The developed prototype code *OLAF* is capable of solving
complex, highly realistic air pollution facility location problems. The code is
able to follow the trace of the pollutant under consideration over different
ecotrophic levels and into different individual organs of a modeled
individual. There, detailed cytodynamic effects are simulated in order
to evaluate the health effect of the pollutant on humans or other beings. A
case study based on real data has shown that highly realistic models of
pollutant fate and effect can be used together with realistic atmospheric
dispersion models and in conjunction with an optimization process as a decision
making tool. As it had to be expected, the computational workload to find
optimal lcoations is far from trivial, and at the present moment the code can
probably not be used in an interactive mode, except for very simple problems.

While we have shown that the overall strategy for estimating and minimizing ecological effects and health effects is feasible, several open questions remain. For example, at the present moment the code is able to compute local minima only. Global minima, are, of course, of high interest to the decision maker. It remains to be seen if highly efficient global optimization techniques can or should be applied to the pollution model developed or if it will be necessary to resort to stochastic techniques of inherently low efficiency. Moreover, several modeling issues, especially with respect to the agglomeration function, remain open. A robust, realistic and computationally effective formulation for the multicriteria problem at hand still needs to be found. If a reformulation of the problem as a stochastic optimization problem is helpful is not clear at all. Other atmospheric dispersion models which include nonlinear chemistry of the pollution should be tested, too, as it is the case with other numerical integration techniques for the ordinary differential equations occuring in the model. Another line of research is to investigate the effect of time-dependent emission rates, a particular interesting subject since these emission rates might be considered as decision variables, albeit heavily constrained ones. Moreover, further work is necessary with respect to the evaluation of synergistic effects caused by several pollutants instead of only one. This, together with other open questions, is a subject of further research.

Wed Dec 22 12:25:31 CET 1999