulties are the tasks of

- achieving reliable and robust transition from one scale to another,
- communicating relevant data from one part of the computational grid to another, and
- preparing the final results, which should be easily understandable by the recipients.

Coupling of Environmental Models with Economical Models. If some critical pollution levels are exceeded, then some measures are to be taken in an attempt to avoid damaging effects. The measures are normally related to some reductions of the human-made (anthropogenic) emissions. The emission reductions can be performed either directly (by using new technologies, filters, etc.) or indirectly (by introducing higher ‘‘green’’ taxes). It is clear, however, that the reductions of the emissions will as a rule cause economical problems; in the worst case, economical crises may be caused by putting too-heavy requirements on emission reductions.

Therefore, it is necessary to combine the need to keep the pollution under the critical levels with the need to preserve the rates of the economical development of the region under consideration. It is worthwhile to combine the environmental models with economical models in the effort to achieve both lower pollution levels and sustainable development of the economy. The aim should be to optimize the process by finding out

- where to reduce the emissions and
- by how much to reduce them (the emissions should be reduced as much as needed, but no more than what is absolutely necessary)

to keep the balance between safe environment and sustainable development. The proper treatment of the task of searching an optimal solution with a combination of large-scale environmental and advanced economical models is both very time consuming, and the storage requirements are extremely high. The efficient solution of this task is a very challenging problem.

Formation and Transportation of Aerosol Particles. Aerosol particles are dangerous for health, modifying radiative fluxes, modifying cloud formation, and changing the chemical composition. Therefore, they have been recognized for their potentially negative impact on human health and ecosystems. This finding implied regulatory legislation regarding emissions and concentration levels of particulate matter all over the world. It was also acknowledged that particles play an important role in the global climate by their influence on earth’s radiative balance.

The goal of aerosol modeling is to establish a detailed description of the aerosol particle concentrations, their composition, and size distribution. This model requires advanced modeling techniques and innovation as well as reliable validation data of particle characteristics.

Aerosol models may provide a predictive capability for future projections of the outcome of policy strategies on emissions. Consequently, aerosol models are needed that properly describe the cycle of formation, dispersion, and removal of particles. Such validated models can be used as cost-effective tools for reliable studies of the current status and predictions for various environmental and health impacts in the future.

Some extra mathematical terms, which described the transport and the transportation of aerosol particles, have to be added to the model described by Equation (1) when it has to be used in aerosol studies. Moreover, it is necessary to add some equations in Equation (1).

Studying Persistent Organic Pollutants. The persistent organic pollutants (POPs) are chemical compounds with different origins but common characteristics, such as semi-volatility, hydro-phobicity, bioaccumulation, toxicity, potential for long-range transport, and a tendency to accumulate in cold regions (‘‘cold condensation’’).

POPs may have adverse health effects on humans and wildlife as well as harmful effects on the immune and reproductive systems. Several POPs currently are either banned or regulated through international treaties, but they are still found in the Arctic environment. Models similar to the model described by Equation (1) can be used to study POPs.

The treatment of such models leads to huge computational tasks because (1) the spatial domains are normally very large (hemi-spherical models and, even better, global models are to be used) and (2) fine resolution of the systems of PDEs is highly desirable.

Implementation of Variational Data Assimilation in Environmental Models. Many uncertainties are related to large environmental models. The lack of reliable knowledge for some underlying physical and chemical processes is introducing great uncertainties, but also other reasons are for uncertainties suggested such as, inaccurate input data. Observations can be used to reduce the influence of the uncertainties.

The variational data assimilation approach is becoming more and more popular. This approach could be viewed as an attempt to adjust globally the results obtained by a given model to a set of available observations. This approach has the theoretical advantage of providing consistency between the dynamics of the model and the final results of the assimilation. Variational data-assimilation procedures are based on the minimization of certain functional.

Assume that (1) an improved initial concentration field must be found (it is important when pollution forecasts are to be computed) and (2) observations are available at time-points t_p, where p = {0,1,2,, P}. These observations can be taken into account in an attempt to improve the results obtained by a given model. This effect can be achieved by minimizing the value of the following functional:

Data assimilation can be used not only to improve the initial values (as in the above example) but also in many other tasks (improving the emissions, boundary conditions, the calculated concentrations, etc.). The variational data-assimilation technique should not be considered as a universal tool that can be used in all situations. No measurements are available when different scenarios are used to study the response of the model to the variation of different parameters (emissions, meteorological conditions, climatic changes, etc).

We have to rely only on the model in such studies. Therefore, it is absolutely necessary to use data- assimilation techniques not only to improve the model results but also to improve some physical and chemical mechanisms implemented in the model. The hope is that the improved model will provide more reliable results in situations where no measurements are available.

Using Ensembles. Ensembles (based on using results from several different models or results from the same model run by using different parameters, say, different initial values) can be applied in an attempt to improve the reliability of the model. The results are normally averaged. One should expect that the averaged results are better in some sense.

This conclusion is based on an assumption (very often not explicitly stated) that no bias is observed in the errors of the results. If, for example, all the models participating in the ensemble procedure strongly overestimate the concentrations of some chemical species, then the results of the ensemble will not be much better because they will also overestimate the species under consideration.

Similar conclusions can also be drawn in the case where one model is run with many values of a selected parameter (in this case the set of the values should be carefully chosen). Another problem may develop if one model produces very different results (say, 100 times higher than the other models). In such a case, the ‘‘bad’’ model will spoil the results of the ensemble (it should be eliminated, before the preparation of the ensemble).

These examples are provided to show that one must be careful: It is necessary (1 ) to analyze somehow the properties of the models participating in the ensemble procedure or (2) to select a good set of values of the parameter which is varied.

The application of ensembles (for different purposes) requires increased computer power. Indeed, the performance of 50-100 runs and preparing an ensemble on the basis of all these runs is a challenging task even for the best high-performance computers currently available. However, the results obtained by using ensembles normally are more reliable than the results obtained in a single run.